Articles
Effect of parental formal education on risk of child stunting
Lancet 2008; 371: 322–28
See Comment page 280
Johns Hopkins University
School of Medicine, Baltimore,
MD, USA (Prof R D Semba MD,
K Sun MS); Nutrition Service,
Policy, Strategy and
Programme Support Division,
World Food Programme, Rome,
Italy (S de Pee PhD,
M W Bloem MD); Johns Hopkins
Bloomberg School of Public
Health, Baltimore, MD, USA
(M W Bloem); Friedman School
of Nutrition Science and Policy,
Tufts University, Boston, MA,
USA (M W Bloem); Helen Keller
International Asia Pacific,
Phnom Penh, Cambodia
(M Sari MSc); and Helen Keller
International Asia Pacific,
Dhaka, Bangladesh
(N Akhter MSc)
Correspondence to:
Prof Richard D Semba, Johns
Hopkins School of Medicine,
550 N Broadway, Suite 700,
Baltimore, MD 21205, USA
rdsemba@jhmi.edu
322
in Indonesia and Bangladesh: a cross-sectional study
Richard D Semba, Saskia de Pee, Kai Sun, Mayang Sari, Nasima Akhter, Martin W Bloem
Summary
Background Child stunting is associated with poor child development and increased mortality. Our aim was to
determine the effect of length of maternal and paternal education on stunting in children under the age of 5 years.
Methods Data for indicators of child growth and of parental education and socioeconomic status were gathered from
590 570 families in Indonesia and 395 122 families in Bangladesh as part of major nutritional surveillance
programmes.
Findings The prevalence of stunting in families in Indonesia was 33·2%, while that in Bangladesh was 50·7%. In
Indonesia, greater maternal formal education led to a decrease of between 4·4% and 5% in the odds of child stunting
(odds ratio per year 0·950, 95% CI 0·946–0·954 in rural settings; 0·956, 0·950–0·961 in urban settings); greater
paternal formal education led to a decrease of 3% in the odds of child stunting (0·970, 0·967–0·974). In Bangladesh,
greater maternal formal education led to a 4·6% decrease in the odds of child stunting (0·954, 0·951–0·957), while
greater paternal formal education led to a decrease of between 2·9% and 5·4% in the odds of child stunting (0·971,
0·969–0·974 in rural settings; 0·946, 0·941–0·951 in urban settings). In Indonesia, high levels of maternal and
paternal education were both associated with protective caregiving behaviours, including vitamin A capsule receipt,
complete childhood immunisations, better sanitation, and use of iodised salt (all p<0·0001).
Interpretation Both maternal and paternal education are strong determinants of child stunting in families in Indonesia
and Bangladesh.
Introduction stunting than is the level of maternal education in
Stunting represents linear growth failure due to poor Bangladesh14 and the Philippines.15
nutrition and infections both before and after birth.1 Further insight is needed into the relation between
Stunting in early childhood is associated with poor maternal and paternal education and other risk factors
cognitive, motor, and socioemotional development,1 and with child stunting, since formal education could be an
increased mortality.2,3 Stunted children do not reach their important key in breaking the intergenerational cycle of
full growth potential and become stunted adolescents stunting. We postulated that the number of years of
and adults.4 The functional consequences of stunting maternal education and paternal education are each
continue in adulthood, with reduced work capacity5 and, strong, independent risk factors for stunting in children.
in women, increased risk of mortality during childbirth6 We also postulated that higher maternal and paternal
and adverse birth outcomes.7,8 Worldwide, about a third education is associated with a greater number of
of preschool children are stunted.9 Although the global protective childcare behaviours. To address these
prevalence of stunting has decreased from nearly 47% hypotheses, we examined the effect of length of maternal
in 1980, most progress in the reduction of stunting has and paternal education and other risk factors on stunting
been made in southeast Asia, with little change in in children from families living in four different settings:
sub-Saharan Africa.10 urban slums and rural areas of both Indonesia and
Child stunting is a result of long-term chronic Bangladesh.
consumption of a low-quality diet in combination with
morbidity, infectious diseases, and environmental Methods
problems. More educated mothers and fathers might Participants and procedures
presumably provide better protection of their child as Households that participated in major nutritional
caretakers than would those with less education. Previous surveillance systems in Indonesia (2000–03) and
studies have shown conflicting results with regard to the Bangladesh (2000–05) were included in this analysis. The
relative importance of maternal and paternal education Nutritional Surveillance System (NSS) in Indonesia was
as a determinant of stunting. In developing countries, established by the Ministry of Health of the Government
maternal literacy,11 the mother having any formal of Indonesia and Helen Keller International (HKI)
education,12 and maternal completion of primary school13 in 1995.16 The Nutritional Surveillance Project (NSP) in
have been associated with a reduced risk of child stunting. Bangladesh was established by HKI and the Institute of
Other studies suggest that the level of paternal education Public Health Nutrition (IPHN) of the Government of
is a stronger and more consistent determinant of child Bangladesh in 1989.17 Both nutritional surveillance
www.thelancet.com Vol 371 January 26, 2008
 Articles

