Report No: AUS13083

Quality and Equity in Basic Health Care Services

in Vietnam:
Findings from the 2015 Vietnam District and Commune Health Facility
Survey

.
June 2016

.
GPV02
EAST ASIA AND PACIFIC
.

Document of the World Bank

.
.

.
Standard Disclaimer:

This volume is a product of the staff of the International Bank for Reconstruction and Development/ The World Bank. The findings,
interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World
Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The
boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of
The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries.
.
Copyright Statement:
.
The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a
violation of applicable law. The International Bank for Reconstruction and Development/ The World Bank encourages dissemination
of its work and will normally grant permission to reproduce portions of the work promptly.

For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470,
http://www.copyright.com/.

All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World
Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail [email protected].
Quality and Equity in Basic Health Care Services
in Vietnam:
Findings from the 2015 Vietnam District and
Commune Health Facility Survey

1
 Acronyms

ALOS Average length of stay

CHS Commune health station
DH District hospital
GRM Generalized Reducible Metrics
HCP Health care provider
HSPI Health Strategy and Policy Institute
IRT Item response theory
MOH Ministry of health
NCD Non-communicable disease
PPS Probability Proportional to Size
SARA Services Availability and Readiness
SDI Service Delivery Indicators
SRHS Self-reported health status
SRS Systematic random selection
TB Tuberculosis
UHC Universal health coverage
US United States
VHLSS Vietnam Household Living Standard Survey
WHO World Health Organization

2
 Acknowledgements

The report was produced by a World Bank team in partnership with the Health Strategy and
Policy Institute (HSPI) of the Ministry of Health of the Government of Vietnam. The main data
underlying the report was collected in a survey designed jointly by the World Bank and HSPI and
collected by an HSPI team.
The World Bank team was led by Gabriel Demombynes (Senior Economist, Poverty and Equity
Global Practice) and Kari Hurt (Senior Operations Officer, Health, Nutrition, and Population
Global Practice). Ha Thi Ngoc Tran (Consultant) led the coordination between the World Bank
and HSPI on survey design and data analysis. Other members of the core team were Benjamin
Daniels (Consultant), Huong Lan Dao (Senior Health Specialist, Health, Nutrition, and Population
Global Practice), Jishnu Das (Lead Economist, Development Economics Research Group), and
Obert Pimhidzai (Economist, Poverty and Equity Global Practice).
The Health Strategy and Policy Institute (HSPI) team was led by Oanh Thi Mai Tran (Director).
Thuy Thu Hoang coordinated the HSPI survey team. Other members of the core team were
Tuan Anh Khuong, Hanh Thi Minh Vu, Thuan Duc Tran, and Van Hong Phan with substantial
contributions from many other HSPI staff.
The project was overseen by Salman Zaidi (Practice Manager, Poverty and Equity Global Practice)
and Toomas Palu (Practice Manager, Health, Nutrition and Population Global Practice), Victoria
Kwakwa (Country Director for Vietnam) and Sandeep Mahajan (Program Leader).

3
 CONTENS

Acronyms ...........................................................................................................................................................2
Acknowledgements ..........................................................................................................................................3
Executive Summary ....................................................................................................................................... 10
Introduction and Overview ......................................................................................................................... 25
Chapter 1: Facility Service Readiness ........................................................................................................ 32
1.1 Introduction ............................................................................................................................................. 33
1.2 General service readiness amenities................................................................................................... 33
1.3. Medical equipment and medication.................................................................................................... 36
1.4. Health staff availability and qualifications .......................................................................................... 40
1.5. Service provision .................................................................................................................................... 41
1.6. Availability of basic services ................................................................................................................. 44
Chapter 2. A profile of health workers ................................................................................................... 48
2.1. Introduction ............................................................................................................................................ 49
2.2. Health workers in commune health stations................................................................................... 49
2.3. Doctors in commune health stations and district hospitals ......................................................... 55
2.4. Doctors’ satisfaction ............................................................................................................................. 63
Chapter 3: Patient experiences .................................................................................................................. 64
3.1. Introduction ............................................................................................................................................ 64
3.2. Patient profile.......................................................................................................................................... 65
3.3. Reasons for Seeking Care at Facility.................................................................................................. 66
3.4. Experiences from Patients’ Perspective ............................................................................................ 68
3.5. Experiences with health insurance ..................................................................................................... 73
3.6. Health expenditure ................................................................................................................................ 75
3.7. Are patients satisfied with healthcare services? .............................................................................. 80
Chapter 4: How knowledgeable is your doctor? An assessment of doctor ability ........................ 82
4.1. Introduction ............................................................................................................................................ 83
4.2. Measuring doctor’s ability .................................................................................................................... 83
4.3. Main findings ............................................................................................................................................ 85
Chapter 5. Knowledge and practice: clinical observation and the know-do gap ............................ 99
5.1. Introduction .......................................................................................................................................... 100

4
5.2. Measuring Effort ................................................................................................................................... 100
5.3. Decomposing the Variance of Effort ............................................................................................... 104
5.4. Comparison of district hospitals and commune health stations ............................................... 107
5.5. Does low effort in district hospitals reflect higher provider knowledge? ............................... 111
5.6. Does low effort in districts reflect higher caseloads? .................................................................. 114
5.7. The quality impacts of low effort: The know-do gap ................................................................... 116
5.8. Putting it together ................................................................................................................................ 119
References ..................................................................................................................................................... 123
Annexes ......................................................................................................................................................... 126
Annex A. Methodology .......................................................................................................................... 126
Annex B. Sampling Design of the Vietnam Household Survey. .................................................... 129
Annex C. Definitions of Selected Indicators in the Report. .......................................................... 132
Annex D. Tables of Medical Vignettes ............................................................................................... 134
Annex E. Additional Tables and Figures ............................................................................................. 140

TABLES

Table O.1: Description of surveyed provinces ....................................................................................... 29

Table 1.2.1: Availability of waste management systems at district hospitals and commune health
stations by province ...................................................................................................................................... 35
Table 1.2.2: General service readiness indicators in Vietnam and Indonesia ................................... 36
Table 1.3.1: Availability of selected equipment at district hospitals by province ............................ 38
Table 1.4.1: The availability of qualified staff at commune health stations........................................ 40
Table 1.5.1: Number of inpatient per doctor per day at district hospitals....................................... 43
Table 1.6.1: Percent of commune health stations that can perform selected services.................. 45
Table 1.6.2: The percent of commune health stations which have at least one patient by case
type in the last year ....................................................................................................................................... 47
Table 2.2.1: Profile of directors of commune health stations by province ....................................... 54
Table 2.2.2: Profiles of directors of district hospitals by province ..................................................... 54
Table 2.3.1: Regression of doctor’s total salary ...................................................................................... 61
Table 3.4.1: Mean travel times from home to facilities for outpatients (in minutes) ..................... 69
Table 3.4.2: Regression of outpatient waiting time ................................................................................ 71
Table 3.4.3: Regression of outpatient consultation time ...................................................................... 72

5
Table 3.5.1: Reasons that inpatients at district hospitals bought medicines outside ...................... 75
Table 3.6.1: Mean health expenditures by patient type and health insurance, in thousands of
Vietnam dong .................................................................................................................................................. 76
Table 4.3.1: Diagnostic accuracy by condition ........................................................................................ 86
Table 4.3.2: Share of doctors prescribing treatment by condition and treatment suitability ....... 88
Table 4.3.3: Share of doctors recommending tests by condition and importance of tests .......... 90
Table 4.3.4: Diagnostic accuracy by doctors’ facility type .................................................................... 93
Table 4.3.5: Correlates of doctors’ ability ............................................................................................... 98
Table 5.2.1. Cross-country comparisons of clinical behaviors across varying effort levels. ....... 101
Table 5.3.1. Cross-province variation in interaction effort and clinical behavior ......................... 104
Table 5.3.2. Effects of patient characteristics on provider effort...................................................... 106
Table 5.4.1 Overall clinical management behaviors and district-commune differences .............. 110
Table 5.4.2. Laboratory diagnostic orders by patient symptom at district facilities ..................... 111
Table 5.5.1. Correlates of interaction effort ......................................................................................... 114
Table 5.8.1. Total cost breakdown of clinical visits ............................................................................. 121
Table AA.1: Outline of the health facility survey questionnaire........................................................ 126
Table AA.2: Sample size information of all modules ............................................................................ 127
Table AB.1: The distribution of sample households according to urban-rural and 6 socio-
economic regions......................................................................................................................................... 130
Table AB.2: The distribution of sample clusters according to urban-rural and 6 socio-economic
regions ............................................................................................................................................................ 131
Table AC.1. Definitions of Health Service Delivery Indicators ......................................................... 132
Table AD.1. Medical vignettes .................................................................................................................. 134
Table AD.2. Answer keys of clinical vignettes for the diagnosis section ........................................ 135
Table AD.3. Answer keys of clinical vignettes for the treatment section ...................................... 136
Table AE.1. Percent commune health stations having set of equipment ........................................ 140
Table AE.2. Percent of district hospitals having selective equipment .............................................. 142
Table AE.3. Percent of commune health stations having selective equipment .............................. 143
Table AE.4. Percent of district hospitals having selective services ................................................... 144
Table AE.5. Percent of commune health stations having selective services ................................... 145
Table AE.6. Percent of district hospitals having selective pharmaceuticals1 ................................... 146
Table AE.7. Percent of commune health stations having selective pharmaceuticals1 ................... 147
Table AE.8. Provider Vignettes IRT Score ............................................................................................. 148
Table AE.9. Provider Average Clinical Effort......................................................................................... 149
Table AE.10. District-Commune Differences - Hawthorne Effect Check ...................................... 149

6
FIGURES

Figure E.1. Private practice by doctors is common among doctors working at public facilities,
particularly district hospitals ....................................................................................................................... 17
Figure E.2. Two-thirds of doctors provided harmful treatment in at least one vignette test
scenario ............................................................................................................................................................ 19
Figure E.3. A higher value of the effort index corresponds to more history questions, more time
with the patient, and a greater number of physical exams ................................................................... 20
Figure E.4. Know-do gap for matched clinical and vignette diarrhea cases ...................................... 22
Figure O.1: Percent of people in each quintile who have used the facility type in the last 12
month ............................................................................................................................................................... 28
Figure 1.2.1: Availability of general service readiness amenities at district hospitals and
commune health stations ............................................................................................................................. 34
Figure 1.2.2: Commune health station access to clean water by province and urban/rural ......... 35
Figure 1.3.1: Average number of available pieces of medical equipment .......................................... 37
Figure 1.3.2: Percentage of commune health stations having all surveyed equipment ................... 38
Figure 1.3.3: Average number of pharmaceuticals available at district hospitals and
communehealth stations .............................................................................................................................. 39
Figure 1.4.1: The ratio of number of nurses and doctors by provinces ............................................ 41
Figure 1.5.1: Caseloads from commune health station clinical observation on the survey day ... 41
Figure 1.5.2: the average number of outpatients per doctor on the survey day............................. 42
Figure 1.6.1: The availability of services at commune health stations by Rural and Urban areas 46
Figure 2.2.1: Frequency of training for health workers at commune health stations..................... 50
Figure 2.2.2: Average monthly total salaries of health workers at commune health stations ...... 51
Figure 2.2.3: Average number of hours per week in private practice by health workers at
commune health stations ............................................................................................................................. 52
Figure 2.2.4: Commune health station health workers’ satisfaction .................................................. 52
Figure 2.2.5: Suggestions of health workers at commune health stations for service
improvement................................................................................................................................................... 53
Figure 2.3.1: Mean age of doctors by province and facility type ......................................................... 55
Figure 2.3.2: Percentage of doctors who are women by facility type ................................................ 56
Figure 2.3.3: Percentage of doctors who are ethnic minorities by facility type, and percentage
who are ethnic minorities in the general population, by province ..................................................... 56
Figure 2.3.4: Percent of doctors with standard university doctor training....................................... 58
Figure 2.3.5: Health staff qualification at district hospitals.................................................................... 59
Figure 2.3.6: Doctors’ clinical experiences in years across different levels of health facilities ..... 59

7
Figure 2.3.7: Training opportunities for doctors at district hospitals and commune health
stations by province ...................................................................................................................................... 60
Figure 2.3.8: Prevalence of private practice by doctors at district hospitals and commune health
stations ............................................................................................................................................................. 62
Figure 2.4.1: Mean satisfaction scores of doctors working at district hospitals and commune
health stations................................................................................................................................................. 63
Figure 3.2.1: Characteristics of outpatients at commune health stations, outpatients at district
hospitals, and inpatients at district hospitals ........................................................................................... 66
Figure 3.3.1: Reasons of patients’ health facility selection .................................................................... 67
Figure 3.3.2: Percent of outpatients rating their health status bad or very bad on the day of visit
to the facility ................................................................................................................................................... 68
Figure 3.4.1: Distribution of travel times from home to facilities for outpatients (in minutes) ... 69
Figure 3.4.2: Outpatients' average waiting time (minutes) ................................................................... 70
Figure 3.4.3: Percent of district hospital outpatients receiving any test or X-ray test .................. 73
Figure 3.5.1: Percent of patients using health insurance ....................................................................... 74
Figure 3.5.2: Percent of patients buying medicine outside the health facilities ................................ 74
3.6. Health expenditure ................................................................................................................................ 75
Figure 3.6.1: Average total out-of-pocket patient expenditures in thousands of Vietnam dong . 76
Figure 3.6.2: Patients' average total expendituresa (in thousand Vietnam dong) by provinces .... 78
Figure 3.6.3: Average medical expenditures to facility by patients with health insurance, by
province ........................................................................................................................................................... 79
Figure 3.6.4: Medical expenditure to facility of children under 6 by provinces (thousand Vietnam
dong) ................................................................................................................................................................. 80
Figure 3.7.1: Fraction of outpatients satisfied with health service during their visit ...................... 81
Figure 4.3.1. Distribution of doctors by number of cases correctly diagnosed .............................. 86
Figure 4.3.2. Median number of history questions asked by condition ............................................. 87
Figure 4.3.3. Median number of physical examination actions by condition .................................... 87
Figure 4.3.4. Number of history question and physical examinations asked by doctors ability ... 87
Figure 4.3.5. Distribution of doctors by number of cases correct or harmful treatment is
prescribed ........................................................................................................................................................ 89
Figure 4.3.6. Average number of tests requested by condition .......................................................... 90
Figure 4.3.7. Share of doctors asking questions by doctors’ ability level .......................................... 91
Figure 4.3.8. Diagnostic accuracy by doctors’ ability levels .................................................................. 91
Figure 4.3.9. Probability of being in a district hospital by doctors ability level ................................ 92
Figure 4.3.10. Share of recommended history and physical examinations by doctors’ facility type
........................................................................................................................................................................... 93

8
Figure 4.3.11. Comparison of distribution of doctors ability by poverty quintile ........................... 94
Figure 4.3.12. Share of doctors making a correct diagnosis and prescribing any correct drugs by
facility area poverty quintile ........................................................................................................................ 95
Figure 4.3.13. Comparison of distribution of doctors ability by province....................................... 96
Figure 4.3.14. Average estimated ability level by individual doctor’s characteristics ..................... 97
Figure 4.3.15. Profile of doctors education by facility type and location characteristics ............... 97
Figure 5.2.1. Variation in clinical behaviors across the distribution of the effort index .............. 102
Figure 5.2.2. Variation in essential checklist completion and medication use by effort index:
diarrhea and cough/cold ............................................................................................................................. 103
Figure 5.3.1. Mean effort index by local poverty rate ......................................................................... 105
Figure 5.3.2. Mean effort index by time of day ..................................................................................... 107
Figure 5.4.1. Probability of a provider being based in a district or commune facility as a function
of demonstrated vignettes ability and average clinical interaction effort ........................................ 108
Figure 5.4.2 Distribution of ability and effort by whether doctors have private practices ......... 109
Figure 5.5.1. Relationship between vignettes ability score and mean clinical interaction effort 112
Figure 5.5.2 Distribution of ability and effort by type of medical degree ....................................... 113
Figure 5.6.1. Total hours worked and number of patients seen per provider .............................. 115
Figure 5.6.2. Time spent with each patient and total workload ........................................................ 116
Figure 5.7.1. Know-do gap for matched clinical and vignette diarrhea cases ................................. 118
Figure 5.7.2. Vignette checklist completion versus clinical checklist completion for matched
providers–diarrhea and cough/cold cases .............................................................................................. 119
Figure 5.8.1. Per-patient salary costs by provider workload ............................................................. 122
Figure AE.1. Average share of recommended history question and physical examinations asked
by doctors ability quintile........................................................................................................................... 141

BOXES

Box 2.3.1: Levels of Doctor Training ........................................................................................................ 57

Box 4.2.1. Medical Vignettes ....................................................................................................................... 84
Box 5.4.1: Do doctors with private practice have different levels of ability or exert different
levels of effort than other doctors?. ........................................................................................................ 109
Box 5.5.1: Do doctors with different medical degrees have different levels of ability or exert
different levels of effort than other doctors?. ....................................................................................... 113

9
 Executive Summary

Vietnam’s grassroots health care system —consisting of commune health stations and district
hospitals — provides nearly all care for the poor and a substantial share of health care services
for all but the wealthiest citizens.1 There are concerns that low quality of health at the grassroots
level may drive patients to seek care at higher levels, driving up out-of-pocket costs and creating
pressure to build more higher level facilities. However, in the absence of systematic evidence,
discussion of health care quality in Vietnam has relied largely on anecdote and partial evidence.
This study helps fill that gap, drawing extensively from an innovative survey of commune health
stations and district hospitals. The survey was designed to be statistically representative of patient
experiences in six provinces drawn from six separate regions, in order to provide a portrait
broadly reflective of the state of the grassroots health care system in Vietnam as a whole. In
addition to providing information about facility readiness, doctor characteristics, and patient
experiences, the resulting data allows for analysis of several different measures of doctor quality,
including 1) a measure of doctor “ability” based on responses to a series of vignette scenarios
testing their knowledge, 2) an “effort” index based on direct observations of time, questions, and
diagnostic examinations per patient, and 3) an assessment of the correctness of practice observed
in direct observation of treatment of simple conditions. Key findings and related policy questions
include the following:

Although the system exhibits a number of weaknesses in terms of quality, the grassroots
health system provides remarkably equitable care. Services and experiences of patients do
not vary greatly by socioeconomic background as much as in other countries. Differentiation
does exist, however, in who accesses the grassroots system. Very few patients in the
wealthiest quintile visit commune health stations, and only rarely do the poorest patients
access provincial and national hospitals, which are above the grassroots system.

Facilities generally have the basic infrastructure, staffing, and equipment needed to provide
quality care, but there are gaps. Many commune health stations in Dien Bien lack clean water,
and greater use of information systems could help improve service delivery at most facilities.
Many facilities lack essential medicines, leading patients to buy medicines outside the facility
at higher cost. Additionally, the low nurse to doctor ratio highlights the need to understand
why more nurses are not being hired by hospitals.

Doctors in district hospitals on average are much more knowledgeable than commune health
station doctors, and the knowledge gap is largely attributable to their education background.
At the same time, in vignette tests a significant share of doctors at both district hospitals and

1
This study concerns the curative care system. District-level preventative care health centers, which are also part of
the grassroots health care system, are outside the scope of this study.

10
commune health stations prescribed unnecessary and potentially harmful treatments. Possible
approaches to address these knowledge gaps include clinical training and other support tools,
greater exchange or supervision between the district and commune doctors, and
standardizing the minimum entry level education competencies.

Doctors in district hospitals on average exert far less up-front examination “effort” than
commune health station doctors. This pattern is not explained by either greater efficiency or
greater time constraints of hospital doctors. Hospital doctors appear to substitute testing
for asking questions and performing physical exams. For simple cases of cough/cold and
diarrhea, doctors at commune facilities are more likely to provide correct treatment than
doctors at district hospitals, despite their lower level of ability. At both types of facilities,
however, many doctors were observed to give unnecessary or potentially harmful treatments,
mirroring the findings from vignette tests. This raises questions about the potential for
commune health stations to be a preferable source of treatment for basic primary health care,
the need to ensure that district hospitals do better at the basics while their focus is on higher
level services, the need for clinical practice and facility-based incentives to put greater
emphasis on initial examination and less on laboratory tests, and the need for quality
assurance checks to ensure that appropriate treatments are followed.

Doctors prescribe an average of 3 medicines per visit, including at least one antibiotic in 45%
of visits at commune health stations and 39% of visits at district hospitals. At district hospitals
doctors order at least one laboratory test for 46% of patients. Patterns of testing and
prescriptions for particular conditions suggest that these rates are excessive. This should be
of concern due to growing antibiotic resistance, the impact on out-of-pocket costs of the
patients, and the health insurance expenditures that could be used for other purposes.

In terms of facility capacity, availability of medicines, and doctor knowledge for diagnosis and
treatment, the grassroots health system performed less well in terms of being able to address
the growing health needs on managing non-communicable diseases.

Off-hours private practice by doctors who work at district hospitals is very prevalent. Thirty-
eight percent of hospital doctors perform some private practice work, and the average time
per week in private practice across all hospital doctors is 11 hours. Doctors in district
hospitals exert lower effort on average and are much more likely to engage in private practice.
It is not clear, however, if some doctors may lower their effort in public facilities in order to
attract patients to their private practice. A better understanding of the private practice
phenomenon could be generated by further analysis based on 1) quantitative analysis of
patient motivations that could be done by matching the facility survey to the household health
survey data that was collected in parallel, and 2) additional qualitative interviews of doctors
and patients.

11
1. The grassroots health care system is the main entry point of the population
into the public health care system and the dominant health care provider for the
lower income population. The network of grassroots health care providers is extensive with
more 11,000 commune health stations and 620 district hospitals, essentially reaching every
administrative jurisdiction of the country. The function of the commune health stations are
evolving from a legacy of providing community public health functions and maternal child health
services to one providing a more comprehensive prevention and curative services for the families
in their community. This includes the need to address changing health needs including rising non-
communicable disease burden and a rapid aging society. District hospitals provide primary
curative services, diagnostic services, and basic inpatient services including for internal medicine,
obstetrics/gynecology, pediatrics and surgery. District hospitals are also evolving providing
increasingly more complicated services due to growing access to technology and more educated
human resources. While under the management of the provincial government authorities--and
certainly there are some variation in resources, capacity, organization and financing around the
country--the basic legislation, regulations and incentive structures driving the performance of the
system are set at the national level by the Ministry of Health and increasingly by the national
health insurance system. The harmony of these central level policies and the implementation of
those policies by the provincial authorities are necessary to see improvements in overall
performance.
2. Concerns have often been raised about the quality and equity of basic health
services provided by the grassroots health care system in Vietnam. Although the
country has made vast progress along a number of measures of health outcomes measured at the
national level, particular deficiencies remain, such as the relatively high rate of infant mortality
among the ethnic minority population—44 per 1000 live births as compared to 10 per 1000 live
births among the Kinh and Hoa ethnic groups. Such gaps are suggestive of possible equity gaps in
the quality of care. Additionally, there are concerns that the system is poorly equipped to provide
quality care to address the changing burden of disease of Vietnam’s rapidly aging society. Assessing
health quality directly is challenging, and a systematic assessment of the quality of care—as
opposed to outcomes—has not previously been conducted in Vietnam (1).2 Critical questions
include 1) do the poor receive substantially worse care than the better off? 2) are quality
deficiencies associated with lack of knowledge or failure to apply knowledge in practice? and 3)
what are critical areas where quality could be improved?
3. In order to assess quality in the grassroots health care system, a new survey
was conducted in 2015 at the facility level. The survey was carried out by the Health
Strategy and Policy Institute of the Ministry of Health in partnership with the World Bank. The
survey was designed to be representative of six provinces in six distinct geographical regions. The
provinces include Dien Bien, which has a large ethnic minority population and is one of the
country’s poorest provinces, as well as Hanoi, one of the wealthiest areas in the country. The

2
The 2001 National Health Survey did collect information on health facilities and simple measures of doctor
knowledge but did not include direct observation of doctor behavior like this study.

12
four other provinces (Binh Dinh, Dak Lak, Dong Nai, and Dong Thap) were selected because
they have socioeconomic characteristics typical of their respective regions. Information was
collected from a representative sample of commune health stations and district hospitals as well
as patients who use those facilities. Elements of the information collected in the study include
the availability of key inputs (infrastructure and medicines) at the facility, patient experiences, the
qualifications and experience of doctors, the knowledge of doctors, and the actual practice of
doctors as recorded in direct observations of clinical practice.
4. The study has two parts. The first half of the study provides a description of various
characteristics of the health care experience: 1) general service readiness in terms of
infrastructure and basic service ability, 2) aspects of the experience from the patients’ point of
view, and 3) the characteristics of healthcare workers. The second half of the study explore three
measure complementary measures of the quality of care. These include a) a measure of doctor
“ability” based on responses to a series of vignette scenarios testing their knowledge, b) an
“effort” index based on direct observations of time, questions, and diagnostic examinations per
patient, and c) an assessment of the correctness of practice observed in direct observation of
treatment of particular conditions. While each of these measures has weaknesses, collectively
they provide a picture of the quality of services delivered. The equity of quality of service delivery
is considered by comparing these three quality measures and other aspects of health care
between district hospitals and commune health stations, between urban and rural facilities,
between care for the poor and non-poor, and between care for ethnic minorities and members
of the Kinh and Hoa ethnic majority.
5. Grassroots level facilities typically have much of the infrastructure required
for delivery of basic health services, although there are important. All surveyed health
facilities have electricity, almost all have toilet facilities, and most have waste water and solid
waste treatment systems. However, 24% of the commune health stations in Dien Bien province
lack a source of clean water. A significant number of district hospitals do not have important
equipment such as anesthesia equipment, child ventilators, electrocardiograms, and blood glucose
analyzers, indicating difficulty in providing emergency response or analyzing newly emerging
diseases such as diabetes. Nearly all facilities have computers and internet access, but only 22%
of district hospitals and almost no commune health stations exploit those tools by using any form
of information management system. The implementation of information management systems
may require policies standards, regulations, and subsidies to increase uptake.
6. The lack of certain essential medicines at the facilities indicate both a
replacement of these medicines with other, sometimes more expensive drugs, and
lack of availability sending patients to purchase drugs outside the facility. The
assessment checked on the availability of a list of 30 essential medicines for the treatment of
primary health care conditions at both district hospitals and commune health stations. On average
district hospitals had half of the medicines on hand, and commune health stations had one-third.
Facilities often lacked the basic medicines for treatment of these conditions but had more
expensive medications available. In exit interviews, a significant share of patients reported that
they were prescribed medicines not available at the facility itself. Survey participants who had

13
purchased medicines outside the facility were asked why they had done so. The most common
reason was that the prescribed drug was not available in the facility for one of two reasons. In
49% of cases, patients responded that the medicine was not on the list of medicines covered by
insurance, and in 15% of cases they indicated it was on the insurance list but out of stock. Given
that the share of out-of-pocket payments out of total health expenditures has remained flat in
Vietnam despite growing health insurance coverage (2), these findings suggest that the out-patient
drug list reimbursed by health insurance should be reviewed. The findings also point to a need
to review the hospital formulary and as well as tools and incentives to standardize prescription
practices.
7. Few commune health stations are provide services to non-communicable
disease patients, even if they report themselves equipped to provide those services.
The study reviewed the number of commune actually providing a list of tracer services. Except
in Dong Thap, very few commune health stations had dispensed diabetes medication or
hypertension medicines. These findings reflect the historical role of commune health stations,
which have traditionally focused on maternal and child health and are less equipped and have
largely not been given the mandate, according to policy and regulations of the health sector, to
address the growing profile of non-communicable disease. In most locations, commune health
stations are limited to being able to describe any medication beyond a few days required for acute
care treatment and need to refer patients to the district hospital for diagnosis as well as on-going
management of a chronic disease. Further, it is not in the financial interest of district hospitals
to refer patients for on-going treatment back to the commune health stations.
8. The health care experience as reported by patients does not differ sharply by
socioeconomic group. Ethnic minorities and the poor are slightly more prevalent at commune
health stations than district hospitals. Travel times for ethnic minorities and the poor to district
hospitals are higher on average than those of other patients, and they may tend to seek care at
district hospitals only when their health status is substantially worse. However, travel times were
under 20 minutes for nearly all patients at commune health stations and most at district hospitals.
However, it is important to note that the facility survey only captures the population of facility
users. Potential users living in the facility catchment area but not using the facility—who may live
on average farther away from the facility—are not captured in the facility survey. The health
access of the remote population groups, which do have the worst health indicators, should be
further explored using the household health survey which was collected in parallel to the facility
survey.
9. One difference by socioeconomic status is in wait times at commune health
stations. Among outpatients at district hospitals, waiting times are invariant to patient
socioeconomic status, averaging close to 33 minutes for all patients across wealth levels. Poorer
patients at commune health stations, however, do wait substantially longer than wealthier
patients, and this difference persists even after controlling for facility effects. In other words this
pattern is not due to poorer patients going to facilities with longer waits. Even the poorest quintile
of patients, however, face fairly moderate wait times—averaging 20 minutes. It is unclear why
wait times would vary with socioeconomic status.

14
10. Patients expressed high rates of satisfaction with their care in exit interviews.
Among hospital outpatients, 72% said they were satisfied or very satisfied with their care, as did
85% of commune health station outpatients. Satisfaction rates were lower for district hospitals
than for commune health stations in every province. Service satisfaction data is difficult to
interpret because it reflects a combination of the care itself and the patient’s perceptions and
expectations. Hanoi district hospital patients expressed relatively low satisfaction rates, which
may be reflect expectations driven by the availability of higher quality care at national and private
hospitals in Hanoi. Satisfaction rates tend to be useful when used by a health facility manager
over time to look at changes over time. According to regulations, hospitals in Vietnam are
supposed to measure patient satisfaction rates but few do it consistently or well.
11. Average expenditures are low for outpatients with insurance at both
commune health stations and district hospitals. Total out-of-pocket expenditures
associated with the facility visit or stay were calculated by summing three categories: i)
expenditures to the facility, ii) expenditure for medical services outside the facility, and iii) gifts,
food, travel costs, and lodging for the patient and relatives, other than costs paid directly to the
facility. Inpatients costs, for both those with and without insurance, were substantial. The largest
component of expenditure for inpatients with insurance was gifts, food, travel costs, and lodging.
The average lengths of stay for inpatients at district hospitals of 6.2 days is quite long, particularly
for the type of basic services offered at district hospitals. The long length of stay is one factor
that drives up out-of-pocket costs. The Government has tried to target these out-of-pocket
costs through cash payments to the poor to cover travel and food expenses. This program is
not universally available in all provinces or even in most provinces. The coverage of this program
as well as clinical practice and incentives to reduce the average length of stay should be reviewed.
12. Doctors have lower levels of qualifications in commune health stations,
particularly in the poorest areas. Doctors at district hospitals are younger but with higher
levels of qualifications than those at commune health stations. All district hospitals are staffed by
fully-qualified doctors, as are 80% of commune health stations. Commune health stations without
fully-qualified doctors are highly concentrated in Dien Bien, the poorest province in the survey.
In Dien Bien, just 24% of commune health stations are staffed by at least one fully-qualified doctor.
The remainder are staffed by assistant doctors, who have a lower level of qualification. Assistant
doctors were trained under an earlier training regime which has been phased out. 3 Most fully-
qualified doctors at commune health stations do not have standard university doctor training but
instead have been promoted from assistant doctor through “twinning” training programs or
pursued degrees through less competitive “direct entry” programs. Given the older age of
doctors at the commune level, that most of these doctors were previously assistant doctors, that
fewer pathways to become a doctor exist now, and that younger doctors prefer to work at

3
Throughout the report, except where otherwise noted, the terms “doctor” and “health care providers” are used
to refer to the set which includes both fully-qualified doctors and assistant doctors.

