Predicting neurodevelopmental delay among two-year-old infants in Northern Uganda

BACKGROUND: Data on neurodevelopmental disabilities in sub-Saharan Africa are scarce. Identifying children at the highest risk of neurodevelopmental delay could enable health workers to target these groups more effectively, improve resource allocation and programmatic focus. OBJECTIVE: To develop and assess a model for predicting neurodevelopmental delay among two-year-old infants in Northern Uganda using easily obtainable predictors. METHODS: We studied children born to women enrolled in a cluster randomized controlled trial (January 2018-March 2019) evaluating the effect of peer counseling on health facility births. At the trial's final visit (day 50), 1,804 children were enrolled. In November 2020, we followed up 1,421 children (78.8%) and assessed their neurodevelopment at 2 years using the Malawi Developmental Assessment Tool. We used least absolute shrinkage and selection operator penalized regression with 10-fold cross-validation to select the best subset of predictors and assessed model performance using the area under the curve (AUC). RESULTS: Of the 1,421 children, 70 (4.9%; 95% CI: 3.6-6.7%) had neurodevelopmental delays. Of the 19 candidate predictors, six were retained in the final model (lambda = 0.0076): moderate wasting, prolonged neonatal hospitalization (up to 28 days), maternal education, place of birth, preterm birth, and absence of skin-to-skin contact at birth. The model showed strong discriminative power (AUC 0.75) and reasonable calibration (calibration slope 1.37; calibration belt 0.536-0.727). CONCLUSION: We developed a promising tool for identifying infants at increased risk of neurodevelopmental delay in low-resource settings, which could help prioritize children for follow-up and specialized interventions.