systems were based on UNICEF’s conceptual framework variables were collected in Indonesian rupiah or
on the causes of malnutrition18 with the underlying Bangladesh taka. For this analysis, expenditures are
principle of monitoring public-health problems and presented in US$ to control for the fluctuation of the
guiding policy decisions.19 The NSS and NSP were based Indonesian rupiah. In Indonesia, monthly exchange
on stratified multistage cluster sampling of households rates in 2000–03 were established using historical data
in subdistricts of administrative divisions of the country publicly available through the Bank of Canada.20 Mean
in rural areas and slum areas of large cities. In exchange rates by data collection round were calculated
Bangladesh, rural data were collected from four on the basis of the months in which data were collected
subdistricts of each six divisions, and urban samples for each round. Expenditure and price variables in US$
consisted of a sample of households selected through
probability proportionate to size sampling in rural areas Stunted (N=205 715) Not stunted (N=384 855) Missing values (%)
and in cities of the six major divisions of Bangladesh— Maternal age (years) 504 (0·08%)
Barisal, Chittagong, Dhaka, Khulni, Rajshahi, and ≤24 59 586 (33·3%) 119 304 (66·7%)
Sylhet—every 2 months.17 In Indonesia, data was collected 25–28 50 754 (34·2%) 97 533 (65·8%)
from about 40 000 randomly selected households every 29–32 43 069 (34·3%) 82 306 (65·7%)
3 months and involved five major urban poor populations 33+ 52 139 (38·0%) 85 215 (62·0%)
from slum areas in the cities of Jakarta, Surabaya, Maternal education (years) 2733 (0·5%)
Makassar, Semarang, and Padang, and rural populations 0 15 192 (46·1%) 17 724 (53·9%)
from the provinces of Lampung, Banten, West Java, 1–6 116 651 (38·0%) 189 634 (62·0%)
Central Java, East Java, the island of Lombok (West
7–9 40 297 (32·3%) 84 277 (67·7%)
Nusatenggara), and South Sulawesi.
≥10 32 293 (26·0%) 91 769 (74·0%)
Methods for data collection were similar in Indonesia
Maternal height (cm) 148·5 (0·02; 205 551) 150·6 (0·02; 84 572) 447 (0·08%)
and Bangladesh. New households were selected every
Paternal education 24 062 (4·1%)
round. Data were collected by two-person field teams. A
0 10 155 (46·3%) 11 790 (53·7%)
structured coded questionnaire was used to record data
1–6 98 608 (38·2%) 159 669 (61·8%)
about children aged 0–59 months, including
7–9 39 116 (33·3%) 78 343 (66·7%)
anthropometric measurements, date of birth, and sex.
≥10 47 054 (27·8%) 121 773 (72·2%)
The mother of the child or other adult member of the
Child sex 0 (0%)
household was asked to provide information on the
Male 110 621 (36·4%) 193 354 (63·6%)
household’s composition, parental education, birthweight
Female 95 094 (33·2%) 191 501 (66·8%)
of children, and weekly household expenditures, along
with other socioeconomic, environmental sanitation, and Birthweight (g) 3038 (3; 152 182) 3215 (2; 308 086) 130 302 (22·1%)

health indicators. In Indonesia, information on Child age (months) 3 (0·0005%)

immunisation history, receipt of vitamin A capsules, use 0–5 5458 (6·2%) 81 825 (93·8%)
of iodised salt, use of contraception, and use of posyandu 6–11 17 647 (16·6%) 88 679 (83·4%)
(local health post) were collected. A history of 12–23 66 764 (40·0%) 100 266 (60·0%)
immunisation with diphtheria-tetanus-pertussis vaccine, 24–35 58 964 (51·8%) 54 778 (48·2%)
oral poliovirus vaccine, and measles vaccine was obtained 36–47 36 838 (49·8%) 37 103 (50·2%)
from the child immunisation card or from the mother, if 48–59 19 827 (47·2%) 22 201 (52·8%)
the card could not be located. A child was deemed to have Number of individuals 46 326 (7·8%)
eating from same kitchen
received a vaccine if the response was yes in the absence
of an immunisation card, or yes as recorded on the 2–4 87 537 (34·8%) 164 126 (65·2%)