15
hospitals, the system may face challenges over the long term in finding doctors to staff commune
health stations after older doctors face retirement.
13. Work by doctors in private practice outside of their jobs at public facilities is
common. Thirty-eight percent of hospital doctors perform some private practice work, and the
average time per week in private practice across all hospital doctors is 11 hours. The prevalence
of private work and average private hours is slightly lower at commune health stations (31% of
doctors, averaging 9 hours per week). Private practice is especially common in the southern
provinces of Binh Dinh, Dong Nai, and Dong Thap. Very few doctors in Dien Bien conduct
private practice, and private practice is much less common for other health care workers. It is
not clear to what extent this is driven by demand for off-hours or care at home by patients, or
whether it is because doctors direct patients to their private hours to earn additional income.
Dual practice can be a concern if it results in shirking or lower effort while the public doctor is
performing their public duties and shifting of patients to their private practice where they perform
more effort. The issue could be explored through further qualitative work as well as analysis
matching the facility survey to a parallel household survey. Dual practice work by doctors is
believed to be common in other countries in South and East Asia, but hard data on its prevalence
is scarce (3). The studies that have examined the issue in other countries are typically very dated
or rely on non-representative samples and qualitative interviews.

16
 Figure E.1. Private practice by doctors is common among doctors working at
public facilities, particularly district hospitals

Percentage of doctors working at district Average number of hours per week that
hospitals and commune health stations district hospital and commune health station
performing private practices doctors spend on private practices

Urban 14% Urban Hanoi 3

8
Hanoi 38%

Rural 21% Rural Hanoi 7

6
Hanoi 22%

26% Dong Nai 9

Dong Nai 53% 17

53% Dong Thap 17

Dong Thap 60% 19

35% Binh Dinh 10

Binh Dinh 55% 17

38% Dak Lak 6

Dak Lak 40% 12

0% Dien Bien 0
Dien Bien 10% 3

0 5 10 15 20

Commune District Commune District

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

14. Doctors’ knowledge was assessed through vignette tests. The vignettes were
developed in close consultation with Vietnamese medical experts, following Ministry of Health
clinical guidelines and standards. (These vignettes are specific to Vietnam and have not been
internationally validated.) The tests gauged (a) doctors’ knowledge of clinical guidelines and
standards of procedure in terms of history taking, physical examinations and laboratory tests, (b)
their ability to process patient information from the guidelines to reach a correct diagnosis, and
(c) their knowledge of appropriate treatment of cases. They covered a wide range of common
conditions (infectious diseases and non-communicable diseases) involving diverse types of
patients (female and male, children and adults as well as elderly). The test were given to more
than 1000 doctors, undertaking 5 different vignettes each. The focus of this assessment is on
what doctors know, which represents the best outcomes one can expect from doctors if they
were to manage cases to the best of their knowledge.
15. Doctors were categorized across a spectrum of “ability” based on their
knowledge of what to ask, what physical exams to perform and what diagnostic
exams to request in order to reach a diagnosis and treatment plan. An aggregate
indicator of doctors’ ability was generated using item response theory (IRT) to produce a ranking

17
or distribution of doctors by levels of ability. The IRT methodology uses maximum likelihood
methodology to estimate the underlying "ability score" of providers based on their performance
during the medical vignettes exercise. This score reduces the history questions and examinations
behavior to a single metric that is comparable across providers. It quantifies their propensity to
ask the history questions and perform the physical examinations that were graded as minimum
or essential by the expert committee.
16. Doctors are generally able to interpret information and reach a correct
diagnosis using information generated by following clinical guidelines, but their
knowledge of these guidelines is limited. Most doctors can give an accurate diagnosis once
they have the necessary patient history and physical examination information typically generated
from following clinical guidelines. Four of the five cases tested in this study were each correctly
diagnosed by more than 70 percent of doctors after they were presented with essential patient
history and physical examination information. The exception was acute diarrhea, which an
overwhelming majority of doctors (81 percent) could only partially correctly diagnose.

17. The challenge for doctors was in their low knowledge of the right history
questions to ask and physical examinations to perform in accordance to the clinical
guidelines. On average, doctors asked less than half of the essential history questions in each of
the 5 cases presented to them. Only for acute diarrhea and hypertension did the average doctor
carry out at least 50 percent of the necessary physical examinations. In this respect knowledge
of clinical guidelines is modest. Doctors who knew more the appropriate questions and physical
examinations were more likely to give an accurate diagnosis and know the correct treatment.
18. In the vignettes tests, a considerable number of doctors also offered harmful
treatment, even after being provided knowledge of the diagnosis and patient history
and physical exam information. This applied to doctors at all levels of the ability scale,
highlighting weaknesses in the knowledge of case management protocols as even a significant
share of doctors most knowledgeable about the clinical guidelines at earlier stages of the
consultation process offered harmful treatment. Harmful treatment in practice can result in part
from doctors who know the correct treatment responding to incentives to offer other
treatments. But the vignettes show that harmful treatment results in part from knowledge gaps,
which could be addressed through improved training and clinical aids. Most doctors have had
some continuous medical education, so it is not clear if the training targets clinical guidelines for
the most common primary conditions.

18
 Figure E.2. Two-thirds of doctors provided harmful treatment in at least one
vignette test scenario
60%

50%
Share of doctors

40%

30%

20%

10%

0%
Harmful treatment Fully correct treatment Supplemental treatment
Number of cases where type of treatment is proposed
None 1 case 2 cases 3 cases 4 cases All 5 cases
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

19. High ability doctors—as identified by the vignettes—are more likely to be

found in district hospitals, Ha Noi and Binh Dinh and in richer areas compared to
poor areas. Differences in the ability of doctors across facility types or location are significantly
attributable to differences in the education profile of doctors in these locations. The estimated
ability of doctors with intermediate training is less than a fifth of the estimated ability of doctors
with a medical degree, and their particular weakness was in the diagnosis and treatment of
diabetes type II. This implies that training and the provision of supportive clinical aids needs to
be targeted to those more difficult provinces, which also have less resources to undertake an
effort like this by themselves. This would be consistent with Ministry of Health’s on-going efforts
that have training programs targeted specifically to the more difficult provinces.
20. Separate from knowledge or ability, doctor “effort” was assessed based on
direct observation of care of actual patients. An “effort” index was calculated using a
composite index based on three variables collected during the survey: consultation time, the
number of questions asked, and the number of physical examinations completed. Both studies
from other countries using standardized patient techniques and a comparison to actual treatment
demonstrate that this measure of effort is a strong predictor of correct treatment. An important
caveat to this analysis is it is possible that due to the "Hawthorne" effect, doctors may exert more
effort because they are being observed. However, previous studies using direct observation of
doctors have found that the Hawthorne effect drops off quite quickly and disappears entirely
after 4-5 patients, suggesting that the effect has minimal effect on the analysis of effort based on
observations over a full day (4).
21. Patterns of the subcomponents of the effort index in Vietnam are broadly
similar to those in other developing countries for which data is available. Health care
providers at grassroots facilities in Vietnam on average spend 5 minutes with patients, ask 7
questions, and complete 2 examinations. This level of effort is higher than what is observed in
19
India, but lower than in similar studies in Paraguay and Tanzania. As shown in the following
graph, those doctors who exert low effort are spending just a few minutes and performing almost
no physical exams. Doctors who exert higher effort spend much more time with patients and ask
many more questions. The average number of drugs prescribed does not vary with the effort
level of doctors.4

Figure E.3. A higher value of the effort index corresponds to more history
questions, more time with the patient, and a greater number of physical exams

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: Box plot indicates 5th, 25th, 50th, 75th, and 95th percentiles of the effort index.

22. Poor patients do not receive lower effort than better off patients. Effort is
essentially identical comparing facilities in wealthier vs. poorer areas, comparing facilities that
receive more vs. fewer patients, and comparing poor vs. non-poor patients visiting the same
doctor. This remarkable result stands in stark contrast to other developing countries for which
similar data are available, where typically large gaps are found in effort for the care provided to
poor as compared to better off patients.
23. District hospital doctors exert substantially less effort, as measured by the
index, than doctors at commune health stations. Total time spent, questions asked and

4
It is not possible to compare the ability measure across countries, because it is based on vignettes which were
designed to correspond to the Vietnam-specific standards of treatment. Likewise, the analysis of quality of care for
two tracer conditions (based on the direct observation of treatment) is based on Vietnam-specific standards of
treatment and thus cannot be directly compared to similar studies in other countries.

20
examinations performed are all significantly lower in district hospitals, although doctors in district
hospitals offer far more tests. Given the importance of this result, two hypotheses were tested.
One possibility is that this “effort deficit” arises because doctors in district hospitals are more
knowledgeable, and hence can exert lower effort without adversely affecting quality. In reality,
the data show precisely the opposite—when district hospitals and commune health posts are
looked at separately, more knowledgeable providers always exert higher effort. Consequently,
controlling for knowledge implies that effort levels are far lower in district hospitals than what
they should be given their levels of provider knowledge. A second possibility is that doctors in
district hospitals exert lower effort because they face heavier caseloads and thus have less time
available per patient. Caseloads (patients per doctor per day) are very low in commune health
stations and much higher in district hospitals. However, 80 percent of doctors in district hospitals
either see fewer than 60 patients a day (which would be a five-hour daily caseload at 5 minutes
per patient). The average provider in a district hospital sees patients for only 2.7 hours a day,
compared to 1.1 hour at commune health stations. Most providers at both levels do not appear
to have a caseload that should adversely affect their effort levels.
24. Medication use is at similar high levels in district hospitals and commune
health stations, with high use of antibiotics in all facilities. On average a patient is
prescribed three different medicines per visit, including at least one antibiotic in 45% of visits at
commune health stations and 39% of visits at district hospitals. At district hospitals doctors order
at least one laboratory test for 46% of patients. Patterns of testing and prescriptions for particular
conditions suggest that these rates are excessive. These patterns suggest that doctors at district
hospitals are substituting tests for medical history and physical examinations, even when the tests
are not necessary for the indications presented.
25. The quality of care was also evaluated using direct observation of actual care
for patients with two specific tracer conditions. This method is a complement to the other
two quality of care measures used in this study--the ability measure derived from the vignette
tests and the effort measure calculated based on direct observation for all patients (regardless of
conditions). The two conditions tracer conditions—cough/cold and diarrhea—are common and
easily identified in direct observation, with clear guidelines on proper care. For the simple tracer
conditions where we have clear metrics of quality (cough/cold and diarrhea),
26. For these two common tracer conditions, commune facilities provide higher
quality care than district hospitals. For diarrhea 94% of all doctors said in vignette tests that
they would give oral rehydration solution (ORS) (a correct treatment). In practice 40% of
commune doctors actually gave ORS to real patients and only 27% of district hospital doctors
did so. In district hospitals, a large number of doctors responded in vignettes that ORS patients
should receive antibiotics (an incorrect treatment) and an even larger number (36%) actually gave
them in practice. Since doctors in district hospitals are far more knowledgeable, the gap between
knowledge and practice is very high for district providers—a pattern that is observed in the data
by comparing provider’s clinical actions with outpatients to their own vignette responses.

21
 Figure E.4. Know-do gap for matched clinical and vignette diarrhea cases

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: The red bars show doctor actions from clinical observations from actual care of diarrhea cases. Blue bars
show responses to vignettes for the diarrhea case.

27. The findings in this report suggest both widespread deficiencies in knowledge
and a wide gap between what doctors know and do. Knowledge gaps are pervasive —
particularly in the failure to follow clinical guidelines in collecting patient history and physical
examination information and in knowing what not to do. The “know-do” gap can be separated
into a “know-can” gap and a “can-do” gap. The “know-can” gap expresses the difference between
what’s doctor’s know to do and what they are able to do, given available tools at the facility. The
“can-do” gap expresses what they are able to do and what they actually do in practice. The
“know-do” gap described in this report is associated with simple conditions, for which doctors
have the necessary tools at their disposal, which the margin for improvement is largely in terms
of “can-do.” At the same time, the lack of essential medicines and access to testing indicates that
the “know-can” gap is also substantial.
28. A number of different approaches could help address the combination of
knowledge deficiencies. One key starting point is to improve and standardize the knowledge
and skills of the new doctors entering in the system. This will require clearing defining what a
doctor graduating with a basic medical education degree should know and do, an education
system that teaches and trains well on basic primary health care, a quality assurance system to
ensure that it happens. Online training and supportive clinical aids could also help doctors follow
clinical guidelines.

22
29. Consideration of a separate set of issues is needed to address the “know-do”
gap. These include the “know-can” gap associated with the lack of access to laboratory tests
and the limitations in medication access. Also, important are the expectations of the population
that may expect a diagnostic procedure or an antibiotic as a sign of quality of care. Additionally,
there is the lack of tools, such as information systems or simple clinical check lists that would
prompt physicians to provide correct treatment. A final critical issue for consideration is the
incentives faced by physicians in hospitals to provide more lab tests, more x-rays and more
ultrasound examinations due to the hospital payment mechanism and the incentive to increase
revenue.
30. Overall, the report shows that Vietnam’s grassroots healthcare system
exhibits remarkable overall equity. Direct observation of doctor performance addressing
two of the most basic curative health conditions shows that the actual quality is no worse at
commune health stations than at district hospitals. This is despite the fact that doctors at
commune health stations have lower qualifications and on average are less knowledgeable,
particularly in the poorest areas. It is likely that this reflects the higher effort exerted by commune
health station doctors.
31. The findings regarding ethnic minorities present a mixed picture. Ethnic
minorities represent 14% of the population and the majority of the poor in Vietnam. The survey
covered two provinces with substantial ethnic minority populations: Dien Bien, where they form
81% of the population, and Dak Lak, where they comprise 29%. In Dien Bien, both district
hospitals and commune health stations have much lower levels of infrastructure than in other
provinces. A large number of doctors in both Dien Bien and Dak Lak are ethnic minorities
themselves, who have been trained through special programs directed at ethnic minority and
remote populations. The ability level of ethnic minority doctors is substantially lower than that
of other doctors, and as a result, patients in those provinces (largely ethnic minorities themselves)
tend to receive care from lower-ability doctors. On average, ethnic minorities tend to receive
care characterized by higher effort, and there is no evidence of discrimination against ethnic
minorities in the care they receive. In some, ethnic minorities receive care that is typically
characterized by lower quality infrastructure, lower ability doctors, and greater effort. Apart from
more general efforts to improve doctor knowledge and effort, targeted efforts to improve
infrastructure in provinces like Dien Bien and Dak Lak could improve the quality of care for ethnic
minorities.
32. The study has various limitations. First, it only covers the grassroots health care
system, and does not consider provincial and central hospitals, where the wealthiest Vietnamese
more often seek care and where quality is likely to be higher. Second, although the survey is
representative of six provinces spread across regions, in order to provide a portrait broadly
reflective of the state of the grassroots health care system in Vietnam, the data is not statistically
representative of the entire country. Third, each of the methods employed to measure aspects
of quality has weaknesses and thus presents a partial view. Vignettes only capture knowledge and
not actual practice. The measure of effort cannot control for the different mix of cases that may
be found across doctors and facilities. And analysis of care based on direct observation can only

23
be conducted for common, easily identifiable cases and consequently is not informative about the
quality of care for more complex and less common cases.
33. Following the technical findings of this report, a series of companion policy
notes will be developed between the World Bank and the Health Strategy and Policy
Institute (HSPI) in consultation with the Ministry of Health and other stakeholders.
The results raised several critical policy issues. First, what are the reasons for lack of information
systems development for grassroots providers? Second, what are the reasons for the shortages
in essential medicines, whether they are replacement with more expensive options, problems
with the procurement process, and lack of incentives for the district hospital to adequately stock
the commune health station? Third, while ability of the doctors is highly correlated with
education, what is the future of commune health station doctors given the change in the education
path, age of the commune health doctor, and current low productivity? Fourth, is private practice
only a matter of convenience for the patients or are providers reducing their effort in public
facilities to send patients to their private office hours? Fifth, why are doctors, particularly district
hospital doctors, performing below their ability? Sixth, what factors explain the suggested pattern
that testing is overused at district hospitals and that medicines—particularly antibiotics—are
over-prescribed at both hospitals and commune health stations?
34. The study also points to areas for future analysis which could address many of
these questions. First, a follow-on study using the technique known as standardized patients
could provide further insights into the quality of care for particular treatments and overcome the
weaknesses of each of the methods employed for this study. Second, a better understanding is
needed of why effort is low at district hospitals and what could be done to boost effort.
Qualitative work could lend insights on this question. Third, a weakness of this study is that it
was unable to collect information on insurance payments, which would make possible an analysis
of total costs (rather than just costs to the patient), and this information could be collected in a
future study. Analysis of such information would make clear the cost implications of the extensive
use of testing and prescription of medication and could point to potential areas of cost savings.
Fourth, it would be informative to match the results of this study with that of households
measuring the health seeking behavior of the population to determine what factors may
contribute to lower or higher utilization and choice of providers, including decisions for seeking
care by public doctors in their private practice. This last line of analysis could be undertaken
without any additional data collection, as a household health utilization survey was conducted in
parallel to the 2015 Vietnam District and Commune Health Facility Survey.

24
 Introduction and Overview
35. This report examines quality of basic health care services and the potential
differences in quality within the grassroots level in Vietnam.5 It is motivated by the
growing concern about the quality of public services generally and the Government’s immediate
policy question as to what it would take to strengthen the delivery of services by health care
providers (HCPs) at the lower tiers of the health system for more efficiency and effectiveness of
the health system overall. It is also motivated by the limited information available to assess actual
quality gaps on any large scale basis (1; 5). The Ministry of Health’s priorities in this area are
outlined in the recently adopted Health Sector Development Plan for 2016-2020 (6), and a
process is underway to reach consensus on the policy and investment choices to reach these
goals. The analysis and findings of this study are intended to inform those discussions and raise
questions for further research.
36. The World Bank has highlighted the need to rebalance Vietnam’s health
service delivery model towards the grassroots health system, in order to respond to
changing health needs, rapid aging, and rising expectations of a growing middle class.
Vietnam will need to adapt its service delivery system at all levels through better policy/regulation,
more value-driven purchasing of health services, and targeted investment strategies (7). The
changing population health needs will demand more high technology and specialized health
services, but that will only be the tip of the health service needs. However, the detection and
management of the chronic conditions, meeting the health needs of elderly close to their homes,
providing timely and responsive outpatient services to a demanding middle class will largely
depend upon the grassroots health system providers as well as private sector providers that are
for the most part absent in Vietnam, except in more urban locations. Therefore, policies and
investments that would improve the quality of basic health services at the grassroots level while
not sufficient are needed. Without being able to measure, support and motivate improvement
and assure the quality of the health services, the population will be left with other signals of quality
such as the availability specialized doctors, high cost technology and availability of expensive
medication. Unfortunately, this would continue to pull Vietnam towards a high-cost health
system and negatively impact its ability to use its scarce health resources to expand public- or
insurance-financed health services to meet the other needs (such as rehabilitation services for
the elderly, drugs for chronic conditions, and high cost cancer treatments). Understanding quality
constraints at the grassroots level and then building a comprehensive policy and investment
approach to addressing those gaps is an important step in this direction.

5
Grassroots is a term used in Vietnam to define the basic health services provided within an administrative district
at the district (about 120,000 population), commune (about 8,000 population) or even village (about 1,500
population) level. In its more general meaning, it would also include the preventive medicine system at the district
and commune level. Given the focus of this study on the quality of care, particularly for curative services, at the
Commune and District Level, the focus of this study is on District Hospital and the Commune Health Stations which
are the primary points of contact for the curative care system.

25
37. Vietnam has a tiered health system largely organized by administrative units
of the country: commune, district, provincial and central level. The commune, district
and provincial health facilities are under the direct authority of the provincial government and its
department of health. The legal and regulatory framework are largely set at the national level
government and its Ministry of Health. As an example, the functions and services at each level
are defined by the Ministry of Health Circular 43 (8),6 and each facility needs to fulfill requirements
from different aspects including location, function, activity, education level of human resources,
facility, equipment, etc.
38. The grassroots health network—commune health stations and district
hospitals—consists of a wide network of facilities. As of 2013 (9), it had more than
11,000 commune health stations reaching 99% of all communes. Commune health
stations are the front line public health service providers of vaccines, family planning, ante-natal
and post-natal care, monitoring for infectious disease control, and providing information and
communication activities to the community. Increasingly, commune health stations—particularly
in rural and remote areas—are being integrated in the national health insurance system and
providing basic illness consultations and treatment services. As of 2014, 80% of communes were
participating in the national health insurance system. The commune health stations, with its
prescribed team of about 5 health staff led by a doctor or assistant doctor, are a potential source
of more comprehensive and family-medicine oriented primary health care. As of 2014, there
were about 620 district hospitals of which about 85 had sub-units – closer to the commune health
stations – called regional polyclinics. These district hospitals (average bed size 112 and average
number of medical doctors 28) provide basic hospital in-patient services related to internal
medicine/trauma, obstetrics/gynecology, pediatrics and surgery as well as specialized outpatient
diagnostic and laboratory services. These facilities also provide primary outpatient services
similar to commune health stations. They do not have the same functions as commune health
stations to provide prevention services.
39. The grassroots level of the Vietnam health system is the main provider of
primary and basic secondary services to the majority of the population. The grassroots
level--commune health stations and district hospitals—is the first point of contact with the public
health system for most patients, and almost the sole provider of care to lower income groups.
As of 2014, 41% of the insured population was registered with a commune health station facility
as a first point of contact and 45% were registered at the district hospital with the remainder at
provincial and central level hospitals (10). District hospitals have some supportive, but often not
a direct supervision responsibility of the commune health stations as well as responsibility to
provide the financing and drugs of the curative services provided at the commune level. The
financing of commune health stations is mixed with the financing of salaries, a basic operating
budget, provision of additional inputs for some of the prevention services and prices per unit
delivered (such as vaccines). They will receive curative care drug supplies from the district

6
Circular 43/2013/ TT-BYT dated December 11, 2013 by the Ministry of Health defining specific services by
technical level for curative care facilities.

26
hospital and a small amount of fees for delivering health insurance related consultations. The
funds from health insurance are channeled through the district hospital and the funds for
prevention through district health center as in fact the commune health stations are not
independent juridical entities that would be able to be contracted directly. District hospitals
receive a minimum operating budget from the provincial budget largely to cover salaries
calculated based on normative allocations linked to their bed capacity. The growing share of
their budget is from health insurance which is largely based on a capped fee-for-service model
(by procedure, bed-day, drug provided). While there is some variation in this mechanism across
the nation, universally the incentive for district hospitals is to increase the range and number of
services provided. This can be odds with the commune health stations providing additional
curative services.
40. The grassroots health system is part of a wider health service system context.
Above the district hospitals are 492 provincial hospitals and 46 central level general
and specialized hospitals under the supervision of the Ministry of Health. The
provincial and central hospitals provide more specialized, secondary, and tertiary-level health
services. They also often provide similar basic health primary, specialized outpatient and
secondary services to those provided by lower level providers. While set up as a network of
health providers within an administrative area, it is important to note that patients have choice
for where they can seek care – at no additional cost or, if at a higher level, with some additional
co-payment. This has meant that when it has been physically and financially within their means,
patients can vote with their feet where they perceive they will receive quality diagnosis and
treatment. Health care providers are largely financed fee-for-service from the national health
insurance system or by the patient out-pocket. They also are reliant on their revenue earning
ability to finance the full cost of their operation, incentivize their staff, and make investments in
their facilities. Therefore, health care providers have the incentive to maximize their revenue
with more patients and more services, refer upwards only when necessary, and rarely transfer
patients back to lower level facilities. These incentives can also impact on perceived quality such
as when patients fear not being able to get a referral and, therefore, proceed directly to higher
level facilities. While analyzing these system incentives are outside of the direct measurement of
this study, they should be considered the study’s findings in broader perspective.
41. Poorer segments of the population rely on the grassroots health care system.
The poorest quintile of Vietnamese overwhelmingly rely on commune health
stations and district hospitals and rarely access higher level facilities. Commune health
stations are used principally by patients from poorer quintiles. District hospitals, however, are
accessed by patients across the socioeconomic spectrum. Therefore, while efforts to improve
the quality of basic health services of the grassroots system will improve the care provided by
the poor, they can also impact the population more broadly.

27
 Figure O.1: Percent of people in each quintile who have used the facility type in
the last 12 month

Source: Authors’ calculations from Vietnam Household Living Standard Survey, 2014.

42. This report takes an internationally accepted and practical approach to

measuring some aspects of quality of care. Defining and measuring quality of health
services are understood to be complex tasks. It is understood that the ultimate desired optimal
health outcome is result of many factors: when and if care is sought, the complexity and severity
of the ill-health condition, the accessibility of the health services able to diagnose and treat the
condition, the care delivered and how the care was delivered often by multiple care providers,
the compliance of the patient and contributing roles of the family and community of that individual
(11). Given the motivation factor of this study, this report uses the often relied upon Donabedian
model of measuring health service delivery quality as a function of structure (availability of necessary
inputs), process (actual practice) and to some extent outcome (as perceived by the patient and the patient
experience) (12). The study starts with looking at the availability of key inputs and the technical
capacity of the health service providers generally. This is often the starting point any policy and
investment plan to address perceived weaknesses in the quality of care, but not necessarily
targeted to most deficient needs. Given that care is most directly a function of the doctor-patient
interaction and the policy choices affecting human resource development and distribution are
different than other investment choices, the study gives particular emphasis to analyzing the
availability, qualifications and the knowledge of the human resources. This study is very rich
providing information not previously available in Vietnam about the process of delivering care at
the grassroots level relying on extensive field work of directly observing doctor and patient
interactions and patient exit interviews. In a few but important circumstances, the study also
compares actual practice with the knowledge leading us and policy makers to discuss whether or
not additional training is always the most necessary investment to address deficiencies in actual
practice or whether other policies affecting motivation is of higher priority. Health outcome,
given its complexity, is limited in this study to what is perceived by the patient, but the patient
perspective is important given that the Vietnam health system is increasingly open to the choice
of provider. Therefore, while not an exhaustive measurement of quality at the grassroots level,

28
this study provides extensive and new information through which policy makers and other
researchers can make reasonable inferences and as well as target further investigations.
43. This study examines quality through an equity lens. There is a perception that
the quality of care in Vietnam is greater at higher level facilities, at urban facilities, and for higher
income patients. The perception of quality differences across facilities drives a tendency for
patients to by-pass or self-refer themselves to higher level facilities, avoiding the grassroots level
health system all-together (13). This study is not all-inclusive of looking at variations of quality
across the different levels of the health system but considers differences within the grassroots
health system. It reviews differences across 6 different geographical zones of the country as
represented by different provinces; by urban and rural areas; between the commune health
station and the district hospital; and through a proxy means test of the wealth of individual
patients seen at the health facility.
44. The data referenced in this study can be used as an indication of the situation
in the country but is not nationally representative. Due to resource constraints, it was
not designed as a nationally representative survey. Six provinces were purposefully selected to
represent the six different geographically distinct regions and following their “representativeness”
of that province of that region. One province was chosen from a less developed and ethnic-
minority concentrated province (Dien Bien). One of the larger urban centers was also selected
(Hanoi). The following table summarizes the selected provinces.

Table O.1: Description of surveyed provinces

Region Province Surveyed Poverty Rate1 Per Capita Income2
Red River Delta Hanoi 1.0 2944.9
Northern Highlands Dien Bien 35.2 819.4
North and South Central Coast Binh Dinh 9.9 1719.0
Central Highlands Dak Lak 12.3 1639.2
Southeast Dong Nai 0.7 2576.7
Mekong Delta Dong Thap 7.5 1665.5
Vietnam Nationally 7.8 1999.8
Source: (1) MOLISA poverty rate 2013; (2) Vietnam Household Living Standard Survey 2012

45. The new data collected was extensive. Health facilities were selected following a
stratified sample of households between the urban and rural areas of those provinces, selecting
the commune (246) or district hospital (78) facilities within the administrative areas of those
households, in order to compare differences in the quality of health care services between urban
and rural areas.7 The data set was based on a facility survey, with an extensive module on the
human resources of the facility, the knowledge of the doctors using vignettes (1000 doctors

7
This survey complements a separate Household Survey being undertaken separately and solely by HSPI. However,
in the future, the results measuring household health seeking behavior and factors of facility quality can be compared.

29
undertaking 5 vignettes each), and directly observed doctor patient interactions (385 doctors
observed treating 8024 patients). Also, there was an extensive outpatient (at the commune and
district level) and inpatient (at the district level) patient exit interview. The specific instruments
were based on Vietnam experience with the National Health Survey in 2001, the World Bank
Service Delivery Indicators Survey as well as other instruments such as WHO Service Area
Readiness Assessment Survey. This study only begins to use this new data set which will be made
available to the public and particularly other researchers following the publication of this study.
For more specific information on the sampling and instrument design, please refer to Annex A
and Annex B.
46. The structure of the report and principal questions addressed by chapter is as
follows:
47. Chapter 1: General Service Readiness – When looking at selective tracers of
infrastructure, medical equipment and pharmaceuticals are there indications that there may be
potential gaps affecting service delivery? Are health care providers equipped with health
information systems that would enable them to manage patients and analyze health information?
Are the healthcare providers able to provide the services currently regulated by the Ministry of
Health, which is a function of human resource and technology capacity? How is this general
service capacity translating to case load at the moment? Are there key findings in differences
across the country or between rural and urban areas?
48. Chapter 2: Healthcare Workers -- What is the distribution and profile of
healthcare workers? What are the key characteristics in terms of qualifications and experience
that may lead to differences in knowledge and competency? What is the difference in terms of
the doctors’ entry into medical schools that is perceived to be a signal of the rigor of the doctors
training? What are differences in gender and ethnic minority status? What do the doctors
indicate as priority investments that would improve their professional satisfaction? Are there key
differences in the profile of healthcare workers, particularly doctors, between the different
grassroots health service providers and across different geographical areas?
49. Chapter 3: Patient Experience – Based on patient exit interviews, what is the
satisfaction with various aspects of the patient visit including waiting time, doctor-patient
interaction, and availability of service? What was the out-of-pocket cost associated with the visit?
Based on the patients using the service and a proxy of their wealth using assets, is there any
difference in the experience and perception of the experience across wealth quintiles? Does this
different by different geographical areas of the country?
50. Chapter 4: Doctors Knowledge: Analysis of Clinical Vignettes – Are the doctors
able to ask for the appropriate medical history and tests, based on Ministry of Health approved
clinical guidelines that would lead towards better diagnostic capability of a few common and
priority health condition? Are doctors able to diagnose accurately based on given facts about
medical history and test results? Are doctors able to prescribe the correct treatment with the
given information on the diagnosis, again on the basis of the Ministry of Health approved

30
guidelines? What may be differences in the knowledge and competency of doctors across
different geographical factors, but also based on their profile?
51. Chapter 5: Doctors Practice and Differences between Knowledge and
Practice – What do doctors do based on actual observed practice of the doctor? How long is
the actual consultation time? Are there any indications of possible under-treatment, over-
treatment or harmful treatment that would lead to quality concerns? In a couple of simple, but
priority cases where we are able to match with the doctors knowledge, are we able to identify
differences between the doctors knowledge and actual practice which could be signaled as a
“Know-Do” gap? Are there any differences in practice or the “know-do” gap between different
kinds of providers, doctors or geographical areas of the country?

31
 Chapter 1: Facility Service Readiness
Vietnam’s grassroots level facilities typically have much of the infrastructure required for delivery of
basic health services, although there are important gaps. All surveyed health facilities have electricity,
almost all have toilet facilities, and most have waste water and solid waste treatment systems. Many
district hospitals lack important equipment such as anesthesia equipment, child ventilators,
electrocardiograms, and blood glucose analyzers. Nearly all facilities have computers and internet
access, but only 22% of district hospitals and almost no commune health stations exploit those tools by
using any form information management system.

Many facilities lack essential medicines. The survey assessed availability of a list of 30 essential
medicines at both district hospitals and commune health stations. On average district hospitals had half
of the medicines on hand and commune health stations had one-third. This could be attributed partly
to a reliance on more expensive medication, as opposed to the essential generic medications. In exit
interviews, patients frequently reported that they were prescribed medicines not available at the facility
itself, and the most common reason cited was that the prescribed medicine were not on the list of
essential medicines covered by insurance.

With the exception of Dien Bien, the prevalence of doctors and specialists in commune health stations
across provinces is similar. In every other province, more than 80% of commune health stations had at
least one fully-qualified doctor, but only 24% of commune facilities in Dien Bien had a fully-qualified
doctor present. Those without doctors had assistant doctors.

