Effects of Parental Education and Wealth on Early Childhood Stunting in Bangladesh

Correlates of child stunting were examined using data generated by a cross-sectional cluster survey conducted in Bangladesh in 2019. The data includes a total of 17490 children (aged < 5 years) from 64400 households. Multiple logistic regressions were used to determine the risk factors associated with child stunting and severe stunting.


Conclusions
After controlling for socioeconomic and demographic factors, parental education and household position in wealth index were found as to be the most important determinants of child stunting in Bangladesh.

Background
Malnutrition represents an insu cient intake of calories and nutrients that results in illness and in extreme cases, death [1]. Malnutrition among children poses serious health threat to survival and it is linked to 3.1 million child deaths globally [8]. Among developing countries, it is estimated ~13 million children who are below ve years of age die annually, of which malnutrition is among the leading cause [2,6].
Malnutrition is an impediment in the successful achievement of Sustainable Development Goals (SDGs) as it reduces the productivity of an individual and consequently negatively effecting the likelihood of future economic growth. As one of the top performing countries in Millennium Development Goals (MDGs), Bangladesh is particularly keen to embrace the new SDGs targets. For example, reducing the proportion of stunting among children under-ve years of age from 36.1-25% is integrated to the 7th Five Year Plan of Bangladesh Government (2016-2020) to achieve the national SDG targets.
Height-for-age is one of the prominent parameters to investigate malnutrition status of a child [3]. According to the World Health Organisation (WHO), a child is regarded as stunted when his/her height < -2 standard deviations from child growth standard median for the same age and the same sex [4]. Stunting represents linear growth delay by a child. Globally, 159 million children are classi ed as stunted, indicating the prevalence rate of 23.8% [5]. Other negative health consequences of stunting among children include, a reduce recovery capacity from disease [10] increased risk of future adult obesity [11], poor cognitive functioning such as loss of memory, lower achievement in learning and increased attention de cit [12]. Stunting is estimated to be responsible for 1.2 million child deaths globally under ve years old [8]. Malaria and pneumonia are more prevalent among children who are stunted and there are strong evidences for that [7,9]. It is associated with unfavourable pregnancy outcome [14]. Stunting has worked and continues as a hindrance to the human capital development because of its negative impact on economic productivity of adults [65].
Recently, public health researchers have examined the factors associated with the risk of stunting. Most studies have shown that socioeconomic and demographic factors are key correlates of child stunting [17]. Among a sample of Nigerian children, low maternal education is associated two-fold increase for the risk of child stunning. [18]. Socioeconomic status is another factor associated with child stunting. Children from a higher socioeconomic status have signi cant lower risk of stunting, compared to children from a low socioeconomic status [2]. Ricci [20] found that among Philippine children, a combination of factors, such as high socioeconomic status, being female, breastfeeding and more frequent prenatal care are important for reducing the risk of child stunting.
While these studies offer some insight into the factors associated with stunting across different countries, to date, little is known about how parental education and socio-economic status in uence stunting among children ve years old and below in Bangladesh. However, data suggest that Bangladesh is among the highest child stunting prevalent countries of the world, with stunting prevalence rate of 36.1%, which ranks it 107 out of 132 countries when the countries are ranked from lowest to highest for the prevalence of stunting [5]. Therefore, identifying the factors associated with child stunting in Bangladesh is a public health priority. In addition to stunting, severe stunting which is de ned as height < -3 standard deviations from child growth standard median for the same age and the same sex [4] is also in our interest since it is a measure of severely growth impairment for a child. Understanding risk factors for both stunting and sever stunting would help policy formulation in order to combat child chronic undernutrition problem. The aim of this study is to explore the effect of parental education and wealth on the risk of stunting and severe stunting among a large sample of children ve years and below in Bangladesh.

Data And Method
Data were drawn from Multiple Indicators Cluster Survey (MICS) 2018-19 [21]. This survey was carried out by the Bangladesh Bureau of Statistics with the support of the United Nations International Children's Emergency Fund (UNICEF). The survey was conducted from January 19 to June 1, 2019 with the purpose of collecting data on important indicators at the national level for the eight divisions of the country. The survey sample was selected using a two-stage strati ed cluster sampling approach, using the 2011 census frame to select clusters.
Standardised questionnaires were used in all data collection. More details on study design, sampling procedure, data collection, data processing and data entry can be found in "Bangladesh Multiple Indicator Cluster Survey 2019, Progotir Pathey: Final Report" [21]. The data include complete information on 64,400 households and the questionnaires were completed by parents of 23099 children, ve years of age and younger. Children with no information about height-for-age-z scores were excluded from this analysis. Households with more than one child under ve years of age, only the youngest one was selected for our study. Our nal sample is composed of 17490 children below ve years of age.