immunisation card. A child was judged to have missed a >4 100 754 (34·4%) 191 827 (65·6%)
vaccine if the response was no in the absence of an Weekly per-head household 0 (0%)
expenditure (quintiles)*
immunisation card or verified as not recorded on the
1 44 460 (37·6%) 73 654 (62·4%)
immunisation card. Birth dates of the children were
2 44 163 (37·4%) 73 951 (62·6%)
estimated with a calendar of local and national events
3 41 371 (35·0%) 76 743 (65·0%)
and converted to the Gregorian calendar.
4 39 232 (33·2%) 78 882 (66·8%)
In both the NSS and NSP, the participation rate of
5 36 489 (30·9%) 81 625 (69·1%)
families in the surveillance system was over 97% in both
urban slum and rural areas, and the main reason for Setting 0 (0%)

non-response was that the family had moved out of the Urban 47 306 (32·9%) 96 501 (67·1%)

area or was absent at the time the interviews were done. Rural 158 409 (35·5%) 288 354 (64·5%)
Non-response because of refusal to participate in the Data are n (%) or mean (SE; n). *Quintile 1 is poorest and quintile 5 is wealthiest; cutoffs were $1·05, $1·83, $2·55,
surveillance system was very low (<1%). and $3·68.
In each household, data were gathered regarding
Table 1: Risk factors for child stunting in families in Indonesia
expenditures in the previous week. Expenditure and price

www.thelancet.com Vol 371 January 26, 2008 323

 Articles

per round were created and calculated using the exchange

–0·8 rates by round. There was less fluctuation in the value of
–0·9 the taka in Bangladesh and expenditures are presented in
US$ based on the average exchange rate in 2000–05.
–1
The study protocol complied with the principles
–1·1 enunciated in the Helsinki Declaration.21 The field teams
were instructed to explain the purpose of the nutrition
Height-for-age Z score

–1·2
surveillance system and data collection to each child’s
–1·3
mother or caretaker, and, if present, the father or
–1·4 household head; data collection proceeded only after
written informed consent. Participation was voluntary,
–1·5
no remuneration was provided to participants, and all
–1·6 participants were free to withdraw at any stage of the
–1·7
interview. The nutritional surveillance system in
Indonesia was approved by the Ministry of Health,
–1·8 Government of Indonesia, and in Bangladesh by the
0 1 2 3 4 5 6 7 8 9 10 11 12+
ethical review committee of the Bangladesh Medical
Parental education (years)
Research Council. The plan for secondary data analysis
was approved by the institutional review board of the
Figure 1: Height-for-age Z score by years of maternal education (blue) and paternal education (pink) in
families in Indonesia Johns Hopkins University School of Medicine.
Figures are unadjusted for any covariates.
Statistical analyses
The youngest child in the household was used as the
Stunted Not stunted Missing values (%)
(N=204 420) (N=190 702) index of child stunting for that particular household (ie,
households were not counted more than once). WHO
Maternal age (years) 3202 (0·8%)
child growth standards were used as the reference growth
≤22 53 219 (50·0%) 53 318 (50·0%)
curves for child height-for-age.22 Stunting was defined as a
23–26 55 493 (50·5%) 54 305 (49·5%)
height-for-age Z score of less than –2.9,10 The US National
27–30 42 375 (51·6%) 39 648 (48·4%)
Center for Health Statistics (NCHS) growth references23
31+ 51 162 (54·7%) 42 400 (45·3%)
were included to compare the prevalence of stunting with
Maternal education (years) 17 113 (4·3%)
the new WHO child growth standard. Maternal age was
0 104 037 (57·9%) 75 719 (42·1%)
divided into quartiles. Maternal and paternal education
1–6 61 550 (50·9%) 59 396 (49·1%) was categorised as 0, 1–3 (first half of primary), 4–6 (second
7–9 23 112 (40·5%) 33 971 (59·5%) half of primary), 7–9 (junior high), and 10 years or more
≥10 6 051 (29·9%) 14 173 (70·1%) (high school or greater). The proportion of mothers and
Maternal height (cm) 149·1 (0·01; 202 043) 151·4 (0·01; 189 552) 3527 (0·9%) fathers who had achieved greater than 12 years (high
Paternal education 27 696 (7·0%) school graduate) was small (2·3% of mothers and 3·8% of
0 96 527 (57·7%) 70 903 (42·3%) fathers in Indonesia; 0·5% of mothers and 2·6% of
1–6 51 882 (51·8%) 48 328 (48·2%) fathers in Bangladesh) and was thus included in the
7–9 25 322 (44·2%) 31 899 (55·8%) category of 10 years or more. Weighting was used to adjust
≥10 15 451 (36·3%) 27 114 (63·7%) for urban as well as rural population size, by city and
Child sex 0 (0%) province, respectively, and all results are weighted except
Male 109 821 (53·1%) 96 844 (46·9%) for families from urban slums of Bangladesh. Weighting
Female 94 559 (50·3%) 93 585 (49·7%) was not used for families from urban slums because of
Child age (months) 34 712 (8·8%) the lack of reliable data on the population size in the
0–5 10 790 (29·9%) 25 336 (70·1%) urban slums of cities in Bangladesh. Birthweight data
6–11 14 646 (34·4%) 27 963 (65·6%) were not used in analyses involving Bangladesh because
12–23 50 742 (56·9%) 38 362 (43·1%) a large proportion of mothers did not know their child’s
24–35 47 314 (59·5%) 32 208 (40·5%) birthweight or it was not recorded. Weekly per-head
36–47 37 667 (57·5%) 27 094 (42·5%) household expenditure was used as the main indicator of
48–59 25 426 (52·7%) 22 862 (47·3%) socioeconomic status. More than four individuals eating
Number of individuals eating 3 (0·0007%) from the same kitchen was used as an indicator of more
from same kitchen crowded households. χ² tests were used to compare
2–4 77 050 (51·9%) 71 363 (48·1%) categorical variables between groups. Mantel-Haenszel
>4 127 367 (51·6%) 119 339 (48·4%) χ² tests were used to examine the relation between
(Continues on next page) parental education level and caregiving behaviours.
Multivariate logistic regression models were used to

