DR Thesis 2015 Yaliso Yaya Balla PDF

Maternal and neonatal mortality in
rural south Ethiopia:

Comparing mortality measurements and assessing obstetric care
<DOLVR<D\D%DOOD
Dissertation for the degree philosophiae doctor (PhD)

at the University of Bergen

Dissertation date: June 25 th

Dedication
To mothers who die during pregnancy and childbirth in resource-limited settings without
accessing essential obstetric services.
“Pregnancy is not a disease but a normal physiological process that women must engage in for the
sake of humanity. Whereas the elimination or eradication of disease is a rational and laudable
endeavour, the same strategy cannot be applied to maternal mortality. There is no pathogen to
control, no vector to eradicate. Women will continue to need care during pregnancy and childbirth as
long as humanity continues to reproduce itself. Failure to take action to prevent maternal death
amounts to discrimination because only women face the risk.” Carla AbouZahr (2003)
Maternal and neonatal mortality in rural south Ethiopia 2015
Acknowledgements
I gratefully acknowledge the Norwegian State Education Loan Fund and the Centre for
International Health, University of Bergen for funding my study. I would also like to
acknowledge the Gamo Gofa Zonal Health Department and Woreda Health Offices in Bonke,
Arba Minch Zuria, Demba Gofa, and Derashe for the important supports during the
fieldwork. I am grateful to the health extension workers who did the birth registry, as well as
the supervisors and data collectors of all of the studies. Thanks to the academic and
administrative staffs and management of Arba Minch College of Health Sciences for support.
I wish to express my deepest gratitude to my supervisor Professor Bernt Lindtjørn. It has

been a great blessing to have been under your supervision for a decade for MPhil and PhD
studies. You have always shown me love and respect, while being strict to the things I should
work and learn. I cannot have sufficient words to thank you for your dedicated and highly
qualified support, amazingly quick feedbacks, and the fatherly understanding. Now I have the
confidence to think critically and work independently. Had it not been your close guidance
and understanding, this work would not have been successful.
To my co-supervisor, Professor Ole Frithjof Norheim, I would like to sincerely thank you for
your close supervision and professional guidance, especially during the write up. It is rare to
find a person both intelligent and humble like you. Thank you co-authors of the studies in this
thesis, Meseret Girma, Tadesse Data, Professor Yemane Berhane, Dr Ewenat Gebrehanna,
and Kristiane Tislevoll Eide, for your important contributions. I learned much working with
you all. I wish to express my sincere appreciation to the staff of the Centre for International
Health, University of Bergen for all-rounded support and friendly environment. To Borgny
Lavik, Unni Kvernhusvik, Ingvild Hope, Øyvind Mørkedal, Solfrid Hornell, thank you for
facilitating all practical and logistical issues.
Families of Solomon Haile, Eyob Nebiyu (Desta Taye), Senait Yigletu, Tadesse Washo, and
all members of Shalom Coventant Evengelical Church in Bergen, thank you for the prayer
and support. God bless you all. My appreciation goes to two special families in Bergen:
Geremew Huluka and wife Mette Øxnevad, and Dr Skjalg Klomstad and wife Dr Grete, you
two families showed me and my family a very specilal love and care. Special thanks!
Yaliso Yaya Balla PhD Thesis i

I owe my sincere gratitude to my best friend, asistant Professor Tuma Ayele, and his family
for the support and encouregement even himself being in difficult situations. As your name
itself mean ‟truth”, you are always committed to preserve faithfulness and integrity, and pay
the resulting cost ─ the reason why I love you so much. Dr Degu Jerene, you are a brother
who always guided me for improvements and here is part of the fruit; and as you often say:
‟the best is yet to come”. Brother Kebede Tefera, you are a person born to help and give love.
I am grateful to you and Selu, your wife. Ato Kare Chawicha, former head of SNNPR Health
Bureau and now minister d’etat, I am grateful to your understanding and encouragement.
Dear friends Misikir Lemma, Mesay Hailu, Dr Eskinder Loha, Dr Daniel Gemechu,
Endashaw Shibru, Taye Gari, Tigist Gebremichael, Berite Belete, Zemach Guma, Tadesse
Data, Damenech Zewdie, Bahilu Merdikios, Amsalu Kusita, Selamawit Dagne, Endrias Olto,
Demissie Admasu, Alemayehu Bekele, and Zekarias Erkola, I thank you and your families
for the lovely friendship and encouragement.
My father Yaya Balla and mother Batale Alaho, thank you for deciding to send me to school
defying the harsh poverty. My brothers Yosef, Lukas, Petros, Bantirgu, and Belachewu, and
sister Amenech, thank you for suppot and caring for our parents during my study. My late
sister Almaz, may your soul rest in peace as you passed away when I prepared to submit this
work. It will be so painful to go home where you are missing. Lame Legesse, Melesse
Minota, Almaz Lema, without your economic support, I could not have even completed my
primary education. Yehuala, Serawit, Enata, Habtamua, Aklile, and Manaye, God bless you.
My wife Tigist (Boche), I can never thank you enough for your love, care, and hard work to
support our subsitence while doing your own study, and reproducing beautiful kids. You are
source of inspiration and reason of sucess in my life, next to God. My four little kids, two
sons and two daughters, Yabets, Abigiya, Magnus, and Lydia, thank you for your love
and endurance. Love you so much and life is beautiful with you. I am grateful to the
love and support from my mother in-law Azalech Wudneh, brothers in-law Solomon,
Yonas, Yared, Dr Ketema, and Biniyam and sisters in-law Hirut, Hanna, Konjit, Selam and
Mesay. Finally, and most importantly, God the Almighty, Alpha and Omega, thank you for
favouring me. ‟ሁሉን ታደርግ ዘንድ ቻይ እንደ ሆንህ፥ አሳብህም ይከለከል ዘንድ ከቶ እንደማይቻል እኔ ያሊሶ አወቅሁ”
(‟I, Yaliso, know that you can do all things, and that no purpose of yours can be thwarted”),
Job 42:2. All I am and I have, it is because of you. Thank you God.
Yaliso Yaya Balla PhD Thesis ii

Summary
Introduction: The aim of the Millennium Development Goals (MDG 4 and 5) is to

substantially reduce maternal and child mortality in the world. However, information is
limited in low-income countries to help oversee progress towards the MDG targets. In
developed countries, quality data are obtained through routine vital registrations.
Unfortunately, registry-based data are lacking or incomplete in most developing countries,
and Ethiopia is no exception. As such, we had a scarcity of information on the level of
maternal and neonatal mortality, as well as the coverage and quality of obstetric services in
south Ethiopia. The information is important because the target for MDGs in 2015 and the
preparation for the Sustainable Development Goals (SDGs) are fast approaching.
Objective: The aim of the thesis was to measure and compare maternal and neonatal
mortality and obstetric services through community- and facility-based methods in southern
Ethiopia.
Methods: We used four different methods to measure maternal mortality, and in one of these
methods assessed one-year obstetric services in all health centres and hospitals in Gamo Gofa
(population 1.8 million in 2010) in south Ethiopia. The methods were: 1) Between January
and December 2010 health extension workers prospectively registered births and birth
outcomes in 75 rural villages in three districts in south Ethiopia (population of 421,639
people) (Paper I); 2) In February 2011, we conducted a survey in 6,572 households that
reported pregnancy and birth outcomes in the previous five years (2006-2010) out of a total
of 11,920 households in 15 out of 30 randomly selected rural villages in the district of Bonke
(Paper II); 3) Using the sisterhood method, we estimated the lifetime risk of pregnancy-
related deaths and MMR through interviewing 8,503 adult siblings from the 15 kebeles where
the survey for Paper II was conducted in Bonke (Paper III); and 4) We reviewed one-year
institutional records on births, birth outcomes (maternal deaths), and signal functions of
emergency obstetric care in all 63 health centres and three hospitals in Gamo Gofa. We then
calculated the population coverage of obstetric care, the rate of skilled birth attendance, and
the quality of obstetric service against the UN’s minimum standards (Paper IV).
Yaliso Yaya Balla PhD Thesis iii

Results: In 2010, we recorded 10,987 births and 53 maternal deaths (an MMR of 489 per
100,000 LBs, with 83% (44/53) of maternal deaths occurring at home. The MMR was
associated with the education level of the father of the baby, access to driveable roads, and
sickness of the mother during pregnancy (Paper I). In the household survey (Paper II), we
enumerated 11,762 births and 49 maternal deaths (an MMR of 425 per 100,000 LBs), and of
the maternal deaths, 87% (43/49) occurred at home. The poorest households and those with
illiterate household heads had a higher MMR. We also found 308 neonatal deaths (an NMR
of 27 per 1,000 LBs). Neonatal mortality was associated with household wealth, a residence
far from a driveable road (≥ 6 km), and narrowly spaced births in the households. We
estimated a lifetime risk (LTR) of 1 in 10 pregnancy-related deaths with a corresponding
MMR of 1,667 per 100,000 LBs by interviewing 8,053 siblings (brothers and sisters).
Because of the indirect nature of the method, the estimate refers to the year 1998 (12 years
before the survey).
In the facility review (Paper IV), we recorded 4,231 pregnancy- and birth-related admissions
(6.6% of an estimated 64,413 births in the area in the same year), and found 79 maternal
deaths in the institutions. This resulted in a quarter of the MMR being measured through the
birth registry (120 vs.489 per 100,000 LBs). Out of the studied 66 health institutions, only
three met the basic-, and two satisfied the comprehensive emergency obstetric care standards.
The coverage and quality of emergency obstetric care (EmOC) was below the UN’s
recommended minimum of five basic and one comprehensive EmOC facility for every
500,000 people. The rate of institutional delivery was very low (on average, 3.7% between
2006 and 2010, and 6% in 2010) in rural villages. Three studies consistently showed that
more than two-thirds of maternal deaths occurred at home (Papers I, II, IV).
Conclusion: Community-based measurement methods (birth registry and household survey)

provided comparable results of the MMR, which was high with most of the deaths occurring
at home. The proportion of skilled birth attendance and EmOC was low. It is possible to
conduct birth registries in rural communities where functional system of community health
workforce is available and use it as a tool to measure birth outcomes.
Yaliso Yaya Balla PhD Thesis iv

List of original papers
The thesis is based on the following original research papers that will be referred by their
respective Roman numerals.
Paper I Yaliso Yaya, Tadesse Data and Bernt Lindtjørn. Maternal mortality in rural
south Ethiopia: Outcomes of community-based birth registration by health
extension workers. PLoS ONE (2015), 10(3): e0119321
Paper II Yaliso Yaya, Kristiane Tislevoll Eide, Ole Frithjof Norheim and Bernt
Lindtjørn. Maternal and neonatal mortality in south-west Ethiopia: Estimates
and socio-economic inequality. PLoS ONE (2014), 9(4):e96294
Paper III Yaliso Yaya and Bernt Lindtjørn. High maternal mortality in rural south-west
Ethiopia: Estimate by using the sisterhood method.
BMC Pregnancy and Childbirth (2012),12:136
Paper IV Meseret Girma, Yaliso Yaya, Ewenat Gebrehanna, Yemane Berhane, and
Bernt Lindtjørn. Lifesaving emergency obstetric services are inadequate in
south-west Ethiopia: A formidable challenge to reducing maternal mortality in
Ethiopia. BMC Health Services Research (2013),13:459
Yaliso Yaya Balla PhD Thesis v

Abbreviations
EmOC Emergency obstetric care
CI Confidence interval
CSA Central Statistical Authority
HEW Health extension worker
HH Household
LB Live birth
LTR Lifetime risk
MMR Maternal mortality ratio
MOH Ministry of Health
NMR Neonatal mortality rate
OR Odds ratio
PCA Principal component analysis
REK Regionale Komiteer for Medisinsk og Helsefaglig Forskningsetikk (Norwegian)
RR Relative risk
SBR Stillbirth rate
SD Standard deviation
SES Socio-economic status
SNNPRS Southern Nations Nationalities and Peoples Regional State
TFR Total fertility rate
UN United Nations
UNFPA United Nations Population Fund
UNICEF United Nations Children's Fund
VHP Volunteer health promoters
WHO World Health Organization
Yaliso Yaya Balla PhD Thesis vi

Table of contents
ACKNOWLEDGEMENTS ............................................................................................................................................. I
SUMMARY ............................................................................................................................................................ III
LIST OF ORIGINAL PAPERS......................................................................................................................................... V
ABBREVIATIONS .................................................................................................................................................... VI
TABLE OF CONTENTS .............................................................................................................................................. VII
1 INTRODUCTION ....................................................................................................................................... 1
1.1 BACKGROUND ON MATERNAL MORTALITY (THE PROBLEM).................................................................................. 1
1.2 DEFINITIONS, CAUSES, TIME, AND INDICATORS OF MATERNAL MORTALITY.............................................................. 2
1.2.1 Definitions of maternal mortality ..................................................................................................... 2
1.2.2 Causes of maternal mortality ........................................................................................................... 3
1.2.3 Time of maternal death .................................................................................................................... 5
1.2.4 Indicators of maternal mortality (statistics) ..................................................................................... 5
1.3 MEASUREMENT METHODS FOR MATERNAL MORTALITY ...................................................................................... 8
1.3.1 Measurement backgrounds .............................................................................................................. 8
1.3.2 Measurement options (empirical measurements vs. analytical estimates)...................................... 9
1.4 HISTORICAL OVERVIEW OF POLICY AND PROGRESS IN MATERNAL MORTALITY ........................................................ 12
1.4.1 Early experiences and lessons learned ............................................................................................ 12
1.4.2 The 1980s to the present: from Alma Ata to the MDGs .................................................................. 13
1.5 DETERMINANTS OF MATERNAL MORTALITY .................................................................................................... 15
1.5.1 Theoretical framework .................................................................................................................... 15
1.5.2 Access to health care or poverty: Which matters most for maternal mortality?............................ 18
1.6 INTERVENTIONS FOR MATERNAL MORTALITY REDUCTION .................................................................................. 19
1.6.1 Single or packages of interventions ................................................................................................ 19
1.6.2 Strategies to distribute proven interventions to the public ............................................................ 19
1.6.3 Skilled birth attendance (SBA) and emergency obstetric care (EmOC) ........................................... 21
1.6.4 Access (availability, acceptability, and affordability) to obstetric services ..................................... 22
1.6.5 The three delays .............................................................................................................................. 23
1.7 NEONATAL MORTALITY AND ITS RELATION TO MATERNAL OUTCOMES ................................................................. 23
2 MATERNAL AND NEONATAL HEALTH IN ETHIOPIA ................................................................................ 24
2.1 ETHIOPIA: THE COUNTRY ........................................................................................................................... 24
2.2 HEALTH SERVICES IN ETHIOPIA .................................................................................................................... 25
2.3 MATERNAL AND CHILD HEALTH PROGRAMMES IN ETHIOPIA .............................................................................. 26
2.4 RATIONALE OF THE STUDY (THE STUDY IN CONTEXT) ........................................................................................ 28
3 OBJECTIVES ........................................................................................................................................... 31
3.1 GENERAL OBJECTIVE ................................................................................................................................. 31
3.2 SPECIFIC OBJECTIVES ................................................................................................................................. 31
4 METHODS ............................................................................................................................................. 32
4.1 STUDY AREA AND SETTING.......................................................................................................................... 32
4.2 STUDY DESIGNS AND DATA COLLECTIONS ....................................................................................................... 34
4.3 DATA ANALYSIS AND STATISTICS .................................................................................................................. 36
4.4 ETHICAL CONSIDERATIONS.......................................................................................................................... 38
5 RESULTS ................................................................................................................................................ 39
5.1 MATERNAL MORTALITY MEASURED THROUGH COMMUNITY-BASED BIRTH REGISTRY (PAPER I) ................................. 39
5.2 MATERNAL AND NEONATAL MORTALITY MEASURED THROUGH HOUSEHOLD SURVEY (PAPER II) ............................... 39
5.3 MATERNAL MORTALITY ESTIMATED THROUGH THE SISTERHOOD METHOD (PAPER III) ............................................ 40
5.4 OBSTETRIC SERVICES AND MATERNAL MORTALITY ASSESSED THROUGH HEALTH FACILITY DATA (PAPER IV) ................. 41
Yaliso Yaya Balla PhD Thesis vii

6 DISCUSSION .......................................................................................................................................... 44
6.1 DISCUSSION OF THE METHODOLOGY (DESIGN AND VALIDITY)............................................................................. 44
6.1.1 Study design and sampling ............................................................................................................. 44
6.1.2 Internal validity ............................................................................................................................... 48
6.1.3 External validity (generalization) .................................................................................................... 53
6.2 DISCUSSION OF MAIN FINDINGS .................................................................................................................. 54
6.2.1 Overview of the work and findings ................................................................................................. 54
6.2.2 Maternal and neonatal mortality ................................................................................................... 56
6.2.3 Inequalities in mortality outcomes ................................................................................................. 58
6.2.4 Skilled birth attendance and emergency obstetric care .................................................................. 59
6.2.5 Why a sign of reduction in maternal mortality in Ethiopia with a low skilled delivery rate? ......... 60
7 CONCLUSION AND RECOMMENDATIONS .............................................................................................. 64
7.1 CONCLUSION ........................................................................................................................................... 64
7.2 RECOMMENDATIONS ................................................................................................................................ 64
8 REFERENCES .......................................................................................................................................... 66
PAPERS I-IV AND APPENDICES ................................................................................................................................ 83
Yaliso Yaya Balla PhD Thesis viii

1 Introduction
1.1 Background on maternal mortality (the problem)
Maternal mortality is still one of the biggest health and societal challenges in the 21st century
in resource-limited countries. For many decades, the number of maternal deaths was over
500,000 per year in the world [1]. In 2013, an estimated 293,000 mothers died in the world in
the process of pregnancy or childbirth, with the maternal mortality ratio (MMR) per 100,000
live births ranging from a high of 957 in south Sudan to a low of 2 in Iceland [2]. The
lifetime risk of maternal death for Sweden is 1 in 30,000, whereas in Sierra Leone it is 1 in 6
[3]. In fact, the difference in the rates of maternal mortality is considered to be the greatest of
all health-related disparities between developed and developing countries [4]. In recent
decades, global maternal deaths dropped by 47% (from 543,000 in 1990 to 287,000 in 2010)
[5]. However, 99% of the current maternal deaths are in developing countries, especially in
sub-Saharan Africa and south Asia [6]. More than 90% of these deaths are preventable with
solutions currently available; particularly in relation to skilled care during labour, at delivery,
and a few days during the postpartum period. Yet, in 2008, 50% of maternal deaths in the
world only occurred in six countries (Ethiopia, Nigeria, Congo DRC, India, Pakistan and
Afghanistan) [7].
Following the initiation of Safe Motherhood Initiatives (SMIs) and Millennium Development
Goals (MDGs), access to interventions improved, and a reduction in maternal mortality has
been observed, even in some low- and middle-income countries [7]. However, in many sub-
Saharan countries, the rate of reduction has not been as planned for the MDG target [8]. In
some of these countries, even previous gains were reversed because of the HIV epidemic [9,
10]. This presents a great challenge in Africa to reducing the MMR (the number of maternal
deaths per 100,000 live births) by 75% in 2015 from the level in 1990 [7]. The reason behind
this is the limited access to and utilization of skilled care during pregnancy, childbirth and
postpartum compounded with the low socio-economic status of women in these settings [11-
13]. As such, many mothers deliver and die out of a health facility, which makes it difficult to
both prevent the unnecessary deaths and identify these deaths in settings where information is
poor.
Yaliso Yaya Balla PhD Thesis 1

In contrast, strong and accountable health systems have a great potential to substantially
decrease the adverse pregnancy and childbirth outcomes [14]. At the individual level,
identifying and classifying an adult woman’s death as maternal or non-maternal is a serious
challenge because of the complex nature of defining the cause of death (medical diagnosis)
and determining the pregnancy status during the death, especially deaths during early
pregnancy [15]. This constraint is a particular concern in low-income settings without vital
registration and where maternal mortality is greatest [16].
1.2 Definitions, causes, time, and indicators of maternal mortality
1.2.1 Definitions of maternal mortality
A clear understanding and universal application of maternal death definitions are crucial for
monitoring progress and comparisons across geographic areas, as the differential use of
definitions may present substantial implications on the credibility of estimates. The
definitions of maternal mortality have changed over time, which resulted in inconsistencies in
measurements and presents problems to oversee trends in maternal mortality. For example, in
Sweden the change of the maternal mortality definition from ICD-8 to the subsequent ICD-9
and ICD-10 editions caused an increase in rates [17]. In addition, the introduction of the
definition of “pregnancy related deaths” instead of “maternal death” has been a source of
measurement variation. For instance, a study from Bangladesh reported 15% more
pregnancy-related deaths compared to maternal deaths [18]. Moreover, clarity on the
definition of life-threatening maternal complications is also important because for every
mother who dies, there are many (30 or more) others who develop lifelong complications and
disabilities known as a “maternal near-miss” [19]. Two basic elements for a definition of
maternal mortality are the cause and the time of death in relation to pregnancy. The current
working definitions of maternal mortality are as follows [20]:

Table 1: Maternal mortality and complication related definitions
Term Definition Requirement ICD

Maternal deaths Death of a woman while pregnant or within Time-of-death ICD-9
42 days of the end of pregnancy, from any and cause-of-
cause related to- or aggravated by the death
pregnancy or its management, but not from
incidental or accidental causes.
Fortuitous Death from unrelated causes that happen to Cause-of-death ICD-9
(incidental) deaths occur in pregnancy or puerperium
Pregnancy-related Deaths occurring in women while pregnant or Time-of-death ICD-10
deaths with 42 days of the termination of pregnancy,
irrespective of the cause of the death
Late maternal Deaths in women occurring between 43 days Time-of-death ICD-10
deaths and 1 year after termination of pregnancy
(abortion, miscarriage, or delivery)
Maternal near- A woman who nearly died but survived a Severity and ICD-10
complication that occurred during pregnancy,
miss* (severe disability
childbirth, or within 42 days of termination of
complication) pregnancy (‘‘nearly dying, but surviving”).
Note: ICD is an international classification of diseases and health-related problems provided by the
World Health Organization (WHO;.* a maternal near- miss is well described elsewhere [19].
1.2.2 Causes of maternal mortality
Several distant and immediate factors determine the survival of a woman from pregnancy and
childbirth-related deaths. However, health services are mainly concerned with medical (direct
and indirect) causes. Direct causes are medical complications that primarily occur because of
pregnancy, child birth or the managements during these periods and account for about 80% of
maternal deaths [21]. The direct causes are bleeding (haemorrhage), infection, hypertensive
disorders, unsafe abortion, and obstructed labour, with bleeding the leading cause of maternal
deaths in developing countries; particularly in sub-Saharan Africa [22]. The combination of
the three dominant direct medical causes of maternal death (bleeding, infection and
hypertensive conditions) caused the largest proportion of MMR per 100,000 live births (500

in sub-Saharan Africa, 300 in south Asia, 100 in Latin America, and four in developed
countries in 2005) [3, 23]. This shows the relative importance of interventions targeting these
problems to reduce maternal deaths.
Indirect medical causes of maternal deaths are conditions that are not unique to pregnancy,
which include anaemia, malaria, HIV, tuberculosis, heart diseases and other existing medical
conditions aggravated by pregnancy to cause a maternal death. They are responsible for
approximately one-fourth of global maternal deaths, with HIV being the leading indirect
cause in countries highly affected by the epidemic [9, 21]. However, the quality of
information on the indirect causes of deaths is particularly problematic because of difficulties
in attributing deaths to these causes [24]. In some areas, targeting the key indirect causes of
maternal deaths may be as crucial as the focus on direct causes. Furthermore, the prevention
of maternal deaths from causes such as malaria [25, 26], anaemia [27] or HIV needs a
strategic investment of resources on the primary and secondary prevention of these causes to
substantially decrease maternal mortality.
Incidental and accidental causes such as suicide, violence (murders) and accidents, while
pregnant or within 42 days, has received little attention in maternal death statistics.
Nonetheless, evidence suggests that these deaths may be related to pregnancy [28-30]. For
example, domestic violence was the second common cause of deaths in pregnancy in India
[29]. In Matlab, Bangladesh, suicide caused 20% of deaths of unmarried pregnant women,
although only 5% among married pregnant women, while the risk of violent death was three-
fold among pregnant girls compared to non-pregnant girls [28, 30]. Though difficult to
measure, these findings suggest that ignoring the number of deaths due to both incidental and
accidental causes might lead to an under-estimation of maternal mortality.
The causes of maternal mortality may vary because of variations in the quality of the health
system in a given country or region as the direct causes of maternal death (obstetric risks) are
particularly sensitive indicators of the level of obstetric services in an area. In other words, in
a country that has quality health service and most women access to services, maternal deaths
from bleeding and infection can be substantially reduced [23]. One cannot expect the same in
areas where most women deliver at home because of poor services and low access to health
care.

Still, experts regularly estimate the proportion of the causes of maternal death for global and
regional comparisons. For example, infection caused a five times higher maternal deaths rates
in sub-Saharan Africa compared to developed countries (2.1% vs. 9.7%), whereas bleeding
caused 34% of all maternal death in Africa and 13% in developed countries in 2006 [23].
Over time, the proportion of maternal deaths caused by bleeding declined to 25% in sub-
Saharan Africa and increased to 16% in developed countries in 2013 [31]. This denotes that
when the health system is strengthened, the share of preventable deaths can decrease.
Consequently, when the number of deaths from a particular cause decreases, the proportion
of other unchanged causes increases despite the number of deaths from the latter not having
changed. The WHO provides the estimates of the causes of maternal deaths for each
geographic area [23, 31].
1.2.3 Time of maternal death
Predicting the time of maternal complications and fatal conditions during pregnancy is
difficult, but most maternal deaths occur during labour, delivery, and shortly after birth [32-
34]. Over 50% of maternal deaths that occur in the postpartum period occur in the first 24
hours after delivery, and over 80% occur in the first week after birth or abortion [35]. Thus,
skilled follow-up is needed during this critical period.
A conventional definition of maternal mortality considers deaths between pregnancy and 42

days after birth [36]. However, data show that some deaths due to maternal causes occur after
42 days, especially in the period up to six months postpartum [33, 34]. Accordingly, a new
category of definition,“ late deaths”, was proposed to include deaths within a year following
the termination of pregnancy [36].
1.2.4 Indicators of maternal mortality (statistics)
Maternal mortality is a major public health problem, but in terms of absolute numbers, it is
rare, which makes maternal mortality a challenging health outcome for statistical
measurement. Hence, several ways of describing the magnitude have been proposed and
used. These indicators are absolute numbers, maternal mortality ratio (MMR), maternal

mortality rate (MMrate), a lifetime risk of maternal mortality (LTR), and the proportion of
maternal deaths among reproductive age female deaths (PMDF). Even so, the meanings
carried by each of these indicators permit slight variations.
A. Number of maternal deaths
This is the method for reporting the number of maternal deaths in a geographic area during a
certain period, commonly per year. As such, absolute numbers in maternal mortality is an
important means of informing and alarming actors in the field. Some scholars expressed these
numbers in a powerful way, including as follows, by Dr Malcolm Potts in the WHO
Interregional Meeting on the Prevention of Maternal Mortality, November 1985 to express
250 maternal deaths very four hour in the world [37]: ''Every four hours, day in, day out, a
jumbo jet crashes and all on board are killed. The 250 passengers are all women, most in the
prime of life, some still in their teens. They are all either pregnant or recently delivered of a
baby. Most of them have growing children at home, and families that depend on them”,
Richard Horton, editor of The Lancet, expressed global maternal deaths as: “the number that
has challenged the maternal health community is 500,000" to reflect the number of maternal
deaths per year in the world over the decades [1]. In fact, the number of maternal deaths has
become the key report of the maternal mortality indicator in global maternal mortality
estimations [2, 5, 7]. Therefore estimating and reporting the number of maternal deaths that
could occur in a country or an area is powerful information, even without complex statistics.
B. Maternal mortality ratio (MMR)
MMR is the number of maternal deaths during a given time period per 100,000 live births
during the same period. It is widely regarded as the conventional measure of maternal
mortality, and measures the risk associated with each pregnancy (obstetric risk) [3]. A
calculation of the MMR can be possible from different sources of measures [38]:
୒୳୫ୠୣ୰୭୤୫ୟ୲ୣ୰୬ୟ୪ୢୣୟ୲୦ୱ
X 100,000
୒୳୫ୠୣ୰୭୤୪୧୴ୣୠ୧୰୲୦ୱ
Or
ൌȀ
ሺ
ሻ
ൌൌͳെሺͳെሻͳȀ
Where LTR= lifetime risk

C. Maternal mortality rate (MMRate)
This is the number of maternal deaths over a given time period per 100,000 women of
reproductive age (15-49) [39]. It measures both the obstetric risk and the frequency with
which women are exposed to this risk [40]. The MMRate can be calculated as follows [38]:

ͳͲͲǡͲͲͲ
ͳͷ െ Ͷͻ
Or
ൌൈ

ൌͳെሺͳെሻͳȀ͵ͷ
D. Lifetime risk (LTR) of maternal death

This measure accounts for both the probability of becoming pregnant and the probability of
dying from pregnancy-related causes, and the risk is accumulated across a woman’s
reproductive year. Consequently, LTR indicates the probability of maternal death throughout
a woman's reproductive life [40]. It takes into account the fertility probability and obstetric
risk together and the likelihood is expressed in odds, example, e.g. one in 10 in a defined area
[38].
ൌͳെሺͳെሻ͵ͷ
ൌͳെሺͳെሻ
ൌ͵ͷൈ
E. Proportion of maternal deaths among female deaths (PMDR)

This measures maternal deaths as a proportion of all female deaths of reproductive age
(usually 15-49 years) for a given time period [38]. In countries and areas with poor obstetric
care, the proportion of maternal deaths among reproductive female deaths is expected to be
high compared to settings where there is effective obstetric care:

‫ͲͲͳݔ‬ǡͲͲͲ
ͳͷ െ Ͷͻ
Note: the formula boxes are obtained from the works of Graham WJ et al; reference [38]

1.3 Measurement methods for maternal mortality
1.3.1 Measurement backgrounds
The efforts of measuring maternal mortality have a long history pointing back to the 17th
century in developed countries [41]. However, critical data to guide policy and interventions
for reductions of avoidable maternal deaths is still limited in developing countries. In fact, all
agree on the need for quality data to oversee changes in maternal mortality [16]. Similarly,
the challenges in measuring maternal mortality are universal, and no country in the world can
give a confident, reliable and valid national estimate of its maternal mortality [42]. Thus,
there is an understanding that maternal mortality is difficult to measure compared to other
health outcomes such as child mortality and fertility [43]. The reason behind this difficulty is
that capturing maternal deaths (finding and counting) at the population level, and then when
found, ascertaining a woman’s death as maternal (confirming the cause of death as a maternal
cause), are challenging [44].
Realizing the challenges in measuring maternal mortality as an outcome, some have used
process indicators such as skilled birth attendance, financial commitments and policy
approval [45, 46]. However, the main MDG5 indicator to assess the goal of 75% reduction in
MMR requires measuring maternal mortality. Accordingly, there is a need to continuing the
effort to find out improved ways of measuring maternal mortality, in addition to the process
indicators [47].
In developed countries, maternal deaths in the population are captured through routine vital
registrations (data sources that regularly record births, deaths, marriages, and divorces).
Autopsy (also referred as post-mortem examination) is an ideal method used to identify the
underlying and immediate causes of a maternal death [48]. Nevertheless, measuring maternal
mortality at population level is a problem, even in settings where sophisticated systems are in
place [49]. As a result, supplementary methods such as a confidential enquiry of maternal
deaths (using different sources of tracking data such as professionals, confidential enquiry
committees, news source) have been used to supplement data from registries [50-52]. In

resource-limited countries, which is where most of the people in the world live, both a routine
registration of deaths and medical autopsies to confirm the deaths are lacking [53].
The weakness of measuring maternal mortality in less developed countries has been widely
recognized since the 1980s; which has been further stimulated by stimulated by the initiation
of safe motherhood programme [54]. The subsequent MDG declarations also helped to put
maternal health and its measurement at the centre of development agendas [55]. The central
aim of these international efforts is to galvanize actions for the reduction of maternal
mortality, e.g. MDG5 to reduce the MMR by three-quarters in 2015 from its 1990 level.
Meanwhile, monitoring progress towards these targets needs quality data, as “what you count
is what you target” [56].
1.3.2 Measurement options (empirical measurements vs. analytical estimates)
Acknowledging a serious demand for information in developing countries where maternal

mortality matters most, several alternative data acquisition methods were proposed. Some of
these methods are empirical (direct searches for primary data), whereas others are analytical
(statistical derivations to reach at estimates) [39].
Empirical methods depend on the collection of first-hand data including:
A) Routine sources that rely on passive data collection based on reports from family or a
health facility. These routine methods are: 1) death registrations (civil registrations), 2)
sample vital registrations, 3) sample vital registrations with verbal autopsy, 4) health
facility statistics, and 5) decennial census.
B) Special opportunities (surveys and surveillances): surveys capture data at a single point
in time, with such methods including direct and indirect sisterhood methods, household
surveys, and sampling at service sites such as antenatal care. Regular surveillances makes
continuous and repeated household visits to note changes in vital events [demographic
surveillance sites (DSS) and active surveillance of reproductive age female deaths
(RAMOS)] [39]. Unfortunately, these passive-routine methods are incomplete or non-
existent in many developing countries. In addition, many of the survey and surveillance
methods classified under special opportunities need a large sample size that demand a
high cost, and are often subject to sample biases.

Analytical methods (modelling to estimate MMR, capture-recapture methods to correct for

under-reporting, and birth-death linkages to find maternal deaths) apply statistical techniques
to reach estimates of maternal mortality from the values of other proxy variables. Statistically
modelled estimates (especially that of the UN and the Institute of Health Metrics at
Washington University) have been the main sources of information for global and national
figures that can be compared across countries [5, 7]. The information has influenced
international organizations and national governments in low-income countries for planning
resources and actions. However, inconsistencies and controversies using these methods
resulted in calls for responsible estimations [57].
The advantages and shortcomings of these alternative methods is well described by Graham
and colleagues [39], and there are many works on measuring maternal mortality in resource-
limited settings [15, 16, 38, 39, 44]. Furthermore, realizing the shortcomings of the routine
and active data collection methods for maternal mortality, modification of the national census
is proposed as a feasible and efficient opportunity [16]. The proposal suggested the inclusion
of additional questions in the census that ask about maternal mortality in the previous 12
months before the national census [58]. Consequently, some countries have used the census
with maternal death questions, and estimates have been released [59].
However, there are no standardized methods that can be universally applied to monitor
maternal mortality in all settings. As such, the methods can be weighed based on the purpose
of the data, time urgency, and feasibility [38]. For example, knowing how mothers die from
bleeding in an area may need a case-by-case assessment of these deaths. Yet, a study to
improve the quality of obstetric health facilities can review maternal deaths and disabilities
that occurred in health institutions. Moreover, aiming to measure the magnitude and
differentials of maternal mortality at the national, regional, or district level requires
community-based studies. The ultimate goal should be routine and complete vital registration
to measure maternal deaths [60]. Figure 1 below summarizes the methods for measuring
maternal mortality.

Empirical measurements Analytical
Routine opportunities Special opportunities
1) Death 3) Decennial 4) Surveys 5) Surveillance

Birth-death
registration census Options: Options: linkages to
Options: 4A. Population-based 5A. Demographic find maternal
1A. Civil survey- asking about Surveillance deaths
registration 2) Health deaths in households Systems (DSS)
Facility
1B.Sample vital Statistics 4B. Population-based 5B. Active Dual method or
registration survey, with Indirect Surveillance of capture-
sisterhood-asking reproductive age recapture,
1C. Sample vital
about deaths of sisters, female deaths corrects for
registration with
without dates under-reporting
verbal autopsy
4C. Population-based
survey, with Direct UN models:
sisterhood-asking estimate
about deaths of sisters, matenal
with dates mortality using
4D. Sampling at regression
Service Sites (SSS)-
using Direct
sisterhood methods
Composite approaches - Reproductive Age Mortality Studies: In-depth review of reproductive-age

female deaths identified from routine and/or special opportunities, and follow-up investigation of maternal deaths
Figure 1: Options (methods) for measuring maternal mortality in developing countries.

Source: Graham et al. BMC Medicine 2008 6:12, doi:10.1186/1741-7015-6-12 [39]

1.4 Historical overview of policy and progress in maternal mortality
1.4.1 Early experiences and lessons learned
Evidence shows that even before modern medical interventions such as antibiotics, blood
transfusions, and caesarean sections were available, effective measures were in place that
reduced maternal mortality in northern European countries such as Sweden, Norway,
Denmark, and the Netherlands. For example, in 1900, the MMR in Sweden was 230 per
100,000 live births, compared to 700 in the US and 400 in England and Wales [61]. In
Sweden, individual health data (births, deaths, marriages, and migrations) have been
registered since 1749, building on pre-existing registrations of births and deaths through
churches. The Swedish Health Commission issued the first national maternal mortality report
in 1751: an MMR of 900 per 100,000 live births. This information system enabled a
recognition for the possibility of avoiding 400 out of every 651 maternal deaths (“avoidable
maternal death”) if trained midwives were available for pregnancy and birth care [62]. As a
result, the Swedish Government responded by training and deploying midwives in every
village, thereby strengthening the information system. This resulted in an MMR of 230 per
100,000 births in 1900. Other developed countries learned that the good experience in
Sweden was the result of an availability of midwives, and subsequently implemented
midwifery training and licensing.
Consequently, in most developed countries the years 1950-1960 marked a uniformly low
level of MMR (<100 per 100,000 live births) [61]. The decline in that period was associated
with the invention of modern medical technologies such as antibiotics, blood transfusions,
and caesareans in addition to a focus on midwifery care. Still, technologies have not been the
only factors responsible for the effective decline in maternal mortality, as other important
elements have helped these countries. On the one hand, there were social movements that
constituted medical professionals and women’s rights groups that seriously campaigned to
bring maternal health into the focus of governments.
For example, in England, confidential inquiries into every maternal death and subsequent
improvements helped to see substantial reductions in maternal deaths [63]. On the other hand,
the availability of information, professional commitments, and policy and technical supports

for maternal health have helped bring an early reduction of maternal mortality in some
developed countries, while causing delays in others [54]. The governments of northern
European countries were successful because of their timely response to the information they
had, as well as the public pressures they faced for a reduction of maternal mortality [64].
1.4.2 The 1980s to the present: from Alma Ata to the MDGs
In developing countries, a high maternal mortality continued for decades in part because of a
lack of information until the late 1980s when the international community begun to
understand the problem. Accordingly, in 1978, the WHO and UNICEF sponsored the
International Conference on Primary Health Care (PHC) in Alma Ata [65]. Countries made
statements to address the social, economic, and political causes of ill health, in addition to
health service delivery. The aim was to provide basic health services that were affordable,
accessible, acceptable, and useful for poor people. However, this novel initiative later shifted
to selective interventions that depended on technological solutions such as child
immunization, growth monitoring, and family planning [66].
The shift resulted in “vertical” programmes such as family planning, and lacked a meaningful
intervention that addressed obstetric (clinical) causes of maternal deaths. As a response to a
lack of information, the WHO, with funding from UNFPA, supported the first community-
based study in 1985 on maternal mortality. The findings of the study and information from
vital registration estimated that approximately 500,000 maternal deaths occur each year in the
world, of which 99% was in developing countries [67]. Halfdan Mahler, the then WHO
director-general, explained that the “main reason for such a striking gap of maternal
mortality rate between developed and developing countries was that until lately the size of the
problem was not known”. Subsequently, he explicitly called for the importance of
information as “sound estimates based on new data are thus the foundations for our current
understanding of the concern’’ [4].
In 1987, the WHO, UNFPA, and the World Bank jointly sponsored the first international safe
motherhood conference in Nairobi, Kenya. The conference marked the formal initiation of
the Safe Motherhood Initiative (SMI), aiming to reduce maternal mortality by 50% in 2000
[68]. At that time, people perceived that maternal mortality was a “neglected tragedy”
compared to attention given to child health, in which Allan Rosenfield and Deborah Maine
produced a powerful article entitled ‘‘Where is the M in MCH?” (the M refers to “maternal”

in MCH, maternal and child health) [69]. In 1989, the heads of states and representatives of
influential organizations attended the World Summit for Children in New York, in contrast to
the small amount of attention given to the maternal conference in Nairobi. In the New York
conference, maternal health was addressed in the context of helping child survival as opposed
to the independent agenda for the mothers [54].
Meanwhile, maternal health advocates considered maternal mortality as a human rights issue,
and campaigned for government accountability and proposed broader approaches such as
reproductive health and rights [54]. Even so, because of the donors’ need to see the measured
effects of the money spent, two components of safe motherhood elements (antenatal
screening and TBA training for delivery attendance) got more attention in the first decade
following the SMI initiation [70]. Nevertheless, in subsequent years it was widely recognized
that both of these strategies had little effect on reducing maternal mortality, and diverted
attention from midwifery care and hospital emergency obstetric services [70-73]. The reason
why antenatal screening and TBA delivery had little effect was that the risks of maternal
mortality are concentrated around labour and delivery, in which antenatal screening cannot
capture these risks. In addition, TBAs had limited skills and technologies to tackle major
killers such as bleeding, eclampsia, infection, and obstructed labour [68]. For this reason, the
500,000 global maternal mortality rates remained for a long period of time. In the meantime,
experts argued for skilled attendance at birth and emergency obstetric care as a critical
strategy to reducing maternal mortality [74, 75], which was later adopted as a key strategy for
MDG5. However, as positive contributions, the SMI helped to increase the commitments of
organisations of health professionals such as nurses, midwives, and gynaecologists for
maternal health [54]. Additionally, non-governmental organizations also contributed and
learned important lessons by working with communities in developing countries [76].
Subsequent to this, the 2000 MDG declaration was considered to be the greatest political
attention that maternal mortality has received, in which 189 heads of states agreed on eight
Millennium Development Goals (MDGs), of which the MDG5 was committed to a reduction
of maternal mortality [77]. The goal of winning global policy attention was to motivate
resource (financial) commitments and health systems strengthening. Yet, studies showed that
there is a large resource gap between what is needed and what is available to help reduce
maternal mortality [78-80]. One study suggested a an annual average increase of 3.9 billion
USD over 10 years to meet combined maternal and new-born health needs [80].

Unfortunately, maternal mortality reduction was not a priority agenda for years in many
developing countries [81], so the general feeling among the maternal health community was
that maternal mortality instead remained an orphan agenda [82-84]. As such, although
general estimates show that global MMR has been declining, the pace is slow and there is no
concrete evidence of acceleration, which runs the risk MDG5 being the least likely achieved
MDG goal by 2015, especially in sub-Saharan Africa [85].
Shiffman and Smith summarize the determinants for a relatively low amount of attention
being given to safe motherhood [86]:
1) Actor power ─ the power of individuals and organizations that deal with maternal
problems;
2) Idea power ─ the way the problem is understood and portrayed to the public and policy
makers;
3) Political context ─ the political environment in which the actors operate; and
4) Issue characteristics ─ the feature of the problem (magnitude, severity, easy solutions).
Maternal mortality received less attention because of a relative rarity in absolute numbers
compared to the high number of deaths from tuberculosis, malaria, AIDS, and child diseases
that compete for policy attention and resources. In addition, it also lacked powerful actors
(leaders) and convincing ways of delivery of message about the problem [86].
1.5 Determinants of maternal mortality
1.5.1 Theoretical framework
The survival of a pregnant woman depends on a diversity of complex factors, and not merely
health services. It starts from the conditions in the time of her own birth to the environment
she was brought up in including nutrition, diseases, culture, education, and mental
satisfaction. Whether her pregnancy is wanted and healthy or not, also affect the outcome.
Furthermore, the socio-economic factors are also important determinants of maternal health
because maternal mortality is associated with factors such as education, economic
backgrounds, and sex discrimination (a lack of reproductive autonomy) [87]. Poor women are
less likely to have a formal education, less power on economic decisions, and are less likely
to receive maternal care [88].

