Maternal undernutrition as proxy indicators of their offspring’s undernutrition: evidence from 2011 Ethiopia demographic and health survey

Anthropometry is widely used tool to assess the nutritional status of children and to monitor their growth [1]. Further, the prevalence of undernutrition in children particularly those aged 6–59 month is used as an indicator for nutritional status of the entire population since they are more sensitive to nutrition related stress [2].

A number of anthropometric indicators have been used assessing nutritional status of children and they include; weight-for-height, weight-for-age and height for age, among others. Multicomponent indicators such as weight for age and height for age require age to be determined accurately since they are more sensitive to random errors in age than in anthropometry [3]. Therefore, in a setting where vital statistics are not recorded or exact ages are not known it will be difficult to assess accurately the nutrition status of children.

Likewise, younger children are more difficult to deal with during weight and height measurements and result in error [4]. In other words, at least two persons are required to measure weight and height accurately: one to take measurements and other to record [5]. In situations where there are no sufficient health workers it may be also difficult to undertake these measurements accurately.

Considering the intergenerational continuity of undernutrition, maternal BMI and height which are independent of age and relatively easier to measure could be a simple way of measuring nutritional status of their children and as a result they can be potential proxy indicators for nutritional status of the entire population.

Therefore, the current study is aimed to determine whether maternal BMI and height can be used to assess child undernutrition and to explore the predictors of child undernutrition by analyzing a heterogeneous study population from a nationally representative sample.

Using a large and representative nationwide sample of children and mothers from Ethiopia, we aimed to examine if we can use maternal BMI and height as a proxy indicator of their children’s undernutrition and explore the predictors of child undernutrition. We observed a significant but weak correlation between mothers and their children’s nutritional status. We also found that child’s age and sex, maternal BMI, place of residence and household wealth index were significantly associated with the three forms of child’s undernutrition. Maternal height and maternal education were significantly associated with child’s stunting and underweight.

One of the main findings is the variability of the strength of correlation when the data was disaggregated according to the age of the child. The correlation between maternal and child nutrition status was lower for under 12 months and 24–35 months children than the older children for most of the indicators. This may be part as a result of increased measurement error among these age groups. However, since WAZ has no height measurement and tends to be less affected by field measurements than HAZ and WHZ, this suggest that the data have been affected by age variation among the under five children. Additionally, this study has shown that HAZ is more strongly correlated with maternal height than WAZ and WHZ. This result is expected since measures of cumulative exposure to socioeconomic and environmental risk factors should be more highly correlated than measures that respond to more immediate environmental change. The attained height of the adult is a consequence of genetic and environmental effect on the growing period of childhood [11]. For instance, the stunted mother may provide an inadequate supply of nutrients for the fetus, which restricts fetal growth and promotes low birth weight and stunting in the children [12–14].

The low sensitivity of the maternal undernutrition as screening test suggests that factors associated with undernutrition tend not to be fully shared by mothers and their off springs. On the other hand, the high specificity suggests that factors which mitigate nutritional risk tend to be fully shared by mothers and their children. In other words, when mothers are well nourished their children tend also to be well nourished. Likewise, the low PPV and the high NPV suggest that knowledge about mother undernutrition only slightly improves the likelihood of correctly classifying the child as being malnourished. Our results are analogous with studies done in Guinea [15] and Ethiopia [16].

The following factors may explain the lack of stronger correlation between mothers and their children: First, the causes of undernutrition are multidimensional and they include both food and non-food related factors that often interact to form a complex web. Social, political, economic and environmental factors by large cause climate change resulting in variation in food supply [17]. Second, the relationship between maternal and child could be also mediated by the dual burden of malnutrition. Even in food secure households some members may be undernourished while others may be overweight, suggesting that with the process of nutritional transition with economic development, it’s possible to see double burden households [18, 19]. This is even evident in our data; 22.3% of stunted, 11.7% of underweight and 7.2% of wasted children are from households with overweight mothers.

Third, using a single BMI cutoff still remain controversial [20]. A healthy BMI is likely to vary with age, sex, pregnancy and lactation, ethnicity and other factors [21]. Therefore, using a BMI < 18.5 kg/m² as the cutoff for everyone and everywhere is likely to be too crude. There are also evidences that showed BMI is not sensitive enough to detect small but clinically significant weight [22].