Caseloads are highly varied across facilities and low at many commune health stations. Caseloads
average just 10 patients per day at commune health stations and 36 patients per doctor per day at
district hospitals. Caseloads per doctor per day at hospitals vary from 18 to 65. However, bed
occupancy rates at district hospitals are universally high—ranging from 83 to 97%. The average length
of stay for inpatients at district hospitals of 6.2 days is quite long, particularly for the type of basic
services offered at district hospitals. The long length of stay is one factor that drives up out-of-pocket
costs.

32
1.1 Introduction

52. This chapter considers a set determinants of the quality of care: facility
characteristics that can be summarized as “service readiness.” Facets of service
readiness include 1) access to infrastructure, 2) medical equipment and pharmaceutical products,
and 3) staff availability and qualification. The survey collected information on these aspects of
service readiness in both district hospitals and commune health stations. The set of information
collected in the survey and presented here is based on policy documents on required
infrastructure, equipment, and medications, with adjustments and additions based on the
informed views of Ministry of Health experts.

1.2 General service readiness amenities

53. In terms of access to infrastructure, the service readiness assessment presents

a mixed picture. Figure 1.2.1 provides a summary of basic infrastructure availability by facility
type. All the surveyed health facilities have electricity, and almost all are equipped with a septic
or semi-septic toilet. Nearly all (97%) of district hospitals and 95% of commune health have access
to clean water, a critical resource for ensuring quality health care services.8 The large majority of
district hospitals have proper health care waste management systems, but they are much less
common at commune health stations. Specifically, among district hospitals, 94% have waste water
treatment system and 92% have a specialized incinerator system or had contracted a sanitation
company for disposal of solid waste. The corresponding figures for commune health stations are
15% and 70%. Nearly all district hospitals owned an ambulance. 9 Access to computers and the
internet was available at almost all facilities, but just 22% of district hospitals and no commune
health stations had any hospital management or information management systems. Most of the
health facilities (92% of commune health stations and 100% of district hospitals) had a landline
phone.

8
Clean water is defined as piped water or water from a protected well.
9
Commune health stations typically do not own ambulances. In the survey, 49% of the commune health stations
that they in case of transfer of patients in case of emergency, they would use an ambulance (sent from a district
hospital), rather than other forms of transportation.

33
 Figure 1.2.1: Availability of general service readiness amenities at district
hospitals and commune health stations

Landline phone 92%

100%

Solid waste treatment 70%

92%

Waste water treatment 15%

94%

Clean water 95%

97%

Ambulance 97%

Internet access 95%

100%

Commune District
Source: Author’s calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (i) a facility is considered to have solid waste treatment system if its medical solid waste is processed by
specialized incinerators or sanitation companies; (ii) a facility is defined to have water waste treatment system if
the waste water is processed by concentrated sewage treatment system or laboratory sewage treatment; (iii)
Clean water is defined as piped water or water from a protected well. (iv) Availability of ambulance for district
hospitals refers to ownership of an ambulance by the hospital, and for commune health stations it refers. For
commune health stations, the relevant question refers to what is the commonly used method of transferring
patients to another facility (where other options are using vehicles other than ambulances.)

54. There are substantial spatial differences in access to some infrastructure.

Facilities in Dien Bien have the lowest levels of infrastructure. Just 76% of commune health
stations in the province have clean water, and among hospitals in the province, only 70% have
waste water treatment system and 50% possess solid waste treatment systems. Few commune
health stations anywhere have waste water treatment systems, across a range from 2% in Binh
Dinh to 30% in Dong Nai. The proportions of commune health stations with solid waste
treatment systems ranged from 21% in Dien Bien to 93% in Dong Nai.

34
 Figure 1.2.2: Commune health station access to clean water by province and
urban/rural
100% 98% 100% 98% 100% 97%
93% 92%

76%

Dien Dak Lak Binh Dong Dong Rural Urban Rural Urban
Bien Dinh Thap Nai Hanoi Hanoi
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: Clean water is defined as piped water or water from a protected well.

Table 1.2.1: Availability of waste management systems at district hospitals and

commune health stations by province
Wastewater Solid waste
District (% of Commune (% District (% of Commune (% of
facilities) of facilities) facilities) facilities)
Dien Bien 70 6 50 21
Dak Lak 100 15 100 37
Binh Dinh 100 2 100 85
Dong Thap 100 13 100 77
Dong Nai 100 30 100 93
Rural Hanoi 89 24 94 78
Urban Hanoi 100 10 100 91
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (a) a facility is considered to have solid waste treatment system if its medical solid waste was processed by
specialized incinerators or sanitation companies; (b) a facility is defined to have liquid waste treatment system if
the waste water was processed by the concentrated sewage treatment system or laboratory sewage treatment.

55. In comparison to a similar assessment of infrastructure in Indonesia,

Vietnam’s performance is stronger in all areas. The conditions of Vietnam grassroots
health facilities were similar or better as compared to those in Indonesia for electricity, clean
water, toilet, and communication conditions (14).

35
 Table 1.2.2: General service readiness indicators in Vietnam and Indonesia
Electricity Clean Toilet Communication
Access water
% % % %
Vietnam rural commune 99 93 92 87
health station
Vietnam urban commune 100 97 98 97
health station
Vietnam district hospital 100 98 97 100
Indonesia rural primary care 97 69 71 81
facilities
Indonesia urban primary 99 81 84 89
care facilities
Indonesia public hospital 98 94 100 100
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: for Vietnam facilities: (1) clean water refers to having water from a piped source or protected well (2) toilet
refers to having a septic or semi-septic for patient utilization; (3) communications refers to having a functioning
landline telephone availability.

1.3. Medical equipment and medication

56. Access to medical equipment is similar across provinces. An inventory of

equipment was carried out in district hospitals and commune health stations. Figure
1.3.1 shows the average number of items available by province and facility type. The average
district hospital had 20 of surveyed 24 items, and the average commune health station had 10 of
16 surveyed items.

36
 Figure 1.3.1: Average number of available pieces of medical equipment

Equipment at district hospitals Equipment at commune health stations

Urban Hanoi 16 Urban Hanoi 10

Rural Hanoi 22 Rural Hanoi 11

Dong Nai 21 Dong Nai 10

Dong Thap 20 Dong Thap 10

Binh Dinh 20 Binh Dinh 10

Dak Lak 21 Dak Lak 11
Dien Bien 19 Dien Bien 10

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: The indicator was calculated as the mean of number of equipment surveyed at district hospitals and
commune health stations. (1) the list of surveyed equipment at district hospitals: adult scale, child scale (250g
gauge), infant scale (100g gauge), thermometer, stethoscope, pinard horn, sphygmomanometer, vaccine cold chain
(refrigerator, vaccine flaks), monitor, portable oxygen concentrator, ventilator, child ventilator, infant incubator,
anesthesia machine, defibrillator and pacemaker, ECG device, C-section toolkit, X-ray, ultrasound scan, CT scan,
electrocardiogram, blood analyzer, blood biochemical analyzer, HbA1C testing; (b) the list of surveyed equipment
at commune health stations: adult scale, child scale (250g gauge), infant scale (100g gauge), thermometer,
stethoscope, pinard horn, sphygmomanometer, oxygen canister, ambu bag, stomach cleansing toolkit,
delivery/natal care table, oral fluid ventouse, antiseptic autoclave/oven, refrigerator, ice box, microscope.

57. Some equipment is only available at a minority of district hospitals. Table 1.3.1
presents the availability of selected equipment at district hospitals 10.While almost all
surveyed hospitals had equipment to perform hematology and biochemical analyzer, only 44%
had HbA1C testing devices for diabetes. The proportions of hospitals with HbA1C testing devices
were quite low in Dak Lak and Dien Bien. Almost half of the hospitals in Hanoi and Dong Nai
had a CT scanner compared to none in Dien Bien. Child ventilator and infant incubator were not
widely available in district hospitals in Dien Bien, Binh Dinh, and Dong Thap. Only one third of
district hospitals in these provinces had a child ventilator and infant incubator was only available
in 46% and 60% of hospitals in Dong Thap and Dien Bien, respectively. Without these equipment,
the hospital cannot establish neonatal care unit to provide essential pediatric intensive care
services. The low figures for urban Hanoi reflect that the fact that the sample of 9 district hospitals
in urban Hanoi includes 3 polyclinics, which do not provide inpatient and delivery services.

10
Some other equipment, including X-ray, Ultrasound, hematology and biochemical analyzer were available in almost
all hospitals so that they were not included in the below table.

37
Table 1.3.1: Availability of selected equipment at district hospitals by province
Basic Resuscitation & Emergency care Diagnostic Labor
equipment atory
testing
Infant Child Infant Anesthesia Defibrillator CT Electrocar- Hb1AC
scale ventilator incubator machine and scanner diogram testing
(100g) pacemaker device
Dien
Bien 90% 30% 60% 80% 90% 0% 30% 20%
Dak
Lak 90% 70% 100% 100% 90% 20% 50% 20%
Binh
Dinh 82% 27% 82% 100% 55% 27% 73% 27%
Dong
Thap 100% 27% 46% 73% 91% 18% 73% 55%
Dong
Nai 100% 44% 67% 78% 89% 44% 78% 89%
Rural
Hanoi 100% 78% 100% 94% 89% 11% 72% 50%
Urban
Hanoi 33% 33% 33% 50% 44% 56% 89% 50%
Total 87% 47% 73% 84% 80% 23% 67% 44%
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

Figure 1.3.2: Percentage of commune health stations having all surveyed

equipment

42%
35%
29%

17%

4% 2% 3%
0%

Dien Bien Dak Lak Binh Dinh Dong Thap Dong Nai Rural Hanoi Urban Total
Hanoi
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (1) 16 surveyed equipment for commune health stations: adult scale, child scale (250g gauge), infant scale
(100g gauge), thermometer, stethoscope, pinard horn, sphygmomanometer, oxygen canister, ambu bag, stomach
cleansing toolkit, delivery/natal care table, oral fluid ventouse, antiseptic autoclave/oven, refrigerator, ice box,
microscope.

38
58. Just 17% of surveyed commune health stations had every piece of surveyed
equipment, with large variation across regions. No commune health stations in Dien Bien
had all the equipment. More than 90% in Hanoi and Dak Lak had a microscope, compared to
only 5% and 6% in Dien Bien and Dong Thap. Oxygen canisters were only available in 3% of
commune health stations in Dien Bien and 15% Dak Lak. Equipment to serve for vaccine cold
chain, such as refrigerator and ice box, were not available across all provinces. The rate of having
a refrigerator ranged from 44% to 70% in Dak Lak, Binh Dinh and rural Hanoi. (See Table AE.3
Annex E.)
59. None of the surveyed health facilities had the full selected list of medicines.
The assessment included a check for the availability of a list of 30 essential medicines at both
district hospitals and commune health stations. The list includes tracer medicines drawn from the
World Health Organization’s SARA guidelines for primary health care. The survey team checked
the facility pharmacy for individual medicine in the list as well as the storage record book to
identify if there was stocking out in the last 12 months. Half of the district hospitals had 50-75%
of the required medicines. Among commune health stations, 42% had 25-50% of the essential
medicines, and 30% of them had less than 25% of the medicine in the surveyed list.

60. On average district hospitals had 16 of 30 medicines and commune health

stations had 11. There was substantial variation across provinces. The medicines were more
available in Dong Nai and Dong Thap than in Dien Bien and Dak Lak.

Figure 1.3.3: Average number of pharmaceuticals available at district hospitals

and commune health stations

District hospitals Commune health stations

Urban Hanoi 15 Urban Hanoi 7

Rural Hanoi 15 Rural Hanoi 10
Dong Nai 19 Dong Nai 13
Dong Thap 19 Dong Thap 16

Binh Dinh 18 Binh Dinh 10

Dak Lak 11 Dak Lak 7

Dien Bien 13 Dien Bien 6

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: the indicator was estimated as mean of available pharmaceuticals (on the survey day) at district hospitals or
commune health stations among 30 surveyed medicines. The list of these 30 pharmaceuticals is presented in
Tables AE.6 & AE.7, Annex E.)

39
1.4. Health staff availability and qualifications

61. With the exception of Dien Bien, the prevalence of doctors and specialists in
commune health stations across provinces is similar (Table 1.4.1). In every other
province, more than 80% of commune health stations had at least one doctor, but only 24% of
commune facilities in Dien Bien had a doctor present. Those without doctors had only assistant
doctors. Only 37% of commune health stations were equipped with a full team of doctor/assistant
doctor, nurse, midwife, traditional practitioner and pharmacist. In average, each commune health
station had 8 staff, include all types of civil servants, contracted staff, and temporary staff.

Table 1.4.1: The availability of qualified staff at commune health stations

Having at least 1 Having at least Having a full
doctor (%) one specialista(%) compositionb(%)
Dien Bien 24 3 3
Dak Lak 96 19 56
Binh Dinh 90 22 37
Dong Nai 92 39 54
Dong Thap 93 25 40
Rural Hanoi 82 22 20
Urban Hanoi 95 29 62
Total 82 23 37
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: The share was calculated for 246 commune health stations. (a) Specialist is a post graduate level (after 6-
year training of medical doctor) on a specific clinical area. The program is 2 year training for Specialist level 1 and
another 2 years for Specialist level 2; (b) Having full staff composition: have at least one doctor or assistant
doctor, nurse, midwife, traditional practitioner, pharmacist.

62. Surveyed district hospitals have relatively few nurses per doctor. Across
hosptials the ratio averaged 1.4, which is similar to figures from administrative sources and is 11
low in international comparison (15). Only 30% of WHO member states have fewer than 2 nurses
per doctor (16). Dien Bien and Dong Thap had the lowest doctor/nurse ratio while Urban Hanoi
had more nurses than the others. The relative scarcity of nurses may explain why inpatients in
hospitals are often cared for principally by the patient’s relatives, who stay in the hospital and
contribute to hospital overcrowding.

11
Ministry of Health, Health Statistics Year Book 20014

40
 Figure 1.4.1: The ratio of number of nurses and doctors by provinces
2.3

1.7
1.4 1.5 1.4
1
0.8 0.8

Dien Bien Dak Lak Binh Dinh Dong Thap Dong Nai Rural Hanoi Urban Total
Hanoi

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

1.5. Service provision

Figure 1.5.1: Caseloads from commune health station clinical observation on the
survey day

Based on direct observation on survey day

Urban Hanoi 4
Rural Hanoi 7
Dong Nai 10
Dong Thap 17
Binh Dinh 7
Dak Lak 12
Dien Bien 9

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (a) a surrogate “caseload” variable was estimated based on the number of patients per doctor on the survey
day. Despite the data we actually collected was only from one day, this visited day was selected randomly based
on the study planning team.

63. Next we consider the caseload at facilities. Caseload is calculated based on the
counts of outpatients directly observed during the clinical observation portion of the study,
considering only the health workers who were observed. There were an average of 10
outpatients per doctor, while the health reports of the commune health station reported 32
patients per commune health station. Many commune health stations have very lower patient

41
loads. At 49 commune health stations in 5 different provinces, the number of outpatients
observed on the day of the survey was zero. Doctors at six commune health stations in Dien
Bien did provide any patient care at all on the day of the survey. These facilities were all close to
a hospital or polyclinic. The low caseload of these facilities raises the question of whether it is
cost effective for these facilities to remain open only to provide services other than patient care
by doctors. This issue is further discussed in Chapter 5.
64. At district hospitals, based on direct observations, outpatient caseload was far
higher than at commune health stations. Observed doctors had an average of 36
consultations per day. The figure was lowest in a polyclinic in Hanoi (7 patients per doctor per
day) and was highest in Lap Vo district Hospital in Dong Thap province (87 patients per doctor
a day).

Figure 1.5.2: Average number of outpatients per doctor on the survey day
65

50
45

33
28
19 18

Dien Bien Dak Lak Binh Dinh Dong Thap Dong Nai Rural Hanoi Urban Hanoi
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

65. The table below shows the average number of inpatient that a doctor had to
manage in a day, as well as the average length of stay and occupancy rate in the
surveyed district hospitals. District hospital doctors managed about four patient beds per
day, except in Dien Bien when each doctor managed 1.8 patient beds. In all surveyed district
hospitals, bed capacity was close to 90%, based on the number of actual beds.

42
 Table 1.5.1: Number of inpatient per doctor per day at district hospitals
Actual Length of stay Occupancy rate by
beds/doctor (day) actual number of
beds (%)
Dien Bien 1.8 5.8 93
Dak Lak 4.4 5.5 97
Binh Dinh 4.0 6.1 83
Dong Thap 3.9 5.4 83
Dong Nai 4.9 7.0 92
Rural Hanoi 3.8 6.2 96
Urban Hanoi 2.3 8.2 90
All district hospitals 3.7 6.2 91
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: Occupancy rate was calculated as the number of inpatient days for a given period divided by number of
actual beds and number of days in the period.

66. The average length of stay (ALOS) at district hospitals was 6.2 days. The ALOS
is one measure of the efficiency with which hospital resources are used. It is normally assumed
that other things being equal, shorter stays reduce treatment cost by shifting care to less
expensive outpatient and ambulatory options. The figure below shows the ALOS across countries
for comparison (17). Note that the ALOS calculated in this study only reflects the situation in the
lowest level of hospital, and normally length of stay is longer on average in higher level facilities.
The extremely long ALOS in Japan and South Korea reflects the aging burden in these countries
with high prevalence of chronic condition and sophisticated comorbidity.

43
 Figure 1.5.3: Average length of stay in hospitals in selected countries

Japan (2010) 18.1

Korea, Republic (2010) 14.1
China (2011) 9.0
Mongolia (2011) 8.0
OECD (2009) 7.3
Vietnam (2003) 6.3
Vietnam (2015) 6.2
Myanmar (2008) 6.1
Australia (2009 5.5
Cambodia (2009) 5.3
Singapore (2011) 5.0
Indonesia (2011) 4.3
Malaysia (2010) 4.1
Thailand (2005) 4.1
Bangladesh (2008) 4.0
Srilanka (2008) 3.2
Source: OECD/World Health Organization (2012) (17)
Note: for the data of Vietnam 2015, the average length of stay was calculated from Equity in health survey, 2015.

1.6. Availability of basic services

67. The survey collected information as to whether particular services are offered
by the facility. The selected set of services covers a number of basic services related to maternal
and child health and non-communicable diseases control and treatment. Overall, 30% of the
district hospitals provide all 100% of the surveyed services, and 8% facilities can perform less than
50% of the required services.12
68. At commune health stations, antenatal care was almost universally available,
but many other services were not available for a substantial fraction of facilities. At
the urban area, a low percentage of commune health station provide childbirth services, which
are now more commonly sought at district hospitals. The fraction of facilities that offer Hepatitis
B vaccination for newborn child within first 24 hours after delivery was just 34%.
69. The study also focused on the provision of non-communicable management
services, specifically diabetes and hypertension, as they are emerging challenges for
Vietnam. Few commune health station implementing NCD related programs. About 50% of the
commune health stations declared that they could register diabetes and hypertension patients,

12
Urban Hanoi was excluded for analysis because of small sample size and a large variation among the selected
facilities.

44
and 40% of the commune health station declared that they could provide treatment to those
patients. Across provinces, very few commune health stations could dispense diabetes medicines
(only 17% in general and none of commune health station in Dien Bien province) (Table 1.6.1).
Figure 1.6.1 compares the availability of services in rural and urban areas. Differences were minor
across each trace service.

Table 1.6.1: Percent of commune health stations that can perform selected services
Binh Dak Dien Dong Dong Rural Urban
Dinh Lak Bien Nai Thap Hanoi Hanoi Total
Antenatal care and
pregnancy management 100 100 100 100 100 98 100 100
Tetanus vaccination for
pregnant women 100 100 100 97 97 98 100 99
Attended vertex
presentation normal
delivery 78 52 70 55 90 89 5 68
Hep B vaccination within
first 24 hrs 34 26 45 17 36 58 0 34
Child acute diarrhea:
diagnosis and treatment 93 89 70 97 100 96 100 92
Child pneumonia 88 93 70 95 100 93 95 91
Diabetes type II: screening
program 32 33 70 40 49 40 38 43
Diabetes type II:
registration based
management only 24 48 85 53 87 71 48 60
Diabetes type II:
management and
monitoring 17 41 64 45 90 42 33 48
Diabetes type II: Periodical
drug dispense 10 7 0 13 56 7 14 16
Hypertension: Screening
program 24 52 82 35 49 62 29 48
Hypertension: registration
based management 24 63 85 55 79 78 43 62
Hypertension:
management and
monitoring 22 63 79 45 92 49 24 54
Hypertension: periodical
medicine dispense 15 63 39 33 64 18 19 35
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

45
 Figure 1.6.1: The availability of services at commune health stations by Rural and
Urban areas
Rural area Urban area
Hypertension: Hypertension:
39% 31%
periodical medicine… periodical medicine…
Hypertension: Hypertension:
60% 49%
management and… management and…
Hypertension: Hypertension:
68% 56%
registration based… registration based…
Hypertension: Hypertension:
53% 43%
Screening program Screening program
Diabetes type II: Diabetes type II:
18% 13%
Periodical drug… Periodical drug…
Diabetes type II: Diabetes type II:
52% 44%
management and… management and…
Diabetes type II: Diabetes type II:
63% 57%
registration based… registration based…
Diabetes type II: Diabetes type II:
39% 46%
screening program screening program
Child pneumonia: Child pneumonia:
92% 90%
diagnosis and treatment diagnosis and treatment
Child acute diarrehea: Child acute diarrehea:
92% 92%
diagnosis and treatment diagnosis and treatment
Hep B vaccination Hep B vaccination
46% 22%
within first 24 hrs within first 24 hrs
Attended vertex Attended vertex
87% 50%
presentation normal… presentation normal…
Tetanus vaccination for Tetanus vaccination for
98% 99%
pregnant women pregnant women
Antenatal care and
100% Antenatal care and
pregnancy management 99%
pregnancy management

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

70. As a cross-check on the health facility’s declaration of their available services,

the study checked the actual patients of these tracer services. Data showed a low
numbers of patients for these tracer services. The table below (Table 1.6.2) shows the percentage

46
of commune health stations that had at least one patient of the tracer diseases in the last year.
Except in Dong Thap, very few commune health stations had dispensed diabetes medication,
although substantial numbers reported that they had provided management and monitoring of
diabetes. Likewise, few commune health stations in most provinces had dispensed hypertension
medicines. These findings reflect the historical role of commune health stations, which have
traditionally focused on maternal and child health and are less well equipped to handle the
growing profile of non-communicable disease.

Table 1.6.2: The percent of commune health stations which have at least one
patient by case type in the last year
Dien Dak Binh Dong Dong Rural Urban Total
Bien Lak Dinh Thap Nai Hanoi Hanoi
Attended natural
childbirths in facility 36 33 17 54 10 62 5 33
Child acute diarrhea 55 74 54 87 77 82 71 72
Child pneumonia 39 74 49 87 53 80 67 64
Diabetes type II:
Management &
monitoring 67 63 22 92 47 49 38 54
Diabetes type II:
medicine dispense 0 7 10 56 12 7 10 15
Hypertension:
Management &
monitoring 67 67 22 92 45 56 29 54
Hypertension: medicine
dispense 33 62 15 64 32 18 14 33
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

47
 Chapter 2. A profile of health workers

Fully qualified doctors comprise just 11% of health care providers at the commune health stations
visited by the survey. Assistant doctors, midwives, nurses, and others provide substantial components of
CHS care. Most CHS health care workers were provided at least some form of in-service training in the
previous 12 months. Considering pay across all professions, CHS doctors are paid the most, and
salaries (including allowances) are highest in Dien Bien province due to the greater allowances paid to
workers living in more remote and ethnic minority areas. CHS health care workers overwhelmingly
express a desire to continue working long-term at their current facilities, except in Dien Bien, where only
60% have such intentions. CHS health care workers suggest additional staffing, more medical
equipment, and more training opportunities as possible measures to improve the quality of CHS service.
A brief profile of health facility directors shows that in many cases—particularly in Dien Bien—facilities
are managed by assistant doctors rather than fully-qualified doctors. CHS directors tend to be older,
with substantial years of clinical experience and are disproportionately men. Directors at district
hospitals typically are doctors with higher level training and are overwhelmingly (94%) male.

Doctors at district hospitals tend to be younger but with higher levels of qualifications than those at
commune health stations. In Dien Bien and Dak Lak provinces, a large fraction of doctors are ethnic
minorities—equaling or exceeding their share of the general population. The large bulk of doctors at
commune health stations (considering those in both commune health stations and district hospitals) do
not have standard university doctor training but instead either have been promoted from assistant
doctor through “twinning” training programs or pursued degrees through less competitive “direct entry”
programs. Nearly half of doctors at district hospitals have bachelor’s level training, and the remainder
have either primary specialist or specialist level 1 training. There is wide variation in the amount of
continuous medical education (in-service) training that doctors receive.
Variation in doctors’ salaries reflects levels of education, experience, and location. On average ethnic
minority doctors are paid more, reflecting the fact that they are chiefly in Dien Bien and Dak Lak,
where doctors are paid greater allowances. No gender difference is found in doctor salaries.

Work by doctors in private practice outside of their jobs at public facilities is very common, particularly
among doctors in the more southern provinces: Binh Dinh, Dong Nai, and Dong Thap. Thirty-eight
percent of hospital doctors perform some private practice work, and the average time per week in
private practice across all hospital doctors is 11 hours. The prevalence of private work and average
private hours is slightly lower at commune health stations (31% of doctors, averaging 9 hours per
week). Few doctors in Dien Bien conduct private practice. Private practice is much less common for
other health care workers.

48
2.1. Introduction

71. This chapter provides a multifaceted profile of health workers in Vietnam. The
first section summarizes the characteristics of all health workers at commune health stations. It
describes their broad professional profile, their training opportunities, the time they spend on
private practice, their satisfaction rates, and their views on how to improve health services. The
second section provides a more detailed profile of doctors, in both commune health stations and
district hospitals. The survey collected general information on the characteristics of staff. The
director of the district hospital and head of the commune health station were asked a separate
set of questions on their management and clinical experience. Ten physicians in each district
hospital and all physicians available in commune health station were interviewed on their working
condition and satisfaction as well as through a vignette interview to measure their clinical
knowledge, which is analyzed in Chapter 4 of this report.

2.2. Health workers in commune health stations

72. Commune health station health care providers are professionally diverse.
Among 1688 commune health station health workers in the survey, 11% are fully qualified
doctors, 35% are assistant doctors, 6% are traditional medicine assistant doctors, 15% are
midwives, 17% are nurses, 12% are pharmacists, and 4% are in other professions. There is notable
variation by profession in the age, gender, and ethnic profile of health workers. Doctors and
assistant doctors are typically the oldest staff—with average ages respectively of 45 and 40,
compared to 35 for other staff. While there is substantial gender balance among doctors (40%
women) and assistant doctors (65% women), nearly all (88%) other health workers are women.
The fractions of ethnic minorities among doctors (12%), assistant doctors (15%), nurses (13%),
midwives (10%) and others (13 %) are higher than the overall fraction of ethnic minorities in the
population of six provinces in the survey (8%), but among traditional medicine assistant doctors
and pharmacists few are ethnic minorities (4% and 7%, respectively).
73. The availability of training is one measure of the opportunities health workers
have to improve their skills. Across professions, doctors at commune health stations have
the most training opportunities, while traditional medicine doctors and pharmacists have the
fewest. Across provinces, Dien Bien, Binh Dinh, Dong Thap have higher proportions of health
workers participating in at least one training, but lower average numbers of trainings per staff.
Among the six provinces, Dak Lak health staff had the fewest training opportunities (Figure 2.2.1).

49
 Figure 2.2.1: Frequency of training for health workers at commune health stations
Proportion of commune health station health Average number of trainings per commune health
workers having any training in last 12 months by station health worker in last 12 months by
profession profession
4.4

74% 73% 74%

62% 65%
3.0
51% 49% 2.5
2.3
1.8
1.4
1.1

Proportion of commune health station health Average number of trainings per commune health
workers having any training in last 12 months by station health worker in last 12 months by
province province
3.3
81% 3.0 3.0
72% 72%
66% 65% 2.4 2.4
58% 58% 2.2

1.4

Dien Dak Binh Dong Dong Rural Urban Dien Dak Lak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi Bien Dinh Thap Nai Hanoi Hanoi
Source: Authors’ calculations from Vietnam District and Commune Health Facility Survey (2015).
Note: TM assistant doctors is “traditional medicine assistant doctors”.

74. Next we consider salaries (including allowances) of health workers at

commune health stations (Figure 2.2.2). Unsurprisingly, doctors’ salaries are on average
the highest, averaging 6.9 million VND per month. Average salaries, considering all professions,
are lowest in urban Hanoi (4.2 million VND) and highest in Dien Bien (6.5 million VND), which
is the poorest province in the sample. This reflects the greater allowances paid to workers in
more remote and ethnic minority dominated areas.

50
 Figure 2.2.2: Average monthly total salaries of health workers at commune
health stations

Average total salary of commune health station Average total salary of commune health
health workers by profession (millions VND) station health workers by provinces
(millions VND)
6.9 6.5

5.3
4.6 4.9 4.9 4.8 4.7
4.3 3.9 4.5
3.7 3.5 4.2

Dien Dak Binh Dong Dong Rural Urban

Bien Lak Dinh Nai Thap Hanoi Hanoi

Source: Authors’ calculations from Vietnam District and Commune Health Facility Survey (2015).
Notes: “TM assistant doctors” are traditional medicine assistant doctors. Total salary includes salary and
allowances due to occupational risk and duty/management.

75. Some commune health station health workers supplement their salaries by
working in private practice. While there are cases of other health workers providing private
care, the practice is overwhelmingly most common among doctors. Surveyed doctors on average
work 9.3 hours per week in their private practice (Figure 2.2.3). Commune health station
workers in more southern provinces including Dong Nai, Binh Dinh, and Dong Thap on average
worked in practice more than in other provinces. Dong Thap is a particular stand out, where
doctors on average work 18 hours a week in private practice.

51
Figure 2.2.3: Average number of hours per week in private practice by health workers
at commune health stations

Private practices (in hours) by profession Private practices (in hours) by provinces

Doctors 9.3 Urban Hanoi

Assistant Rural Hanoi

doctors 1.2

TM assistant Dong Nai

doctors 2.0
Dong Thap
Midwives 1.5
Binh Dinh

Nurses 1.2 Dak Lak

Pharmacists 3.1 Dien Bien

0 2 4 6 8 10 12 14 16 18 20
Others 0.7
Doctors Assistant doctors Others
Source: Authors’ calculations from Vietnam District and Commune Health Facility Survey (2015).
Note: TM assistant doctors is “traditional medicine assistant doctors”.

Figure 2.2.4: Commune health station health workers’ satisfaction

Mean satisfaction scores by profession Percent commune health station doctors
who want to work long term at the facility
3.3 3.1 3.1 3.1 3.1 3.2 93% 97% 100%
3.0 88% 91%
81%

60%

Dien Dak Binh Dong Dong Rural Urban

Bien Lak Dinh Thap Nai Hanoi Hanoi
Source: Authors’ calculations from Vietnam District and Commune Health Facility Survey (2015).
Notes: The satisfaction score was calculated as the mean of ten satisfaction indicators for salary, allowance, training
opportunity, promotion opportunity, occupational safety, hospital security, working environment, availability of
medicines, availability of equipment, working pressure, and having enough staff. Each indicator was measured using a
Likert scale with from very dissatisfied (1) to very satisfied (5).

52
76. Next we consider measures of job satisfaction for workers at commune health
stations. An overall index of satisfaction, calculated as a mean of satisfaction indicators, shows
similar values across professions—all close to the middle of the 1 to 5 scale. Another measure
of worker satisfaction is whether workers expressed a desire to work long term at the current
facility. Nearly all workers other than doctors said they intended to work long-term at their
current health facilities. Among doctors, however, there was substantial variation by geography.
Smaller fractions of doctors in poorer provinces said they want to work long term at their current
facility. Only 60% doctors in Dien Bien and 81% doctors in Dak Lak fall in this category (Figure
2.2.4).

77. Health workers were also asked about their recommendations for improving
services at their facilities. Responses were similar across different professions. The most
common answers were that health facilities should have more training opportunities for staff,
more medical equipment, and more health workers (Figure 2.2.5).

Figure 2.2.5: Suggestions of health workers at commune health stations for

service improvement

Better examination procedures

Better hospital security Doctors

Assistant doctors
Others
Better information system

More financial incentives

Sufficient medicine/medical supplies

Better facility infrastructure

More health workers

More medical equipment

More training opportunity

0% 5% 10% 15% 20% 25% 30%

Percent doctors choosing above suggestion
as the most important factor for service improvement
Source: Authors’ calculations from Vietnam District and Commune Health Facility Survey (2015).