324 www.thelancet.com Vol 371 January 26, 2008

 Articles

examine the relation between maternal and paternal (Continued from previous page)
education and other variables and the risk of stunting. Weekly per-head household 0 (0%)
Variables were included in the multivariate models if expenditure, (quintile)*
significant in univariate analyses; p values less than 0·05 1 47 282 (59·8%) 31 742 (40·2%)
were considered significant. Covariance matrices were 2 44 649 (56·5%) 34 375 (43·5%)
used to examine multi-colinearity between independent 3 41 250 (52·2%) 37 775 (47·8%)
variables in the models. Data analyses were done with 4 38 141 (48·3%) 40 883 (51·7%)
SAS Survey (version 9.13). 5 33 098 (41·9%) 45 927 (58·1%)
Setting 0 (0%)
Role of the funding source Urban 28 383 (54·3%) 23 927 (45·7%)
The funding source had no role in the design, conduct, Rural 176 037 (51·4%) 166 775 (48·6%)
or analysis of this study. KS had full access to all the data.
RDS had final responsibility for the decision to submit Data are n (%) or mean (SE; n). *Quintile 1 is poorest and quintile 5 is wealthiest; cutoffs were $0·40, $0·94, $1·67, and
$2·89.
the manuscript for publication.
Table 2: Risk factors for child stunting in families in Bangladesh
Results
In Indonesia, the prevalence of stunting in 590 570 families –1·1
was 33·2% when measured with WHO child growth –1·2
standards, and 30·7% with NCHS child growth standards. –1·3
The proportion of stunted children increased with
–1·4
increasing maternal age and with increasing child age
Height-for-age Z score

–1·5
(table 1). Stunting also seemed to be associated with
lower levels of maternal education, shorter maternal –1·6

height, lower paternal education, lower birthweight, and –1·7

lower weekly per-head household expenditure (table 1). –1·8
Higher proportions of boys were stunted than of girls, –1·9
and children were less likely to be stunted if they lived in –2
a crowded household (table 1). Higher levels of maternal –2·1
and paternal education were associated with higher
–2·2
height-for-age Z scores (figure 1; p<0·0001 for both
–2·3
maternal and paternal education).
0 1 2 3 4 5 6 7 8 9 10 11 12+
In Bangladesh, the prevalence of stunting in Parental education (years)
395 122 families was 50·7% with WHO child growth
standards, and 57·5% with NCHS standards. The Figure 2: Height-for-age Z score by years of maternal education (blue) and paternal education (pink) in
prevalence of child stunting increased with increasing families in Bangladesh
maternal age and increasing child age (table 2). Stunting Figures are unadjusted for any covariates.