James McCarthy and Deborah Maine [89] give the theoretical framework for the
determinants of maternal mortality in three categories:
1) Distant (socio-economic and cultural) factors such as woman's status in the family and
community, her family’s status in the community, and the development of her community
2) Immediate determinants such as a woman’s health status, reproductive status (age,
parity, marriage), access to health services, and utilization behaviour
3) Pregnancy outcomes (complications)
Thus, efforts to reduce maternal mortality need to focus on: 1) reducing the chances of a
woman becoming pregnant; 2) reducing the probability of a pregnant woman developing
complications; and 3) improving the outcomes for women with complications [89].
Figure 2 below describes the relationship between the three determinants of maternal
mortality:

Distant determinants Immediate determinants Outcomes
Health status of women

Pregnancy
Women's status in x nutrition (anaemia, height,
family and community weight)
x Infections (malaria,
x Education
hepatitis, TB, HIV)
x income
x Chronic diseases
x occupation Complications
(hypertension, diabetes,
x social and legal cardiac diseases) x Haemorrhage
autonomy
x previous pregnancy x Infection
complications x pregnancy-
induced
hypertension
Reproductive status x obstructed
x age labour
Family status in x parity x uterine
community x marital status
x family income
x education of others
Access to health services
x occupation of
others x Location of services (family
x Land planning, antenatal care,
emergency obstetric care, Death/disability
and other primary cares)
x Quality and diverse care
x access to information
about health services
Community's status
x aggregate wealth Health care behaviour
x Community x use of family planning
resources ( doctors, x use of antenatal care
midwives, x use of skilled care for
ambulance) delivery
x Driveable road x use of harmful traditional
practices
x use of illicit induced abortion
Unknown or unpredicted factors
Figure 2: A framework for analysing the determinants of maternal mortality and morbidity.
Source: McCarthy J, Maine D (1992), reference [89]

1.5.2 Access to health care or poverty: Which matters most for maternal mortality ?
Research findings demonstrate the links between poverty and maternal mortality, i.e.
maternal mortality is higher among women in lower socio-economic classes [90].
Nevertheless, poverty does not always explain the level of maternal mortality. Historic
evidence from western countries suggest that the overall standard of maternal care is more
important than poverty in determining the level of maternal mortality [91]. There were
exemplary community-based house-to-house maternal care services, which remarkably
reduced maternal mortality in people of a low socio-economic condition in developed
countries. These services include the midwifery service of the “Queen’s Institute of Nursing”
in England and Wales from 1920s to the 1940s, and US midwives in the “Kentucky Frontier
Nursing Service” from the 1920s to the1930s [91]. In the communities where these services
are provided, maternal mortality exhibited a marked decline compared to mortality levels in
better-off social classes who received hospital services given by physician services (MMR of
66 vs. 800-900) [61, 91].
As such, mere access to health services in which modern technologies are available, does not
guarantee good maternal outcomes. Other evidence also shows that in Britain in the years
from 1870 to the 1940s, interference of labour by physicians in terms of an overuse of
forceps and chloroform has been suggested to have caused higher maternal mortality rates
among higher social classes [92].
In some developing countries as well, better commitment to health service provision resulted
in good outcomes. Evidence from developing countries showed that despite an equally low
level of economic development, some countries achieved a low level of maternal mortality,
while others with a similar economy had up to 10 times a higher mortality. For example, in
the 1990s the gross national product (GNP) in terms of purchasing power parity (PPP) for
Vietnam, Lesotho, Central African Republic, and Nepal was between 1,000 and 1,200 USD
(almost similar), although their MMR varied from 160 in Vietnam to 1,500 in Central African
Republic (CAR) [61].
In general, poverty plays an important role by denying availability and access to health
services that put poor mothers at a greater risk of death without any prioritized action to make
health services accessible to the poor [93]. Consequently, a study done in over 50 countries

shows that more than 80% of births for the richest women is attended by skilled assistants,
while only 34% of poorest women received the similar service [94]. In conclusion, access and
utilization of standard maternal care have the potential to reduce maternal mortality,
irrespective of household or national poverty levels.
1.6 Interventions for maternal mortality reduction
The knowledge of what works in reducing maternal mortality has two components: first, the
knowledge of what interventions work to prevent a particular woman from dying (single
intervention or packages of interventions for a single woman) and second, what strategies
work to distribute these proven interventions to the public to help reduce maternal mortality
at the population level [47].
1.6.1 Single or packages of interventions
Interventions proven to increase the survival of individual mothers include: infection

prevention, antibiotics, blood transfusions, anticonvulsants, drugs that enhance the uterine
contraction to reduce bleeding, and operative interventions (caesarean sections and repair of a
ruptured uterus) [95]. Packages of interventions used to save the life of a woman suffering
from severe bleeding include skills such as the manual removal of retained products,
administering oxytocic drugs, blood transfusions, and caesarean sections. Because of this, the
introduction of successful interventions (especially antibiotics, caesarean sections, and blood
transfusions), and an effective public health strategy (midwifery care), helped to reduce
maternal MMR from more than 1,000 to less than 10 per 100,000 live births in developed
countries [96].
1.6.2 Strategies to distribute proven interventions to the public
There is no single strategy to supply these proven interventions to many mothers who need
them. In addition, translating what works in one place to another is complex because of the
diverse nature of area contexts and the multiple determinants of maternal health. Thus,
merely ensuring the availability of essential maternal care at health facilities may not

guarantee the reduction of maternal mortality. Other factors such as a low utilization of the
services and over-medicalization (iatrogenic factors) may also play a role [97].
Maternal mortality reduction will work by using strategies that combine proven interventions
with distribution mechanisms that achieve a high coverage in the community [96]. To achieve
a high coverage, a given intervention should be distributed through several strategies, e.g.
contraceptive pills through health facilities, community health workers, social marketing,
misoprostol for severe bleeding through community health workers and health facilities
instead of restricting these to only health institutions. Instead of sticking to only facility-based
care [96], skilled intrapartum care at home, health centres, and hospitals, are some of the mix
of distribution strategies for a better achievement.
Accordingly, some developing countries that implemented effective distribution strategies for
these interventions have rapidly reduced their maternal mortality [98]. For instance, Thailand,
Malaysia, and Sri Lanka halved their MMR in less than 25 years [98-101]. In other
developing countries a fast reduction of MMR has been registered over the period of less than
a decade; Egypt reduced its MMR by 50% between 1992/93 and 2000 [102], whereas
Honduras decreased by 40% between 1990 and 1997 [103].
However, some strategies used in developing countries to reduce maternal mortality were less
effective and had a limited impact. Examples include community-based primary care, such as
antenatal screening and the use of traditional birth attendants (TBAs). Primary care strategies
can help to treat acute child diseases such as acute diarrhoea by oral rehydration, but it is still
difficult to manage obstructed labour or severe the bleeding at the remote area [104]. No
amount of antenatal screening can predict bleeding, infection, and high blood pressure that
will occur during labour and delivery, and TBAs are not able to manage these acute
conditions [105]. Community strategies that provide emergency obstetric care, such as health
centre-based intrapartum midwifery care backed by comprehensive care at hospitals, is
currently underscored [106].

1.6.3 Skilled birth attendance (SBA) and emergency obstetric care (EmOC)
The current globally recommended strategy to reduce maternal mortality is to provide access
to basic emergency obstetric care. These are primarily intrapartum care strategies focusing on
health centre-based skilled attendance at birth backed by a referral mechanism to connect to
comprehensive obstetric care at hospitals. The World Health Organization (WHO), the
International Confederation of Midwives (ICM), and the International Federation of
Gynaecology and Obstetrics (FIGO) jointly define a skilled birth attendant as follows: “A
skilled attendant is an accredited health professional ─ such as a midwife, doctor, or nurse ─
who has been educated and trained to proficiency in the skills needed to manage normal
(uncomplicated) pregnancies, childbirth, and the immediate postnatal period, and in the
identification, management, and referral of complications in women and newborns” [74].
Using skilled attendants, health centres are required to provide seven signal functions of basic
emergency obstetric care (BEmOC) recommended by the WHO [107].
These signal functions are:
1) Assisted virginal delivery;

2) Manual removal of the placenta;
3) Giving uterotonic drugs such as parenteral oxytocin, ergometrine and misoprostol
(medicines that make uterus contract and prevent bleeding);
4) Administration of anti-consultants such as magnesium sulphate for women diagnosed
with pre-eclampsia and eclampsia;
5) Removal of retained products using manual vacuum (MVA) aspiration or dilatation and
curettage (DIC);
6) Administration of parenteral antibiotics; and
7) Newborn resuscitation.
In addition, hospitals should be available for referrals, and be able to provide comprehensive
emergency obstetric care (CEmOC). CEmOC is the provision of all the seven functions
mentioned above as well as two advanced functions:
1) Caesarean section (operative delivery); and
2) Blood transfusion.

An accessible health centre that consistently provides these signal functions, using
professionals with midwifery skills backed by referral systems to CEmOC, can help reduce
maternal mortality [96].
Misoprostol is another important intervention proposed for the prevention and treatment of
postpartum haemorrhage (PPH) in remote areas. Studies have shown that the drug is
clinically non-inferior to oxytocin (the current standard treatment), and can be distributed to
mothers using community health workers [108, 109]. Moreover, misoprostol has two
important advantages over oxytocin for use in rural areas: first, it does not require
refrigeration for storing; and second, it has oral and sublingual forms that do not need a
skilled person to administer unlike injections [110]. However, the fear of its potential abuse
for purposes such as abortion and induced labour limits misoprostol’s wider use for PPH
[111].
1.6.4 Access (availability, acceptability, and affordability) to obstetric services
Strategies designed to distribute emergency obstetric services to a larger community of

people in the rural areas of developing countries are problematic and often achieve low
coverage. The principal factors are a shortage of skilled staff and equipped facilities for the
large proportion of the population [112, 113]. Furthermore, the experience by women of
services provided during facility delivery can define their future use [114]. As such, because
of a low acceptability of services to the community, existing services are not effectively used.
Women do not like to receive facility delivery services that are disrespectful, and in which
their privacy concerns are not addressed [114]. In some places, mothers and the community
perceive delivering at home as normal, and may not see the importance of going to a facility
for childbirth [115].
Moreover, transportation is limited, and in many places getting to health facilities may be
“too far to walk” [116], in addition to financial constraints [117-119]. Financial barriers are
not only the fees for services, but also include a demand for side costs (non-facility costs).
Such costs outside of the health service may become as high as half the cost of normal
delivery, and introducing fee-free delivery service cannot change these non-service costs [76,

120]. Therefore, accessing and utilizing maternal services is affected by several factors, such
as an availability of service, acceptability of the standard (quality and ethical handling) of
service, and whether the users can afford to pay. Consequently, these access barriers may
lead to either a delay in seeking skilled care or abandoning the services.
1.6.5 The three delays
Timely and quality treatment has the potential to avoid most maternal deaths. However, in
resource-limited settings, delays before receiving health services are common. The three
delays related to maternal health service utilization are [116]:
1) Delay at home before making decision to seek health care. This depends on an awareness
of the benefits of service, the availability of financial sources, and family care.
2) Delays on the way before arriving at health facilities (between the decision and reaching
the health facility), in which transport access and road conditions determine the speed.
3) Delays in health institutions before receiving adequate care (health provider delay), in
which the availability of supply, staff attitude, and rules and procedures determine timely
care.
1.7 Neonatal mortality and its relation to maternal outcomes
Neonatal mortality (newborn deaths in the first four weeks after birth) is a major public
health problem, and its reduction depends on the quality of the health-care system. In 2009,
an estimated 3.3 million neonates died in the world. Africa had the highest rate of death, and
the slowest progress in terms of reduction [121]. The aforementioned amount is a sign of
decline in neonatal mortality from 4.6 million in 1990. Nonetheless, because of a slow
decline in neonatal deaths compared to an overall decrease in child mortality, the proportion
of child deaths during the neonatal period has increased in all parts of the world. For instance,
the neonatal share of child mortality increased from 37% in 2000 to 41% in 2008 [122].
Most newborn deaths (75% in 2008) occur in the first week of life, with the largest
proportion taking place within 24 hours following birth (also known as early neonatal death).
This highlights the critical importance of clinical care for the survival of the high-risk

newborn babies [122]. The main causes of neonatal mortality are preterm birth, severe
infections, asphyxia, and neonatal tetanus [123]. Fortunately, there are effective solutions
such as antibiotics, oxygen and assisted respiration, and incubation techniques that help
tackle these deaths. The main challenge here is how to make available these proven
interventions to the poor people in developing countries. For example, access to incubation
and infection prevention and treatment techniques that help the survival of at-risk babies in
advanced countries are limited in developing countries.
The health of newborn babies is also associated with the well-being of their mothers, which
determines the level of feeding and the overall care the babies receive [124]. The implication
of the association between maternal health and neonatal survival is that an integrated
approach for the care of mothers and newborns is critical.
2 Maternal and neonatal health in Ethiopia
2.1 Ethiopia: The country
Ethiopia is a large county, with a total area of 1.1 million square kilometres in the eastern part
of Africa. It shares borders with Sudan, south Sudan, Kenya, Somalia, Djibouti, and Eritrea.
Ethiopia has the second largest population (88 million in 2014), after Nigeria [125], and the
country is one of the ancient civilizations and is the oldest independent country in Africa.
Even so, most of the population live in rural areas where health and other services are
difficult to access. Economically, agriculture is the base of the economy (47% of the GDP
share in 2013), which accounts for 85% of employment. Currently, there are pushes for
diversifying the economy into manufacturing, textiles, and hydropower energy [126]. In
recent years, the country has achieved a fast economic growth. However, according to the
2014 reports of the Human Development Index (HDI) [127] and the Multi-dimensional
Poverty Index (MPI) [128], Ethiopia remains among the poorest countries in the world.

2.2 Health services in Ethiopia
Medical treatment was introduced to the country during the reign of Emperor Libne Dingel
(1508-40) through foreign missionaries and travellers [129]. Later, Menelik II and Emperor
Haile Sellassie promoted treatment in the same way. Haile Sellassie established the Ministry
of Health, the first national health service, in 1947, and subsequent to this, the College of
Public Health was opened at Gondar Town in 1952 with support from the WHO, UNICEF,
and US-AID [129]. The Gondar Public Health College started to train a team of the first
Ethiopia health professionals, which was comprised of public health officers at the degree
level and community nurses and sanitarians at the diploma levels, and assigned in teams to
health centres throughout the country.
Following the military overthrow of the Haile Sellassie government in 1974, the overall
economic and social conditions deteriorated, and the plan to expand primary health care
principles to the rural majority mostly failed. In 1991, the current Ethiopian Peoples’
Revolutionary Democratic Front (EPRDF) government came to power and devised a new
health policy that promoted a priority to health promotion, disease prevention, and equitable
health service distribution to the areas with poor health service [130].
The current Ethiopian health service delivery system has four tiers (primary health-care units,
district hospitals, zonal hospitals, and referral hospitals) [131]. The primary health-care unit
is a combination of five satellite health posts and a health centre. Health posts are the lowest
level, two-room buildings, which serve as offices and treatment places constructed in villages
for an average population of 5,000 people, and are staffed by HEWs. Health centres provide
curative and preventive services for approximately 25,000 people, with staff composition of
health officers (people with four-year clinical and preventive health education at a
university), nurses, midwives, laboratory technicians, and pharmacy technicians. Hospitals
have medical doctors, in addition to other professionals with the specialty depending on the
status of the hospital.
The Ministry of Health follows the work of Regional Health Bureaus and specialized referral
hospitals. Regional Health Bureaus supervise regional referral hospitals and zonal health

departments; Zonal health departments follow zonal hospitals, district hospitals, and district
health offices, while district health offices are responsible for the proper functioning of
primary health-care units.
Since 2003, the year the Health Extension Program (HEP) was implemented, the outcome of
the current policy has been widely appreciated. The HEP is based on placing two women,
trained for one year in general health, in every village (with an average population of 1,000
households) in the country. As a result, Ethiopia made remarkable reductions in child
mortality, and achieved an MDG4 target in 2012 [132]. What makes Ethiopia’s HEP unique
from community health worker systems in other countries is that the HEP receives an
important amount of attention from the government. Approximately 30,000 rural HEWs
(38,000 HEWs, including the urban version that followed later), were trained and deployed in
five years. The government pays a salary, in addition to an established system of training,
career, and supervision [133].
Furthermore, two other strategies of public mobilization linked to the HEP play an important
role. The first strategy is a model family initiative, in which HEWs identify, train, and
graduate families assumed to have adopted a healthy living behaviour. The second is the
creation of the Health Development Army (HDA), a network of women who volunteer to
identify bottlenecks in the use of essential health services within the community, and find
local solutions for the problems by coordinating the efforts of the people [133].
2.3 Maternal and child health programmes in Ethiopia
According to the UN organizations (WHO, UNICEF, and UNFPA) and the World Bank joint
estimate, MMR in Ethiopia decreased to 350 in 2010 from estimated 950 per 100,000 live
births in 1990 [134], while the mortality ratio of children under-five years of age was reduced
by two-thirds, from 204 to 68 per 1,000 live births between 1990 and 2012 [135]. However,
the maternal mortality reports often conflict with each other and have wide confidence
intervals, hence causing difficulty in observe the progress. For example, the DHS reported an
MMR per 100,000 live births of 671 in 2005 and 676 in 2010, which is an unchanged rate.
Moreover, according to the 2008 global maternal mortality estimate by the Institute of

Health Metrics and Evaluation (IHME), Ethiopia is one of the six countries that contributed
to more than half of the maternal deaths in the world; the other countries were Afghanistan,
Pakistan, India, Nigeria, and Democratic Republic of Congo [7].
Promisingly, however, the Ethiopian government puts a strong emphasis on the reduction of
maternal mortality in its strategic health sector development plan (HSDP-IV), which is the
fourth phase for 2010-2015 out of a 20-year vision [136]. The aim of the strategic plan was to
strengthen skilled birth attendance, family planning, antenatal and postnatal cares, and
emergency obstetric care. Emphasized interventions include an accelerated training of
midwives, and improving the capacity of health extension workers. The accelerated training
of midwives was comprised of two parts: Many who join regional health science colleges for
three-year midwifery diploma training, and other who join universities for a four-year
midwifery education. In addition, recognizing the skill gap of the HEWs for intrapartum care,
the Ministry of Health began to upgrade HEWs to diploma level, with additional one-year
training in midwifery, starting 2011.
Moreover, the implementation of a plan to accelerate the expansion of district hospitals to

provide comprehensive emergency surgery is currently taking place. The new expansion aims
to build 800 district hospitals, one for every district of 100,000 to 150,000 people. To respond
to human resource shortages, two new programmes have been started: 1) Speciality training
in surgery for non-physician clinicians (NPCs) who are health professionals with a bachelor’s
degree before joining the programme for three years 2) Four-year accelerated training of
health professionals with bachelor’s degree as medical doctors. In addition, there is an
increased specialization of medical doctors.
Through these efforts, the Ministry of Health aims to cut the MMR per 100,000 live births to
a level between 100 and 260 in 2025, further reducing this to a range of 45 to 53 in 2035
from an assumed MMR of 420 in 2014 [137].

2.4 Rationale of the study (the study in context)
Studies consume resources, and should be assessed for their importance (relevance). In other
words, we should be able to answer questions such as: Is the topic of investigation really a
problem?, Are there solutions to the problem?, Is the information obtained useful for any
short- or long-term effects of the problem?, Can the study be applied in places other than the
particular study setting? Regarding the aforementioned concerns, this thesis attempts to look
into two key issues related to MDG5 targets. These are: 1) measuring the level of maternal
mortality (the agreed indicator of progress towards the goal) and 2) assessing emergency
obstetric services (the strategy emphasized to attain the goal). We conducted the studies
under this thesis in areas in Ethiopia where information on these subjects was not well
documented.
Maternal mortality is a great problem in developing countries; especially sub-Saharan Africa

where Ethiopia is no exception. The reasons for high mortality are: 1) limited information
about the problem for policy makers to give a priority to reducing the problem; and 2) a poor
quality and low utilization of existing health services for pregnancy and childbirth care. The
aim of MDG5 is to reduce the 1990 level of maternal mortality by 75% in 2015. Towards
that end, the internationally emphasized strategy is emergency obstetric care. Unfortunately,
with a few years remaining for the deadline of the MDG target when we conducted the
studies, there was no sustainable community-based, real-time data to oversee progress in the
level of maternal mortality and the status of skilled obstetric care in Ethiopia in general, and
in south Ethiopia in particular.
From a practical perspective, measuring maternal mortality is difficult everywhere in the

world and there is no universally agreed upon single method for obtaining perfect data [42].
However, the standard method (referred to as the gold standard) used by the developed
countries is the regular registration of population events (births, marriages, population
movements, and deaths) known as vital registration. In developing countries, vital
registrations are rare and incomplete. The reason for a lack of registry-based data on births
and birth outcomes is that the large proportion of population in these countries live in rural
villages where health service and household information is difficult to access [138].

As a result, sample based household surveys that ask family members and relatives about
deaths that took place several years ago were widely implemented. However, these small-
scale surveys were not able to permit international and national estimates to help compare
regions and countries. As such, another source of information, the statistical modelling of
proxy data for national and international estimation, is widely practiced [5, 7]. The modelled
data are important in providing an overview of the problem at the national level, though often
with controversial results [57]. Moreover, they mask the local variations in mortality, and
limit the efforts to establish sustainable data sources at local levels.
Ethiopia adopted the Health Extension Programme (HEP) as the Primary Health Care (PHC)
strategy by training and deploying HWEs in every village (kebele) starting in 2003, which
increased access to basic preventive and promotive health services. Consequently, Ethiopia
registered remarkable health outcomes, such as an increased uptake of methods of fertility
control and a reduction in child mortality that achieved the MDG4 target before the 2015
schedule [132]. Nevertheless, using that opportunity as a source of sustainable registry-based
data to monitor births and birth outcomes has not been explored. Fortunately, Ethiopia is
preparing towards the universal registration of vital events with the approval of a new
national legislation in 2012 [139], and the findings in this thesis may be used as a learning
step for implementation. The thesis reports the prospective community-based birth and birth
outcome registration as a feasible tool to measure maternal mortality from a large rural
community study using the Ethiopian HEP as an opportunity (Paper I). Given the universal
presence of HEWs in all parts of Ethiopia, as well as sustainable and timely nature of data
obtained through a prospective registration of births and birth outcomes, it can be applied
throughout the country and in other developing countries with a similar community health
system.
Yet due to the inherent problem in the complexity of measuring maternal mortality because
of its under-reporting problem, the use of mixed methods is recommended to reach sound
estimates [39]. Accordingly, with the aim of supplementing and comparing results from the
birth registry with alternative methods, we used two additional surveys: A large household
survey to measure maternal and neonatal mortality (Paper II) and the sisterhood method,

which is considered as simple and cheap for resource-limited settings. Several Asian and
African countries applied the sisterhood method to estimate maternal mortality indicators
[140, 141]. Still, it had a limited use in Ethiopia, and we did not know how a result from the
sisterhood study compared to findings through other community-based methods (birth
registry and household survey). To that effect, the thesis also applied the sisterhood method
(Paper III) by including the simplified indirect sisterhood questions into a large household
survey, as recommended by the developers of the method. Observations into inequalities
(socio-economic and service-access variations) associated with mortality outcomes provide
important information to help plan targeted interventions. However, such information is
scarce in the study area and in the country at large. Two papers in this thesis (Papers I and II)
attempt to look at selected factors associated with the maternal and neonatal mortality.
Most maternal deaths are preventable, and there are effective solutions to cut the problem.
Nevertheless, we do not know the status (population coverage, quality, and utilization) of the
UN-emphasized emergency obstetric care (EmOC) in the study area. Measuring maternal
mortality without assessing the status of obstetric services does not give a complete image of
maternal health conditions. To achieve that effect, the WHO prepared a universal guideline
to help measure the status of EmOC and recommended the minimum standards for the
service. Using the UN tool, we assessed the status (coverage, quality, and utilization) of
EmOC in a population of nearly two million people (Paper IV). In addition, the studies in the
thesis provide estimates of neonatal mortality (Paper II), and describe that neonatal survival
was associated with maternal outcomes. Furthermore, we report socio-economic inequalities
associated with mortality. As such, information in the thesis has important significance in
helping understand the current level of mortality, to planning an improvement of the obstetric
service, and to monitor future progress. Additionally, achievability of a high-coverage birth
registry is an important lesson for future efforts in obtaining real-time, country-owned,
locally available data.

3 Objectives
3.1 General objective
The overall aim of the thesis is to measure and compare maternal mortality using four different
methods, and assess the coverage and quality of obstetric services in south Ethiopia in 2010.
3.2 Specific objectives
I. To measure maternal mortality through establishing a community-based birth registry

in all rural villages in three districts (Paper I);
II. To measure the maternal and neonatal mortality and socio-economic inequalities in
mortality outcomes using a household survey (Paper II);
III. To estimate maternal mortality indicators using the sisterhood method (Paper III);
IV. To assess the status (coverage, quality, and utilization) of emergency obstetric
services and maternal mortality through a review of health facilities (Paper IV)

4 Methods
4.1 Study area and setting
The Federal Democratic Republic of Ethiopia has nine ethnic-based regional states and two
city administrations. Regions are subdivided into zones (provinces), zones into woredas
(districts) and woredas into kebeles (villages). A kebele is the lowest administrative unit in
Ethiopia, with an average population of 5,000 people (equivalent to 1,000 households),
whereas a woreda is a group of about 20-70 kebeles. Papers in this thesis used studies done in
two zones of the Southern Nations Nationalities and Peoples' Regional State (SNNPRS) in
Ethiopia (Gamo Gofa and Segen Area People’s zones); Figure 3 shows a map of the area:
Figure 3: The map of the study area in Southern Nations Nationalities and Peoples Regional State
(SNNPRS) in Ethiopia.

Gamo Gofa is one of the remote areas in Ethiopia with Arba Minch, the central town, located
505 km south of Addis Ababa. The Segen Area Peoples' zone is also a similar adjacent zone
with Segen, the centre, being 575 km from Addis Ababa. In 2010, the Gamo Gofa zone had a
population of 1,740,828 people living in two town administrations and 15 woredas, while the
Segen Area Peoples' zone had 636,794 people in five woredas [142]. The birth registry study
(Paper I) was conducted in three woredas, two in Gamo Gofa (Bonke and Arba Minch Zuria)
and a third (Derashe) in the Segen Area Peoples' zone. Two studies (Paper II and III) were
conducted in the Bonke woreda in Gamo Gofa. One study (Paper IV) was conducted in all
three hospitals and 63 health centres in Gamo Gofa.
Bonke has a health centre in Geresse, the central town, and three new health centres in
Gezeso, Shalakaye, and Dembile. However, no health institution provided comprehensive
emergency obstetric care, and there were no doctors when we collected the data. The nearest
comprehensive EmOC service was at the Arba Minch Hospital, which is 70-120 km
depending on which village you are driving from in Bonke. However, most of the Bonke
population live in inaccessible mountainous villages.
The Arba Minch Zuria woreda has two distinctly different populations: Most of the
population live in mountainous highlands with poor roads and limited health service access
from the health centres in Maze-Doysa and Zigiti. By contrast, the lowland population live
closer to an asphalt road and closer to a relatively better-equipped health centres in Lante and
Shele, as well as Arba Minch Hospital.
Derashe is 60 km from Arba Minch. The central town of Gidole has a district hospital with a
maternity waiting area, where mothers with high-risk pregnancy can wait until delivery [143].
In addition, Derashe has four health centres in Gidole, Gato, Holte, and Busa.
The Health Extension Programme and the agents
In 2003, Ethiopia started the community-based health extension programme by training the
health extension workers (HEWs). The first batch of the HEWs were deployed in 2004, and
over 38,000 HEWs are currently working in the country (30,000 rural and 8,000 urban) [144].

The work of HEWs focuses on health promotion and disease prevention through regular
home visits in their catchment area. In addition, they give a prioritized follow-up to
households with pregnant women, newborn babies, and sick people. HEWs also provide
antenatal examinations and delivery services both at home and health posts. Five to 10
laywomen from sub-kebeles, known as volunteer health promoters (VHP), assist the HEWs.
VHPs receive a few days general training from the Woreda Health Departments, and work to
provide timely information and organize schedules for HEWs [145].
4.2 Study designs and data collections
Table 2 below presents the summary of the design, the participants, and the period of the
studies. For Paper I, we prospectively registered birth and birth outcomes involving over 200
HEWs in 75 kebeles in three woredas, covering a total of 421,639 residents. Each HEW is
responsible for about 500 households. Accordingly, a HEW is expected to register six-seven
births per week. Starting in January 2010, we kept a printed registry book for each HEW.
They registered information in two copies, transferring the first copy to the Research and
Training Centre at Arba Minch Hospital, with the second remaining in the kebeles with the
book. HEWs registered most of the births on the date of birth except for a delayed
registration of births at hospitals and in places other than the villages of residence. HEWs
made follow-up household visits after delivery on the 28th day to record the neonatal outcome
(surviving or died), and on the 42nd to 45th day to learn about maternal survival. The
conventional last day of observation is day 42, but we allowed three more days to give room
for HEWs when they were busy. However, most maternal deaths were reviewed and recorded
on the day they occurred.
The study in Paper II used a cross-sectional household survey in 15 randomly selected of 30

rural kebeles in Bonke. In February 2011, data collectors (who all completed the 12th grade)
visited all households, asking if the household had births and pregnancy outcomes in the
previous five years (between January 2006 and December 2010). They completed interviews
in 6,572 households that had births and pregnancy outcomes out of 11,920 households
visited. We purposely selected data collectors from their respective kebeles of residence
because by living and participating in social events, they remember most of the births and
birth outcomes that have occurred in their kebeles. We aimed to minimize an under-reporting

of deaths, especially those of neonatals because of recall-bias and the cultural values attached
to the death of a newborn baby in the area. In the ethnic “Gamo” culture, the birth of a dead
foetus and the early death of a newborn are not publicized, and are not publicly mourned.
Only close relatives and family members are informed, while other people are told about the
incident that “something went wrong”, as expressed in local terms. This can make it difficult
for an outsider to distinguish between a neonatal death and a stillbirth. The Principal
Investigator (PI) and five college graduates supervised the interviewers.
In Paper III, we conducted a cross-sectional interview among 8,870 respondents, of which

8,503 (96%) provided complete answers. To collect the data, we used the standard questions
of the indirect sisterhood study [146] to ask both adult men and women during household
visits. A retrospective cohort analysis of risk exposure among respondents' sisters was done
using adjustment factors developed by demographic survival methods. The information about
risk exposure and the number of pregnancy-related sister deaths was used to determine the
lifetime risk of a woman’s death. By using the lifetime risk and total fertility rate, we
calculated the corresponding MMR.
For the fourth study (Paper IV), we conducted a one-year (July 2009 to June 2010)
retrospective record review in all three hospitals and 63 health centres in the Gamo Gofa
zone. We used a questionnaire developed by the WHO to monitor the coverage and quality of
emergency obstetric care in health facilities [107], and 15 graduating class health officer
students (one in each woreda) collected the data after two-days of training. The data
collectors reviewed birth registry logbooks, admission and discharge books, charts, and
monthly reports to the government. In addition, they interviewed professional and
coordinators responsible for obstetric services on the availability of important service
packages, drugs, and procedures performed in the previous three and 12 months. Data
collectors visually observed the drug and equipment stores, and recorded the availability of
items such as blood and drugs. We calculated the catchment population of the respective
facilities based on the population information used by the Zonal Health Department and
woreda Health Offices.

Table 2: Summary of study designs and participants
Paper Title Study design Participants Data

collection
I Maternal mortality Prospective 10,987 births among 421, January to

measured through birth registry 639 residents in 75 kebeles December
community -based in 3 woredas of 2 zones 2010
birth registry
II Maternal and Cross-sectional 11,762 births between February

neonatal mortality study January 2006 and 2011
measured through December 2010 in 6,572
household survey households in randomly
selected 15 kebeles
III Maternal mortality Cross-sectional 8,503 respondents reported February

indictors estimated study 22,473 sisters; resulting in 2011
through the indirect 8,068 sister-units exposed to
sisterhood method risk
IV Assessment of the Retrospective All 63 health centres and 3 November to

coverage and quality record review hospitals in Gamo Gofa December
of obstetric services were assessed for obstetric 2010
services provided between
July 2009 to June 2010
4.3 Data analysis and statistics
We used descriptive statistics to present data as proportions (education among mothers,

fathers, proportion of received antenatal care, delivered in health facilities...), ratios (MMR,
NMR), and means with standard deviations (ages, number of pregnancies, distances from
roads, health facilities). We also did a logistic regression analysis to determine the effect of
selected independent (predictor) variables on the binary dependent (outcome variables). To
test goodness of the model fit to the data, we used Hosmer-Lemeshow statistic. The
regression model had good fit to the data if Hosmer-Lemeshow had a significance value >
0.05.We reported results in terms of OR with confidence intervals (95% CI). A predictor
variable with the effect size (value), for example, OR, and a CI that did not cross the null
value of 1, was considered to have a statistically significant association with the outcome
variable (Paper II).

We performed a principal component analysis (PCA) based on 10 household-asset variables

to construct a wealth index for each of the 6,572 households studied (Paper II). For each asset
variable, we calculated a mean, standard deviation, communality, and its factor score on a
component that the variable was associated with. In the model, we included asset variables
with the highest mean score ( >0.20), as recommended by Vyas and colleagues [147]. We
used the wealth index based on the first component, which explained 20.6 % of the variations
explained by the variables, and the Eigen value was 2.06. We also used a chi-square (F2) test
to assess the relationship between the maternal mortality outcome and selected variables
(Paper I). The result is reported by a F2 value with a degree of freedom (df), and a P-value. A
significant relationship of a variable with maternal mortality was set at a P-value of 0.05 or
smaller.
In Paper III, we used a lifetime risk analysis to determine the risk of a maternal death in the
study area. We used an inflation factor to determine the final number of surviving adult-
sisters for the younger respondents (15-24 years of age). This was done by multiplying the
number of respondents in the young age groups by the average number of sisters among the
older respondents (25-49 years of age) [146]. The inflation factor was used with the
assumption that the younger respondents had sisters who had yet to reach reproductive age.
This provided the expected proportion of sisters that would have finished their reproductive
age for respondents in each age category. Thus, 90% of the sisters of respondents from 45-49
years of age are expected to have passed through their reproductive life, but only 10.7% of
the sisters of 15-19 year-old respondents. The purpose of the adjustment was to determine the
number of sister units exposed to the risk of maternal death, which serves as a denominator
for lifetime risk calculation [148].
To assess the correlation of midwife- to-population ratio against the proportion of

institutional deliveries in woredas, we used Pearson's product moment correlation test and the
result was reported by the correlation coefficient (r) and the P-value (Paper IV). We entered,
checked and analysed most of the data using the Statistical Package for Social Sciences
(SPSS-16) [149] (Papers I, II, and IV). We also used Open Epi ─ Open Source
Epidemiologic Statistic for Public Health (openepi.com) for chi-square tests and two-by-two

table analysis (Paper I). Excel spread sheets and manual calculators were also used in Paper
III (Table 3 below summarizes the main statistical methods used):
Table 3: The main statistical analysis used in the papers of the thesis
Statistical methods Papers
Logistic regression analysis Paper II

Principal component analysis (PCA) Paper II
Lifetime risk of mortality analysis Paper III
Pearson product-moment correlation Paper IV
Descriptive analysis Papers I, II, III, and IV
4.4 Ethical considerations
The Ethical Review Committee for Health Research in the Southern Nations Nationalities
and Peoples’ Regional State (SNNPRS) Health Bureau in Ethiopia, and the Regional
Committee for Health Research Ethics of North Norway (REK Nord), approved the studies
included in this thesis. In addition, the research review committee of Gondar University in
Ethiopia approved the fourth study (Paper IV). We also obtained permission from the Gamo
Gofa Zonal Health Department in writing to the Woreda Health Offices where they provided
good cooperation (ethical approvals are attached in the appendix).
In Paper I, the HEWs, as part of their routine work, are expected to visit all households in
their catchment area, with a priority to households with pregnant mothers, newborns, and sick
persons. Hence, they are expected to record vital events such as births and deaths. We further
systematized and standardized the HEWs’ routine work by printing a birth registry book,
supervising their registration, and giving feedbacks and refresher training. For the household
and sisterhood surveys in Paper II and III, we obtained informed verbal consent from the
heads of households or other adult respondents. Personal identifiers were removed from the
data, and all the data are stored in a secured computer at the Research and Training Centre in
the Arba Minch Hospital in south Ethiopia. We have not included minors in our studies.

5 Results
5.1 Maternal mortality measured through community-based birth registry

(Paper I)
We aimed to establish and assess the feasibility of birth registration in three districts, with
nearly half a million people and use the registry as a tool to measure maternal mortality.
We registered 10,987 births, 81% of the expected 13,492 births (annual CBR of 32 per 1,000
population) among 421,639 people in 2010. Out of the registered births, 90% (9,863)
delivered at home, 4% (430) at health posts, 2.5% (282) at health centres, and 3.5% (412) in
hospitals. We recorded 10,833 live births and 53 maternal deaths, (an MMR of 489 per
100,000 LBs). Among the maternal deaths, 83% (44/53) occurred at home. Five mothers
(9.4%) died during pregnancy, 21 (39.6%) during labour, and 27 (51 %) during the post
partum period. Most of the dead mothers [19 (35.8%)] had bleeding with two of them related
to abortion, 12 (22.6%) to fever, 9 (17%) to convulsion, and 7 (13.2%) with a history of
obstructed or prolonged labour.
The MMR increased if male partners were illiterate (609 vs. 346, p= 0·051), the villages had
no road access (946 vs. 410, p= 0·039), and mothers complained of illness during pregnancy
(1763 vs. 306, p< 0·0001). A validation study of house-to-house survey in 15 of the registry
villages after eight months of registration showed that births in 71·6% (1,718) of the
surveyed 2,401 households were registered with similar a MMR between registered and
unregistered (474 vs. 439). The findings of the validation study therefore helped to improve
registration coverage based on feedback discussions.
5.2 Maternal and neonatal mortality measured through household survey

(Paper II)
The aim of this study was to measure maternal and neonatal mortality and socio-economic
inequalities in these outcomes through a household survey of pregnancy and birth outcomes
in the five previous years before the survey.

We found 11,762 births that occurred in 6,572 households within five years before the survey
(average crude birth rate of 3.2% per year), including 11,536 live and 226 stillbirths. There
were 308 neonatal deaths with a neonatal mortality ratio (NMR) of 27 per 1,000 LBs (95%
CI: 24-30). We also identified 49 maternal deaths, yielding an MMR of 425 per 100,000 LBs
(95% CI: 318 - 556).
Neonatal mortality was more likely to occur in the poorest households, with an adjusted odds
ratio (AOR) = 2.84 (95% CI: 1.92-4.22); households headed by illiterates, AOR = 1.40 (95%
CI: 1.06-1.86); far away from a driveable road (≥6km), AOR =2.82 (95% CI: 1.85-4.29); and
households that had three or more births in five years, AOR= 3.35(95% CI: 2.56-4.39).
MMR was high in households in the poorest wealth quartile compared to the richest (850 vs.
250) per 100,000 LBs, OR= 3.35 (95% CI: 1.33-9.42). Households that had a maternal
mortality were 11 times more likely to have stillbirths, OR =11.6 (95% CI: 6.00-22.7), and
seven times more likely to have neonatal deaths, OR= 7.2 (95% CI: 3.6-14.3). The average
institutional delivery rate was only 3.7 % in the period between 2006 and 2010.
5.3 Maternal mortality estimated through the sisterhood method (Paper

III)
The objective of the study was to estimate maternal mortality indicators (life time risk and
maternal mortality ratio, as well as the proportion of maternal death out of the deaths of adult
females [PMDF]) through using the simple and cheap indirect sisterhood.
We analysed 8,503 of 8,870 (96%) respondents who provided complete responses. Of the
responding siblings, 5,262 (62%) were men (brothers) and 3,241 (38%) were women
(sisters). The mean age of the respondents was 26.4 (SD, 8.7) years, ranging from 15–49. The
8,503 respondents reported 22,473 sisters (an average of 2.6 sisters per respondent) who
survived to reproductive age and 2,552 died from all causes. Of the 2,552 sisters who had
died, 819 (32%) occurred during pregnancy and childbirth. This retrospective cohort analysis
provided 8,068 sister units exposed to the risk of maternal death, which served as the
denominator for calculating the lifetime risk (LTR) of maternal deaths. The lifetime risk of

death from maternal causes was 0.102 (95% CI, 0.096-0.108) or 1 in 10. Using a total fertility
rate (TFR) of 6.4 for south Ethiopia during the reference period for the estimate, we
converted the LTR into an MMR of 1667 (95% CI, 1564–1769) per 100,000 live births.
Because of the indirect nature of the sisterhood method, the time for this estimate goes back
to 1998 (about 12 years before the data collection). A separate analysis of data from male and
female respondents provided similar estimates of maternal mortality.
5.4 Obstetric services and maternal mortality assessed through health

facility data (Paper IV)
In this study, we aimed to assess the coverage, quality, utilization status of EmOC and
institutional maternal mortality using a standard tool prepared by the UN for EmOC
assessment.
There were 4,231 pregnancy- and birth-related admissions over a period of one year (between
July 2009 and June 2010) in all 66 health facilities (63 health centres and three hospitals) in
Gamo Gofa. This provided a skilled birth attendance rate of 6.6% out of a total of 64,413
births expected during the year (CBR of 32 per 1,000 population). We recorded 79 (1.9%)
maternal deaths out of 4,231 mothers admitted to the institutions. The approach through
health facility data provided an MMR of 120 per 100, 000 LBs (out of 64,413 expected
births). This shows that approximately 25% of expected maternal deaths occur at health
institutions if the MMR of 489 from the birth registry (Paper I) is considered a reference.
However, the variation in the proportion of maternal admissions and maternal mortality
between institutions and districts was large. Districts with a higher proportion of midwives
per capita, and where hospitals and health centres were capable of doing emergency
caesarean sections, had higher institutional delivery rates. We counted 521 caesarean sections
(0.8% of the 64,413 expected deliveries, and 12.3% among 4, 231 admitted to institutions).
Of the 66 health facilities, three health institutions (a hospital in Chencha, and two health
centres at Kamba and Mirab Abaya) met basic emergency obstetric care standards, and two
hospitals (Arba Minch and Sawla) satisfied the comprehensive emergency obstetric care
quality during the time of the study. These institutions served 1,740,885 people in Gamo
Gofa in 2010.

