Prev Med Rep. 2026 Nov 2;71:103684. doi: 10.1016/j.pmedr.2026.103684. eCollection 2026 Nov.
ABSTRACT
OBJECTIVE: To develop and internally validate explainable machine-learning models for identifying current overweight or obesity among preschool children in Heilongjiang, China.
METHODS: This cross-sectional analysis used monitoring data collected in Harbin, Shuangyashan, and Suihua from late August through late November 2020. After excluding 35 children with missing weight measurements, 1228 children were divided into training (860) and internal validation (368) sets before imputation. Overweight or obesity was defined as a body mass index standard deviation score ≥ +1. Least absolute shrinkage and selection operator regression selected predictors for seven models.
RESULTS: Five predictors were retained: sleep sufficiency, birth weight, maternal body mass index, weekend moderate-to-vigorous physical activity, and weekend screen time. Random forest had the highest numerical area under the receiver operating characteristic curve (0.82; 95% confidence interval, 0.77, 0.86), lowest Brier score (0.16), highest accuracy (0.76), and highest F1 score (0.62). Weekend physical activity ranked highest in global importance, followed by maternal body mass index.
CONCLUSIONS: Explainable machine learning may support pre-screening prioritization when anthropometric assessment is unavailable but should not replace measured height and weight or standard body mass index classification. External validation is required.
PMID:43034615 | PMC:PMC13769008 | DOI:10.1016/j.pmedr.2026.103684