53
78. At commune health stations, substantial numbers of women are directors, but
male directors outnumber female directors in all provinces except Dong Nai. The
mean age of the commune health station head is 48 years. They have an average of 23 years of
clinical experience and 11 years of management experience. Almost 56% of the commune health
station heads have university (bachelor’s or higher) degrees. The remainder are assistant doctors.
Dien Bien has the highest numbers of commune health station head who are assistant doctors.

Table 2.2.1: Profile of directors of commune health stations by province

Province
Dien Dak Binh Dong Dong Rural Urban Total
Bien Lak Dinh Thap Nai Hanoi Hanoi
Count by gender
Male 19 16 30 22 18 32 5 142
Female 14 11 11 17 22 13 16 104
Average age 41 46 49 48 49 50 48 48
Count by highest training
Specialist, master,
PhD 0 3 6 11 5 5 4 34
Bachelor 3 14 18 19 19 19 12 104
Assistant doctor 30 10 17 9 16 21 5 108
Years clinical experience
(mean) 18 20 25 25 26 25 23 23
Years management
experience (mean) 7 10 15 9 10 12 12 11
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

79. Extremely few hospital directors are women. Only 5 out of 78 hospital directors
in the surveyed hospitals are female. The mean age of the district hospital directors was 52.
Almost all had post-graduate training, and they average 26 years of clinical experience and 6 years
of management experience.

Table 2.2.2: Profiles of directors of district hospitals by province

Province
Dien Dak Binh Dong Dong Rural Urban All
Bien Lak Dinh Thap Nai Hanoi Hanoi
Count by gender
Male 9 10 11 11 8 17 6 72
Female 1 0 0 0 1 0 3 5
Average age 47 51 52 53 52 53 52 52
Count by highest
training
Specialist, master,
PhD 9 9 11 11 9 18 8 75
Bachelor 1 1 0 0 0 0 1 3

54
 Years clinical
experience (mean) 22 25 26 31 28 25 25 26
Years management
experience (mean) 5 8 8 6 6 5 6 6
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

2.3. Doctors in commune health stations and district hospitals

80. This section provides an additional focus on doctors, who play a particularly
critical role in the health system, leading in patient consultations, diagnosis, and
treatment decision. We examine their demographic characteristics, their training and
experience, and their compensation. Unlike the previous section, which apart from the discussion
of directors considers health workers at commune health stations only, the following analysis for
doctors is presented for both those at district hospitals and those at commune health stations.

81. As noted in the previous section, doctors are older on average than other
health care staff, and across most provinces the mean age of doctors is in the range
40-45 at both commune health station and district hospitals. The one major exception
is Dien Bien, the poorest province covered by the survey, where doctors are much younger,
averaging age 30 in commune health stations and age 35 in district hospitals. Doctors in district
hospitals are on average slightly younger than those in commune health stations, and this age gap
is more prominent in Hanoi (Figure 2.3.1).

Figure 2.3.1: Mean age of doctors by province and facility type

47 48
46 45 46
42 43 44 44
41 41
37
35
30
Years

Dien Bien Dak Lak Binh Dinh Dong Thap Dong Nai Rural Hanoi Urban Hanoi

District Commune
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

82. A substantial number of doctors are women, but with wide variation by facility
type and location. While overall 39% of doctors are women, they are nearly half of doctors at
urban commune health stations and a smaller fraction (31%) of doctors at rural commune health

55
stations. At district hospitals of various levels, the percentage who are women falls between those
extremes (Figure 2.3.2).

Figure 2.3.2: Percentage of doctors who are women by facility type

46% 49%
42% 39%
36%
31%

District hospital District hospital District hospital Rural Urban All doctors
level 1 level 2 level 3 Commune Commune
Health Station Health Station
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

Figure 2.3.3: Percentage of doctors who are ethnic minorities by facility type, and
percentage who are ethnic minorities in the general population, by province
100%

80%

60%

40%

20%

0%
Dien Bien Dak Lak Binh Dinh Dong Thap Dong Nai Rural Hanoi Urban Hanoi

District Commune General population

Source: Authors’ calculations from the Vietnam District and Commune Health Facility Survey (2015) for percent
ethnic minority doctors at commune health stations and district hospitals and from 2014 Vietnam Household
Living Standards Survey 2014 for the general population.

56
 Box 2.3.1: Levels of Doctor Training
Each doctor in Vietnam has one of three levels of training, all of which correspond to a
bachelor’s degree:
(i) Standard university doctor training: after students complete upper secondary
school, they can be admitted to a doctor training program at medical
universities based on a competitive entrance exam. Formal medical doctor
training usually takes 6 years.

(ii) Promotion from assistant doctor: under an earlier training regime, students
could train to become assistant doctors through a four year program after
upper secondary school (with admissions also based on a competitive exam).
These assistant doctor training programs required 4 years of full-time study.
Assistant doctor training programs no longer exist. As a transitional measure,
medical schools now have a program known as “twinning” by which assistant
doctors can become full doctors through an additional four years of study.

(iii) Direct entry programs: some medical schools offer less competitive admissions
for particular populations, with no entrance exam. Such programs exists for
applicants who i) are ethnic minorities or living in poor areas, and ii) willing to
commit to work in remote areas after graduation. They receive scholarships.
Top ranked medical universities generally do not offer these programs.
Additionally, some doctors have various levels of post-graduate training (beyond bachelor’s
degree-level):
Some doctors pursue a three year internship program. The internship program is very
selective and its graduates work largely at central level hospitals. No graduates from this
program were identified in the survey.

Doctors can also be designated as primary specialists with 1 year of training.

With additional training they can achieve designation as a specialist level 1 and with
further training as specialist level 2. These specialized trainings are hospital based and
focused on clinical practice, while the internship program is more comprehensive, including
both clinical and research skills.
A small number of doctors pursue master’s degrees and doctoral degrees (Ph.D.)

83. Next we consider the ethnic composition of doctors. In the population of Vietnam
as a whole, ethnic minorities constitute 14% of the population and are geographically
concentrated. They form a substantial share of the population in two of the provinces covered
by the Equity in Health Survey—Dien Bien and Dak Lak. In other provinces the share of doctors

57
who are ethnic minorities is small, except at commune health stations in urban Hanoi and Binh
Dinh, where they constitute 14% and 13% of doctors, respectively. In Dien Bien, where ethnic
minorities are a majority (81%) of the population, they make up a similar share of commune
health station doctors (80%) and a lower but substantial share of district hospital doctors (47%).
In Dak Lak, ethnic minorities make up larger shares of the doctor population in both commune
health stations (62%) and district hospitals (31%) than they do in the population overall. Notably,
all ethnic minority doctors in Dien Bien at commune health stations are under age 30, which
suggests that their presence is the result of recent efforts to promote medical education for
ethnic minorities (Figure 2.3.3).

84. Doctors’ education levels vary by province and type of facility. While 100% of
doctors at level 1 district hospitals have standard university doctor training, only one fifth of
commune health station doctors have such training (Figure 2.3.4). The remainder were either
promoted from assistant doctors or were trained in less competitive “direct entry” programs
(see Box 2.3.1). Somewhat surprisingly, doctors at more sophisticated facilities have doctors who
on average have less experience (Figure 2.3.6). At district hospitals, almost half of doctors (47%)
have bachelor level training, 12% are primary specialists, and 36% are Specialists Level 1.
Extremely few doctors at district hospitals have higher qualifications beyond Specialists Level 1
(Figure 2.3.5.)

Figure 2.3.4: Percent of doctors with standard university doctor training

100%

83%

52%

28%
21%

District hospital District hospital District hospital Urban commune Rural commune
level 1 level 2 level 3 health station health station
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

58
 Figure 2.3.5: Health staff qualification at district hospitals
3.8% 0.1%
1.3%
Bachelor

Primary speacialist

35.8% 46.7% Speacialist level 1

Speacialist level 2

Master

12.2%
PhD

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

85. Surprisingly, experience levels are higher among doctors at commune health
station than those at district hospitals. In urban commune health stations, nearly all (87%)
of doctors have 10 or more years of experience. In the district hospitals that offer the most
comprehensive range of services (level 1 district hospitals), the fraction of doctors with this level
of experience is much lower (61%), and a substantial fraction (29%) has less than five years of
experience. Because commune health stations have low turnover of staff and are not expanding
over time, their doctors are largely those with longer tenure. District hospitals have expanded
over time, recruiting many young doctors and thus bringing down the average age of their
doctors.

Figure 2.3.6: Doctors’ clinical experiences in years across different levels of health
facilities

Urban Commune health station

Rural Commune health station
District hospital Level 3
District hospital Level 2
District hospital Level 1
0% 20% 40% 60% 80% 100%

Doctors' clinical experiences

< 5 years >=5 years & < 10 years >=10 years
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

59
86. There is wide variation in the amount of continuous medical education (in-
service) training that doctors receive (Figure 2.3.7). Roughly half of doctors at district
hospitals report having received some form of training in the previous 12 months, and the average
number of trainings received was 1-2, with small variation by province. Surprisingly, commune
health station doctors in most provinces receive much more training than district hospital
doctors, but with notable variation by province. Trainings average 3 or more per year among
commune health station doctors in all provinces with the exception of Dien Bien, where they
average one per year.

Figure 2.3.7: Training opportunities for doctors at district hospitals and

commune health stations by province

Percent of district hospital doctors having any Percent of commune health station doctors
training in the last 12 months having any training in the last 12 months
87%
81% 78% 79%
70%
62% 63% 60%
57% 56%
52%
48% 48% 45%

Dien Dak Lak Binh Dong Dong Rural Urban Dien Dak Binh Dong Dong Rural Urban
Bien Dinh Thap Nai Hanoi Hanoi Bien Lak Dinh Thap Nai Hanoi Hanoi

Average number of trainings that district Average number of trainings that commune
hospital doctors had in last 12 months health station doctors had in last 12 months
6.6

5.6

4.3
3.8
3.6
3.0
2.0
1.7
1.3 1.2 1.1 1.2 1.2
1.0

Dien Dak Binh Dong Dong Rural Urban

Bien Lak Dinh Thap Nai Hanoi Hanoi Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

60
87. Variation in doctors’ salaries reflects levels of education, experience, and
location. Table 2.3.1 shows results from a regression of salary (including allowances) on a variety
of characteristics. The results in the first column, without controls for province, show that
doctors who hold management positions, who have more experience, and who have qualifications
at the highest levels (specialist 1 and specialist 2) are paid more. There is no difference by gender
in salaries. Without controlling for location, ethnic minorities are paid 11% more than other
doctors. However, in the results in column 2, which include controls for province, this “ethnic
minority premium” disappears. This indicates that the apparent higher average salaries for ethnic
minorities are a reflection of the fact that they are chiefly in the Dien Bien and Dak Lak, where
doctors are paid greater allowances. Other regression results are similar in the analysis with and
without province controls.

Table 2.3.1: Regression of doctor’s total salary

Dependent variable: log of monthly total salarya
(1) (II)
Variable Coefficient Standard error Coefficient Standard error
Ethnic Minority 0.109*** 0.024 0.033 0.026
Female -0.024 0.016 -0.014 0.016
Holding management position 0.176*** 0.019 0.163*** 0.018
Years in clinical experience 0.031*** 0.003 0.029*** 0.003
Years in clinical experience
squared -0.000*** 0.000 -0.000*** 0.000
Highest qualification (base: less
than or equal bachelor degree)
Primary/Intern specialist -0.017 0.027 0.001 0.027
Specialist level 1 0.076*** 0.020 0.087*** 0.019
Specialist level 2, Master,
PhD 0.071* 0.040 0.125*** 0.041
Working at district hospital 0.026 0.021 0.004 0.021
Province (Base: Urban Hanoi)
Rural Hanoi 0.001 0.030
Dong Nai 0.118*** 0.032
Dong Thap 0.027 0.031
Binh Dinh 0.081** 0.032
Dak Lak 0.149*** 0.034
Dien Bien 0.184*** 0.037
Constant 15.150*** 0.030 15.110*** 0.038
Number of observation 920 920
R-squared 0.534 0.565
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Significance: *** p<0.01, ** p<0.05, * p<0.1; (a) monthly total salary includes monthly salary and allowances of
occupational risk and duty/management.

61
88. Work by doctors in private practice outside of their jobs at public facilities is
common, particularly among doctors in the more southern provinces: Binh Dinh,
Dong Nai, and Dong Thap. In each of those provinces, more than half of doctors at district
hospitals did some private practice work, and average time in private work across doctors
exceeded 11 hours per week. The prevalence of private work and average private hours are
lower at commune health stations. Very few doctors in Dien Bien conduct private practice (Figure
2.3.8). Such “dual practice” work by doctors is believed to be common in other countries in
South and East Asia, but hard data on its prevalence is scarce. The studies that have examined
the issue in other countries are typically very dated or rely on non-representative samples and
qualitative interviews (Hipgrave et al 2013).

Figure 2.3.8: Prevalence of private practice by doctors at district hospitals and

commune health stations

Percent doctors working at district hospitals Average number of hours (per week) that
and commune health stations performing district hospital and commune health station
private practices doctors spending on private practices

Urban Hanoi Urban Hanoi

Rural Hanoi Rural Hanoi

Dong Nai Dong Nai

Dong Thap Dong Thap

Binh Dinh Binh Dinh

Dak Lak Dak Lak

Dien Bien Dien Bien

0% 10% 20% 30% 40% 50% 60% 70% 0 5 10 15 20

Commune District Commune District

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

62
2.4. Doctors’ satisfaction

89. As noted earlier, at commune health stations, doctors’ overall levels of

satisfaction falls in the middle of a scale. We consider separate satisfaction results for
doctors in district hospitals and commune health stations by individual satisfaction items. District
hospital doctors express lower satisfaction with staff levels. Commune health station doctors
express lower levels of satisfaction with hospital security. (In news reports, there have been
numerous accounts of irate relatives of patients threatening or attacking doctors.) Doctors at
both types of facilities gave lower satisfaction scores for their salary and allowances and relative
high satisfaction scores to working environment and training opportunities.

Figure 2.4.1: Mean satisfaction scores of doctors working at district hospitals and
commune health stations

Working Pressure

Working environment

Training opportunity

Salary

Promotion opportunity

Occupational safety

Facility security

Availability of medicines

Availability of equipment

Allowance

Adequate staffing

1 2 3

Satisfaction score
Commune District
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: satisfaction score was calculated as the mean of ten satisfaction indicators including salary, allowance,
training opportunity, promotion opportunity, occupational safety, hospital security, working environment,
availability of medicines, availability of equipment, working pressure, having enough staff. These ten satisfaction
indicators were measured by a Likert scale with 5 level from very dissatisfied (1) to very satisfied (5).

63
 Chapter 3: Patient experiences
This chapter presents information on the health care experience from the patient’s perspective, based
on exit surveys conducted at commune health stations and district hospitals as part of the survey.
Analysis of the characteristics of patients confirm findings from other surveys that although the
socioeconomic profile of patients across the two facility types is not sharply different, for outpatient
care, ethnic minorities and those classified as poor are more prevalent at commune health stations than
district hospitals.

Most patients said they chose the facility for the visit because it was their primary facility registered for
health insurance purposes, although substantial numbers indicated that quality of care and health
worker attitudes drove their choice, particularly among patients at district hospitals. Ethnic minority and
poor patients at district hospitals tended to rate their own health more negatively than other patients
on average, which may reflect selection among which of those patients visit district hospitals. Travel
times for ethnic minorities and the poor to district hospitals are higher on average than those of other
patients, and they may tend to seek care at district hospitals only when their health status is
substantially worse. Travel times were under 20 minutes for nearly patients at commune health stations
and most at district hospitals. Extremely few patients have travel times of over 1 hour.
Among outpatients at district hospitals, waiting times are invariant to patient socioeconomic status,
averaging close to 33 minutes for all patients across wealth levels. Poorer patients at commune health
stations, however, do wait substantially longer than wealthier patients, and this difference persists even
after controlling for facility effects. In other words this pattern is not due to poorer patients going to
facilities with longer waits. Analysis of consultation times mirror results from those concerning the
composite effort index in Chapter 5: controlling for differences by facility, there is no difference in
consultation times by wealth or ethnic minority status. Those with health insurance, do receive slightly
shorter consultation times on average, however. The frequency of testing at district hospitals varies quite
substantially across provinces (commune health stations do not typically provide testing.) In rural Hanoi
hospitals, 74% of outpatients received at least one test, compared to 17% in Dong Thap.

Purchase of medicines outside the facility as a consequence of the visit was common for wealthier
patients at district hospitals. The most common reason patients cited for purchasing medicines off-site
was that specific medicines prescribed by the doctor were not covered by insurance.
Total out-of-pocket expenditures associated with the facility visit or stay were calculated by summing
three categories: i) expenditures to the facility, ii) expenditure for medical services outside the facility,
and iii) gifts, food, travel costs, and lodging for the patient and relatives, other than costs paid directly to
the facility. Average expenditures were low for outpatients with insurance at both commune health
stations and district hospitals. In particular, medical expenditures to facilities are minimal for children
under 6. Inpatients costs, for both those with and without insurance, were substantial. Notably, the
largest component of expenditure for inpatients with insurance was gifts, food, travel costs, and lodging.

3.1. Introduction

90. This chapter considers health care from the patient’s perspective. Drawing from
the exit interviews collected as part of the survey, it considers the demographic profile of the
patients, why they seek care, their reports of their experiences as patients, and their use of

64
 health insurance. The exit interviews were conducted at the end of inpatients’ stays and
outpatients’ visits. To ensure the high response and completion rates, the exit interviews were
designed to take less than 15 minutes. Response and completion rates were 90% for inpatients
and 80% for outpatients (82% among district hospital outpatients and 90% among commune
health station outpatients). A total of 948 inpatients and 4989 outpatients were interviewed at
district hospitals along with 1759 outpatients at commune health stations.

3.2. Patient profile

91. In considering the profile of patients, we distinguish between 1) district

hospital inpatients, 2) district hospital outpatients, and 3) commune health station
outpatients. The very small group consisting of commune health station inpatients were not
covered by the survey.

92. Three findings emerge from a comparison of basic demographic

characteristics across the three groups. First, relative to the overall population, patients in
all three categories are more likely to be children or older individuals, reflecting the greater health
demand of those groups. Second, ethnic minority outpatients are more prevalent at commune
health stations than district hospitals. However, substantial numbers of ethnic minorities are
found among inpatients at district hospitals. Third, paralleling the ethnic minority findings, among
outpatients the poor are more likely to receive care at commune health stations (see Figure
3.2.1.)

65
 Figure 3.2.1: Characteristics of outpatients at commune health stations,
outpatients at district hospitals, and inpatients at district hospitals

Age

Outpatients, commune 25% 44% 31%

Outpatients, district 33% 42% 25%

Inpatients, district 30% 45% 25%

Age: <15 Age: 15-59 Age: 60+

Poor versus nonpoor

Outpatients, commune 20% 80%

Outpatients, district 8% 92%

Inpatients, district 14% 86%

Poor Nonpoor

Ethnic minority versus ethnic majority

Outpatients, commune 17% 83%

Outpatients, district 8% 92%

Inpatients, district 19% 81%

Ethnic Minority Ethnic Majority

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: “General population” refers to the overall country population. (a) “Ethnic majority” consists of Kinh and
Hoa people. (b) The poor and nonpoor classification is that of the “poor list” of the Ministry of Labour, Invalids,
and Social Affairs, which is used to determine eligibility for various benefits. Patients were asked whether or not
they were classified as poor.

3.3. Reasons for Seeking Care at Facility

93. Patients have the freedom to choose their place of medical care, but those
enrolled in the social insurance system must designate a single facility as their
primary destination of care. At the time of the survey in 2015 patients who went to other
66
 facilities faced larger co-payments for their care.13 In the survey, patients were asked why they
chose a particular facility for their current care (on the day of the survey). Unsurprisingly, for all
three types of patients, by far the most common response was that the patient chose the facility
because it was registered as the patient’s primary place of care in the health insurance system
(Figure 3.3.1). Quality of care was the second most common response among both inpatients and
outpatients at district hospitals. Among patients at commune health stations, attitudes of health
workers was the second most common response, followed by quality of care. Overall, the results
indicate that while choice of facility is principally driven by health insurance registration, patients
are sensitive to the quality of care, particularly for district hospital visits.

Figure 3.3.1: Reasons of patients’ health facility selection

Inpatients at district hospitals Outpatient at district hospitals Outpatients at commune health

stations

63%
58%
45%

23% 20%
18%
7% 8% 12%
1% 3% 5%

Health Quality of Health Cost Health Quality of Health Cost Health Quality of Health Cost
Insurance care workers' Insurance care workers' Insurance care workers'
registration attitude registration attitude registration attitude

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: “health insurance registration” means that the health facilities were the place that patients registered their health insurance. This
figure shows responses to the question “Why do you seek care this health facility?”.

94. Surveyed patients were asked to give a subjective evaluation of their own
health, rating their health status from 1 (very bad) to 5 (very good). Among visitors to
commune health stations, 42% of patients overall rate their health as poor or very poor, and
similar figures are found for ethnic minorities and the poor. Among patients at district hospitals,

13
Nationally, 70% of the population has health insurance (based on analysis of the 2014 Vietnam Household Living
Standards Survey.) For basic treatment, the insurance is effective when the insured patients receive care from the
primary care facilities (commune health stations or district hospitals) or the higher level on referral. Insured people
have the option to register their health insurance at eligible commune health stations and district hospitals defined
by Ministry of Health (or provincial and central level hospitals for some specific groups such as senior civil servants.)
Before January 1st, 2016 if insured patients received healthcare services at other facilities, they paid higher co-payment
rates (30%, 50% or 60% at district, provincial, or central levels respectively instead of 20% at their health insurance
registered facilities.) As of January 1st, 2016, insured patients can seek care at any health facility at the same level with
no extra co-payment.

67
ethnic minority and poor patients were more likely than the general population to rate their
health as poor or very poor. This could be a result of selection in who visits district hospitals: for
the typical poor or ethnic minority patient living in a more remote area, travelling to a district
hospital involves greater cost, and thus such patients are more likely to make the journey only
when they have substantial illness or injury.

Figure 3.3.2: Percent of outpatients rating their health status bad or very bad on
the day of visit to the facility

Poor 48%
57%

Ethnic Minority 41%

55%

All patients 42%

43%

Commune District
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: “commune” refers to commune health stations and “district” refers to district hospitals. This analysis is for
outpatients at both district hospitals and commune health stations only because this data is not available for
inpatients.

3.4. Experiences from Patients’ Perspective

95. We consider several aspects of the patient experience: travel time, waiting
time, consultation time, and whether doctors provided instructions to patients. Each
of these outcomes can vary as a consequence of a number of factors, but other things being equal
a higher quality patient experience will have a shorter waiting time, a longer consultation time,
and include provision of instruction to patients. (Travel time is not determined directly by the
facility itself but is nonetheless included as an important element of the nonmonetary cost of
seeking care.) This analysis considers only outpatients, in both district hospitals and commune
health stations.

96. Mean patient travel times from home to commune health stations were 11
minutes, half that for district hospitals (considering only outpatients.) Travel times
were longest in Dien Bien and shortest in urban Hanoi. Similarly, for both commune health
stations and district hospitals, travel times were substantially less for Kinh and Hoa people
compared to ethnic minorities. Figure 4.4.1 shows a breakdown of the overall distribution of
times. While nearly visitors to commune health facilities and most to district hospitals travel less

68
than 20 minutes a substantial share of district hospital patients travel more than 20 minutes. Very
few travel more than an hour.

Table 3.4.1: Mean travel times from home to facilities for outpatients (in
minutes)
District Hospital Commune Health Station
Dien Bien 32 15
Dak Lak 21 12
Binh Dinh 21 11
Dong Thap 25 12
Dong Nai 22 10
Rural Hanoi 19 9
Urban Hanoi 15 8

Ethnic Minority 36 16
Kinh & Hoa 21 11

Total 22 11
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

Figure 3.4.1: Distribution of travel times from home to facilities for outpatients
(in minutes)

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
97. At district hospitals, waiting times before receiving care show very little
variation by patient wealth. Across wealth quintiles, patients wait on average for 33 minutes
before receiving care at district hospitals. At commune health stations, average waiting times are
lower overall but with variation by wealth of the patient. Patients in the lowest wealth quintile
face waiting times of 19 minutes on average, while those in the top quintile wait an average of 7

69
minutes (Figure 3.4.2). Further analysis shows that this not simply because poorer patients seek
care at facilities with longer waiting times. There are small differences in waiting times by wealth
within facilities. In a regression of waiting time on wealth index, controlling for facility using fixed
effects, an increase of one standard deviation of the wealth index is associated with 3 minutes
less waiting time (Table 3.4.2).14

Figure 3.4.2: Outpatients' average waiting time (minutes)

40

30

20

10

Poorest0 Richest
1 Outpatients,
2 commune 3 Outpatients,4 district 5

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (i) “outpatients, district” refers to “outpatients at district hospitals”, and “outpatients, commune” refers to
“outpatients at commune hospitals”. (ii) This analysis is based on outpatients’ data only. Waiting time of inpatient
data is not included in this figure.

14
The “poorest” and “richest” were defined as quintile 1 and quintile 5 determined by the wealth index of patients.
This wealth index was developed using the principal component analysis method based on the patients’ household
assets (washing machine, (bath) water heater, computer, refrigerator, gas/magnetic cooker, cell phone, electric
(rice/pressure) cooker, desk/chair/long bench/dressing table, motorbike, color TV.) The questionnaire designers
selected these ten assets from the durable list of Vietnam Living Household Standard Survey (2014) that were most
correlated to households’ per capita total expenditure. In this regression analysis, wealth index was used under
standardization form

70
 Table 3.4.2: Regression of outpatient waiting time
Dependent variable: waiting time (in minutes)
(Without facility fixed (With facility fixed
effects) effects)
Standard Standard
Variable Coefficient error Coefficient error
Have health insurance -0.535 1.858 -0.036 1.829
Wealth index -1.377*** 0.457 -2.634*** 0.485
Self-reported health status (base = normal)
Bad or very bad 0.939 0.855 -0.310 0.840
Good or very good -4.206 2.579 -5.178** 2.414
Ethnic Minority 3.351** 1.508 -0.063 2.003
Female 1.707** 0.838 0.598 0.774
Age 0.320*** 0.056 0.384*** 0.055
Age squared -0.002*** 0.001 -0.003*** 0.001
Seeking care at district hospital (base:
seeking care at commune health station) 22.490*** 0.961
Constant 2.249 2.103 19.170 1.992
Number of observation 6678 6678
R-squared 0.087 0.290
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (i) Significance: *** p<0.01, ** p<0.05, * p<0.1; (ii) This analysis is based on outpatients’ data only.

98. Consultation times are another rough measure of quality. Table 3.4.3 presents
results from a regression of consultation time on a variety of characteristics. Column 1 shows
results from the basic specification, column 2 adds controls for facility fixed effects, and column
3 includes doctor fixed effects. Without facility or doctor controls, longer consultation times are
observed for those without health insurance, those of higher wealth levels, those with worse self-
rated health, ethnic minorities, and patients who are female and older. The results for wealth,
ethnic group, and female are not statistically significant after controlling for facility, which suggests
that on average women, ethnic minorities, and the wealthy are more likely to go to facilities
where consultation times are longer. Even after controlling for facility, those with health insurance
have shorter consultation times.

71
 Table 3.4.3: Regression of outpatient consultation time
Dependent variable: consultation time (in minutes)
(Without fixed (With facility fixed (With doctor fixed
effects) effects) effects)
Standard Standard Standard
Variable Coefficient error Coefficient error Coefficient error
Having health insurance -1.352*** 0.128 -0.730*** 0.179 -0.673*** 0.171
Wealth index a
0.753*** 0.053 -0.035 0.047 -0.027 0.045
Self-reported health
status (base = normal)
Feeling bad/very bad 1.077*** 0.099 0.405*** 0.082 0.377*** 0.078
Feeling good/very good 0.257 0.300 -0.754*** 0.235 -0.620*** 0.224
Ethnic Minority 1.762*** 0.175 -0.207 0.195 0.002 0.188
Female 0.241** 0.097 0.105 0.075 0.061 0.072
Age -0.044*** 0.007 -0.016*** 0.005 -0.001 0.006
Age squared 0.001*** 0.000 0.000*** 0.000 0.000*** 0.000
Visiting district hospital
(base: visiting commune
health station) -2.255*** 0.112
Constant 7.558*** 0.247 5.607*** 0.195 5.338*** 0.197
Number of observation 6667 6667 6667
R-squared 0.114 0.517 0.576
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (i) significance: *** p<0.01, ** p<0.05, * p<0.1. (ii) As consultation time is specific for outpatients only, this analysis was
not performed for inpatients.

99. Next we examine the proportion of patients who received any tests during
their health care visits as outpatients at district hospitals. Inclusion of a test does not
necessarily indicate higher quality of care. We would expect the frequency of testing to be similar
across provinces, assuming a fairly homogenous profile of conditions. If that is the case, varying
rates of testing may reflect both undertesting (failure to test when tests are called for) and
overtesting (unnecessary testing.) The variation in testing rates is remarkable, both for testing
overall and when specifically considering x-rays. At the low end, in Dong Thap just 17% percent
of patients received any test, and 5% received an x-ray. At the high end, in rural Hanoi, 74% of
patients received a test, and 31% received x-rays.

72
 Figure 3.4.3: Percent of district hospital outpatients receiving any test or X-ray
test

74%
63% 66%
60%
56%

39%

17%
31% 28%
13% 19% 16% 16%
5%
Dien Bien Dak Lak Binh Dinh Dong Thap Dong Nai Rural Hanoi Urban Hanoi
Any test X-ray
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: This analysis was performed for outpatients at district hospitals only. Commune health stations generally
do not provide examination tests.

3.5. Experiences with health insurance

100. Nationally in Vietnam, 70% of individuals have health insurance. Among patients
in the survey, the fractions reporting that they used health insurance during their visit were
higher—82% of commune health station outpatients, 93% of district hospital outpatients, and 88%
of district hospital inpatients (Figure 3.5.1.) Slightly higher fractions of patients who are poor (and
thus qualify for free insurance) reported using health insurance.15

15
The large majority of patients who report that they have health insurance report that they used it during their
visit. Specifically, among those who have health insurance 99% of district hospital inpatients, 97% of district hospital
outpatients, and 90% of commune health station outpatients reported using their insurance during their visit. It is
unclear why a patient with health insurance might not use it during a visit. A patient with insurance but who had
left his or her insurance card at home might pay non-insurance costs rather than the retrieve the card, particularly
when visiting a commune health station, where fees are modest.

73
 Figure 3.5.1: Percent of patients using health insurance

Inpatients at district hospitals Outpatients at district Outpatients at commune

hospitals health stations

98% 95%
93% 93% 94% 90%
88% 88%
82%

All Ethnic Poor All Ethnic Poor All Ethnic Poor

patients Minority patients Minority patients Minority
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: The “all patients” category includes all three populations of patients including (i) outpatients at commune
health stations, (ii) outpatients at district hospitals, (iii) inpatients at district hospitals. The “poor patients” were
individuals from poor households classified by local authority.

Figure 3.5.2: Percent of patients buying medicine outside the health facilities
30%

Inpatient, district
20%
Outpatient, district

Outpatient, commune
10%

0%
1 2 3 4 5
Quintiles (1=poorest, 5=richest)
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: : “inpatients, district” refers to “inpatients at district hospitals”, “outpatients, district” refers to “outpatients
at district hospitals”, and “outpatients, commune” refers to “outpatients at commune hospitals”. This analysis was
only performed for patients who used health insurance.
101. Next we consider the extent to which patients buy medicine outside the
health facilities. Purchase of medicine outside of facilities implies a higher cost for patients,
because such medicines are generally not covered by health insurance. Inpatients are more likely
74
to buy medicines outside the hospital. This may reflect a greater tendency for inpatients to have
more complicated conditions which require medicines not available at the hospital pharmacy.
Across all three patient groups, wealthy patients are more likely to purchase medicines outside
the hospital.
102. Survey participants who had purchased medicines outside the facility were
asked why they had done so. The most common reason was that the prescribed drug was
not available in the facility for one of two reasons. In 49% of cases, patients responded that the
medicine was not on the list of medicines covered by insurance, and in 15% of cases they indicated
it was on the insurance list but out of stock (Table 3.5.1.)