Statistical analysis
Nutritional status of a child is examined using three anthropometric measures: height-for-age z scores, weightfor-age z scores and weight-for-height z scores [50]. We used the Multiple Indicator Cluster Survey 2019 (MICS) provided for height-for-age z scores for generating dependent variable (stunting condition), as long-term growth of children can be re ected by height-for-age z scores and it also can capture the effects of chronic malnutrition.
Using WHO child growth standards as reference [51], the prevalence of childhood stunting in our sample was assessed using two classi cations: (i) 'stunted': de ned as height-for-age z-scores less than -2 standard deviations; and (ii) 'severely stunted': de ned as children with height-for-age z-scores less than -3 standard deviations.
We developed two separate models; (i) one to examine the associations of parental education and wealth with child stunting status [  [55] and division [2]. To determine if there was any association between child stunting and the size of household, the household size was categorized in three categories based on the number of household member in each household (up to three members, four to six members and seven or more). This was included because family size is an important determinant of child malnutrition as is caring for children and availability of food for consumption dependant on the number household members [48,49].
Children's ages were divided into six categories (0 to < 6, 6 to <12, 12 to < 24, 24 to <36, 36 to <48 and 48 to <60) with children less than six months (0 to <6) as reference category. Children's age was included in the model as it has been reported by many studies as a signi cant predictor of child stunting status [40,52].
We included gender as an independent variable to examine if there was any signi cant difference of likelihood on stunting between male and female children. Mymensingh. Several studies on child under-nutrition in Bangladesh used geographical location (division) as a predictor of child stunting [2,53]. Area (Urban/rural) variable was also included in the model. It is the most common variable included in almost all studies conducted in Bangladesh for identifying socioeconomic risk factors of malnourished children in Bangladesh [2,33,40,53].
We included education of the mother and education of the father in our regression model to examine their role on the risk of child stunting and severe stunting. Parental education, both for education level of father and that of mother, was categorized into four categories based on the number of years of schooling; none for no schooling, primary schooling (years 1 to <6), secondary incomplete (years 6 to <10) and secondary complete or higher (≥10).
Wealth index was used as the indicator of socioeconomic status of households. Wealth index was constructed by principal components analysis in Bangladesh Multiple Indicators Cluster Survey 2019. In calculating wealth index, source of water, type of housing, type of toilet facility, type of fuels for cooking, electricity, bank account, some durable goods and animals were taken into consideration [21]. Households were assigned wealth score depending on the asset they owned. We used wealth scores of our sampled households to rank them and the households were then divided into ve equal portions (quintiles) from poorest to wealthiest, such as poorest, second, middle, fourth and wealthiest. To assess the impact of latrine facilities on child stunting and severe stunting, we included type of toilet facility as predictor variable in our model. Iodine is essential micronutrient and su cient intake of iodine is necessary for normal growth [54]. Farebrother et.al. [54] also argued that su cient intake of iodine prevents stunted growth and if the required daily minimum intake of iodine is not met, growth will be hampered. We therefore included salt iodization test outcome in our regression model to accommodate the role of iodized salt in reducing the risk of stunting among children who were ve years and below age. In multiple indicators cluster survey 2019, salt iodization test was carried out in each household and the result was classi ed into four categories namely, 0 PPM (not iodized), more than 0 PPM and less than 15 PPM, 15 PPM or more and no salt in the household. 15 PPM means salt containing 15 parts per million (PPM) of iodate or iodide. In this survey, a cut-off point was set at 15 ppm indicating that salt containing 15 ppm or more of iodate or iodide will be considered as adequately iodized [21].
As our dependent variable in the rst model was binary in nature ('stunted' vs 'not stunted'), we used binary logistic regression model to determine the relationship between socio-demographic factors and child stunting.
Likewise, dependent variable in the second model was dichotomous ('severely stunted' vs not 'severely stunted').
Therefore, we also used a binary logistic regression model to determine the association between sociodemographic factors and child severe stunting. We found no evidence for multicollinearity using a test for variance in ation factor (VIF) with estimated mean VIF of 1.72, implies that level of multicollinearity is within tolerance limit [63].

Results
Descriptive characteristics of the sample Table 1 shows the socioeconomic and background characteristics of the sample and the bivariate distribution of stunting and severe stunting with sample characteristics. Data were available for 17490 children aged <5 years.
A total 25.96% children were estimated to be stunted, and 7.97% severely stunted. More than half of the children were male (52.53%). Children from Dhaka division comprised the highest proportion (18.93%) and the lowest percentage of study children were from Mymensingh division (6.11% The prevalence rate of stunting was highest for those children whose parents had no education and whose family used a hanging toilet. Half of the children classi ed as stunted were aged between 24-35 months. The greatest proportion of children stunted, were from Sylhet division (34.24%) and the lowest proportion from Khulna division (20.01%).
In bivariate analysis, the following variables were signi cantly associated with stunting prevalence: age (p<0.001), gender (p<0.001), area (p<0.001), division (p<0.001), education of mother (p<0.001), education of father (p<0.001), wealth index quintile (p<0.001), type of toilet facility (p<0.001) and salt iodization test outcome (p<0.001).   Risk factors for severe stunting Table 1 shows the proportion of children classi ed as severely stunted. The proportion of male children severely stunted (8.18%) was similar to that of female children (7.73%). A slightly higher percentage of rural children were severely stunted (8.15%) compared with the urban children (7.23%). The highest proportion of severely stunted children was found in the Sylhet division (10.93%) and the lowest in the Khulna division (3.85%). Approximately one-tenth of children, with their mothers had no education were severely stunted (10.93%), and ~6% were severely stunted if their mothers completed secondary education or higher. Similarly only 5.88 percent children were severely stunted whose fathers had completed secondary or higher education and more than one-tenth (10.72%) children were severely stunted whose Table 3 should be pasted here father didn't go to school. Only 5.85% children from wealthiest families were severely stunted and 11.03% children from poorest families were severely stunted. The highest rate of severe stunting prevalence was found for children age between 24-35 months (10.37%) and the lowest rate for age less than six months (6.21%).
In bivariate analysis, the following variables were signi cantly associated with the prevalence of severe stunting: age (p<0.001), division (p<0.001), education of mother (p<0.001), education of father (p<0.001), wealth index (p<0.001), type of toilet facility (p<0.001) and salt iodization test outcome (p<0.001). Table 4 shows the results on the binary logistic regression analysis on the risk factors that associated with the risk for severe stunting among the children ve years and below. Children aged twenty-four to less than thirty six months were almost two times more likely to be severely stunted