Women were most often admitted to the health institutions because of postpartum
haemorrhage (42%), obstructed labour (15%), and puerperal sepsis (15%). Remote districts
far from the capital of the zone had a lower proportion of institutional deliveries (<2% of
expected births compared to an overall average of 6.6%). Moreover, some remotely located
institutions had very high maternal deaths (>4% of facility deliveries, which was much higher
than the average of 1.9%, and the WHO minimum expectation of < 1% maternal deaths
among facility deliveries to achieve the MDG5 goal).
Table 4: Summary of maternal mortality measurements described in the papers
Paper targets At risk Maternal MMR Maternal Reference

deaths per deaths at year
100,000 home x
LB SBA
I (birth 421,639 y
registry) people 10,833 LBs 53 489 44 (83%) 6.3% 2010
I
(validation 2,401
study) households 1,698 LBs 8 474 ------ -------- 2010
II
(household 6,572 2006-
survey households 11,536 LBs 49 425 43 (87%) 3.7% 2010
8,068
1667
III sister units
(LTR of 1
(sisterhood 8,503 exposed to z
survey) siblings risk 819 in10) -------- -------- 1998
64,413 births 120

out of (from
1,740,885 births in 237*
IV (facility 66 health
people 79 the area (75%) 6.6% 2010
study) facilities
4,231 1867
Maternal (among
admissions 79 admitted) 2010
Note: SBAx = skilled birth attendance (institutional delivery), LBsy = live births
LTRz = lifetime risk of pregnancy related deaths, * estimated number of maternal deaths

Additional results (not included in the papers)
A further stratified analysis on the household survey data (Paper II) showed that both
education and wealth were independently associated with neonatal mortality. Nevertheless, a
stratification of the data revealed that education modified the effect of wealth on neonatal
mortality. In other words, when the association between wealth and neonatal mortality was
analysed separately, for educational level, wealth was associated with neonatal mortality in
only in households headed by illiterate adults, and there was no such association in
households where the heads had education. This means that in the illiterate households,
wealth may have had a positive effect in preventing neonatal deaths, but in the educated
households whether wealthy or poor may have had no effect (Table 5). However, education
had not modified the effect of distance from driveable road on neonatal mortality (data not
presented).
Table 5 (additional results): The modifying effect of education on the association between
wealth and neonatal mortality from household survey data, Bonke, Gamo Gofa, 2010
Strata Wealth Neonatal mortality (in OR 95% CI

(Education) category the household)
Yes No
Over all
(N= 6,453)
Richest 25% 31 1,322 Ref Ref
Rich 25% 124 1,685 3.14 2.12, 4.74
Poor 25% 46 1,662 1.18 0.74, 1.89
Poorest 25% 57 1,526 1.59 1.02, 2.48
Illiterate
(n= 3,766)
Richest 25% 19 907 Ref Ref
Rich 25% 99 1,254 3.77 2.29, 6.20
Poor 25% 28 805 1.66 0.92, 2.99
Poorest 25% 32 622 2.46 1.38, 4.37
Educated
(n= 2,687)
Richest 25% 12 415 Ref Ref
Rich 25% 25 431 2.00 0.99, 4.05
Poor 25% 18 857 0.73 0.34, 1.52
Poorest 25% 25 904 0.96 0.47, 1.92
Note: The household data had 6,572 cases (households) of which 6,453 had complete assets
used in the wealth index construction

6 Discussion
6.1 Discussion of the methodology (design and validity)
6.1.1 Study design and sampling
Studies included in this thesis used different observational study designs. The main study
design was a prospective birth registry (cohort) to determine maternal mortality (Paper I).
Other designs were comprised of a cross-sectional household survey that used descriptive and
analytical approaches (Paper II), a survey of a retrospective event of pregnancy-related deaths
of sisters reported by surviving adult siblings (sisterhood method) (Paper III), and a
retrospective record review in health facilities to help assess the status of emergency obstetric
services and outcomes (Paper IV).
In general, epidemiological studies have two broader designs (experimental trials and
observational studies). Experimental studies (trials) involve a random allocation of study
participants (clusters in community trials and individuals in clinical trials) into intervention
(exposed) and control (unexposed) groups before looking for outcomes. By contrast,
observational studies observe the outcomes in naturally exposed and unexposed groups
(cohort studies), examine the presence or absence of exposure in study participants who
already have the outcome (cases) and do not have outcomes (controls) or look into both
exposure and outcome at the same time (cross-sectional studies) [150].
For a particular study, the choice of the design depends on its suitability to answer a
particular research question, the cost it demands, the length of time needed for the study, and
whether a particular design can be applied ethically and logistically [151]. Accordingly, when
the fundamental objective of a study is to assess the outcome of an intervention, randomized
controlled trials (RTCs) are considered as the gold standards to determine causal
relationships between exposure and outcome [152]. This is because randomization minimizes
the problem of confounding by randomly distributing potential or known confounders into
intervention and control arms.

Conversely, observational studies are perceived to be prone to the problem of confounding

and bias. As such, evidence obtained through observational studies is considered weak
compared to information from randomized trials. Even so, there are conditions in which
randomization is unnecessary or in appropriate [153]. For example, in Paper II we aimed to
describe the effect of household variables such as wealth, distance from a road, and
educational status on maternal and neonatal mortality outcomes where randomizing
households across these variables is not practical.
Observational studies can provide useful information about incidence (cohort studies),
prevalence (cross-sectional surveys), risk factors (case-control), and health care needs of the
people. Consequently, information from properly conducted observational studies can be
useful for planning public health interventions, priority setting, and resource distribution
[154]. In addition, findings from observational studies can help as initial steps to formulate
intervention studies.
The cohort design is a natural equivalent of experimental studies because exposure happens
naturally, unlike experimental studies where the investigator assigns the exposure. A cohort
study follows exposed and unexposed groups until they develop an outcome of interest or the
study time is completed and observation is stopped (censored), in addition to a loss-to-follow
up of some participants [155]. Two directions can be used to observe cohorts from the time of
a study (retrospective ─ where information recorded before the outcome occurs is reviewed
and prospective ─ in which exposed and unexposed groups followed towards the future). A
cohort study has two purposes: describing the incidence of outcomes (descriptive) and
analysing associations between risk factors and outcomes (analytical) [156].
We used a cohort study design to prospectively follow women who terminate their
pregnancy (by birth or abortion) until maternal death occurred or the follow up ended six
weeks after the termination of pregnancy (Paper I). Due to the difficult nature of following
what happens at each household in a short interval of time for the HEWs, the study was not
concerned with time-to-event, but instead was concerned with the outcome of maternal death
or survival within a six-week period. The predictor variables such as distance from health

facility, availability and quality of road facility between households and health institutions,
maternal age, education of parents, and similar others were measured at baseline.
Active registration of births and birth outcomes is a standard procedure used by developed
countries to measure pregnancy or birth outcomes. The birth registration can help the health
system to conduct continued epidemiological studies on health services provided during
pregnancy, childbirth, and neonatal period with the aim of improving the quality. Moreover,
it can provide good data to study maternal and neonatal deaths, stillbirths, birth defects, and
related adverse perinatal outcomes, as well as the causes of these outcomes [157].
The disadvantages of the cohort studies are that they take a longer time to do (especially
prospective), are prone to a loss-to-follow up, and are expensive [155]. In the prospective
birth registry study (Paper I), mothers were observed in their usual place of residence, while
information on births that took place anywhere (at home, health institutions, or at the parents’
home) was possible to record through their home with little chance for a loss-to-follow up. As
a result, we have not lost registered births until the observation of maternal survival.
Furthermore, the fact that this study was integrated into an existing health extension system
and has not required additional salary for data collectors means it was not an expensive
exercise. However, it required the low cost of a few days of training for HEWs, supervisors,
and local health authorities gathered in their respective woreda centres. It also required a
small cost related to the printing of registry books and supervisions.
The thesis also used a cross-sectional study design (Paper II) of a household survey. The
challenge related to analytical results from cross-sectional studies is the limitation to
determine whether the exposure or the outcome occurred first (temporality) [158]. Yet, with
exposures that have known to be in place for predictably long time such as sex, race, and
blood group, it is easier to infer the causality of an association when it exists. For example, in
Paper II, having three or more births in five years was associated with neonatal mortality.
However, it was difficult to know whether the death of the first or second children led to
further births or whether the frequent births caused deaths due to poor care because of a
competition for scarce resources. Similarly, the same study has showed that household wealth
was linked with maternal and neonatal mortality. Nevertheless, we cannot clearly claim

whether the deaths required expenses for treatment before deaths occurred and caused poor
wealth or whether a prior poverty existed in households, which created little resources for
treatment and resulting deaths.
Still, household exposure variables, such as the educational status of the heads and distance
from driveable road were likely to have been there before deaths occurred. This is because
there were no observable adult education programmes and new road constructions in the
study area in the same reference period of the study. As such, these variables were more
likely to have existed before maternal and neonatal deaths and lack of temporality may not be
a concern. Conversely, in Paper IV, the rate of facility delivery was higher in districts with a
higher ratio of midwife-to-population. Nonetheless, it is difficult to characterize whether the
presence of midwives increased facility deliveries or districts with higher number of facility
delivery received more midwives.
Even so, the difficulty of ensuring temporality alone cannot rule out the importance of these
associations. Some mechanisms to improve the worthiness of the findings are comparing the
results for consistency with other findings and measuring the outcomes through different
methods [156]. Consequently, though describing associated risk factors was mainly exercised
in the household survey (Paper II), the results from the birth registry (Paper I) also showed
that factors such as education and road access were associated with maternal mortality,
highlighting consistency of associations.
Sampling issues
The size of the sample and selection methods is important to achieve enough power for a
given study to reach reliable findings (precision). Studies included in this thesis have
considered sample size, and in all the studies large number of participants were included. In
the birth registry (Paper I) we registered around 11,000 births, in 75 kebeles in three districts
with the exception of a few that had administrative problems (transferred to new districts), as
well as in kebeles where HEWs were sick or in maternity leave. In the birth registry our aim
was to find a perceived MMR of 531 (95% CI: 413, 669) and to determine group differences
in maternal mortality based on socio-economic and access-related variables. However, the
result after the study showed that the MMR was 489 (95% CI: 366, 628), which is within the

95% CI of what was perceived, and the MMR in villages without a driveable-road compared
to villages with such a road was twice as high. We also used sufficient sample size in the
household survey (Paper II) to find out about 11,000 births and their outcomes.
However, in both papers (Papers I and II), the unavoidable problem related to rarity (in
absolute numbers) of maternal deaths caused wider confidence intervals. Practically
speaking, the wider confidence interval around the estimates of MMR is a well-understood
challenge, and precise estimates would be much expensive. Paper III used over 8,500 siblings
interviewed in 15 randomly selected of 30 rural kebeles within Bonke. This amount of study
size was well over the recommended 5,000 siblings for the indirect sisterhood study in a
setting with a similar magnitude of maternal mortality and fertility [159]. In Paper IV, all of
the 66 health facilities that were expected to provide basic and comprehensive emergency
obstetric care (health centres and hospitals) were studied because the WHO suggests
assessing all such institutions when the number of facilities are fewer than 100 [107].
Concerning the technique of sample selection, probability sampling (simple random,

stratified, cluster, and systematic), is the gold standard. For this reason, making inferences
about the population from observations made in samples assumes this standard technique has
been applied [160]. However, health science studies also use non-random designs such as
convenience sampling. Papers in this thesis used random techniques (Papers II and III) and
included all eligible participants in purposely-selected areas (Papers I and IV).
6.1.2 Internal validity
Epidemiologic studies are measurement exercises that aim to obtain valid results (accurate
and precise) with minimized error [150]. Precision (reliability) can be achieved by using a
large sample size to minimize chance error. However, increased sample size cannot solve the
problem of accuracy. Accuracy needs a careful design and analysis of the study to help
minimize bias and confounding. Unfortunately, epidemiologic studies often appear to have
random errors (chance) that affect precision, as well as systematic errors (bias and
confounding) that influence accuracy. Systematic errors arise from three important sources:

selection bias (biased sampling), information bias (biased measurements), and confounding
(false inferences on associations).
Random error (chance)
Random error is the extent to which sampling variability (chance) explains the observed
association in the data. The role of chance error can be assessed using statistical techniques
(significance tests and confidence intervals) and this type of error can be reduced by
increasing the sample size [161]. Papers in this thesis addressed the role of chance by
performing statistical tests appropriate for the measured effect sizes. Effect sizes such an
Odds-ratio were reported together with a 95% CI, and chi-square tests were reported with the
P-values as necessary. Hence, in Paper II, maternal mortality was statistically associated with
few variables, which unlike neonatal mortality that showed a statistical association with more
variables when assessed using the P-value of a cut-off point of ≤ 0.05. The reason for this
may be that the maternal mortality had a lower chance of yielding a statistically significant
association, thereby causing chance error to be the likely explanation.
Selection bias
Selection bias is a systematic error due to a problem in the study design, where participants
are deferentially enrolled into the study. In other words, when a fewer or larger number of
participants with a potential outcome than the normal average in the community are selected,
the result is artificially biased towards lower or higher estimates [162]. Selection bias has a
clear potential for estimating maternal mortality through the facility study (Paper IV) because
often it is only a few selected severe cases visit health facilities for delivery or for treatment
of complications. This is particularly a prevalent problem in settings where the awareness of
users and the quality of health service providers are poor. As such, the MMR would be too
high among those who visited health facilities because already complicated cases visit these
facilities, and many die.
The MMR could also be too low when facility deaths are directly translated to the catchment
population because many deaths occur at home. In the birth registry study (Paper I), we
registered 81% of births estimated for 2010 in the study communities and reported the MMR
outcomes. However, the measurement concern is whether the outcome (maternal mortality) is

differentially lower or higher in the remaining unregistered 19% of the expected births. The
validation study showed that HEWs were more likely to register births to mothers who
attended antenatal services and deliveries in households physically closer to the HEW
stations (health posts), which may highlight the introduction of a selection bias. The same
validation study also showed that the MMR between registered and unregistered births was
similar: RR, 1.06 (95% CI: 0.28, 3.98). The similarity in maternal mortality and a high
coverage in the registration of estimated births may imply a minimized influence of selection
bias in affecting the overall value of the MMR.
In addition, the maternal mortality estimate obtained from birth registry (Paper I) was
consistent with a finding from another community-based measurement through household
survey (Paper II). In conclusion, even if selection bias may have been difficult to completely
avoid in the birth registry and the validation studies, it may have been less likely to be
differential in terms of the outcome (maternal mortality).
Another important source of selection bias is refusal (self-selection bias), in which

participants who are perceived to have less or more likely to have the outcome of interest, fail
to participate in a study. For example, if many households refused to participate in the
maternal mortality study and the MMR was different between those who responded and
refused, this could lead to an artificially lower or higher estimate of MMR. However, since
the studies in this thesis (Papers II, III and IV) had more than a 95% response rate, selection
bias due to refusal may not have been a problem. In the birth registry study (Paper I), given
the difficulty of registering births in rural communities who often live in scattered
households, the 81% registration coverage we achieved is high.
In Paper III (the sisterhood paper), more male siblings were interviewed than females about
the deaths of their sisters. This may have been because women often travel to rural open
markets, collect firewood, fetch water, while men stay in the garden cultivating the farms,
which makes men more available for interviews. Additionally, the difference could also be
due to a self-selection of women who shy away from interviews, as women usually stay the
secluded parts of the home, while men respond to questions from outsiders who visit their
homes. However, we found that maternal mortality indicators obtained through male and

female respondents were similar, which was also confirmed in a recent study by the original
founders of the sisterhood method [163]. Therefore, the sisterhood estimate of MMR in our
study (Paper III) may not have been greatly affected by selection bias.
Information (measurement bias)

Information bias is a systematic error in epidemiologic studies, in which differential
measurements of exposures or outcomes are made between groups. Information bias has
different sources: 1) Observer bias ─ related to a bias introduced by interviewers in surveys,
readers of measuring gauges, assessors of diagnosis, and readers of laboratory and x-ray
results. 2) Subject bias ─ is a respondent bias caused by a recall bias, a socially desirable
response, or intentionally biasing of information by respondents. 3) Instrument bias ─ is
related to the problem of materials used for measuring the exposure or outcome (the change
of a reader machine at different settings), and confusing (non-standardized) guidelines for the
classification of cases and deaths.
Recall bias is a concern in the household survey and sisterhood studies (Papers II& III)
because these studies required memorizing past events of maternal deaths in the families.
Observer bias also has a potential because people who respond to interviews about maternal
deaths may not be able to properly classify the cause of a woman’s death as maternal or not,
or may not recognize early pregnancy when a woman has died.
We attempted to minimize information bias by selecting interviewers who had lived and
knew about most of the vital events in their respective villages for a long time (Paper II).
Moreover, pregnancy and maternal death are memorable events that cannot be easily
forgotten within five years. We also believe that there was a minimal chance for information
bias related to misclassifications deaths during early pregnancy because the study was done
in a socially connected rural community. In these communities, unlike with urban dwellers,
information about important events such as pregnancy is closely shared with family members
and communities.
Even so, for the sisterhood study (Paper III), it is difficult to know how much information
bias has played a role in the unexpectedly high estimate, though due to the inbuilt nature of

the method, the reference time for the estimate was more than a decade before the study. In
the birth registry study (Paper I), data collectors were not aware how the predictor variables
(road distance and type, literacy of father and mother, sickness during pregnancy, antenatal
follow-up) would be used. As a result, the HEWs had little reason for a biased classification
of maternal mortality outcome based on these variables.
In summary, however, I admit that the maternal mortality estimation is often affected by
under-estimation due to misclassifications and recall biases [164], and that studies in this
thesis might not have completely avoided these problems.
Confounding and effect modification
Confounding is an important concern in epidemiologic studies and is defined as a confusion

or mixing of the effect of one exposure variable with the effect of another exposure variable
on the outcome [150]. For this to happen, a confounding variable should be associated with
both the exposure (predictor) and the outcome variables. Nonetheless, a confounder is not an
intermediate variable in the pathway between exposure and outcome. The household survey
(Paper II) showed that exposure variables such as the education of the head of household, the
household distance from a driveable road, three or more births in five years, and the wealth
status of households were associated with neonatal mortality outcomes.
However, the heads of households closer to driveable roads or in wealthy households may
have been more educated than those living in poor or distant households. Consequently,
independent of the distance from a driveable road, or the wealth of the household, educated
heads of household are more likely to seek medical care for their sick neonates and reduce
mortality. In other words, education may have been the true cause of the difference in
mortality, and others (non-spaced birth, wealth, road distance) may have been potential
confounders.
Several design and analysis techniques have been developed for epidemiological studies to
control for confounding. During the design, randomization, specification of inclusion criteria,
and matching may be applied. At the time of analysis, we may use stratifications and

adjustments (regression models) to control for the measured confounder variables in a study
[150, 151, 165]. In the household study (Paper II), we used the logistic regression to adjust
for the potential confounders mentioned above, and we also did a stratified analysis to assess
whether education had an effect on the modification of the association between wealth and
neonatal mortality. The analysis showed that education modified the effect of wealth on
neonatal mortality, in which wealth was only associated with neonatal mortality among
illiterates, and the neonatal mortality was similar between the wealthy and poor households
headed by educated adults (Table 5, under additional results). Conversely, as all studies
included in the thesis were observational studies, except for using some specific inclusion
criteria, we did not use randomization in the design as a solution for confounding problem.
6.1.3 External validity (generalization)
External validity refers to whether or not a research finding can be generalized to a

population in different settings [166]. The birth registry study (Paper I) was conducted using
the existing community health workers, in which household visit is their routine activity, with
priority given to households with pregnancies and births. The same study involved people
from different cultural backgrounds and three ethnic groups in south Ethiopia (Gamo, Zeisse,
and Derashe). Moreover, health extension workers who performed the birth registration and
identified maternal deaths are distributed in the same proportion to the population and have
the same training background, all are females and all have similar working conditions in the
country. This may highlight that HEWs can conduct a birth registration of high coverage and
optimum quality, and can classify pregnancy and birth outcomes in any rural community in
Ethiopia.
Furthermore, the area we conducted the studies in were similar in demographics, health
services, road access, and economic structure with most rural communities in Ethiopia. These
facts may indicate that the findings of maternal and neonatal mortalities, along with the
associated access and socio-economic factors may be generalized to many parts of rural
Ethiopia. Nevertheless, because of specific local differences in health service quality,
utilization, and cultural contexts, the rate of skilled birth attendance and the MMR may vary
in a few places. However, the findings in three studies in this thesis which showed over 80%
of maternal deaths occurring at home may be true for many rural communities in Ethiopia. In

fact, this generalization may also apply to developing countries at similar stages of health
service access and socio-economic development, as a study from Mozambique also found
similar results (86% of maternal deaths found through community-based study were not
registered in the existing passive civil registration) [167].
6.2 Discussion of main findings
6.2.1 Overview of the work and findings
To the best of our knowledge, we assessed maternal mortality for the first time by using
different methods that combined community and facility data in south Ethiopia. More
importantly, we describe the possibility of birth registration and a registry-based
measurement of maternal mortality in rural communities using the existing primary health
care workers (HEWs). Although we admit that measuring maternal mortality in rural
communities in developing countries is difficult, our findings suggest that it may be possible
in places where functional system of community health service exists. As a result, we
challenge the idea about whether to stop measuring maternal mortality in resource-limited
countries and depend on process indicators [45]. Reducing maternal mortality depends on
sustainable information from a measured magnitude of mortality, understanding socio-
economic factors associated with the problem, as well as information on the coverage,
quality, and utilization of obstetric services.
Alternative information for tackling maternal mortality can be obtained from process
indicators such as financial commitments and skilled birth attendance (SBA) to help facilitate
planning, implementation, and monitoring services. By contrary, evidence show that the level
of process indicators may not necessarily match maternal mortality rate. For example,
Ethiopia had one of the lowest skilled birth attendance rates in the world (< 10 until 2010) but
the MMR was lower than some countries that had over 50% facility deliveries.
Thus, it is important to describe the measured changes in terms of the ultimate goal of
reducing maternal and neonatal mortality. The MDG5 aims to see the measured reduction in
maternal mortality as the main goal (goal 5A). The same goal also emphasizes monitoring of

the level of skilled birth attendance as a key indicator of progress. In developing countries, as
is also true in Ethiopia, there is limited routine data from vital registration, which is
considered the gold standard in measuring maternal mortality. The major reason for a lack of
routine vital data is shortage of information systems accessible to the largest proportion of
people living in remote rural areas in resource-poor settings (e.g. 85% in Ethiopia) [168].
Fortunately, Ethiopia has partly solved this problem through the health extension programme
that placed two trained and salaried women in every village of the country. This innovative
approach has met the goal of achieving MDG4 by reducing the mortality of children under-
five years of age [132]. However, this programme has not been tested for its effect on the
much needed registry-based data.
In this thesis, these health extension workers were able to register more than 80% of expected
births (Paper I) in a country where only 7% of births were registered in 2013 [169], and with
almost non-extent registry in rural areas. By so doing, we describe that the registration of
births and birth outcomes in rural communities in developing countries where functional
community health workers exist is feasible. This feasibility of birth registration served as an
important tool to measure maternal mortality. The community birth registration (Paper I) and
other community and facility studies in the thesis (Paper II and IV) also demonstrated that for
every one maternal death that occurred at a health facility, four others died at home without
any notice by the health system. Furthermore, we believe that the primary health-care
workers (HEWs) were able to note the true proportion and types of skilled care received by
mothers during pregnancy and at birth, and the distribution of major causes and time of
maternal deaths through the prospective data.
The thesis also validated the birth registry data, and supplemented the findings with other
community and health facility studies. As such, measuring maternal and neonatal mortality
using a large household survey (Paper II) showed that the MMR is close to the estimate
obtained through the birth registration. One major weakness of the paper-based birth registry
is its limited capacity to link the outcomes with socio-economic factors because of recording
of mainly pregnancy- and birth-related variables without any details about socio-economic
conditions of households. Consequently, the household survey has complemented the
weakness of birth registry by informing about socio-economic factors associated with
maternal and neonatal mortality. Moreover, these two community-based approaches (Papers I

and II) and the assessment of health facilities (Paper IV) have shown consistent results that
most maternal deaths occurred at home. Additionally, because of its complexity, measuring
the level of maternal mortality itself requires a mixed methods approach (vital registrations,
confidential inquiries, and other data sources), even in developed countries.
Another simple and cheap method of estimating maternal mortality in high fertility and
mortality settings is the sisterhood method (Paper III), which helped to provide an overview
of the problem in the area a decade before 2010. Approaches through community-based
methods (Paper I and II) and the health facility study (Paper IV) have also shown that many
existing health institutions do not provide the level of delivery care expected. Moreover, the
quality of emergency obstetric care institutions was far below the UN recommended
minimum to attain MDG5 goal (Paper IV).
6.2.2 Maternal and neonatal mortality
The knowledge of the magnitude and differentials of maternal and neonatal mortality is
critical for the policy and programme response. Reporting the level of maternal mortality at
this period is especially important because the result would indicate how the area is
progressing towards the MDG5 target of reducing the MMR by 75% in 2015 from the 1990
level [7]. The MMR in Ethiopia was estimated to be 968 (95% CI: 600-1507) in 1990. This
level needs to be reduced to 242 per 100,000 live births in 2015 for Ethiopia to achieve the
MDG5 goal. Community-based studies in this thesis have shown measured value of an
MMR between 425 in the household survey (Paper II) and 489 in the birth registry (Paper I),
which are nearly 500 maternal deaths per 100,000 live births. The sisterhood study (Paper III)
has also showed a higher rate of maternal mortality (an MMR of 1667 per 100,000 LBs) that
refers to a time of a decade ago (Paper III).
These values indicate that there was a high MMR in 2010, when there was five years
remaining for the final target in 2015. However, the findings also highlight that maternal
mortality in rural Ethiopia might have been decreasing over time compared to what we
estimated through the sisterhood method and other global estimates for the MMR in Ethiopia
in the 1990s. The sisterhood method (Paper III) resulted in unexpectedly high estimate with a

lifetime risk of maternal mortality of 1 in 10, which translated to a very high MMR of 1,667,
close to the natural MMR expected to occur without any modern intervention. However,
because of the in-built characteristics of the method, the estimate points to the year 1998
which is when the MMR for Ethiopia was similarly high from other reports [170]. In fact, the
district of Bonke where the sisterhood study was conducted had few safe motherhood
interventions that could have prevented maternal deaths before 1998. Consequently, maternal
mortality might have been very high in the area before 1998, and is currently showing a
decline. Nevertheless, since there was no previous empirical data from the area, it is difficult
to describe the trends in decline.
The maternal mortality rate was not previously reported from community-based birth
registration in Ethiopia, although there have been few community-based household and
sisterhood surveys conducted in other rural districts in Ethiopia. The MMR per 100,000 LBs
was 440 in Butajira in 1996, 570 in Illubabor in 1995, 402 in Jimma in 1990, and 566 in
Addis Ababa in 1983 [17, 171-173]. Unfortunately, many of these studies were conducted a
decade before our studies. However, the community-based studies mentioned above reported
an MMR similar to what we measured in this thesis. However, the time difference between
our studies and previous studies is long. For this reason, the similarity of these estimates with
our findings may highlight that the previous studies in other parts of Ethiopia might have
been under-estimates. Alternative explanation could be that the MMR in the area we did our
studies in had been higher than in those other districts.
An alternative source of community-based data is the DHS, which reported an MMR of 671
in 2005 [168] and 676 in 2010 [174], thereby indicating no change. However, the DHS uses
very few households as being representative to the entire country and this, compounded with
the rarity in absolute numbers of maternal deaths, might have resulted in unrealistically
higher estimates with wide confidence intervals. Modelled national estimates are also
available from the UN and the Institute of Health Metrics and Evaluation (IHME). However,
the UN-estimate for the MMR in Ethiopia is much lower (350 per 100,000 LBs in 2010) [5]
compared to our findings of MMR of 489 and 425 (Paper I & II). The IHME estimate of
MMR in Ethiopia was 590 for 2008 [7], which is relatively higher than our findings, possibly
because it was two years earlier than our studies.

We also measured an NMR of 27 per 1,000 LBs in Bonke (Paper II), which was lower than
the DHS national estimate of 37 per 1,000 LBs in 2010 [174] and the UN Inter-agency group
estimation of 31 in 2011 [175] in Ethiopia. Yet, there was no previous local estimate from the
area, which makes it difficult to conclude whether the rate we found was under-reported or
represents reality. In fact, the child mortality rate has substantially declined in Ethiopia from
204 in 1990 to 68 in 2012 [135], and there may have been reductions in neonatal mortality in
the same period. If we consider international estimates that attribute 30-40% of under-five
mortality to neonatal deaths, the NMR we found (27) was 40% of the national under-five
mortality rate (68) during the same period, highlighting a consistency in our results with other
reports. Even so, concrete evidence for the rate of neonatal mortality requires further studies.
In summary, our community-based MMR estimates were consistent with each other, as well
as agreeing with information from health facilities that showed high proportion of maternal
deaths outside of health facilities. Moreover, we believe that the MMR estimate (489)
obtained through the prospective birth registry represents the best estimate, as we have also
validated this finding. Furthermore, the MMR measured through the birth registry was in
between the controversially high DHS (676) and low UN (350) estimates for the same period
(2010) in Ethiopia. Therefore, the community-based measurement methods, particularly the
findings from the birth registry (Paper I), may have represented the reality on the ground.
6.2.3 Inequalities in mortality outcomes
Even within a narrow area such as districts and provinces, maternal and neonatal mortalities
can vary depending on access to health service and socio-economic conditions. While
obstetric causes such as bleeding or infection are responsible for the mortality of an
individual woman, at the population level, these causes can have a differential effect for a
group of women with certain common backgrounds (illiterate, poor, far from road, and with
poor access to health services) compared to others without these exposures. In our studies
(Papers I and II), we report inequalities in maternal and neonatal mortality based on
household wealth, educational status, distance to and quality of driveable roads, and frequent
births.

The MMR was higher in the poorest households, among households in distant villages, and
where the head of households were illiterate (Paper II). The birth registry study (Paper I) has
also shown that the educational status of the heads of household and the type of road to the
villages was associated with the MMR. The importance of the educational level of the father
of babies as a factor for maternal mortality has also been reported before, and may be related
to the importance of the male partner’s knowledge for seeking health services, and affording
the related costs. Similarly, the distance from road and the poor condition of roads may
decide the availability and speed of transport when families want to receive medical care.
The implication of these findings, from the health service point of view, is that health
professionals should give a priority to pregnant women with poor socio-economic
backgrounds and in difficult to access villages, if a programme is to reduce maternal
mortality without waiting for changes in education and wealth in these households. A similar
explanation may also be true for the neonatal mortality, which was associated with household
wealth, education of the head of household, road distance from the household, and the
frequency of births within five years in households (Paper II).
6.2.4 Skilled birth attendance and emergency obstetric care
In Ethiopia, an aggressive expansion of health centres took place in the past few years with
the aim of making a health centre available within a 10 km distance of households. As such,
we found that the median distance of households was 10 km from health centres, and 57% of
households were within 10 km of a health centre in 2010 (Paper I). Nonetheless, the quality
standards and the public utilization of existing health facilities was very low.
The community and facility studies included in the thesis have shown that the overall rate of
delivery in skilled health facilities (health centres and hospitals) was less than 10% until
2010. The rate was 3.7% in between 2006 and 2010 in rural villages (Paper II), 6% in 2010 in
rural villages (Paper I), and 6.6% in 2010 through the facility study that includes cases from
urban areas (Paper IV). Our result is consistent with the findings of a nation-wide assessment
of the status of emergency obstetric care in the country in 2008 [176]. The aforementioned

national assessment reported that 7% of deliveries in Ethiopia took place in institutions of any
type and 3% in facilities that provided all the signal functions of emergency obstetric care.
In fact, the rate of skilled birth attendance in Ethiopia has stagnated at a very low level for
decades. Furthermore, the institutional delivery rate was widely varied between urban and
rural areas within Ethiopia (84% in Addis Ababa and 7.2% in the Afar region in 2010) [174].
According to national DHS studies, skilled birth attendance was 5.7% in 2005 [168] and 10%
in 2010 [174]. Encouragingly, an improvement to 15% was reported in 2014 [121], but in the
same DHS report in 2014, the rate was still 10% in rural areas and 12% in the southern region
where we conducted the studies. Overall, this rate is far lower than the 45.5% average for
Africa and 35.3% for least developed countries in 2008 according to a WHO update [177].
Review of the evidence suggest that emergency obstetric care should be a critical component
of an effort to reduce maternal mortality [178]. Consequently, the WHO has developed
assessment tools and minimum standards for monitoring the status of EmOC services [179].
This WHO tool was tested and proven effective in reflecting the level of EmOC in the field
[180]. In the facility study (Paper IV) we used the UN standard and the assessment
instrument to describe the physical coverage, quality, and utilization of EmOC services
because measuring maternal mortality without assessing the status of obstetric services
cannot adequately inform policy makers and programme planners. We found that the number
of health centres and their physical proximity to the population was promising. However,
many of these institutions did not meet the minimum standards expected to provide
emergency obstetric services [179]. As a result, fewer mothers use the facilities for delivery
and the institutional MMR was high, thus highlighting a formidable challenge to rapidly
reducing maternal and neonatal mortality.
6.2.5 Why a sign of reduction in maternal mortality in Ethiopia with a low skilled
delivery rate?
Ethiopia has achieved the MDG 4 target for reducing the mortality of children under the age
of five by two-thirds through general preventive and health promotion measures by
implementing community-based primary health care [132]. However, reducing the neonatal

component of child mortality and the maternal mortality needs better clinical care, and in the
future further reductions will depend on improving clinical excellence. Because of statistical
challenges related to wide confidence intervals surrounding MMR due to a relative rarity in
the numbers of maternal deaths, it is often difficult to comment on changes in the MMR. In
addition, a lack of real-time and continuous data makes it hard to regularly update
information.
Model-based estimates [5, 7] show that there has been a decline in maternal mortality in
Ethiopia in the face of exceptionally low levels of skilled birth attendance and access to
emergency obstetric care. Recently, the Federal Ministry of Health of Ethiopia and the United
Nations Population Fund (UNFPA-Ethiopia) have released a joint statement which suggests
Ethiopia is making good progress in reducing maternal mortality to achieve its MDG 5 goal
[181]. However, the statement was not backed by concrete data for the claim.
By contrast, Yifru Berhan and Asres Berhan, in their recent review argue that there has been
no significant change in maternal mortality over the last 30 years, citing the overlaps in 95%
CIs and inconsistencies in several estimates issued since 1980 [182]. However, because of the
nature of a relative rarity in absolute numbers of maternal mortality, which often fails to
satisfy the power of statistical analysis, entirely depending on the 95% CI may be difficult to
witness a progress in MMR reduction. To obtain a reliable 95% CI with adequate statistical
power may need a large number of maternal deaths, which is much costly. Hence, it may be
wise to assess the central level of estimate of MMR at a given time for a given area.
Our findings (Paper I and II) also consistently measured an MMR below 500 per 100,000
LBs compared to the MMR of over 1,000 in Ethiopia in 1990s, although the paradox of a
maternal mortality reduction without marked improvements in clinical obstetric care remains
unexplained. Ethiopia’s rapid expansion of the physical coverage of health centres since 2005
has had little impact on increasing the utilization of facility intrapartum care, which is the key
strategy for reducing maternal mortality. In addition, the HEWs were expected to be involved
in delivery services, but they had little clinical skills. Consequently, the government is now
upgrading them with one year of additional training on the maternal and child health

curriculum. Hence, the effect of the upgrading of HEWs may help reduce the MMR in the
future by improving the skilled birth rate, but cannot explain the current declines in MMR.
However, an unexplained reduction of maternal mortality in places with a low utilization of

professional care at birth is not a new phenomenon only seen in Ethiopia. It has been seen in
other countries and is the subject of debate [183]. For example, skilled birth attendance was
10% in Ethiopia and 50% in Tanzania in 2010, though both countries had a similar MMR
(420 in Ethiopia and 410 in Tanzania) [184]. On the other hand, Somalia had three times
higher skilled birth attendance (33%) compared to Ethiopia; even so, this is a contradiction
given the MMR of 850 in Somalia [184]. As such, some experts do not agree on a high
emphasis on clinical services to reduce maternal mortality, arguing in favour of some
“context-specific interventions” such as women’s support groups, vitamin-A distributions,
and related community interventions. The reason cited for this is that achieving a high
coverage of skilled birth attendance in remote communities is unrealistic and cannot be
achieved in the foreseeable future [185].
The Ministry of Health of Ethiopian attributes recent health gains to the decentralized
community-based HEP. However, what specific and measureable interventions HEWs
conduct to reduce maternal mortality is a question that remains to be answered. Our studies
have demonstrated that HEWs conducted a low proportion of births that took place in their
villages, 13.4% in 2010 (Paper I) and 4% on average between 2006 and 2010 (Paper II).
Rather, the following four preventive interventions might have contributed to the beginning
of a decline in maternal mortality in Ethiopia.
First, fertility has markedly declined in rural Ethiopia in the past decade, from an annual
crude birth rate (CBR) of 43 births per 1,000 population in 2000 [186] to 28 in 2014 [121] in
which HEWs may have played important role by providing birth control medications.
Similarly, the total fertility rate (TFR, the average number of births a woman would give
throughout her reproductive life span) has decreased to 4.1 in 2014 from 5.9 in 2000. This
fertility decline was simultaneous, with an increased uptake of contraceptive technologies
[121]. Second, general improvements in living conditions, nutrition, and hygiene that have

implications for deaths related to anaemia and infections may have played some role, as
Ethiopia has shown economic growth in recent years.
Third, the decline of infectious diseases such as malaria [187] and HIV [188] might have also
contributed to the beginning of a decline in maternal mortality in Ethiopia. Fourth, Ethiopia
passed a law in May 2005 that liberalized abortion, and the Ethiopian Ministry of Health
issued a technical and procedural guideline in 2006 [189]. The new law gives women more
authority to request a safe abortion if the pregnancy was due to rape or incest (from a closely-
related person or family), or when the mother or the foetus has severe medical problem.
Moreover, girls under the age of 18 can obtain an abortion without any conditional
requirements if they make a claim.
A five-year follow-up study in a hospital setting in Addis Ababa on abortion-related

admissions before and after the 2005 legislation has highlighted a decreased trend of abortion
admissions and abortion related maternal mortality [190]. A national study also estimated the
incidence of abortion in Ethiopia at 23 per 1,000 women of reproductive age in 2008, which
is a lower rate compared to the WHO estimate of 39 per 1,000 reproductive women in East
Africa [191].
In our facility-based study (Paper IV), abortion-related admissions in 2010 were only 291
cases among 1.8 million people served by the surveyed facilities, which may also indicate a
lower rate of admission, though this could be somewhat biased because of a low rate of
abortion service utilization. Similarly, in the northern Ethiopia in Tigray, a pilot study
reported decreased admission of abortion complications after the introduction of the new
abortion law [192]. However, more concrete evidence is needed to suggest whether a
reduction in abortion is contributing to maternal mortality decline in Ethiopia.

7 Conclusion and recommendations

7.1 Conclusion
Community-based measurement methods (birth registry and household survey) provided

consistent maternal mortality estimates, in addition to describing inequalities in the
mortality outcomes across selected variables. Moreover, the birth registry provided a
more valid MMR compared to the alternative measurement methods. The finding of the
feasibility of prospective birth registration and its usefulness as a tool to measure
maternal mortality can be considered an important learning step towards implementation
of the recently approved law of universal vital registration in Ethiopia. Through using
different methods of assessment, the studies in the thesis characterize the magnitude,
main causes, time, and associated factors related to maternal and neonatal mortality. The
thesis also describes the coverage, quality, and utilization status of obstetric service. The
findings should therefore serve as a baseline for the future monitoring of progress in
maternal and neonatal mortality and obstetric services in the study area, which can be
replicated in other rural settings in Ethiopia, as well as elsewhere in developing countries
with a functional community health system.
7.2 Recommendations
For further research
1. Long-term prospective studies of birth registration may be important to oversee

whether such follow-ups have any effect on reducing maternal mortality.
2. Replication of the birth registry study in other areas with a different cultural and
health service status may help to strengthen the evidence for its wider application.
3. Such an effort may take further a step towards establishing a system of starting the
registry during pregnancy to improve the quality of the data.
4. Future birth outcome studies through the use of a community birth registry should
consider the concerns of the diagnostic qualities of cause-of-death ascertainment. A
mechanism of linking HEWs to medical specialists could be an option.
5. Future integration of birth registration with the WHO’s “maternal death surveillance
and response’’ (MDSR) strategies may help improve the overall results.

6. Complementing the quantitative finding with qualitative analysis on the effects of

maternal mortality on the surviving babies and families may help to provide full
images of the consequences of maternal mortality.
7. Intermittent use of household and sisterhood methods may be useful to help compare
the findings and to learn about the strengths and weaknesses of measurement
methods.
8. A repeated assessment of the quality and utilization of obstetric services using the
standard document produced by the UN may help to oversee changes in obstetric
services.
For programme implementation
1. Programme planners and implementers may need to base their work on the
information produced, and support the generation of new data such as birth
registration in their respective localities.
2. The fact that maternal and neonatal mortality is differentially high in extremely poor
households, households with lower education for the head, and with frequent births
highlights the importance of a focused follow-up of pregnancies in such households.
3. Strengthening the quality of the emergency obstetric service facilities is crucial to
both increase the very low utilization rate, and to reduce high maternal mortality
rates in health facilities. Special attention needs to be given to the critical problem of
lack of blood for transfusions and a shortage of parenteral drugs.
4. The initiative of upgrading the HEWs to midwifery should be backed with supplies
because the current shortage of supplies could hamper their practice. Such initiatives
should also be backed by clear strategies and guidelines to link HEWs with referral
facilities.
Policy recommendations
1. Policy makers may consider the presence of HEWs as opportunity and prioritise the
universal registration of births and birth outcomes. The information can provide
important data that can be useful to the health sector and others such as education,
legal services, and equity analysis.
2. The policy attention given to expanding health centres for physical access should be
backed by a similar commitment to help improve the quality of these facilities.

8 References
1. Horton R: Maternal mortality: surprise, hope, and urgent action. The Lancet 2010,
375(9726):1581-1582.
2. Kassebaum NJ, Bertozzi-Villa A, Coggeshall MS, Shackelford KA, Steiner C, Heuton
KR, Gonzalez-Medina D, Barber R, Huynh C, Dicker D et al: Global, regional, and
national levels and causes of maternal mortality during 1990 to 2013: a systematic
analysis for the Global Burden of Disease Study 2013. The Lancet 2014.
3. Ronsmans C, Graham WJ: Maternal mortality: who, when, where, and why. The
Lancet, 368(9542):1189-1200.
4. Mahler H: The Save Motherhood Initiative: A Call to Action. The Lancet 1987,
329(8534):668.
5. UNFPA: UN Inter-agency Group for Maternal Mortality Estimation. Trends
inmaternal mortality 1990 to 2010: WHO, UNICEF, UNFPA, and the World Bank
Estimates. Geneva: World Health Organization, 2012. available at:
https://www.unfpa.org/webdav/site/global/shared/documents/publications/2012/Trend
s_in_maternal_mortality_A4-1.pdf (accessed 5 May, 2014). 2012.
6. Donnay F: Maternal survival in developing countries: what has been done, what can
be achieved in the next decade. International Journal of Gynecology & Obstetrics
2000, 70(1):89-97.
7. Hogan MC, Foreman KJ, Naghavi M, Ahn SY, Wang M, Makela SM, Lopez AD,
Lozano R, Murray CJL: Maternal mortality for 181 countries, 1980–2008: a
systematic analysis of progress towards Millennium Development Goal 5. The Lancet
2008, 375(9726):1609-1623.
8. Kinney MV, Kerber KJ, Black RE, Cohen B, Nkrumah F, Coovadia H, Nampala PM,
Lawn JE, on behalf of the Science in Action: Saving the lives of Africa's mothers n,
children working g: Sub-Saharan Africa's Mothers, Newborns, and Children: Where
and Why Do They Die? PLoS Med 2010, 7(6):e1000294.
9. Bicego G, Boerma JT, Ronsmans C: The effect of AIDS on maternal mortality in
Malawi and Zimbabwe. AIDS 2002, 16(7):1078-1081.
10. Fawcus SR, van Coeverden de Groot HA, Isaacs S: A 50-year audit of maternal
mortality in the Peninsula Maternal and Neonatal Service, Cape Town (1953–2002).
BJOG: An International Journal of Obstetrics & Gynaecology 2005, 112(9):1257-
1263.

