Public Health Nutrition: 22(1), 85–94 doi:10.

1017/S1368980018002926

Individual-, maternal- and household-level factors associated with

stunting among children aged 0–23 months in Bangladesh
Sabuj Kanti Mistry1,*, Md. Belal Hossain1, Fouzia Khanam1, Fahmida Akter1,
Mahmood Parvez1, Fakir Md Yunus2, Kaosar Afsana3 and Mahfuzar Rahman1
1
Research and Evaluation Division, BRAC, BRAC Centre, 75 Mohakhali, Dhaka 1212, Bangladesh: 2College of
Pharmacy and Nutrition, The University of Saskatchewan, Saskatoon, SK, Canada: 3Health Nutrition and Population
Programme, BRAC, Dhaka, Bangladesh

Submitted 28 March 2018: Final revision received 21 September 2018: Accepted 25 September 2018: First published online 8 November 2018

Abstract
Objective: Childhood stunting remains a major public health concern in
Bangladesh. To accelerate the reduction rate of stunting, special focus is required
during the first 23 months of a child’s life when the bulk of growth takes place.
Therefore the present study explored individual-, maternal- and household-level
factors associated with stunting among children under 2 years of age in
Bangladesh.
Design: Data were collected through a nationwide cross-sectional survey
conducted between October 2015 and January 2016. A two-stage cluster random
sampling procedure was applied to select 11 428 households. In the first stage, 210
enumerations areas (EA) were selected with probability proportional to EA size
(180 EA from rural areas, thirty EA from urban slums). In the second stage, an
average of fifty-four households were selected from each EA through systematic
random sampling.
Setting: Rural areas and urban slums of Bangladesh.
Participants: A total of 6539 children aged 0–23 months.
Results: Overall, 29·9 % of the children were stunted. After adjusting for all
potential confounders in the modified Poisson regression model, child’s gender,
birth weight (individual level), maternal education, age at first pregnancy, nutrition
(maternal level), administrative division, place of residence, socio-economic
status, food security status, access to sanitary latrine and toilet hygiene condition Keywords
(household level) were significantly associated with stunting. Bangladesh
Conclusions: The study identified a number of potentially addressable multilevel Childhood stunting
risk factors for stunting among young children in Bangladesh that should be World Health Assembly
addressed through comprehensive multicomponent interventions. Urban slums

Linear growth faltering in childhood, also familiarly known environmental issues; socio-economic status; and cultural
as stunting, is a major global health concern(1). Stunting influences with childhood stunting(4). Causes of malnutri-
has both short-term and long-term consequences on tion in early childhood have also been extensively ana-
health and development throughout the life cycle(2–4) as lysed by both Fenske et al.(9) and Goudet et al.(10). These
well as across generations(5). Nearly half of all deaths authors have classified the determinants of childhood
among children under 5 years of age (under-5s) are attri- stunting into immediate (individual level), intermediate
butable to undernutrition, which has both direct and (individual/household level) and underlying (maternal,
indirect impacts on economic productivity and growth(6). household and regional level) factors.
It has been estimated that stunting can reduce a country’s Globally, although the prevalence of stunting is drop-
gross domestic product by up to 3 %(7) and that as adults, ping, there are still around 156 million under-5s who are
stunted children earn 20 % less than the non-stunted(8). stunted(11). In 2015, Asia was the home to 56 % of all
The causes of stunting are multisectoral and multi- stunted under-5 children, while Africa’s share was 37 %(11).
factorial. The WHO conceptual framework on childhood The prevalence of stunting among under-5s has decreased
stunting describes the complex interaction of household by an average of 2·7 % per year (from 51 to 36 %) between
characteristics; water, sanitation and hygiene; 2004 and 2014 in Bangladesh(12). This is far below the