Likewise, child undernutrition is mainly as a result of poor feeding practices and infection and unlike children, the predominant cause of reduced body weight in adults is reduced food intake to meet the energy requirements [23]. In nutshell, the above discussed factors may explain lack of stronger correlation between maternal undernutrition (BMI and height) and child undernutrition indicators.

The estimation of three forms of undernutrition (stunting, underweight and wasting) was comparable to that of EDHS report but we could not exactly replicate the figures except for wasting. We believe the main source of our discrepancy from the report is that we used the child recode file (KR file), whereas the DHS report uses the household recode file (PR file) for stunting, underweight, and wasting. Height and weight of all children under five and all women 15–49 are measured in the household interview. If the mother is alive and in the same household as the child, then the child will also be in the KR file. Otherwise, the child will not be in the KR file. In other words, the children who are in the PR file but not in the KR file are more likely to be stunted.

Age and gender are factors that influence children susceptibility to malnutrition. For that reason, our results show that the male children are underprivileged across all three forms of undernutrition. There studies that documented higher prevalence of undernutrition among male children compared to female children [11, 24–26]. These studies corroborate our findings and there could be a biological explanation. For instance, Wamani and his colleagues found from epidemiological and cohort studies that neonatal and preterm babies’ mortality and morbidity to be consistently higher in males and females in early life though the underlying mechanism is poorly understood. However, the reported male predominance suggests that boys are generally vulnerable [26]. The child’s age is also an important predictor for child undernutrition. We found undernutrition to be more prevalent in older than in younger children. Children among 24–35 months have the highest levels of undernutrition, and the youngest age group (0–11 months) has the lowest levels except for wasting and were significantly associated. The higher proportion of undernutrition among older children could be due to inappropriate child feeding practices [24].

Results on maternal education and place of residence and child undernutrition are similar with those of other studies [11, 24, 27–29]. Education and place residence are socio-economic indicators and are expected to reflect both current and ongoing access to resources. In our result, undernutrition is common among children from rural areas and whose mothers have lower levels of education and were significantly associated. Additionally, our data has shown that child undernutrition is more common among children in the poorest wealth quintiles than in the richest quintiles and almost two times more at risk of being undernourished than their richest counter parts. Our result is consistent with findings of studies that found that children in the poorest households are more at risk of being malnourished compared to their counterparts in the richest households [11, 24, 30]. This could be explained by the fact the rich are able to afford better living conditions that may improve child nutrition.

The following limitations need to be considered while interpreting the above findings. First, though the DHS data is of high quality, recall bias in reporting children birth histories remains a potential concern [31]. Moreover, a quality analysis of height data in 81 DHS surveys also showed clustering on whole and half centimeters, with some additional heaping at multiples of 5 and 10 cm [32]. Second, we have used all children and repeated their mothers and we did not match randomly a mother with her child. This is likely to bias our standard errors because we are not getting a full amount of new information by adding new children to the same mother. Instead we used the DHS clustering method (svyset) which accounted for this. The basic intuition is that, without DHS survey design, we would want to “cluster” on the household (all those with the same mother). But, since DHS requires us to cluster at the Primary Sampling Unit (PSU) level, and all children in a household are in the same PSU, we are already taking the within-household serial correlation in maternal BMI into account. Lastly, this study is limited by its cross-sectional nature and hence, we cannot determine the findings of the current study and elsewhere represent evidence of causal relationship between maternal nutrition status and child undernutrition. In addition, the direction of possible causality in maternal nutritional status and child undernutrition may not be the same across the life course.

Despite these limitations, our study has some persuasive strength. First, EDHS data are considered to be of high quality and reliable source of maternal and child health data. Second, the standardized procedures used by these surveys increases the quality of the data and also allows for comparability across countries. Third, it provides a nationally representative sample, allowing for drawing conclusions about the entire country. Moreover, we believe we used the appropriate statistical methods to investigate these relationships and controlling for important covariates.

In conclusion, maternal BMI and height are significantly associated with the child undernutrition but not accurate proxy indicators of child undernutrition. Therefore, regardless of the specific explanation, our findings show that the sensitivity and specificity of maternal anthropometric indicators to identify growth deficits among children were too low to justify using maternal indicators as a replacement for child growth measurements.