Dietary diversity is associated with overall quality and nutrient adequacy of the diet in low-income countries. We determined the association between dietary diversity and stunting among children aged 6–59 months in rural Bangladesh.
In total, 165 111 under-fives who participated in the National Surveillance Project in 2003–2005 were included in the analysis. Dietary diversity score (DDS) was constructed through the summation of the number of days each of the nine food groups was consumed in the previous week. The association between stunting and DDS was determined adjusting for confounders using logistic regression models. All analyses were performed separately for children aged 6–11, 12–23 and 24–59 months.
One-half of the children were stunted. In multivariate analyses, compared with low DDS, high dietary diversity was associated with a 15, 26 and 31% reduced odds of being stunted among children aged 6–11, 12–23 and 24–59 months, respectively, after adjusting for all potential confounders (odds ratio (OR)=0.85, 95% confidence interval (CI): 0.76–0.94; OR=0.74, 95% CI: 0.69–0.79; OR=0.69, 95% CI: 0.66–0.73). In all groups, children who were still breastfed were more likely to have limited diversity (OR=1.88, 95% CI: 1.32–2.67; OR=1.71, 95% CI: 1.52–1.92; OR=1.15, 95% CI: 1.11–1.19). Those having diarrhea in the past week and coming from families with low socioeconomic status were more likely to have decreased diversity (P<0.05).
Reduced dietary diversity is a strong predictor of stunting in rural Bangladesh. The inclusion of a variety of food groups into complementary foods may be essential to improve child nutritional status.
Stunting reflects linear growth retardation accumulated before and after birth (Victora, 1992). Worldwide, stunting affects nearly one-third of children under 5 years of age, with the prevalence being higher in low-resource countries in sub-Saharan Africa and South Asia (United Nations Children's Fund, 2008). In Bangladesh, 43% of under-fives are stunted and a higher proportion is affected in rural areas (Measure DHS, 2007). Growth failure during infancy and early childhood is often irreversible, leading to short stature during adolescence and adulthood (Martorell et al., 1994; Victora et al., 2008). Stunting is associated with an elevated risk of child mortality, increased susceptibility to infection and poor cognitive and psychomotor development (Pollitt et al., 1995; Victora et al., 2008). The long-term consequences of stunting include deficits in school achievement, reduced work capacity and adverse pregnancy outcomes (Martorell et al., 1981; Victora et al., 2008). The etiology of stunting is complex and yet to be fully elucidated. Nevertheless, it may be largely attributable to chronic intake of a low-quality diet lacking in both macro- and micronutrients, as well as to frequent infections (Waterlow, 1994). These are all strongly associated with poverty (Waterlow, 1994).
Dietary diversity is an important component of dietary quality: consumption of a higher number of food items and food groups is associated with improved nutritional adequacy of the diet (Hatloy et al., 1998; Torheim et al., 2004). Dietary diversity is usually assessed using simple tools such as the dietary diversity score (DDS), which counts the number of food groups consumed over a given reference period (Ruel, 2003). Several studies have shown that DDS is positively associated with overall dietary quality, micronutrient intake of young children and household food security (Steyn et al., 2006; Kennedy et al., 2007). A higher DDS has also been associated with better nutritional status of children in developing countries (Arimond and Ruel, 2004; Sawadogo et al., 2006). This association has not yet been explored in Bangladesh, where the prevalence of stunting is high and overall dietary quality is likely to be poor.
This analysis was undertaken to determine the association between dietary diversity and stunting among children 6–59 months of age in rural Bangladesh using the National Surveillance Project (NSP) data. All analyses were performed separately for children aged 6–11, 12–23 and 24–59 months, because the prevalence and intensity of breastfeeding tends to differ substantially according to age during this period. We also tried to identify child, maternal and household characteristics associated with low dietary diversity among under-fives in rural Bangladesh.