*Corresponding author: Email [email protected] © The Authors 2018

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 86 SK Mistry et al.
targeted annual reduction of 3·3 %, necessary to meet the union (rural areas) or ward (urban slums), which is the
World Health Assembly target of 40 % reduction by 2025. lowest administrative unit in Bangladesh. A complete list
Therefore, Bangladesh is less likely to achieve the World of unions and wards, collected from the Bangladesh
Health Assembly target for reducing childhood stunting Bureau of Statistics, was used as the sampling frame for
with the current pace of reduction(6). the first stage of sampling. In the second stage, starting
Bangladesh has experienced rapid urbanization in from the north-west corner of an EA with systematic ran-
recent years and is expected to have more than 50 % urban dom sample of five households, on average, fifty-four
population by 2050(13). As a result, more people are households were selected per EA to provide statistically
residing in urban slum settlements – a number currently reliable estimates for rural areas and urban slums sepa-
estimated to be 2·2 million(14,15). Congested living condi- rately. Finally, the random selection covered all adminis-
tions, poverty, lack of knowledge on nutrition, and trative divisions and fifty-eight out of sixty-four districts of
improper hygiene and sanitation facilities all make slum Bangladesh. If a household contained more than one
dwellers more vulnerable to recurrent infections which under-5 child, the smallest child was included in the sur-
result in health and nutritional deprivation(16,17). Evidence vey. The present analysis was performed among all the
also suggests that the prevalence of stunting among under- children (n 6539) aged 0–23 months of the selected
5s is higher among children living in these slums com- households (Fig. 1).
pared with the overall urban population of Bangladesh
(40 v. 26 %)(18).
Data collection tools and techniques
Given the situation, careful exploration of the most
A pre-tested structured questionnaire was administered to
important determinants of stunting is critical for govern-
gather information pertaining to the child, mother and
ments and development partners. This would enable them
household through face-to-face interview with the child’s
to make strategic decisions to reduce the stunting pre-
mother. Anthropometric measurements were taken using
valence as well as to plan, design and monitor effective
standardized procedures. Data collection was carried out
interventions(12,19). However, only a few studies have
electronically in ODK (Open Data Kit), an Android-based
been conducted to date focusing on comprehensively
open-source mobile platform software(24).
explaining the myriad factors associated with childhood
Skilled female interviewers (with health survey experi-
stunting(19–22). Furthermore, it is broadly accepted that the
ence) were recruited to minimize reporting bias as the
first 1000 d from conception through the first 2 years of life
respondents were women(25). Fifteen days of intensive
is the most critical period for stunting(3,23). Therefore, the
training was provided to the interviewers which included
present study was carried out to identify the risk factors
lectures, mock interviews, role play and field practice at
of stunting at different levels (i.e. individual, maternal
the community level. A multilayer monitoring system was
and household) among Bangladeshi children aged
employed to maintain the quality of data and several tasks
0–23 months.
such as spot checking, thorough checking and back
checking of the questionnaire were performed. Any dis-
crepancies arising were resolved through re-interviewing
Methods
the respective mothers.
Study design and participants
The data used for the present study were part of a Variables assessed and measured
nationwide cross-sectional health-care survey carried out Child characteristics considered for the present study
between October 2015 and January 2016. The base survey were: age in months; gender (boy/girl); birth order (1–2,
was conducted by Building Resources Across Commu- 3–4, ≥5); birth weight in kilograms (<2·5, ≥2·5); immedi-
nities (BRAC) to explore the overall health status of chil- ate breast-feeding after birth (yes/no); prelacteal feeding
dren, mothers and senior citizens of rural areas and urban (yes/no); exclusive breast-feeding up to 6 months
slums in Bangladesh. The survey excludes urban non-slum (yes/no); and suffering from diarrhoea within the last
areas as BRAC operates its health interventions in rural 3 months (yes/no). Although the study did not collect data
areas and urban slums only. A slum is defined as a cluster on exact birth weight, mothers were asked to report,
of compact settlements of five or more households grown according to their perception, whether their child was
unsystematically and haphazardly on government and/or ‘very small’, ‘smaller than average’, ‘average’, ‘larger than
private vacant land(14). A two-stage cluster random sam- average’ or ‘very large’ at birth. A birth size of ‘very small’
pling procedure was applied and a total of 11 428 or ‘smaller than average’ was categorized as low birth
households (response rate 94·9 %) having at least one weight (<2·5 kg), while a birth size of ‘average’ to ‘very
under-5 child was selected in the base survey. In the first large’ was categorized as normal birth weight (≥2·5 kg).
stage, 210 enumeration areas (EA) were selected randomly Maternal characteristics were: age in years; level of edu-
with probability proportional to EA size, with 180 EA from cation (no education, primary incomplete: grade 1–4,
rural areas and thirty EA from urban slums. An EA is a primary complete/secondary incomplete: grade 5–9,

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 Factors of early childhood stunting 87

Total households surveyed

(n 11 428)

Total children aged 0–59 months

(n 11 428)

Total children aged 23–59 months

(n 4876)

Total children aged 0–23 months

(n 6552)

Anthropometric measurement not taken

(n 13)

Anthropometric measurement taken

(n 6539)

Male (n 3334) Female (n 3205)

Fig. 1 Study profile and enrolment of participants

secondary or higher: grade 10 or higher); occupation supine position. Weight was measured with an electronic
(housewife/working outside), BMI in kg/m2 (under- bathroom scale at 0·1 kg precision. Height and weight of
weight: <18·5, normal: 18·5–24·9, overweight: 25·0–29·9, the children and mothers were taken separately. BMI of
obese: ≥30·0), age at first pregnancy in years (<20, ≥20); the mothers was calculated using the formula(27): [weight
vitamin A supplementation immediately after delivery (kg)]/[height (m)]2.
(yes/no); Fe supplementation during pregnancy (yes/no); The UNICEF and WHO Joint Monitoring Programme’s
using soap before eating (yes/no); and using soap after operational definition for improved water supply, treated
defecation (yes/no). Household-level characteristics were: water and improved sanitation facilities was used(28). The
administrative division (Barisal, Chittagong, Dhaka, households were considered food insecure if they faced
Khulna, Rajshahi, Rangpur, Sylhet); place of residence any food insufficiencies in the 30 d prior to the survey.
(rural areas/urban slums); household wealth quintile
(lowest, low, middle, high, highest); size of household
Outcome measurement
(≤4, >4); household food security status (secure/inse-
The outcome variable was stunting of children aged 0–23
cure); household source of drinking-water (safe/unsafe);
months. A child with height-for-age more than 2 SD below
type of latrine (improved/unimproved); and toilet hygiene
the median height-for-age of the WHO reference popu-
condition (hygienic/unhygienic).
lation (height-for-age Z-score < − 2) was considered
Construction of the wealth index was based on factor
stunted(29).
analysis of principal component analysis of key socio-
economic variables(26). The variables considered were:
type of wall, floor and roof of the house; ownership of a Statistical analysis
radio, television, computer, bicycle, mobile/telephone, Descriptive analysis was performed to assess the dis-
refrigerator, wardrobe, table, chair, watch, bed, sewing tribution of the variables. The χ2 test was used to compare
machine, bike, motor vehicle and livestock; and access to the prevalence of stunting within different categories of a
solar electricity. variable with 5 % level of significance. As the EA were
Age of the child was recorded from the immunization scattered throughout the country with different demo-
card or birth certificate. A local event calendar was used graphic and social characteristics, there was a potential
when this information was unavailable. Height was mea- chance for certain variability in stunting prevalence among
sured to the nearest 0·1 cm using a board with a wooden clusters. Moreover, mothers from the same EA could share
base and a movable headpiece, with the participant in certain types of unobserved cultural and environmental