11. Paul BK: Maternal mortality in Africa: 1980–87. Social Science & Medicine 1993,
37(6):745-752.
12. Rush D: Nutrition and maternal mortality in the developing world. The American
Journal of Clinical Nutrition 2000, 72(1):212s-240s.
13. Karlsen S, Say L, Souza J-P, Hogue C, Calles D, Gulmezoglu AM, Raine R: The
relationship between maternal education and mortality among women giving birth in
health care institutions: Analysis of the cross sectional WHO Global Survey on
Maternal and Perinatal Health. BMC Public Health 2011, 11(1):606.
14. Freedman LP, Waldman RJ, de Pinho H, Wirth ME, Chowdhury AMR, Rosenfield A:
Transforming health systems to improve the lives of women and children. The Lancet,
365(9463):997-1000.
15. Royston E, AbouZahr C: Measuring maternal mortality. BJOG: An International
Journal of Obstetrics & Gynaecology 1992, 99(7):540-543.
16. Hill K: Measuring maternal mortality. The Lancet 2006, 368(9553):2121.
17. Berhane Y, Andersson T, Wall S, Byass P, Högberg U: Aims, options and outcomes
in measuring maternal mortality in developing societies. Acta Obstetricia et
Gynecologica Scandinavica 2000, 79(11):968-972.
18. Hill K, El Arifeen S, Koenig M, Al-Sabir A, Jamil K, Raggers H: How should we
measure maternal mortality in the developing world? A comparison of household
deaths and sibling history approaches. Bulletin of the World Health Organization
2006, 84:173-180.
19. Say L, Souza JP, Pattinson RC: Maternal near miss – towards a standard tool for
monitoring quality of maternal health care. Best Practice & Research Clinical
Obstetrics & Gynaecology 2009, 23(3):287-296.
20. WHO: Beyond the numbers: reviewing maternal deaths and complications to make
pregnancy safer. Available at:
http://www.who.int/maternal_child_adolescent/documents/9241591838/en/ (accessed
20 June, 2013). . World Health Organization 2004.
21. WHO: Maternal mortality: Fact sheet No 348, WHO-media centre. available at:
http://www.who.int/mediacentre/factsheets/fs348/en/ (accessed 15 May 2014). 2014.
22. WHO-PMNCH: The Partiner for Maternal, Newborn and Child Health (PMNCH) fact
sheet: Maternal mortality. available at:
http://www.who.int/pmnch/media/press_materials/fs/en/ (accesssed 14 May 2014).
2011.

23. Khan KS, Wojdyla D, Say L, Gülmezoglu AM, Van Look PFA: WHO analysis of
causes of maternal death: a systematic review. The Lancet 2006, 367(9516):1066-
1074.
24. Say L, Chou D: Better understanding of maternal deaths—the new WHO cause
classification system. BJOG: An International Journal of Obstetrics & Gynaecology
2011, 118:15-17.
25. ANYA SE: Seasonal variation in the risk and causes of maternal death in The
Gambia: Malaria appears to be an important factor. The American Journal of Tropical
Medicine and Hygiene 2004, 70(5):510-513.
26. Romagosa C, Ordi J, Saute F, Quintó L, Machungo F, Ismail MR, Carrilho C, Osman
N, Alonso PL, Menendez C: Seasonal variations in maternal mortality in Maputo,
Mozambique: the role of malaria. Tropical Medicine & International Health 2007,
12(1):62-67.
27. Brabin BJ, Hakimi M, Pelletier D: An Analysis of Anemia and Pregnancy-Related
Maternal Mortality. The Journal of Nutrition 2001, 131(2):604S-615S.
28. Fauveau V, Blanchet T: Deaths from injuries and induced abortion among rural
Bangladeshi women. Social Science & Medicine 1989, 29(9):1121-1127.
29. Ganatra B, Coyaji K, Rao V: Too far, too little, too late: a community-based case-
control study of maternal mortality in rural west Maharashtra, India. Bulletin of the
World Health Organization 1998, 76(6):591.
30. Ronsmans C, Khlat M: Adolescence and risk of violent death during pregnancy in
Matlab, Bangladesh. The Lancet 1999, 354(9188):1448.
31. Say L, Chou D, Gemmill A, Tunçalp Ö, Moller A-B, Daniels J, Gülmezoglu AM,
Temmerman M, Alkema L: Global causes of maternal death: a WHO systematic
analysis. The Lancet Global Health 2014.
32. AbouZahr C, Wardlaw T: Maternal mortality at the end of a decade: signs of
progress? Bulletin of the World Health Organization 2001, 79:561-573.
33. Høj L, da Silva D, Hedegaard K, Sandström A, Aaby P: Maternal mortality: only 42
days? BJOG: An International Journal of Obstetrics & Gynaecology 2003,
110(11):995-1000.
34. Pradhan EK, West Jr KP, Katz J, Christian P, Khatry SK, LeClerq SC, Dali SM,
Shrestha SR: Risk of death following pregnancy in rural Nepal. Bulletin of the World
Health Organization 2002, 80:887-891.

35. Li XF, Fortney JA, Kotelchuck M, Glover LH: The postpartum period: the key to
maternal mortality. International Journal of Gynecology & Obstetrics 1996, 54(1):1-
10.
36. WHO: International Statistical Classification of Diseases and Health Related
Problems. tenth revision (ICD-10).Geneva. online version available
at:http://apps.who.int/classifications/icd10/browse/2010/en. (accessed 15 May 2014).
World Health Organization 1992.
37. WHO: Maternal mortality: helping women off the road to death WHO Chronicle
1986, 40(5):175-183. available at:
http://whqlibdoc.who.int/chronicle/1986/WHO_Chronicle_1986_40%285%29_175-
183_eng.pdf (accessed 1910 May 2014)
38. Graham WJ, Foster LB, Davidson L, Hauke E, Campbell OMR: Measuring progress
in reducing maternal mortality. Best Practice & Research Clinical Obstetrics &
Gynaecology 2008, 22(3):425-445.
39. Graham W, Ahmed S, Stanton C, Abou-Zahr C, Campbell O: Measuring maternal
mortality: An overview of opportunities and options for developing countries. BMC
Medicine 2008, 6(1):12.
40. WHO: Reduction of maternal mortality: A joint WHO/UNFPA/UNICEF/World Bank
statement. Available at:
http://www.who.int/reproductivehealth/publications/monitoring/9789241561952/en/
(accessed 20 June 2014). WHO Sexual and Reproductive Health publications 1999.
41. Högberg U: The decline in maternal mortality in Sweden: the role of community
midwifery. American Journal of Public Health 2004, 94(8):1312.
42. WHO: The world health report 2005: Make every mother and child count, World
Health Organization (2005). available at:
http://www.who.int/whr/2005/whr2005_en.pdf?ua=1 (accessed 9 May 1024).
43. Stanton, Cynthia, Noureddine Abderrahim, and Kenneth Hill (1997)
DRSMaternal Mortality Indicators: AnAssessmentofData Quality
and Implications for Data Use. DHS Analytical Reports No.4. Calverton,
Maryland: Macro International Inc. available at:
http://pdf.usaid.gov/pdf_docs/PNACB889.pdf (accessed June 20,2014).
44. Hill K: Making deaths count. Bulletin of the World Health Organization 2006,
84:162-162.

45. Pitt C, Greco G, Powell-Jackson T, Mills A: Countdown to 2015: assessment of

official development assistance to maternal, newborn, and child health, 2003–08. The
Lancet, 376(9751):1485-1496.
46. Countdown Working Group on Health Policy and Health Systems. Assessment of the
health system and policy environment as a critical complement to tracking
intervention coverage for maternal, newborn, and child health. The Lancet,
371(9620):1284-1293.
47. Graham WJ: Now or never: the case for measuring maternal mortality. The Lancet
2002, 359(9307):701-704.
48. Hoyert D: The autopsy, medicine, and mortality statistics National Center for Health
Statistics Vital Health Stat, 3 (33) (2001)
http://www.cdc.gov/nchs/data/series/sr_03/sr03_032.pdf. Centers for Disease Control
2001.
49. Deneux-Tharaux C, Berg C, Bouvier-Colle M-H, Gissler M, Harper M, Nannini A,
Alexander S, Wildman K, Breart G, Buekens P: Underreporting of Pregnancy-Related
Mortality in the United States and Europe. Obstetrics & Gynecology 2005,
106(4):684-692 610.1097/1001.AOG.0000174580.0000124281.e0000174586.
50. Arthure H: Confidential inquiries into maternal deaths. British medical journal 1975,
1(5953):322.
51. Saucedo M, Deneux-Tharaux C, Bouvier-Colle M: Confidential enquiries into
maternal deaths. BJOG: An International Journal of Obstetrics & Gynaecology 2010,
117(7):894-895.
52. Gibb D: Confidential enquiry into maternal death. International Journal of Obstetrics
& Gynaecology 1990, 97(2):97-99.
53. Mathers CD, Ma Fat D, Inoue M, Rao C, Lopez AD: Counting the dead and what they
died from: an assessment of the global status of cause of death data. Bulletin of the
World Health Organization 2005, 83:171-177c.
54. AbouZahr C: Safe Motherhood: a brief history of the global movement 1947–2002.
British Medical Bulletin 2003, 67(1):13-25.
55. WHO: Health and the Millennium Development Goals: available
at:http://www.who.int/hdp/publications/mdg_en.pdf?ua=1 (accessed 15 May 2014).
World Health Organization (2005).

56. Cross S, Bell JS, Graham WJ: What you count is what you target: the implications of
maternal death classification for tracking progress towards reducing maternal
mortality in developing countries. Bulletin of the World Health Organization 2010,
88:147-153.
57. Graham WJ, Adjei S: A Call for Responsible Estimation of Global Health. PLoS Med
2010, 7(11):e1001003.
58. Hill K, Stanton C, Gupta N: Measuring maternal mortality from a census: guidelines
for potential users (2001). http://www.cpc.unc.edu/measure/publications/pdf/ms-01-
04.pdf#search=%22stanton%20hill%20census%20maternal%20mortality%22
(accessed May 15, 2014)
59. Stanton C, Hobcraft J, Hill K, Kodjogbé N, Mapeta WT, Munene F, Naghavi M,

Rabeza V, Sisouphanthong B, Campbell O: Every death counts: measurement of
maternal mortality via a census. Bulletin of the World Health Organization 2001,
79:657-664.
60. Byass P: The Imperfect World of Global Health Estimates. PLoS Med 2010,
7(11):e1001006.
61. de Brouwere V, Van Lerberghe W: Safe motherhood strategies: a review of the
evidence Studies in Health Services Organisation and Policy, 17, 2001. iTG Press
Antwerp, Nationalestraat 155, B-2000 Antwerp, Belgium.
62. Hogberg U, Innala E, Sandstrom A: Maternal mortality in Sweden, 1980-88 Obstet
Gynecol 1994, 84: 2404.
63. Dawson I: The confidential enquiry into maternal deaths: its role and importance for
pathologists. Journal of clinical pathology 1988, 41(8):820-825.
64. Van Lerberghe W, De Brouwere V: Of blind alleys and things that have worked:
history’s lessons on reducing maternal mortality. Studies in Health Services
Organisation and Policy 2001, 17:7-35.
65. Horder J: General practice in 2000. Alma Ata declaration. British medical journal
(Clinical research ed) 1983, 286(6360):191.
66. Werner D, Saunders D, Schweiger M: Questioning the solution: the politics of
primary health care and child survival. BMJ-British Medical Journal-International
Edition 1997, 315(7107):555.
67. WHO: Prevention of Maternal Mortality: Report of the WHO Interregional Meeting,
November 1985 World Health Organization.

68. Starrs AM: Safe motherhood initiative: 20 years and counting. The Lancet,
368(9542):1130-1132.
69. Rosenfield A, Maine D: Maternal Mortality ─A Neglected Tragedy: Where is the M
in MCH? The Lancet 1985, 326(8446):83-85.
70. De Brouwere V, Tonglet R, Van Lerberghe W: Strategies for reducing maternal
mortality in developing countries: what can we learn from the history of the
industrialized West? Tropical Medicine & International Health 1998, 3(10):771-782.
71. Greenwood AM, Bradley A, Byass P, Greenwood B, Snow R, Bennett S, Hatib-N'Jie
A: Evaluation of a primary health care programme in The Gambia. I. The impact of
trained traditional birth attendants on the outcome of pregnancy. The Journal of
tropical medicine and hygiene 1990, 93(1):58-66.
72. Hall M, Chng PK, Macgillivray I: Is Routine Antenatal Care worthwhile? The Lancet
1980, 316(8185):78-80.
73. McDonagh M: Is antenatal care effective in reducing maternal morbidity and
mortality? Health Policy and Planning 1996, 11(1):1-15.
74. WHO: Making pregnancy safer: the critical role of the skilled attendant. A joint
statement by WHO, ICM, and FIGO. WHO, International Confederation of Midwives,
and International Federation of Gynaecologists and Obstetricians,Geneva: 2004.
75. Graham WJ, Bell JS, Bullough CH: Can skilled attendance at delivery reduce
maternal mortality in developing countries. In: Safe motherhood strategies: a review
of the evidence (available at:
http://www.jsieuropeorg/safem/collect/safem/pdf/s2934e/s2934e.pdf (accessed 30
June, 2014) 2001, 17:97-130.
76. Freedman LP, Graham WJ, Brazier E, Smith JM, Ensor T, Fauveau V, Themmen E,
Currie S, Agarwal K: Practical lessons from global safe motherhood initiatives: time
for a new focus on implementation. The Lancet, 370(9595):1383-1391.
77. Haines A, Cassels A: Can the millennium development goals be attained? BMJ:
British medical journal 2004, 329(7462):394.
78. Borghi J, Ensor T, Somanathan A, Lissner C, Mills A: Mobilising financial resources
for maternal health. The Lancet, 368(9545):1457-1465.
79. Powell-Jackson T, Borghi J, Mueller DH, Patouillard E, Mills A: Countdown to 2015:
tracking donor assistance to maternal, newborn, and child health. The Lancet,
368(9541):1077-1087.

80. Johns B, Sigurbjörnsdóttir K, Fogstad H, Zupan J, Mathai M, Tan-Torres Edejer T:

Estimated global resources needed to attain universal coverage of maternal and
newborn health services. Bulletin of the World Health Organization 2007, 85(4):256-
263.
81. Shiffman J: Generating Political Priority for Maternal Mortality Reduction in 5
Developing Countries. American Journal of Public Health 2007, 97(5):796-803.
82. Maine D, Rosenfield A: The Safe Motherhood Initiative: why has it stalled? American
Journal of Public Health 1999, 89(4):480-482.
83. Weil O, Fernandez H: Is safe motherhood an orphan initiative? The Lancet 1999,
354(9182):940-943.
84. AbouZahr C: Cautious champions: International agency efforts to get safe
motherhood onto the agenda. Stud Health Serv Organ Pol 2001, 17:384-411.
85. Lozano R, Wang H, Foreman KJ, Rajaratnam JK, Naghavi M, Marcus JR, Dwyer-
Lindgren L, Lofgren KT, Phillips D, Atkinson C et al: Progress towards Millennium
Development Goals 4 and 5 on maternal and child mortality: an updated systematic
analysis. The Lancet, 378(9797):1139-1165.
86. Shiffman J, Smith S: Generation of political priority for global health initiatives: a
framework and case study of maternal mortality. The Lancet, 370(9595):1370-1379.
87. Shen C, Williamson JB: Maternal mortality, women's status, and economic
dependency in less developed countries: a cross-national analysis. Social Science &
Medicine 1999, 49(2):197-214.
88. Garg B: Safe Motherhood: Social, Economic, and Medical Determinants of Maternal
Mortality: available at:http://www.the-
networktufh.ys.be/sites/default/files/attachments/basic_pages/WHLP%20Safe%20Mo
therhood.pdf (accessed 30 May 2014). TUFH Women and Health Taskforce
September 2006, Second edition
89. McCarthy J, Maine D: A framework for analyzing the determinants of maternal
mortality. Studies in family planning 1992:23-33.
90. Graham WJ, Fitzmaurice AE, Bell JS, Cairns JA: The familial technique for linking
maternal death with poverty. The Lancet 2004, 363(9402):23-27.
91. Loudon I: Maternal mortality in the past and its relevance to developing countries
today. The American Journal of Clinical Nutrition 2000, 72(1):241s-246s.
92. Topping A: Maternal mortality and public opinion. Public Health 1935, 49(0):342-
349.

93. Kunst AE, Houweling T: A global picture of poor-rich differences in the utilisation of
delivery care. Stud Health Serv Organ Policy 2001, 17:297-315.
94. Gill K, Pande R, Malhotra A: Women deliver for development. The Lancet,
370(9595):1347-1357.
95. Bhutta ZA, Ali S, Cousens S, Ali TM, Haider BA, Rizvi A, Okong P, Bhutta SZ,
Black RE: Interventions to address maternal, newborn, and child survival: what
difference can integrated primary health care strategies make? The Lancet,
372(9642):972-989.
96. Campbell OMR, Graham WJ: Strategies for reducing maternal mortality: getting on
with what works. The Lancet 2006, 368(9543):1284-1299.
97. Buekens P: Over-medicalisation of maternal care in developing countries. Stud Health
Serv Organ Policy 2001, 17:195-206.
98. Liljestrand J, Pathmanathan I: Reducing Maternal Mortality: Can We Derive Policy
Guidance from Developing Country Experiences? Critical Elements in Reducing
Maternal Mortality. Journal of Public Health Policy 2004, 25(3/4):299-314.
99. Chowdhury ME, Ahmed A, Kalim N, Koblinsky M: Causes of maternal mortality
decline in Matlab, Bangladesh. Journal of health, population, and nutrition 2009,
27(2):108.
100. W Van Lerberghe, V De Brouwere. Of blind alleys and things that have worked:
history's lessons on reducing maternal mortality. Studies Health Serv Organ Policy
2001, 17:7-33.
101. Seneviratne HR, Rajapaksa LC: Safe motherhood in Sri Lanka: a 100-year march.
International Journal of Gynecology & Obstetrics 2000, 70(1):113-124.
102. Campbell O, Gipson R, Issa AH, Matta N, El Deeb B, El Mohandes A, Alwen A,
Mansour E: National maternal mortality ratio in Egypt halved between 1992-93 and
2000. Bulletin of the World Health Organization 2005, 83(6):462-471.
103. Shiffman J, Stanton C, Salazar AP: The emergence of political priority for safe
motherhood in Honduras. Health Policy and Planning 2004, 19(6):380-390.
104. Lucas AO: Introductory remarks: History of the Prevention of Maternal Mortality
Network. International Journal of Gynecology & Obstetrics 1997, 59, Supplement
2(0):S11-S13.
105. Rosenfield A, Min CJ, Freedman LP: Making Motherhood Safe in Developing
Countries. New England Journal of Medicine 2007, 356(14):1395-1397.

106. Koblinsky M, Matthews Z, Hussein J, Mavalankar D, Mridha MK, Anwar I, Achadi

E, Adjei S, Padmanabhan P, van Lerberghe W: Going to scale with professional
skilled care. The Lancet 2006, 368(9544):1377-1386.
107. WHO: Monitoring emergency obstetric care: a hand book. WHO, UNICEF, UNFPA,
and AMDD. available at:
http://www.unfpa.org/webdav/site/global/shared/documents/publications/2009/obstetr
ic_monitoring.pdf (accessed 13 Nov. 2013). 2009.
108. Derman RJ, Kodkany BS, Goudar SS, Geller SE, Naik VA, Bellad MB, Patted SS,
Patel A, Edlavitch SA, Hartwell T et al: Oral misoprostol in preventing postpartum
haemorrhage in resource-poor communities: a randomised controlled trial. The
Lancet, 368(9543):1248-1253.
109. Mobeen N, Durocher J, Zuberi NF, Jahan N, Blum J, Wasim S, Walraven G, Hatcher
J: Administration of misoprostol by trained traditional birth attendants to prevent
postpartum haemorrhage in homebirths in Pakistan: a randomised placebo-controlled
trial. BJOG: An International Journal of Obstetrics & Gynaecology 2011,
118(3):353-361.
110. Winikoff B, Dabash R, Durocher J, Darwish E, Ngoc NTN, León W, Raghavan S,
Medhat I, Chi HTK, Barrera G et al: Treatment of post-partum haemorrhage with
sublingual misoprostol versus oxytocin in women not exposed to oxytocin during
labour: a double-blind, randomised, non-inferiority trial. The Lancet, 375(9710):210-
216.
111. Starrs A, Winikoff B: Misoprostol for postpartum hemorrhage: Moving from
evidence to practice. International Journal of Gynecology & Obstetrics 2012,
116(1):1-3.
112. ten Hoope-Bender P, Liljestrand J, MacDonagh S: Human resources and access to
maternal health care. International Journal of Gynecology & Obstetrics 2006,
94(3):226-233.
113. Anand S, Bärnighausen T: Human resources and health outcomes: cross-country
econometric study. The Lancet, 364(9445):1603-1609.
114. Kabakian-Khasholian T, Campbell O, Shediac-Rizkallah M, Ghorayeb F: Women's
experiences of maternity care: satisfaction or passivity? Social Science & Medicine
2000, 51(1):103-113.
115. Shiferaw S, Spigt M, Godefrooij M, Melkamu Y, Tekie M: Why do women prefer
home births in Ethiopia? BMC Pregnancy and Childbirth 2013, 13(1):5.

116. Thaddeus S, Maine D: Too far to walk: Maternal mortality in context. Social Science
& Medicine 1994, 38(8):1091-1110.
117. Kowalewski M, Mujinja P, Jahn A: Can Mothers Afford Maternal Health Care Costs?
User Costs of Maternity Services in Rural Tanzania. African Journal of Reproductive
Health / La Revue Africaine de la Santé Reproductive 2002, 6(1):65-73.
118. Borghi J, Hanson K, Acquah CA, Ekanmian G, Filippi V, Ronsmans C, Brugha R,
Browne E, Alihonou E: Costs of near-miss obstetric complications for women and
their families in Benin and Ghana. Health Policy and Planning 2003, 18(4):383-390.
119. Nahar S, Costello A: The Hidden Cost of ‘Free’ Maternity Care in Dhaka,
Bangladesh. Health Policy and Planning 1998, 13(4):417-422.
120. Afsana K: The Tremendous Cost of Seeking Hospital Obstetric Care in Bangladesh.
Reproductive Health Matters 2004, 12(24):171-180.
121. CSA-Ethiopia: Ethiopia Mini Demographic and Health Survey 2014. Central
Statistical Agency, Addis Ababa, Ethiopia July 2014 (available at:
http://www.unicef.org/ethiopia/Mini_DHS_2014__Final_Report.pdf (accessed 15
August 2014) 2014.
122. Black RE, Cousens S, Johnson HL, Lawn JE, Rudan I, Bassani DG, Jha P, Campbell
H, Walker CF, Cibulskis R et al: Global, regional, and national causes of child
mortality in 2008: a systematic analysis. The Lancet, 375(9730):1969-1987.
123. Lawn JE, Cousens S, Zupan J: 4 million neonatal deaths: When? Where? Why? The
Lancet, 365(9462):891-900.
124. Souza JP, Gülmezoglu AM, Vogel J, Carroli G, Lumbiganon P, Qureshi Z, Costa MJ,
Fawole B, Mugerwa Y, Nafiou I et al: Moving beyond essential interventions for
reduction of maternal mortality (the WHO Multicountry Survey on Maternal and
Newborn Health): a cross-sectional study. The Lancet 2013, 381(9879):1747-1755.
125. CSA-Ethiopia: Central Statistica Agency of Ethiopia: Population projection of
Ethiopia for all Regions at wereda level from 2014 - 2017. Available at:
http://www.csa.gov.et/images/general/news/pop_pro_wer_2014-2017 (accessed 09
July 2014). 2014.
126. CIA-USA: The World Fact Book: Africa- Ethiopia, Economy: Ethiopia. available at:
https://www.cia.gov/library/publications/the-world-factbook/geos/et.html (accessed
09 July 2014).

127. UNDP: Human Development Report 2014. Sustaining Human Progress: Reducing
Vulnerabilities and Building Resilience. available at:
http://hdr.undp.org/sites/default/files/hdr14-report-en-1.pdf (accessed 25 July 2014).
United Nations Development Programme 2014.
128. OPHI-MPI: Oxford Poverty and Human Development Initiative (OPHI): Global
multi-dimensional poverty index (MPI) Interactive Data Bank (2014) available
http://www.dataforall.org/dashboard/ophi/index.php (acessed 09 July 2014).
129. Kloos H: Primary Health Care in Ethiopia: from Haile Sellassie to Meles Zenawi.
Northeast African Studies 1998 (new series), 5 (1):83-113.
130. TGE: Transitional Government of Ethiopia. Health Policy of transional government
of Ethiopia, Addis Ababa, TGE. 1993.
131. Teklehaimanot H, Teklehaimanot A: Human resource development for a community-
based health extension program: a case study from Ethiopia. Human Resources for
Health 2013, 11(1):39.
132. UNICEF: In Ethiopia, a far-reaching health worker programme has helped reduce
child mortality across the country. available at:
http://www.unicef.org/infobycountry/ethiopia_70372.html [accessed 03 October
2013]. 2013.
133. FMOH-Ethiopia: H.E Dr. Keseteberhan Admasu (Minister of Health) Speech in the
Two Years Anniversary of: ''A Promised Renewed: Celebrating Progress and
Mapping the Road Ahead'', Washington, D.C. available at:
http://www.moh.gov.et/home/ (accessed 16 July 2014). 2014.
134. WHO: Trends in maternal mortality:1990 to 2010. WHO, UNICEF, UNFPA and The
World Bank estimates. Available at :
http://whqlibdoc.who.int/publications/2012/9789241503631_eng.pdf [accessed 1o
September 2012]
135. Girma M, Yaya Y, Gebrehanna E, Berhane Y, Lindtjorn B: Lifesaving emergency

obstetric services are inadequate in south-west Ethiopia: a formidable challenge to
reducing maternal mortality in Ethiopia. BMC Health Services Research 2013,
13(1):459.
136. FMOH-Ethiopia: Federal Democratic Rpublic of Ethiopia, Ministry of Health: Health
Sector Development Program IV 2010/11 – 2014/15 Final Draft, October 2010.

137. FMOH-Ethiopia: Federal Democratic Republic of Ethiopia Ministry of Health Policy

and Practice Information for Action. Ministry of Health Quarterly Bulletin April
2014, Vol. 6 No. 1.
138. Graham WJ, Campbell OMR: Maternal health and the measurement trap. Social
Science & Medicine 1992, 35(8):967-977.
139. FDR-Ethiopia: Registration of Vital events and National Identity Card Proclamation
(proclamation no.760/2012). . Ethiopian Federal Democratic Republic Federal
Negarit Gazeta 2012.
140. Shahidullah M: The Sisterhood Method of Estimating Maternal Mortality: The Matlab
Experience. Studies in family planning 1995, 26(2):101-106.
141. Font F, González MA, Nathan R, Lwilla F, Kimario J, Tanner M, Alonso P: Maternal
mortality in a rural district of southeastern Tanzania: an application of the sisterhood
method. International Journal of Epidemiology 2000, 29(1):107-112.
142. CSA-Ethiopia: Statistical Tables for the 2007 Population and Housing Census of
Ethiopia: SNNPR1 available at www.csa.gov.et (accessed February 24, 2012). In.:
CSA-Ethiopia; September 2010.
143. Gaym A, Pearson L, Soe KW: Maternity waiting homes in Ethiopia--three decades
experience. Ethiopian Medical Journal 2012, 50(3):209-219.
144. Argaw H: The Health Extension Program (HEP) of Ethiopia: Summary of Concepts,
Progress, Achievements and Challenges. Addis Ababa, Ethiopia. 2007
145. FMOH-Ethiopia: Federal Ministry of Health of Ethiopia: Guidelines for Health

Extension Programme; Addis Ababa, 2007.
146. Graham W, Brass W, Snow RW: Estimating Maternal Mortality: The Sisterhood
Method. Studies in Family Planning 1989, 20(3):125-135.
147. Vyas S, Kumaranayake L: Constructing socio-economic status indices: how to use
principal components analysis. Health Policy and Planning 2006, 21(6):459-468.
148. Hanley JA, Hagen CA, Shiferaw T: Confidence Intervals and Sample-size
Calculations for the Sisterhood Method of Estimating Maternal Mortality. Studies in
Family Planning 1996, 27(4):220-227.
149. IBM: SPSS software: Predictive analytics software and solutions. available at:
http://www-01.ibm.com/software/analytics/spss/ (accessed 20 November 2013).
150. Rothman KJ: Epidemilogy: An Introduction: Oxford University press, 198 Madison
Avenue, New York, New York 10016; 2002.

151. Hennekens CH, Buring JE, Mayrent SL, Doll R: Epidemiology in medicine. Ed.
Sherry L. Mayrent. Reprinted by Lippincott Williams & Wilkins. 530 Walnut St.
Philadelphia, PA 19106 USA; 1987.
152. Harbour R, Miller J: A new system for grading recommendations in evidence based
guidelines. BMJ 2001.
153. Black N: Why we need observational studies to evaluate the effectiveness of health
care. BMJ 1996, 312(7040):1215-1218.
154. Grimes DA, Schulz KF: Descriptive studies: what they can and cannot do. The Lancet
2002, 359(9301):145-149.
155. Grimes DA, Schulz KF: Cohort studies: marching towards outcomes. The Lancet
2002, 359(9303):341-345.
156. Cummings SR, Newman TB, Hulley SB: Designing and Observational study: Cohort
studies. In: Designing Clinical Research: An Epidemiologic Approach by Stephen B
Hulley MD, Steven R Cummings MD, Warren S Browner MD and Deborah Grady.
Volume 2nd edn.: Lippncott Williams & Wilkins; 2001.
157. Irgens LM: The Medical Birth Registry of Norway. Epidemiological research and
surveillance throughout 30 years. Acta Obstetricia et Gynecologica Scandinavica
2000, 79(6):435-439.
158. Rothman K, Greenland S: Causation and causal inference in epidemiology. Am J Publ
Health 2005, 95:S144 - 150.
159. Hanley JA, Hagen CA, Shiferaw T: Confidence intervals and sample-size calculations
for the sisterhood method of estimating maternal mortality. Studies in family planning
1996:220-227.
160. Hulley SB, Newman TB, Cummings SR: Choosing the study subjects: Specification,
Sampling and recruitment. In: Designing Clinical Research: An Epidemiologic
Approach by Stephen B Hulley MD, Steven R Cummings MD, Warren S Browner MD
and Deborah Grady. Volume 2nd edn.: Lippncott Williams & Wilkins; 2001.
161. Browner WS, Newman TB, Cummings SR, Hulley SB: Estimating Sample Size and
Power: The Nitty-gritty. In: Designing Clinical Research: An Epidemiologic
Approach by Stephen B Hulley MD, Steven R Cummings MD, Warren S Browner MD
and Deborah Grady Volume 2nd edn.: Lippncott Williams & Wilkins; 2001.
162. Boyko EJ: Observational research — opportunities and limitations. Journal of
Diabetes and its Complications 2013, 27(6):642-648.

163. Merdad L, Hill K, Graham W: Improving the Measurement of Maternal Mortality:

The Sisterhood Method Revisited. PLoS ONE 2013, 8(4):e59834.
164. Kao S, Chen L-M, Shi L, Weinrich MC: Underreporting and misclassification of
maternal mortality in Taiwan. Acta Obstetricia et Gynecologica Scandinavica 1997,
76(7):629-636.
165. Newman TB, Browner WS, Hulley SB: Enhancing Causal Inferene in Observational
studies. In: Designing Clinical Research: An Epidemiologic Approach by Stephen B
Hulley MD, Steven R Cummings MD, Warren S Browner MD and Deborah Grady.
Volume 2nd edn.: Lippncott Williams & Wilkins; 2001.
166. Bobby JC, Phillips LW, Tybout AM: The Concept of External Validity. Journal of
Consumer Research 1982, 9(3):240-244.
167. Songane FF, Bergström S: Quality of registration of maternal deaths in Mozambique:
a community-based study in rural and urban areas. Social Science & Medicine 2002,
54(1):23-31.
168. CSA-Ethiopia: Ethiopian Demographic and Health Survey 2005. Addis Ababa,
Ethiopia and Calverton, MD,USA, Ethiopian Central Statistical Agency and ORC
Marco, 2006. (http://dhsprogram.com/pubs/pdf/FR179/FR179[23June2011].pdf ).
accessed 11 August 2014. 2006.
169. UNICEF: United Nations Children’s Fund, Every Child’s Birth Right: Inequities and
trends in birth registration, UNICEF, New York, 2013. Available at:
http://www.unicef.org/mena/MENA-Birth_Registration_report_low_res.pdf (accessed
10 January 2014). . 2013.
170. Hill K, AbouZahr C, Wardlaw T: Estimates of maternal mortality for 1995. Bulletin of
the World Health Organization 2001, 79:182-193.
171. Shiferaw T, Tessema F: Maternal mortality in rural communities of Illubabor,
southwestern Ethiopia: as estimated by the" sisterhood method". Ethiopian medical
journal 1993, 31(4):239-249.
172. Kwast BE, Rochat RW, Kidane-Mariam W: Maternal mortality in Addis Ababa,
Ethiopia. Studies in family planning 1986:288-301.
173. Mersha A, Jira C, Demessie S: Community based study on maternal mortality in
Jimma town, south western Ethiopia. Indian journal of public health 1995, 40(2):30-
34.

174. CSA-Ethiopia, ICF-International: Ethiopia Demographic and Health Survey 2011

Central Statistical Agency of Ethiopia and ICF International Addis Ababa, Ethiopia
and Calverton, Maryland, USA available at:
http://www.measuredhscom/publications/publication-fr255-dhs-final-reports.cfm
(accessed 25 June 2013) 2012.
175. WHO: Levels & Trends in Child Mortality Report 2012. Estimated Developed by the
UN Inter-agency Group for Child Mortality Estimation. In. New York, USA.
available at:
http://www.who.int/maternal_child_adolescent/documents/levels_trends_child_mortal
ity_2012.pdf (accessed 20 June 2013); 2012.
176. Admasu K, Haile-Mariam A, Bailey P: Indicators for availability, utilization, and
quality of emergency obstetric care in Ethiopia, 2008. International Journal of
Gynecology & Obstetrics 2011, 115(1):101-105.
177. WHO: Proportion of births attended by a skilled health worker 2008 updates.
Department of Reproductive Health and Research 2008, WHO/RHR/08.22(available
at: http://apps.who.int/iris/bitstream/10665/69950/1/WHO_RHR_08.22_eng.pdf?ua=1
(accessed 11 August 2014)).
178. Paxton A, Maine D, Freedman L, Fry D, Lobis S: The evidence for emergency
obstetric care. International Journal of Gynecology & Obstetrics 2005, 88(2):181-
193.
179. WHO, UNICEF, UNFPA, AMDD: Monitoring Emergency Obstetric Care, a
handbook. WHO publications, 2009 Geneva, Switzerland: World Health
Organization; 2009 available at:
http://whqlibdoc.who.int/publications/2009/9789241547734_eng.pdf (Accessed 10
August 2014).
180. Paxton A, Bailey P, Lobis S: The United Nations Process Indicators for emergency
obstetric care: Reflections based on a decade of experience. International Journal of
Gynecology & Obstetrics 2006, 95(2):192-208.
181. UNFPA-Ethiopia: Joint Press conference discloses Ethiopia on track to achieve
MDG5 (available at
:http://countryoffice.unfpa.org/ethiopia/2014/05/12/9693/joint_press_conference_disc
loses_ethiopia_on_track_to_achieve_mdg5/ (accessed 17 September 2014). UFPA
anf FMOH of Ethiopia Joint statement 2014.

182. Berhan Y, Berhan A: Review of Maternal Mortality in Ethiopia: A Story of the Past
30 Years. Ethiopian Journal of Health Sciences 2014, 24(0 Suppl):3-14.
183. Christian P, Sommer A, West Jr KP: Reducing maternal mortality where rates are
greatest. The Lancet, 368(9553):2122-2123.
184. WHO: Maternal, newborn and child health score cards (by country). World Health
Organization Document Centre (Maternal, newborn, child and adolescent health)
2014, available at:
http://www.who.int/maternal_child_adolescent/documents/countries/indicators/en/
(accessed 15 August, 2014).
185. Costello A, Azad K, Barnett S: An alternative strategy to reduce maternal mortality.
The Lancet, 368(9546):1477-1479.
186. CSA-Ethiopia: Ethiopia Demographic and Health Survey 2000. Central Statistical
Agency of Ethiopia and ORC Macro, Calverton, Maryland, USA, 2001 (available at:
http://dhsprogram.com/publications/publication-FR118-DHS-Final-Reports.cfm
(accessed 20 August 2014). 2000.
187. Otten M, Aregawi M, Were W, Karema C, Medin A, Bekele W, Jima D, Gausi K,
Komatsu R, Korenromp E: Initial evidence of reduction of malaria cases and deaths in
Rwanda and Ethiopia due to rapid scale-up of malaria prevention and treatment.
Malaria Journal 2009, 8(1):14.
188. Wolday D, Meles H, Hailu E, Messele T, Mengistu Y, Fekadu M, Parekh BS,
Wuhib T: Temporal trends in the incidence of HIV infection in antenatal clinic
attendees in Addis Ababa, Ethiopia, 1995–2003. Journal of Internal Medicine 2007,
261(2):132-137.
189. MOH-Ethiopia: Technical and Procedural Guidelines for Safe Abortion Services in
Ethiopia. . Federal Democratic Republic of Ethiopia, Ministry of Health, Addis
Ababa, June 2006.
190. Gebrehiwot Y, Liabsuetrakul T: Trends of abortion complications in a transition of
abortion law revisions in Ethiopia. Journal of Public Health 2009, 31(1):81-87.
191. Singh S, Fetters T, Gebreselassie H, Abdella A, Gebrehiwot Y, Kumbi S, Audam S:
The Estimated Incidence of Induced Abortion In Ethiopia, 2008. International
Perspectives on Sexual and ReproductiveHealth 2010, 36(1):16–25.
192. Prata N, Gessessew A, Campbell M, Potts M: “A new hope for women”: medical
abortion in a low-resource setting in Ethiopia. Journal of Family Planning and
Reproductive Health Care 2011, 37(4):196-197.