Table 3.5.1: Reasons that inpatients at district hospitals bought medicines outside
Percent
Medicine was not on the “health insurance medicine list”a 49%
Medicine was in “health insurance medicine list” but out of stock 15%
Patient didn’t trust the quality of “health insurance medicine” 2%
Health workers requested patients to buy medicine outside 17%
without explanation
Other 21%
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: this analysis was performed for district hospital inpatients who used health insurances for their stays and
bought medicine outside. This data is not available for outpatients. The study inpatients can have multiple answers
(among the above options) for the question “Why you bought medicines outside the facility?”. (a) “health
insurance medicine list” is referred to the medicines that covered by health insurance in the hospitals.

3.6. Health expenditure

103. Next we consider expenditure associated with facility visits. Only information of
health expenditure paid out-of-pocket by patients was collected in the survey. For all studied
patients, total expenditures paid out-of-pocked consisted of three categories: (i) expenditure paid
to facility, (ii) expenditure for medical services outside the facility, and (iii) gifts, food, travel costs,
and lodging for the patient and relatives, other than costs paid directly to the facility.16 Table 3.6.1
shows the mean expenditure of outpatients and inpatients (showing those with and without
health insurance separately.) For outpatient care, expenditure at district hospitals were much
higher than those at commune health stations. Patients with insurance paid much less out-of-
pocket than those without insurance, but total out-of-pocket expenditures were still substantial
for hospital inpatients with insurance (Figure 3.6.1.) Notably, the largest component of
expenditure for inpatients with insurance was gifts, food, travel costs, and lodging.

16
Gifts include informal payments made to the doctor or health staff.

75
Table 3.6.1: Mean health expenditures by patient type and health insurance, in
thousands of Vietnam dong
Patients with Patients
health without
insurance health
insurance
Inpatients, district hospitals
Expenditure paid to facility 206 1561
Medical expenditure for services outside the facility 48 25
Gifts, food, travel costs, and lodging 545 679
Total expenditure2 803 2285
Outpatients, district hospitals
Expenditure paid to facility 12 157
Medical expenditure for services outside the facility 6 35
Gifts, food, travel costs, and lodging 16 33
2
Total expenditure 33 226
Outpatients, commune health stations
Expenditure paid to facility 1 34
Medical expenditure for services outside the facility 1 11
Gifts, food, travel costs, and lodging 2 3
Total expenditure2 4 48
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (1) “health insurance patients” refers to the patients who owed and actually used the health insurances in their
visits (for outpatients) or stays (for inpatients). For the patients who owed at least one health insurance but didn’t use
the health insurances in the visits/stays, they were considered as “non health insurance patients”. (2) “Total
expenditure” was the mean of summation of the expenditure components listed in the previous rows.

Figure 3.6.1: Average total out-of-pocket patient expenditures in thousands of

Vietnam dong

48.4
Outpatients, commune 3.6

225.7
Outpatients, district 32.8

2284.6
Inpatients, district 802.8

Non health insurance patients Health insurance patients

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (i) “inpatients, district” refers to “inpatients at district hospitals”, “outpatients, district” refers to “outpatients
at district hospitals”, and “outpatients, commune” refers to “outpatients at commune hospitals”. (ii) “Health
insurance patients” refers to the patients who owed and actually used the health insurances in their visits (for
outpatients) or stays (for inpatients).

76
104. Outpatient total expenditure are highly varied across provinces. Patients in the
wealthiest province (urban Hanoi) both with and without insurance spent substantially more than
patients elsewhere. The variation of total health expenditure was more notable among patients
without health insurance. While commune health station outpatients in Dien Bien with insurance
spent just 6000 VND on average, urban Hanoi patients spent almost 100,000 VND (Figure 3.6.2).
105. Among health insurance outpatients at both district hospitals and commune
health stations, direct expenditure to facilities varied substantially across provinces.
The charges were very low at commune health stations but were substantial for inpatients at
hospitals. Direct expenditures to facilities were highest in urban Hanoi for both inpatients and
outpatients at hospitals. Notably, health insurance hospital outpatients in Dien Bien, one of three
poorest provinces in the country, had to pay more than those in Binh Dinh, Dak Lak, and Dong
Thap.
106. Next we consider medical expenditures to the facility for patients less than six
years of age, who are registered at the visited facility. Care should have been provided
for free to such patients with presentation of their health insurance cards. Accordingly, medical
expenditures to the facility of commune health station outpatients under 6 were low. Outside of
rural Hanoi, average costs were less than 5,000 Vietnam dong. The same observation were also
found for district hospital outpatients under 6 in Dong Nai, Dong Thap, Binh Dinh. The average
expenditure to the facility for hospital outpatients in Dien Bien and Dak Lak was higher than in
the three other southern provinces. District hospitals in both urban and rural areas of Hanoi
charged patients under 6 the most (averaging 22,000 dong and 26,000 dong per patient.)

77
Figure 3.6.2: Patients' average total expendituresa (in thousand Vietnam dong)
by provinces
Health insurance inpatients at district hospitals Non health insurance inpatients at district
hospitals
5,235

2,494
1,985 2,054 2,081 2,133
1,721
967 1,174
652 699 864
588 565

Dien Dak Binh Dong Dong Rural Urban Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi Bien Lak Dinh Thap Nai Hanoi Hanoi

Health insurance outpatients at district Non health insurance outpatients at district

hospitals hospitals
470

268
236
147
123
95
34 37 51 52 29
28 17 26

Dien Dak Binh Dong Dong Rural Urban Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi Bien Lak Dinh Thap Nai Hanoi Hanoi

Health insurance outpatients at commune Non health insurance outpatients at

health stations commune health stations
86

57

40
37 34
23 23
10
6 3 2
0 0 0

Dien Dak Binh Dong Dong Rural Urban Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi Bien Lak Dinh Thap Nai Hanoi Hanoi

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: “Health insurance” patients refers to the patients who owed and used health insurance for their visit or
stay.

78
Figure 3.6.3: Average medical expenditures to facility by patients with health
insurance, by province

Inpatients at district hospitals

Urban Hanoi 888

Rural Hanoi 183
Dong Nai 323
Dong Thap 222
Binh Dinh 102
Dak Lak 95
Dien Bien 33

Outpatients at district hospitals

Urban Hanoi 25
Rural Hanoi 20
Dong Nai 19
Dong Thap 4
Binh Dinh 4
Dak Lak 10
Dien Bien 17

Outpatients at district hospitals

Urban Hanoi 1
Rural Hanoi 4
Dong Nai 1
Dong Thap 0
Binh Dinh 0
Dak Lak 0
Dien Bien 0

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: ‘Medical expenditure to facility” is the expenditure that patients (inpatients at district hospitals, outpatients
at district hospitals, outpatients at commune health stations) paid directly to the facilities. “Health insurance
patients” refers to the patients who owed and actually used the health insurances in their visits (for outpatients)
or stays (for inpatients). For the patients who owed at least one health insurance but didn’t use the health
insurances in the visits/stays, they were considered as “non health insurance” patients

79
 Figure 3.6.4: Medical expenditure to facility of children under 6 by provinces
(thousand Vietnam dong)

Urban Hanoi
Rural Hanoi
Dong Nai
Dong Thap
Binh Dinh
Dak Lak
Dien Bien

0 5 10 15 20 25 30
Outpatient, Commune Outpatient, District
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: “outpatients, district” refers to “outpatients at district hospitals”, and “outpatients, commune” refers to
“outpatients at commune hospitals”. This analysis was only performed for outpatients who were children under
6 utilizing their health insurances at the visited facilities.

3.7. Are patients satisfied with healthcare services?

107. Outpatients were also asked for their level of satisfaction with service received
at the facility. The fraction of patients answering that they were satisfied with their care were
high (72% among hospital outpatients and 85% among commune health station outpatients.)
Satisfaction rates were lower for district hospitals than for commune health stations in every
province. Patients were the least satisfied at district hospitals in rural and urban Hanoi. Service
satisfaction data is difficult to interpret because it reflects a combination of the care itself and the
patient’s perceptions and expectations. We can speculate that Hanoi district hospital patients
may be less satisfied because they have higher expectations, driven by the availability of higher
quality care at national and private hospitals in Hanoi.

80
Figure 3.7.1: Fraction of outpatients satisfied with health service during their visit
100%

80%

60%

40%

20%

0%
Binh Dinh Dak Lak Dien Bien Dong Nai Dong Thap Rural Hanoi Urban Hanoi

Outpatient, district Outpatients, commune

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: “outpatients, district” refers to “outpatients at district hospitals”, and “outpatients, commune” refers to
“outpatients at commune hospitals”. The analysis was not performed for inpatients at district hospitals and at
commune health stations because the data is not available. For the patients’ satisfaction, this indicator was
estimated using the data from the question “are you satisfied with the services of the facility for this visit?”. Among
five scales of this question (from “very dissatisfied” to “very satisfied”), one patient was considered as “satisfied
with the facility” if she answered “satisfied” or “very satisfied”.

81
 Chapter 4: How knowledgeable is your doctor?
An assessment of doctor ability

In this chapter, we provide an assessment of the clinical knowledge of 1,010 healthcare providers using
a series of structured medical vignettes covering five common outpatient conditions. We use item
response theory (IRT) to score providers on their use of history questions, physical examinations, and
diagnostic tests for each case. We directly evaluate diagnostic accuracy and the use of necessary,
harmful, and unnecessary medications for each case. The IRT scores are compared against case
management behaviors to understand how diagnostic knowledge and clinical management behaviors
are related in practice.
We find large inequities in diagnostic knowledge across providers, with the top half of the diagnostic
knowledge distribution completing substantially more of the appropriate history questions, physical
exams, and diagnostic tests for each case. The median provider asked 4.0 questions during each
vignette, while the top 20% of providers asked 5.8. Top providers are overwhelmingly concentrated in
district facilities as opposed to communes, and significantly lower levels of knowledge are found in
providers from minority ethnic groups and lower education levels. However, there are no systematic
differences in knowledge by geographic location or local poverty rates once these provider
characteristics are accounted for.

Correct diagnosis rates were above 75% for all cases at all levels of diagnostic knowledge, and providers
with more knowledge were more likely to give the correct diagnosis in all cases but hypertension.
Correct treatment was indicated in 52% to 97% of vignettes, depending on the case, once providers
were given the necessary information, but were only significantly correlated with diagnostic knowledge in
the diabetes case. The use of harmful treatment ranged from 9% to 68% depending on the case, and
providers with higher knowledge of the recommended diagnostic procedures were more likely to give
harmful treatments. The use of unnecessary treatments, by contrast, typically fell sharply as diagnostic
knowledge increased.

Although we are unable to directly determine the link between knowledge of history questions, physical
exams, and laboratory diagnostics and the quality of diagnosis or treatment due to the structure of the
vignettes exercise, this survey demonstrates a great diversity of diagnostic knowledge throughout the
health care system with real consequences. Even though they are much more likely to correctly diagnose
any given case, highly knowledgeable providers are as likely as less knowledgeable providers to believe
that harmful treatments are medically necessary. Therefore, we find on the one hand that increasing
knowledge has a strong positive effect on the diagnostic actions a provider knows how to take and a
small positive effect on the best case treatment, but on the other hand this knowledge is not linked to
any reduction in the use of harmful treatments or excess testing.

82
4.1. Introduction

108. This chapter provides an assessment of doctors’ knowledge, an important

determinant of the quality of care provided at health facilities. In this case, doctors are
assessed on (a) their knowledge of clinical guidelines and standards of procedure in terms of
history taking, physical examinations and laboratory tests, all of which are required to obtain
essential information to diagnose conditions, (b) then their ability to process patient information
from the guidelines to reach a correct diagnosis and (c) their knowledge of appropriate treatment
of cases. The focus of this assessment is on what doctors know, which represents the best
outcomes one can expect from doctors if they were to manage cases to the best of their
knowledge. Key findings include the following:

4.2. Measuring doctor’s ability

109. Doctors’ knowledge is assessed using medical vignettes for five conditions
selected to match the morbidity profile of Vietnam. The five conditions are child diarrhea
(acute diarrhea without dehydration), child pneumonia, tuberculosis, diabetes type II and
hypertension category 1. For each of these, doctors were presented with specific symptoms from
a carefully constructed script about a patient suffering from the condition (see Box 4.2.1) and
then requested to provide the specific questions they would ask or actions they would take at
various stages of the patient consultation process – namely history taking, physical examination,
lab testing, diagnosis and treatment – in sequence. Answers to the relevant questions/actions
from each stage are given to the doctors before they are asked to provide questions/actions for
the next stage. The description of vignettes are provided in Table AD.1, Annex D. Such vignettes
have been employed in several studies (see Peabody et.al. 2004 and Connor et.al. 2014) and found
to be a good proxy of the quality of care (18; 19). However, it should be kept in mind that doctors
knowledge is being measured, not what they actually do (see next chapter) and thus results from
this chapter should be interpreted in this context as they may be biased towards overestimating
the care doctors would provide (see Shah, et.al, 2010) (20). The vignettes were administered to
1010 doctors in total. Of these 749 were from district hospitals and the other 261 form commune
health centers. A maximum of 10 doctors were randomly selected for the administration of the
medical vignettes from a roster of doctors at district hospitals and 1-2 doctors/assistant doctors
at health commune facilities.
110. Doctors’ performance was evaluated using their responses to the medical
vignettes, comparing them to the set of clinical guidelines for the management of
these cases in Vietnam. The clinical guidelines outline a set of essential history questions,
physical examinations, mandatory laboratory tests and protocols for treatment that doctors
should follow when dealing with a patient suffering from each of the five conditions. A score of 1
if recorded, for each item (i.e. history question, physical examination, test or recommended
drugs) in the clinical guidelines that a doctor proposed, otherwise a zero is recorded. Thus
doctors were assessed on whether they would ask or perform these essential questions or

83
actions, request mandatory laboratory tests, prescribe the correct treatment and whether given
all the essential information, they could correctly diagnosed the condition.
Box 4.2.1. Medical Vignettes
Acute diarrhea without dehydration: The mother of a 15-month old child takes the infant
to the clinic as her child has had diarrhea for two days and the condition does not go away after
the child took medicine at home.
Child pneumonia: A 3.5-year child has cough and fever for three days and has been given
medicines bought from a private drugstore but did not get better. The mother takes him in for
examination and care.
Tuberculosis: A 37-year old male patient, with sporadic cough and fever for the last three
weeks, fatigue and weight loss, came in for check-up. The patient said that sometimes he had mild
fever during the day.
Diabetes (Type 2): A 58-years woman has signs of fast weight loss recently (3kg loss within 2
months) and frequently feel energy-less hungry despite eating more than normal. She doesn’t
know why, so she came for a check-up.
Hypertension (Category 1): A 65-year male patient, who sometimes has headache and
burning face, came to your clinic for check-ups.

111. An aggregate indicator of doctors’ ability was generated using item response
theory (IRT) to produce a ranking or distribution of doctors by levels of ability. The
IRT methodology uses maximum likelihood methodology to estimate the underlying "ability
score" of providers based on their performance during the medical vignettes exercise. This score
reduces the history questions and examinations behavior to a single metric that is comparable
across providers. It quantifies their propensity to ask the history questions and perform the
physical examinations that were graded as minimum or essential by the expert committee.
112. IRT also assesses the characteristics of each of the essential behaviors
themselves. Each item receives three estimated parameters that together shape the "estimated
response curve", or the way in which average performance on a particular item improves with
ability. The first parameter is the item's discrimination power, which is its ability to distinguish
between high ability providers and low ability providers (questions or actions that high ability
doctors are likely to ask or take but which low ability doctors are unlikely to). The second
parameter is the difficulty of the question, or the ability level at which the item is usually mastered.
The third parameter is the guessing rate, or the expected probability of correct behavior for an
individual of the lowest ability level. Combining these parameters with the providers' actual
behaviors on the items allows for a reliable estimate of their underlying ability levels with respect
to the vignettes questioning and examination.
113. The composite indicator of ability necessarily correlates with the number of
questions and exams correctly performed, but it is not guaranteed to predict the
diagnostic accuracy or treatment quality of the providers. The structure of the vignettes
makes it hard to draw a link between the questioning and the diagnosis or treatment outcomes.
All doctors were only provided with the same information prior to being asked to provide a
diagnosis, since at that point, only the key history and physical examination information and key

84
test results contained in the vignette was provided regardless of whether they asked more or
less questions and tests at each stage. Thus the measured indicator or ability should have no
correlation with the diagnostic accuracy, as in the vignette setting, a high-ability provider does
not necessarily have more information than a lower-ability provider would have at the diagnosis
and the treatment stages. Any differences in diagnosis or treatment may only serve to reflect the
inability of low ability doctors to even make sense of the basic information provided to them.
Otherwise because of the structure of the vignette, it is not possible to look at the entire link
from checklist to diagnosis to treatment as in Das and Hammer (2007). In fact, each of these
vignettes can be conceptually treated as different cases, although the basic information and
examination can be put together in the checklist to create a composite score as we do.

4.3. Main findings

114. Doctors are generally able to interpret information generated from clinical
guidelines and reach a correct diagnosis, but their knowledge of these guidelines is
limited. Most doctors can give an accurate diagnosis once they have the necessary patient
history and physical examination information typically generated from following clinical guidelines.
Four of the five cases tested in this study were each correctly diagnosed by more than 70 percent
of doctors after they were presented with essential patient history and physical examination
information (see Table 4.3.1). The exception was acute diarrhea, which an overwhelming majority
of doctors (81 percent) could only partially correctly diagnose. Many diagnosed the case as “acute
diarrhea, level A dehydration” (as per the outdated guidelines) or “acute diarrhea” instead of
acute diarrhea without dehydration in accordance to the revised guidelines in Decision no. 4121
of Ministry of Health. Overall, close to half (48 percent) of the doctors accurately diagnosed 4 of
the 5 cases each and a third correctly diagnosed 3 of the 5 cases. Close to 90 percent of the
doctors either correctly or partially correctly diagnosed 4 of the 5 cases presented to them. Thus
doctors are highly likely to correctly diagnose cases if clinical guidelines are followed to obtain
relevant information.

85
Table 4.3.1: Diagnostic accuracy by condition Figure 4.3.1. Distribution of
doctors by number of cases
correctly diagnosed
Diagnosis Accuracy 56
Condition Statistic Partial Incorrect Correct
or full Partially Fully
48
Correct Correct correct
Child Share 88 81 6 12
Diarrhea (%)

Share of doctors
36
Child Share 97 2 95 3 33

Pneumonia (%)
Share 92 2 90 9
Tuberculosis (%)
Share 86 14 72 14
Diabetes 11
(%) 9
Type II
2 3
0 0 1
Hypertensio Share 81 6 75 19 0
n Category I (%) None One Two Three Four All five
case cases cases cases cases
only
Number of cases correctly
diagnosed

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: the shares were calculated for the sample of 1,010 doctors.

115. The challenge for doctors was in their low knowledge of the right history
questions to ask and physical examinations to do in accordance to the clinical
guidelines. On average, doctors asked less than half of the essential history questions in each of
the 5 cases presented to them. Only for acute diarrhea (58 %) and hypertension (50%) did doctors
ask at least 50 percent of the necessary physical examinations on average. In this respect
knowledge of clinical guidelines (SoPs) is modest (see Figure and Figure 4.3.3).

86
Figure 4.3.2. Median number of history Figure 4.3.3. Median number of
questions asked by condition physical examination actions by
condition
Hypertension

Diabetes Type 2

Tuberculosis

Child Pneumonia

Acute Child Diarrhea

0 5 10 15
Number of history questions asked 0 2 4 6 8 10
Number of questions
Median number of questions asked Median number actions taken
Number of recommended questions Number of recommended actions
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
116. Indeed, the variation in the knowledge of the clinical guidelines on history
questions and physical examinations between low ability doctors and high ability
doctors was high (Figure 4.3.4). To give an example, the top 20 percent ability doctors on
average asked between 7 (for Child Pneumonia) and 12 necessary questions per case, compared
to an average ranging from 3 to 6 questions for the bottom 20 percent. The gap between high
ability and low ability doctors was wider on physical examinations where doctors in the highest
ability quintile asked more than twice as many necessary physical examination questions when
compared to doctors in the ability lowest quintile.

Figure 4.3.4. Number of history question and physical examinations asked by doctors
ability

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: 95% confidence intervals shown

87
117. Knowledge of proper treatment of cases is a challenge, even when the
diagnosis is known. A plurality of doctors could only propose full treatment for just one or
two of the five cases after being presented with both the diagnosis and other essential patient
information about the case. A significant number also offered harmful treatment (see Figure 4.3.5).
Just 2 percent and 4 percent of doctors proposed full treatment for diabetes type II and
hypertension category one – the two cases where doctors were much more likely to prescribe
partially correct treatment (63 and 73 percent respectively) and also prescribe harmful treatment
(Table 4.3.2). Most doctors proposed harmful treatment for diabetes type II by prescribing a
combination of two anti-hyperglycemic medicines and by prescribing two antihypertensive
medicines for hypertension category 1. Knowledge of proper treatment was very low for child
pneumonia for which only 52 percent proposed any correct treatment, be it full or partial, despite
child pneumonia being correctly diagnosed by 95 percent of the doctors. Nearly all doctors
prescribed ORS for diarrhea, but a quarter of them offered harmful treatment by prescribing
antibiotics. Overall, most doctors could only prescribe fully treatment for just child diarrhea (58
percent) or child diarrhea and TB (58 percent). None of the doctors offered full treatment in all
five cases.

Table 4.3.2: Share of doctors prescribing treatment by condition and treatment

suitability
..of which Any
Any Any
Condition Statistic Partly Fully harmfu
correct other
correct correct l
Child Diarrhea Share (%) 96.7 0.1 96.6 25.5 69.7
Child Pneumonia Share (%) 52.2 35.6 16.5 7.9 18.9
Tuberculosis Share (%) 61.8 - 61.8 4.0 2.8
Diabetes Type II Share (%) 65.4 63.4 2.1 28.7 1.9
Hypertension
Share (%)
Category I 78.2 73.4 4.8 30.1 11.6
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: the shares were calculated with the whole sample of 1,010 doctors.

118. Even high ability doctors also proposed harmful treatment. For example, nearly
a third of doctors in the highest ability quintile proposed harmful treatment for diabetes type II
and hypertension category I while more than 20 percent of these top ability doctors also offered
antibiotics for child diarrhea. No statistically significant difference in prescription of harmful
treatment is observed between low and high ability doctors in all cases, suggesting the
understanding of clinical guidelines on treatment is a challenge faced by doctors at all levels of
ability.

88
 Figure 4.3.5. Distribution of doctors by number of cases correct or harmful
treatment is prescribed
60.0

50.0
Share of doctors

40.0

30.0

20.0

10.0

0.0
Harmful treatment Fully correct treatment Supplemental treatment
Number of cases where type of treatment is proposed

None 1 case 2 cases 3 cases 4 cases All 5 cases

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

119. Doctors rely excessively on tests. Just 6 percent and 3 percent of doctor’s correctly
indicated that no test would be required for child diarrhea and child pneumonia respectively if
even though tests are not required to reach a correct diagnosis of these conditions. Instead, most
doctors (around 95 percent) recommended optional tests in each of these cases (see Table 4.3.3),
with a median of 3 tests being recommended for child diarrhea and 2 tests for child pneumonia.
Doctors recommended more tests for TB (average of 4 tests), diabetes type II (5 tests) and
hypertension (5 tests, see Figure 4.3.6). The mandatory tests for TB (at least one of AfB, TB
culture or straight/side cardiopulmonary X-ray) were recommended by nearly everyone (98
percent of doctors). Of these, the Straight X ray was the easiest or most common, with almost
all doctors in district hospitals likely to request it, followed by Afb test, whose likelihood of being
asked was greater than 80 percent even among low ability doctors and equally likely to be
requested in both district and commune facilities. No one requested for a TB culture test. About
90 percent of the doctors recommended a blood glucose test for the case of diabetes type II,
mainly in additional to other optional tests. However, only 64 percent of doctors recommended
at least 3 of the mandatory tests for hypertension, even though doctors recommended more
than 5 tests for this case on average.

89
Table 4.3.3: Share of doctors recommending Figure 4.3.6. Average number of tests
tests by condition and importance of tests requested by condition
Share requesting category
of tests
Hypertension
Condition Statistic Mandatory Optional Other
Child Share
Diarrhea (%) 6a 94 57
Diabetes Type II
Child Share
Pneumon (%)
ia 4a 95 36 Tuberculosis
Tubercul Share
osis (%) 98 82 32
Diabetes Share Child pneumonia
Type II (%) 90 92 47
Hyperten Share
sion (%) Child diarrhea
Category
I 64 74 25 0 1 2 3 4 5 6
Number of tests
Median number of tests

Average number of tests

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: The shares were calculated with the whole sample size of 1,010 doctors. (a) No mandatory tests required for Child
Diarrhea and Child Pneumonia.

120. Doctors in district hospitals were more likely to ask for mandatory tests than
commune doctors at all levels of the ability scales, except at the very top ability
levels. At all ability levels, doctors in district hospitals favored requesting for a complete blood
count test, whether it is essential or not. At least 88 percent of doctors in district hospitals asked
for this test in the cases of child diarrhea (91 percent), child pneumonia (94 percent) and TB (89
percent), while close to half requested for the test in the cases of diabetes (45 percent) and
hypertension (49 percent). For these last two cases, district hospitals – especially at high ability
levels - were more likely to use a blood lipids tests, which were rarely requested by commune
health centers doctors, especially those at the low ability scale.
121. Low ability doctors have limited knowledge of diagnosis and management of
chronic conditions, beginning from low knowledge of recommended guidelines form
history and physical examinations and how to interpret obtained information. Besides
checking for vital signs, high ability doctors were more likely to ask or undertake essential history
questions or physical examinations, particularly for diabetes type II. Figure 4.3.1 shows other
history questions or physical examinations (other that vital signs), that distinguished low and high
ability workers. The difference between the shares of low and high ability asking these questions
were relatively smaller (even though still significant) for the two child illness conditions when
compared to the other three cases. The largest differences were observed on diabetes type II. In

90
 this case, high ability doctors were much more likely to ask for an assessment of the patient’s
weight and height, check for numbness in limps and ask for the history of diabetes in the patient’s
family. Less than 10 percent of doctors in the lowest ability quintile for asked the first 3 of these
questions, compared to at least 55 percent in the highest ability quintile who did. In contrast, the
discriminating essential physical examination for diarrhea included observation of sunken eyes
and asking about the child’s vomiting, which respectively, 45 percent and nearly 40 percent of
doctors in the lowest ability quintile asked.

Figure 4.3.1. Share of doctors asking questions by Figure 4.3.2. Diagnostic accuracy by
doctors’ ability level doctors’ ability levels
Straight cardiopulmonary X test
Hypertension

1.0
Exam peripheral blood vessel
Family history 0.9
Headace along with vomitin 0.8
Blood lipid panel test
Diabetes Type 2

0.7
Measure height
Measure weight 0.6
Diabetes in family
0.5
Numbness in hand/food
Complete blood count test 0.4
Tuberculosis

Percus to check pulmonary 0.3

Exam pleural fremitus
0.2
Family history
Sputum's color 0.1
Observe nasal flaring
Pneum
Child

Diarrhea onia

0.0
Observe rib cage
Bottom 20%

Bottom 20%

Bottom 20%

Bottom 20%

Bottom 20%
Highest 20%

Highest 20%

Highest 20%

Highest 20%

Highest 20%
Observation of sunken eyes
Acute

Eat/feed well
Vomit Acute Child Tuberclousis Diabetes Hypertension
Diarrhea Pnemonia Type 2
Ability quintile 5th 0
Ability quintile 1st 20 40 60 80 100
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: 95% confidence intervals shown.

122. Lower ability doctors were also significantly less likely to correctly diagnose
cases of diabetes type II and hypertension, even with the knowledge of the relevant
patient history, physical examination and test results (Figure 4.3.2). Thus high ability
workers both knew the recommended questions and actions to ask and were able to process
this information to reach a correct diagnosis when compared to low ability doctors. In addition,
a significantly larger share of high ability doctors (90 percent among the top ability quintile)
recommended at least 3 of the 8 mandatory tests for hypertension compared to low ability
doctors (only 38 percent among the low ability quintile). Likewise, a statistically significant
difference in the likelihood recommending the mandatory test for diabetes type II was also
observed between low and high ability doctors (77 percent of doctors in the bottom quintile

91
compared to 98 percent of those in the top quintile). Even the optional tests for these conditions
were more known to high ability doctors.
123. In fact, diabetes type II and hypertension are the only two cases where the
likelihood of correct diagnosis significantly increased with doctors’ ability. However,
once aware of the diagnosis, no statistically significant differences are observed in the treatment
offered by low and high ability workers. It seems the most difficult aspect for low ability workers
in terms of managing the two chronic conditions is low knowledge of clinical guidelines for history
taking and physical examination for these cases and interpreting the information that following
these guidelines reveals in order to reach a correct diagnosis.
124. There are significant differences in doctors’ ability between commune and
district hospitals and between poor and well to do areas. There are proportionately
more high ability doctors in district hospitals than in commune health centers (see
4.3.9). Estimated ability for three
in every five doctors in commune Figure 4.3.9. Probability of being in a district
hospital by doctors ability level
health centers was below the
national median, and only 7
percent of doctors working in
commune health centers were
among the top 20 percent in
terms of ability. The share of
recommended history and
physical examination questions by
doctors in commune health
facility is statistically significantly
and consistently lower than the
share of questions asked by
doctors in district hospitals
(Figure 4.3.10). Knowledge of Source: Calculations from the Vietnam District and Commune Health
Facility Survey (2015).
clinical guidelines for patient
consultations for these cases is not only lower for the chronic conditions, but for child illnesses
which they are routinely expected to manage at that level too.
125. While some differences in treatment are observed, no statistically significant
differences between doctors in commune and district facilities are observed when it
comes to providing correct diagnosis once the key history and physical examination
information is provided to the doctors. The diagnosis of diabetes type II is the only
exception (Table 4.3.4). In all other cases, commune health facilities doctors were equally likely
to reach a correct diagnosis as doctors in district hospitals. However, 69 percent of doctors in
district hospitals prescribed some correct treatment for diabetes type II compared to 54 percent
of doctors in commune health centers. district hospitals doctors were also less likely to offer
harmful treatment for TB (by 4 percentage points) but they were more likely to prescribe harmful

92
 treatment for acute diarrhea (9.1 percentage points difference) and hypertension category 1 (6.4
percentage points higher) than doctors in commune health centers.

Figure 4.3.10. Share of recommended Table 4.3.4: Diagnostic accuracy by doctors’

history and physical examinations by facility type
doctors’ facility type
Facility type
Condition Variable district CHCb Differencec
Hypertension

Physical Examination
hospitala
Correct
5.9 7.3 -1.4
History questions diagnosis
Acute Any correct
96.9 96.2 0.8
Diarrhea treatment
Diabetes Type II

Physical Examination Any harmful

27.9 18.8 9.1***
treatment
Correct
95.5 93.9 1.6
History questions
diagnosis
Child Any correct
52.2 52.1 0.1
Pneumonia treatment
Physical Examination
Any harmful
Tuberculosis

8.3 6.9 1.4

treatment
Correct
90.4 88.5 1.9
diagnosis
History questions
Any correct
Tuberculosis 62.1 60.9 1.2
treatment
Child Pneumonia

Any harmful
Physical Examination 2.9 6.9 -4.0**
treatment
Correct
74.2 66.3 8.0**
diagnosis
History questions
Diabetes Any correct
69.4 54.0 15.4***
Type 2 treatment
Any harmful
Acute diarrhea

Physical Examination 28.8 28.4 0.5

treatment
Correct
75.8 74.3 1.5
diagnosis
History questions
Hypertensio Any correct
77.7 79.7 -2.0
n treatment
Any harmful
Commune District 0 10 20 30 40 50 60 31.5 26.1 5.5*
treatment
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: (a) district hospital –district Hospital (b) CHC –commune Health Center, (c) *- p-value < 0.1; **-p-value < 0.05; ***-p-value
< 0.01.