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 88 SK Mistry et al.
factors. This nested source of variability was addressed by Table 1 Demographic and socio-economic characteristics of the
incorporating random effects with the usual fixed effects in sample of Bangladeshi children under 2 years old (n 6539), October
2015–January 2016
the model(30). Therefore, considering EA as clusters, the
modified Poisson regression was executed to explore the Characteristic n %
factors truly associated with early childhood stunting(31). Child’s age (months)
EA were considered as random effects with exchangeable 0–5 1550 23·70
correlation among clusters, while other variables were 6–11 2032 31·08
12–23 2957 45·22
considered as fixed effects. The parameters of the model Child’s gender
were estimated through the generalized estimating equa- Female 3205 49·01
tions approach to estimate the true effect of associated Male 3334 50·99
Maternal age (years)
factors and the unknown correlation of the outcome < 20 1055 16·13
among clusters(32). Due to the convergence problem with 20–29 4147 63·42
many covariates in the binomial model, the modified ≥ 30 1337 20·45
Maternal education
Poisson regression model was performed, which is No education 888 13·58
equivalent to the binomial model while estimating the risk Primary incomplete* 1109 16·96
ratio(31). The crude risk ratio (cRR), adjusted risk ratio Primary or secondary incomplete† 3605 55·13
Secondary or higher‡ 937 14·33
(aRR) and 95 % CI were estimated at 5 % level of sig- Maternal occupation
nificance. The children’s anthropometric index was con- Housewife 6175 94·43
structed in WHO AnthroPlus software and all statistical Working outside 364 5·57
Maternal BMI (kg/m2)
analyses were performed using the statistical software < 18·5 1348 20·61
package Stata version 13.0. 18·5–24·9 4102 62·73
25·0–29·9 944 14·44
≥ 30·0 145 2·22
Ethical approval Division
Barisal 453 6·93
The guidelines of the Declaration of Helsinki were fol- Chittagong 746 11·41
lowed while conducting the study and ethical clearance Dhaka 2435 37·24
was sought from the Bangladesh Medical Research Khulna 859 13·14
Rajshahi 745 11·39
Council (reference number BMRC/NREC/2013-2016/802) Rangpur 801 12·25
before the study. Both written and verbal consent were Sylhet 500 7·65
taken from the mothers of the children prior to the Place of residence
Rural 5679 86·85
interview. Urban slum 860 13·15
Household wealth quintile
Lowest 1323 20·23
Low 1297 19·83
Results Middle 1268 19·39
High 1322 20·22
Sample description Highest 1329 20·32
n 6539 100·0
Of the 6539 children less than 2 years of age sampled,
23·7 % were 0–5 months old, 31·1 % were 6–11 months *Primary incomplete = grade 1–4.
†Primary or secondary incomplete = grade 5–9.
and 45·2 % were 12–23 months old (Table 1). The male: ‡Secondary or higher = grade 10 or higher.
female ratio was close to unity (49·0 v. 51·0 %, respec-
tively). Meanwhile, 63·4 % of mothers were 20–29 years of
birth order (29·2 % for the first or second child, 30·4 % for
age and 20·5 % were aged 30 years or above. In addition,
the third or fourth child, and 37·2 % for the fifth or later
13·6 % of children’s mothers had no education while
child; P = 0·006). Stunting prevalence was also sig-
14·3 % had completed grade 10 or higher. Most of the
nificantly higher among low-birth-weight babies (weight
mothers (94·4 %) were housewives. Also, 20·6 % of the
<2·5 kg) than babies with normal birth weight (42·0 v.
mothers were underweight, 14·4 % were overweight and
29·0 %; P < 0·001). The study further showed that the
only 2·2 % were obese. More than one-third of children
prevalence of stunting was considerably lower among
were sampled from Dhaka division and 86·9 % lived in
children who were exclusively breast-fed up to 6 months
rural areas. A total of 40·0 % of children belonged to poor
compared with those who were not (17·1 v. 32·9 %;
households (20·2 % lowest, 19·8 % low; Table 1).
P < 0·001). Children suffering from diarrhoea episodes
within the last 3 months were also more likely to be
Prevalence of childhood stunting stunted (34·7 v. 29·4 %; P = 0·006).
Overall, 29·9 % of children were stunted (Table 2). The Although not significant, the prevalence of childhood
prevalence of stunting increased with age of the child stunting was found to be increasing with mother’s age.
(P < 0·001) and was higher among boys than girls (32·6 v. However, in contrast, the prevalence of stunting
27·1 %; P < 0·001). The prevalence increased with higher decreased significantly with increasing maternal education

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 Factors of early childhood stunting 89
Table 2 Bivariate analysis of child, maternal and household char- Table 2 Continued
acteristics with childhood stunting in Bangladeshi children under 2
years old (n 6539), October 2015–January 2016 Stunting