Papers I-IV and Appendices

I
RESEARCH ARTICLE
Maternal Mortality in Rural South Ethiopia:

Outcomes of Community-Based Birth
Registration by Health Extension Workers
Yaliso Yaya1,2*, Tadesse Data3, Bernt Lindtjørn1
1 Centre for International Health, University of Bergen, Bergen, Norway, 2 Arba Minch College of Health
Sciences, Arba Minch, Ethiopia, 3 Gamo Gofa Zone Health Department, Arba Minch, Ethiopia
* [email protected]
Abstract
OPEN ACCESS Introduction

Citation: Yaya Y, Data T, Lindtjørn B (2015) Maternal Rural communities in low-income countries lack vital registrations to track birth outcomes.
Mortality in Rural South Ethiopia: Outcomes of We aimed to examine the feasibility of community-based birth registration and measure ma-
Community-Based Birth Registration by Health ternal mortality ratio (MMR) in rural south Ethiopia.
Extension Workers. PLoS ONE 10(3): e0119321.
doi:10.1371/journal.pone.0119321
Academic Editor: Kristina Gemzell-Danielsson,

Methods
Karolinska Institutet, SWEDEN
In 2010, health extension workers (HEWs) registered births and maternal deaths among
Received: April 4, 2014
421,639 people in three districts (Derashe, Bonke, and Arba Minch Zuria). One nurse-su-
Accepted: January 26, 2015
pervisor per district provided administrative and technical support to HEWs. The primary
Published: March 23, 2015 outcomes were the feasibility of registration of a high proportion of births and measuring
Copyright: © 2015 Yaya et al. This is an open MMR. The secondary outcome was the proportion of skilled birth attendance. We validated
access article distributed under the terms of the the completeness of the registry and the MMR by conducting a house-to-house survey in 15
Creative Commons Attribution License, which permits
randomly selected villages in Bonke.
unrestricted use, distribution, and reproduction in any
medium, provided the original author and source are
credited.
Data Availability Statement: Data cannot be made

Results
publicly available as the dataset contains sensitive We registered 10,987 births (814% of expected 13,492 births) with annual crude birth rate
and identifying information. The authors confirm that of 32 per 1,000 population. The validation study showed that, of 2,401 births occurred in the
the data will be made available upon request.
Requests may be sent to the corresponding author,
surveyed households within eight months of the initiation of the registry, 716% (1,718) were
or to Bernt Lindtjørn ( [email protected]). registered with similar MMRs (474 vs. 439) between the registered and unregistered births.
Funding: This study was funded by the Centre for
Overall, we recorded 53 maternal deaths; MMR was 489 per 100,000 live births and 83%
International Health (CIH), University of Bergen, (44 of 53 maternal deaths) occurred at home. Ninety percent (9,863 births) were at home,
Norway. The funders have no role in data acquisition, 4% (430) at health posts, 25% (282) at health centres, and 35% (412) in hospitals. MMR in-
analysis, writing, and decision to submit for
creased if: the male partners were illiterate (609 vs. 346; p= 0051) and the villages had no
publication. The views expressed are that of the
authors and do not necessarily express the policies of road access (946 vs. 410; p= 0039). The validation helped to increase the registration cov-
the funders and the institutions of the authors. erage by 10% through feedback discussions.
PLOS ONE | DOI:10.1371/journal.pone.0119321 March 23, 2015 1 / 17

Maternal Mortality through Birth Registration in South Ethiopia
Competing Interests: The authors have declared Conclusion

that no competing interests exist.
It is possible to obtain a high-coverage birth registration and measure MMR in rural commu-
nities where a functional system of community health workers exists. The MMR was high in
rural south Ethiopia and most births and maternal deaths occurred at home.
Introduction
The global maternal mortality ratios (MMRs) were halved between 1990 and 2010. However,
of all maternal deaths in the world, 99% occur in low-income countries; 36 of the 40 countries
with the highest MMR are in sub-Saharan Africa [1]. The MMR is the conventional key indica-
tor to help monitor progress towards the MDG5 target of reducing maternal mortality by 75%
in 2015 from the level in 1990 [2]. Unfortunately, measuring maternal mortality is difficult in
low-income countries because of limited registration of births and deaths [3]. In 2013, UNICEF
reported that 44% of births in sub-Saharan Africa, and only 7% in Ethiopia, were registered [4].
The continued failure of vital registration in low-income countries was noted as “the single
most critical development failure over the past 30 years”[5].
Following the safe motherhood initiative (SMI) in 1987 and MDG declaration in 2000, sev-
eral alternative approaches, such as survey methods [6–8] and statistical modelling of proxy
data for national and international use [1,9], have been devised to estimate maternal mortality
indices in low-income countries. While these methods provide important information for glob-
al and national planning, the evidence obtained using these techniques are often inconsistent
and sometimes contradictory [10]. Progress towards the planned MDG goal and equity in
health outcomes as well as post-MDG efforts require concrete data from population-based reg-
istries [11].
The study in context

We conducted this study at a time when Ethiopia acknowledged the importance of and made
practical movements towards improving maternal health and formulating a law for compulso-
ry vital registration. As such, the five year (2010–2015) National Health Sector Strategic Plan
emphasizes the intention to improve maternal and newborn health [12]. In addition, the gov-
ernment of Ethiopia was amidst discussions to pass a law for compulsory vital registration,
which was approved in 2012 [13]. To implement the new registration law, the Central Statisti-
cal Agency (CSA) suggested the use of health extension workers (HEWs) as registrars of vital
events in rural Ethiopia [14]. On the other hand, the World Health Organization (WHO) is-
sued the maternal death review (MDR) and maternal death surveillance and response (MDSR)
guidelines as part of international efforts to reduce maternal mortality [15,16]. When we con-
ducted the current study, however, there was a limited such committees or systems for review-
ing maternal deaths in rural Ethiopia. As a result, we did not know whether the continuous
registration of births and surveillance of maternal deaths could result in intended outcomes of
complete birth registration and maternal mortality measurement in rural areas of Ethiopia.
In 2003, Ethiopia adopted a community-based health extension programme (HEP) [17],
and currently over 38,000 HEWs are working in all rural villages. However, the opportunity for
using these community health workers to obtain useful data for public health policy and action
has not been explored. Consequently, the objective of this study was to assess whether HEWs
can effectively register births and actively identify maternal deaths in the rural villages to

measure the magnitude and associated factors for maternal mortality through a community-
based birth registration system.
Methods
Ethics statement
The Ethical Review Committee for the Health Research of Southern Nations Nationalities and
Peoples’ Regional State (SNNPRS) Health Bureau in Ethiopia, and the Regional Committee for
Health Research Ethics of North Norway (REK Nord) in Norway approved the study. Birth
and birth-outcome registration is part of the routine work of the HEWs in Ethiopia, which is
acknowledged by the government. We systematized the registry by preparing a standardized
format and providing technical support. Personal identifiers were removed from the stored
data used for research. We obtained informed verbal consent from respondents for the valida-
tion study of house-to-house survey and the responses were recorded on the questionnaire as
“accepted” or “declined” to participate. Written consent was not considered because a large
number of the respondents were illiterate and the Ethics Committee approved the verbal
consent procedure.
Study area
The Ethiopian government has autonomous regional states within the Federal Republic. In
turn, regional states are subdivided into zones (provinces), Woredas (districts), and Kebeles
(villages). A zone is a cluster of 10–15 districts, and a district is a group of 20–50 villages. A
Kebele is the lowest administrative structure and is comprised of 1,000–1,500 households. This
study was conducted in three districts (Arba Minch Zuria, Bonke, and Derashe) in two zones
(Gamo Gofa and Segen Area Peoples') in the Southern Nations, Nationalities, and Peoples’ Re-
gion (SNNPR, Fig. 1). The Gamo Gofa Zone (population = 1,740,828 people in 2010) [18], the
centre of which is at Arba Minch, is 505 km from Addis Ababa to the southwest and the Segen
Area Peoples’ Zone (636,794 residents in 2010) [18] is 575 km from Addis Ababa.
Bonke, with a population of 166,913 people in 2010, had no hospital providing comprehen-
sive emergency obstetric care at the time of the study. The nearest such service was at Arba
Minch Hospital, which is 50–150 km from the villages of Bonke. Arba Minch Zuria, with a
population of 179,785 people, has a hospital, although the largest proportion of the population
lives in the highlands far from the hospital and driveable roads. Derashe, with a population of
141,589 has a district hospital in the main town of Gidole, as well as well-functioning maternity
waiting homes, traditional thatched huts built in the hospital compound, where mothers with
high-risk pregnancies are referred and observed until delivery [19].
Sampling and study participants

Fig. 2 presents the study profile. In 2008, the MMR in Ethiopia was 590 per 100,000 live births
(LBs) [9]. Assuming this would be comparable for the study area, we expected there could be a
10% decline in two years resulting in an MMR of 531 (95% CI: 413, 669) per 100,000 LBs in
2010. Thus, we expected 70 maternal deaths in a year (95% CI: 55, 88) out of estimated 13,492
births (13,223 LBs) in a population of 421,639 people. LBs were approximated 98% of all births
in the area [20]. To estimate the expected number of births, we used an annual crude birth rate
(CBR) of 32 per 1,000 population based on the following two sources of birth rate information:
a finding from a household survey in 2010 in one of the study districts (Bonke) [20], and the
same estimate by The World Bank of CBR in Ethiopia for 2010 [21]. To identify group

Fig 1. The map of the study area within southern Ethiopia.

doi:10.1371/journal.pone.0119321.g001
differences in the MMR, we assumed the number of maternal deaths amongst births deter-
mined above would provide sufficient data.
We purposely selected three districts with the number of residents expected to produce the
above estimated births and maternal outcomes. The districts were assumed to represent the
area in terms of health services, demographics, and road access. In these districts, we included
all kebeles (villages), except those where the HEWs were sick or on maternity leave at the time
of starting the registration. We used OpenEpi software (Open Source Epidemiologic Statistics
for Public Health version 3.01,www.openepi.com) to calculate the sample size.
The HEP and the HEWs

The HEP is a community-based healthcare system with two female HEWs serving a rural vil-
lage of 1,000–1,500 households. Most of the HEWs have completed a 10th grade education and
received one year of general health training. Their work focuses on family health (child vacci-
nations, family planning, antenatal care, and assisting normal deliveries) and health promo-
tion. HEWs are expected to routinely visit each household in their catchment once a month,
prioritizing households with pregnancies, newborns, and sick persons. HEWs are part of the
permanent health workforce and receive a monthly salary of 40–50 USD from the government
based on their years of service. In addition, 5–10 lay-women known as volunteer health pro-
moters (VHPs), assist the work of HEWs by informing of households with a recent delivery,
sick people, and deaths in the sub-villages.
Data collection procedures (the birth registry)

We conducted one week training at each woreda centre for HEWs, supervisors, and the district
health authorities before the registry started. Supervisors were experienced nurses (one per dis-
trict), who helped the HEWs in reviewing and classifying deaths, monitoring the quality of
data, and transferring the registered information from HEWs to the central data clerk. During

Fig 2. The profile of the birth registry.

the training, we clarified the WHO ICD-10 definition and classification of maternal deaths
[22]. Accordingly, if a woman died during ante- or intra-partum periods, or within six weeks
after termination of a pregnancy and her pregnancy status was known, her death was consid-
ered a maternal death if the death was not because of an accident or incident such as suicide.
We also used extractions from the WHO maternal death review (MDR) manual published in
2004 to determine the cause of deaths [15]. As such, diagnosing the cause of death was based
on symptomatic approaches such as convulsions attributed to hypertensive disorders, fevers to
infections, and excessive bleeding due to haemorrhage.

The specific registration and maternal death ascertainment procedure is presented as fol-
lows. HEWs visited homes within hours or days after the pregnancy ended depending on the
distance and the speed of notification from the sub-village VHPs or families. At the household,
HEWs assessed and registered birth and births conditions. The HEWs continued the follow-up
until a maternal death was occurred or six-week post-partum. This collection of information
was similar to births that occurred at home and in health facilities because all births were avail-
able for recording at homes. In addition, in households in which a woman of reproductive age
died without giving birth, HEWs critically reviewed the conditions at the time of death to de-
termine the pregnancy status of the deceased and determine the probable cause of death. Hus-
bands or fathers of the baby (FOBs) were primary sources of information for maternal deaths;
however, in the cases where obtaining information from the husbands or FOBs was not possi-
ble, adult members of the family helped in providing information.
HEWs registered the data in printed birth registry books (Fig. 3). The book contained im-
portant socio-demographic variables, such as the distance of the village from the nearest health
centre and the nearest hospital recognized by the respective district health offices, as well as the
type (quality) of road to the village as a general heading information. The actual body of the
book rows contained personal background information, such as education of the mother and
father and age of the mother. In addition, the woman’s parity, the place of birth, the attendant
of birth, the condition of the newborn at birth (alive or stillbirth), the gender of the foetus, and
maternal deaths (including the place, cause, and time) were among the variables. Registration
was made in duplicate and the first copy was detached and sent to the Research and Training
Centre at Arba Minch Hospital, while the second copy remained with the book in the village.
Most births were registered within 24 hours of delivery, unless there was a special reason for a
delay (births in distant health institutions, where the household was far from the HEW station
or HEWs were not informed in a timely manner). Similarly, most maternal deaths were identi-
fied immediately. Nevertheless, HEWs made a final follow-up home visit six weeks after birth
or abortion when death information was not obtained prior to the stated deadline.
Outcomes
The primary outcomes were the coverage of birth registration (percentage registered out of the
estimated) and the MMR. The secondary outcome was the proportion of skilled birth atten-
dance, facility deliveries supervised by skilled professionals.
Data quality control (the validation study)

To check the validity of the registration eight months after the start of the registration, we con-
ducted a house-to-house survey in 15 of the 30 rural villages in the Bonke. Data collectors who
had completed the 12th grade visited every household and searched for a birth or pregnancy
outcome since the start of the birth registry. For births already registered in the birth registry,
they checked the content (date of birth, date of death, and baby’s gender). The unregistered
were recorded and the data were transferred to the registry book. Based on the findings of the
validation study, we discussed the feedback with the HEWs and supervisors to improve the
coverage of the registration.
Data analysis
We entered, checked, and analyzed the registry and validation data using the statistical package
for social sciences (SPSS-16) describing the results in tables showing proportions and means.
To show the variation in maternal mortality, we used a chi-square test. For the validation
study, we produced a descriptive table showing the proportion of births registered and

Fig 3. The birth registry format.


Table 1. Socio-demographic data on parents, services, and infrastructures in the birth registry dis-
tricts of south Ethiopia in 2010.
Background variable Value

Age of mother (years): mean (SD) 28.1 (45)
Education of fathers N = 10,987*
Illiterate: no. (%) 6,001 (546)
Literate: no. (%) 4,986 (454)
Education of mothers N = 10,987*
Illiterate: no. (%) 8,454 (770)
Literate: no. (%) 2,533 (230)
Health Centre distance (km): median** (IQR) 10 (5–18)
Hospital distance (km): median (IQR) 40 (24–67)
Pregnancy (gravidity): mean (SD) 3.7 (22)
Antenatal visits: mean (SD) N = 10,987 2.4 (14)
No antenatal: no. (%) 1,655 (151)
1–2 visits: no. (%) 3,767 (343)
3 or more: no. (%) 5,565 (507)
Place of Delivery (%): N = 10,987
Home: no. (%) 9,863 (90)
Health post: no. (%) 430 (4.0)
Health Centre: no. (%) 282 (25)
Hospital: no. (%) 412 (35)
Road access (%): N = 10,987
All-weather: no. (%) 4,214 (384)
Dry weather: no. (%) 5,446 (496)
No car road: no. (%) 1,327 (120)
Note:
* Education: illiterate are those cannot read/write and had no formal education, literate include those who
can read/write and completed higher education,
** 57% of households with births were within 10 km of health centres (10 km is the government target for
access)
doi:10.1371/journal.pone.0119321.t001
unregistered out of the births found during the validation survey. We made a cross-tabulation
for crude analysis to determine the risk of maternal deaths among registered births compared
to unregistered and the effect of antenatal follow-up and distance from HEW station on the
likelihood of births being registered.
Results
Socio-demographic characteristics of registered births
Table 1 presents the background information about the parents and the maternal services re-
ceived during pregnancy and delivery. We registered 10,987 births (5,612 [([511%] boys and
5,375 [489%] girls). The average age of the mothers was 281 (SD = 45) years, and the median
number of pregnancies was 3 (IQR = 2–5). The illiteracy rate was 77% (8,454/10,987) among
mothers, and 546% among the husbands and FOBs (6,001/10,987). The median distance to
health centres was 10 km (IQR = 5–18), and 57% (6,236/10,987) of the births had a health cen-
tre within 10 km. The median distance to hospitals was 40 km (IQR = 24–67).

Table 2. District distribution of births, birth outcomes, and services during pregnancy and delivery in south Ethiopia in 2010.
Districts
Derashe Bonke Arba Minch Overall

All births 3,134 3,571 4,282 10,987
Live births 3,099 3,529 4,205 10,833
Maternal deaths (number) 14 18 21 53
Antenatal visits: Mean (SD)* 25 (16) 21 (14) 25 (13) 24 (14)
No antenatal: no (%) 546 (174) 669 (187) 440 (103) 1,655 (151)
1–2 visits: no (%) 826 (264) 1,418 (397) 1,523 (356) 3,767 (343)
3 and more: no (%) 1,762 (562) 1,484 (416) 2,319 (541) 5,565 (506)
Place of delivery no (%) N = 10,987
Home 2843 (907) 3,183 (891) 3,837 (896) 9,863 (90)
Health post 40 (13) 290 (82) 100 (24) 430 (40)
Health Centre 25 (08) 81 (22) 176 (41) 282 (25)
Hospital 226 (72) 17 (05) 169 (39) 412 (35)
Delivery attendant no (%)
Family member or relative 234 (75) 1,823 (511) 1,937 (453) 3994 (364)
Traditional birth attendant 2,434 (777) 1,027 (288) 1,364 (318) 4825 (440)
Health Extension Worker 215 (68) 623 (174) 636 (149) 1,474 (134)
Health professionals (skilled) 251 (8.0) 98 (2.7) 345 (8.0) 694 (6.3)
Note:
* SD, standard deviation
Completeness of birth registration

The 10,987 registered births were 81.4% of the 13,492 expected births based on the annual CBR
for the year. Because the CBR was estimate-based, we made sensitivity analysis of the expected
births by increasing the annual CBR by two to 34 and again by decreasing by two to 30 per
1000 population. If the CBR was 34, the number of expected births was 14, 336 of the 421, 639
residents in the registry villages. This decreased the registration coverage from 81.4% to 77%,
and if the CBR was 30, the registration coverage rose to 87% from our best estimate of 81.4%
because the expected number of births was 12,649.
Maternal mortality outcomes

We registered 53 maternal deaths (Table 2), yielding an MMR of 489 per 100,000 live births
The MMR of 489 per 100,000 live births was within the 95% confidence interval of the expected
MMR of 531 (95% CI: 413–669). Table 3 compares the MMR estimates between the birth regis-
try (489), the validation study (474), the previous household survey finding from the area
(425), and the national estimates by IMHE (590), UN (350), and DHS (676) per 100,000 live
births. In the birth registry study, five mothers (94%) died during pregnancy, 21 (396%) died
during labour and 27 (51%) died within six weeks post-partum. Of the maternal deaths, 35.8%
(19/53) were due to bleeding, 226% (12 /53) due to infection, 17% (9/53) because of hyperten-
sive disorders, 132% (7/53) because of obstructed labour, and 11% (6/53) registered as “oth-
ers”. Additional inquiries indicated that three of the six cases categorized as ‘other causes’ and
two cases in the haemorrhage category were probable complications of abortion. Of all 53 ma-
ternal deaths identified, 83% (44) occurred at home and 17% (9) were in health institutions.

Table 3. Variations in maternal mortality across variables, south Ethiopia, 2010.
Variable Maternal deaths Live births MMR*(95% CI) p-value (2-tail)

District
Bonke 18 3,529 510 (312, 789)
Arba Minch 21 4,205 500 (318, 750)
Derashe 14 3,099 452 (257, 740)
Maternal age
20 4 709 564 (180, 1,355)
21–35 45 9,475 475 (351, 630)
36 4 649 616 (196, 1,480)
Parity (no. of births)
3 32 5,760 556 (386, 774)
4 21 5,073 414 (263, 621)
Mother’s education**
Illiterate 40 8,331 480 (347, 647)
Literate 13 2,502 520 (290, 865)
Father’s education**
Illiterate 36 5,913 609 (433, 833) 0051
Literate 17 4,920 346 (208, 541)
Antenatal visits
No ANC 11 1,628 676 (356, 1,172)
1–2 visits 17 3,711 458 (276, 717)
3 or more 25 5,494 455 (301, 661)
Place of birth
Home 44 9,757 451 (332, 599)
Institution 9 1,066 844 (412, 1,544)
Distance to health centre
10 35 6,151 569 (403, 782)
11 18 4,649 387 (237, 600)
Distance to hospital
25 14 3196 438 (250, 716)
26 39 7,604 513 (370, 693)
Road to the village§
All weather drive 17 4,150 410 (247, 642)
Dry weather drive 24 5,377 446 (293, 653)
No driveable road 12 1,269 946 (513, 1,602) 0039
Sickness in pregnancy
Yes 24 1,361 1,763(1,159, 2,572) 0000
No 29 9,472 306 (209, 433)
Total 53 10,833 489 (366, 628)
Note:
*maternal mortality ratio per 100,000 live births. Cells with P-value > 0.05 (with no statistical significance) are left empty. MMR in the parenthesis are 95%
CIs.
**Education: illiterate are those who cannot read/write and had no formal education, literate include those who can read/write and more educated up to
higher education.
§
Compared only all-weather road against no driveable road

Place of birth and assistance during labour

Table 2 also shows the place of delivery and who assisted the births. Of the registered births,
85% of the mothers (9332/10987) received antenatal care. However, 90% of the mothers (9863/
10987) delivered at home, 4% (430) at health stations (two-room buildings staffed by HEWs),
25% (282) at health centres (staffed by nurses and midwives), and 35% (412) in hospitals.
One-third of the births were (364% [3,994/10,987]) were assisted by family members and rela-
tives, 44% (4,825) by traditional birth attendants (TBAs), 134% (1,474) by HEWs, and 63%
(694) by skilled professionals (physicians, health officers, midwives, and nurses). Regarding
road access, 616% of the births (6,773/10,987) took place in villages without driveable-road or
access to driveable road during the dry season only.
Variations in maternal mortality

The MMR was similar in the three districts, and there was no difference between educated and
illiterate mothers (Table 3). However, the MMR was higher in the villages where there was no
driveable road access compared to villages where there was at least a dry-weather road
(946 vs. 446) and an all-weather road [946 vs. 410; χ2 (df): 611 (2), p = 0039]. Maternal mor-
tality was six times higher among mothers who had illness of any kind during pregnancy, com-
pared to those who had not complained of any illness [1,763 vs. 306; χ2 (df): 486 (1),
p < 00001], and increased when the male partners were illiterate, compared to those who were
literate [609 vs. 346; χ2 (df: 38 (1), p = 0051].
Results from the validation study

Table 4 shows findings from a house-to-house validity study compared to the birth registry. Of
the 2,401 births identified during house-to-house checks, 1,718 (716%) were registered and
683 (284%) were not registered. Births to women who attended antenatal clinics (ANCs) were
more likely to be registered compared to mothers who did not have any ANC visits (746%
[1413/1895] vs. 603% [305/506]; RR = 124 [95% CI: 115–133]). Births that had occurred
within 5 km from a HEW station were also more likely to be registered compared to births that
occurred > 6 km from a HEW station (752% [1072/1425] vs. 662% [646/976]); RR = 114
Table 4. Results from house-to-house validation study to check completeness of birth registry in south Ethiopia during 2010.
Variables registered number (%) unregistered number (%) checked (100%) RR** (95% CI)
All births 1,718 (716) 683 (284) 2,401
Live births 1,698 (715) 675 (285) 2,373
Maternal death 8 3 11
MMR* 474 439 464 1.06 (0.28, 3.98)§
Antenatal care visit
Yes 1,413 (746) 482 (254) 1,895 1.24 (1.15, 1.33)
No 305 (603) 201 (397) 506 Ref
Distance to HEWβ station
5 km 1,072 (752) 353 (248) 1,425 1.14 (1.08, 1.20)
6 km 646 (662) 330 (338) 976 Ref
Note:
* MMR, maternal mortality ratio per 100,000 live births,
** RR, relative risk,
§ relative risk of MMR among registered births compared to unregistered. HEWβ, health extension worker

Table 5. Estimates of the maternal mortality ratio per 100,000 live births in the study area compared with national estimates in Ethiopia during the
years around 2010.
MMRβ Year Source Estimated for Remarks

489 2010 Birth registry Study area This paper
474 2010 Validation of birth registry study area This paper
425 2010 Household survey Study area Yaya et al [20]
350 2010 UN modelled estimate National level WHO and co [1]
676 2010 DHS* National level CSA-Ethiopia** [6]
590 2008 IHME modelled estimate*** National level Hogan et al [9]
Note: MMR,
β
maternal mortality ratio per 100,000 live births, DHS,
* Demographic and Health Survey, CSA,
** Central Statistical Agency IHME,
*** Institute of Health Metrics and Evaluation, Washington University, USA.
[95% CI = 108–120]). The MMR was similar among registered and unregistered (474 vs. 439)
per 100,000 LBs, (RR = 106 [95% CI: 028–398]).
Comparing maternal mortality in the study area with national estimates

To provide a better understanding of the MMR in the study area obtained by using different
methods of measurement with the national estimates for Ethiopia, we provide a summary table
(Table 5).
Discussion
In 75 rural kebeles, we registered 81% of expected births in nearly half a million residents. The
MMR was 489 per 100,000 LBs. In addition, four of every five maternal deaths occurred at
home without the attention of health facilities in the area. A validation study also showed a
high initial coverage of birth registration out of births identified through checks by household
visits, and there was a similar MMR between registered and unregistered births. These findings
suggest that it is possible to register a high percentage of births expected in rural communities
using community health workers. In addition, the community-based birth registry appears to
be a useful tool to identify and measure maternal mortality. Majority (90%) of the registered
births took place at home while > 90% of maternal deaths occurred during intra- and post-
partum periods. Haemorrhage and infections were the leading causes of maternal deaths. The
MMR was higher in remote areas without roads, amongst couples in which the male partners
were illiterate, and among mothers who experienced illnesses during pregnancy.
We are not aware of other studies that have tested the feasibility of registration of births and
pregnancy outcomes such as maternal deaths in rural Ethiopia. The 2013 UNICEF report
showed that only 7% of births are registered in Ethiopia, which is the third lowest in the world
ahead of Liberia (4%) and Somalia (3%) [4].
Unfortunately, because of resource constraints we were not able to conduct a baseline sur-
vey of fertility for 2010 in the study area. Consequently, we assessed the completeness of the
birth registry using the coverage of the registry out of expected (estimated) births in the area,
depending on a finding from a previous survey (CBR = 32 per 1,000 population between 2006
and 2010) [20]. Expecting the number of births through the annual CBR estimation may
under- or over-estimate the true number of births, and the coverage may vary from the 81.4%

we report. Additionally, we conducted a validation study for births occurring in 8 months, and
this presents a problem in measuring the annual CBR. As such, the validation study only serves
the purpose of determining what proportion of actually observed births at households were
registered and whether or not there was a difference in an outcome of interest (MMR) between
registered and unregistered births. The validation study cannot help to estimate annual CBR
that needs 12 months of data on births.
Given the continuous decline in fertility in rural areas, as indicated by a decrease in CBR per
1,000 population (43 in 2000 to 28 in 2014) according to the DHS showed by findings from
DHS [23], the CBR in 2010 may have been < 32 per 1,000 people or we are uncertain whether
or not the CBR was even higher. We did a sensitivity analysis by using annual CBRs of 34 and
30 to compare the registration coverage against the CBR of 32 used in the analysis. The analysis
provided registration coverage between 77% and 87%. The true CBR in the area may be be-
tween 30 and 32 per 1,000 population, which implies that the actual coverage of birth registra-
tion may have been > 81%. Therefore, we argue that the registration coverage for the study
was high for a beginning community-based birth registry in rural settings.
Nevertheless, the uncertainty on the number of expected births does not affect our MMR
because we calculated MMR from registered births rather than expected births. The concern,
however, is whether there was a difference in MMR between registered and unregistered births.
We attempted to address this partly by doing a validation study, which resulted in a similar
MMR between the two groups. Theoretically, both the continuous birth registration and the
validation survey may give an under- or over-estimate of the MMR. In practice, the problem
related to measuring maternal mortality is under-reporting rather than over-reporting. As
such, reported maternal deaths under estimate up to 30% of the actual MMR worldwide [16].
Important factors related to under-reporting of maternal mortality among registered preg-
nancy outcomes are problems of identifying early pregnancy maternal deaths before the preg-
nancy is clearly recognized and deaths due to abortion because of secrecy and stigma [24].
Thus, we made a rigorous effort to review every adult woman’s death to determine whether it
was pregnancy-related, mainly by using the advantage of a close relationship between HEWs
and the people in their villages. Nevertheless, we recognize that it is difficult to avoid under-re-
porting and we cannot estimate how many early pregnancy, and abortion-related deaths went
unnoticed. Future studies may consider beginning the registration of pregnancies (instead of
pregnancy outcomes used in this study) to better capture maternal deaths during pregnancy.
Another important concern of under-reporting is the limited knowledge of the proportion of
maternal deaths among unregistered pregnancy outcomes. Our validity check showed no dif-
ference in the MMR between registered and unregistered births. Nevertheless, because the vali-
dation study took place in 15 of 75 villages during the 8th of a 12-month registration, there may
have been limitations to representing a complete picture of similarity.
While these limitations are acknowledged, we have reported an MMR estimate closer to re-
ality and the MMR herein may be one important step in measuring maternal mortality through
a community-based birth registry in rural Ethiopia. Furthermore, the current finding of the
proportion of maternal deaths at facilities and homes is also consistent with our previous re-
port [20,25]. In addition, our findings showed that 83% of maternal deaths occurred at home
without access to a health institution, which is similar with the results of a study from Mozam-
bique that tracked maternal mortality through active community-based approaches, and
showed that health institutions did not identify 86% of maternal deaths that occurred in their
area [26]. The percentage of maternal deaths at home was less than the proportion of births in
the same place (83% vs. 90%). The reason could be that some severe cases selectively visited
health facilities and died in the facilities. The severity-based selective facility utilization argu-
ment is also supported by the findings of a higher MMR among facility births compared to

births at home (844 vs. 451) in the current study. These results have two implications. First, the
finding of four of five maternal deaths occurring at home means health facilities had no means
to avoid these deaths and were not able to identify and record when the deaths occurred. Sec-
ond, the higher MMR among facility deliveries implies that the health institutions had low abil-
ity to save mothers who had sought help.
In the study area, we have limited documentation to describe progress in terms of maternal
mortality in previous years. A household survey in one of the districts to measure maternal
mortality between 2006 and 2010 provided an MMR of 425 per 100,000 LBs [20], which is sim-
ilar to the current finding. In addition, we published maternal mortality indices (MMR and
lifetime risk) from the area using the indirect sisterhood method [27]. Nonetheless, because of
the indirect nature, the estimate refers to a time of more than a decade ago and indirect sister-
hood estimates provide the order of magnitude with limited value to oversee trends.
The MMR we report (489 per 100,000 LBs in 2010) is higher than the level of reduction re-
quired to attain the MDG5 target (from 968 in 1990 to 242 in 2015 in Ethiopia) [9]. The MMR
was also higher than the joint estimate provided by WHO, UNFPA, UNICEF, and The Wold
Bank (350) for 2010 in Ethiopia [1], but lower than the DHS reported national estimate (676)
for a similar period [28]. However, our finding was similar to most results from previous com-
munity-based studies in Ethiopia. The MMR per 100,000 LBs was 402 in Jimma (1990) using a
cross-sectional survey [29], 440 in Butajira (1996) using a surveillance approach [30], 570 in
Illubabor (1991) using the indirect sisterhood method [31], and 566 in Addis Ababa (1983)
using a household survey [32].
Nevertheless, given decades of time between the aforementioned studies and the current
study, the MMR may have been different in this study area compared to those provinces. An al-
ternative explanation could be that the survey and periodic surveillance techniques used in the
aforementioned studies resulted in under-estimations of MMR at that time. This could also be
explained by the fact that Ethiopia had a high MMR estimate nationally during the time when
these studies were conducted [33]. As such, the prospective method we used and the presence
of the HEWs within the villages that increases the awareness of important events might have
helped our study provide a better estimation of the current MMR.
We showed that there was no difference in the MMR between districts with better health in-
stitutions (Arba Minch and Derashe) and a district with less poor health facilities (Bonke).
This may be explained by the low utilization rate of existing health institutions for deliveries in
all the districts. Hence, this emphasizes the importance of improving people's behaviours with
respect to utilizing institutional delivery by skilled professionals in addition to distributing and
strengthening health facilities. The findings of a high MMR in the remotest villages without
driveable roads highlight the inequalities in health outcomes experienced by these women. Ma-
ternal mortality was also higher where husbands and FOBs were illiterate. This may explain the
importance of male partners as decision makers for receiving care (health), as well as their gen-
eral contribution as providers to improved living conditions (wealth). This finding is in agree-
ment with the suggestions of the Oxford Multi-dimensional Poverty Index (MPI) Study, which
described the importance of an educated person in a household for a positive health outcome
[34]. In addition to the aimed outcomes, the study also showed the existence of substantial
missed opportunities in maternal care by the health system. Fortunately, 85% of pregnant
women visited health workers for antenatal check-ups at least once, whereas only 10% returned
for delivery at the health facilities for supervised delivery.
The weaknesses of the current study follow. First, although we made a rigorous effort to
identify early maternal deaths, an important limitation of the study was our inability to explain
the amount of unnoticed early pregnancy-related maternal deaths. Specifically, we might have
missed some of the abortion-related deaths because of the stigma and secrecy associated with

abortion. Second, although we achieved a high coverage of > 80% registration, not all expected
births were registered, thus leaving concerns about whether or not the MMR is higher or at
least different among unregistered births. Our finding was a significant increase from the 7%
national birth registration in Ethiopia [4], and almost non-existent in rural areas. Third, despite
similar MMRs between registered and non-registered births, our validation study showed that
HEWs were more likely to register births that occurred near their station and those who at-
tended antenatal care, highlighting potential selection bias. Fourth, the death ascertainment
method through the HEWs was less valid compared to the standard techniques of death confir-
mation by physicians or using autopsy [35].
Nevertheless, we believe that the diagnostic technique we used is better than asking a family
member about a death that occurred years ago in survey studies. Furthermore, the standard
methods of ascertaining maternal death are less likely to reach rural communities in low-in-
come countries. Hence, it may be important to use and improve available opportunities, such
as community health workers (HEWs in Ethiopia) to prospectively identify, review, and record
maternal deaths, even when it means less accurate diagnosis of the cause of death in resource-
limited settings.
Finally, because of the short period of observation, we cannot show in this paper whether
community-based registration contributes for the reduction of MMR. Future studies may try
to address these concerns. In addition, although it is difficult to suggest similar achievements
in all areas, the study can be repeated in any part of Ethiopia because all villages have the same
HEWs working under similar conditions. The same can also be applied in low-income coun-
tries that have an organized community health workforce.
Conclusion
It is possible to register most births in rural Ethiopia through the HEWs and use the registry as
a tool to measure maternal mortality. The MMR was high in the study area compared to the re-
ductions needed to attain MDG5 and most births and maternal deaths occur at home without
the attention of the health service.
Acknowledgments
The Centre for International Health (CIH) of the University of Bergen in Norway funded this
study. The funders had no role in data acquisition, analysis, writing, and decision to submit for
publication. The views expressed are those of the authors and do not necessarily express the
policies of the funders and the institutions of the authors. We would like to thank the health ex-
tension workers and supervisors in Arba Minch Zuria, Derashe, and Bonke who conducted
and supervised the birth registration. We kindly appreciate the support we received from the
health authorities in the Derashe, Bonke, and Arba Minch Zuria districts. We are also grateful
to the encouragement and support provided from the Gamo Gofa Zone Health Department
and the Southern Nations Nationalities and Peoples’ Regional State Health Bureau in Ethiopia.
Author Contributions
Conceived and designed the experiments: YY TD BL. Performed the experiments: YY TD BL.
Analyzed the data: YY TD BL. Wrote the paper: YY TD BL.
References
1. WHO, UNICEF, UNFPA, World-Bank Trends in maternal mortality:1990 to 2010. WHO, UNICEF,
UNFPA and The World Bank estimates. Available: http://whqlibdoc.who.int/publications/2012/
9789241503631_eng.pdf. Accessed 10 September 2012.

2. Bhutta ZA, Chopra M, Axelson H, Berman P, Boerma T, et al. (2010) Countdown to 2015 decade report
(2000–10): taking stock of maternal, newborn, and child survival. Lancet 375: 2032–2044. doi: 10.
1016/S0140-6736(10)60678-2 PMID: 20569843
3. Graham W, Ahmed S, Stanton C, Abou-Zahr C, Campbell O (2008) Measuring maternal mortality: An
overview of opportunities and options for developing countries. BMC Medicine 6: 12. doi: 10.1186/
1741-7015-6-12 PMID: 18503716
4. UNICEF United Nations Children’s Fund, Every Child’s Birth Right: Inequities and trends in birth regis-
tration, UNICEF, New York, 2013. Available: http://www.unicef.org/mena/MENA-Birth_Registration_
report_low_res.pdf (accessed 10 January 2014).
5. AbouZahr C, Cleland J, Coullare F, Macfarlane SB, Notzon FC, et al. (2007) The way forward. Lancet
370: 1791–1799. PMID: 18029003
6. CSA-Ethiopia, ICF-International Ethiopia Demographic and Health Survey 2011 Central Statistical
Agency of Ethiopia and ICF International Addis Ababa, Ethiopia and Calverton, Maryland, USA 2012.
Available: http://wwwmeasuredhscom/publications/publication-fr255-dhs-final-reportscfm (accessed
25 June 2013).
7. Campbell O, Gipson R, Issa AH, Matta N, El Deeb B, et al. (2005) National maternal mortality ratio in
Egypt halved between 1992–93 and 2000. Bulletin of the World Health Organization 83: 462–471.
PMID: 15976898
8. Graham W, Brass W, Snow RW (1989) Estimating Maternal Mortality: The Sisterhood Method. Studies
in Family Planning 20: 125–135. PMID: 2734809
9. Hogan MC, Foreman KJ, Naghavi M, Ahn SY, Wang M, et al. (2010) Maternal mortality for 181 coun-
tries, 1980–2008: a systematic analysis of progress towards Millennium Development Goal 5. The Lan-
cet 375: 1609–1623.
10. Graham WJ, Adjei S (2010) A Call for Responsible Estimation of Global Health. PLoS Medicine 7:
e1001003. doi: 10.1371/journal.pmed.1001003 PMID: 21151346
11. Jamison DT, Summers LH, Alleyne G, Arrow KJ, Berkley S, et al. (2013) Global health 2035: a world
converging within a generation. The Lancet 382: 1898–1955.
12. FMOH-Ethiopia (2010) Health Sector Development Program IV 2010/11–2014/15 Final Draft, October
2010. Federal Democratic Rpublic of Ethiopia, Ministry of Health.
13. FDR-Ethiopia (2012) Registration of Vital events and National Identity Card Proclamation (proclamation
no.760/2012). Ethiopian Federal Democratic Republic Federal Negarit Gazeta.
14. Samia ZG (2011) Source of Vital Statistics in Ethiopia: United Nations Expert Group Meeting on Inter-
national Standard for Civil Registration and Vital Statistics System, New York, 27–30 June 2011. Avail-
able: https://unstats.un.org/unsd/demographic/meetings/egm/CRVS2011/EGM%20PowerPoints/
Session%201-04%20Ethiopia.ppt (accessed 25 July 2014).
15. WHO (2004) Beyond the numbers: reviewing maternal deaths and complications to make pregnancy
safer. Available: http://www.who.int/maternal_child_adolescent/documents/9241591838/en/ (ac-
cessed 20 June, 2013). World Health Organization.
16. WHO (2013) Maternal death surveillance and response: technical guidance. Information for action to
prevent maternal death. Available: http://www.who.int/maternal_child_adolescent/documents/
maternal_death_surveillance/en/) accessed 15 August 2014. World Health Organization and partner
organizations.
17. Koblinsky M, Tain F, Gaym A, Karim A, Carnell M, et al. (2010) Responding to the maternal health care
challenge: The Ethiopian Health Extension Program. Ethiopian Journal of Health Development 24.
18. CSA-Ethiopia (2010) Statistical Tables for the 2007 Population and Housing Census of Ethiopia/CEN-
SUS 2007 TABLES /SNNPR1. Available: www.csa.gov.et (accessed February 24, 2013). Central Sta-
tistical Agency of Ethiopia.
19. Gaym A, Pearson L, Soe KW (2012) Maternity waiting homes in Ethiopia—three decades experience.
Ethiopian Medical Journal 50: 209–219. PMID: 23409404
20. Yaya Y, Eide KT, Norheim OF, Lindtjørn B (2014) Maternal and Neonatal Mortality in South-West Ethio-
pia: Estimates and Socio-Economic Inequality. PLoS ONE 9: e96294. doi: 10.1371/journal.pone.
0096294 PMID: 24787694
21. World-Bank Crude Birth Rate in Ethiopia (2010) World Bank report 2012. Available: http://www.
tradingeconomics.com/ethiopia/birth-rate-crude-per-1-000-people-wb-data.html. Accessed 10 August
2013.
22. WHO International Statistical Classification of Diseases and Related Health Problems 10th Revision,
ICD-10: 2010 Version. Available: http://apps.who.int/classifications/icd10/browse/2010/en. Accessed
09 December 2011.