seemed to be associated with lower maternal education,

OR (95% CI)* p*
shorter maternal height, lower paternal education, and
Maternal age (years) 0·992 (0·990–0·993) <0·0001
lower monthly per-head household expenditure (table 2).
A greater proportion of male children were stunted than Maternal education (years) in rural 0·950 (0·946–0·954) <0·0001
setting†
were girls, and there was a smaller proportion of stunted
Maternal education (years) in urban 0·956 (0·950–0·961) <0·0001
children in more crowded households (table 2). Higher setting†
levels of maternal and paternal education were associated Maternal height (cm) 0·917 (0·915–0·919) <0·0001
with higher height-for-age Z scores (figure 2; p<0·0001 Paternal education (years) 0·970 (0·967–0·974) <0·0001
for both maternal and paternal education). Child age (months) 1·048 (1·047–1·049) <0·0001
Multivariate logistic regression models were used to Child sex (boy vs girl) 1·246 (1·223–1·269) <0·0001
examine effect of maternal and paternal education on Birthweight (per 100 g) 0·935 (0·933–0·937) <0·0001
stunting in families in both Indonesia and Bangladesh.
Setting (rural vs urban) 1·136 (1·075–1·202) <0·0001
In both countries, after adjustment for maternal age and
Weekly per-head expenditure (per $10) 0·819 (0·792–0·847) <0·0001
height, child sex, birthweight, child age, weekly per-head
household expenditure, and setting, increasing lengths *Adjusted for all other variables in the table, which were selected by backward
of formal maternal and paternal education were stepwise regression. †Maternal education shown for both rural and urban areas
because significant modification found between maternal education and setting
significantly associated with reduced odds of stunting in Indonesia.
(table 3 and table 4). No multi-colinearity was found for
independent variables in the model. Table 3: Multivariate logistic regression model for maternal education,
paternal education, and other risk factors in relation to stunting in
Analyses restricted to mothers and fathers who had children in Indonesia
some level of primary school education showed that,

www.thelancet.com Vol 371 January 26, 2008 325

 Articles

The effect of maternal education on child stunting was

OR (95% CI)* p*
much the same in Bangladesh as it was in Indonesia; by
Maternal age (years) 0·990 (0·989–0·992) <0·0001 contrast, the effect of paternal education on stunting was
Maternal education (years) 0·954 (0·951–0·957) <0·0001 significantly stronger in Bangladesh than it was in
Maternal height (cm) 0·915 (0·914–0·917) <0·0001 Indonesia (data not shown).
Paternal education (years) in rural 0·971 (0·969–0·974) <0·0001 Higher levels of both maternal and paternal education
setting†
were significantly associated with higher uptake of each
Paternal education (years) in urban 0·946 (0·941–0·951) <0·0001
setting†
of the eight caregiving behaviours that we assessed
Child age (months) 1·019 (1·018–1·019) <0·0001
(table 7). However, the eight caregiving behaviours, and a
composite score of the eight behaviours, did not seem to
Child sex (boy vs girl) 1·113 (1·095–1·131) <0·0001
mediate a direct effect on child stunting when assessed
Setting (rural vs urban) 0·846 (0·820–0·872) <0·0001
by linear regression analyses (data not shown).
Weekly per-head expenditure (per $10) 0·998 (0·997–0·998) <0·0001