Stunting Characteristic n % P value

Characteristic n % P value Rajshahi 745 26·85

Rangpur 801 28·96
Child characteristics Sylhet 500 41·40
Age (months) Place of residence
0–5 1550 15·42 < 0·001 Rural 5679 29·57 0·099
6–11 2032 21·70 Urban slum 860 32·33
12–23 2957 43·19 Wealth quintile
Gender Lowest 1323 33·18 < 0·001
Female 3205 27·11 < 0·001 Low 1297 34·77
Male 3334 32·63 Middle 1268 28·79
Birth order High 1322 28·06
1–2 4438 29·18 0·006 Highest 1329 24·91
3–4 1746 30·36 Household size
≥5 355 37·18 ≤4 2482 29·65 0·704
Birth weight (kg) >4 4057 30·10
< 2·5 476 42·02 < 0·001 Food security status
≥ 2·5 6063 28·98 Insecure 1673 34·91 < 0·001
Breast-feeding immediately after birth Secure 4866 28·22
No 938 30·38 0·742 Drinking-water safe
Yes 5601 29·85 No 6049 30·07 0·375
Prelacteal feeding Yes 490 28·16
No 3179 29·10 0·154 Type of latrine
Yes 3360 30·71 Unimproved 5067 31·42 < 0·001
Exclusive breast-feeding up to 6 months Improved 1472 24·80
No 5319 32·88 < 0·001 Toilet hygiene condition
Yes 1220 17·05 Unhygienic 5155 31·79 < 0·001
Diarrhoea within 3 months Hygienic 1384 22·98
No 5923 29·43 0·006 n 6539 29·93
Yes 616 34·74
Maternal characteristics
Age (years)
< 20 1055 28·06 0·111
20–29 4147 29·76
≥ 30 1337 31·94 (P < 0·001). A total of 38·7 % of children whose mothers
Education had no education were stunted, compared with 28·3 and
No education 888 38·74 < 0·001 22·0 %, respectively, among children whose mothers had
Grade 1–4 1109 34·99
Grade 5–9 3605 28·27 completed grade 5–9 and grade 10 or more. Maternal
Grade 10 or more 937 21·99 nutritional status, measured by BMI, also predicted
Occupation decreased childhood stunting (P < 0·001). Also, the pre-
Housewife 6175 29·89 0·808
Working outside 364 30·49 valence of stunting was higher among children whose
BMI (kg/m )2
mothers became pregnant for the first time before the age
< 18·5 1348 35·53 < 0·001 of 20 years (P = 0·020). At the same time, the prevalence of
18·5–24·9 4102 29·28
25·0–29·9 944 25·42 stunting was also significantly lower among children
≥ 30·0 145 25·52 whose mother received vitamin A supplements immedi-
Age at first pregnancy (years) ately after delivery (P = 0·036), received Fe supplements
< 20 4873 30·70 0·020
≥ 20 1666 27·67 during pregnancy (P < 0·001), used soap before eating
Vitamin A supplementation (P = 0·003) and used soap after defecation (P < 0·001).
No 3208 31·14 0·036 Disparity in childhood stunting prevalence was
Yes 3331 28·76
Fe supplementation observed among households from different administrative
No 1591 33·69 < 0·001 divisions (P < 0·001). The prevalence was highest in Sylhet
Yes 4948 28·72 division (41·4 %) and lowest in Khulna (24·5 %). Also, the
Using soap before eating
No 5223 30·75 0·003 prevalence was higher among the children from urban
Yes 1316 26·67 slums than among those from rural areas (32·3 v. 29·6 %;
Using soap after defecation P = 0·099). Stunting prevalence was significantly higher
No 1530 33·73 < 0·001
Yes 5009 28·77 among children from poor households (poorest: 33·2 %,
Household characteristics richest: 24·9 %; P < 0·001). However, the prevalence was
Division considerably lower among children from food-secure
Barisal 453 26·05 < 0·001
Chittagong 746 29·49 households (P < 0·001), and among households with
Dhaka 2435 31·62 improved (P < 0·001) and hygienic (P < 0·001) latrines
Khulna 859 24·45 (Table 2).

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 90 SK Mistry et al.
Risk factors for childhood stunting Table 3. We found that the children aged 6–11 months had
Individual assessments of each of the covariates with 34 % higher risk of stunting than those aged 0–5 months
childhood stunting are presented in Table 3. Significant (aRR = 1·34; 95 % CI 1·11, 1·62), while children aged 12–
associations were observed between stunting and different 23 months had 2·65 times higher risk (aRR = 2·65; 95 % CI
child characteristics such as age, gender, birth order, birth 2·20, 3·20). Meanwhile, boys had 21 % higher risk than
weight, exclusive breast-feeding up to 6 months and girls (aRR = 1·21; 95 % CI 1·12, 1·30). It was also notable
diarrhoea episodes. Maternal characteristics such as level that low-birth-weight babies had more than 50 % higher
of education, BMI, age at first pregnancy, Fe supple- risk of stunting than babies with normal birth weight
mentation during pregnancy, use of soap before eating (aRR = 1·51; 95 % CI 1·35, 1·68).
and after defecation were also significantly associated with The risk of stunting was 12 % lower among children
childhood stunting. Moreover, household-level factors whose mothers had completed grade 5–9 (aRR = 0·88;
such as administrative division, wealth quintile, food 95 % CI 0·80, 0·97) and was 22 % lower among children of
security status, type of latrine and toilet hygiene condition mothers who had completed grade 10 or more (aRR =
were significantly associated with childhood stunting. 0·78; 95 % CI 0·67, 0·92) compared with the children of
The final adjusted modified Poisson regression model uneducated mothers. The children of underweight
captured 3 % of variation (P <0·001) in outcome among mothers had 11 % higher risk of stunting than those of
clusters. The adjusted estimated effects for the factors normal-weight mothers (aRR = 1·11; 95 % CI 1·02, 1·20).
associated with childhood stunting are also presented in Meanwhile, there was 9 % higher risk of stunting among

Table 3 Association of child, maternal and household characteristics with childhood stunting in Bangladeshi children under 2 years old
(n 6539), October 2015–January 2016

Crude Adjusted

Characteristic RR 95 % CI P value RR 95 % CI P value

Child characteristics
Age (months)
0–5 1·00 Ref. – 1·00 Ref. –
6–11 1·40 1·18, 1·64 < 0·001 1·34 1·11, 1·62 0·003
12–23 2·77 2·37, 3·23 < 0·001 2·65 2·20, 3·20 < 0·001
Gender
Female 1·00 Ref. – 1·00 Ref. –
Male 1·20 1·11, 1·30 < 0·001 1·21 1·12, 1·30 < 0·001
Birth order
1–2 1·00 Ref. – 1·00 Ref. –
3–4 1·02 0·94, 1·12 0·593 0·94 0·85, 1·03 0·195
≥5 1·19 1·02, 1·39 0·029 0·96 0·81, 1·14 0·624
Birth weight (kg)
≥ 2·5 1·00 Ref. – 1·00 Ref. –
< 2·5 1·51 1·35, 1·69 < 0·001 1·51 1·35, 1·68 < 0·001
Breast-feeding immediately after birth
No 1·00 Ref. – 1·00 Ref. –
Yes 0·96 0·87, 1·07 0·454 1·05 0·95, 1·16 0·337
Prelacteal feeding
No 1·00 Ref. – 1·00 Ref. –
Yes 1·02 0·95, 1·10 0·524 1·03 0·96, 1·11 0·458
Exclusive breast-feeding up to 6 months
No 1·00 Ref. – 1·00 Ref. –
Yes 0·51 0·44, 0·59 < 0·001 0·96 0·81, 1·15 0·664
Diarrhoea within 3 months
No 1·00 Ref. – 1·00 Ref. –
Yes 1·15 1·03, 1·28 0·011 1·02 0·92, 1·14 0·662
Maternal characteristics
Age (years)
< 20 1·00 Ref. – 1·00 Ref. –
20–29 0·94 0·84, 1·06 0·338 0·97 0·86, 1·10 0·622
≥ 30 1·05 0·96, 1·15 0·279 1·05 0·94, 1·17 0·413
Education
No education 1·00 Ref. – 1·00 Ref. –
Grade 1–4 0·93 0·83, 1·04 0·221 0·99 0·88, 1·10 0·798
Grade 5–9 0·76 0·69, 0·84 < 0·001 0·88 0·80, 0·97 0·009
Grade 10 or more 0·61 0·52, 0·70 < 0·001 0·78 0·67, 0·92 0·003
Occupation
Housewife 1·00 Ref. – 1·00 Ref. –
Working outside 1·02 0·89, 1·18 0·755 1·00 0·87, 1·16 0·966