23. CSA-Ethiopia (2014) Ethiopia Mini Demographic and Health Survey 2014. Central Statistical Agency,
Addis Ababa, Ethiopia July 2014. Available: http://wwwuniceforg/ethiopia/Ethiopia_Mini_
Demographic_and_Health_Survey_2014pdf. Accessed 15 August 2014.
24. Major B, Gramzow RH (1999) Abortion as stigma: Cognitive and emotional implications of conceal-
ment. Journal of Personality and Social Psychology 77: 735–745. PMID: 10531670
25. Girma M, Yaya Y, Gebrehanna E, Berhane Y, Lindtjorn B (2013) Lifesaving emergency obstetric ser-
vices are inadequate in south-west Ethiopia: a formidable challenge to reducing maternal mortality in
Ethiopia. BMC Health Services Research 13: 459. doi: 10.1186/1472-6963-13-459 PMID: 24180672
26. Songane FF, Bergström S (2002) Quality of registration of maternal deaths in Mozambique: a commu-
nity-based study in rural and urban areas. Soc Sci Med 54: 23–31. PMID: 11820679
27. Yaya Y, Lindtjørn B (2012) High maternal mortality in rural south-west Ethiopia: estimate by using the
sisterhood method. BMC Pregnancy Childbirth 12: 136. doi: 10.1186/1471-2393-12-136 PMID:
23176124
28. CSA-Ethiopia, ICF-International (2012) Ethiopia Demographic and Health Survey 2011 Central Statisti-
cal Agency of Ethiopia and ICF International Addis Ababa, Ethiopia and Calverton, Maryland, USA.
Available: http://wwwmeasuredhscom/publications/publication-fr255-dhs-final-reportscfm Accessed 25
June 2013.
29. Mersha A, Jira C, Demessie S (1996) Community based study on maternal mortality in Jimma town,
south western Ethiopia. Indian J Public Health 40: 30–34. PMID: 9090901
30. Berhane Y, Andersson T, Wall S, Byass P, Högberg U (2000) Aims, options and outcomes in measur-
ing maternal mortality in developing societies. Acta Obstet Gynecol Scand 79: 968–972. PMID:
11081682
31. Shiferaw T, Tessema F (1993) Maternal mortality in rural communities of Illubabor, southwestern Ethio-
pia: as estimated by the"sisterhood method". Ethiop Med J 31: 239. PMID: 8287858
32. Kwast BE, Rochat RW, Kidane-Mariam W (1986) Maternal mortality in Addis Ababa, Ethiopia. Stud
Fam Plann 17: 288–301. PMID: 3798492
33. WHO, UNICEF, UNFPA, World-Bank (2012) Trends in maternal mortality:1990 to 2010. WHO, UNI-
CEF, UNFPA and The World Bank estimates. Available: http://whqlibdoc.who.int/publications/2012/
9789241503631_eng.pdf. Accessed 10 September 2012.
34. Alkire S, Santos ME (2011) Acute Multidimensional Poverty: A New Index for Developing Countries.
Proceedings of German Development Economics Conference (Berlin). Available: http://www.econstor.
eu/handle/10419/48297. Accessed 25 May 2013.
35. Menéndez C, Romagosa C, Ismail MR, Carrilho C, Saute F, et al. (2008) An autopsy study of maternal
mortality in Mozambique: the contribution of infectious diseases. PLoS Med 5: e44. doi: 10.1371/
journal.pmed.0050044 PMID: 18288887

II
Maternal and Neonatal Mortality in South-West Ethiopia:
Estimates and Socio-Economic Inequality
Yaliso Yaya1,3*, Kristiane Tislevoll Eide2, Ole Frithjof Norheim1,2, Bernt Lindtjørn1
1 Centre for International Health, Faculty of Medicine and Dentistry, University of Bergen, Bergen, Norway, 2 Department of Global Public Health and Primary Care, Faculty
of Medicine and Dentistry, University of Bergen, Bergen, Norway, 3 Arba Minch College of Health Sciences, Arba Minch, Ethiopia
Abstract
Introduction: Ethiopia has achieved the fourth Millennium Development Goal by reducing under 5 mortality. Nevertheless,
there are challenges in reducing maternal and neonatal mortality. The aim of this study was to estimate maternal and
neonatal mortality and the socio-economic inequalities of these mortalities in rural south-west Ethiopia.
Methods: We visited and enumerated all households but collected data from those that reported pregnancy and birth
outcomes in the last five years in 15 of the 30 rural kebeles in Bonke woreda, Gamo Gofa, south-west Ethiopia. The primary
outcomes were maternal and neonatal mortality and a secondary outcome was the rate of institutional delivery.
Results: We found 11,762 births in 6572 households; 11,536 live and 226 stillbirths. There were 49 maternal deaths; yielding
a maternal mortality ratio of 425 per 100,000 live births (95% CI:318–556). The poorest households had greater MMR
compared to richest (550 vs 239 per 100,000 live births). However, the socio-economic factors examined did not have
statistically significant association with maternal mortality. There were 308 neonatal deaths; resulting in a neonatal mortality
ratio of 27 per 1000 live births (95% CI: 24–30). Neonatal mortality was greater in households in the poorest quartile
compared to the richest; adjusted OR (AOR): 2.62 (95% CI: 1.65–4.15), headed by illiterates compared to better educated;
AOR: 3.54 (95% CI: 1.11–11.30), far from road ($6 km) compared to within 5 km; AOR: 2.40 (95% CI: 1.56–3.69), that had
three or more births in five years compared to two or less; AOR: 3.22 (95% CI: 2.45–4.22). Households with maternal
mortality had an increased risk of stillbirths; OR: 11.6 (95% CI: 6.00–22.7), and neonatal deaths; OR: 7.2 (95% CI: 3.6–14.3).
Institutional delivery was only 3.7%.
Conclusion: High mortality with socio-economic inequality and low institutional delivery highlight the importance of
strengthening obstetric interventions in rural south-west Ethiopia.
Citation: Yaya Y, Eide KT, Norheim OF, Lindtjørn B (2014) Maternal and Neonatal Mortality in South-West Ethiopia: Estimates and Socio-Economic Inequality. PLoS
ONE 9(4): e96294. doi:10.1371/journal.pone.0096294
Editor: Shannon M. Hawkins, Baylor College of Medicine, United States of America
Received October 9, 2013; Accepted April 4, 2014; Published April 30, 2014
Copyright: ß 2014 Yaya et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was funded by the Centre for International Health, University of Bergen, Norway. The funders had no role in study design, data collection
and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: [email protected]
Introduction during pregnancy, childbirth, and the post-natal period increase

the risk of death for both the baby and the mother [6].
In 2010, there were 287,000 maternal deaths in the world from Ethiopia has achieved MDG-4 by reducing two-thirds of under
causes of pregnancy related complications, which is 50% down 5 mortality per 1000 live births from 204 in 1990 to 68 in 2012 [7].
from the 1990 baseline [1]. Every year, four million newborns die Ethiopia’s achievement has been attributed to the country’s
during the neonatal period (28 days after birth) [2], while 3.2 community-based health promotion and disease prevention
million pregnancies end with stillbirths [3]. Among these deaths, programme through a health extension package [8]. However in
99% of the neonatal deaths and stillbirths as well as 98% of terms of neonatal mortality, maternal mortality, and stillbirth
maternal deaths, occur in low- and middle-income countries. reduction, challenges still remain. Ethiopia is among 10 countries
Moreover, 40% of global, and 29% of African child mortality is that contributed to two-thirds of global neonatal deaths in 2005 [2]
caused by neonatal deaths [4]. The Millennium Development and stillbirths in 2009 [9]. UN agencies (WHO and UNFPA) and
Goals (MDG-4 and 5) aim to reduce child mortality by two-thirds, World Bank data also indicate a decline in the maternal mortality
and maternal mortality by three-quarters, between 1990 and 2015. ratio (MMR) per 100,000 live births in Ethiopia from 950 in 1990
High maternal and neonatal deaths and stillbirths often occur to 350 in 2011 [1]. Nevertheless, there are often controversies over
because of inadequate care during pregnancy and childbirth. these estimates. For example, the 2011 national Demographic and
Accordingly, these deaths are considered sensitive indicators of the Health Survey (DHS) reported an MMR of 676 per 100,000 live
quality of a healthcare system in an area [5]. High neonatal deaths births, which is close to two-folds higher compared to the UN
and stillbirths are often related to maternal health; complications estimate [10]. According to the 2007 National Census, over 84%
of the estimated 90 million people in Ethiopia live in rural areas
PLOS ONE | www.plosone.org 1 April 2014 | Volume 9 | Issue 4 | e96294

Maternal and Neonatal Mortality in Gamo Gofa, Ethiopia
with limited access to quality health care[11]. In 2008, a Maternal mortality ratio (MMR). Is the number of
nationwide study showed that 7% of all deliveries took place at maternal deaths in a population during a given time period per
health institutions while only 3% in facilities that could provide 100,000 live births during the same period.
comprehensive essential obstetric care [12]. Neonatal mortality ratio (NMR). Is the number of
In places such as rural Ethiopia, where there is no birth and newborn deaths (within 0–28 days) in a population per 1,000 live
death registration and the majority of births and deaths take place births in the same population.
outside of health institutions, measuring maternal and neonatal Stillbirth rate. Is the number of births of dead fetuses after
mortality is difficult [13]. In some places, methods such as 28 weeks of gestation per 1,000 births.
demographic surveillance systems help to find and measure
maternal and neonatal mortalities [14]. Unfortunately, this Study area
alternative method does not exist in the Gamo Gofa province of We conducted this study in Bonke, one of the 15 woredas in the
Ethiopia. Maternal mortality can also be estimated through low- Gamo Gofa zone in south-west Ethiopia. In 2010, the woreda had
cost innovative options of the sisterhood method, which asks adult a population of 173,240 people [11]. A kebele is the lowest
siblings about their sisters’ death related to pregnancy or childbirth administrative structure with 5,000 to 7,000 residents in the
during reproductive age [15]. Nevertheless, results from the Ethiopian government system. Bonke has 31 kebeles and one of
sisterhood method refer to many years before the survey and may these, the administrative centre, has a town status with a
not show the current magnitude of the problem [16]. Findings population of 6,347 people in 2007. Table 1 shows the profile of
from well-planned household surveys that use large samples in the 15 kebeles included in this study. Bonke is 618 km from Addis
high fertility and high mortality areas can be useful in providing Ababa, and 68 km from Arba Minch (zonal capital) where the
real-time data to motivate actions [17]. The aim of this study was nearest hospital is situated. Nevertheless, over half of the remote
to estimate maternal and neonatal mortality, the stillbirth rate, the areas of Bonke are more than 100 km (20 hours walking distance)
institutional delivery rate, and household risk factors associated away from the hospital in Arba Minch. An estimated three-fourths
with these mortality outcomes in rural south-west Ethiopia. of the population also live in villages far from the motorable road
($6 km). The only road to the woreda is the road from Arba
Minch to Kamba, which crosses parts of Bonke. Overflowing
Methods and Materials
rivers during the rainy season often interrupt the road. So, often
Ethics statement people have to carry critical patients or use transport animals such
The Ethical Review Committee for the Health Research of as horses and mules to go to the hospital.
Southern Nations Nationalities and Peoples’ Regional State Health care is provided by a health centre in the town (Geresse)
(SNNPRS) Health Bureau in Ethiopia, and the Regional as well as three other rural health centres. There are no medical
Committee for Health Research Ethics of North Norway (REK doctors working in the woreda; a few health officers (people with a
Nord) approved the study. We obtained informed verbal consent bachelor’s degree in medical training), nurses, and midwives staff
from all respondents and the response was recorded on the the health centres. In Bonke, there is no access to lifesaving
questionnaire as ‘‘accepted’’ or ‘‘declined’’ to participate. Almost comprehensive essential obstetric care that can provide caesarean
sections, blood transfusions, and effective care to sick and low
all approached households were willing to be interviewed, and
birth-weight newborns. This study was part of an implementation
written consent was not considered because a large number of the
project to reduce maternal mortality in Gamo Gofa. The project
respondents were illiterates. The study involved only interview and
trains health officers in emergency obstetric services, community
the ethics committee approved the verbal consent procedure.
health workers in identifying and referring high-risk mothers, in
Additionally, minors were not included in this study.
addition to equipping health centres and hospitals with essential
instruments.
Definitions
Verbal autopsy for maternal deaths. A method for finding
Study design and period
out the medical causes of death and ascertaining the factors that
The study was a cross-sectional household survey with a five-
may have contributed to the death in women who died outside of a
year recall of events prior to the data collection. We collected the
medical facility. It consists of interviewing people (family members, data in February 2011 from all households that had births and
neighbours, traditional birth attendants) who had knowledge pregnancy outcomes between January 2006 and December 2010.
about the events leading to the death [18]. We purposely selected January 2006 as the starting reference
Neonatal mortality. A death within 28 days of an alive born period for the recall because it was the immediate period after the
baby. 2005 National Election, an event well known to all respondents.
Maternal mortality. A death of a woman while pregnant, in
labour, or within 42 days of the termination of pregnancy,
Sampling
irrespective of the duration and site of the pregnancy, from any
We based our sample size calculation on the assumptions of a
cause related to or aggravated by the pregnancy or its
crude national birth rate of 35 per 1,000 population, and a
management, but not from accidental or incidental causes (ICD-
neonatal mortality ratio of 35 per 1,000 live births [20]. We aimed
10) [19]. to detect a minimum of 350 neonatal deaths to make empirical
Stillbirth. A birth of a dead fetus after 28 weeks of gestation.
estimates and assess the household risk factors associated with
We did not use the baby-weight criteria of classifying stillbirths, as neonatal deaths. To find 350 neonatal deaths we needed to find at
it was not possible to measure weight in the rural area. least 10,000 live births in five years (an average of 2,000 per year)
Household. A person or a group of people living in a room or within the population. With a fertility rate of 35 births per 1,000
rooms and sharing common things together. In cases of polygamy people, 2,000 live births per year could be obtained from an
(more than one wife for a man, we considered each wife as a estimated population of 57,143. Assuming a constant birth rate
separate household as they culturally have separate houses. over the five years, we projected the population of 57,143 in 2006
to be 67,244 in 2010 (half the rural population in Bonke).

Table 1. Background information comparing study population with national census data.
Bonke 2010
(current study) Census 2007 [16] (adjusted to 2010)
Number of households in 15 Kebeles 11 920 12 681

Population of the 15 Kebeles 72 7121 78 181
Average persons per household 6.14 6.17
Crude birth rate (per 100 population) 3.22 3.60{
Percent of under-5 year population 15 16m
A
Percent of illiterate adults ($15 years) 58. 5 64.5B
1
= obtained by multiplying average persons per household in studied households (6.1) with the total households counted during the study (11 920).
{
= DHS 2011 for rural Ethiopia (no separate data for Bonke) [10].
m
= rural Bonke.
A
= Heads of interview households, Bonke, 2010.
B
= Adult men ($15 years of age) Gamo Gofa zone (rural).
We used OpenEpi, open source calculator (www.openepi.com), Respondents were asked about whether the newborn was alive
and calculated the minimum sample needed based on information or dead at the end of the 4th week after delivery. If the response
that three-fourths of the study households resided far from the was ‘‘dead’’, then we asked about the timing of death (in weeks) in
motorable road ($6 km), thereby expecting a neonatal mortality relation to the birth. However, we did not investigate the causes of
prevalence twice that among households far from the road. We deaths for neonatal deaths and stillbirths, assuming it would be
used a statistical power of 80%, a 95% confidence interval, and an difficult for rural respondents to answer it properly. In the
assumption that 4% of households far from motorable road could households that had deaths of women -in reproductive age (15–49
experience neonatal mortality to calculate the number of years), we used questions modified from the WHO manual for
households needed for the study. This provided 5,187 households verbal autopsy for maternal death to investigate the causes of
with expected births in the five years before the survey. On deaths [18].
average, we expected two births per household over five years The questions included: whether the mother was pregnant, in
yielding 10,374 births. The neonatal mortality rate from the the process of giving birth, or in postnatal period after birth, what
estimated number of households was also assumed to provide main medical condition or symptom was associated with her
enough power to detect other risk factors (wealth, education, non- death, what assistance she received, and from whom she received
spaced births). help. A nurse decided on pre-coded choices of the major causes of
We also assumed that a number of maternal deaths among the maternal deaths (bleeding, prolonged labour, fever and convul-
estimated 10,374 births would give an optimum MMR estimate, sions, including the option of ‘‘others’’) based on quick algorithmic
and a similar assumption was applied for stillbirths. Taking into analysis of information provided by the respondents. Sensitive
account a potential 10% of non-responders, we decided to study questions related to abortion deaths were placed at the end of the
50% of the rural population in the Bonke woreda and we interview to minimize the intentional hiding of information. We
randomly selected 15 of the 30 rural kebeles in Bonke. Data collected information on the estimated walking distance (in hours)
collectors visited all of the households in the selected 15 kebeles from each house to the nearest health centre, the nearest
asking about any pregnancy and birth outcomes (abortion, alive motorable road, and the nearest hospital. Based on the local
and stillbirths, neonatal and maternal deaths) in the households experience of one hour of walking time per 5 km for an average
over the previous five years. Enumerators noted the number of person, we converted the walking distance into kilometers.
households in each kebele, collecting data from the households
that had pregnancy and birth outcomes during the stated time Data collection
period. As projected from the Ethiopian 2007 census [11], the We recruited 15 natives from the respective study villages who
selected kebeles had a population of 78,181 people in 2010. had completed the 12th grade for data collection. The purpose of
selecting data collectors from their respective kebele of data
Variables and questions collection was to reduce the potential recall bias by the
The primary outcome variables were maternal and neonatal respondents. Data collectors are aware of many vital events in
mortality and stillbirth rates while a secondary outcome variable the villages they collected data by living and participating in social
was rate of skilled delivery service utilization. We also used the events such as birth celebrations, mourning rituals, and burials at
following household predictor variables: household wealth (assets the time of the deaths. Five diploma graduates who had a
index), educational level of heads of households, the number of thorough knowledge of the culture and language of the area
births in a household over five years, and household distance to a supervised the data collection.
motorable road. The data collectors were trained for two days on pre-testing field
Survey questions included: where each delivery took place interviews, translating the questions from ‘‘Amharic’’ (the official
(home, health post, health centre, and hospital) and who attended Ethiopian state language) to ‘‘Gamotho’’ (the language of the
the birth (family member, community health extension worker, ethnic ‘‘Gamo’’ community) and how to introduce the simplified
skilled health professional-doctor, midwife). We also asked about verbal autopsy questions. Depending whoever was present at
the place where the maternal death occurred (at home or within a home during the visit, the respondent was the father or mother for
health facility) as well as what had happened to the fetus (stillbirth, a recently deceased newborn. In cases of death of a married
neonatal death, or alive at the time of data collection). woman, we interviewed a husband while in the absence of a

husband, an adult relative or an adult child of the deceased was Health centres are medium level institutions with health officers
interviewed. For those who were unmarried, we asked parents or and nurses, and some have midwives. The overall institutional
siblings. If the respondents were not present at home during the delivery rate (health centre and hospital combined) was 3.7%
first visit, the data collectors re-visited the next day in the early (431/11762).
morning. Less than 1% of households were missed after two visits.
Health centre births
The wealth-index creation Births that occurred within a 10 km distance of health centres,
For the wealth index, we selected 10 variables of household were five-times more likely to use health centres for delivery,
assets with the highest standard deviation (.0.20), as recom- 4.75% (274/5766), compared to households $11 km away, 1.07%
mended by Seema Vyas and colleagues [21]. The types of asset (64/5996); OR: 4.62 (95% CI: 3.51–6.09) (Table 2). A 10 km
variables and their standard deviations are presented in Table S1 distance is the Ethiopian government plan to achieve health centre
in File S1. We transformed the categorical variables into physical access to the population.
dichotomous (0–1) indicators: 0 for indicators of poor wealth
and 1 for indicators of good wealth. We examined the Description of deaths
dichotomous variables by using the principal component analysis There were 49 maternal deaths, resulting in a maternal
(PCA) to produce a factor score for each household with mortality ratio (MMR) of 425 (95% CI: 318–556) per 100,000
households being assigned a rank according to the factor score. live births. Among the 49 maternal deaths, 6 (12%) occurred
Because of the low number of maternal deaths in the socio- during pregnancy, 18 (37%) during labour, and 25 (51%) after
economic classes for calculation, we divided households into four birth within six weeks (Figure 1). The primary causes of death
equal categories (quartiles), rather than the widely used five classes. were: fever 14 (29%), bleeding 13 (27%), prolonged labour 8
Each category was comprised of 25% of the households studied. (16%), convulsion 8 (16%), and others 6 (12%) (Table 3). Other
Table S2 in File S1 shows the mean score, standard deviations, causes included two abortions and one anemia while three deaths
communalities, and correlations of the variables to the first (main) were not classified. Regarding the places of maternal death, most
component. The total variance explained by the first component [88% (43/49)] of the maternal deaths occurred during home
was 20.58%, with an eigenvalue of 2.06. deliveries, and health facilities were able to identify only 12% (6/
49) of the maternal deaths found in this study (Table 3).
Data analysis We found 308 neonatal deaths, which yields a neonatal
We used two units of analysis (household and birth). By using mortality ratio (NMR) of 27 (95% CI: 24–30) per 1,000 live
births as the unit of analysis, we presented descriptive tabulations births. Out of the 308 neonatal deaths reported, 143 (46.4%) died
of outcomes in the form of rates and ratios. By using the household in the first week, 72 (23.4%) in the second week, 63 (20.5%) in the
as a unit of analysis, and applying logistic regression, we present third week and 30 (9.7%) in the fourth week (Figure 2). There were
household risk factors associated with the mortality outcomes 226 stillbirths out of 11,762 total births yielding a stillbirth rate
(Tables 2, 3, and 4). We used SPSS 16 (Statistical Package for (SBR) of 19 (95% CI: 17–22) per 1,000 births.
Social Sciences) for the data entry and analysis [22]. Data are
freely available from the corresponding author on request. Household risk for mortality outcomes
Maternal mortality. Table 4 shows the MMR differences
Results across different risk factors. The MMR was increased in
households in the poorest quartile compared to the richest (550
Demographic description vs 239 per 100,000 live births); OR: 2.29 (95% CI: 0.91–6.44).
Table 1 describes households and population of the study area. However, socio-economic factors examined (wealth, distance from
Data collectors enumerated all households (11, 920) in the selected road, education, and non-spaced births) did not have statistically
15 kebeles but collected data from 6,572 households that had significant association with maternal mortality because of the
pregnancy and birth outcomes in the last five years before the data relative rarity of maternal deaths in terms of absolute numbers.
collection. In the 6,572 households that had pregnancy and birth Neonatal mortality. Table 5 describes the household risk
outcomes, there were 40,357 persons, an average of 6.1 persons factors associated with neonatal mortality. Neonatal mortality was
per household. greater among the poorest quartile households compared to the
Of the 6,572 household heads, 3,842 (58.5%) were not able to richest; adjusted OR (AOR): 2.62 (95% CI: 1.65–4.15). However,
read and write (illiterate), 2,446 (37.2%) had an elementary the highest risk was in the wealthy class; AOR: 3.57 (95% CI:2.37–
education (grade 1–8), and 279 (4.2%) had completed 9th grade or 5.38). The poorest were in the second highest at risk groups. The
more. Regarding the occupations of the head of the households, likelihood of neonatal mortality was also increased among
6,289 (95.7%) engaged in farming, 204 (3.1%) in farming mixed households with illiterate heads compared to where the heads
with small trade, while 79 (1.2%) were salary employed ($ 25–50 had a higher education (9th grade or more); AOR: 3.54 (95% CI:
per month) (Table 2). 1.11–11.30), in households far from a motorable road ($6 km)
compared to those within 5 km of a road; AOR: 2.40 (95% CI: 1.
Description of births 56–3.69), and greater among households that had three or more
Table 2 presents the backgrounds of households with pregnancy births in five years compared to those that had two or less births;
and birth outcomes and description of births. We found 11,762 AOR: 3.22 (95% CI: 2.45–4.22).
births in 6572 households over 5 years (3.2% annual crude birth Stillbirths. Table 6 shows the factors associated with stillbirth
rate), of which 11,536 were live births. Of the 11,762 total births, which were greater in households in the poorest quartile compared
10,861 (92.3%) took place at home, 470 (4%) at health posts, 338 to the richest; AOR: 3.13 (95% CI: 1.66–5.87). However, similar
(2.9%) at health centres and 93 (0.8%) at hospitals. A health post is to neonatal mortality, the highest risk of stillbirth was in the
a two-room building which is the lowest ranked health facility and households in the rich category compared to the richest; AOR:
staffed by two community agents known as health extension 6.40 (95% CI: 3.69–11.11). Stillbirth was greater among
workers whom received one year of general health training. households far from motorable road ($6 km) compared to those

Table 2. Socio-demographic backgrounds of household-survey participants, Bonke, Gamo Gofa, south-west Ethiopia, 2011.
Variables Category number %
Visited households With pregnancy and birth outcomesA 6572 55

Without pregnancy and birth outcomes 5348 45
Total (visited HHs) 11920 100
Births Live births 11536 98
Still births 226 2
Total (all births) 11762 100
Households with events Stillbirths 192 3
Neonatal deaths 265 4
Maternal deaths 49 0.7
Without events 6066 92.3
Total (HHs) 6572 100
Education of heads of Illiterate (cannot read/write) 3842 58.5
households Primary (1–8th) 2446 37.2
Higher (9th +) 279 4.2
Missing 5 0.1
Occupation of heads of households Subsistence farming 6289 95.7
Farming and small trade 204 3.1
Salary employed 79 1.2
Distances of HHs from #5 km 1624 24.7
motorable road $6 km 4948 75.3
Health centre distance #10 kmB 3258 49.6
$11 km 3314 50.4
Place of delivery Home (assisted by family or relative) 10861 92.3
Health post (attended by HEWs)C 470 4.0
Skilled institutions (HC and Hospital) 431 3.7
Total (births) 11762 100
Health centre delivery vs HC deliveries among #10 km from HC 274/5766 4.75
HC distance HC deliveries among $11 km from HC 64/5996 1.07
Total HC births out of all births 338/11762 2.87
Note: AHouseholds that had the outcomes over five years before the survey (focus of this study), B10 km distance is Ethiopian government plan for physical access to
health centres, HC = Health Centre, HHs = households, HEWs = CHealth Extension Workers (non-skilled birth attendants).
at a distance of 5 km to the road; AOR: 3.40 (95% CI: 1.91–6.06), compared to households without maternal death; OR: 11.6 (95%
and greater in households that had three or more births in five CI: 6.0–22.7).
years compared to two or fewer births; adjusted OR: 4.55 (95%
CI: 3.35–6.16). Discussion
Clustering of mortality in similar households In this household study we found a maternal mortality ratio of
425 per 100,000 live births, a neonatal mortality ratio of 27 per
Table 7 presents the concentration of maternal and newborn
1,000 live births, and a stillbirth rate of 19 per 1,000 births. The
mortality in certain households. Of the 49 households that had
risk of neonatal mortality was associated with the wealth status of
maternal deaths, nearly half (46.9%) also experienced either a
the households, literacy status of the head, non-spaced births, and
stillbirth or neonatal death; 12 (24.5%) had stillbirths, and 11
the distance from motorable road of households. The risk of
(22.4%) had neonatal deaths. The likelihood of neonatal death in
stillbirth was also associated with the wealth, distance of the
households that had maternal deaths was seven times higher
household from motorable road, and non-spaced births. The
compared to households that had no maternal mortality; OR: 7.2
maternal mortality was also high among the poorest households
(95% CI: 3.6–14.3). The similar likelihood of having a stillbirth in
compared to the richest and households that had maternal
households that had maternal mortality was 11 times greater
mortality also experienced a clustering of neonatal mortality or

Table 3. Causes and places of maternal deaths, Bonke, Gamo Gofa, south-west Ethiopia, 2006–2010.
Variables number %
Causes of death Fever 14 29

Bleeding 13 27
Prolonged labour 8 16
Convulsion 8 16
Others* 6 12
Total (deaths) 49 100
Places of death Home (community) 43 88
Health institution 6 12
Total (deaths) 49 100
Note: *other causes include: two abortions, two anaemia, and two cause not reported.
stillbirths. Moreover, the institutional delivery rate was unaccept- and the national level estimates for Ethiopia. As a result, we do not
ably low. know whether our findings were under-reported or represent the
To the best of our knowledge, this study is the first to describe reality of the area. The maternal mortality ratio of 425 per
three mortality estimates and associated household-risk factors 100,000 live births was similar to the findings of community-based
using a large sample with high response rate in Gamo Gofa. In studies decades ago in other parts of Ethiopia: the MMR per
fact, there is a limited amount of evidence of maternal and 100,000 live births was 402 in Jimma in 1990 [23], and 440 in
neonatal mortality and stillbirths using community-based data Butajira in 1996 [24]. However, our estimate is higher than the
from southern Ethiopia [14]. We used data collectors who had UN and World Bank’s estimate for Ethiopia of 350 per 100,000
experience and were sensitive to the cultural taboos of the live births in 2010 [1]. If we adjust our maternal mortality estimate
respective villages where they collected the data. Being the upward by a factor of 1.6, it yields an MMR of 680 per 100,000
residents in their villages, they participated in all social events, live births. The 2011 DHS reported MMR of 676 per 100,000 live
including involvements in vital events such as celebrating births, births for Ethiopia [25] which is similar to our upwardly adjusted
caring for the sick, and funerals for the dead. The experience and estimate. The suggestion as well as the factor of adjustment were
deep knowledge of the area enabled them with the skill to handle conducted according to the recommendation by Santon and
sensitive questions and recall many of the deaths that occurred in colleagues for the correction of the potential under-reporting of
their villages. We also used an experienced nurse as a field demographic studies [26]. The under-reporting of maternal deaths
technical supervisor to help in classifying deaths by using the is a well-recognized global problem [27–29], and our study may
verbal autopsy method. not have escaped the challenge. However, given a general
Because of lack of previous reports from the area, we compare downward trend of Ethiopia’s MMR estimate by the UN inter-
our findings with community-based studies from other provinces
Table 4. Maternal mortality across socio-economic factors, south-west Ethiopia, 2006–2010.
Variable Category Maternal deaths Live births MMRA OR (95% CI)

B
Wealth Richest 25% 6 2506 239 Ref
Rich 25% 11 3146 350 1.46 (0.54, 4.28)
Poor 25% 14 2953 474 1.98 (0.77, 5.59)
poorest 25% 15 2725 550 2.29 (0.91, 6.44)
EducationC Higher (9th +) 0D 468 --- Ref
Primary (1–8th) 18 4492 401 2.85 (0.42, 77.9)
Illiterate 31 6568 472 3.35 (0.52, 89.9)
Distance to road #5 km 9 2754 327 Ref
$6 km 40 8782 455 1.40 (0.67, 2.88)
No. of births (in 5 yr) #2 39 8620 452 Ref
$3 10 2916 343 0.76 (0.38, 1.52)
Place of deaths Home (comm.) 43 11160 385 Ref
Health institution 6 376 1596 4.14 (1.75, 9.79)
Note: Amaternal mortality ratio per 100,000 live births.

B
119 households, including 3 with maternal deaths, have missing value on the wealth index due to incomplete asset variables.
C
Education of head of household. Dzero was replaced by 0.5 during analysis to make calculation defined.

Figure 1. Time of maternal deaths, Bonke woreda, south-west Figure 2. Time in weeks of neonatal deaths, Bonke woreda,
Ethiopia, 2010. south-west Ethiopia, 2010.
doi:10.1371/journal.pone.0096294.g001 doi:10.1371/journal.pone.0096294.g002
agency [1], and Hogan et al [30], our direct (unadjusted) estimate land, cattle, and crops may not be owned by young people who
of an MMR of 425 per 100,000 may be close to reality. returned to rural residence after certain years of education in
Our neonatal mortality estimate of 27 per 1,000 live births is urban areas. In addition, educated residents such as government
lower than the DHS national estimate of 37 [31], and the UN employees who usually do not have these rural assets may have
Inter-agency estimate of 31 for Ethiopia in 2011 [20]. However, been wrongly classified as poor. Our finding of the association of
since mathematically modeled UN estimations often have education with neonatal mortality could also support the idea that
inconsistencies with survey findings, and we have limited locally the educated, but classified as poor and poorest in terms of rural
available estimates, we cannot judge whether our finding is due to wealth, may have been relatively better-off compared to those in
under-reporting. Regarding stillbirths, Cousens et al estimated the rich category because of the knowledge advantage.
Ethiopia’s stillbirth rate as 26 per 1,000 births in 2009 [9], which is On the other hand, households in the richest wealth category
still higher than our finding of 19 per 1,000 births. In the ethnic had the lowest risk of mortality possibly because of their economic
‘‘Gamo’’ culture, a birth of a dead fetus and the early death of a access to health services and expected better living conditions. In
newborn are not publicized and not publicly mourned. Only close general, the findings that the richest households had advantages
relatives and family members are informed and other people are while households in the three other categories experience greater
told the incident as ‘‘something went wrong’’ in local terms. This risk could be due to the situation of wealth where few households
can make it difficult for an outsider to distinguish between a have greater possession of assets and the other majority is
neonatal death and a stillbirth. Accordingly, though we tried to homogeneously poor. The finding agrees with a previous analysis
minimize under-reporting by carefully choosing the data collec- of Ethiopian rural asset data that classified the majority (up to
tors, some of the neonatal deaths and stillbirths may have been 60%) of households as having low SES, thereby suggesting a
missed in this study. Global estimates suggest that 75% of neonatal homogeneity of most households in asset ownership [21].
deaths take place in the first week of life [4]. In our finding, first- Economists measure economic status indicators through infor-
week deaths were approximately half of the neonatal deaths mation from income or expenditure, which is difficult to gather in
(46.4%) which may also suggest an under-reporting of early low-income countries, and asset-based wealth is an alternative
neonatal deaths and stillbirths. proxy in less developed areas [32]. Practically speaking, the wealth
The present study showed that indicators of the socio-economic index is equally valid to income or expenditure data for health
status (SES) of households (wealth, education of head of surveys in Africa [33], and the effect of household wealth on health
household, the distance of households from a motorable road), outcome is well known [34–36]. Less clear, however, is how wealth
and a factor related to reproductive health (non-spaced births) causes mortality difference in areas where the overall access to
were associated with stillbirths and neonatal mortality outcomes. health services and service utilization is very low to all people such
We examined the effect of household wealth (poor-rich differences) as those in our study area. For instance, in our results, there is the
on mortality outcomes and found a significant variation. The poorest-richest difference in maternal mortality and stillbirths.
poorest households were more likely to have mortality outcomes Even so, very few households (including the richest) utilize health
compared to the richest. Nevertheless, when compared to the institutions for delivery service. This indicates that household
richest quartile, the greatest risk of both neonatal mortality and wealth contributes to maternal mortality in mechanisms other
stillbirth was to the rich households, not to the poorest. than those that create economic access to health service. These
Households in the poorest category were the second most at-risk mechanisms may include, e.g., improved, clean housing and better
group for the mortality outcomes. This finding needs further nutrition [37]. Unfortunately, this is beyond the aim and
investigation with a focus on how the true wealth status can be limitations of our study.
determined at the particular stage of socio-economic development Regarding the association of wealth with neonatal mortality, the
in the rural Ethiopia [21]. Our opinion is that the asset variables poorest-richest difference may be due to economic access to
selected to indicate wealth status may not have correctly reflected antibiotics and other drugs from rural private venders, where a
the actual wealth status of all households. For example, farming wealthy family often has a better access in addition to better

Table 5. Factors associated with neonatal mortality, south-west Ethiopia, 2006–2010.
Neonatal deaths in the

Predictors Category household?A (n = 6572 HHs) Crude AdjustedE
Yes No OR 95% CI OR 95% CI
WealthB Richest 25% 31 1322 Ref Ref

Rich 25% 124 1685 3.14 2.12, 4.74 3.57 2.37, 5.38
PLOS ONE | www.plosone.org

Poor 25% 46 1662 1.18 0.74, 1.89 1.92 1.19, 3.10
Poorest 25% 57 1526 1.59 1.02, 2.48 2.62 1.65, 4.15
EducationC Higher (9th +) 3 276 Ref Ref
Primary (1–8th) 79 2367 3.07 1.08, 12.37 2.86 0.89, 9.18
IlliterateD 182 3660 4.57 1.64, 18.24 3.54 1.11, 11.30
Distance to road #5 km 28 1596 Ref Ref
$6 km 237 4711 2.87 1.93, 4.26 2.40 1.56,3.69
No. of birthsF #2 170 5388 Ref Ref
$3 95 919 3.28 2.53, 4.25 3.22 2.45, 4.22
A
HH = households (because there were more than one events in some households, the number of households having neonatal deaths are different from the number of neonatal deaths; 308 neonatal deaths in 265 households).
B
119 households, including 7 with neonatal deaths, have missing value on the wealth index due to incomplete asset variables.
C
Education of head of household. DIlliterate = cannot read and write Eadjusted to the other variables in the table. Fnumber of births in five years.
8
Note: Hosmer-Lemeshow Test of Model fit: X2 (df) = 9.14 (7); p = 0.24. A p-value greater than 0.05 shows that the model well fit the data.
April 2014 | Volume 9 | Issue 4 | e96294

Table 6. Factors associated with stillbirths, south-west Ethiopia 2006–2010.
Predictors groups Stillbirths in household? (n = 6572 HHA) Crude AdjustedE
Yes No OR 95% CI OR 95% CI
WealthB Richest 25% 16 1337 Ref Ref

Rich 25% 106 1703 5.20 3.12, 9.11 6.40 3.69, 11.11
Poor 25% 35 1673 1.75 0.97, 3.25 3.28 1.76, 6.11
PLOS ONE | www.plosone.org

Poorest 25% 32 1551 1.72 0.95, 3.23 3.13 1.66, 5.87
EducationC Higher (9th+) 5 274 Ref Ref
Primary (1–8) 52 2394 1.19 0.50, 3.40 1.03 0.40, 2.65
IlliterateD 134 3708 1.98 0.87, 5.52 1.25 0.50, 3.16
Distance to road #5 km 14 1610 Ref Ref
$6 km 178 4770 4.29 2.48, 7.42 3.40, 1.91, 6.06
No. of birthsF #2 107 5451 Ref Ref
$3 85 929 4.66 3.48, 6.25 4.55 3.35, 6.16
A
HH = households (because there were more than one such events in some households, the number of households having stillbirths are different from the number of stillbirths; 226 stillbirths in 192 households).
B
119 households, including 3 with stillbirths, have missing value on the wealth index due to incomplete asset variables.
C
Education of head of household DIlliterate = cannot read and write Eadjusted to the other variables in the table Fnumber of births in five years.
Note: Hosmer-Lemeshow Test of Model fit: X2 (df) = 7.98 (7); p = 0.33. A p-value greater than 0.05 shows that the model well fit the data.
9
April 2014 | Volume 9 | Issue 4 | e96294

Table 7. Clustering of maternal and neonatal mortality and stillbirth in households, south-west Ethiopia 2006–2010.
Maternal death in the household?X OR (95% CI)

Yes (n = 49) No (n = 6523)
Stillbirths in household? yes 12 (24.5%) 180 (2.8%) 11.6 (6.0, 22.7)

No 37 (75.5%) 6343 (97.2%) Ref
Neonatal deaths in household? yes 11(22.5%) 254 (3.9%) 7.2 (3.6, 14.3)
No 38 (77.5%) 6269 (96.1%) Ref
Note: XMaternal mortality in a household was considered an exposure variable for stillbirth and neonatal death outcomes.
nutrition and improved housing since access to antibiotics plays an (abortions, stillbirths, livebirths, neonatal and maternal deaths)
important role in newborn survival [38]. In addition, the may be the ideal option. In addition to providing sustainable data
association of neonatal mortality with the education of the heads for evaluation of effectiveness of policy and programmes, ongoing
of a households and the distance to a motorable road further registration can provide evidence for rights- based advocacy for
suggests the importance of these variables as tools to access health improvement of health services, and many other benefits outside of
interventions. The education of parents has a positive correlation the health sector [46]. For prospective registration to occur,
with better health of children through better knowledge of Ethiopia must utilize the privilege of the available two health
solutions and critical decisions during crisis, in addition to the extension workers (HEWs) responsible for each kebele (average
opportunity education creates for job and economic access [39– 500 households per HEW). Registration-based information may
42]. However, in the rural area where this study took place, few have benefits of reduced risk of recall bias and the cost of surveys
people had achieved better jobs despite higher levels of education. by using the already available community health workers for active
The knowledge advantage may have played a role related to access data collection.
and utilization of treatments in households with educated heads. In the following, we address some of the limitations of the study.
Our study also demonstrated an association of household First, recall bias and under-reporting are widely recognized
distance from a motorable road with neonatal death. Travel problems in studies that ask respondents about past events. The
distance is clearly an important factor once the decision is made to intensity of the bias depends on the time interval between the
seek medical care during critical conditions and distance to a event and the sensitivity of the event to memory [3]. We tried to
motorable road has dual effects: 1) as a disincentive to seek health reduce recall bias in two ways: 1) by selecting data collectors from
care, and 2) as a barrier to reach the relevant facility [43]. As such, the respective villages of the data collection that the data collectors
people in households closer to a motorable-road are more likely to helped the respondents to recall the events through their in depth
seek health care and save lives. We also found an increased knowledge of social events that happened in their villages, and 2)
likelihood of neonatal mortality and stillbirths among households by choosing a memorable and short time reference period for the
where there was a maternal death. This is in agreement with a event to be recalled. However, there might have been some deaths
finding in a WHO multi-country maternal and newborn health
that were missed due to recall bias in the current data. Second, as
survey, which showed a seven fold greater early neonatal
it is well-known in survey studies, we cannot show the temporality
mortality, in which mothers had died or developed nearmis
(time sequence) of the occurrence of exposures and outcomes [47].
(dangerous illness) compared to mothers without these events [44].
For example, having more than three births in the last five years in
The clustering of maternal and neonatal mortality, as well as
households was associated with both neonatal mortality and
stillbirths in certain households, illustrates how impoverished
stillbirth. Nevertheless, we cannot assure whether neonatal
households are trapped in many adverse outcomes.
mortality and stillbirths led to more births or whether more births
The findings of a low utilization of facilities for skilled delivery,
led to a greater risk.
compounded with high mortality rates, call for educating mothers
and other family members the importance of seeking skilled Third, we used reported information from family members on
delivery service (institutional deliveries). Furthermore, improving mortality outcomes and mid-level expert decisions of the cause-of-
the quality of the existing poorly equipped health institutions in the death classifications using a simplified verbal autopsy technique,
area [45] might help to increase the willingness and trust of which also has the potential for misclassification (misdiagnosis).
families to utilize these institutions. The Ethiopian health The existence of misclassification in using the verbal autopsy was
extension workers have the opportunity to educate women and reported from a well-designed prospective study of maternal
family, as they have close contact during the antenatal care and deaths in Guinea-Bissau in Africa as 30% of maternal deaths were
the routine household visits. In summary, our findings highlighted left unclassified [48]. The confirmatory diagnostic method used to
that households living in the villages far from road access, poor ascertain the cause of maternal death is an autopsy test which has
SES, and having illiterate household heads and non-spaced births, been used in a hospital setting in Mozambique in Africa [49].
had greater risks of mortality. Therefore, a targeted follow-up of However, such a modern technology cannot be applied in a rural
pregnancies in these households could help to achieve reduced community such as Bonke. Fourth, we were not able to show
mortality outcomes. Interventions such as family planning yearly changes in mortality. We aimed to describe at least
education and the availability of FP technology choices to support aggregated measures by using a single reference period in a
women in poor households may help prevent deaths caused due to community where the date of the event is not easy to identify.
the risk of non-spaced births. From a practical point, it is difficult and inappropriate to expect a
In order to obtain ongoing data to help monitor progress, the specific time-related response from a largely illiterate rural society
community-based registration of pregnancy and birth outcomes where there is no vital registration system.

Conclusions Analysis (PCA) for wealth index creation. Table S2. Background
descriptions of the variables included in the PCA analysis.
Mortality rates are still lagging behind the MDG targets for (DOCX)
Ethiopia. There also exist socio-economic inequalities in maternal
and neonatal mortalities, as well as stillbirths in the area. The
socio-economic inequality in mortality and the low utilization of
Acknowledgments
existing institutions for delivery care highlight the importance of We would like to thank the study participants for providing the information
quality emergency obstetric care service. The services need to and committing their time for the interviews. Thanks also goes to the
target the poorest households where mortalities cluster and Southern Nations Nationalities and People’s Regional State (SNNPRS)
disproportionately high. It is important to address barriers to Health Bureau, the Gamo Gofa zone Health Department, and the Bonke
Woreda Health Office in Ethiopia for cooperating during the data
accessing institutional delivery services in a way that is acceptable
collection.
for rural women.
Author Contributions
Supporting Information
Conceived and designed the experiments: YY OFN BL. Performed the
File S1 Supporting information file (two tables inform- experiments: YY OFN BL. Analyzed the data: YY KTE OFN BL. Wrote
ing variables included in Principal Component Analysis the paper: YY KTE OFN BL. Reviewed and approved the final version of
(PCA). Table S1. Variables included in Principal Component the manuscript: YY KTE OFN BL.
References
1. WHO, UNICEF, UNFPA, World-Bank Trends in maternal mortality:1990 to York, USA. Available: http://www.who.int/maternal_child_adolescent/
2010. WHO, UNICEF, UNFPA and The World Bank estimates. Available: documents/levels_trends_child_mortality_2012.pdf. Accessed 20 June 2013.
http://whqlibdoc.who.int/publications/2012/9789241503631_eng.pdf. Ac- 21. Vyas S, Kumaranayake L (2006) Constructing socio-economic status indices:
cessed 10 September 2012. how to use principal components analysis. Health Policy and Planning 21: 459–
2. Lawn JE, Cousens S, Zupan J (2005) 4 million neonatal deaths: When? Where? 468.
Why? The Lancet 365: 891–900. 22. IBM SPSS software for data entry and analysis. SPSS version-16. Available:
3. Lawn J, Yakoob M, Haws R, Soomro T, Darmstadt G, et al. (2009) 3.2 million http://www-01.ibm.com/software/analytics/spss/.
stillbirths: epidemiology and overview of the evidence review. BMC Pregnancy 23. Mersha A, Jira C, Demessie S (1996) Community based study on material
and Childbirth 9: S2. mortality in Jimma town, south western Ethiopia. Indian J Public Health 40: 30–
4. Black RE, Cousens S, Johnson HL, Lawn JE, Rudan I, et al. (2010) Global, 34.
regional, and national causes of child mortality in 2008: a systematic analysis. 24. Berhane Y, Andersson T, Wall S, Byass P, Högberg U (2000) Aims, options and
The Lancet 375: 1969–1987. outcomes in measuring maternal mortality in developing societies. Acta
5. Lawn JE, Blencowe H, Pattinson R, Cousens S, Kumar R, et al. (2011) Obstetricia et Gynecologica Scandinavica 79: 968–972.
Stillbirths: Where? When? Why? How to make the data count? The Lancet 377: 25. WHO International Statistical Classification of Diseases and Related Health
1448–1463. Problems 10th Revision, ICD-10: 2010 Version. available at : http://apps.who.
6. Ngoc NTN, Merialdi M, Abdel-Aleem H, Carroli G, Purwar M, et al. (2006) int/classifications/icd10/browse/2010/en (accessed 09 December 2011).
Causes of stillbirths and early neonatal deaths: data from 7993 pregnancies in six 26. Stanton C, Hobcraft J, Hill K, Kodjogbé N, Mapeta W, et al. (2001) Every death
developing countries. Bulletin of the World Health Organization 84: 699–705. counts: measurement of maternal mortality via a census. Bulletin of the World
7. UNICEF (2013) Committing to Child Survival: A Promise Renewed – Progress Health Organization 79: 657–664.
Report 2013. Available: http://www.unicef.org/publications/files/APR_ 27. Kao S, Chen LM, Shi L, Weinrich MC (1997) Underreporting and
Progress_Report_2013_9_Sept_2013.pdf. Accessed 03 October 2013]. misclassification of maternal mortality in Taiwan. Acta Obstetricia et
8. UNICEF (2013) In Ethiopia, a far-reaching health worker programme has Gynecologica Scandinavica 76: 629–636.
helped reduce child mortality across the country. Available: http://www.unicef. 28. Karimian-Teherani D, Haidinger G, Waldhoer T, Beck A, Vutuc C (2002)
org/infobycountry/ethiopia_70372.html. Accessed 03 October 2013]. Under-reporting of direct and indirect obstetrical deaths in Austria, 1980–98.
9. Cousens S, Blencowe H, Stanton C, Chou D, Ahmed S, et al. (2011) National, Acta Obstetricia et Gynecologica Scandinavica 81: 323–327.
regional, and worldwide estimates of stillbirth rates in 2009 with trends since 29. Deneux-Tharaux C, Berg C, Bouvier-Colle M-H, Gissler M, Harper M, et al.
1995: a systematic analysis. The Lancet 377: 1319–1330. (2005) Underreporting of pregnancy-related mortality in the United States and
10. CSA-Ethiopia, ICF-International (2012) Ethiopia Demographic and Health Europe. Obstetrics & Gynecology 106: 684–692.
Survey 2011 final report. Central Statistical Agency of Ethiopia and ICF- 30. Hogan MC, Foreman KJ, Naghavi M, Ahn SY, Wang M, et al. (2010) Maternal
International Addis Ababa, Ethiopia and Calverton, Maryland, USA. Available: mortality for 181 countries, 1980–2008: a systematic analysis of progress towards
http://wwwmeasuredhscom/publications/publication-fr255-dhs-final- Millennium Development Goal 5. The Lancet 375: 1609–1623.
reportscfm Accessed 25 June 2013. 31. CSA-Ethiopia, ICF-International (2012) Ethiopian Demographic and Health
11. CSA-Ethiopia (2010) Statistical Tables for the 2007 Population and Housing Survey 2011 final report. Available: http://www.measuredhs.com/
Census of Ethiopia/CENSUS 2007 TABLES/SNNPR1. Available: www.csa. publications/publication-fr255-dhs-final-reports.cfm. Accessed 25 June 2013.
gov.et. Accessed February 24, 2012. Central Statistical Agency of Ethiopia. Central Statistical Agency of Ethiopia and ICF-International Addis Ababa,
12. Admasu K, Haile-Mariam A, Bailey P (2011) Indicators for availability, Ethiopia and Calverton, Maryland, USA.
utilization, and quality of emergency obstetric care in Ethiopia, 2008. 32. Montgomery MR, Gragnolati M, Burke KA, Paredes E (2000) Measuring living
International Journal of Gynecology & Obstetrics 115: 101–105. standards with proxy variables. Demography 37: 155–174.
13. Graham WJ, Ahmed S, Stanton C, Abou-Zahr CL, Campbell OMR (2008) 33. Morris SS, Carletto C, Hoddinott J, Christiaensen LJM (2000) Validity of rapid
Measuring maternal mortality: an overview of opportunities and options for estimates of household wealth and income for health surveys in rural Africa.
developing countries. BMC medicine 6: 12. Journal of Epidemiology and Community Health 54: 381–387.
14. Docherty M, Smith R (1999) The case for structuring the discussion of scientific 34. Pollack CE, Chideya S, Cubbin C, Williams B, Dekker M, et al. (2007) Should
papers. British Medical Journal 318: 1224–1225. Health Studies Measure Wealth? A Systematic Review. American Journal of
15. Graham W, Brass W, Snow RW (1989) Estimating maternal mortality: the Preventive Medicine 33: 250–264.
sisterhood method. Studies in family planning 20: 125–135. 35. Duncan GJ, Daly MC, McDonough P, Williams DR (2002) Optimal Indicators
16. Yaya Y, Lindtjørn B (2012) High maternal mortality in rural south-west of Socioeconomic Status for Health Research. American Journal of Public
Ethiopia: estimate by using the sisterhood method. BMC Pregnancy and Health 92: 1151–1157.
Childbirth 12: 136. 36. Menchik PL (1993) Economic Status as a Determinant of Mortality Among
17. Kwast BE, Rochat RW, Kidane-Mariam W (1986) Maternal mortality in Addis Black and White Older Men: Does Poverty Kill? Population Studies 47: 427–
Ababa, Ethiopia. Studies in family planning: 288–301. 436.
18. WHO (2004) Beyond the numbers: reviewing maternal deaths and complica- 37. Christian P (2002) Maternal nutrition, health, and survival. Nutrition reviews 60:
tions to make pregnancy safer.World Health Organization 2004. Available: S59–S63.
http://www.who.int/maternal_child_adolescent/documents/9241591838/en/. 38. Dawodu AH, Effiong CE (1985) Neonatal Mortality: Effects of Selective
Accessed 20 June, 2013. Pediatric Interventions. Pediatrics 75: 51–57.
19. Ronsmans C, Graham WJ (2006) Maternal mortality: who, when, where, and 39. Alkire S, Santos ME Acute Multidimensional Poverty: A New Index for
why. The Lancet 368: 1189–1200. Developing Countries. Proceedings of German Development Economics
20. WHO (2012) Levels & Trends in Child Mortality Report 2012. Estimated Conference, 2011 (Berlin). Available: http://www.econstor.eu/handle/10419/
Developed by the UN Inter-agency Group for Child Mortality Estimation. New 48297. Accessed 25 May 2013.