*Adjusted for all other variables in the table, which were selected by backward Discussion
stepwise regression. †Paternal education shown for both rural and urban areas Our data suggest that both maternal and paternal
because significant modification found between maternal education and setting
in Bangladesh. education are strong predictors of child stunting: greater
levels of formal education achieved by both mothers and
Table 4: Multivariate logistic regression model for maternal education fathers were associated with a decreased odds of child
and other risk factors in relation to stunting in children in Bangladesh
stunting. Previous reports have examined the effect of no
education versus some formal education, or the
OR (95% CI) in OR (95% CI) in completion of primary or secondary school, with child
Indonesia Bangladesh height or stunting; there is also a paucity of data for the
Completion of primary school effect of paternal education on child development. The
Mother 0·74 (0·71–0·77) 0·78 (0·76–0·79) large sample sizes studied here allowed sufficient power
Father 0·84 (0·80–0·88) 0·81 (0·79–0·82)
to discern the effect of levels of formal parental education
Some primary school education
on the risk of child stunting. Results were consistent
Mother 0·87 (0·84–0·91) 0·85 (0·83–0·88)
between two different populous developing countries,
Father 0·91 (0·86–0·96) 0·86 (0·84–0·88)
although there were some differences in the size of the
effect in urban areas compared with rural settings, and
Table 5: Effect of length of primary school education on the odds of child in Indonesia compared with Bangladesh.
stunting
The prevalence of stunting was higher in Bangladesh
than in Indonesia. The median level of formal education
Odds ratio (95% CI) in Odds ratio (95% CI) in achieved by both mothers and fathers was lower in
Indonesia Bangladesh Bangladesh than in Indonesia, and the mean maternal
Level 2 0·85 (0·79–0·92) 0·83 (0·81–0·85) age was lower in Bangladesh than in Indonesia (data not
Level 3 0·77 (0·68–0·86) 0·74 (0·71–0·76)
shown). These findings reflect that girls in Bangladesh
Level 4 0·65 (0·61–0·70) 0·76 (0·74–0·78)
are less likely to complete primary and secondary school
Level 5 0·56 (0·52–0·61) 0·63 (0·62–0·65)
and are more likely to marry at an earlier age than are
Level 6 0·41 (0·38–0·44) 0·46 (0·44–0·47)
girls in Indonesia. In Indonesia, increasing maternal
education was associated with greater reductions in the
Level of education: 1=both parents no education; 2=one with no education, one odds of stunting than was paternal education. Mothers
with primary school; 3=one with no education, one with secondary school and
are generally the primary caregivers for the children and
above; 4=both with primary school; 5=one with primary school, one with
secondary school and above; and 6=both with secondary school and above). Odds their behaviour, and level of education might be expected
ratio calculated relative to level 1 education. to have a stronger effect on child stunting than that of
fathers. However, in Bangladesh, increasing paternal
Table 6: Effect of combined level of parental education on risk of child
stunting education was associated with greater decreases in the
odds of stunting than was maternal education. Thus, the
relative importance of maternal and paternal education
compared with parents who had no formal education, might vary in different settings. A study that involved
completion of primary school was associated with over 5000 preschool children in Bangladesh found a
reduced odds of child stunting, as was having had some strong and significant relation between paternal
primary school education (table 5). In both Indonesia education level, but not maternal education level, and
and Bangladesh, increasing levels of combined parental child stunting,14 and a study of 18 544 children from Metro
education were associated with reduced odds of stunting Cebu, Philippines, showed that paternal education was a
after adjustment for maternal age, maternal height, child more consistent determinant of child stunting than was
sex, child age, setting, and weekly per-head household maternal education in families in urban and rural areas.15
expenditure (table 6). The level of paternal education is usually related to

326 www.thelancet.com Vol 371 January 26, 2008

 Articles

Mothers Fathers
N Years of education p* N Years of education p*
0 1–6 7–9 ≥10 0 1–6 7–9 ≥10
Child received vitamin A capsule in past 6 months 584 659 41·1% 54·1% 58·9% 61·2% <0·0001 563 478 41·3% 53·9% 57·2% 60·3% <0·0001
Child received measles vaccine 481 384 55·8% 75·3% 84·5% 88·5% <0·0001 463 501 57·2% 75·5% 82·2% 86·5% <0·0001
Child received all three doses of polio vaccine 467 061 51·3% 73·4% 84·6% 89·8% <0·0001 458 894 51·9% 73·6% 82·0% 87·6% <0·0001
Child received all three doses of DTP vaccine 475 784 47·7% 69·8% 81·1% 86·7% <0·0001 458 666 48·1% 70·1% 78·3% 84·3% <0·0001
Use of iodised salt (by interview) 587 206 51·4% 79·1% 89·8% 94·5% <0·0001 565 909 51·7% 79·0% 87·3% 92·1% <0·0001
Place to defecate, use of closed latrine 587 408 28·2% 44·2% 64·1% 81·3% <0·0001 566 090 23·7% 41·7% 60·9% 77·5% <0·0001
Use of family planning methods 583 582 55·9% 72·8% 77·3% 75·6% <0·0001 562 330 57·4% 73·9% 77·2% 76·1% <0·0001
Has taken child to posyandu in the past year 584 412 70·6% 84·5% 88·3% 88·5% <0·0001 563 120 71·2% 84·6% 87·4% 87·6% <0·0001

Data are proportion of children who received caregiving behaviour. *Mantel-Haenszel χ² test.