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 Factors of early childhood stunting 91

Table 3 Continued

Crude Adjusted

Characteristic RR 95 % CI P value RR 95 % CI P value

2
BMI (kg/m )
< 18·5 1·19 1·10, 1·30 < 0·001 1·11 1·02, 1·20 0·011
18·5–24·9 1·00 Ref. – 1·00 Ref. –
25·0–29·9 0·89 0·79, 1·02 0·090 0·96 0·85, 1·09 0·541
≥ 30·0 0·91 0·69, 1·18 0·472 1·02 0·78, 1·32 0·909
Age at first pregnancy (years)
≥ 20 1·00 Ref. – 1·00 Ref. –
< 20 1·13 1·03, 1·24 0·012 1·09 0·99, 1·19 0·097
Vitamin A supplementation
No 1·00 Ref. – 1·00 Ref. –
Yes 0·94 0·87, 1·02 0·165 1·01 0·93, 1·09 0·857
Fe supplementation
No 1·00 Ref. – 1·00 Ref. –
Yes 0·86 0·80, 0·94 < 0·001 0·96 0·89, 1·04 0·343
Using soap before eating
No 1·00 Ref. – 1·00 Ref. –
Yes 0·86 0·78, 0·96 0·005 0·95 0·86, 1·06 0·366
Using soap after defecation
No 1·00 Ref. – 1·00 Ref. –
Yes 0·87 0·81, 0·94 0·001 1·01 0·93, 1·09 0·883
Household characteristics
Division
Barisal 0·84 0·69, 1·04 0·108 0·89 0·74, 1·07 0·217
Chittagong 0·94 0·81, 1·08 0·384 0·98 0·84, 1·14 0·779
Dhaka 1·00 Ref. – 1·00 Ref. –
Khulna 0·78 0·65, 0·93 0·007 0·84 0·72, 0·99 0·038
Rajshahi 0·87 0·73, 1·03 0·116 0·93 0·79, 1·10 0·404
Rangpur 0·93 0·78, 1·11 0·411 0·97 0·83, 1·13 0·708
Sylhet 1·34 1·07, 1·68 0·011 1·35 1·12, 1·64 0·002
Place of residence
Rural 1·00 Ref. – 1·00 Ref. –
Urban slum 1·09 0·91, 1·30 0·345 1·19 1·00, 1·41 0·048
Wealth quintile
Lowest 1·00 Ref. – 1·00 Ref. –
Low 1·02 0·91, 1·13 0·748 1·03 0·93, 1·15 0·546
Middle 0·85 0·76, 0·94 0·002 0·90 0·81, 1·01 0·063
High 0·83 0·73, 0·94 0·003 0·93 0·82, 1·05 0·231
Highest 0·68 0·59, 0·79 < 0·001 0·84 0·72, 0·98 0·030
Household size
≤4 1·00 Ref. – 1·00 Ref. –
>4 0·98 0·90, 1·06 0·593 1·03 0·95, 1·13 0·436
Food security status
Insecure 1·00 Ref. – 1·00 Ref. –
Secure 0·81 0·74, 0·88 < 0·001 0·93 0·85, 1·00 0·064
Drinking-water safe
No 1·00 Ref. – 1·00 Ref. –
Yes 0·86 0·73, 1·02 0·081 1·03 0·88, 1·21 0·729
Type of latrine
Unimproved 1·00 Ref. – 1·00 Ref. –
Improved 0·76 0·69, 0·84 < 0·001 0·88 0·79, 0·98 0·025
Toilet hygiene condition
Unhygienic 1·00 Ref. – 1·00 Ref. –
Hygienic 0·74 0·66, 0·82 < 0·001 0·90 0·79, 1·02 0·086

RR, risk ratio; Ref., reference category.

children whose mothers became pregnant for the first time children from the richest households had 16 % lower risk
before the age of 20 years (aRR = 1·09; 95 % CI 0·99, 1·19). of stunting than those from the poorest households
It was found that children from Khulna division had (aRR = 0·84; 95 % CI 0·72, 0·98). Moreover, children from
16 % lower risk of stunting compared with those of Dhaka food-secure households had 7 % lower risk (aRR = 0·93;
(aRR = 0·84; 95 % CI 0·72, 0·99), while the risk was 35 % 95 % CI 0·85, 1·00), households having improved toilet had
higher among children from Sylhet division (aRR = 1·35; 12 % lower risk (aRR = 0·88; 95 % CI 0·79, 0·98) and
95 % CI 1·12, 1·64). It was also evident that children from households having hygienic toilet had 10 % lower risk
urban slums had 19 % higher risk of stunting than those (aRR = 0·90; 95 % CI 0·79, 1·02) of stunting compared with
from rural areas (aRR = 1·19; 95 % CI 1·00, 1·41). Also, their corresponding reference (Table 3).