40. Cleland JG, van Ginneken JK (1988) Maternal education and child survival in 45. Girma M, Yaya Y, Gebrehanna E, Berhane Y, Lindtjørn B (2013) Lifesaving
developing countries: The search for pathways of influence. Social science & emergency obstetric services are inadequate in south-west Ethiopia: a formidable
medicine 27: 1357–1368. challenge to reducing maternal mortality in Ethiopia. BMC health services
41. Asefa M, Drewett R, Tessema F (2006) A birth cohort study in South-West research 13: 459.
Ethiopia to identify factors associated with infant mortality that are amenable for 46. Okonjo-Iweala N, Osafo-Kwaako P (2007) Improving health statistics in Africa.
intervention. Ethiopian Journal of Health Development 14: 161–168. The Lancet 370: 1527–1528.
42. Cutler DM, Lleras-Muney A (2006) Education and health: evaluating theories 47. Grimes DA, Schulz KF (2002) Bias and causal associations in observational
and evidence the National Bureau of Economic Research. NBER Working research. The Lancet 359: 248–252.
Paper No.12352. 48. Høj L, Stensballe J, Aaby P (1999) Maternal mortality in Guinea-Bissau: the use
43. Gabrysch S, Campbell OMR (2009) Still too far to walk: Literature review of the of verbal autopsy in a multi-ethnic population. International Journal of
determinants of delivery service use. BMC Pregnancy and Childbirth 9: 34. Epidemiology 28: 70–76.
44. Souza JP, Gülmezoglu AM, Vogel J, Carroli G, Lumbiganon P, et al. (2013) 49. Menéndez C, Romagosa C, Ismail MR, Carrilho C, Saute F, et al. (2008) An
Moving beyond essential interventions for reduction of maternal mortality (the autopsy study of maternal mortality in Mozambique: the contribution of
WHO Multicountry Survey on Maternal and Newborn Health): a cross-
infectious diseases. PLoS Medicine 5: e44.
sectional study. The Lancet 381: 1747–1755.

III
Yaya and Lindtjørn BMC Pregnancy and Childbirth 2012, 12:136
http://www.biomedcentral.com/1471-2393/12/136
RESEARCH ARTICLE Open Access
High maternal mortality in rural south-west

Ethiopia: estimate by using the sisterhood
method
Yaliso Yaya1,2* and Bernt Lindtjørn1
Abstract
Background: Estimation of maternal mortality is difficult in developing countries without complete vital
registration. The indirect sisterhood method represents an alternative in places where there is high fertility and
mortality rates. The objective of the current study was to estimate maternal mortality indices using the sisterhood
method in a rural district in south-west Ethiopia.
Method: We interviewed 8,870 adults, 15–49 years age, in 15 randomly selected rural villages of Bonke in Gamo
Gofa. By constructing a retrospective cohort of women of reproductive age, we obtained sister units of risk
exposure to maternal mortality, and calculated the lifetime risk of maternal mortality. Based on the total fertility for
the rural Ethiopian population, the maternal mortality ratio was approximated.
Results: We analyzed 8503 of 8870 (96%) respondents (5262 [62%] men and 3241 ([38%] women). The 8503
respondents reported 22,473 sisters (average = 2.6 sisters for each respondent) who survived to reproductive
age. Of the 2552 (11.4%) sisters who had died, 819 (32%) occurred during pregnancy and childbirth. This
provided a lifetime risk of 10.2% from pregnancy and childbirth with a corresponding maternal mortality ratio of
1667 (95% CI: 1564–1769) per 100,000 live births. The time period for this estimate was in 1998. Separate
analysis for male and female respondents provided similar estimates.
Conclusion: The impoverished rural area of Gamo Gofa had very high maternal mortality in 1998. This
highlights the need for strengthening emergency obstetric care for the Bonke population and similar rural
populations in Ethiopia.
Keywords: High maternal mortality, Maternal mortality, Sisterhood method, Bonke, Gamo Gofa, Southwest
Ethiopia, Ethiopia, Sub-Saharan Africa
Background [5]. The indicator chosen to measure the progress is the

Maternal mortality is defined as the death of a woman maternal mortality ratio (MMR; number of maternal
during pregnancy or within 42 days after termination of deaths per 100,000 live births). Unfortunately, the pro-
pregnancy from any cause related to or aggravated by gress in many sub-Saharan African countries has been
the pregnancy or management of the pregnancy [1]. Ma- slow or non-existent [6].
ternal mortality is particularly high in developing Ethiopia is one of the six countries where > 50% of the
countries [2], where 98% of the yearly 500,000 maternal total maternal deaths worldwide occur; the other coun-
deaths occur [3,4]. Of the 20 countries with the highest tries are the Democratic Republic of Congo, Nigeria,
maternal mortalities in the world, 17 are in Africa. The India, Pakistan, and Afghanistan [6]. Since 1990, Ethiopia
Millennium Development Goals aim to reduce maternal has reduced child mortality [7]. There are also reports of
deaths by 75% by 2015 from the 1990 baseline (MDG-5) reductions in MMR, but not statistically significant. The
MMR for Ethiopia was 1061 (665–1639) in 1980, 968
* Correspondence: yalisoyaya@ gmail.com
(600–1507) in 1990, 937 (554–1537) in 2000, and 590
1
Centre for International Health, University of Bergen, Bergen, Norway (358–932) in 2008 [6]; however, these estimates have
2
Arba Minch College of Health Science, Arba Minch, Ethiopia
© 2012 Yaya and Lindtjørn; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the
Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.
Yaya and Lindtjørn BMC Pregnancy and Childbirth 2012, 12:136 Page 2 of 7
wide and overlapping confidence intervals that highlight zone and had a population of 173,240 in 2010 [16]. The
the difficulty in detecting real changes. It was as recent woreda consists of 31 kebeles; 1 of these kebeles is a
as 2008 that the upper uncertainty limit of the MMR town. Geresse, the administrative centre of Bonke, is 618
decreased to < 1000. Also, there are discrepancies be- km from Addis Ababa and 68 km from the zonal town,
tween estimates from different sources and methods. For Arba Minch. However, greater than two-thirds of the
example, the MMR was 590 (358–932) for 2008 accord- people in Bonke live in highlands, which are far from
ing to the Institute for Health Metrics and Evaluation roads. The only road to the woreda is the road from
by Hogan et al. [6], while the UN agencies and The Arba Minch to Kamba. The road is often interrupted be-
World Bank estimated the MMR to be 470 (270–790) cause of overflowing rivers during the rainy season and
[8]. Estimates of the MMR from community-based studies most of the population lives in remote villages far from
also vary; specifically, for 1982/83 the MMR was 566 in the road.
Addis Ababa [9], 570 (420–720) for Illubabor in western The district is divided into the cold and mountainous
Ethiopia in 1991 [10], and between 440 (314–598) and highlands, and hot lowlands with malaria endemic to the
665 (558–785) by surveillance and sisterhood method re- lowland area. Healthcare is provided by a health centre
spectively for Butajira in south central Ethiopia in 1996 at the town, and three other rural health centres. There
[11]. These surveys showed lower estimates than the are no medical doctors working in the district, and the
mathematically-modelled estimates for the country. health institutions are staffed by a few health officers
African countries, unlike developed nations, lack reli- and nurses. In the woreda, there is no access to compre-
able vital registrations to provide good MMR estimates. hensive emergency obstetric care providing caesarean
Developed countries use birth registries and link such deliveries and blood transfusions. There are villages that
registries to causes of death registries, which are the gold are as far as a 14-h walk (approximately 72 km) from a
standard by which maternal mortality is estimated. An road and a 20-h walk (100 km) from the nearest com-
alternative source of information includes health service prehensive emergency obstetric care at Arba Minch
data, which depends on reports of health institutions; Hospital.
however, the health service reports in developing coun- We conducted this study as part of an intervention
tries are often biased, as only few people use these ser- project to reduce maternal mortality in Gamo Gofa. The
vices. Also, information gathered through health services work also included studies on the estimation of maternal
is incomplete. It is thus difficult to estimate the accurate mortality through a community-based birth registry, a
MMRs based on institutional data [12]. Therefore, devel- retrospective 5-year recall period household survey, and
oping countries with limited heath service coverage at- a health facilities obstetric care quality study.
tempt to include maternal mortality-related questions in
household surveys, such as the Demographic and Health
Survey (DHS). Although these surveys have contributed The sisterhood method
important information for monitoring interventions, the In the sisterhood method, adult men and women report
surveys are expensive and do not provide the regional the proportion of their adult sisters (born to the same
and local estimates which are needed to improve health mother) dying during pregnancy, childbirth, or within 6
services. weeks following pregnancy [17]. The main objective of
For countries with high maternal mortality and fertility this method is to create a retrospective cohort of women
rates, Graham and colleagues [13] developed an indirect at risk of pregnancy-related death, and to estimate the
sisterhood method for calculating maternal mortality in- lifetime risk (LTR; the chance of a woman dying from
dices. This method is widely used in Africa and Asia to pregnancy-related causes during her entire reproductive
provide community-based maternal mortality estimates period). Then, the LTR is translated into the more con-
[13-15]. Unfortunately, there are no such reports from ventional MMR.
south Ethiopia. Our study aimed to determine the life- The MMR estimate obtained through the indirect
time risk of death of women from pregnancy-related sisterhood method using respondents 15–49 years of age
causes and to calculate the MMR in a rural area in refers to events approximately 10–12 years before the
Gamo Gofa. collection of data. The time of estimation for the MMR
extends up to 35 years from the time of data collection,
Methods when the respondents are older (if included, > 50 years
The setting of age). Therefore, the information obtained from such
We conducted this study in 15 of 30 randomly selected surveys is used as a quick reference of past mortality
rural kebeles (lowest administrative units) in the Bonke rather than of recent events. This method is not recom-
woreda (district) of the Gamo Gofa zone in south-west mended for overseeing the trend over the long period of
Ethiopia. Bonke is one of 15 woredas in the Gamo Gofa maternal mortality or for geographic comparisons [18].
To translate the lifetime risk into the MMR, the present during the first visit, the data collectors
method recommends that the total fertility rate (TFR; re-visited the household the following morning.
the average number of children that would be born to a
woman over her lifetime) should be ≥ 5. In 2000, the Sample size and sampling technique
TFR for the rural Ethiopian population was 6.4 [19]. The sample size recommended by Graham and colleagues
Because this rural area has a high illiteracy rate, and is a was 3000–6000 adult respondents [17]. A more precise
densely-populated, subsistent-farming community, we recommendation of the sample size estimation, which
assumed the population to have similar fertility with considers the margin of error, confidence level, power of
other rural areas in Ethiopia. Therefore we used a TFR the estimate, and the required number of maternal deaths
of 6.4 in the current study. of sisters, suggests a more detailed sample size determin-
ation [20]. The formula which calculates the number of
maternal deaths required for reporting by respondents
The data collection was determined as follows: r ≥ [Zα/2]2 * [100÷% ME]2,
We recruited data collectors who had completed the where r is the number of sister deaths due to maternal
12th grade, lived in the area, and were familiar with the causes that were required, Zα/2 is the standard normal
local language and culture. Five diploma graduates who deviate at a two-sided confidence level of 100[1-α], and
also had a thorough knowledge of the culture and the% ME is the percentage margin of error tolerated by
language of the area supervised the data collectors. Each the investigators.
enumerator was trained for 2 days. The training We used a tolerable margin of error of 10%, and an α
included pre-test field interviews, translation of the value of 5% (two-sided 95% CI). From the formula we
questions, and understanding the different interpreta- calculated [1.96]2 * [100/10]2 = 384 sister deaths due to
tions of the questions by the respondents. pregnancy, childbirth, or 6 weeks after the pregnancy
We asked men and women 15 – 49 years of age the terminated. Hanely and colleagues [20] have suggested
following standard questions using the sisterhood that with 80% statistical power for a community with a
method [17]: MMR > 750 per 100,000 live births, a report of ≥ 384
maternal deaths is expected from interviewing 8000
1. How many sisters (born to the same mother) have adult siblings. In 2000, the MMR estimate was 937 for
you had who survived to reproductive age (15 years Ethiopia [6]. To account for non-responses and missed
of age)? information, we decided to interview 9000 respondents.
2. How many sisters who reached reproductive age We grouped the 30 kebeles of Bonke Woreda into
(15 years of age) are alive now? three climatic zones (hot, temperate, and cold). To ensure
3. How many sisters died? fair representation of all three climatic conditions, we
4. How many sisters died during pregnancy, childbirth, selected one-half of the kebeles in each climatic zone
or 6 weeks after delivery or termination of pregnancy using a lottery method. Thus, we selected 8 of 16 Dega
(cold weather), 4 of 8 Woinadega (moderate temperature),
In addition, we collected data on the age, gender, and and 3 of 6 Kolla (hot temperature) kebeles. Then, the 9000
education of each respondent. Fifteen years of age was respondents were distributed to the study kebeles propor-
considered the common age at which women are expected tionate to the population size.
to undergo menarche. Therefore, we used 15 years as the
proxy age for reaching reproductive age with additional Data analysis
probing of a reproductive age phrase itself. Data collec- SPSS 16 (SPSS, Inc., Chicago, IL, USA) was used for data
tors were carefully trained not to include the responding entry and analysis [21]. We used an inflation adjustment
woman in the reported number of sisters born to her to determine the final number of surviving adult sisters
mother. for the younger respondents (15–24 years of age. This
The questions were translated to Amharic (Ethiopian was done by multiplying the number of respondents in
official state language), and the enumerators adminis- the young age groups by the average number of sisters
tered Amharic using the local Gamotho language. The among the older respondents (25–49 years of age),
enumerators visited each household in the selected which was 2.65 in this data. For example, 2.65* 2443=
communities that had at least one pregnancy during the 6471 adjusted sisters for the 15–19 year old respondents
5 years prior to the study. The enumerators asked the [17]. This factor was used with the assumption that the
four questions (vide supra) to the husband and wife, and younger respondents had sisters who had yet to reach
to the children, if any, who were 15–49 years of age. reproductive age.
Other extended adult family members in the house- Using standard adjustment factors [17], we adjusted
hold were also interviewed. If an adult person was not for the expected proportion of sisters that would have
finished their reproductive age for respondents in each

40
age category. Thus, 90% of the sisters of respondents
Percentage of population
45–49 years of age are expected to have passed through Bonke
their reproductive life, but only 10.7% of the sisters of 30
Gamo Gofa
15–19 year old respondents. The adjustment was imple-
mented so as to determine the number of sister units
20
exposed to maternal death.
This retrospective cohort analysis provided 8,068 sister
units exposed to the risk of maternal death that served 10
as the denominator for calculating the lifetime risk of
maternal death.
The lifetime risk (Q) of maternal death was calculated 0
by Q=r/ β, where r is the number of maternal deaths
9
-1
-2
-2
-3
-3
-4
-4
15
20
25
30
35
40
45
and β is the sister units exposed to the risk of maternal
Age-group (in years)
death. We calculated the MMR as MMR =1-(P) 1/TFR ,
Figure 1 Age distribution of men and women respondents for
where P is the probability of surviving, which equals
Bonke sisterhood study 2011 versus Gamo Gofa zone
(1-Q), and TFR is the total fertility rate [20]. population of the same age group in 2007 Ethiopian National
Census.
Ethics approval
This study was approved by the Ethical Review Commit-
1667 (95% CI, 1564–1769) per 100,000 live births for
tee for Health Research of the Southern Nations Nation-
1998.
alities and the Peoples' Regional State (SNNPRS) Health
Table 2 also shows estimates obtained from male and
Bureau in Ethiopia, and the Regional Committee for
female sibling respondents separately. The lifetime risk
Medical and Health Research Ethics of North Norway
estimate based on male respondents was 0.095 (95% CI,
(REK Nord). We obtained informed oral consent from
0.086-0.105) with a corresponding MMR of 1547 (95%
all of the respondents.
CI,1395-1718) per 100,000 live births (LB). A similar
estimate based on information from female respondents
Results provided a slightly higher lifetime risk of 0.121 (95% CI,
We interviewed 8870 people of the 9000 sample (98.5% 0.104-0.127) and MMR of 1995 (95% CI, 1701–2099)
response rate), and included 96% (8503/ 8870) of per 100,000 LB.
respondents in the analysis. The missing information
from the excluded 4% (367 people) of the respondents
was mainly because of misclassification of age (outside Discussion
the 15–49 year age range) and missing information We calculated a lifetime risk of maternal mortality of
regarding the gender of the respondents. There were no 10.2%, which corresponded to a MMR of 1667 per
maternal deaths reported by those excluded from the 100,000 LB in 1998. There have been no prior
analysis. community-based maternal mortality estimations from
Of the 8503 respondents in the analysis, 5262 (62%) Gamo Gofa, and our study presents the highest estimate
were men and 3241 (38%) were women. The mean age for community-based studies using the sisterhood
of the respondents was 26.4 (SD = 8.7) years (range, method in Ethiopia.
15–49 years). The most frequently reported age of the In Butajira, which is in south central Ethiopia, the
respondents was 30 years, followed by 20 and 18 years MMR was estimated to be 665 per 100, 000 LB in 1996
(Figure 1). using the sisterhood method [11]. The Butajira study
The 8503 respondents reported 22,473 sisters born might have been methodologically more robust than the
to the same mother who survived to the reproductive current study as it was linked to demographic surveil-
age. The average number of adult sisters per respondent lance and probably had a more precise age estimation.
was 2.6. Of the 22,473 sisters who survived to repro- However, Butajira also had better access to health ser-
ductive age, 2,552 (11.35%) had died. Among the sisters vices, and this could also explain the differences in
who had died, 32% (819/2552) were pregnancy-related MMR compared with Bonke. Shiferaw et al. [10]
deaths. reported a MMR of 570 per 100,000 LB from Illubabor
The lifetime risk of death from maternal causes was in western Ethiopia in 1991; however, both studies
0.102 (95% CI, 0.096-0.108) or 10.2% (Table 1). Using a reported MMR rates below the international estimates
TFR of 6.4 for south Ethiopia, we calculated a MMR of for Ethiopia at that time.
Table 1 Maternal mortality estimate using the sisterhood method for the reference period 1998 in rural Bonke, Gamo
Gofa, south-west Ethiopia, 2011
Age of No. of sisters survived age Dead from all Maternal adjustment Sisters units exposed Lifetime
respondents respondents ≥ 15 yrs causes deaths (r) factor ( f) to risk (β) risk (Q )
k e C r f β = ef Q=r/ β
15-19 2443 6471* 428 240 0.107 693 0.346
20-24 1625 4306* 370 152 0.206 887 0.172
25-29 1450 3889 375 152 0.343 1334 0.114
30-34 1235 3135 358 103 0.503 1576 0.065
35-39 812 2201 331 89 0.664 1461 0.061
40-44 523 1397 255 52 0.802 1120 0.046
45-49 415 1074 225 31 0.900 997 0.031
Total 8503 22473 2342 819 8068 0.102
*inflated number of sisters obtained by multiplying the average number of sisters survived for respondents aged 25–49 (which is 2.65 in this data) by the number
of respondents in the younger age groups (age 15–19 and 20–24). Originally reported sisters by the young group were: 5425 for aged 15–19 and 4230 for 20–24
years old respondents.
Hill and colleagues [3] estimated the MMR for Ethiopia been important as the HIV prevalence was < 1% in rural
in 1995 to be 1814 per 100,000 LB, which was similar areas [24].
to our finding. Our estimate was close to the natural An alternative explanation for the high MMR in the
MMR expected without access to contemporary obstet- current study may be that the sisterhood method pro-
ric care. We believe the impoverished and rural Bonke vides a biased estimate through selection or information
area had a high MMR in recent decades when the errors and data adjustments. With respect to selection
population had no access to basic and comprehensive bias, we could have obtained information from many sib-
emergency obstetric care because the population resided lings on a death that involved a single woman. Such mul-
in isolated villages with limited transportation. A recent tiple counting is considered the basis for over-estimation.
national survey in 2008 showed that 7% of all deliveries Potential information biases include misreporting of
take place with health care facilities, and only 3% of facil- age or recall errors on the timing of maternal deaths, or
ities provide comprehensive emergency obstetric care even non-recognition of early pregnancy-related deaths.
[22]. Taking in to account the year of the estimate (1998) To ensure correct age determination, we asked several
and the typical rural location of Bonke, our estimate may probing questions, such as the number of children the
have reflected the reality the Bonke women experienced. respondents had, the year of marriage, and past events
Also, between 1996 and 2000 there were severe malaria (local calendar) to determine the respondent's age. Be-
epidemics in southern Ethiopia, and the Bonke lowlands cause Ethiopia has no system of birth registry, determin-
was no exception, which might have caused additional ation of age data is uncertain, which could lead to
maternal deaths. High MMRs have been associated with errors, such as digit preference, as observed in our data.
high HIV prevalence rates elsewhere [23]. In Ethiopia, Some respondents may also claim to be younger than
however, the effect of the HIV epidemic might not have their real age, as suggested in Figure 1.
With respect to multiple counting as a potential basis
for overestimation, Graham et al. [17] argued that be-
Table 2 Maternal mortality indicators estimated
separately for male and female respondents using the cause the sisterhood method is based on a proportional
sisterhood method for year 1998 in rural Bonke, Gamo relationship, multiple counting in the denominator is
Gofa, South-west Ethiopia, 2011 offset by counting sister deaths in the numerator, thus
Male Female total there is no biased result. Trusell et al. [25] emphasized
Number of respondents 5262 (62%) 3241(38%) 8503
multiple counting of siblings who fall in the sample as
essential for the success of the sisterhood method.
Sisters survived 15 years+ 11235 8838 22473
Therefore, because we did not restrict the siblings dur-
Sisters died of all causes 1483 859 2342
ing data collection and analysis, we cannot rule out mul-
Pregnancy related deaths 482 337 819
tiple counting, but we believe this is not a major source
Sister units of risk exposure 5094 2785 8068 of bias influencing our estimates.
Lifetime risk of maternal death 0.095 0.121 0.102 People often forget past dates of events when respond-
MMR* 1,547 1,995 1,667 ing to research questions. We asked the respondents to
* per 100,000 live births. recall and report the time and cause of maternal deaths
of their adult sisters. Two potential forms of error are Although the estimate obtained by the sisterhood
of concern. First, the respondent could forget the method cannot be used to make geographic comparisons
exact time of the death. This could incorrectly in- and time trend changes, it is useful in providing the
crease the MMR if the respondents reported that the magnitude of the situation in a given area. There have
sisters died 6 weeks after pregnancy termination. Sec- been several policy interventions implemented by the
ond, underreporting could occur if the cause of death Ethiopian government during the past decade. We con-
was misclassified without recognizing early-pregnancy- sider it encouraging that the public health authorities
and abortion-related deaths. However, in rural areas are using Emergency Obstetric Care Guidelines for im-
there are strong social ties, and events such as preg- proving health care in resource-poor settings, and are
nancy are announced early, suggesting a reduced risk working to strengthen the referral system. The Ethiopian
of missed early pregnancy-related maternal deaths. We Government is also setting up primary hospitals for
attempted to probe respondents, especially those reporting every 100,000 population, and thus improving access to
maternal deaths, to ensure the death was within 6 weeks health care. Other important interventions include a
after the pregnancy was terminated. The information pro- malaria prevention campaign through the distribution of
vided is most likely accurate because the 6-week period is bed nets for households, the introduction of two health
the time that most mothers remain at home. The Gachino extension works to all rural villages, training and posting
tradition of women staying at home after delivery is of midwives and health offices, and rapid expansion of
strictly followed by the rural Bonke population. health centres. Therefore, the findings of this study may
The 95% CI of our MMR estimate was narrower com- help establish a baseline to assess the current situation
pared to some other reports using similar methods. The and the effects of the interventions using other methods,
interval was calculated from the 95% CI of the lifetime such as household surveys.
risk, which in turn depends on the number of maternal
deaths counted in the study. In the current study we Conclusion
expected a minimum number of maternal deaths of 384 Our findings suggest that people living in remote and
for 8 000 respondents; however, there was actually 8503 underprivileged Bonke have high MMRs. This highlights
respondents and 819 reported maternal deaths. The the importance to strengthen lifesaving comprehensive
large number of maternal deaths that resulted in the emergency obstetric care in this area, and in similar
narrow 95% CI of the lifetime risk may have caused a rural areas in Ethiopia. Because of uncertainties in our
narrow interval in the MMR. estimates, we also advise using alternative sources of in-
We also showed separate estimates based on informa- formation, such as birth registries and short recall-
tion from male and female respondents (Table 2). In the period household interviews to improve the accuracy of
current study there were more male respondents than the MMR estimate.
women respondents, which may have occurred for the
following reasons: women usually travel to rural open Competing interests
We declare we have no competing interests. YY receives a PhD stipend from
market places that are often far from home in Bonke; The Norwegian state loan for higher education, and BL receives a salary from
and in rural Ethiopia women usually sit in a hidden part the University of Bergen in Norway.
of their home ('Guada' in Amharic) as men chat in the
living room, thus women may not be available for inter- Authors' contributions
YY designed the study, organized the data collection, analyzed the data, and
view as they shy away from interviewers. Nevertheless, wrote the first draft of the manuscript. BL participated in the design of the
the estimates were similar with a slight increase for fe- study, supervised the entire process, and reviewed and modified the drafts
male respondents, which could be because of the close of the manuscript. Both authors revised and approved the final draft of the
manuscript.
relationship sisters have with each other regarding the
sharing of information, such as pregnancy. Most previ- Acknowledgements
ous studies have only enrolled female respondents, des- We would like to thank the Regional Health Bureau in the Southern Nations
pite the recommendation by Graham and colleagues Nationalities and Peoples Regional State in Ethiopia, the Gamo Gofa Zone
Health Department, and the Bonke woreda Health Office for their support
[17] in their original introduction of the sisterhood during the study. We are grateful to the participants for committing their
method to include men in subsequent studies. Male time to interview and providing information. The Centre for International
respondents can more easily be accessed for interviews Health at the University of Bergen in Norway funded this study.
in rural places where they gather for social meetings Received: 21 March 2012 Accepted: 31 October 2012
than women who mostly stay at home or travel to mar- Published: 23 November 2012
ket places. Thus, in future studies interviewing men
alone may be an efficient way to reduce the house- References
1. WHO: International statistical classification of diseases and related health
to-house visits in search of women respondents among problems, tenth revision instruction manual. Geneva: World Health
the often scattered households in rural areas. Organization; 2004.
2. Ronsmans C, Graham WJ: Maternal mortality: who, when, where, and

why. Lancet 2006, 368(9542):1189–1200.
3. Hill K, AbouZahr C, Wardlaw T: Estimates of maternal mortality for 1995.
Bull World Health Organ 2001, 79:182–193.
4. Hill K, Thomas K, AbouZahr C, Walker N, Say L, Inoue M, Suzuki E: Estimates
of maternal mortality worldwide between 1990 and 2005: an
assessment of available data. Lancet 2007, 370(9595):1311–1319.
5. WHO: Health and the Millennium Development Goals. Geneva: World Health
Organization; 2005.
6. Hogan MC, Foreman KJ, Naghavi M, Ahn SY, Wang M, Makela SM, Lopez
AD, Lozano R, Murray CJL: Maternal mortality for 181 countries,
1980–2008: a systematic analysis of progress towards Millennium
Development Goal 5. Lancet 2010, 375(9726):1609–1623.
7. Bhutta ZA, Chopra M, Axelson H, Berman P, Boerma T, Bryce J, Bustreo F,
Cavagnero E, Cometto G, Daelmans B, et al: Countdown to 2015 decade
report (2000–10): taking stock of maternal, newborn, and child survival.
Lancet 2010, 375(9730):2032–2044.
8. Ngoc NTN, Merialdi M, Abdel-Aleem H, Carroli G, Purwar M, Zavaleta N,
Campódonico L, Ali MM, Hofmeyr GJ, Mathai M: Causes of stillbirths and
early neonatal deaths: data from 7993 pregnancies in six developing
countries. Bull World Health Organ 2006, 84(9):699–705.
9. Kwast BE, Rochat RW, Kidane-Mariam W: Maternal Mortality in Addis
Ababa, Ethiopia. Stud Fam Plann 1986, 17(6):288–301.
10. Shiferaw T, Tessema F: Maternal mortality in rural communities of
Illubabor, southwestern Ethiopia: as estimated by the sisterhood
method. Ethiop Med J 1993, 31(4):239–249.
11. Berhane Y, Andersson T, Wall S, Byass P, Hogberg U: Aims, options and
outcomes in measuring maternal mortality in developing societies. Acta
Obstet Gynecol Scand 2000, 79(11):968–972.
12. Graham W, Ahmed S, Stanton C, Abou-Zahr C, Campbell O: Measuring
maternal mortality: an overview of opportunities and options for
developing countries. BMC Med 2008, 6(1):12.
13. David P, Kawar S, Graham W: Estimating maternal mortality in djibouti: an
application of the sisterhood method. Int J Epidemiol 1991, 20(2):551–557.
14. Font F, González MA, Nathan R, Lwilla F, Kimario J, Tanner M, Alonso PL:
Maternal mortality in a rural district of southeastern Tanzania: an
application of the sisterhood method. Int J Epidemiol 2000, 29(1):107–112.
15. Bhat PNM: Maternal mortality in India: an update. Stud Fam Plann 2002,
33(3):227–236.
16. Central Statistical Authority of Ethiopia: Statistical Tables for the 2007
Population and Housing Census of Ethiopia: SNNPR vol.1, September 2010;
2012. www.csa.gov.et (accessed February 24, 2012).
17. Graham W, Brass W, Snow RW: Estimating maternal mortality: the
sisterhood method. Stud Fam Plann 1989, 20(3):125–135.
18. Olsen ØE, Ndeki S, Norheim OF: Human resources for emergency
obstetric care in northern Tanzania: distribution of quantity or quality?
Hum Resour Heal 2005, 3(1):5.
19. Central Statistical Authority AA: Demographic and Health Survey Report 2000.
Ethiopia and ORC Macro, Calverton, Maryland, USA: Ethiopia: DHS, 2000,
Final Report (English); 2001.
20. Hanley JA, Hagen CA, Shiferaw T: Confidence Intervals and Sample-size
Calculations for the Sisterhood Method of Estimating Maternal Mortality.
Stud Fam Plann 1996, 27(4):220–227.
21. Olsen ØE, Ndeki S, Norheim OF: Availability, distribution and use of
emergency obstetric care in northern Tanzania. Health Policy Plan 2005,
20(3):167–175.
22. Admasu K, Haile-Mariam A, Bailey P: Indicators for availability, utilization,
and quality of emergency obstetric care in Ethiopia, 2008. Int J Gynecol
Obstet 2011, 115(1):101–105. Submit your next manuscript to BioMed Central
23. Bicego G, Boerma JT, Ronsmans C: The effect of AIDS on maternal and take full advantage of:
mortality in Malawi and Zimbabwe. AIDS 2002, 16(7):1078–1081.
24. WHO, UNICEF, UNFPA, The World Bank: Trends in maternal mortality:1990 to • Convenient online submission
2010. WHO, UNICEF, UNFPA and The World Bank estimates. Geneva: World
• Thorough peer review
Health Organization; 2012.
25. Trussell J, Rodriguez G: A note on the sisterhood estimator of maternal • No space constraints or color figure charges
mortality. Stud Fam Plann 1990, 21(6):344–346. • Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
doi:10.1186/1471-2393-12-136
Cite this article as: Yaya and Lindtjørn: High maternal mortality in rural • Research which is freely available for redistribution
south-west Ethiopia: estimate by using the sisterhood method. BMC
Pregnancy and Childbirth 2012 12:136.
Submit your manuscript at
www.biomedcentral.com/submit
IV
Girma et al. BMC Health Services Research 2013, 13:459
RESEARCH ARTICLE Open Access
Lifesaving emergency obstetric services are

inadequate in south-west Ethiopia: a formidable
challenge to reducing maternal mortality in
Ethiopia
Meseret Girma1, Yaliso Yaya2,4*, Ewenat Gebrehanna3, Yemane Berhane3 and Bernt Lindtjørn2
Abstract
Background: Most maternal deaths take place during labour and within a few weeks after delivery. The availability
and utilization of emergency obstetric care facilities is a key factor in reducing maternal mortality; however, there is
limited evidence about how these institutions perform and how many people use emergency obstetric care
facilities in rural Ethiopia. We aimed to assess the availability, quality, and utilization of emergency obstetric care
services in the Gamo Gofa Zone of south-west Ethiopia.
Methods: We conducted a retrospective review of three hospitals and 63 health centres in Gamo Gofa. Using a
retrospective review, we recorded obstetric services, documents, cards, and registration books of mothers treated
and served in the Gamo Gofa Zone health facilities between July 2009 and June 2010.
Results: There were three basic and two comprehensive emergency obstetric care qualifying facilities for the
1,740,885 people living in Gamo Gofa. The proportion of births attended by skilled attendants in the health facilities
was 6.6% of expected births, though the variation was large. Districts with a higher proportion of midwives per
capita, hospitals and health centres capable of doing emergency caesarean sections had higher institutional
delivery rates. There were 521 caesarean sections (0.8% of 64,413 expected deliveries and 12.3% of 4,231 facility
deliveries). We recorded 79 (1.9%) maternal deaths out of 4,231 deliveries and pregnancy-related admissions at
institutions, most often because of post-partum haemorrhage (42%), obstructed labour (15%) and puerperal sepsis
(15%). Remote districts far from the capital of the Zone had a lower proportion of institutional deliveries
(<2% of expected births compared to an overall average of 6.6%). Moreover, some remotely located institutions
had very high maternal deaths (>4% of deliveries, much higher than the average 1.9%).
Conclusion: Based on a population of 1.7 million people, there should be 14 basic and four comprehensive
emergency obstetric care (EmOC) facilities in the Zone. Our study found that only three basic and two
comprehensive EmOC service qualifying facilities serve this large population which is below the UN’s minimum
recommendation. The utilization of the existing facilities for delivery was also low, which is clearly inadequate to
reduce maternal deaths to the MDG target.
* Correspondence: [email protected]
2
Centre for International Health, University of Bergen, Bergen, Norway
4
Arba Minch College of Health Sciences, Arba Minch, Ethiopia
Full list of author information is available at the end of the article
© 2013 Girma et al.; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.
Girma et al. BMC Health Services Research 2013, 13:459 Page 2 of 8
Background functions): administration of parenteral antibiotics, par-

