The aim of this study was to model the predictors of stunting using a dataset that is representative of the nation, and the findings could possibly be put in to practice. To the best of our knowledge, this is the first study that analyzed the 2016 EDHS and considered nutritional factors and dietary patterns, developed a model, and rigorously tested for its fitness. The model will open the door for future studies to be conducted and improve the Area under the curve (AUC), sensitivity and specificity of the model by including other important factors that are not measured in this study. Therefore, the prevalence of stunting in included children of 6 to 59 months found to be 39.1%. More importantly, we found that children aged above 24 months, low weight, small size, being male at birth, short maternal stature, overweight and obese maternal status, rural residence, no education and primary education of husband, and mild anemia of the mother were predictors of stunting. Among this child age, low weight, and maternal stature were the strongest predictors with AOR above 2.
Child age above 24–59 months was strongly associated with stunting with the AOR above 2. A number of studies have found similar associations [14, 17, 19]. This evidence corroborates that stunting is malnutrition, which starts during pregnancy and continues until the second year of life with the most frequent appearance after the second year of life . However, this finding does not mean intervening after the second year of life is not effective as there are numerous immediate avoidable factors that, if intervened, may reduce the effects of stunting after the second year of life. Studies have shown that factors associated with stunting in late childhood and adolescence are different, and there are other windows of opportunities to stop chronic malnutrition if the first 1000 years window is missed . Regardless, our study finding reaffirms that, to reduce stunting, intervention strategies ideally will focus on the child before 24 months. Therefore, policies and interventions should focus on the first one thousand days to prevent the stunting incidence. Equally important, the period between 24 and 59 months is crucial to provide a personalized intervention, such as providing specific nutritional needs for children with stunting, to accelerate catch-up growth.
The current weight of the child was the strongest predictor of stunting, according to the 2016 EDHS data. The continuous form of the variable was shown in a non-linear relationship (quadratic) with stunting. Consequently, the variable is categorized into quartiles. There are no previous studies that assessed the effect of child weight on stunting directly, although studies have shown that wasting (weight to height) is associated with stunting [22, 23], low birth weight is associated with stunting , and weight to-age are indicators which reflect cumulative effects of wasting and stunting . The association between child weight and stunting could be very strong because there are a few children with a large weight and being stunted at the same time; hence, this could inflate the association. Children’s weights reflect body composition with a recent study finding that stunted mothers have low body composition, including small kidneys and other organs, are thin, and give birth to small infants . The implication of this finding is that weight monitoring of children is critical in preventing and intervening in cases of stunting. Prospective future studies need to assess the effectiveness of weight monitoring overtime on stunting incidence.
Maternal stature was one of the independent predictors of stunting. Different study findings were in line with our findings [21, 27]. Maternal stature reflects the intergenerational effects of stunting, maternal malnutrition, and its consequences on childhood outcomes . This knowledge suggests that focusing on mothers of short stature might help in improving birth outcomes and in following-up of children after birth. However, the maternal stature cut-off point we used was arbitrary as there is no universally agreed cut-off point in the literature. Intuitively, the cut-off point should be locally derived as genetics may also have a role in determining one’s height. This observation could be similarly applied to stunting’s definition which might not be a perfect indicator of malnutrition as linear growth failure may be caused by different biological causes apart from inadequate nutrition .
In addition, maternal BMI, rural residence, paternal education, being male, and child’s size were important independent predictors. These findings are similar to those of recent literature [14, 15, 19, 20, 24]. Although there is some contrasting evidence regarding the association of sex with stunting , most studies’ findings indicated male children are more likely to be affected by stunting [15, 16, 19, 26], which could be due to biological or metabolic differences across sexes . A meta-analysis of 16 demography health surveys conducted in SSA concluded that male children under 5 years of age are more likely to become stunted than females, which might suggest that boys are more vulnerable to health inequalities than their female counterparts in the same age groups . Further studies are required to uncover the reason behind this finding. Policies and interventions that on stunting reduction need to place a due emphasis on gender differentials.
Education is also a well-established factor for the likelihood stunting. Previous studies have found a similar association [19, 26]. Stunting is a complex problem correlated with the socioeconomic level of a society and a nation at large. Achieving the sustainable development goals is addressing the problem of stunting indirectly . However, since stunting affects generations, hundreds of millions of individuals are already stunted, so intervention strategies should focus not only first 1000 years but also on these different population groups to stop the stunting syndrome as there might be other windows of opportunities for intervention .
The prevalence of stunting has decreased by about 1% every year from 58% in 2000 to 38% in 2016 [11, 36] and unexpectedly has shown almost no reduction between 2016 and 2019 at 37% [11, 12]. It has increased in some regions of Ethiopia, such as in Tigray region; it has increased by almost 10% from 38% in 2016 to 48.7% 2019 [11, 12]. This probably indicates the failure to design workable policies and locally based interventions. Particularly, data on dietary intake were not collected and analyzed in previous demographic surveys. Therefore, this study gives an important insight to design programs and policies based on the local evidence to prevent and address stunting.
The main limitation of this study is missing data on some variables, especially with respect to our intent to include all nutritional factors, including dietary habits which was limited due to missing data (almost 40%) in the EDHS data set. Multiple imputation was beyond the scope this analysis. Moreover, AUC of the model we developed is not high because of missing or unmeasured variables which indicate the need for future supportive studies to improve its predictive ability. However, it evident from the descriptive statistics that, on average, very few (i.e., one in five of children) received adequate and diverse nutrition which included grains, animal and plant-based proteins, fruits and vegetables.
Our model indicates that being born male, being from a mother of short stature, living in a rural area, small child size, mother with mild anemia, father with no education or primary education only, low weight of the child, and over 24 months of age increases the likelihood of stunting. On the other hand, being born to an overweight or obese mother decreases the likelihood of stunting.
The model fits the data very well with an AUC of 77%. Improving the nutritional and socioeconomic status of women pre-pregnancy and during pregnancy might reduce the burden of stunting.
Future studies are needed which focus on determining the association between body composition of the mother, medical conditions, and childhood stunting, finding another window of opportunity to minimize/reverse the consequences in already stunted children, and impact of low child weight should receive due emphasis on future endeavors of reducing chronic malnutrition. Moreover, we recommend the DHS to include more robust, and a contextually customized objective dietary/ nutritional patterns questionnaires in upcoming surveys.