Table 7: Maternal and paternal education and caregiving behaviours in families in Indonesia

household income, since more educated fathers usually
earn more money and marry women of a comparable
level of education. Our multivariate analyses adjusted for
weekly per-head household expenditure, which is closely
related to paternal education. Paternal education might
be a stronger predictor of child stunting than is maternal
education in Bangladesh because of the low social status
of women in Bangladesh and more limited influence on
household decision making.24,25
Our data from Indonesia suggest that higher levels of
maternal and paternal education were strongly associated
with protective childcare behaviours such as use of a
closed latrine, receipt of childhood immunisations,
receipt of vitamin A capsules, use of iodised salt, and use
of the local health post. Although higher parental
education and better childcare behaviours can lead to
improved child health, those behaviours assessed here
might not be the factors with the greatest effect on child
stunting. The causes of stunting are complex and reflect
poor quality diet, breastfeeding practices, long-term
burden of infectious disease morbidity, and chronic
adverse environmental exposures, and such factors are
difficult to measure in any cross-sectional survey. The
caregiving behaviours assessed however reflect a parent’s
knowledge of, and ability to implement, those behaviours
that are good for child health and development, and
hence contribute to growth. However, one must interpret
these data with caution, since collection of information
on these behaviours relied almost exclusively on maternal
report.
Our data are consistent with studies that have shown
that shorter maternal height is associated with stunting
of their children.14,25 Furthermore, our data corroborate
earlier evidence that birthweight is a determinant of
stunting.11,12,26,27 In Indonesia, but not in Bangladesh,
stunting was more prevalent in families in rural areas
than in those in urban slum areas (data not shown),
consistent with a study from the Philippines that showed
a higher prevalence of stunting in rural areas than in the
urban setting.27 Boys were at higher risk of stunting than
girls in both Indonesia and Bangladesh.
Our data suggest that the odds of child stunting
decrease with increasing levels of formal education of the
mother or father. This effect was consistent across rural
and urban settings and between two large countries that
have considerable economical, cultural, and social
differences. One should note, however, that the
cross-sectional design of this observational study limits
the inferences about causality between parental education
and child stunting.
Although we cannot deduce exactly the reason why
greater formal education would make such a difference
to child stunting, these findings emphasise the
importance of completing as much formal schooling as
possible. Although not assessed here, non-formal
education is available in Bangladesh and previous
analyses suggest that, in relation to child nutritional
status, non-formal education is equivalent to about
3 years of formal education.17
Long-term approaches to the elimination of stunting
will require consideration of the inter-generational nature
of stunting. Children who are stunted have a reduced
learning ability in school,28 and poor scholastic
achievement could increase the risk that they do not
complete primary or secondary education, thus
perpetuating the cycle. Even in families where the
mothers or fathers had completed 10 years of education
or more, the prevalence of stunting was still high. Thus,
achievement of high levels of formal education is not
enough to achieve the goal of eliminating stunting in one
generation. The causes of stunting are complex and
involve constitutional determinants (maternal height,
birthweight), proximal factors (dietary intake, infectious
disease morbidity), and underlying factors (maternal
education, socioeconomic status, poverty). Although
better education of women and men would be expected
to reduce child stunting, there are other social, economic,
and political factors on the societal and national level that
also have an effect.29
The promotion of higher levels of formal education for
both women and men is part of the second Millennium
Development Goal. Promotion of higher levels of