The fifth Millennium Development Goal (MDG 5) is to enteral oxytocic drugs, parenteral anticonvulsants for
reduce maternal mortality by 75% between 1990 and pre-eclampsia, manual removal of retained placentas, re-
2015. Although there are good tools available to help removal of retained products of conception and assisted
duce maternal deaths [1], the limited availability and vaginal delivery (vacuum extractions or forceps deliveries)
poor quality of services cause nearly 300,000 maternal [8]. Institutions providing comprehensive EmOC should
deaths in the world every year, with approximately 85% also be capable of performing caesarean sections, blood
of the 287,000 global maternal deaths taking place in transfusions and services provided by the basic EmOC
both Sub-Saharan Africa (56%) and southern Asia (29%) institutions.
[2]. In 2008, more than half of all maternal deaths in the The findings of an assessment regarding the avai-
world occurred in six countries: Afghanistan, Demo- lability, quality and distribution of EmOC services is
cratic Republic of the Congo, Ethiopia, India, Nigeria important for health professionals and policymakers in-
and Pakistan [3], with most of these preventable and un- volved in maternal health services. With high maternal
acceptable deaths occurring around delivery or a few mortality rates and a mostly rural population, it is im-
days after [4]. Bleeding during pregnancy and birth, portant to evaluate emergency obstetric care provided at
obstructed and prolonged labour and pregnancy-related public health institutions in Ethiopia.
hypertension represent the leading causes of deaths A recent study has shown there are too few health in-
among women of reproductive age in resource-poor stitutions providing EmOC to meet the UN standards of
countries [5]. at least five (four basic and one comprehensive) EmOC
The maternal mortality ratio (MMR) for Ethiopia was institutions per 500,000 population in Ethiopia [9]. Only
1,061 (665–1,639) in 1980, 968 (600–1,507) in 1990, 937 7% of deliveries took place in institutions, including only
(554–1,537) in 2000 and 590 (358–932) in 2008 [3]. 3% in institutions that routinely provided all signal func-
Nevertheless, the results of the 2011 Demographic and tions. Six percent of women with obstetric complications
Health Survey (DHS) revealed that there has been little were treated in health institutions, whereas only one-
progress in reducing maternal mortality [6]. The DHS half of these women were treated in fully functional
estimate of the MMR for 2011was 676 (541–810) per comprehensive EmOC facilities [9]. The study concluded
100,000 live births. A study has also showed that in sub- that far too few public institutions in Ethiopia meet the
Saharan African countries, the progress towards achie- indicators set by the UN standards.
ving MDG5 has been slow because of a poor quality of Ethiopia therefore faces many challenges, not only be-
care, low access, inadequate skilled personnel and finan- cause of a limited number of adequately functioning ob-
cial barriers to care [3]. stetric facilities, but also because of its large population
The WHO recommends the use of process indicators and mountainous topography, with large parts of the
on emergency obstetric care (EmOC) facilities to assist populations living in remote areas. We conducted this
in monitoring the progress in maternal mortality re- study to assess the availability (coverage), quality (func-
duction efforts, which are considered necessary for tionality) and utilization of emergency obstetric care fa-
planning, implementing and monitoring initiatives to cilities in Gamo Gofa in south-west Ethiopia.
improve maternal health [7]. Unfortunately, there is li-
mited evidence regarding how these institutions are dis- Methods
tributed, how well the existing facilities perform and Setting
how many people use them in Gamo Gofa, Ethiopia. The study was conducted in the Gamo Gofa Zone in
The investment in maternal health programmes can be south-west Ethiopia (see map in Figure 1). Nearly 1.7
evaluated by measuring input indicators (midwifery million people live in the area, with 90% living in rural
training), process (the number of midwives posted) and communities. The Zone has 15 woredas (districts) and
outcomes (the uptake of skilled delivery care). However, two town administrations, each being directly adminis-
the assessment of impacts such as the reduction in mor- tratively responsible to the Zone. However, people in the
tality in a community can show the effects of long-term surrounding districts of the towns, as well as the towns
interventions. themselves, use the health facilities/services/ in these
The availability and use of emergency obstetric care towns. The Zone represents three climatic zones (cold,
services is important for reducing maternal morbidity temperate and hot), where most of the people live in
and mortality. Based on the capacity to provide lifesa- highlands 2,000 metres above sea level and practice sub-
ving emergency obstetric procedures, a health institution sistence farming. There are few all-weather roads in the
can be classified as basic or comprehensive emergency area, although most of the population lives in the high-
obstetric care facility [8]. Basic EmOC institutions are lands without access to roads. Health care is provided by
expected to provide the following six services (signal three hospitals, 63 health centres and by rural health
Figure 1 Administrative map of Gamo Gofa Zone and its Woredas, south-western Ethiopia, 2010.
extension workers in 483 kebeles, which are Ethiopia’s caesarean sections. The programme aims to support
lowest administrative units, with an average coverage of public health services to help reduce maternal and neo-
1,000 households (population of 5,000). Hospitals are ex- natal deaths [10], and is primarily a support to govern-
pected to provide comprehensive emergency obstetric ment institutions with training, supervision and providing
care, while the health centres are expected to provide the institutions with basic equipment. Thus, while the
basic emergency obstetric care. Due to limited access to population in 2007 had only one hospital capable of doing
hospitals, senior staff (health officers) are given minimal comprehensive EmOC for approximately 1.7 million
training, and provide services such as caesarean sections people, the services such as caesarean section delivery had
in some health centres. Four (6%) of the health insti- improved to three hospitals and two health centres (one
tutions in the area are accessible by asphalt roads, 21 fa- institution per 350,000 people) by 2010. The project also
cilities (32%) are accessed by all-weather gravel roads, 30 includes studies on estimating maternal and neonatal
health centres (46%) are only accessible by car during mortality through community-based birth registries, esti-
the dry season and 11 institutions (17%) could not be mations of maternal mortality through the sisterhood
accessed by a vehicle at the time of the survey. method, large-sample household survey to estimate ma-
We conducted this study as part of a public health ternal and neonatal deaths and a health facilities obstetric
intervention project aimed at reducing maternal morta- care quality study (the current study).
lity in Gamo Gofa. A few years prior to the study, the
intervention programme (“Reducing Maternal Mortality Data collection and instruments
in south-west Ethiopia”) had started training non- We collected data using questionnaires and procedures
physician clinicians (NPCs) to provide EmOC, including developed according to UN guidelines [8], and assessed
the performance of health institutions using the same Gondar, so ethical clearance was therefore obtained from
guidelines. We recruited eight health officers (people the University of Gondar. After obtaining the clearance,
with bachelor’s degrees in clinical and community medi- we received written permission to carry out the study
cine) to collect the data, and the health officers were from the Gamo Gofa Zone Health Department and each
trained for two days before visiting the institutions. If of the woreda health authorities. Before starting to
deemed necessary, key health personnel at each institu- record information about the health institutions, we in-
tion were interviewed for the clarification of any re- formed the leaders of each of the health institutions
corded data. about the study. Lastly, we received a written consent
Between September and November 2010, we visited 66 from the head of each facility to allow us to conduct the
health institutions, the three hospitals in Arba Minch, study at the institution. The Regional Committee for
Chencha and Sawla and 63 health centres throughout Medical and Health Research Ethics of North Norway
the Zone. When visiting the institutions, we retrospec- (REK Nord) also approved this study.
tively reviewed one year of available obstetric services,
records, documents, cards and registration books related Results
to delivery services. As a result, we collected information Availability of EmOC
from records and registers such as admission registers, We visited and reviewed all of the 66 health institutions
delivery registers, delivery log books, referral registers (hospitals and health centres) in Gamo Gofa. Of these,
and death registers. We also registered the number of only the two hospitals in Arba Minch and Sawla (3% of
staff available for obstetric care at each of the health in- institutions) provided all signal functions, and were thus
stitutions we reviewed. As recommended by the WHO designated as providing comprehensive EmOC. Three
guidelines for areas with fewer than 100 facilities, we in- health centres (4.5%) provided basic EmOC, but did not
cluded all hospitals and health centres in Gamo Gofa in have a blood bank, while 61 (92%) facilities lacked some
the current study [8]. or all signal functions and 40 (60.6%) institutions lacked
> 5 of the signal functions. Only 36 (54.5%) institutions
Data analysis provided parenteral antibiotics when needed, 61 of 66
We used SPSS (version 16; SPSS, Inc., Chicago, IL, USA) (92%) performed assisted vaginal deliveries, 47 (71%)
for data entry and statistical analysis, and we performed performed the manual removal of placentas, 23 (35%)
a descriptive analysis to present rates and ratios. We used parenteral oxytocin and 14 (21%) used anticonvul-
calculated the expected number of deliveries for each sants during eclampsia when indicated in the last three
woreda using the Central Statistical Authority (CSA) months (Figure 2).
estimates for birth rates (3.7%) and woreda population
size [11]. Delivery, complications, and deaths
A total of 4,231 deliveries and related admissions took
Operational definition place at the health institutions over the course of 1 year.
An EmOC facility refers to whether or not an institution Furthermore, there was an annual average of 522 de-
is fully functioning as a basic or comprehensive facility liveries at each hospital, 213 deliveries at two health
[8]. Functioning is defined by nine signal functions, as centres capable of providing emergency obstetric care,
follows: administering parenteral antibiotics, administer- including caesarean sections, and an average of 32 deli-
ing parenteral oxytocic drugs, administering parenteral veries at each of the remaining 61 health centres. Five
sedatives, manual removal of the placenta, removal of health centres did not have any recorded deliveries and
retained products of conception, vacuum-assisted vagi- 24 health centres had one delivery per month during the
nal deliveries or forceps deliveries, performing caesarean year surveyed. A total of 521 deliveries were done by
sections, performing newborn resuscitation and the caesarean section (0.8% of 64,413 expected births and
availability of a blood transfusion service. An institu- 12.3% of 4,231 facility births), and over the one year, we
tion that had not performed any or only some of the recorded 10 neonatal deaths and 178 stillbirths.
signal functions during the past three months was de- We reviewed 1,031 of 4,231 (24.3%) births and preg-
fined as a non-functioning EmOC. The reasons for nancy-related admissions at the health facilities as
not performing signal functions may vary, and include complicated cases. The complications were further ca-
a lack of equipment or medications or a lack of avai- tegorized as complications associated with abortions
lable skilled personnel. (28.2%), obstructed labour (18%), prolonged labour
(16.9%), post-partum haemorrhage (7.3%), antepartum
Ethical issues haemorrhage (6.3%), pre-eclampsia or eclampsia (4%)
The data for this study was collected as a part of and unclassified (7.5%). We recorded 79 maternal deaths,
Meseret Girma’s master thesis at the University of with the primary causes of deaths being haemorrhage
70
5
60
19
50 30 28
No of facilities
43 45
40 52
61 58 57
64 63
30 61
47
20 38 Do not provided
36
10 23 21
14 provided
5 8 9
0 2 3
signal functions
Figure 2 Signal functions provided at all health centres and hospitals in Gamo Gofa Zone between July 2009 and June 2010.
(42%), obstructed labour (15%), puerperal sepsis (15%), with the woredas with the largest towns (Arba Minch
prolonged labour (8%) and complications from abortions and Sawla), having the highest proportion of institutional
(8%, Table 1). Table 2 shows that the proportion of institu- deliveries. When analysing the proportion of institu-
tional deaths varied between districts. Very high mortality tional deliveries per institutional catchment area, we de-
rates (61% and 28%) were recorded in two rural and re- termined that the proportion varied from zero to an
mote woredas; these woredas also had very low institu- average of > 20% in the two woredas with towns having
tional delivery rates, and few midwives worked at the hospitals. The institutional delivery rate was approxi-
institutions (Table 2). mately 3% in areas with health centres not fulfilling the
criteria of basic EmOC, while in contrast, areas such as
Proportions of births in all facilities and caesarean Kamba, with health centres capable of providing EmOC
sections and performing caesarean sections, had a higher rate of
Over the course of one year, we recorded 4,231 births at institutional deliveries. We used a Pearson product–
the health institutions. Consequently, 6.6% of the ex- moment correlation analysis to determine the correla-
pected 64,413 deliveries occurred at institutions in tion of the rate of institutional deliveries in the districts
Gamo Gofa. Table 2 shows the variations in institutional to the proportion of midwives in the catchment popula-
deliveries between the different administrative districts, tion of the district and the number of physicians in the
districts (where possible). Woredas with a higher ratio of
Table 1 Major causes of pregnancy and birth midwives per population (r = 0.71; p < 0.01), and where
complications and maternal deaths in hospitals and doctors worked (r =0.66; p < 0.01), were associated with
health centres in Gamo Gofa, south-west Ethiopia, July a higher proportion of institutional deliveries.
2009 to June 2010
Causes Complications Deaths Discussion
No. % No. % Based on the total population of 1,740,885, there should
Ante partum haemorrhage 65 6.3 1 1 have been 14 basic and four comprehensive EmOC fa-
Post-partum haemorrhage 75 7.3 33 42 cilities in the Zone. There was a sufficient number of
Prolonged labour 174 16.9 6 8
health facilities in the Zone, if functional, that could
serve as a basic EmOC, and the current study showed
Obstructed labour 186 18.0 12 15
that only three basic and two comprehensive EmOC fa-
Puerperal sepsis 57 5.5 10 13 cilities served the population, which is clearly inadequate
Complication of abortion 291 28.2 6 8 and below the UN’s minimum recommendations [8].
Pre-eclampsia 41 4.0 4 5 The proportion of institutional deliveries varied greatly,
Ruptured uterus 65 6.3 5 6 and five of the health centres did not offer delivery ser-
Others 77 7.5 2 3
vices to the catchment population, whereas 24 health
centres provided one or fewer deliveries per month for
Total 1,031 100 79 100
at least one year. Hence, the population in the area has
Table 2 Expected births, institutional deliveries and health human resource distributions for 66 health institutions,
2010, Gamo Gofa, south-west Ethiopia
Woreda Population No. of Expected Institutional Midwives Nurses and health Doctors† Maternal
institutions no. of births§ deliveries officers† deaths
No. % No. Per 100,000 No. Per 100,000 No.
population population
Melokoza 131,009 5 4,847 102 2.1 2 1.5 20 15.3 0 0
Denba Goffa** 114,309 5 4,230 417 9.9 11 9.6 8 7 1 17
Kucha 162,513 6 6,013 108 1.8 2 1.2 13 8 0 5
Boreda 74,008 4 2,738 30 1.1 5 6.8 22 29.7 0 1
Merab Abaya 81,819 4 3,027 287 9.5 4 4.9 22 26.9 0 5
Arba Minch Zuria* 264,927 7 9,802 1,809 18.5 17 6.4 54 20.4 5 3
Chencha 122,193 5 4,521 409 9 8 6.5 45 36.8 2 6
Dita 91,433 4 3,383 163 4.8 3 3.3 16 17.5 0 7
Daramalo 88,232 2 3,265 54 1.7 3 3.4 3 3.4 0 0
Zala 80,931 5 2,995 58 1.9 3 3.7 36 44.5 0 2
Ubadebretsehay 75,377 3 2,789 37 1.3 2 2.7 15 19.9 0 0
Kemba 169,756 7 6,281 411 6.5 4 2.4 13 7.7 0 3
Bonke 173,240 5 6,410 276 4.3 4 2.3 11 6.3 0 3
Geze Goffa 74,951 3 2,773 56 2 3 4 2 2.7 0 21
Oyida 36,187 1 1,339 14 1 2 5.5 1 2.8 0 6
Gamo Gofa Zone 1,740,885 66 64,413 4,231 6.6 73 4.2 281 16.1 8 79
*Includes Arba Minch Town; **Includes Sawla Town; †Available for obstetric care means those who participated in obstetric care.
§
Expected number of births = 3.7% (national annual crude birth rate) of the woreda catchment population.
unequal access to obstetric care. Areas with hospitals possible explanation could be the idea that basic emer-
and health centres providing comprehensive EmOC had gency obstetric care is rather new, and that these ser-
higher rates of institutional deliveries. The number of vices have not yet been given sufficient priority within
midwives per population was also an important deter- the health system. A study conducted in Tanzania has
mining factor for institutional delivery rates. also revealed that there were fewer basic EmOC facilities
Another important finding was that 28% of women compared to comprehensive EmOC facilities, which is in
had abortion complications. Ethiopian law allows an contrast to UN standards [13]. A study conducted in a
abortion when the pregnancy is due to rape (without the low-income country has shown that many women re-
woman being asked to provide evidence of rape), when ported dissatisfaction with unprofessional and careless
there is a medical threat to the mother and when the behaviour at health facilities, and preferred the care of
foetus has serious irreversible malformations [12]. The traditional birth attendants or relatives [14].
current findings suggest that the women are either un- Our study demonstrated that only 6.6% of expected
aware of such services, or may have limited access to- deliveries occurred at health institutions, while a caesa-
and use of contraceptive services. rean section was performed for only 0.8% of the ex-
Our study represents the first mapping of delivery ser- pected births. The overall rate of facility deliveries was
vices in a rural Ethiopian district, and the strength of the even lower without the relatively higher contribution of
study was that it included all health institutions in births in the five better EmOC facilities. The UN mini-
the Gamo Gofa Zone. Although we attempted to record mum is that 10% of expected deliveries should take place
the relevant work being done at the institutions, our data in EmOC facilities to help reduce maternal mortality in
may be incomplete, as facility records of delivery compli- an area [8]. Additionally, the rates of institutional deliv-
cations and deaths are often incomplete. eries varied from one area to the other, thus suggesting
Approximately one-fourth of facilities did not provide unequal use and access to obstetric care. Using the UN
at least three or four of the signal functions, while 60% guidelines as a reference, both the number of institu-
did not provide > 5 of the signal functions. This could be tional deliveries and caesarean sections were far below
because of an inadequate number of trained staff and a what is regarded as adequate in order to reduce mater-
lack of the necessary supplies such as medications, blood nal deaths [8]. These results agree with earlier research
transfusion bags and resuscitation equipment. Another conducted in Ethiopia, as well as in other countries
[9,14]. The possible reasons for these findings could be and that such information should be used to improve
because people live far away from adequately func- the coverage and quality of health services.
tioning institutions or because of social and cultural res-
trictions for women to use health institutions during Conclusion
deliveries [15]. Our data suggest that most of the institu- Our study showed that the availability, use and quality
tions do not provide essential delivery services, and a of basic and comprehensive EmOC facilities fell below
lack of availability of services near the patients’ homes the accepted standard. This poses a formidable challenge
probably explains the low caesarean section rates. to achieving the MDG related to maternal health. Many
Seventy-nine maternal deaths occurred at the institu- women visiting health facilities with complications after
tions during the one year surveyed, and the case fatality abortions need closer attention. Nonetheless, we find it
rate among women with obstetric complications was encouraging that current efforts by the public health
higher than the minimum standard set by the UN [8]. authorities to use emergency obstetric care guidelines
The causes of deaths are similar to studies elsewhere for improving health care in resource-poor settings, and
[16], and if the maternal mortality ratio is 590 per the works to help strengthen the referral system. It is
100,000 live births [3], only one in five of expected ma- also encouraging to learn a new effort by the Ethiopian
ternal deaths are recorded at health institutions. government to set up primary hospitals for every 100,000
Approximately one-fourth of all deliveries were com- of the population, thereby improving access to health care.
plicated. This high proportion of complicated deliveries
Competing interests
shows that the population seeks care when compli- The authors declare that they have no competing interests.
cations arise during home deliveries. It is therefore re-
commended that all women with complicated deliveries Authors’ contributions
MG conceived the study, coordinated data collection, analysed and
should be treated in obstetric emergency care facilities; interpreted the data, and prepared the draft manuscript. YY conceived the
however, with an expected complication rate of 10%, far study, helped to organize the data collection, analysed and interpreted the
too few women with complications received adequate data, and prepared the draft manuscript. EG and YB supervised MG’s master
thesis, and took part in the data collection, data analysis, and writing of the
care. A qualitative study with informants of 42 maternal paper. BL conceived the study, advised on the data collection, interpreted
deaths in the Gambia highlighted the challenges mothers the data, and helped to write the manuscript. All the authors have read and
face to reach lifesaving health facilities. The major bar- approved the submitted version of the manuscript.
riers described were as follows: an under-estimation of
Acknowledgements
the severity of the complications, a bad experience with We would like to thank the data collectors and staff at all the health
the health-care system, a lack of transportation and pro- institutions in the Gamo Gofa Zone for their relentless effort exerted during
longed transportation [17]. Moreover, large parts of the the data collection. The study was done with financial support from the
NORAD funded project, “Reducing Maternal Mortality (RMM) in south-west
population in our study area live in remote mountainous Ethiopia”.
areas, far away from the health institutions. The high
proportion of obstetric complications and high maternal Author details
1
Department of Public Health, College of Medicine and Health Sciences,
death rates in some institutions suggest that health cen- Arba Minch University, Arba Minch, Ethiopia. 2Centre for International Health,
tres do not refer such cases to places where the proper University of Bergen, Bergen, Norway. 3Department of Reproductive Health
management of complicated births are available. It may and Nutrition, Addis Continental Institute of Public Health, Addis Ababa,
Ethiopia. 4Arba Minch College of Health Sciences, Arba Minch, Ethiopia.
also show that there is a lack of trained personnel who
can provide correct interventions. Received: 10 May 2012 Accepted: 31 October 2013
We have noted that remote rural districts without in- Published: 4 November 2013
stitutions doing comprehensive or basic EmOC have References

lower institutional deliveries, which is consistent with 1. Campbell OMR, Graham WJ: Strategies for reducing maternal mortality:
previous a study in a low-income country [13]. The dis- getting on with what works. Lancet 2006, 368(9543):1284–1299.
2. WHO, UNICEF, UNFPA, The World Bank: Trends in maternal mortality:1990 to
tance to health services exerts a dual influence on use, 2010. WHO, UNICEF, UNFPA , and The World Bank estimates. Geneva,
as a disincentive to seeking care first and as an obstacle Switzerland: World Health Organization. available at: http://whqlibdoc.who.int/
to reaching care after a decision has been made to seek publications/2012/9789241503631_eng.pdf [accessed 1o September 2012].
3. Hogan MC, Foreman KJ, Naghavi M, Ahn SY, Wang M, Makela SM,
care [14]. Although we showed an inadequate coverage Lopez AD, Lozano R, Murray CJL: Maternal mortality for 181 countries,
of delivery services, these results can be used as baseline 1980–2008: a systematic analysis of progress towards Millennium
data for planning, improving and implementing delivery Development Goal 5. Lancet 2010, 375(9726):1609–1623.
4. Ronsmans C, Graham WJ: Maternal mortality: who, when, where, and
services in rural Ethiopia. Studies have shown that the why. Lancet 2006, 368(9542):1189–1200.
UN guidelines to assess the process indicators have been 5. Khan KS, Wojdyla D, Say L, Gülmezoglu AM, van Look PFA: WHO analysis of
proven to be generally effective in identifying the level of causes of maternal death: a systematic. Lancet 2006, 367:1066–1074.
6. CSA-Ethiopia, ICF-International: Ethiopia Demographic and Health Survey
emergency obstetric care [18]. We suggest that similar 2011. Maryland, USA: Central Statistical Agency of Ethiopia and ICF
studies should be conducted at all zones in the country, International Addis Ababa, Ethiopia and Calverton; 2012.
7. Wardlaw TMD: Process indicators for maternal mortality programmes.

In Reproductive Health Matters. Blackwell: Safe Motherhood Initiatives: Critical
Issues Oxford; 1999:24–30.
8. WHO, UNICEF, UNFPA, AMDD: Monitoring Emergency Obstetric Care, a
handbook. WHO publications, 2009. Geneva, Switzerland: World Health
Organization; 2009. available at: http://whqlibdoc.who.int/publications/2009/
9789241547734_eng.pdf (Accessed 10 March 2012.
9. Admasu K, Haile-Mariam A, Bailey P: Indicators for availability, utilization,
and quality of emergency obstetric care in Ethiopia, 2008. Int J Gynaecol
Obstet 2011, 115(1):101–105.
10. Project, RMM: Project document for Reduction of maternal mortality (RMM) in
south-west Ethiopia. http://www.lindtjorn.no/page1/page11/page11.html
(accessed December 30, 2011).
11. Federal Demographic Republic Ethiopia, Population Census Commission:
Summary and Statistical Report of the 2007 Population and Housing Census.
Addis Ababa: UNFPA; 2008.
12. FDR-Ethiopia: Technical and Procedural Guidelines for Safe Abortion Services in
Ethiopia. Addis Ababa: Ministry of Health; 2006.
13. Olsen ØE, Ndeki S, Norheim OF: Availability, distribution and use of
emergency obstetric care in northern Tanzania. Health Policy Plan 2005,
20(3):167–175.
14. Gabrysch S, Campbell OM: Still too far to walk: literature review of the
determinants of delivery service use. BMC Preg Childbirth 2009, 9:34.
15. Berhane Y, Gossaye Y, Emmelin M, Hogberg U: Women’s health in a rural
setting in societal transition in Ethiopia. Soc Sci Med 2001, 53(11):1525–1539.
16. Gaym A: Maternal mortality studies in Ethiopia–magnitude, causes and
trends. Ethiop Med J 2009, 47(2):95–108.
17. Cham M, Sundby J, Vangen S: Maternal mortality in the rural Gambia, a
qualitative study on access to emergency obstetric care. Reprod Health
2005, 2(3):1–8.
18. Paxton A, Bailey P, Lobis S: The United Nations Process Indicators for
emergency obstetric care: reflections based on a decade of experience.
Int J Gynecol Obstet 2006, 95(2):192–208.
doi:10.1186/1472-6963-13-459
Cite this article as: Girma et al.: Lifesaving emergency obstetric services
are inadequate in south-west Ethiopia: a formidable challenge to
reducing maternal mortality in Ethiopia. BMC Health Services Research
2013 13:459.
Submit your next manuscript to BioMed Central

and take full advantage of:
• Convenient online submission

• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution
Submit your manuscript at

www.biomedcentral.com/submit
Appendices I: Questionnaires
Questionnaire for Paper I part I (The birth registry format and variables)
Woreda (name) __________________ Kebele (name) _____________________
x Kebele distance from: 1) the nearest health centre ____km (______minutes walking distance)
2) the nearest hospital ________km (_______hours walking distance)
x Road type from Kebele to the nearest health facility:
a) asphalt b) all-season functional gravel c) dry season gravel d) no motorable road
Parity of mother (number of births including this)
Liveborn on 28th day after birth? 1) alive 2) died
(F)
Mother on 42nd day after birth? 1) alive 2) died
How many born in this birth 1) single 2) twins
Was there a problem in this birth? 1 ) Yes 2) No
Any illness in this pregnancy? 1) Yes 2) No
Gravidity of mother (number of pregnancies
If mother died, main cause (symptom/sign)

How many ANC visit in this pregnancy?
(E)
(B)
Education of mother (completed grade)
Education of father (completed grade)
(G)
(C)
What was fetal condition at birth?

Place of birth for the current birth
(D)
If mother died, time of death ?

Who helped the current birth?
No. of alive children now
Have you been referred?

(A)
Mode of delivery?
Name of mother
Sex of newborn
Name of father
Age of mother
Date of birth
Serial no.
Descriptions and coded choices

(A) mode of delivery (B) place of birth (C) who helped the birth
1) Normal labour (Vaginal) 1) at home 1) Traditional birth attendant
2) Vaginal instrument (vacuum, forceps) 2) at family home 2) family (friend)
3) Caesarean section 3) at a health post 3) Health Extension
(D) Referred? 4) at a health centre 4) professional (midwife, nurse, Dr.)
1) not referred 5) in a hospital (E) fetus at birth?
2) Yes to health centre 6) in private clinic 1) alive and normal
3) Yes to hospital 2) alive but malformed
(F) Cause (mother death)? (G) Time (mother death) 3)stillborn
1) bleeding (haemorrhage) 1) during pregnancy
2) fever (infection) 2) during labour and delivery
3) convulsion 3) postnatal period (birth to 42 days)
4) Others (mention)
1
Validation of the birth registry study questionnaire (Paper I part 2)
N.B: To data collector: please collect the birth registry book from the health extension worker
before you start home visits and cross-check with answers you receive at household with the
registration book. At the end register the unregistered births before you leave the household.
Informed oral consent

I am (name) ------------------------------- working at -----------------------institution.
The following study is about measuring maternal and neonatal mortality in this area. Specifically
this study is about whether birth registration in this kebele is going well or not. I appreciate
study. The result of the study may help to improve services. We will keep the confidentiality of
all information you give us. Your names and any identifier will not be used publicly. We will
only use the aggregate data. It will take approximately half an hour discussion. Participation is
voluntary but I hope you will be willing to participate as your idea may have important benefits.
Do you have any question? Can I start interviewing?

1) Yes, I can participate
2) No I will not want to participate
If the person approached does not want to participate, give thanks and proceed to the next
household
Kebele………………………………………………….
Sub-kebele (village) _____________________________ Date---------------

2
Name of the head of HH__________________________ sex……….age………..education……
Name of the mother ----------------------------------

The following are the questions to be tick (the format to be filled) in box in front of the answer
(Coding) by asking the head of the household where there is birth since last 12 months.
S. Questions Choice Coding Skip

no
101 Distance of the
household from health
post (km) hint 1 hr walk Write number in km--------------
of normal person= 5 km
102 Is the head of household 1. Yes
model farmer in that 2. No
locality?
103 Do you know that there 1. Yes

is a birth registration in
your kebele? 2. No(I do not know)
104 If yes for Q103 from 1. HEW(health extension worker)

whom you heard?
2. CHP(community health
promoters)
3. Kebele leaders
105 1.Yes
Do you think 2. No
registration of all births 3.I do not know
is important?
106 Have you followed 1. Yes

antenatal care service
during your pregnancy? 2. No
107 How many pregnancies

have you had(gravidity) Write the number………………
108 The level of this birth in (first ,second or------------------)

all your births is…..
109 1.Yes
Do you think this birth
registered in the birth 2.Not registered
registry?
3. do not know
3
110
What is the age of the
newborn today (in days) Write the number of days…………
111 1) Myself visiting the HEW

If your newborn 2) HEW visited our home
registered, how was it
and registered it.
registered
3) CHP( volunteer health
promoters) visited and
registered
4) Other (specify)______
112 1. Since we do not know

If not registered, why do
you think? 2. Since we do not want

3. Nobody asked us
113 If your current newborn

was not registered, do
you have intention for 1 Yes
future registration? 2.No, I do not want
114 Who attended the 1.Parent

delivery? 2.TBA

3.HEW
4.Health works
115 Place of birth

1.Home
2.Health post
3.Health center
4. Hospital
5. Family home

4
116 Is the mother alive now? 1.Yes
2 No
117 If died for Q116 when? 1. During pregnancy
2. During child birth
3. After birth until 42 days

4. After 42 days of birth
118
If the mother died cause 1. Fever
of the death (use 2. Bleeding(retained placenta)

simplified verbal 3. Convulsion
autopsy questions) 4. obstructed/prolonged labour
5. other (specify)--------------
119 Is the neonate (infant) 1.Yes

alive?
2.No
120 If not for Q 119 (dead), 1.During birth

when?
2.After birth until 28 days
3.After 28 days of birth

121 For data collector: Was 1.Yes
the birth registered in
birth registry? 2.No
122
If registered, the date of
registration Date registered………………
123 If registered, write

registration number Serial number on the
book…………………..
Thank you so much for responding

1
Questionnnaire for household survey (Paper II)
House code--------------------------
Date ---------------------------------
Time --------------------------------
Informed Oral consent
I am (name) ------------------------------- working at -----------------------institution.

The following study is about measuring maternal and neonatal mortality in this area. I appreciate your participation. The result of the
study may help to improve services. We will keep the confidentiality of all information you give us. Your names and any identifier
will not be used publicly. We will only use the aggregate data. The interview will take approximately 45 minutes. Participation is
voluntary but I hope you will be willing to participate as your idea may have important benefits.
Do you have any question? Can I start interviewing?
1) Yes, I can participate
2) No I will not want to participate
If the person approached does not want to participate, give thanks and proceed to the next household
2
General backgrounds
1. Name of Kebele-----------------------------
2. Name of Sub-Kebele (Mender) -----------------------
3. Distance of the house from all-weather driveable road ------------km
4. Distance of the house from the nearest health center -------------km
5. Distance of the house from the nearest hospital -------------------km
6. Is there an all-weather driveable road reaching this household? 1- Yes 2-No
Household information
7. Name of head of HH (respondent) ---------------------------
8. Age---------- (years)
9. Sex---------
10. Educational level--------- (completed grade)
11. Occupation (main job) --------------------- 1. Farming 2. Mixed (farming +trading), 3. Trading 4. Employed
12. Number of people living in the house since last 1 year: total __________ Male_____Female______ under 5 year____
3
13. Had there been any birth in this house in the last 5 years (January 2006 to December 2010) HINT: Election 2005
1- Yes 2- No
1. 13. If "yes" to question "9"_______1) yes 2) no births
14. If yes to Q.13, how many to total births?____________________
15. How many live births _____________
16. How many still births_______________
17. How many of these children born alive are still living -------------------
4
18. Fill out the information in the box for each birth that occurred in the household in the previous five years.
Birth Place of birth Who assisted Child Mother
order the delivery?
1: home 1.TBA Month/ye Sex of Type of Live if Mother if died Mother died
2: family 2.HEW ar of the birth : born died, 1) month/yea died where
home 3.SBA birth (if newborn 1) still week alive r during ? ?
3: Health post (nurse, possible) livebirth alive of 2) died (if 1) 1) at
4: health midwife, HO, 2) or death possible) pregnanc home
center Doctor) stillbirth died after y 2) on
5: hospital 1)alive birth 2) child way
st n
2)died (1 ,2 birthing to HI
d
, 3rd 3) in 6 3) HI
,4th weeks
week) after birth
4) after 6
weeks
Birth 1
Birth 2
Birth 3
Birth 4
Birth 5
5
19. What was the main problem that caused the death of the mother? Ask several respondents about the conditions preceding the
death ( husband, sister, any relevant adult), and use the verbal autopsy extraction paper
1- Bleeding (hemorrhage) 2- Fever (sepsis) 3- Convulsion (hypertension) 4- prolonged/obstructed labour 5- Others

(Specify other cause of death)________________________________________________
20. Could the cause be related to abortion? (be skillful on this issue, approach systematically as it is sensitive)
Proceed to the asset (wealth) questionnaire in the following pages

Verbal autopsy extraction for cause of maternal death assessment (Papers I & II)
REF Questions and filters Coding categories
Q 01 How many children had (NAME) 1. Live births______
given birth to when she died? 2. Don't know
Q02 Did (NAME) die during pregnancy 1. Yes
or childbirth or within 6 weeks of giving birth? 2. No
3. Do not know
Q03 Did (NAME) have her periods 1. Yes
coming regularly? 2. No
3. Do not know
Q04 Did (NAME) have bleeding from 1. Yes
the vagina? 2. No
3. Do not know
Q05 How many months was she pregnant when she ________months
died?
Q06 Did she suffer from any complaints during her 1. Yes (specify)
last pregnancy? 2. No
3. Do not know
Q07 Did she attend antenatal clinics during her last 1. Yes (_____times)
pregnancy? 2. No
3. Do not know
Q08 Did (NAME) have high blood pressure during 1. Yes
pregnancy? 2. No
3. Do not know
Q09 Was there bleeding during pregnancy? 1. Yes
2. No
3. Do not know
Q10 Did (NAME) have oedema of the limbs during 1. Yes
pregnancy? 2. No
3. Do not know
Q11 At what stage of the pregnancy 1. During pregnancy
did (NAME) die? 2. During labour
3. In 6 weeks after
birth/abortion
Q12 Was there excessive bleeding 1. Yes
during delivery? 2. No
3. Do not know
Q13 Was she complaining of severe 1. Yes
Headaches during delivery? 2. No
3. Do not know
Q14 Did the placenta come out within half an hour of 1. Yes
the birth of the child? 2. No
3. Do not know
Q15 Did (NAME) have convulsions 1. Yes
during delivery? 2. No
3. Do not know
Q16 Was there high fever starting after delivery? 1. Yes
2. No
3. Do not know
Q 17:
Please discuss deeply with the respondent and describe conditions preceding the death of
the mother in the following free space
_______________________________________________________________________
_______________________________________________________________________
_______________________________________________________________________
_______________________________________________________________________
_______________________________________________________________________
__________________________________________________________________
Indirect sisterhood questions for (Paper III)
The informed oral consent in the household survey questionnaire (Paper II) is also used for
this interview as both studies administered together.
Serial No. Age sex How many sisters have How many of How many How many of these
of you ever had who these sisters of these dead sisters died
respondents were born to your who reached sisters who while pregnant,
mother and who the age of 15 reached the during child birth, or
reached the age of 15 are still alive? age of 15 in 6 weeks after
years? have died? pregnancy ended?
1
10
1
Questionnaire for Paper IV

A Format for Data Collection for a survey on Assessment of Availability and Utilization of Emergency
obstetric care service in Gamo Goffa Zone 2010.
Part one: Facility review
1. Name of facilities___________________________Woreda____________________
1.1 Facilities type 1. hospital 2. health center 3. Others (specify) ______
1.2 Distance from 1. Woreda center__________Km
2. Zone center____________Km
3. from the nearest hospital_______Km
4. from the nearest all-weather road______KM
1.3 Driveable road access to the facility is: 1. Asphalt 2. All-weather gravel road
3. Dry whether gravel road 4. No road access
1.4 Facility status______________1. Basic EMoC 2. Comp EMoC
1.5 Functional status of the Basic EOC facility is 1) fully functional 2) partially functional
3) not functional
2
2. Were the following signal functions (services) performed at least once during the last 3 months & 12
months?
Last3 Last12 (If “yes”), are If not performed

month month medicines/equipmen in past 3 months,
No Items ts available today why?
(1 April (1 July 2009
2010 to 30 to 30 June
June 2010) 2010)
2.1 Parentral antibiotics 1.Yes 1.Yes 1.Yes

2.No 2.No 2.No
2.2 Parentral oxytocin 1.Yes 1.Yes 1.Yes
2.No 2.No 2.No
2.3 Parentral sedatives 1.Yes 1.Yes 1.Yes
2.No 2.No 2.No
2.4 Manual removal of 1.Yes 1.Yes 1.Yes
placenta 2.No 2.No 2.No
2.5 Removal of retained 1.Yes 1.Yes 1.Yes

product 2.No 2.No 2.No
2.6 Assisted vaginal delivery 1.Yes 1.Yes 1.Yes

2.No 2.No 2.No
2.7 Perform newborn 1.Yes 1.Yes 1.Yes
resuscitation(e.g. with bag 2.No 2.No 2.No
and mask)
2.8 Blood transfusion 1.Yes 1.Yes 1.Yes

2.No 2.No 2.No
2.9 Caesarean section 1.Yes 1.Yes 1.Yes
2.No 2.No 2.No
2.10 Others(currtage D &C 1.Yes 1.Yes 1.Yes
2.No 2.No 2.No
E& C 1.Yes 1.Yes 1.Yes
2.No 2.No 2.No
Hysterectomy…. 1.Yes 1.Yes 1.Yes
2.No 2.No 2.No
Note: Please choose 1 or more reasons for ‘’If not performed in past 3 months,(if “NO”” why?’
1. Training issues 2. Supplies, equipment, drugs issue 3. Management issue 4. Policy issues
5. No indication
3
3. For each health facility (fill the following information)
1 July 2009_ 30 June 2010
No Types of services
3.1 Catchment population (use data officially used by the facility)
3.2 Expected number of delivery from catchment population
(With CBR of 32 per 1000 population)
3.3 No of births and pregnancy related admissions
3.4 No of vaginal deliveries
3.5 Caesarean sections performed
3.6 No of total maternal deaths
3.7 No of still births
3.8 No of neonatal deaths
3.9
No of referral
4
4. Number of clinical staffs working in obstetrics in the facility
Is there
24hrs
Staffs During the last 3 month 1 July 2009- 30
services?
June 2010
No (1 April 2009 to 30 June 2010) (Yes/No)
4.1 Doctors(specialist)
4.2 Doctors (generalists)
4.3 Health officers (HO)
Nurses Bsc
4.4 Diploma
Midwives Bsc
4.5 Diploma
4.6 Anesthetics nurses
Total
5
5. Complicated obstetric cases during 12 month period (1 July 2009 to 30 June 2010)
Complicate Months
d obstetric
November
cases
December
Septembe
July 2009
February
October
January
S. Tot
August
March
2010
April
June
May
No al
r
Hemorrhag
5.1
e(APH)
Hemorrhag
5.2
e(PPH)
Prolonged
5.3
labor
obstructed
5.4
labor
Post partum
5.5
sepsis
5.6 Complication
of abortion
Pre –
eclampsia
5.7
or
Eclampsia
Ectopic
5.8
pregnancy
Ruptured
5.9
uterus
5.1
Others
0
Total of the
month
6
6. Maternal deaths from complicated obstetric cases during 12 month period (1 July 2009 to 30 June 2010)
Maternal Months
death from
September
S.No each
November
December
July 2009
February
October
January
complicatio Tot
August
March
2010
April
June
May
n al
Hemorrhage
6.1 (APH)
Hemorrhage
6.2 (PPH)
Prolonged
6.3 labor
obstructed
6.4 labor
Post
partum
6.5 sepsis
Complicatio
6.6 n of abortion
Pre –
eclampsia
or
6.7 Eclampsia
Ectopic
6.8 pregnancy
Ruptured
6.9 uterus
6.10 Others
7
7. What sources of data were used to complete this form?
a. maternal ward register b. delivery book
c. general admission register d. operating theatre register
e. others (specify)_____
8. In your opinion (from taking to staff, the record system etc), what proportion of the complications
treated in this facility is recorded on this form?
a. none b. some(less than half) c. most (more than half) d. all
9. In your informed opinion (from talking to staff, looking at the record system, etc.), what
Proportion of the maternal deaths that occurred in the facility is recorded on this form?
a. none b. some c. most d. all
Thank you very much for your corporation!
10. Date of review_________________
11. Reviewed by (Data collector):
Name______________________Sign._______
N.B to the data collector: Please receive a stamped letter of confirmation from
the head of the facility that describes you have reviewed in the particular facility
Appendices II: Ethical approvals
Region: Saksbehandler: Telefon: Vår dato: Vår referanse:
REK nord 2011/2495/REK nord
31.01.2012
Deres dato: Deres referanse:
13.12.2011
Vår referanse må oppgis ved alle henvendelser
Yaliso Yaya Balla

Overlege Danielseens Hus
Universitetet i Bergen
2011/2495 Measuring Maternal and neonatal mortality in southwest Ethiopia
We refer to form for application for approval of research for this project.
Regional Committees for Medical and Health Research Ethics, Northern Norway (REC north) processed the
application in the meeting January 12 th 2012.
Institution responsible for the research: University of Bergen, by Prof. Rune Nilsen
Chief Investigator: MPhil Yaliso Yaya Balla (PhD candidate)
Chief Investigator's research project description

Introduction: The maternal mortality rate remains high in many countries. Half of maternal deaths in the
world occurred in six countries of which one is Ethiopia. There is need and demand for reliable empirical
evidence to monitor effects of interventions. Objective: The overall aim of the study is to estimate maternal
and neonatal mortality by using different approaches that links the data from community birth registry with
health institution registries and demographic surveys in south Ethiopia. Materials and methods: 1. We shall
use data from population based birth and birth outcome registry by the health extension workers (HEWs). 2.
Institutional birth registry: we shall assess past one year birth and its outcome data in 68 health institutions
in Gamo Goffa. We shall also assess the the quality of obstetric care. 3. Using the sisterhood and household
death survey method we shall collect data from Bonke district (population 173,015) to estimate maternal
deaths in the past years.
Remarks from the Committee:

It is presupposed that approval is given by Ethiopian REC. The project is considered Justifiable and can be
conducted as requested. The Committee also presuppose that other necessary approval from Ethiopian
authority is given.
Decision:
The project is approved.
It is presupposed that the project is approved by other relevant authorities before it is implemented. The
project must be presented before the committee again, if complications or changes to the conditions the
committee had based its decision on, arise during implementation. The committee has to be notified if the
project is not implemented.
The approval is valid until December 20 th 2013. Data collected through the project can be stored until
December 20 th 2018. The project manager is obliged according to the Health Research Act §12 to inform
the committee when the project is completed.
Besøksadresse: Telefon: 77646140 All post og e-post som inngår i Kindly address all mail and
TANN-bygget Universitetet E-post: [email protected] saksbehandlingen, bes e-mails to the Regional Ethics
i Tromsø 9037 Tromsø Web: adressert til REK nord og ikke til Committee, REK nord, not to
http://helseforskning.etikkom.no/ enkelte personer individual staff
Data collected through the project must be treated without names or other recognizable information.
Information and the list linking names and ID-numbers must be stored at separate locations.
The decision of the Committee can be appealed by a part or others with judicial appeal interest in the case
cf. Public administration Act §28. The time-limit of the appeal is three weeks from the time when the party
has been notified of the decision, cf. Public administration Act § 29. The court of appeal is the national
committee for research ethics, but an appeal should be addressed to the Regional Committee for Research
Ethics, Northern Norway.
Letters from REK are approved transmitted electronic without signature.
Sincerely
May Britt Rossvoll

Sekretariat leader
Monika Rydland Gaare

Executive officer
Copy: Rune Nilsen: [email protected]

DR Thesis 2015 Yaliso Yaya Balla PDF

Uploaded by

Copyright:

Available Formats

DR Thesis 2015 Yaliso Yaya Balla PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DR Thesis 2015 Yaliso Yaya Balla PDF

Uploaded by

Copyright:

Available Formats

Maternal and neonatal mortality in

rural south Ethiopia:

Dissertation for the degree philosophiae doctor (PhD)

Dissertation date: June 25 th

I wish to express my deepest gratitude to my supervisor Professor Bernt Lindtjørn. It has

Yaliso Yaya Balla PhD Thesis i

Yaliso Yaya Balla PhD Thesis ii

Introduction: The aim of the Millennium Development Goals (MDG 4 and 5) is to

Yaliso Yaya Balla PhD Thesis iii

Conclusion: Community-based measurement methods (birth registry and household survey)

Yaliso Yaya Balla PhD Thesis iv

List of original papers

Yaliso Yaya Balla PhD Thesis v

CSA Central Statistical Authority

HEW Health extension worker

LTR Lifetime risk

MMR Maternal mortality ratio

MOH Ministry of Health

NMR Neonatal mortality rate

PCA Principal component analysis

REK Regionale Komiteer for Medisinsk og Helsefaglig Forskningsetikk (Norwegian)

SBR Stillbirth rate

SES Socio-economic status

SNNPRS Southern Nations Nationalities and Peoples Regional State

TFR Total fertility rate

UNFPA United Nations Population Fund

UNICEF United Nations Children's Fund

VHP Volunteer health promoters

WHO World Health Organization

Yaliso Yaya Balla PhD Thesis vi

Yaliso Yaya Balla PhD Thesis vii

Yaliso Yaya Balla PhD Thesis viii

1.1 Background on maternal mortality (the problem)

Yaliso Yaya Balla PhD Thesis 1

1.2 Definitions, causes, time, and indicators of maternal mortality

1.2.1 Definitions of maternal mortality

Yaliso Yaya Balla PhD Thesis 2

Table 1: Maternal mortality and complication related definitions

Term Definition Requirement ICD

1.2.2 Causes of maternal mortality

Yaliso Yaya Balla PhD Thesis 3

Yaliso Yaya Balla PhD Thesis 4

1.2.3 Time of maternal death

A conventional definition of maternal mortality considers deaths between pregnancy and 42

1.2.4 Indicators of maternal mortality (statistics)

Yaliso Yaya Balla PhD Thesis 5

A. Number of maternal deaths

B. Maternal mortality ratio (MMR)

Yaliso Yaya Balla PhD Thesis 6

C. Maternal mortality rate (MMRate)

D. Lifetime risk (LTR) of maternal death

E. Proportion of maternal deaths among female deaths (PMDR)

Yaliso Yaya Balla PhD Thesis 7

1.3 Measurement methods for maternal mortality

1.3.1 Measurement backgrounds

Yaliso Yaya Balla PhD Thesis 8

1.3.2 Measurement options (empirical measurements vs. analytical estimates)

Acknowledging a serious demand for information in developing countries where maternal

Empirical methods depend on the collection of first-hand data including: