Predictors of Preschool Neurodevelopmental Outcomes and Trajectories in a South African Birth Cohort.

Background: Early childhood is a critical period for brain development, yet many children in low- and middle-income countries face adversities that put them at risk of not reaching their developmental potential. To address gaps in understanding factors which predict developmental outcomes at 5 years in a South African birth cohort, this research assessed which early-life factors shaped developmental outcomes at 5 years and trajectories between 2-5 years. Methods: 728 mother-child dyads from the Drakenstein Child Health Study, were followed from the antenatal period through 5 years. Neurodevelopmental outcomes (cognition, receptive and expressive language, fine motor skills) were measured between 2-5 years using standardized instruments. Random forest models identified key predictors of each domain. Longitudinal trajectories from 2-5 years were identified using latent class growth analysis; and random forest modelling examined predictors of class membership. Results: Fifty theoretically motivated predictors of neurodevelopment accounted for over 90% of variance across neurodevelopmental domains at 5 years, in this South African birth cohort. Cross-validated models retained 9 for neurocognition, 15 key predictors for receptive language, 3 for expressive language and 1 for fine motor skills. The greatest variance explained was observed for neurocognition (23%), with modest variance explained for other domains. Critical predictors of neurocognition in this cohort included socioeconomic status indicators, maternal gravida and childhood trauma, sex, early anthropometry, and parenting practices. Predictors that demonstrated overlap between neurocognition and receptive language domains were socioeconomic status indicators, maternal gravida and parenting practices. Receptive language also included domain-specific predictors including maternal mental health, intimate partner violence, social support and childbirth anthropometry. Expressive language was best explained by 3 predictors, including in-utero HIV exposure, maternal smoking and gravida, whilst fine motor skills were best predicted exclusively by child sex. Trajectory analysis revealed 8.8-16% of children demonstrated developmental catch-up between 2-5 years. Across both univariate and multivariate classification models, 8 top-ranked predictors discriminated recovery-class membership; where child anthropometry, parenting practices and household crowding were shared predictors across models. Conclusion: Alongside factors predicting neurodevelopment variance at age 5-years, trajectory analysis identified developmental catch-up among a group of at-risk children, with maternal and environmental factors predicting this recovery trajectory.