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 92 SK Mistry et al.
Discussion under 2 years of age. It was evident that children of
undernourished mothers were more likely to be stunted
The present study recognized several modifiable risk fac- than those of well-nourished mothers. Maternal under-
tors of stunting at individual, maternal and household nutrition is one of the major causes of intra-uterine growth
levels among children aged 0–23 months in Bangladesh. retardation which contributes about 20 % of the global
Although the overall stunting prevalence among children burden of childhood stunting(3). Moreover, stunted girls
has been declining in Bangladesh over the last decade, the are likely to grow to be stunted women in their later life,
reduction rate is not satisfactory(12). The findings of the which adversely affects growth of the fetus during their
present study suggest that an intervention considering pregnancy. Suboptimal growth of the fetus often results in
multilevel factors might increase the reduction rate, which low birth weight at delivery(46), which contributes to
in turn would help to achieve the World Health Assembly stunting in early childhood. However, the present study
targets on reducing childhood stunting. showed that maternal age at first pregnancy played a
Overall, about one-third of children were stunted and pivotal role in stunting among the children, which is also
there was marked variation in stunting prevalence among supported by other studies(37,47). Furthermore, mother’s
children based on place of residence, region and eco- education was identified as an important predictor of
nomic status. This agrees with the findings of the most stunting. Studies conducted in Bangladesh(22,48) and
recent national Demographic and Health Survey(12). The India(37) also confirmed that higher maternal education
present study also found that the prevalence of stunting significantly reduces stunting among children. Another
was comparatively higher among the children from urban multi-country study(49) showed that 43 % of the total
slums, which may be due to lack of adequate public ser- reduction in stunting can be attributed to improvement in
vices as well as health and sanitation facilities in slums. maternal education. Therefore, certain nutrition-sensitive
This is in line with the findings of some other studies(33–35). and nutrition-specific interventions targeting adolescent
Furthermore, as reported elsewhere(22), stunting pre- girls and mothers may also be central to the national
valence was considerably higher in Sylhet division, which efforts in reducing childhood stunting and achieving the
is characterized by high rate of poverty, low literacy and World Health Assembly target(6,19,50). Since mothers are
lack of access to safe water and sanitation(36). the main caregivers in households for young children,
High prevalence of stunting among young children can women’s empowerment in terms of education and deci-
be attributed to chronic nutrition deprivation and poor sion making should also be given attention when
linear growth in infancy and early childhood(37,38). The designing nutrition intervention programmes.
demand for nutrients increases with age and inappropriate At household level, children from poor and food-
complementary feeding may result in inadequate intakes insecure households were more likely to be stunted(33).
of vital nutrients, which is one of the major causes of This is similar to the findings of the most recent national
stunting in early childhood(39,40). However, evidence of Demographic and Health Survey(12). Besides, lack of
the positive role of exclusive breast-feeding on linear access to a sanitary latrine and unhygienic toilet condition
growth is surprisingly weak(41). The current study also were significantly associated with higher risk of stunting
found no significant association between exclusive breast- among children. This association is in line with the find-
feeding and risk of early childhood stunting. ings of some other previously conducted studies(51,52).
It is well documented in several studies(37,42) that chil- Therefore, to reduce stunting prevalence, policy makers
dren born with a birth weight of less than 2·5 kg possess a should focus on reducing poverty and household food
higher risk of being stunted in early childhood. The pre- insecurity through ensuring easy access to credit, pro-
sent study also found that children born with low birth moting small investments and social safety net transfers,
weight had 50 % higher risk of being stunted. The asso- subsidizing raw materials of production such as seeds and
ciation between low birth weight and stunting in child- fertilizer, subsidizing the cost of medicines, and encoura-
hood can be explained by the increased vulnerability of ging homestead gardening, among others.
children to infections and increased risk of complications However, merely implementing nutrition interventions
including sleep apnoea, jaundice, anaemia, chronic lung may not necessarily ensure the desired level of reduction
disorders, fatigue and loss of appetite(43,44). As reported in in childhood stunting prevalence in Bangladesh. It is also
an earlier study carried out in Bangladesh(37), the present important that a suitable environment is being ensured
study also found that stunting prevalence was higher where successful programme implementation can be
among boys than girls. The explanation for this gender achieved. A recent systematic review also highlighted that
variation in stunting is still unclear, although some certain factors such as political stability, community
researchers have argued that this disparity is more pre- mobilization, a multisectoral approach and programme
valent among poor households compared with those delivery using community-based platforms can contribute
better-off socio-economically(45). to achieve higher rate of reduction(53). It is also very cru-
Several maternal factors were also identified to be cial that the poorest of the poor(2) and geographically
significantly associated with stunting among children vulnerable people(22) are given the highest priority.