Discussion
This study examined the prevalence rate of stunting among a large sample of Bangladeshi children aged < 5 years. Our data showed that national prevalence rate for child stunting is at 25.96%. This prevalence is remarkably lower than the 41% observed in 2007 [22], suggesting that child stunting in Bangladesh has declined over time.
A key nding was that our adjusted analysis showed that children living in Sylhet division had the highest risk for being stunted. This was somewhat surprising because a large proportion of Bangladeshi's total income is received by the people of Sylhet division. In addition, food security status in this region has been suggested to be relatively better than the other divisions of Bangladesh [22]. Furthermore, the rate of poverty is lower in Sylhet division compared with other divisions in Bangladesh [23]. Poor educational opportunities and access, adverse maternal health [64] might explain why children from Sylhet division had high odds of child stunting. Even though Sylhet is suggested to have better economic condition than any other region of the country, it is important to note that there are still inequalities in this region. For example, the Sylhet region is geographically heterogeneous with tea gardens, hills and haors. In those areas transport facilities are poor which hinders students to attend schools and hence there are few opportunities for educational attainment by children. Our ndings suggest that future research is needed to determine the factors associated with the high prevalence rate of child stunting in the Sylhet region.
A further key nding of the present study was that wealth index was a signi cant protective factor against child stunting. This nding is consistent with other studies showing that household socioeconomic status was a signi cant predictor of child stunting [24][25][26][27][28][29]. A study conducted by Talukder, reported that Bangladeshi children living in the poorest household had 38% higher odds for malnourishment compared to the children from wealthiest family [30]. The lower odds for child stunting associated with higher wealth index might be explained by the fact that children from wealthier families have better access to health facilities, better environmental conditions such as potable water, sanitation and access to su cient food. It might also be that parents from wealthiest families have a higher education and more responsible for the health of their children when compared with parents from the poorest households. Other studies have shown that low quality housing and no access to potable water were signi cantly associated with child stunting [31,32]. According to UNICEF (1990), major determinants of nutritional condition for a child are access to su cient food supplies, improved health facilities and access to safe potable water supplies that are determined by household economic status [30]. The key ndings from the present study suggest that improvement in household wealth status can possibly have signi cant impact on the reduction of reduce the probability of child stunting.
Consistent with other studies, we showed that maternal education has signi cant impact in reducing the chance for a child to be stunted [33-35; 2, 25, 4]. The presence of an educated mother might lead to a better understanding of nutritional conditions for her children. Better child feeding practices, for instance, exclusive breast feeding during rst six months of a newborn child and timely initiation of complementary foods for children have been identi ed to be signi cantly associated with lower risk of child stunting [36]. More educated mothers may earn more money and therefore they may have more opportunity to invest in the health of their children [37]. Moreover, educated mothers are more likely to make better use of health care for their children [38,39], to make e cient use of family resources [40,41] and more willing to utilise family planning [18].
A further outcome from the present study was that a child's age had a signi cant impact on child stunting. The chances of a child to being stunted increased with age and reached at peak at 24-35 months. The nding is supported by other studies [41,44,45]. A potential cause of this might be that as the age of a child increases, biological factors and socioeconomic factors become greater determinants of stunting. It has been reported that male children had the higher chance of being stunted compared with female children [46,47]. However, our study failed to establish signi cant association by gender for increased odds of stunting or severe stunting.
Children from households where salt was found to be adequately iodized (≥15 PPM in salt iodization test) had signi cantly lower odds on stunting, compared with the children from households where salt was found as not iodized (0 PPM). The role of iodized salt as an important protective factor against stunting was supported by other studies which also suggest iodized salt reduces the risk of stunting [55,56]. Research has shown that iodine is crucial for normal physical growth [54] and insu cient intake of iodine by children during infancy and childhood could cause impaired growth [57,58].
There are some limitations in the study that should be acknowledged. First, the cross-sectional design limits the possibility of determining causal relationships. Moreover, many studies have shown a seasonal impact on the prevalence of stunting especially in the rural areas of developing countries [59,60]. Given our study design, we were not able to assess seasonal impact. A further limitation was that there was no information on childbirth weight, paternal height, maternal height and maternal age, all of which have been reported as signi cant predictor of child malnutrition (20,40-). Strengths of the study include the use of a large representative sample size with a standardized questionnaire. This allows for comparisons with future data collected as a part of the Bangladeshi Multiple Indicators Cluster Survey.

Conclusion
Among a large sample of Bangladeshi children aged < 5 years, ~25.96% were classi ed as stunted. Key risk factors for child stunting included increased age, living in Sylhet division and more under ve children in a household using a pit toilet or a hanging toilet. In contrast, protective factors against child stunting were higher level of parental education and living in higher wealth index. Importantly, the risk factors as well as protective factors were similar between stunting and severe stunting implying intervention targeting the reduction of child stunting could eventually reduce the prevalence of severe stunting among children below ve years of age. Our study suggested that parental education is a major protective factor that plays a crucial role in reducing the risk for a child to be stunted or severely stunted, therefore, we recommend promotion of education both for men and women. Higher level of maternal education would improve child nutritional conditions through its role on in uencing child feeding and childcare.
Promotion of paternal education would improve child nutritional condition primarily through a generation of higher income and living in better neighbourhoods that is living in areas where there are greater medical facilities and cleaner environment. Another crucial protective factor indicated by the study is that socioeconomic status measured by the wealth index quintile. As regional differences appeared as a signi cant predictor of child stunting and severe stunting, for instance children living in Sylhet division are in higher risk for being stunted and severely stunted, geographical targeting should be adopted by governments and stakeholders on reducing stunting prevalence.
To reduce stunting prevalence in Bangladesh, policies should be implemented that focus on the risk factors determined by the study. Interventions regarding parental education and reducing socio-economic inequality may have long lasting bene cial effects on child malnutrition.  Level 1= both parents no education; Level 2= one with no education, one with primary school; Level 3= one with no education, one with secondary school and above; Level 4= both with primary school; Level 5= one with primary school, one with secondary school and above; and Level 6= both with secondary school and above [40]  Level 1= both parents no education; Level 2= one with no education, one with primary school; Level 3= one with no education, one with secondary school and above; Level 4= both with primary school; Level 5= one with primary school, one with secondary school and above; and Level 6= both with secondary school and above [40]