Materials and methods
The NSP in Bangladesh has been carried out since 1990 by Helen Keller International together with the Institute of Public Health Nutrition of the Government of Bangladesh. Details of the NSP are described elsewhere (Bloem et al., 2003; Shafique et al., 2007). Briefly, the surveillance system is founded on UNICEF's conceptual framework on the causes of malnutrition (de Pee and Bloem, 2001) and collects relevant information from households with children under-five in both urban and rural areas. The NSP is undertaken bimonthly throughout the year. It employs a stratified multistage cluster sampling method to identify eligible rural households from four subdistricts in each of the six divisions of Bangladesh, including Barisal, Chittagong (including Chittagong Hill Tracts), Dhaka, Khulna, Rajshahi and Sylhet. A household is defined as a group of individuals eating from the same kitchen. New households are selected for every round of the survey. The sampling scheme was designed with the goal of obtaining a nationally and divisionally representative sample. This analysis included 165 111 children, 6–59 months of age in the rural areas of Bangladesh, who participated in the NSP in 2003–2005.
Visits were paid to eligible respondents in each of the selected households, to conduct interviews and take anthropometric measurements. Before the assessment, written consent was sought from each respondent. For children under-five, consent was obtained from parents or guardians. Interview and assessments were conducted only after consent was obtained.
Data were collected by two-person field teams on a structured coded questionnaire. The mother of the child or other caretaker was requested to provide information on the child's age, gender, morbidity in the past week and breastfeeding status. Children's dietary intake was determined by asking the number of days each of the nine food groups was consumed in the previous week. These food groups included: rice, lentils, green leafy vegetables, yellow/orange fruits, eggs, fish, chicken, meat other than chicken and milk other than breast milk. Data on maternal characteristics such as age, education, reproductive history and morbidity were collected. Information on the household's composition, expenditures and other socioeconomic, environmental and health indicators was collected.
Anthropometric measurements including height/length and weight of the children and mothers were assessed following standard procedures (Gibson, 1990). Height/length was measured to the nearest 0.1 cm using the wooden height-and-length board locally manufactured. Weight was assessed using the digital scale (Tanita, Tokyo, Japan) to the nearest 0.1 kg. Immunization cards or home records of date of birth, if available, were used to determine the age of the children. When these documents were unavailable, mother's recall was taken or the local events calendar was used to elicit the age.
All field staff collecting the interview and anthropometric data were from local non-government organization partners. They received training from Helen Keller International before each round of data collection. The performance of field staff during data collection was supervised by monitoring teams from Helen Keller International. Quality control teams from Helen Keller International re-collected 5–10% of anthropometric and interview data the following day to ensure the reliability of the data.
The vast majority (97%) of the households selected for the surveillance participated in the NSP. Most of the non-responses were owing to migration or being absent at the time of interview. Refusals to participate in the surveillance system was very low (<1%).
When more than one under-fives were available in a given household, only the youngest child in the household was included in the analysis. All analyses were weighted according to the population size of each of the six divisions of Bangladesh and adjusted for the multistage cluster design of the NSP.
Children's dietary diversity was assessed by creating a simple DDS through the summation of the total number of days each of the nine food groups was consumed in the previous week. The sum of the number of days could range from 0 to 63. Tertiles of DDS for each age group were created to classify children into low, average and high diversity, using the following cutoffs: children aged 6–11 months (low 0–6; average 7–14; high 15–48), 12–23 months (low 0–14; average 15–20; high 21–51) and 24–59 months (low 0–15; average 16–21; high 22–58). Different cutoffs were applied for different age group, because the distribution of the total number of days each of the nine food groups was consumed varied across the three age groups.
Stunting and wasting were defined as height-for-age (HAZ) and weight-for-height z-scores less than −2, respectively, using the WHO growth standards (World Health Organization, 2006) in AnthroPlus 2009 software. Maternal body mass index was defined as a ratio of weight (kg)/height2 (m) and thinness was defined as body mass index <18.5 kg/m2.
Household ownership of cultivable land and monthly per-capita expenditure were used as the primary indicators of socioeconomic status (SES). Details on the estimation of household expenditure are described elsewhere (Thorne-Lyman et al., 2010). Briefly, the sum of household expenditures on non-food items such as medical care, education and housing in the previous month was calculated. The weekly expenditures on 13 non-grain food items was added and multiplied by 30.4/7 to calculate monthly household expenditure on non-grain food items (30.4: average number of days in a given month; 7: number of days in a week). Further, the monetary value of in-kind rice produced on a household's land, received as in-kind payment for labor or as a gift was added to the monthly cash expenditure on grain foods, including rice and wheat. Subsequently, the total monthly expenditures on all non-food items, grain and non-grain foods were summed and divided by household size to arrive at an estimate of monthly per capita total expenditure. The household size was measured as a simple count of household members. Quintiles of monthly per-capita expenditure were created.
Descriptive statistics were used to examine the distribution of the full range of variables. Appropriate cutoffs were applied to create categorical variables for birth order (1–3 vs 4); child diarrhea (no incidence vs at least once); maternal height (< vs 150 cm); and maternal education (no schooling, primary school, secondary school or higher). The land ownership was categorized into possession of no land, 1–249 and 250 decimals (one decimal is equivalent to ∼40 m2).
Child feeding practices were examined by estimating the proportion of children who consumed each food item at least once in the week preceding the interview. Predictors of low dietary diversity were determined by performing bivariate analyses for all the various risk factors using χ2 tests. Child, maternal and household characteristics that were significantly associated with low dietary diversity of children were included in the logistic regression models.
Multiple logistic regression analyses were used to examine the association between DDS and the risk of stunting. Potential confounders were first selected on the basis of their known association with child stunting in developing countries. Subsequently, bivariate analyses were conducted for the selected potential confounders using χ2 tests and only factors that were significantly associated with DDS were included in the multiple logistic regression models. Stunting was included as the dependent variable and tertiles of DDS as the independent variables, along with all the potential confounding factors, including child sex, birth order, diarrhea in the past week, breastfeeding status, maternal education, and household land ownership and monthly expenditures. All analyses were performed using STATA version 9.0 (Stata Corp., College Station, TX, USA).
The study protocol was approved by the ethical review committee of the Bangladesh Medical Research Council. Secondary data analysis was approved by the institutional review board of the Johns Hopkins University School of Medicine.
The mean (standard error) age of children included in the analysis was 29.9 (0.04) months and 52% were male (Table 1). One-half of the children were stunted and 14% wasted. The mean (standard error) age of the mothers of under-fives was 26.9 (0.02) years. More than half the mothers were literate and had been enrolled in school (Table 1). Maternal short stature was prevalent with 46% shorter than 150 cm, and more than one-third thin. Approximately 60% of the households owned no cultivable land (Table 1).
The vast majority of children aged 6–11 (98%) and 12–23 months (94%) and 43% of children aged 24–59 months were breastfed at the time of the interview. Two-thirds of children 6–11 months ate rice and one-third received fish, leafy vegetables and lentils in the past week (Figure 1). On the other hand, most children aged 12–23 and 24–59 months consumed rice, fish and vegetables at least once in the past week. Intake of animal foods including chicken and meat was rare, with less than a quarter having received them in the previous week. In general, more children in the older vs younger age groups consumed each food group (all P<0.01).
In all age groups, low dietary diversity was more frequent among children who were still breastfed (odds ratio (OR)=1.88, 95% confidence interval (CI): 1.32–2.67; OR=1.71, 95% CI: 1.52–1.92; OR=1.15, 95% CI: 1.11–1.19) and had episodes of diarrhea in the week preceding the interview (Table 2). On the other hand, high levels of maternal education and household monthly per capita expenditure, as well as ownership of cultivable land were all protective against poor dietary diversity (Table 2).
Dietary diversity tertiles were significantly associated with stunting in all age groups. In multivariate analyses, compared with low DDS, high dietary diversity was associated with a 15, 26 and 31% reduced odds of being stunted among children aged 6–11, 12–23, 24–59 month, respectively, after adjusting for all potential confounders (OR=0.85, 95% CI: 0.76–0.94; OR=0.74, 95% CI: 0.69–0.79; OR=0.69, 95% CI: 0.66–0.73) (Table 3).
In this cross-sectional analysis of data collected from the NSP in Bangladesh, we examined whether dietary diversity assessed using a simple scoring system of food groups consumed over a reference period was associated with child stunting in rural Bangladesh. Lack of diversity was a strong predictor of stunting across all age groups of children under-five, regardless of breastfeeding status, morbidity, gender, and maternal and household characteristics. In all age groups, children who were still breastfed, had diarrhea, came from families with low SES, and whose mothers had limited education were more likely to have low diversity.
Overall, our results of the inverse association between stunting and dietary diversity tend to confirm the findings of previous research using HAZ as the indicator of nutritional status (Arimond and Ruel, 2004; Sawadogo et al., 2006). Despite the inconsistencies in age ranges and types and methods of dietary indices applied, most studies showed a positive association between dietary diversity and HAZ using national or multinational samples. Using data from multiple countries in Africa, Asia and Latin America, Arimond and Ruel (2004) showed that improved dietary diversity was associated with a higher HAZ among children aged 6–23 months. Similarly, Sawadogo et al. (2006) found a positive relationship between infant and child feeding index and HAZ in all age groups of children 6–35 months in rural Burkina Faso. On the other hand, in a study conducted in rural China, infant and child feeding index was shown to be associated with weight-for-length and weight-for-age z-scores, but not with HAZ (Zhang et al., 2009). This lack of association may be explained by the low prevalence of stunting (3.2%) and sample consisting only of children aged 6–11 months. The findings of this analysis and previous research reinforce the notion that improved food variety may indeed reflect a greater likelihood of meeting daily energy and nutrient requirements, which would result in improved nutritional status among young children.
Children who were still breastfed at the time of the interview were more likely to consume complementary foods with limited diversity. As breast milk replaces part of the energy and nutrient requirements of young children, the relative demand for complementary foods is likely to be reduced, resulting in decreased dietary diversity. The positive association between dietary diversity and SES corroborates previous studies in developing countries (Hatloy et al., 1998; Torheim et al., 2004). Recently, Thorne-Lyman et al. (2010) also showed that the DDS is associated with total household expenditures. This indicates that the household's capability to acquire necessary foods and the general availability of food is a prerequisite to achieve the diversification of child diets (Torheim et al., 2004). Overall, complementary foods in rural Bangladesh seemed to be monotonous, being primarily cereal-based. Animal food consumption was extremely restricted.
A few limitations to this study must be considered. This was a cross-sectional analysis exploring the association between low dietary diversity and stunting prevalence. Thus, a causal association between the two factors cannot be established. The DDS was constructed based on a simple count of the number of food groups consumed. No information on portion sizes and amount of intake was collected. Also, the DDS used a reference period of the previous 7 days, reflecting diversity in the recent diet. Given that stunting reflects long-term cumulative nutritional status of individuals, it should be noted that this analysis was carried out based on the assumption that recent diversity is a good proxy of long-term dietary diversity. The household SES was assessed using indicators of monthly per capita expenditures and ownership of cultivable land. However, every indicator likely has its limitations, leaving the possibility of the adjustment for SES to be imperfect (Arimond and Ruel, 2004).
In conclusion, this analysis revealed that limited diversity in complementary foods is a strong predictor of stunting among rural Bangladeshi children under-five, after controlling for all potential confounders. This reinforces the growing evidence of the association between dietary diversity and child nutritional status. Optimizing the overall quality of complementary foods through the inclusion of a variety of food groups may be more essential to improve child nutritional status, rather than prolonged breastfeeding itself, particularly after the second year of life. Further research on the methodological issues related to the construction of better dietary diversity indicators is warranted.