126. Poor areas are served with relatively more low ability workers than well to do
areas, as evidenced by the disproportionate share of doctors working in the poorest
40 percent areas who are at the bottom of the ability distribution (Figure 4.3.11). For
instance, nearly half (47 percent) of doctors working in areas in the poorest poverty quintile are

93
in the bottom 40 percent of the national ability scale, compared to only 37 percent of doctors
working in the richest 20 percent areas. Compared to doctors in the richest 20 percent areas,
those working in the poorest areas (highest poverty quintile) were 12 percentage points less
likely to correctly diagnose diabetes type II, even with test results and patient history and physical
examination information provided to them, but no statistically significant differences in diagnostic
accuracy of other conditions is found (Figure 4.3.12a). Only 44 percent of them prescribed any
correct treatment for TB, compared to 70 percent of doctors in the richest 20 percent areas
(see Figure 4.3.12b). However doctors in poor arears were most likely to prescribe a correct
treatment of child pneumonia (about 60 percent of doctors in the poorest 20 percent) than
doctors in the richest 20 percent areas (45 percent). Doctors in poor and well off areas were
statistically equally likely to prescribe harmful treatment for all conditions however.

Figure 4.3.11. Comparison of distribution of doctors ability by poverty

quintile
100
90
80
Share in ability quintile

70
60
50
40
30
20
10
0
Lowest poverty 2nd Quintile 3rd Quintile 4th Quintile Highest poverty
quintile quintile

Ability quintile Bottom 20% Ability quintile 2nd Ability quintile 3rd
Ability quintile 4th Ability quintile Highest 20%
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

94
Figure 4.3.12. Share of doctors making a correct diagnosis and prescribing any
correct drugs by facility area poverty quintile
(a) Share of doctors making a correct diagnosis by case and poverty quintile

1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1st 3rd 5th 1st 3rd 5th 1st 3rd 5th 1st 3rd 5th 1st 3rd 5th
Acute Diarrhea Child Pnemonia Tuberclousis Diabetes Type 2 Hypertension

(b) Share of doctors prescribing any correct treatment by case and poverty quintile

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0
1st 3rd 5th 1st 3rd 5th 1st 3rd 5th 1st 3rd 5th 1st 3rd 5th
Acute Diarrhea Child Pnemonia Tuberclousis Diabetes Type 2 Hypertension
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: 95% confidence intervals shown.

95
127. Of the six provinces in Figure 4.3.13. Comparison of distribution of
the sample, the highest ability doctors ability by province
doctors can be found in Ha Noi
and Binh Dinh. Doctors in other
provinces like Dien Bien, Dong Nai,
and Dong Thap show substantially
lower ability. Not only were the
median ability levels substantially
lower in Dak Lak and Dien Bien, but
even the ability of their best
performers were substantially lower
than the best of Ha Noi and Binh
Dinh (see Figure 4.3.13). The
estimated average ability of the top
25 percent performers in Dien Bien
Source: Calculations from the Vietnam District and Commune
was 40 percent lower than the top
Health Facility Survey (2015).
25 percent performers in Ha Noi for
example.
128. Medical training received is the primary determinant of doctors’ ability and to
a large extent accounts for the variation in doctors ability between facility types and
between poor and well to do areas. Doctors with advanced medical training have a higher
estimated levels of ability. The estimated ability of doctors with intermediate training is less than
a fifth of the estimated ability of doctors with a medical degree (see Figure 4.3.14). As generally
established for doctors at the low ability scale, a particular weakness of doctors with an
intermediate qualification was in the diagnosis and treatment of diabetes type II for which these
doctors were substantially less likely to correctly diagnose the condition or prescribe any correct
treatment.

96
Figure 4.3.14. Average estimated ability level Figure 4.3.15. Profile of doctors education
by individual doctor’s characteristics by facility type and location characteristics
0.4 100
0.2
90
0.0
-0.2 80

Share of doctors
-0.4 70
-0.6 60
-0.8 50
-1.0 40
Specialists or… -1.2
30
Male

0 -2
Other ethnic groups

16 or more
Bacherlors

8-15
Intermediate

Primary / Intern specialist

3-7
Female

Kinh

20
10
0

Dong Nai
District

Dong Thap

Second poorest
Middle
Binh Dinh
All

Ha Noi

Lowest poverty

Second highest
Dak Lak
Dien Bien
Commune

Highest poverty
Facility type Province Poverty quintile
Gender Ethinicity Highest medical Clinical Intermedi Bacherlor
qualification experience Primary /Intern Sopecialist Specialists or postgraduate
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: 95% confidence intervals shown.

129. District hospitals, well-off areas and Ha Noi, all have relatively more doctors
with advanced medical training which to a greater extent, accounts for the observed
higher performance of doctors in these areas (see Figure 4.3.15). About 40 percent of
doctors in commune facilities have an intermediate qualification, compared to none in district
hospitals, which have more doctors with specialist training instead. Similarly, facilities in areas
with the lowest poverty rates have more specialist doctors (40 percent) compared to the poorest
areas (20 percent), while both Binh Dinh and Ha Noi – with the best doctors in terms of doctors
ability – also have the highest share of specialist doctors. Once the education profile of doctors
is taken into account, no statistically significant correlation is found between doctor’s ability and
both facility type and the rate of poverty the facilities is located in (see Table 4.3.5). Thus doctors
with similar level of medical training are found to have similar levels of ability on average whether
they work in commune or district facilities and poor or non-poor areas.

97
 Table 4.3.5: Correlates of doctors’ ability
Standard
Variable Coefficient
Error
Gender 0.097 0.075
Age -0.095 0.060
Age squared 0.001 0.001
Ethnic Minority -0.413*** 0.141
Years in clinical experience 0.048** 0.024
Years in clinical experience squared -0.001* 0.001
Tertiary education branch 0.036 0.106
Highest qualification (Base) intermediate)
Bachelor’s degree 0.663*** 0.171
Primary / Intern specialist 0.490** 0.229
Specialists or postgraduate degree 0.825*** 0.176
district hospital 0.097 0.141
Rural -0.088 0.141
Poverty quintile (Base: Poorest Quintile)
Second -0.006 0.102
Third -0.047 0.127
Fourth -0.252 0.179
Fifth 0.059 0.171
Province (Base: Ha Noi)
Binh Dinh 0.047 0.130
Dak Lak -0.260 0.181
Dien Bien -0.054 0.162
Dong Nai -0.139 0.135
Dong Thap -0.097 0.119
Constant 1.255 1.174
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

130. Differences in education qualification does not explain all the differences in
ability however. Clinical experience has a non-linear relationship with doctor’s ability that
suggested that ability increases with experience only up to a certain point (implied peak of 24
years) after which experience matters little. Intriguingly, ethnic minority doctors are found to
have low estimated ability levels (see Figure 4.3.14), even after controlling for the level of
education (Table 4.3.5). This could be an indication of unobserved heterogeneity. Ethnic minority
doctors may have attended less prestigious schools where the quality of medical training is lower
for example. Unfortunately not enough data is available in the survey to test this hypothesis.

98
 Chapter 5. Knowledge and practice: clinical observation and
the know-do gap

In this chapter, we evaluate the clinical behaviors of 385 healthcare providers in actual patient
interactions. Using all-day clinical observations and matched patient exit interviews, we produce an
effort score based on the providers’ time spent with the patient and use of questions and examinations.
While clinical observation does not allow direct assessment of the appropriateness of treatment and
testing decisions, we assess how clinical effort responds to patient characteristics, how it varies with
provider knowledge as demonstrated in the vignettes exercise, and how workload affects providers’
clinical effort. We measure “effort” using a composite index based on three variables collected during
the survey: consultation time, the number of questions asked, and the number of physical examinations
completed.
Vietnam’s highest-effort providers are on par with Paraguay, and its lowest-effort providers are on par
with India. Our analysis suggests that 52% of that variation is explained by community, facility, provider,
and patient characteristics, and we also observe highly equitable clinical effort. Patients receive similar
effort at the systemic, facility, and provider level regardless of their wealth, gender, age, or ethnicity.
However, the provision of unnecessary care seems to be endemic. Medication use is high, with 83% of
patients receiving medication and 38% receiving an antibiotic. The use of diagnostic testing is also high
at district facilities, with 51% of patients being ordered a laboratory diagnostic.

We observe large differences in clinical practice between district and commune facilities, even once
caseload and location are accounted for, with communes exerting significantly more effort. Despite
lower levels of clinical knowledge, communes perform closer to their knowledge levels and exhibit
similar or better actual quality of care among two common cases. We investigate whether these
differences are due to “effort conservation” among higher-knowledge providers (who tend to be located
in district facilities) and find instead that more knowledgeable providers systematically exert higher
effort. We also investigate whether caseload is a limiting factor on the effort providers allocate to
patients. We find that providers reduce effort substantially as caseload increases; however, this
reduction in effort occurs well before providers appear to be constrained by the duration of the
workday.
The combination of low caseloads at commune facilities and high use of medication and diagnostics at
district facilities – even though we find they are typically geographically proximate – suggests that costs
are poorly contained across the system once salaries, insurance payments, and patient expenditures are
taken into account. We observe substantial excess capacity throughout the system, especially at
commune facilities where providers exert somewhat more diagnostic effort and cannot access
laboratory tests. These findings are promising for the overall equitable availability of capacity throughout
the Vietnamese health care system, while simultaneously drawing attention to barriers to the efficient
use of resources that have already been invested in.

99
5.1. Introduction

131. This chapter analyzes the actual practice of doctors based on direct
observation of patient care. It examines a measure of provider “effort”, considers variation
in patterns of effort across geography, patients, and types of patients, and considers possible
explanations for the relatively low effort observed in district hospitals. It also assesses the
correctness of observed treatment for specific conditions and the interrelationship between
effort, ability as measured using the vignettes, and correctness of treatments.

5.2. Measuring Effort

132. What happens when a real patient visits a provider? To what extent does medical
knowledge demonstrated in hypothetical interactions like those examined in the previous chapter
translate to practice with real patients? To examine this question, we draw from an extensive
database of clinical observations, with over 8,000 patients observed at both district hospitals and
commune health posts. We measure “effort” using a composite index based on three variables
collected during the survey: consultation time, the number of questions asked, and the number
of physical examinations completed.
133. To measure clinical effort, providers were observed in their place of practice
over the course of a full day, with an individual record for each patient who sought
treatment. As patients exited, they were interviewed by enumerators outside the facility, who
asked the patient about their symptoms, basic demographic information, and other questions
about the interaction with the provider so they could be matched back to the clinical observation
record. This produced a record of 1,961 observed interactions at commune health posts, of
which 1,757 (90%) were matched to exit interviews and 6,063 observed interactions at district
health posts, of which 4,988 (82%) were matched to exit interviews. The combined information
allows us to look at key provider behaviors as well as reference these against more detailed
patient demographic information when assessing the determinants of diagnostic effort.
134. We measure “effort” using a composite index based on three variables
collected during the survey: consultation time, the number of questions asked, and
the number of physical examinations completed. Using the data on clinical interactions,
we extract the first principal component of these three variables, which represent visible costly
actions the providers can take during each interaction. This index ranges over an approximately
standard normal distribution ranging from -2 to +3 (just 3.8% of interactions fall outside this
range), and the index is normalized so that the mean effort in the overall sample is zero. The
mean effort in the matched sample is 0.02 and a t-test for difference gives p=0.25, strongly
suggesting that a representative subsample of patients were interviewed on exit. Table 5.2.1
shows how various components of practice varies with the effort index and compares it to other
countries where we have similar data (21; 22; 23; 24). Figure 5.2.1 shows the continuous variation
in the components of the index over its range, as well as the shape of the full distribution.

100
Table 5.2.1. Cross-country comparisons of clinical behaviors across varying effort
levels.
International Comparisons
(1) (2) (3) (4)
Time (min) Questions Physical Exams Polypharmacy
Vietnam Low Effort 2.96 3.63 1.03 2.90
Medium Effort 4.69 6.08 2.12 3.21
High Effort 8.55 10.57 3.12 3.08
All (Average) 5.40 6.75 2.09 3.06

Delhi Low Effort 1.90 1.36 14% 2.13

Medium Effort 3.36 2.94 78% 2.72
High Effort 6.15 5.32 98% 3.05
All (Average) 3.80 3.20 63% 2.63

Paraguay Low Effort 5.79 5.33 1.38 1.36

Medium Effort 7.90 7.50 2.93 1.55
High Effort 11.34 11.91 3.64 1.65
All (Average) 8.33 8.23 2.65 1.52

Tanzania Low Effort 3.00 2.00 0.00

All (Average) 6.32 3.96 1.51

Other Tanzania 3.00 2.20

Nigeria 6.30 2.80
Malawi 2.30 1.80
United Kingdom 9.40
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015) and various other
studies summarized in Das and Hammer, 2007 (India); Das and Sohnesen, 2007 (Paraguay); Leonard, Mimeo
(Tanzania); Hogelzeir et al., 1993 (International Comparisons), and Deveugele et al. .2003 (21; 22; 23; 24).
Notes: Interactions are divided into effort terciles based on the distribution of behaviors within their own
countries. Polypharmacy reports the number of distinct medications given to each patient during the interaction.
Tanzania I and Tanzania II refer to separate studies.

101
 Figure 5.2.1. Variation in clinical behaviors across the distribution of the effort
index

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Box plot indicates 5th, 25th, 50th, 75th, and 95th percentiles of effort. Effort distribution truncated at +5 (46
interactions excluded).

135. The bottom third of patients received effort similar to average doctors in
India, with providers spending on average 2.96 minutes per patient, completing one
physical exam, and asking 3.6 questions. The top third of interactions were similar to those
done by Paraguayan providers, spending 8.6 minutes per patient, asking 8.8 questions, and
conducting three exams. One outcome of note is that polypharmacy, defined as the total number
of distinct medicines given, is high and does not vary with effort.
136. It is natural to ask whether these large real disparities in terms of diagnostic
effort have real consequences for the quality of care received by patients. Could high
effort simply reflect the fact that the provider knows the patient personally and asks them about
general questions and spends more time, in which case effort does not affect the quality of care?
Could increases in effort correlate perfectly with the seriousness or difficulty of the case, meaning
that the variation in effort reflects appropriate time rationing by providers?
137. The appropriateness of providers’ management choices is difficult to assess
directly through patient observation alone because the underlying condition is
unknown for most patients observed for such a short time. However, there are good
reasons to believe that there is in fact a strong correlation between clinical effort and appropriate
treatment.

102
138. Both studies from other countries and analysis with the Vietnam survey show
that higher provider effort is associated with more correct treatment. Studies using
standardized patients from other countries, where the illness and correct treatment
are pre-specified by the research team, almost always show that higher provider
effort is associated with more frequent correct treatment choices (25; 26; 27; 28).
We can also investigate directly whether a similar relationship holds in our data, using two
common conditions with well-defined diagnostic checklists, and for which the observers
specifically noted the questions asked and the examinations completed. In Figure 5.2.2, we use
these two conditions—clinical diarrhea and cough/cold—to show several things. First, the
histogram of the effort index is shown in the background, to show the underlying variation in
effort even within two common and well-known conditions. The figure then shows the non-
parametric relationship between (a) the effort index and the fraction of medically recommended
checklist items that were completed and (b) the effort index and the use of medicines for these
cases. As is clear, the effort index is strongly associated with the greater likelihood of completing
medically necessary checklists for these two conditions. However, as found in other studies,
more effort does not lead to lower use of medicines. To the extent that some of these medicines
are unnecessary, it implies that effort and over-treatment are not necessarily correlated; this is
much like the lack of correlation between knowledge and use of unnecessary medicines
documented in the previous chapter.

Figure 5.2.2. Variation in essential checklist completion and medication use by

effort index: diarrhea and cough/cold

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: N = 203 clinical diarrhea interactions and 1,889 clinical cough/cold interactions. Box plot indicates 5th, 25th,
50th, 75th, and 95th percentiles of effort. Effort distribution truncated at +5 (11 interactions excluded).

103
139. One reason that the effort we observe may be even larger than what patients
truly receive is the so-called "Hawthorne" effect, whereby doctors put in more effort
because they are being observed. In previous studies, the Hawthorne effect appears
to drop-off quite quickly and after 4-5 patients, disappears entirely. The clearest
evidence we have on this comes from Leonard and Masatu's (2008) study, where the researchers
interviewed patients outside the clinic before they went in to observe the doctor without the
doctors' knowledge (4). They show that immediately on entering the clinic and starting
observation, effort increases, but within 5 patients or so it returns to what it was before. Analysis
shown in Annex Table AE.10. shows that the Vietnam survey data show patterns consistent with
this behavior, with slightly higher effort in the first 5 patients. To the extent that the Hawthorne
effect is at play in these results, we are estimating an upper bound of effort.

5.3. Decomposing the Variance of Effort

140. There are large variations in the effort index. To understand where this variation
comes from, a simple ANOVA decomposition shows that 4.1% is due to the community
characteristics of province and poverty rate; 38% is due to the facility and community
characteristics; 49% is due to community, facility and doctor characteristics (identifiable because
we have multiple HCPs in many facilities) and the remainder due to patient-level variation and
idiosyncratic error. Table 5.3.1 below shows, for instance, the variation across provinces.

Table 5.3.1. Cross-province variation in interaction effort and clinical behavior

Dien Dak Binh Dong Dong Urban Rural
Total
Bien Lak Dinh Thap Nai Hanoi Hanoi
Interaction Effort 0.04 -0.07 -0.21 0.01 -0.17 0.45 0.37 0.02
Time with Doctor (Minutes) 6.55 5.94 4.05 3.29 4.64 10.42 8.75 5.48
Number of Questions 7.11 7.48 6.53 6.33 5.92 7.59 7.65 6.78
Number of Exams 1.90 1.66 1.89 2.51 2.11 2.05 2.10 2.11
Test Ordered 23% 48% 43% 12% 48% 56% 60% 38%
Patient Satisfied 77% 71% 74% 84% 72% 66% 66% 75%
Referral 21% 2% 10% 2% 8% 14% 19% 9%
Price 10,520 12,381 4,728 10,562 31,827 92,316 52,939 24,354
Number of Medications 2.48 3.42 3.35 4.01 3.23 2.70 2.63 3.31
Antibiotics 51% 39% 41% 40% 38% 34% 41% 41%
Number of Observations 531 1,017 760 1,845 1,194 301 1,097 6,745
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Price includes all expenses incurred inside and outside the facility, as reported by the patient in exit
interview.

141. Health care providers in rural and urban Hanoi are among the highest
performing, while those in Binh Dinh are among the worst performing. The average
patient interaction in urban Hanoi lasts 6.4 minutes longer than Binh Dinh, with one additional
question and 0.2 more physical exams. More patients are ordered tests and patients are also
more likely to be referred to higher order care. They are given fewer medicines and fewer
104
antibiotics. All of this comes at a price—on average, patients in urban Hanoi can be expected to
pay out of pocket VND 87,000 more than in Binh Dinh (note though that urban Hanoi patients
also report the lowest satisfaction in any province.)
142. Remarkably, none of this variation is correlated with poverty at any level. We
can examine both whether effort is correlated with the poverty index in the community, as well
as whether poor/rich patients receive different types of care from providers. Figure 5.3.1 shows
first that is relatively stable over most of the poverty distribution with surprisingly small absolute
differences between even very poor areas and the richest within the top 75% of the poverty
distribution. The poorest 25%, however, may face substantially lower levels of effort from
providers on average.

Figure 5.3.1. Mean effort index by local poverty rate

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: N = 214 commune providers and 171 district providers. 27 facilities with poverty rates over 50% excluded
from visualization for scale. Box plot indicates 5th, 25th, 50th, 75th, and 95th percentiles of effort in visualized facilities.

143. There is very little discriminatory correlation between effort and either
ethnicity, education or wealth. Table 5.3.2 shows results from a series of regressions of
effort on patient characteristics. Because of the exit surveys, we have data on a rich set of patient
characteristics including age, sex, measures of health status (Self-Reported Health Status or SRHS)
as well as measures of ethnicity, education and wealth (computed using an asset index) and day
of week and time of day that the patient was seen. Column 1 shows correlations without any

105
fixed effects; Column 2 with facility fixed effects and Column 3 with provider fixed effects
(because we have many patients observed with each provider). Sicker patients as well as those
who are older receive more effort.

Table 5.3.2. Effects of patient characteristics

reg1on providerreg2
effort reg3
(1) (2) (3)
Facility Fixed Provider Fixed
No Fixed Effects
Effects Effects
Patient Asset Index 0.013 -0.007 -0.010
(0.009) (0.008) (0.008)
Patient Health Status (Higher = Better) -0.134*** -0.106*** -0.106***
(0.026) (0.023) (0.021)
Male 0.005 0.007 0.014
(0.030) (0.024) (0.023)
Patient Age 0.001 -0.000 0.002**
(0.001) (0.001) (0.001)
Ethnic Minority 0.092* 0.030 0.088
(0.052) (0.061) (0.057)
Hour of Interaction Start -0.025*** -0.017*** -0.022***
(0.005) (0.004) (0.004)
Can Read or Write / Primary School -0.110*** -0.076** -0.035
(0.042) (0.035) (0.033)
Secondary School and Above -0.203*** -0.207*** -0.109***
(0.046) (0.039) (0.038)
Day of Week Controls x x x
Patient Symptoms Controls x x x
Facility Fixed Effects x
Provider Fixed Effects x
Number of Observations 6,649 6,649 6,649
R2 0.053 0.414 0.512
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Results reported from multiple regression models with interaction effort on the left hand side. Education
effects are relative to patients who reported that they were illiterate. *** p<0.01, ** p<0.05, * p<0.1.

144. These results also show sharp declines in effort through the day. Figure 5.3.2
shows that patients come in two shifts, the 7am to 11am shift and the 2pm to 5pm
shift with a break in between. Patient load peaks at 8:30am. The decline in effort over the
entire day is striking, with patients who arrive in the evening receiving up to 0.75 standard
deviations lower effort than those who come early in the morning, with the biggest decline in
district facilities. In this figure we also introduce the difference in effort between district hospitals

106
and commune health stations. Effort declines over the course of the day in both type of faciilty.
However, at all points in the day, effort in the commune facilities is higher than in district facilities.

Figure 5.3.2. Mean effort index by time of day

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Controlled for patient age, gender, asset index, health status, ethnic minority status, day of week,
symptomatic presentation, and education, with provider fixed effects.

5.4. Comparison of district hospitals and commune health stations

145. District hospitals are staffed with the most medically knowledgeable providers
in Vietnam. Figure 5.4.1 shows this bifurcation very clearly; 90% of all providers who are two
standard deviations below average in medical knowledge are in the communes and 90% of all
those above average in medical knowledge are in district hospitals. Virtually all the doctors at the
top of the knowledge distribution are in districts.

107
 Figure 5.4.1. Probability of a provider being based in a district or commune
facility as a function of demonstrated vignettes ability and average clinical
interaction effort

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: Box plots indicate 5th, 25th, 50th, 75th, and 95th percentiles of ability and average effort.

146. Effort is much lower in district hospitals than in commune health stations. This
difference is (a) present in every province and (b) remains virtually the same after controlling for
patient characteristics, patient symptoms, and location characteristics such as poverty rate, which
are all frequently cited as key driving forces for differential provider behavior. Overall, 70% of
providers who average at least one standard deviation below the mean are in districts, while 70%
of providers who average at least one standard deviation above the mean are in communes.
147. Similar differences between commune health stations and district hospitals
are found when considering the components of the effort index. Table 5.4.1
summarizes the behaviors and reports the difference estimates both with and
without controls, including an additional control for provider workload, the number
of patients seen by the provider in a day. These differences illustrate the portion of the
difference is attributable to external characteristics. The effort difference between commune and
district is larger after controlling for patient characteristics. However, the difference in workload
accounts for two-thirds of that as workload because workload is strongly correlated with the
effort.

108
Box 5.4.1: Do doctors with private practice have different levels of ability or exert
different levels of effort than other doctors? The high prevalence of private practice
raises the question of which doctors are in private practice and how private practice relates
to behavior in doctors’ work in public facilities.

Overall, doctors who have some private practice have on average higher levels of
ability. Figure 5.4.2(a) shows the distribution of ability by whether doctors report that they
have any private practice. Overall, doctors who have some private practice have on average
higher levels of ability, as measured by the vignettes IRT score. Average ability levels are 0.37
standard deviation higher among those with private practice. In particular, very few doctors
with extremely low ability have private practice. This difference is diminished but persists in a
multivariate regression analysis that controls for doctor and facility characteristics. Even
controlling for other factors, private practice doctors have ability scores that are 0.128 sd
higher. (A comparison of doctor ability by various characteristics with and without controls is
shown in Annex E Table AE8a.) This suggests that patients who seek private practice are at
least somewhat sensitive to doctor ability and that lower ability doctors are less able to attract
customers for private practice.
Figure 5.4.2 Distribution of ability and effort by whether doctors have private
practices
a) Ability B) Effort

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

Doctors who have some private practice on average exert lower levels of effort,
but this difference disappears after controlling for other factors. Figure 5.4.2(b)
presents a plot with the distribution of effort by whether doctors have any private practice.
Doctors with private practice exert notably lower levels of effort on average. Average levels
of effort are 0.27 standard deviation lower among doctors with private practice. It is unclear
why this would be the case. It is conceivable that some doctors exert lower effort in public
facilities in order to direct patients to their private practice to receive care with higher effort.
However, there is no statistically significant difference after controlling for facility and doctor
characteristics. (Doctor effort by various characteristics with and without controls is shown
in Annex E Table AE9b.) This suggests that the apparent lower effort of doctors with private
practice may be an artefact of other variables. Most notably, doctors at district hospitals on
average both exert lower effort and are more likely to engage in private practice. Further

109
 investigation is needed to understand the causes of their lower effort and whether private
practice may play a role.

148. Controlling for these factors, commune providers spent 2.6 more minutes
with patients than district providers (1.2 minutes with controls); asked 0.9 more
questions (0.4 with controls); and conducted 0.4 more clinical examinations (0.2 with
controls). Note that, as before, effort and the use of medicines are uncorrelated. While both
district and commune providers gave a similar number of medications to each patient, commune
providers were 6% more likely to give antibiotics (8% with controls).

Table 5.4.1 Overall clinical management behaviors and district-commune

differences
(1) (2) (3) (4) (5) (6)
District - Adjusted With
Overall Commune District Adjusted
Commune Workload
Mean Mean Mean Difference
Difference Control
Interaction Effort 0.02 0.39 -0.11 -0.50*** -0.66*** -0.22***
Time with Doctor (Minutes) 5.48 7.37 4.81 -2.56*** -2.85*** -1.24***
Number of Questions 6.78 7.47 6.54 -0.93*** -1.47*** -0.42***
Number of Exams 2.11 2.39 2.01 -0.38*** -0.49*** -0.16***
Test Ordered 38% 4% 50% 0.46*** 0.42*** 0.46***
Patient Satisfied 75% 83% 72% -0.12*** -0.12*** -0.12***
Referral 9% 5% 10% 0.05*** 0.05*** 0.08***
Price 24,354 8,823 29,794 20,970*** 17,469*** 12,069***
Number of Medications 3.31 3.31 3.30 -0.00 0.16*** 0.04
Antibiotics 41% 45% 39% -0.06*** -0.03*** -0.08***
N 6,745 1,757 4,988

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: Adjusted differences are controlled for province, local poverty, provider ability, patient age, gender, asset
index, health status, ethnic minority status, day of week, symptomatic presentation, and education. Workload
control accounts for the number of patients seen by the provider during the observation period. *** p<0.01, **
p<0.05, * p<0.1.

149. Commune health providers do not generally provide tests to patients. Although
we do see 4% of patients receiving some sort of test in the commune health posts, this is negligible
relative to district hospitals, where providers ordered tests in 46% more interactions (46% with
controls). It is difficult to assess the need for these tests in the outpatient setting with more
information through, for instance, patient charts. Table 5.4.2 below shows the different symptoms
that patients reported with as well as the fraction who received different tests. Blood tests are
the most frequently performed, but ultrasounds and X-rays are also used frequently. While blood
tests for fever in the outpatient setting is consistent with clinical guidelines for a variety of
presenting symptoms, it is harder to understand why a quarter of patients with a cough/cold and

110
a third of patients with dermatological symptoms receive a blood test. Similarly, a third of all
outpatients with diarrhea receive an ultrasound; again, this may be well warranted for some
conditions but may seem excessive for a regular outpatient load.

Table 5.4.2. LaboratorySymptomatic

diagnostic orders
Presentation by patient
and Testing symptom at district facilities
Behavior (Districts)
(1) (2) (3) (4) (5) (6) (7) (8)
Percentage of patients
presenting with __ who
X-ray Ultrasound ECG Blood Test Urine Test Endoscopy Other Test Any Test
received…

Fever (N=933) 11.6% 10.7% 3.2% 41.1% 3.4% 1.2% 1.6% 49.7%
Cough/Cold (N=1379) 19.2% 7.6% 4.6% 26.7% 2.5% 1.9% 1.7% 41.9%
Diarrhea (N=128) 7.1% 31.0% 2.4% 38.9% 4.8% 2.4% 12.6% 55.9%
Weakness (N=1216) 16.5% 17.2% 14.7% 32.4% 6.1% 2.1% 3.8% 56.3%
Injury (N=9) 66.7% 0.0% 0.0% 33.3% 0.0% 0.0% 0.0% 100.0%
Vomiting (N=245) 14.3% 25.3% 2.9% 37.6% 2.4% 5.7% 2.9% 59.6%
Dermatological (N=137) 2.9% 7.3% 2.9% 33.6% 4.4% 0.0% 0.0% 38.7%
Pregnancy (N=7) 0.0% 57.1% 0.0% 28.6% 0.0% 0.0% 14.3% 57.1%
Pain (N=2695) 20.2% 22.6% 9.1% 23.6% 5.5% 2.0% 3.1% 53.6%

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Test results may not sum to 100% due to the exclusion of surgery and acupuncture which were used for
less than 1% of all conditions.

5.5. Does low effort in district hospitals reflect higher provider knowledge?

150. We explore two potential avenues through which lower effort in the districts
could be consistent with higher quality for average patients. The two possibilities are
higher knowledge and higher caseloads. Under the first hypothesis, more knowledgeable doctors
would exert less effort because their knowledge allows them to correctly treat efforts with less
effort.
151. In both communes and districts, providers with higher knowledge exert higher
effort, and the relationship is linear over the range of effort. Figure 5.5.1 shows the non-
parametric relationship between knowledge (measured as the vignettes IRT score described in
the previous chapter) and observed average clinical effort. Moving across the range of knowledge
increases effort by +0.5 standard deviation.

111
 Figure 5.5.1. Relationship between vignettes ability score and mean clinical
interaction effort

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Box plots indicate 5th, 25th, 50th, 75th, and 95th percentiles of ability scores.

152. The correlation between the effort index and knowledge (measured using the
vignettes) is robust to controlling for other factors. Table 5.5.1 shows a series of multiple
regressions using the effort index as the dependent variable. The first column shows just the
association with knowledge and suggests that a 1sd increase in knowledge is associated with a
0.25sd increase in effort. As before providers in the commune exert more than 0.6 standard
deviations higher effort than those in districts. Column 2 then includes a host of patient
characteristics that may be arguably correlated to effort. Column 3 is our most exacting
specification with facility fixed-effects. Because we have multiple providers in each facility, with
variation in knowledge among them, we can check whether in the same facility, providers with
higher knowledge also exert higher effort. Note that, to the extent that there is queuing with
random allocation to available providers, this should approximate the causal impact of knowledge
on effort and would thus go beyond correlations. Providers who are more knowledgeable in
district hospitals and in communes exert higher effort, suggesting that the district hospital “effort
gap” cannot be explained by the higher knowledge among these providers.

112
Box 5.5.1: Do doctors with different medical degrees have different levels of ability
or exert different levels of effort than other doctors? As outlined in Box 2.3.1, there
are three different pathways to become a doctor with training corresponding to a bachelor’s
degree. These include standard university doctor training, promotion from assistant doctor
through “twinning” training, and direct entry programs, which less competitive admissions for
particular populations.