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 Factors of early childhood stunting 93
We used the modified Poisson regression model with design of the study. M.B.H. conducted the data analysis
the generalized estimating equations approach in the and participated in interpretation of the results. S.K.M.,
present study while estimating the unknown correlation of M.B.H., M.P., F.K. and F.A. contributed to write the first
outcome among clusters. Future research may also employ draft of the manuscript. F.M.Y. commented extensively on
mixed-effect models as an alternative to using generalized the manuscript draft. All authors read and approved the
estimating equations as mixed-effect models are capable final manuscript. Ethics of human subject participation:
of estimating the variance of multiple random or nested All procedures performed in studies involving human
random effects(54). participants were in accordance with the ethical standards
The present study has several strengths over other of the institutional and/or national research committee and
similar studies. First, it followed a rigorous sampling with the 1964 Helsinki Declaration and its later amend-
method covering a large sample size of children aged less ments or comparable ethical standards. Ethical approval
than 2 years. Second, along with the rural areas, it also was obtained from the Bangladesh Medical Research
covered urban slums of the country where living standards Council (reference number BMRC/NREC/2013-2016/802).
as well as health and sanitation parameters are worse than The purpose of the study was described and written
in rural areas. Thus, the inclusion of children from urban informed consent was obtained from the child’s parents
slums alongside those from rural areas greatly increases prior to the interview. The respondents were ensured of
the comprehensiveness of the study. However, the study the confidentiality of information provided.
was subjected to certain limitations too. Being cross-
sectional in nature, it was unable to draw out the temporal References
relationship between stunting and its determinants. Third,
since the base study was carried out in rural areas and 1. Humphrey JH & Prendergast AJ (2017) Population-level
urban slums where BRAC operates its health and nutrition linear growth faltering in low-income and middle-income
interventions, we were unable to include the children of countries. Lancet Glob Health 5, e1168–e1169.
2. de Onis M & Branca F (2016) Childhood stunting: a global
urban non-slum areas of the country. Lastly, we had to rely perspective. Matern Child Nutr 12, 12–26.
on mother’s perception of the size of her child at birth 3. Black RE, Victora CG, Walker SP et al. (2013) Maternal and
instead of the exact birth weight. This is due to the fact that child undernutrition and overweight in low-income and
birth weight is not being taken for many newborn babies, middle-income countries. Lancet 382, 427–451.
4. Stewart CP, Iannotti L, Dewey KG et al. (2013) Con-
particularly in cases of home deliveries in Bangladesh. textualising complementary feeding in a broader framework
for stunting prevention. Matern Child Nutr 9, 27–45.
5. Martorell R & Zongrone A (2012) Intergenerational influ-
Conclusions ences on child growth and undernutrition. Paediatr Perinat
Epidemiol 26, 302–314.
6. World Health Organization (2014) Global Nutrition Targets
The present study findings urge the importance of admin- 2025: Policy Brief Series. Geneva: WHO.
istering a multicomponent, multidisciplinary intervention 7. The World Bank (2006) Repositioning Nutrition as Central
to Development – A Strategy for Large-Scale Action.
to address the issue of stunting among children aged 0–23
Washington, DC: World Bank.
months in Bangladesh. Policy makers and public health 8. Grantham-McGregor S, Cheung YB, Cueto S et al. (2007)
practitioners should focus on specific initiatives such as Developmental potential in the first 5 years for children in
quality health-care services, promotion of breast-feeding, developing countries. Lancet 369, 60–70.
9. Fenske N, Burns J, Hothorn T et al. (2013) Understanding
sanitation facilities, household food security and especially child stunting in India: a comprehensive analysis of socio-
on maternal education and nutrition. economic, nutritional and environmental determinants
using additive quantile regression. PLoS One 8, e78692.
10. Goudet SM, Griffiths PL, Bogin BA et al. (2015) Nutritional
interventions for preventing stunting in children (0 to 5
Acknowledgements years) living in urban slums. Cochrane Database Syst Rev
issue 5, CD011695; available at https://www.cochrane
Acknowledgements: The authors acknowledge the role of library.com/cdsr/doi/10.1002/14651858.CD011695/epdf/full
the relevant staff of the Health Nutrition and Population 11. UNICEF, WHO & World Bank Group (2017) Joint Child
Malnutrition Estimates – Levels and Trends (2017 edition).
Programme, BRAC, for initial review of the tools and http://www.who.int/nutgrowthdb/estimates2016/en/ (acces-
smooth arrangement of the data collection initiatives. sed October 2018).
Financial support: This research was funded by the Stra- 12. National Institute of Population Research and Training,
tegic Partnership Arrangement (SPA) between BRAC, the Mitra and Associates & ICF International (2014) Bangladesh
Demographic and Health Survey 2014. Dhaka and Rock-
UK Department for International Development (DFID) ville, MD: NIPORT, Mitra and Associates, and ICF
and the Australian Department of Foreign Affairs and International.
Trade (DFAT). The funders had no role in the design, 13. United Nations (2014) World Urbanization Prospects: The
2014 Revision. Washington, DC: UN, Department of Eco-
analysis or writing of this manuscript. Conflict of interest:
nomic and Social Affairs, Population Division.
The authors have no conflict of interest to disclose. 14. Bangladesh Bureau of Statistics (2015) Census of Slum Areas
Authorship: S.K.M., M.R. and K.A. contributed to the and Floating Population 2014. Dhaka: BBS.