www.thelancet.com Vol 371 January 26, 2008 327

 Articles
education should help promote gender equality, empower
women, and—through better informed caregiving
practices—reduce child stunting and, over the long term,
reduce the risk of child mortality.
Contributors
All authors participated in the conception of the study, data analysis and
interpretation, and writing of the manuscript. All authors have seen and
approved the final version.
Conflict of interest statement
We declare that we have no conflict of interest.
Acknowledgments
This research was supported by a Lew R Wasserman Merit Award from
Research to Prevent Blindness to RDS.
References
1 Grantham-McGregor S, Cheung YB, Cueto S, et al. Developmental
potential in the first 5 years for children in developing countries.
Lancet 2007; 369: 60–70.
2 Pelletier DG, Frongillo EA, Habicht J-P. Epidemiologic evidence for
a potentiating effect of malnutrition on child mortality.
Am J Public Health 1993; 83: 1130–33.
3 Pelletier DL, Frongillo EA. Changes in child survival are strongly
associated with changes in malnutrition in developing countries.
J Nutr 2003; 133: 107–19.
4 Martorell R, Kettle Kahn L, Schroeder DG. Reversibility of stunting:
epidemiological findings in children from developing countries.
Eur J Clin Nutr 1994; 48 (suppl): S45–57.
5 Spurr GB, Barac-Nieto M, Maksud MG. Productivity and maximal
oxygen consumption in sugar cane cutters. Am J Clin Nutr 1977;
30: 316–21.
6 Royston E, Armstrong S. Preventing maternal deaths. Geneva:
World Health Organization, 1989.
7 Habicht JP, Yarbrough C, Lechtig A, Klein RE. Relationships of
birth weight, maternal nutrition and infant mortality. Nutr Rep Int
1973; 7: 533–46.
8 Martorell R, Delgado H, Valverde V, Klein RE. Maternal stature,
fertility, and infant mortality. Hum Biol 1981; 53: 303–12.
9 de Onis M, Blössner M, Borghi E, Morris R, Frongillo EA.
Methodology for estimating regional and global trends of child
malnutrition. Int J Epidemiol 2004; 33: 126–27.
10 de Onis M. Is malnutrition declining? An analysis of changes in
levels of child malnutrition since 1980. Bull World Health Organ
2000; 78: 1222–33.
11 Chopra M. Risk factors for undernutrition of young children in a
rural area of South Africa. Public Health Nutr 2003; 6: 645–52.
12 Sakisaka K, Wakai S, Kuroiwa C, et al. Nutritional status and
associated factors in children aged 0–23 months in Granada,
Nicaragua. Public Health 2006; 120: 400–11.
13 Wamani H, Åstrøm AN, Peterson S, Tumwind JK, Tylleskär T.
Predictors of poor anthropometric status among children under 2
years of age in Uganda. Public Health Nutr 2006; 9: 320–26.
328
14 Rahman A, Chowdhury S. Determinants of chronic malnutrition
among preschool children in Bangladesh. J Biosoc Sci 2006;
39: 161–73.
15 Ricci JA, Becker S. Risk factors for wasting and stunting among
children in Metro Cebu, Philippines. Am J Clin Nutr 1996;
63: 966–75.
16 de Pee S, Bloem MW, Sari M, Kiess L, Yip R, Kosen S. High
prevalence of low hemoglobin concentration among Indonesian
infants aged 3-5 months is related to maternal anemia. J Nutr 2002;
132: 2115–221.
17 Bloem MW, Moench-Pfanner R, Panagides D, eds. Health and
nutritional surveillance for development. Singapore: Helen Keller
Worldwide, 2003.
18 de Pee S, Bloem MW. Assessing and communicating impact of
nutrition and health programs. In: Semba RD, Bloem MW, eds.
Nutrition and health in developing countries. Totowa, NJ: Humana
Press; 2001: 483–506.
19 Mason JB, Habicht JP, Tabatabai H, Valverde V. Nutritional
surveillance. Geneva: World Health Organization, 1984.
20 Bank of Canada. Exchange rates: monthly average rates: 10-year
look-up. http://www.bankofcanada.ca/en/rates/exchange-avg.html
(accessed Dec 18, 2005).
21 World Medical Association. World Medical Association Declaration
of Helsinki: ethical principles for medical research involving
human subjects. Bull World Health Organ 2001; 79: 373–74.
22 WHO. WHO child growth standards: methods and development.
Geneva: World Health Organization, 2006.
23 Dibley MJ, Goldsby JB, Staehling NW, Trowbridge FL. Development
of normalized curves for the international growth reference:
historical and technical considerations. Am J Clin Nutr 1987;
46: 736–48.
24 Khan SR. The socio-legal status of Bengali women in Bangladesh:
implications for development. Dhaka: University Press, 2001.
25 Bushamuka VN, de Pee S, Talukder A, et al. Impact of a homestead
gardening program on household food security and empowerment
of women in Bangladesh. Food Nutr Bull 2005; 26: 17–25.
26 Adair LS, Guilkey DK. Age-specific determinants of stunting in
Filipino children. J Nutr 1997; 127: 314–20.
27 Keller W. The epidemiology of stunting. In Waterlow JC, ed. Linear
growth retardation in less developed countries. New York: Nestle
Nutrition/Raven Press, 1988.
28 Pollitt E. Malnutrition and infection in the classroom. Belgium:
United Nations Educational, Scientific, and Cultural Organization
(UNESCO), 1992.
29 Milman A, Frongillo EA, de Onis M, Hwang JY. Differential
improvement among countries in child stunting is associated with
long-term development and specific interventions. J Nutr 2005;
135: 1415–22.
www.thelancet.com Vol 371 January 26, 2008