Differences in ability level by type of medical degree are small. A plot of the
distribution of ability for each degree type is shown in Figure 5.5.2a. Doctors with direct entry
degrees have slightly lower ability on average, and those who have been promoted from
assistant doctors have slightly higher ability on average. These gaps are relatively small,
however. They are not statistically significant, with or without controls for other variables.
This suggests that the alternative pathways (direct entry and the twinning program) are
reasonably successful in training doctors to a level of knowledge similar to that achieved by
doctors who have gone through standard training.

Differences in effort by type of medical degree are negligible after controlling for
doctor and facility characteristics. The distribution is shown in Figure 5.5.2b Those with
direct entry degrees exert lower effort on average, and those with twinning degrees exert
the highest level of effort. These differences, however, shrink to near zero and are statistically
insignificant after controlling for facility and doctor characteristics. This suggests that
differences in effort by degree reflect differences in other variables.

Figure 5.5.2 Distribution of ability and effort by type of medical degree

a) Ability B) Effort

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

113
 Table 5.5.1. Correlates of interaction effort
reg1 reg2 reg3
(1) (2) (3)
Patient Facility
No Controls
Controls Controls
Vignettes IRT Ability Score 0.231*** 0.234*** 0.111***
(0.016) (0.016) (0.027)
District Facility -0.640*** -0.664***
(0.034) (0.035)
Patient Asset Index 0.033*** -0.007
(0.009) (0.008)
Patient Health Status (Higher = Better) -0.157*** -0.103***
(0.025) (0.023)
Male 0.012 0.007
(0.029) (0.025)
Patient Age -0.000 -0.000
(0.001) (0.001)
Ethnic Minority 0.134*** 0.050
(0.051) (0.061)
Hour of Interaction Start -0.017*** -0.018***
(0.005) (0.004)
Day of Week Controls x x
Patient Symptoms Controls x x
Patient Education Controls x x
Facility Fixed Effects x
Number of Observations 6,663 6,587 6,587
R2 0.064 0.116 0.415
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Results reported from multiple regression models with interaction effort on the left hand side. *** p<0.01,
** p<0.05, * p<0.1.

5.6. Does low effort in districts reflect higher caseloads?

153. A second potential reason for low effort in the districts is that they see more
patients. Figure 5.6.1 plots each provider in our sample, showing both the patients per day and
the total hours seeing patients. Commune health posts are shown in red crosses and district
hospitals in blue dots. As expected, district hospitals are busier than commune health posts; the
blue dots tend to lie to the right of the red crosses and the red crosses themselves are clustered
between 0 and 10, highlighting the very low caseloads in these facilities.
154. However, patient loads are sufficiently low in both district hospitals and
commune health posts that most doctors spend less than 5 hours a day seeing

114
patients. Only 2 providers in our sample of 383 providers sees patients for more than 8 hours.
Although 8 hours may be too high of a workload target given the administrative work that
providers need to complete during the day, it is reasonable to assume that they have at least 5
hours in their work schedule during which they can actively see patients in the outpatient setting.
Still, only 12 providers (10 in districts and 2 in communes) are above this limit.

Figure 5.6.1. Total hours worked and number of patients seen per provider

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Dashed reference line shows 5 hours per day.

155. Another way to consider workload is by comparing number of patients per

day and time spent with each patient. Figure 5.6.2 shows the number of patients per day
on the x-axis and the time spent with each patient on the y-axis. The dashed black line shows the
number of minutes that providers would spend with patients if they spent exactly 5 hours a day
seeing patients and evenly divided up the number of patients across those 300 minutes.

115
 Figure 5.6.2. Time spent with each patient and total workload

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Curved reference line illustrates an even spread of five hours among all patients at all levels of workload.
Dashed lines indicate 5 and 30 patients per day.

156. In rural health posts, the majority of providers see very few patients, but they
do not use the opportunity to increase effort with their patients. There are a small
number of providers with fewer than ten patients who spend 15 minutes or more with every
patient (with ten patients that is 150 minutes in total), but the vast majority spend less than 10
minutes and many even spend less than 5 minutes per patient. A similar pattern is seen in other
developing countries.
157. Most district hospital providers see more than 10 patients a day. Until the
provider has around 50 patients, there is ample room to spend more time with patients, but most
choose not to do so. This is seen in the solid blue linear trendline that lies significantly below the
full time-utilization hyperbola until about 50 patients. After 50 patients, many more providers are
close to the 5-hour limit line, but nearly always below it. There are 10 providers in district
hospitals who are above the hyperbola, and they are spread out along a wide caseload, ranging
from 30 to 160. In short, even though district hospitals see more patients than communes,
virtually no provider in the sample in district hospitals either sees sufficient patients or spends
sufficient time with the patients that they do see to “max out” their 5 hours

5.7. The quality impacts of low effort: The know-do gap

158. The impact of low effort can be evaluated using analysis of clinical observations
of doctor treatment for two tracer conditions: diarrhea and cough/cold. For these

116
conditions, the observers specifically noted whether the provider completed key checklist items
including questions and examinations. For diarrhea, for instance, the observer noted whether the
provider asked about fever, vomiting and nature of stool, which are all necessary questions to
assess the degree of dehydration and whether the diarrhea has a bacterial or viral origin. The
observer also checked whether the provider examined the patient’s temperature, pulse rate, and
respiration rate among other exams. In the medical vignettes, the diarrhea tracer condition also
checked whether the provider asked these questions and completed the exams, allowing us an
exact match between what the providers said would do and what they actually did when faced
with a similar patient in the clinic. For cough/cold, the match is less exact. In observations, we
note whether the patient had a cough or cold while in the vignettes, the tracer condition used
was pneumonia. Since the majority of cough and colds are viral in nature and hence self-resolving,
we may expect larger antibiotic use in the vignettes than in real life. Nevertheless, the questions
that observers noted (chest congestion, expectoration and fever) as well as the exams should
match fairly well.
159. What doctors actually did when faced with a patient with diarrhea differed
substantially from what the same doctors said they would do in the vignettes. This
comparison is shown in Figure 5.7.1. (A comparison for cough/cold are included in the appendix.)
We have a sample of 44 providers in communes and 79 in district hospitals who we can match
between observations and vignettes. There are large gaps between what doctors say they would
do (in blue) and what they actually do. Further, the gaps are larger in district hospitals for most
actions, except for specific tests that are not available in commune health posts. Providers in the
clinic are much less likely than in the vignettes they were to ask necessary questions, less likely
to perform key exams like checking the temperature or pulse and less likely to suggest blood,
urine and stool tests.

117
 Figure 5.7.1. Know-do gap for matched clinical and vignette diarrhea cases

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

160. The most striking gaps are in treatment. When asked, 94% of providers said they
would give ORS to the patient. In practice, only 40% did so in commune health posts and only
27% in district hospitals. Equally surprising is the dramatic increase in use of antibiotics. In the
communes, 9% said they would give antibiotics while 33% gave them in practice. In district
hospitals, a larger number (24%) appeared to believe that the patient should receive antibiotics
and an even larger number (36%) used them in practice. There are large gaps between knowledge
and practice, and these gaps are larger in the district hospitals than the communes. The gaps in
the district hospitals are so large that even though the providers are medically more
knowledgeable, their low effort implies that the ultimate quality of care that they deliver for this
case is lower overall.
161. The know-do gap implies that the benefits of further training may be quite
small. To see this, suppose that knowledge if is a function of training and quality is a function of
knowledge. We know from the previous chapter that moving from intermediate qualification to
primary specialist training correlates with increased knowledge by 0.5 standard deviations. The
usual assumption is that therefore training will also improve quality. But this depends critically on
how quality increases with knowledge.
162. Comparison of observed clinical behavior and vignette responses shows a
large “do-know” gap. Figure 5.7.2 plots a comparison of the percentage of the checklist
providers completed in the vignettes with the percentage they completed in the clinical
observations, for diarrhea and cough/cold cases. If knowledge translates fully into quality, we

118
would observe all points to lie along the 45-degree line—providers are doing all that they know
to do. In fact, for most of the range of the knowledge variable, the curve lies far below the 45-
degree line, and as knowledge increases, so does the gap. The slope of the line is only 0.14, so
that combining the two derivatives suggests that 4 full years of training will increase practice by
0.07 standard deviations.

Figure 5.7.2. Vignette checklist completion versus clinical checklist completion

for matched providers–diarrhea and cough/cold cases

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Scatter plot shows actual observed values with random offset to show density. Black line shows perfect
correspondence between vignette and clinical checklists with identical items.

5.8. Putting it together

163. Like most health systems, the Vietnamese system is intended to provide
health care through a multi-stage process. First, the system uses a less-qualified cadres of
medical professionals working at commune health posts to provide appropriate primary care at
low cost, both task-shifting the caseload away from expensive highly trained doctors and
restricting the use of laboratory diagnostic tests. Then, district hospitals provide high quality care
with testing as required for patients referred up the chain, and these are staffed with highly trained
providers. In this model, the commune health posts also serve a gatekeeping function—they see
a large number of outpatients, and providers triage the ones who require more complex care to
district hospitals. Unfortunately, no health system works the way it is supposed to because both
doctors and patients, like all humans, tailor their behavior to the incentives and options that they
face.

119
A summary of the portrait of the Vietnamese health system provided by the analysis is as follows:

As planned, the most highly trained providers work in district hospitals. They also see
more patients than in commune health posts.

o However, the effort of these providers is very low and significantly lower than in
commune health posts. For the cases where we can directly check, their delivered
quality is lower than in the commune health posts.
o They also provide a large number of medicines, antibiotics and (especially)
laboratory tests, which inflate the final cost to patients.

Providers in commune health posts are less knowledgeable and less trained, but they see
very few patients overall. This means that the public cost per patient in commune health
posts tends to be quite high.

o Effort in commune health posts is higher, but the quality of care they provide in
practice is still lower than what their knowledge levels would allow.
o Although tests are prohibited, significant medication and antibiotic use leads to
higher costs and could contribute to antimicrobial resistance in the future.
To the extent that health systems are trying to provide access to quality at a reasonable cost, we
can also use the data to provide some estimates of the cost per patient in district and commune
health posts. The table below provides a breakdown.

120
 Table 5.8.1. Total cost breakdown of clinical visits

Cost to Cost to
Cost to Insurance Total Cost
Government Patient

Salary cost of Other Average out-

provider, per Testing Medicines reimbursed of-pocket
patient costs expenditure
Insured patients

District 9,609 Unknown Unknown Unknown 21,276 >30,000

Commune 34,055 0 Unknown Unknown 5,234 >39,000

Uninsured patients

District 7,677 0 0 0 213,910 >221,000

Commune 51,266 0 0 0 53,429 >104,000

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: Salary cost per patient calculated at the provider level as self-reported monthly salary times twelve months,
divided by 250 working days, divided by the daily number of patients seen by the provider during patient
observation. 17

164. As is clear from the table, non-salary costs are higher in district hospitals. It
would be ideal to obtain data on the costs reimbursed by insurance for each of the patients we
observed; in the absence of these data, the out-of-pocket expenditures provide a potential
approximation. This approximation suggests that these other costs may be more than
VND200,000 per patient which would make district hospital visits significantly more expensive
to the health system in total than commune health posts. On the other hand, salary costs per
patient are much higher in commune health posts relative to district hospitals, primarily due to
the very small patient load in such places. In fact, we have 4 commune facilities were the single
provider’s per-patient salary costs exceed VND400,000 per patient (see Figure 5.8.1).

17
The table is incomplete since we have not costed the tests provided, the medicines given or the other costs of
the interaction that are reimbursed through insurance. Neither does it include the salaries of other staff at the
institution; including all of these will considerably inflate the cost estimates that we provide here. One way to see
this is to look at the total costs for uninsured patients, who will bear directly the costs of all medicines and tests,
and we present this in the last two rows of the table.

121
 Figure 5.8.1. Per-patient salary costs by provider workload

Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: Horizontal reference lines indicate average per-patient salary costs.

165. One reason for the low patient loads (and hence high salary costs) in commune
health posts is that they are located in remote rural regions and therefore are of
value who cannot travel easily to district hospitals. While this is the case in specific
instances, in our sample it appears that commune health posts and district hospitals cluster close
to each other. For instance, we computed the distance from each commune to the closest district
hospital and find that 50% of commune health posts are within 3Km of district hospitals and 75%
are within 5.5Km. In general, only a few commune health posts (<5%) are more than 12Km from
a district hospital. That commune health posts are seldom the first contact for primary care can
also be seen from referral rates which are twice as high in district hospitals (10%) compared to
commune health posts (5%). Although we do not have data from household choices, all
indications suggest that patients are actively bypassing commune health posts to go directly to
district hospitals, inflating the per-patient average costs in the communes and increasing the cost
of the visit through greater testing and insurance reimbursables in district hospitals.
166. It is striking that the quality that they receive for the two tracer conditions we
are able to directly assess is also lower in district hospitals. This suggest that either
patients value the tests that they receive very highly or value the option value of other services
at the district hospital that they cannot receive at the commune. It could also be that the quality
difference varies by tracer condition, and for more severe conditions, the district hospital is a
better bet. These are questions that we currently cannot answer, but that studies using
standardized patients can help with in the future.

122
 References

1. General Statistics Office (Viet Nam), Ministry of Health (Viet Nam). 2002. Vietnam National
Health Survey 2001-2002. Hanoi, Vietnam.

2. World Bank and Ministry of Finance (Vietnam). 2015. Public Expenditure Review, draft.
3. World Health Organization. Dual Practice by Health Workers in South and East Asia.
Available at:
http://www.wpro.who.int/asia_pacific_observatory/resources/policy_briefs/Policy_Brief_Dual_P
ractice_19072013.pdf?ua=1.

4. Leonard KL, Matasu MC. 2010. Professionalism and the know-do gap: exploring intrinsic
motivation among health workers in Tanzania. Health Econ. 19: 1461–1477.
5. Ministry of Health (Vietnam), Health Partnership Group (Vietnam). 2012. Joint Annual Health
Review 2012: Improved quality of medical services. Hanoi, Vietnam.

6. Ministry of Health (Vietnam). Decision 139/KH-BYT dated March 1st, 2016. Available at:
http://thuvienphapluat.vn/van-ban/The-thao-Y-te/Ke-hoach-139-KH-BYT-bao-ve-cham-soc-nang-
cao-suc-khoe-nhan-dan-2016-2020-304953.aspx.

7. World Bank, Ministry of Planning and Investment of Vietnam. 2016. Vietnam 2035: Toward
Prosperity, Creativity, Equity, and Democracy, draft. Hanoi, Vietnam.
8. Ministry of Health (Vietnam). Circular 43/2013/TT-BYT dated December 11, 2013. Available
at: http://thuvienphapluat.vn/van-ban/The-thao-Y-te/Thong-tu-43-2013-TT-BYT-phan-tuyen-
chuyen-mon-ky-thuat-co-so-kham-chua-benh-217263.aspx.

9. —. 2013. Health Statistical Yearbook. Hanoi, Vietnam.

10. Ministry of Health (Vietnam), the Health Parnership Group. 2015. Joint Annual Health
Report 2015: Strengthening primary health care in the grassroots healthcare network towards
universal health coverage. Hanoi, Vietnam.
11. Donabedian A. 1988. The Quality of Care. How it can be assessed. Journal of Amedican
Medical Association, Sep 23;260(12):1743–1748.

12. World Health Organization. 2015. Quality of care: what are effective policy options for
governments in low and middle income countries to improve and regulate the quality of
ambulatory care?. Policy brief, Vol. 4 No. 1 2015.
13. Health Strategy and Policy Institute. Hospital Overcrowding and Undercapacity - Issues and
Solutions. Ministerial Research (unpublished).

14. World Bank. 2014. Universal Maternal Health Coverage - Assessing the Readiness of Public
Health Facilities to provide Maternal Health Care in Indonesia. No.90510-ID.

15. Ministry of Health (Vietnam). 2014. Health Statistics Year Book 2014. Hanoi, Vietnam.

123
16. World Health Organization. Global Health Observatory (GHO data). Available at:
http://www.who.int/gho/health_workforce/nursing_midwifery_density/en/.

17. OECD, World Health Organization. 2012. Health at a Glance: Asia/Pacific 2012. Available
at: http://dx.doi.org/10.1787/9789264183902-en.
18. Peabody JW, Luck J, Glassman P, Jain S, Hansen J, Spell M, Lee M. 2004. Measuring the
Quality of Physician Practice by Using Clinical Vignettes: A Prospective Validation Study. Annals
of Internal Medicine. Vol. 141. 10 pp771-781.

19. Spreng CP, Ojo IP, Burger NE, Sood N, Peabody JW, Demaria LM. 2014. Does stewardship
make a difference in the quality in the quality of care? Evidence from clinics and pharmacies in
Kenya and Ghana. International Journal for Quality in Health Care. Vol. 26. 4: pp. 388-396.

20. Shah R, Edgar DF, Evans BJ. A comparison of stadardised patients, record abstraction and
clinical vignettes for the purpose of measuring clinical practice. Ophthalmic and Physiological
Optics. Vol. 30. pp209-224.

21. Deveugele M, Derese A, Van den Brink-Muinen A, Bensing J, De Maeseneer J. 2002.

Consultation length in general practice: cross sectional study in six European countries. British
Medical Journal. 31 August: 325:472.

22. Hogerzeil H, Bimo D, Ross-Degnan DG, Laing RO, Orofi-Adjei D, Santoso B, Chowdhury,
AK. 1993. Field tests for rational drug use in twelve developing countries. Lancet. 342(8884):
1408-1410.
23. Das J, Hammer J. 2007. Money for nothing: The dire straits of medical practice in Delhi,
India. J Development Economics. 83(1): 1-36..

24. Das J, Sohnesen T. 2007. Variations in Doctor Effort: Evidence from Paraguay. Health Affairs.
26, No. 3: w324-w337.

25. Sylvia S, Shi Y, Xue H, Tian X, Wang H, Liu Q, Medina A, Rozelle S. 2015. Survey using
incognito standardized patients shows poor quality care in China's rural clinics. Health Policy
Plan. 2015 Apr;30(3):322-33.
26. Glassman PA, Luck J, O'Gara EM, Peabody JW. 2000. Using standardized patients to
measure quality: evidence from the literature and a prospective study. The Joint Commission
journal on quality improvement. 26(11): 644-53.

27. Rethans JJ, Gorter S, Bokken L, Morrison L. Unannounced standardised patients in real
practice: a systematic literature review. Medical education. 41(6): 537-49.
28. Das J, Holla A, Das V, Mohanan M, Tabak D, Chan B. 2012. In urban and rural India, a
standardized patient study showed low levels of provider training and huge quality gaps. Health
Aff (Millwood). 31(12): 2774-84.

29. World Bank. Service Delivery Indicators. Available at: http://www.sdindicators.org/.

124
30. World Health Organization. Service Availability and Readiness Assessment . Available at:
http://www.who.int/healthinfo/systems/sara_introduction/en/.

31. Leonard KL, Masatu MC. 2005. The use of direct clinician observation and vignettes for
health services quality evaluation in developing countries. Soc Sci Med. Nov;61(9):1944-51.
32. General Statistical Office (Vietnam). 2007. GSO projection 2007.

33. General Statistical Office (Vietnam). 2014. Vietnam Household Living Standard Survey.
Available from the General Statisticas Office of Vietnam.

125
 Annexes
Annex A. Methodology
Instrument
The 2015 Vietnam health survey consists of 5 components including: (1) facility questionnaire; (ii)
health worker interviews; (iv) exit patient interviews (iv) clinical vignettes; (v) clinical observation.
Except clinical observation, the core instruments of four remaining modules were modelled along
the Service Delivery Indicators (SDI), with the integration of the Service Availability and Readiness
Assessment (SARA) and 2001-2002 Vietnam National Health Survey tools, and adapted to
Vietnam contexts (1; 29; 30). The module clinical observation, specifically, used Generalizable
Reducible Metrics (GRM) method which was based on direct observation of clinical practice. The
clinical observation analysis was mostly based on data collection instrument tools implemented
successfully in other settings including India and Tanzania (23; 31). Table 1 describes contents of
these five modules.

Table AA.1: Outline of the health facility survey questionnaire

Module Key respondent Description
1.Facility Administrative staff in Collected general information about the health
Questionnaire charge/hospital facility, utilization, waste management, facility
director/most senior infrastructure, availability of equipment,
medical officer in materials, drugs and supplies, offered
charge/facility laboratory and diagnostic services.
financing senior staff in Collected revenues and expenditure by source,
charge information on clinical audits, supervision visits,
availability of guidelines.
2. Health Health workers Collected data of district hospital doctors’ and
worker commune health station all staff’s
interviews characteristics, training opportunities, income,
dual practices, satisfaction, and policy
suggestions.
3. Clinical Doctors/assistant Assessed the clinical knowledge of doctors
vignettes doctors and/or assistant doctors using medical
vignettes.
4. Clinical Doctors and/or Assessed the practices of doctors and assistant
observations assistant doctors doctors. Collected information on consultation
time, number of history questions,
performance of examinations, prescribed
medicines, given treatments, given tests.
5.Exit patient Facility service users Collected information on patient experience
interviewsa (waiting time, services of receive, procedures
carried out, payments, etc.), socio-economic
characteristics, source of health financing, and

126
 provider preferences and expectations (reason
for choosing facility.)
Note: (a) Exit patient interviews were performed for outpatients at district hospitals and commune health stations,
and inpatients at district hospitals (the very small group consisting of commune health station inpatients were not
covered by the survey.)

Sampling strategy
The 2015 Vietnam health quality survey was conducted in the same locations with the 2015
household survey (which was simultaneously conducted to collect information on demand side
of Vietnam health system) to ensure the linkage in analyzing the relationship between the health
seeking behavior and the quality of local providers (the sampling design and calculation of the
household survey is described in Annex B.)
The study consists of six provinces locating in six geographical regions of Vietnam: Dien Bien,
Hanoi, Binh Dinh, Dak Lak, Dong Nai, and Dong Thap. Four and a half provinces (Binh Dinh, Dak
Lak, Dong Nai, Dong Thap, and the new half of Hanoi which was the “formal Ha Tay” 18) were
selected as a “typical” of their corresponding regions based on criteria of provincial average
income per capita and provincial poverty rates. To assess the equality of healthcare services, one
poor and ethnic minority province (Dien Bien), and a major city (the original half of the capital
Hanoi1) were also included.

Sampling design and sample size

The sample of the health facility survey were commune health stations and district hospitals
locating in the communes and districts that were corresponding with the selected enumeration
areas (clusters) in the household survey. Specifically in urban areas of Hanoi, where multiple
central level hospitals concentrate, some districts do not have district hospitals. In this case, the
corresponding city level hospitals or polyclinics were selected. In each facility, besides facilities’
overall information, data of a sample of doctors and inpatients and outpatients were collected.
The table below describes the sample size of each module.
Table AA.2: Sample size information of all modules
Module District Hospital Commune Health Stations
1.Facility All district hospitals or All commune health stations
Questionnaire replacements locating in the locating in the communes
districts selected in the selected in the household survey.
household survey.

18
Hanoi is the capital of Vietnam and the country’s second largest city. In 2008, another province in the same region
(Red River Delta) Ha Tay and one district of Vinh Phuc and four communes of Hoa Binh were merged into the
metropolitan areas of Hanoi. Hanoi after 2008 approximately was doubled in size of population. Therefore, Hanoi
can be considered as “two provinces”: (i) the original half that had been the capital of Vietnam since a thousand of
years ago; (ii) the second half (“formal Ha Tay”) that can “represent” for Red river delta in term of annual income
per capita and poverty rate.

127
 Total sample size: 78 district Total sample size: 246 commune
hospitals. health stations.
2. Health worker At most 10 randomly selected All commune health stations’
interviews doctors per facility. health workers including
Total sample size: 749 doctors. doctors, assistant doctors,
midwives, nurses, pharmacists,
and others.
Total sample size: 1688 health
workers.
3. Clinical vignettes At most 10 doctors per facility 1-2 doctors/assistant doctors
(the same doctors participated responsible for patient
in the module 2 “Health examination/consultation per
3
workers interviews”.) These commune health station .
included 2 doctors1 selected in Total sample size: 251
the clinical observation module doctors/assistant doctors.
and 8 randomly selected from
the doctor list.
Total sample size: 749 doctors.
4. Clinical 2 randomly selected doctors1 1-2 doctors/assistant doctors
observations among whom provided who are responsible for patients’
outpatient examination services examination and treatments per
per district hospital. commune health stations3.
Total sample size: 171 doctors Total sample size: 214
(6063 outpatients2.) doctors/assistant doctors (1961
outpatients.3)
5. Exit patient Inpatients: 12-16 randomly Inpatients: No inpatient care
Interviews selected inpatients (who exit on services at almost all commune
the survey) per facilities. health stations.
Outpatients: all outpatients Outpatients: all outpatients
examined by 2 selected doctors examined by doctors/assistant
in the clinical observation doctors in the clinical
module. observation module.
Total sample size: 948 inpatients; Total sample size: 1759
4989 outpatients. outpatients.
Note: (1) Averagely, each district hospital has 3-5 doctors performing outpatient examinations per working day.
In this study, to capture the diversity of patient population, two tables (one doctor per table) were selected. In
general, one adult and one pediatric examination tables were randomly selected. In case the patient examination
separation was on health insurance utilization basis instead, one health insurance and one non-health insurance
examination tables were randomly chosen; (2) In the clinical observation module, despite doctors were the objects
to be observed for practice assessment, the analysis unit was interactions between doctors and outpatients; (3)
At commune health stations, mostly there is one doctor in each commune health station. For commune health
station do not have any doctor, there is an assistant doctor playing “doctor” role for examination. In this study,
these doctors (assistant doctors) were selected to participate in both clinical vignettes and clinical observation
modules.

128
Annex B. Sampling Design of the Vietnam Household Survey.

The 2015 Vietnam household survey was implemented simultaneously with the 2015 Vietnam
health facility survey. This provides a chance to look at the relationship between health utilization
patterns and the quality of local health services. Despite this relationship has not been explored
in this report, the close linkage in sampling design of these two survey components are
acknowledged. The 2015 Vietnam health facility survey comprised commune health stations and
district hospitals for all communes and districts selected in the sample that the household survey
was implemented. The sampling design and sample size estimation of the household survey was
described below.
Sampling Design

Multi-stage sampling method with 3 stages was used in this survey:

Stage 1: Province selection. Six provinces Dien Bien, Hanoi, Binh Dinh, Dak Lak, Dong Nai, Dong
Thap were selected as a “typical” of their corresponding regions based on criteria of provincial
average income per capita and provincial poverty rates.
Stage 2: Cluster selection using the Probability Proportional to Size (PPS) method. The sample
clusters were selected based on the sampling frame of six provinces used in the 2014 Intercensal
survey (stratified by rural/urban for each province.) The household lists and maps were checked,
reviewed and updated to use in this survey.
Stage 3: Household selection using the Systematic Randomly Selection (SRS) method. In each
cluster, 25 households were systematic randomly selected from the list of households.

Sample size estimation

The sample of the 2015 Household Survey was based on the sampling frame of the 2014
Intercensal survey. The sample was ensured to be representative for urban and rural areas, as
well as for six geographical regions. Sample size was calculated follow as:
4 r (1 r ) deff
n
( RME r ) 2 pb Hsize RR
In which:
•n : The estimation for needed sample size, it’s shown by number of households
based on the key indicator.
•4 : Factor for 95% statistical significance.
•r : Estimated ratio for the key indicator.
• deff : Design effect
• RME*r : Error limits are allowed with a 95% confidence level; 0.13 (13%) (to
guarantee the reliability in the conditions shortage of resource)
• pb : Proportion of population will be used to calculate for r
• Hsize : Household size (the average members per one household).
• RR : Estimated respondent percentage.

129
The key indicator used in the study was “the percentage of the illness/diseases/injuries among
population in last 4 weeks”. According to the data of Vietnam Living Standards Survey 2008, the
proportion of sickness/injury in 4 weeks preceding the survey was 16.34%. Using the above
formula (deff=8; respondent percentage=90%; household size=3.8 persons; RME=0.13), the
sample size needed for each domain (urban and rural) was 2800 households (Table B.1.)
For the clusters numbers calculation, to increase the reliability, the number of selected clusters
was increased in regions with low population and decreased relatively in high population- regions.
For this adjustment, the formula used to estimate the number of clusters in urban and rural
domains for six regions was:
3 H ij
nij Ni 6
3 H ij
In which: j 1

nij
: The number of clusters allocated for area i (i=1 (urban), 2 (rural)), region j
(j=1÷6)
Ni
: The total number of sample clusters needed of domain i (i=1 (urban), 2 (rural))
H ij
: The total number of households of domain i, region j (j=1÷6)

In this study, the purposes of sampling were not only to select a “representative” province for
each region, but also to compare the access to and use of health services between big cities and
other socioeconomic regions. Due to this objective, Hanoi was included in the sample. Hanoi
now comprises the former Capital Hanoi with 3.2 million inhabitants and “formal Ha Tay”
province with 2.7 million people (32)19. “Formal Ha Tay” was considered as the “average”
province of the Red River Delta region (in terms of poverty rate and income per capita.) In the
study sample, 44 sample clusters (22 urban and 22 rural clusters) was selected in “formal Ha
Tay”. To increase the reliability of the estimation for the “new” Hanoi, 22 clusters (25 households
per cluster), including 15 urban and 7 rural clusters were added. With this addition, the final
sample size was 246 clusters (66 clusters in Hanoi) with 6150 households (Table B.1 and B.2.)
Table AB.1: The distribution of sample households according to urban-rural and
6 socio-economic regions
Regions (6 socio-economic regions)1
Total 1 2 3 4 5 6
Sample size of Clusters
Total 224 33 44 41 27 40 39
Urban 112 15 22 20 13 24 18

19
Data source: GSO projection 2007. 2007 was the last year before Ha Tay was merged into Hanoi.

130
Rural 112 18 22 21 14 16 21
Sample size of households
Total 5600 825 1100 1025 675 1000 975
Urban 2800 375 550 500 325 600 450
Rural 2800 450 550 525 350 400 525
Note: (1) 1=Northern mountain and Midlands (selected province= Dien Bien); 2=Red River Delta (selected
province=Ha Noi”); 3=North Central and Coasted Central (selected province=Binh Dinh); Central Highland
(selected province=Dak Lak); South East (selected province=Dong Nai); Me Kong River Delta (selected
province=Dong Thap).

Table AB.2: The distribution of sample clusters according to urban-rural and 6

socio-economic regions
Region Clusters/HHs Total Regions (6 socio-economic regions)1
1 2 3 4 5 6
Urban Number of 127 15 37 20 13 24 18
clusters
Number of 3175 375 925 500 325 600 450
households
Rural Number of 119 18 37 20 13 24 18
clusters
Number of 2975 375 925 500 325 600 450
households
Total Number of 246 33 66 41 27 40 39
clusters
Number of 6150 825 1650 1025 675 1000 975
households
Note: (1) 1=Northern mountain and Midlands (selected province= Dien Bien); 2=Red River Delta (selected
province=Ha Noi”); 3=North Central and Coasted Central (selected province=Binh Dinh); Central Highland
(selected province=Dak Lak); South East (selected province=Dong Nai); Me Kong River Delta (selected
province=Dong Thap).

