A large body of research investigates the determinants of stunting in young children, but few studies have considered which factors are the most important predictors of stunting. We examined the relative importance of predictors of height-for-age z-scores (HAZ) and stunting among children under 5 years of age in seven of the most food-insecure districts in South Africa using data from the Grow Great Community Stunting Survey of 2022. We used dominance analysis and variable importance measures from conditional random forest models to assess the relative importance of predictors. We found that intergenerational and socioeconomic factors-specifically maternal height (HAZ: Coef. 0.02, 95% CI 0.01-0.03; stunting: OR 0.96, 95% CI 0.94-0.98), birth weight (HAZ: Coef. 0.3, 95% CI 0.16-0.43; stunting: OR 0.5, 95% CI 0.35-0.72) and asset-based measures of socioeconomic status (HAZ: Coef. 0.17, 95% CI 0.10-0.24; stunting: OR 0.77, 95% CI 0.67-0.89)-were the most important predictors of HAZ and stunting in these districts. We explored whether any other factors moderated (weakened) the relationship between these intergenerational factors and child height using conditional inference trees and moderation analysis. We found that being on track for vitamin A and deworming, adequate sanitation, a diverse diet and good maternal mental health moderated the effect of birth weight or mother's height. Though impacts are likely to be small relative to the impact of intergenerational risk factors, these moderating factors may provide promising avenues for helping to mitigate the intergenerational transmission of stunting risk in South Africa.
Keywords: South Africa; birth weight; random forest; sanitation; socioeconomic factors; stunting; vitamin A.
© 2024 The DG Murray Trust and The Author(s). Maternal & Child Nutrition published by John Wiley & Sons Ltd.