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press

 94 SK Mistry et al.
15. National Institute of Population Research and Training, 37. Aguayo VM, Nair R, Badgaiyan N et al. (2016) Determinants
International Centre for Diarrhoeal Disease Research Ban- of stunting and poor linear growth in children under 2 years
gladesh & MEASURE Evaluation (2015) Bangladesh Urban of age in India: an in-depth analysis of Maharashtra’s com-
Health Survey 2013. Dhaka and Chapel Hill, NC: NIPORT, prehensive nutrition survey. Matern Child Nutr 12, 121–140.
icddr,b and MEASURE Evaluation. 38. Jones AD, Ickes SB, Smith LE et al. (2014) World Health
16. National Institute of Population Research and Training, Organization infant and young child feeding indicators and
International Centre for Diarrhoeal Disease Research Ban- their associations with child anthropometry: a synthesis of
gladesh, MEASURE Evaluation et al. (2008) Bangladesh recent findings. Matern Child Nutr 10, 1–17.
Urban Health Survey 2006. Dhaka and Chapel Hill, NC: 39. Shrimpton R, Victora CG, de Onis M et al. (2001) Worldwide
NIPORT, icddr,b and MEASURE Evaluation. timing of growth faltering: implications for nutritional
17. Ahsan KZ, El Arifeen S, Al-Mamun MA et al. (2017) Effects of interventions. Pediatrics 107, e75–e75.
individual, household and community characteristics on 40. Dewey KG & Adu-Afarwuah S (2008) Systematic review of the
child nutritional status in the slums of urban Bangladesh. efficacy and effectiveness of complementary feeding inter-
Arch Public Health 75, 9. ventions in developing countries. Matern Child Nutr 4, 24–85.
18. Bangladesh Bureau of Statistics & UNICEF (2016) Child 41. Bhutta ZA, Das JK, Rizvi A et al. (2013) Evidence-based
Well-Being Survey in Urban Areas of Bangladesh – Key interventions for improvement of maternal and child nutri-
Results. Dhaka: BBS & UNICEF. tion: what can be done and at what cost? Lancet 382,
19. Ahmed T, Hossain M, Mahfuz M et al. (2016) Imperatives for 452–477.
reducing child stunting in Bangladesh. Matern Child Nutr 42. Mishra K, Kumar P, Basu S et al. (2014) Risk factors for
12, 242–245. severe acute malnutrition in children below 5 y of age in
20. Fuchs C, Sultana T, Ahmed T et al. (2014) Factors associated India: a case–control study. Indian J Pediatr 81, 762–765.
with acute malnutrition among children admitted to a diarrhoea 43. Haque SMR, Tisha S & Huq N (2015) Poor birth size a badge
treatment facility in Bangladesh. Int J Pediatr 2014, 267806. of low birth weight accompanying less antenatal care in
21. Fakir AMS & Khan MWR (2015) Determinants of malnutri- Bangladesh with substantial divisional variation: evidence
tion among urban slum children in Bangladesh. Health from BDHS-2011. Public Health Res 5, 184–191.
Econ Rev 5, 1–11. 44. Khanal V, Sauer K, Karkee R et al. (2014) Factors associated
22. Hossain MB & Khan MHR (2018) Role of parental education with small size at birth in Nepal: further analysis of Nepal
in reduction of prevalence of childhood undernutrition in Demographic and Health Survey 2011. BMC Pregnancy
Bangladesh. Public Health Nutr 21, 1845–1854. Childbirth 14, 32.
23. Leroy JL, Ruel M, Habicht J-P et al. (2014) Linear growth 45. Wamani H, Åstrøm AN, Peterson S et al. (2007) Boys are
deficit continues to accumulate beyond the first 1000 days more stunted than girls in sub-Saharan Africa: a meta-
in low- and middle-income countries: global evidence from analysis of 16 demographic and health surveys. BMC
51 national surveys. J Nutr 144, 1460–1466. Pediatr 7, 17.
24. DeRenzi B, Borriello G, Jackson J et al. (2011) Mobile phone 46. Young MF, Nguyen PH, Addo OY et al. (2015) The relative
tools for field-based health care workers in low-income influence of maternal nutritional status before and during
countries. Mt Sinai J Med 78, 406–418. pregnancy on birth outcomes in Vietnam. Eur J Obstet
25. Davis RE, Couper MP, Janz NK et al. (2010) Interviewer Gynecol Reprod Biol 194, 223–227.
effects in public health surveys. Health Educ Res 25, 14–26. 47. Varela‐Silva MI, Azcorra H, Dickinson F et al. (2009) Influ-
26. Rutstein SO (2015) Steps to Constructing the New DHS ence of maternal stature, pregnancy age, and infant birth
Wealth Index. Rockville, MD: ICF International. weight on growth during childhood in Yucatan, Mexico: a
27. Gibson RS (2005) Principles of Nutritional Assessment. New test of the intergenerational effects hypothesis. Am J Hum
York: Oxford University Press. Biol 21, 657–663.
28. UNICEF & World Health Organization (2015) Progress on 48. Headey D, Hoddinott J, Ali D et al. (2015) The other Asian
Sanitation and Drinking Water: 2015 Update and MDG enigma: explaining the rapid reduction of undernutrition in
Assessment. https://www.unicef.org/publications/index_ Bangladesh. World Dev 66, 749–761.
82419.html (accessed October 2018). 49. van den Bold M, Quisumbing AR & Gillespie S (2013)
29. World Health Organization (2006) WHO Child Growth Women’s Empowerment and Nutrition: An Evidence
Standards: Length/Height for Age, Weight-for-Age, Weight- Review (vol. 01294). Washington, DC: International Food
for-Length, Weight-for-Height and Body Mass Index-for- Policy Research Institute; available at https://pdfs.semantic
Age, Methods and Development. Geneva: WHO. scholar.org/e43b/1432ddc7f87327fc0d086893da802df7a61a.
30. Agresti A (2003) Categorical Data Analysis, vol. 482. New pdf
York: John Wiley & Sons. 50. Haddad L, Achadi E, Bendech MA et al. (2015) The Global
31. Zou GY & Donner A (2013) Extension of the modified Nutrition Report 2014: actions and accountability to accel-
Poisson regression model to prospective studies with cor- erate the world’s progress on nutrition. J Nutr 145, 663–671.
related binary data. Stat Methods Med Res 22, 661–670. 51. Fink G, Gunther I & Hill K (2011) The effect of water and
32. Zeger SL, Liang K-Y & Albert PS (1988) Models for long- sanitation on child health: evidence from the demographic
itudinal data: a generalized estimating equation approach. and health surveys 1986–2007. Int J Epidemiol 40,
Biometrics 44, 1049–1060. 1196–1204.
33. Awasthi S & Agarwal S (2003) Determinants of childhood 52. Rah JH, Cronin AA, Badgaiyan B et al. (2015) Household
mortality and morbidity in urban slums in India. Indian sanitation and personal hygiene practices are associated
Pediatr 40, 1145–1161. with child stunting in rural India: a cross-sectional analysis
34. Ghosh S & Shah D (2004) Nutritional problems in urban of surveys. BMJ Open 5, e005180.
slum children. Indian Pediatr 41, 682–696. 53. Hossain M, Choudhury N, Abdullah KAB et al. (2017)
35. Islam MM, Sanin KI, Mahfuz M et al. (2018) Risk factors of Evidence-based approaches to childhood stunting in low
stunting among children living in an urban slum of Ban- and middle income countries: a systematic review. Arch Dis
gladesh: findings of a prospective cohort study. BMC Public Child 102, 903–909.
Health 18, 197. 54. Hubbard AE, Ahern J, Fleischer NL et al. (2010) To GEE or
36. Center for Environmental and Geographic Information Ser- not to GEE: comparing population average and mixed
vices (2012) Master Plan of Haor Area, vol. I. Dhaka: Center models for estimating the associations between neighbor-
for Environmental and Geographic Information Services. hood risk factors and health. Epidemiology 21, 467–474.

https://doi.org/10.1017/S1368980018002926 Published online by Cambridge University Press