131
Annex C. Definitions of Selected Indicators in the Report.

Table AC.1. Definitions of Health Service Delivery Indicators

Chapter 1: Facility Service Readiness
Clean water “Clean water” was defined as piped water or water from a
protected well.
Waste water treatment A facility was considered to have waste water treatment system
system if the waste water was processed by the concentrated sewage
treatment system or laboratory sewage treatment.
Solid waste treatment A facility was considered to have solid waste treatment system if
system its medical solid waste was processed by specialized incinerators
or sanitation companies.
Electricity access A facility was considered to have electricity access if its power
source was national grid power.
Toilet access A facility was considered to have toilet access if septic or semi-
septic toilet(s) was (were) available for patient utilization.
Caseload A surrogate “caseload” variable was estimated based on the
number of patients per doctor on the survey day. Despite the data
we actually collected was only from one day, this visited day was
selected randomly based on the study planning team.
Occupancy rate by actual Occupancy rate was calculated as the number of inpatient days
number of beds for a given period divided by number of actual beds and number
of days in the period.
Chapter 2: A profile of health workers
Satisfaction score Satisfaction scores were calculated as the mean of ten satisfaction
indicators: salary, allowance, training opportunity, promotion
opportunity, occupational safety, hospital security, working
environment, availability of medicines, availability of equipment,
working pressure, having enough staff. These ten satisfaction
indicators were measured by a Likert scale with 5 levels from very
dissatisfied (1) to very satisfied (5).
Chapter 3: Patient experiences
Satisfied with the facility This indicator was estimated using the data from the question “are
you satisfied with the services of facility for this visit?”. Among five
scales of this question (from “very dissatisfied” to “very satisfied”),
one patient was considered as “satisfied with the facility” if she
answered “satisfied” or “very satisfied”.
Chapter 4: How good is your doctor? Assessment of doctor availability
Doctors’ ability This aggregate indicator was generated using item response
theory (IRT) to produce a ranking or distribution of doctors by
levels of ability. The IRT methodology used maximum likelihood
methodology to estimate the underlying “ability score” of
providers based on their performance during the medical
vignettes exercise. This score quantified doctors’ propensity to
ask the history questions and perform the physical examinations

132
 that were graded as minimum or essential by the medical expert
committee based on Ministry of Health technical guidelines.
Correct diagnosis Doctors’ correct diagnosis was estimated based on their diagnosis
performance during five cases of vignettes exercise (acute
diarrhea without dehydration, child pneumonia, tuberculosis,
diabetes, and hypertension.) For each case, doctors’ diagnosis was
graded into three categories: fully correct, partially correct, and
incorrect. The grading process was done following Ministry of
Health technical guidelines and through multiple rounds of
consultations with Vietnam medical experts.
Correct treatment For every cases of the vignette exercise, doctors’ correct
treatment was graded into three categories: completely correct,
partially correct, and incorrect. The grading process was done
following Ministry of Health technical guidelines and through
multiple rounds of consultations with Vietnam medical experts.
Harmful and unnecessary Doctors’ harmful and unnecessary treatment was estimated based
treatment on the treatments given by doctors during the vignette exercise.
It was graded into three categories: harmful, unnecessary, and not
harmful and unnecessary. The grading process was done following
Ministry of Health technical guidelines and through multiple
rounds of consultations with Vietnam medical experts.
Supplemental and optional This indicator was also derived from the vignette exercise. It was
treatment graded into three categories: supplemental, optional, not
supplemental and optional. The grading process was done
following Ministry of Health technical guidelines and through
multiple rounds of consultations with Vietnam medical experts.
Chapter 5: Knowledge and practice: clinical observation and the know-do gap
Doctors’ effort index The effort index of doctors was developed based on three
variables collected during the survey day: consultation time, the
number of questions asked, and the number of physical
examinations completed. This index was normalized so that the
mean effort in the overall sample is zero.

133
 Annex D. Tables of Medical Vignettes
Table AD.1. Medical vignettes
Box 1B Child Diarrhea Child Pneumonia Tuberculosis Diabetes Type 2 Hypertension category Provider’s Analysis
1 action output
Introduction The mother of a 15- A 3.5-year child has cough A 37-year old male A 58-years woman has signs of A 65- year male Provider
month old child takes and fever for three days patient, with sporadic fast weight loss recently (3kg patient, who asks
the infant to the clinic and has been given cough and fever for loss within 2 months) and sometimes has questions
as her child has had medicines bought from a the last three weeks, frequently feel emery less headache and burning
diarrhea for two days private drugstore but did fatigue and weight loss, hungry despite eating more than face, came to your
and the condition does not get better. The mother came in for check-up. normal. She doesn’t know why. clinic for check-ups
not go away after the takes him in for The patient said that So she came for the check-up
child took medicine at examination and care sometimes he had mild
home fever during the day. IRT
Basic The child looks Seeing that the child The patient has been The patient does not have Asking results with Provider Score
information lethargic fatigue, slow wheezes and is in fatigue, smoking for 10 years. cough, fever and she look bulky. normal answers. describes
to react. The mother what steps of physical Recently, the fever has She has headache, dazzle and Headache sometimes physical
said her child still examination will you take? occurred often in the dizzy happens during 1 last examination
drinks breast mild but afternoon and evening. month. Medical
less than normal Cough and sputum history is healthy
contains simply blood
Examinations When you examine the Examination indicates the Physical examination The patient has blood pressure The patient has pulse Provider Testing
child, he conscious, but child’s temperature is 390C; shows no special measurement of 130/80 mmHg, rate of 80 beats per orders tests grading
shows signs of fatigue, no signs of chest indrawing; symptoms pulse of 80 strokes per minute. minute, blood
negative meningeal respiratory rate is 42 beat pressure of 155/95
syndrome, negative per minutes; inspiratory mmHg. Other organs
infection syndrome and crackles and moist rales are normal
skin fold retracts sporadically exist
immediately
Testing With all information With all information you X-ray shows opaque The indicated tests result with Testing results are Provider Diagnosis
you have, how do you have, how do you diagnose sections at the top of normal value of complete blood normal. What gives grading
diagnose this cases? the patient? the right lung. What count, blood glucose (random presumptive diagnosis diagnosis
presumptive diagnosis test) at 8.6 mmol/l, normal ECG, would you give?
would you give? normal straight cardiopulmonary
X-ray. What presumptive
diagnosis would you give
Treatment If your hospital/CHS is Given that bacteriology or With diagnosis as With diagnosis as diabetes type With diagnosis as Provider Treatment
not able to do stool antibiotic susceptibility pulmonary TB, how 2, what medication will you give hypertension category describes grading
testing in this case, how testing is not available at will you treat the the patient? 1, what initial treatment
will you treat the child? grassroots level, what patient? medication would you
medication do you choose give (please specify)
for the patients
Note: the medical vignettes were constructed consistently with 5 stages (history questions, physical examinations, tests, diagnosis, and treatments) across 5 cases. The vignettes were designed
in the “stage structure”. This means the respondents did not know about the next question until they completed their answers for the previous one.

134
Table AD.2. Answer keys of clinical vignettes for the diagnosis section
Case Answer keys Reference source1
Completely correct Partly correct Incorrect
Child’s acute Acute diarrhea Acute diarrhea, Level A Not completely and Decision no.
diarrhea without dehydration dehydration partly correct 4121/QĐ-BYT dated
OR 28/10/2009
Acute diarrhea
Child’s pneumonia Pneumonia Bronchopneumonia Not completely and Decision no. 101/QĐ-
OR partly correct BYT dated 09/01/2014
Mild pneumonia
Pulmonary Pulmonary Suspected chronic Not completely and Decision no. 979/QĐ-
Tuberculosis Tuberculosis obstructive pulmonary partly correct BYT dated 24/03/2009
disease (COPD)
OR
Suspected lung cancer
Diabetes type 2 (Suspected) diabetes Prediabetes Not completely and Decision no.
type 2 OR partly correct 3879/QĐ-BYT dated
Disorders blood glucose 30/09/2014
intolerance
OR
Fasting glucose
disorders
Hypertension (Suspected) (Suspected) Not completely and Decision no.
category I hypertension hypertension partly correct 3192/QĐ-BYT dated
category I 31/08/2010
Note: (1) The answer keys were developed based on the Ministry of Health technical guidelines and were gone through three rounds of clinical
experts’ consultations.

135
Table AD.3. Answer keys of clinical vignettes for the treatment section
Case Answer keys

Correct treatment
Completely correct Partly correct Incorrect
Child’s acute diarrhea Oresol Offset orally dehydration and electrolytes loss No completely and partly
that are not oresol correct treatments
AND
No oresol
if there is oresol in treatment, coded as
“completely correct”
Child’s pneumonia Only 01 antibiotic among: Cephalosporin or co-trimoxazol No completely and partly
- Amoxicillin AND correct treatments
- Amoxicillin + acid clavulanic Paracetamol or ibuprofen
- Clarithromycin
- Erythromycin
AND
Paracetamol
Pulmonary Tuberculosis One of two below options: None No completely and partly
- Diagram 1 correct treatments
- Referral to specific departments or
hospitals
Diabetes type 2 One of two below options: One of below options: No completely and partly
- A+B - Only A correct treatments
Note: - A+B+C - Only B
A: consulting about diet, - Only C
living regime and physical - A+C
activities - B+C
B: scheduling periodic re-
examinations
C: Using orally
antihyperglycenmic drug
(alone)

Hypertension category I One of two below options: One of below options: No completely and partly
- A+B - Only A correct treatments
Note: - A+B+C - Only B
A: consulting about diet, - Only C
living regime, and physical - A+C
activities - B+C

136
B: guiding patients how to
measure blood pressure
daily
C: only 01 of following
medicines:
- ACE inhibitors
- Long-activing calcium
channel blockers
- Beta blockers
Harmful and unnecessary treatment
Harmful Unnecessary Not harmful and
unnecessary
Child’s acute diarrhea At least one of below options: - Vitamin and/or paracetamol No harmful and
- Antibiotics - (and/or) Hidrasec or smecta unnecessary treatments
- IV AND
No harmful treatment
if there is any harmful treatment, coded as
“harmful”
Child’s pneumonia Corticoid One of below options: No harmful and
- Antibiotics that are not “correct” or unnecessary treatments
“partly correct”
treatments
- Combination of at least two types of
antibiotics
- Salbutamol
- Aceylcystein (mucomyst, exomuc,
ACC,…)
AND
No harmful treatment
if there is any harmful treatment, coded as
“harmful”
Pulmonary Tuberculosis At least one of below options: One of below options: No harmful and
- Antibiotic - Diagram 2 unnecessary treatments
- Corticoid - Diagram 3
AND
No harmful treatment
if there is any harmful treatment, coded as
“harmful”
Diabetes type 2 At least of below options: Consulting patients to take blood glucose test No harmful and
- Combination of more than one again unnecessary treatments
antihyperglycemic medicines at the same AND
time No harmful treatment
- Insulin if there is any harmful treatment, coded as
“harmful”

137
Hypertension category I Combination of more than one antihypertensive At least one of below options No harmful and
medicines at the same time - Cerebral circulation medicines unnecessary treatments
- Analgesic
- Lipid lowering medicines
AND
No harmful treatment
if there is any harmful treatment, coded as
“harmful”
Supplemental and optional treatment
Supplemental Optional Not supplemental and
optional
Child’s acute diarrhea Consultation At least one of below options: No supplemental and
- Digestive ferment optional treatments
- Zinc
AND
No consultation
If there is “consultation” in treatment, coded as
“supplemental”
Child’s pneumonia Consultation At least one of below options: No supplemental and
- Alphachymotripsin optional treatments
- Vitamin
AND
No consultation
If there is “consultation” in treatment, coded as
“supplemental”
Pulmonary Tuberculosis Consultation At least one of below options: No supplemental and
- Paracetamol optional treatments
- Vitamin
AND
No consultation
If there is “consultation” in treatment, coded as
“supplemental”
Diabetes type 2 Consultation Vitamin No supplemental and
AND optional treatments
No consultation
If there is “consultation” in treatment, coded as
“supplemental”
Hypertension category I At least one of below options: At least one of below options: No supplemental and
- Consultation - Neuroleptics optional treatments
- Aspirin 81mg - Vitamin
AND
No “supplemental” treatment
If there is “supplemental” treatment, coded as
“supplemental”

138
Note: The answer keys were developed based on the Ministry of Health technical guidelines and were gone through three rounds of clinical experts’
consultations. The Ministry of Health technical guidelines were listed as below:
- Child’s acute diarrhea: Decision no. 4121/QĐ-BYT dated 28/10/2009
- Child’s pneumonia: Decision no. 101/QĐ-BYT dated 09/01/2014
- Pulmonary Tuberculosis : Decision no. 979/QĐ-BYT dated 24/03/2009
- Diabetes type 2: Decision no. 3879/QĐ-BYT dated 30/09/2014
- Hypertension category I: Decision no. 3192/QĐ-BYT dated 31/08/2010

139
Annex E. Additional Tables and Figures

Table AE.1. Percent commune health stations having set of equipment

Dien Dak Binh Dong Dong Rural Urban Total
Bien Lak Dinh Thap Nai Hanoi Hanoi
< 25% 0% 0% 0% 0% 0% 0% 0% 0%
25-
49.99% 0% 0% 0% 0% 0% 0% 0% 0%
50-
74.99% 18% 4% 24% 8% 3% 0% 0% 9%
75-100% 82% 93% 73% 90% 63% 58% 71% 74%
100% 0% 4% 2% 3% 35% 42% 29% 17%
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

140
Figure AE.1. Average share of recommended history question and physical
examinations asked by doctors ability quintile
(a) History questions

0.7
Share of recommneded questions

0.6
0.5
0.4
0.3
0.2
0.1
0.0
2nd
2nd

2nd

2nd

2nd
1st

1st

1st

1st

1st
4th
5th

4th
5th

4th
5th

4th
5th

4th
5th
3rd
3rd

3rd

3rd

3rd
Acute Diarrhea Child Pnemonia Tuberclousis Diabetes Type 2 Hypertension

(b) Physical examinations

0.9
Share of recommned examinations

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th
Acute Diarrhea Child Pnemonia Tuberclousis Diabetes Type 2 Hypertension
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Notes: 95% confidence intervals shown

141
Table AE.2. Percent of district hospitals having selective equipment
Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi
Essential Equipment
Adult scale 100 100 100 100 100 100 100
Child scale (250g gauge) 70 80 91 82 56 83 44
Infant scale (100g gauge) 90 90 82 100 100 100 33
Thermometer 100 100 100 100 100 100 100
Stethoscope 100 100 100 100 100 100 100
Pinard horn 100 100 91 100 100 100 89
Sphygmomanometer 100 100 100 100 100 100 100
Vaccine cold chain
(refrigerator, vaccine
flask) 100 100 100 91 78 94 67
Emergency and resuscitation care equipment
Monitor 90 100 100 100 100 100 56
Portable oxygen
concentrator 90 90 100 64 56 78 38
Medical ventilator 70 100 91 91 100 100 56
Child ventilator 30 70 27 27 44 78 33
Infant incubator 60 100 82 45 67 100 33
Anesthesia machine 80 100 100 73 78 94 50
Defbrillator and
pacemaker 90 90 55 91 89 89 44
ECG device 100 100 100 100 100 100 78
C-section toolkit 70 100 91 73 89 100 22
Diagnostic imaging/probe testing
X-ray machine 90 100 91 100 100 100 89
Ultrasound 100 100 100 91 100 100 100
CT scanner 0 20 27 18 44 11 56
ECG 30 50 73 73 78 72 89
Laboratory testing equipment
Blood analyzer 90 100 100 91 100 100 100
Bloog biochemical
analyzer 90 100 100 91 100 100 100
1
HbA1C testing 20 20 27 55 89 50 50
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015). Note: (1) For
HbA1C testing, the facilities answered “Yes” if they have equipment and capacity to provide this test.

142
Table AE.3. Percent of commune health stations having selective equipment
Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi
Adult scale 58 59 76 54 53 71 67
Child scale (250g gauge) 79 89 78 74 70 82 71
Infant scale (100g gauge) 70 81 54 59 80 89 67
Thermometer 76 93 85 87 70 93 90
Stethoscope 79 44 78 82 72 69 62
Pinard horn 55 74 76 69 65 69 76
Sphygmomanometer 61 70 73 74 57 64 62
Oxygen canister 3 15 32 59 82 49 48
Ambu bag 85 44 37 69 63 47 48
Stomach Cleansing toolkit 67 74 39 64 47 47 33
Delivery/natal care table 76 48 54 64 45 69 52
Oral fluid ventouse 52 81 71 64 70 64 57
Antiseptic autoclave/oven 48 74 61 72 72 71 67
Refrigerator 55 70 73 44 47 71 62
Ice box 85 74 59 79 70 64 43
Microscope 6 89 49 5 63 91 90
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

143
Table AE.4. Percent of district hospitals having selective services
Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi
Obstetric care services
Attended vertex presentation
normal delivery 100 100 100 100 100 100 22
Attended childbirth with twins
or more 100 100 91 73 67 89 22
First C-section 70 90 91 73 67 100 22
Second C-section 60 90 73 64 67 100 22
Neonatal resuscitation 80 100 100 100 89 83 44
Hepatitis B vaccination for
infants within the first 24 hours
after birth 100 100 91 100 89 94 33
Obstetric complications care services
Management of uterine rupture
risk and uterine rupture 80 100 91 91 67 100 44
Emergency management of
placental expulsion phase
metrorrhagia 90 100 100 100 100 100 44
Emergency management of
postpartum infection 80 100 100 100 100 100 44
Emergency management of
eclampsia 80 100 100 91 100 100 44
Emergency management of
neonatal tetanus 80 100 82 82 56 67 44
Child care services
Removal of airway foreign body 70 60 55 91 56 78 67
Respiratory advanced cardiac life
support 90 100 100 100 100 94 89
Anti-shock resuscitation 100 100 100 100 100 94 89
Diabetes care services
Management and treatment of
diabetes Type II 60 90 82 82 78 100 100
Insulin dependent diabetes care 60 90 73 100 89 83 56
Insulin dependent diabetes
management 60 60 55 65 67 78 56
Hypertension care services
Treatment of hypertension with
complications 70 100 100 100 100 89 78
Ambulatory hypertension
management 70 60 100 82 67 83 100
Appendicitis care services
Appendicitis operation 80 90 91 73 89 100 44
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).

144
Table AE.5. Percent of commune health stations having selective services
Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi
Obstetric care services
Antenatal care and pregnancy
management 100 100 100 100 100 98 100
Tetanus vaccination for pregnant
women 100 100 100 97 97 98 100
Attended vertex presentation formal
delivery 70 52 78 90 55 89 5
Hepatitis B vaccination for infants
within the first 24 hours after birth1 45 26 34 36 17 58 0
Child care services
Child acute diarrhea diagnosis and
treatment 70 89 93 100 97 96 100
Child pneumonia diagnosis and
treatment 70 93 88 100 95 93 95
Diabetes care services
Involvement in community based
screening 70 33 32 49 40 40 38
Management of risk groups on papers 85 48 24 87 53 71 48
Care management and observation 64 41 17 90 45 42 33
Periodical drug dispense 0 7 10 56 13 7 14
Hypertension care services
Involvement in community based
screening 82 52 24 49 35 62 29
Management of risk groups on papers 85 63 24 79 55 78 43
Care management and observation 79 63 22 92 45 49 24
Periodical drug dispense 39 63 15 64 33 18 19
Vaccination services
Hepatitis B vaccination 56 56 49 72 55 84 67
Rubella vaccination 100 100 100 100 100 100 100
Diphtheria – Pertussis – Tetanus
vaccination 97 100 100 100 100 100 100
Cholera vaccination 12 7 12 18 3 33 10
TB vaccination 100 100 100 100 100 98 100
Oral polio vaccination (OPV) 100 100 100 100 100 100 100
Japanese encephalitis vaccination 100 93 100 100 100 100 100
Typhoid vaccination 42 7 12 64 3 9 14
DPT-VGB-Hib vaccination 100 96 100 100 100 100 100
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (1) The service “Hepatitis B vaccination for infants within the first 24 hours after birth” is only surveyed
for facilities having service “Attended vertex presentation formal delivery”.

145
Table AE.6. Percent of district hospitals having selective pharmaceuticals1
Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi
Albendazole (200mg, capsule) 0 0 64 0 33 6 0
Altorvastatin (20mg, capsule) 10 20 55 91 100 11 0
Amitriptyline (25 mg, capsule) 10 10 27 36 0 0 22
Amlodipine (5mg, capsule) 100 80 82 100 100 83 100
Amoxicillin (500 mg, tablet) 90 30 100 100 100 94 78
Amoxicillin powder for oral
suspension sachets (Sachet 250mg) 70 30 64 36 67 22 22
Atenolol (50 mg, capsule) 10 10 36 73 89 17 22
Captopril (25 mg, capsule) 40 50 64 100 100 28 22
Ceftriaxone injection (1 g/vial) 0 10 36 27 44 56 56
Cephalexin (500 mg, tablet) 70 70 91 82 100 94 78
Ciprofloxacin (500 mg, capsule) 80 70 91 91 100 78 78
Co-trimoxazoles suspension
(40+200mg/5ml) 10 0 9 0 0 11 0
Diazepam (5 mg, capsule) 90 90 82 100 100 78 89
Diclofenac (50 mg, capsule) 20 40 64 73 89 17 67
Enalapril (10 mg, capsule/tablet) 10 30 91 27 56 17 33
Furosemide (40 mg, capsule) 90 90 91 100 100 83 78
Glibenclamide (5 mg, capsule) 0 0 36 9 22 6 0
Gliclazide (80 mg, capsule) 30 40 36 82 100 67 44
Ibuprofen (400 mg, capsule) 10 10 18 36 44 11 22
Insulin (100UI/ml, vial 10ml) 20 40 55 91 33 67 44
Metformin (500 mg, capsule) 80 40 73 91 78 72 100
Metronidazole (250 mg, capsule) 60 60 82 73 78 94 78
Nifedipine Retard (20 mg, capsule) 20 10 73 82 78 67 89
Omeprazole (20 mg, capsule) 70 90 100 73 56 94 78
Oresol (Sachet 1 liter) 40 20 64 55 44 50 67
Paracetamol (500mg, capsule) 100 90 100 91 100 100 100
Paracetamol
Suspension (24 mg/ml (120mg/5ml),
syrup) 40 20 64 55 33 39 44
Salbutamol inhaler (100mcg/dose) 70 40 9 82 78 67 56
Simvastatin (20 mg , capsule) 10 0 36 9 22 28 33
Valproic acid (200 mg, capsule) 0 0 9 0 0 0 11
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (1) This list of pharmaceuticals were developed based on SARA.

146
Table AE.7. Percent of commune health stations having selective pharmaceuticals1
Dien Dak Binh Dong Dong Rural Urban
Bien Lak Dinh Thap Nai Hanoi Hanoi
Albendazole (200mg, capsule) 24 15 6 47 28 20 5
Altorvastatin (20mg, capsule) 5 0 0 5 21 2 0
Amitriptyline (25 mg, capsule) 5 0 0 15 15 22 0
Amlodipine (5mg, capsule) 46 52 42 82 77 33 14
Amoxicillin (500 mg, tablet) 95 70 64 88 92 98 67
Amoxicillin powder for oral
suspension sachets (Sachet 250mg) 32 33 30 53 67 53 43
Atenolol (50 mg, capsule) 2 0 0 25 59 2 0
Captopril (25 mg, capsule) 66 48 0 80 90 16 10
Ceftriaxone injection (1 g/vial) 5 0 0 0 0 20 0
Cephalexin (500 mg, tablet) 78 59 42 95 92 91 52
Ciprofloxacin (500 mg, capsule) 66 30 33 78 85 64 43
Co-trimoxazoles suspension
(40+200mg/5ml) 0 4 6 13 8 13 14
Diazepam (5 mg, capsule) 17 11 48 15 49 31 14
Diclofenac (50 mg, capsule) 44 37 30 90 97 49 52
Enalapril (10 mg, capsule/tablet) 34 19 0 17 62 2 14
Furosemide (40 mg, capsule) 29 7 15 38 79 60 10
Glibenclamide (5 mg, capsule) 5 0 0 5 10 0 0
Gliclazide (80 mg, capsule) 7 4 0 33 82 2 5
Ibuprofen (400 mg, capsule) 12 0 0 40 74 13 5
Insulin (100UI/ml, vial 10ml) 2 0 0 0 5 0 0
Metformin (500 mg, capsule) 10 0 0 28 85 4 5
Metronidazole (250 mg, capsule) 63 59 64 82 79 71 33
Nifedipine Retard (20 mg, capsule) 61 11 3 68 59 44 38
Omeprazole (20 mg, capsule) 80 67 42 75 51 67 19
Oresol (Sachet 1 liter) 80 44 48 80 85 82 86
Paracetamol (500mg, capsule) 93 70 73 97 95 91 90
Paracetamol
Suspension (24 mg/ml (120mg/5ml),
syrup) 20 33 21 55 59 49 48
Salbutamol inhaler (100mcg/dose) 2 0 3 7 21 36 67
Simvastatin (20 mg , capsule) 5 0 0 0 0 0 0
Valproic acid (200 mg, capsule) 0 0 0 15 0 2 0
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: (1) This list of pharmaceuticals were developed based on SARA.

147
Table AE.8. Provider Vignettes IRT Score
Bivariate Correlations Multivariate
Regression
Estimate Standard Estimate Standard
Error Error
Commune -1.299*** (0.095) -0.580*** (0.102)
District 1.299*** (0.095) . .
Poverty rate -0.003 (0.002) 0.003 (0.003)
Age -0.014*** (0.004) -0.018*** (0.004)
Ethnic Minority -0.587*** (0.121) -0.322*** (0.103)
Male 0.188** (0.081) -0.022 (0.060)
Private Practice 0.369*** (0.070) 0.128** (0.058)
Intermediate -1.934*** (0.146) -1.673*** (0.412)
College -1.338* (0.711) -1.316* (0.782)
Bachelors 0.127* (0.077) -0.434 (0.382)
Primary/Intern Specialist 0.287*** (0.097) -0.527 (0.378)
Specialist level 1 0.533*** (0.072) -0.213 (0.365)
Specialist level 2 0.519 (0.350) . .
Masters 0.524*** (0.081) -0.199 (0.382)
PhD 0.709*** (0.050) . .
Binh Dinh -0.062 (0.163) 0.038 (0.189)
Dak Lak -0.250* (0.137) -0.058 (0.222)
Dien Bien -0.415** (0.188) -0.137 (0.271)
Dong Nai -0.078 (0.113) 0.035 (0.181)
Dong Thap 0.225** (0.087) 0.175 (0.171)
Rural Hanoi 0.326*** (0.111) 0.256 (0.168)
Urban Hanoi 0.061 (0.168) . .
Formal Education -0.101 (0.082) 0.089 (0.126)
Twinning Programs 0.131 (0.080) 0.146 (0.127)
Direct Entry -0.156 (0.127) . .
Satisfaction – Salary -0.200*** (0.037) -0.045 (0.034)
Satisfaction – Allowance -0.142*** (0.037) -0.018 (0.041)
Satisfaction – Training Opportunity -0.032 (0.035) -0.002 (0.032)
Satisfaction – Promotion Opportunity -0.030 (0.037) 0.019 (0.036)
Satisfaction – Occupational Safety -0.010 (0.038) 0.035 (0.034)
Satisfaction – Hospital Security -0.121*** (0.033) -0.046 (0.030)
Satisfaction – Working Environment -0.060 (0.041) 0.013 (0.038)
Satisfaction – Availability Of Medicines -0.159*** (0.039) 0.016 (0.034)
Satisfaction – Availability of Equipment -0.180*** (0.037) -0.037 (0.038)
Satisfaction – Working Pressure -0.115*** (0.038) 0.035 (0.041)
Satisfaction – Adequate Staffing -0.188*** (0.037) -0.029 (0.035)
Constant . . 1.071 (0.510)
Number of Observations . . 963 .

148
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: significance: *** p<0.01, ** p<0.05, * p<0.1.

Table AE.9. Provider Average Clinical Effort

Bivariate Correlations Multivariate
Regression
Estimate Standard Estimate Standard
Error Error
Commune 0.505*** (0.109) 0.754*** (0.139)
District -0.505*** (0.109) . .
Poverty rate -0.003 (0.002) 0.000 (0.004)
Age -0.001 (0.005) -0.010 (0.007)
Ethnic Minority -0.316** (0.148) -0.367** (0.182)
Male -0.291*** (0.110) -0.226** (0.112)
Private Practice -0.270** (0.118) -0.099 (0.129)
Intermediate -0.087 (0.140) -1.054** (0.446)
College 0.039 (0.488) -0.447 (0.603)
Bachelors 0.219** (0.105) -0.441 (0.401)
Primary/Intern Specialist -0.093 (0.133) -0.569 (0.411)
Specialist level 1 -0.217 (0.141) -0.463 (0.414)
Specialist level 2 0.230 (0.338) . .
Masters -0.285*** (0.083) -0.313 (0.458)
Binh Dinh -0.521*** (0.137) -0.856*** (0.243)
Dak Lak -0.203 (0.160) -0.474* (0.259)
Dien Bien -0.020 (0.172) -0.243 (0.367)
Dong Nai 0.149 (0.197) -0.414 (0.262)
Dong Thap -0.130 (0.134) -0.666*** (0.229)
Rural Hanoi 0.336*** (0.127) -0.183 (0.222)
Urban Hanoi 0.436** (0.198) . .
Formal Education -0.201* (0.112) . .
Twinning Programs 0.234** (0.115) 0.092 (0.130)
Direct Entry -0.183 (0.268) 0.025 (0.249)
Satisfaction – Salary -0.004 (0.050) 0.028 (0.072)
Satisfaction – Allowance -0.041 (0.050) -0.108 (0.072)
Satisfaction – Training Opportunity 0.013 (0.048) 0.038 (0.055)
Satisfaction – Promotion Opportunity 0.022 (0.055) 0.021 (0.067)
Satisfaction – Occupational Safety -0.011 (0.047) -0.009 (0.056)
Satisfaction – Hospital Security -0.054 (0.044) -0.015 (0.049)
Satisfaction – Working Environment 0.020 (0.051) -0.009 (0.058)
Satisfaction – Availability Of Medicines 0.024 (0.050) -0.031 (0.061)
Satisfaction – Availability of Equipment 0.070 (0.053) 0.031 (0.060)
Satisfaction – Working Pressure 0.047 (0.056) 0.068 (0.063)
Satisfaction – Adequate Staffing 0.152*** (0.046) 0.050 (0.056)
Constant 1.368 (0.558)
Number of Observations 365

149
Source: Calculations from the Vietnam District and Commune Health Facility Survey (2015).
Note: significance: *** p<0.01, ** p<0.05, * p<0.1.

Table AE.10. District-Commune Differences - Hawthorne Effect Check

(1) (2) (3) (4) (5)

All patients
District -
Overall Commune District Controlled
Commune
Mean Mean Mean Difference
Difference
Interaction Effort 0.02 0.39 -0.11 -0.50*** -0.24***
Time with Doctor
5.48 7.37 4.81 -2.56*** -1.43***
(Minutes)
Number of Questions 6.78 7.47 6.54 -0.93*** -0.41***
Number of Exams 2.11 2.39 2.01 -0.38*** -0.17***
Test Ordered 38% 4% 50% 0.46*** 0.46***
Patient Satisfied 75% 83% 72% -0.12*** -0.11***
Referral 9% 5% 10% 0.05*** 0.08***
Price 24,354 8,823 29,794 20,970*** 11,646***
Number of Medications 3.31 3.31 3.30 -0.00 0.06
Antibiotics 41% 45% 39% -0.06*** -0.08***
Number of observations 6,745 1,757 4,988

(6) (7) (8) (9) (10)

Excluding first 5 patients
District -
Overall Commune District Controlled
Commune
Mean Mean Mean Difference
Difference
Interaction Effort -0.11 0.21 -0.18 -0.39*** -0.15***
Time with Doctor
4.82 6.27 4.49 -1.78*** -0.95***
(Minutes)
Number of Questions 6.42 6.92 6.30 -0.62*** -0.21
Number of Exams 2.05 2.33 1.98 -0.35*** -0.11**
Test Ordered 0.39 0.03 0.47 0.44*** 0.40***
Patient Satisfied 0.74 0.84 0.71 -0.13*** -0.11***
Referral 0.09 0.04 0.10 0.06*** 0.08***
Price 24,852 5,951 29,143 23,192*** 11,595***
Number of Medications 3.35 3.43 3.33 -0.09* 0.09
Antibiotics 0.41 0.46 0.40 -0.06*** -0.08***
Number of observations 5,247 971 4,276

150
As a check on the Hawthorne effect in the direct observation data, we can examine
how effort changes after the 5 patients. The comparison in the above table provides
the mean of overall effort and effort after 5 patients for districts and communes
respectively (note that some communes are not in this sample due to the low patient
load). When we exclude the first 5 patients, effort declines by 0.1 standard-deviations overall,
but more so in the commune (0.18sd) than the district (0.07sd). Consequently, the district-
commune difference in clinical effort, which is one of our key coefficients of interest, reduces
from -.50sd to -0.39sd (unadjusted) and from -.24sd to -.15sd (adjusted). The table also shows
other key outcomes of interest, with similarly small or zero changes. Note that this alters the
sample as in some communes, providers did not see 5 patients and these are therefore excluded
entirely. To the extent that even without a Hawthorne effect, effort is higher in these communes,
the difference in coefficients reflects both the difference in sample composition and the
Hawthorne effect, which cannot be identified separately.

151