Predictive value of birth weight and age for preliminary community-level stratification of undernutrition risk among children under 5: a cross-sectional study in rural Indonesia

PURPOSE: We evaluated the predictive value of low birth weight (LBW), child age, and maternal education for stunting, wasting, and underweight among children aged 0-59 months. METHODS: We analyzed cross-sectional surveillance data from 394 children in Pesisir Selatan District, West Sumatra, Indonesia, and classified nutritional status according to the 2006 World Health Organization Child Growth Standards. For each outcome, we fitted a parsimonious logistic model (LBW, age, maternal education, and sex) and an expanded model incorporating 8 additional variables (9 regression coefficients). We assessed discrimination using the area under the receiver operating characteristic curve (AUROC) and calibration using the Hosmer-Lemeshow test and calibration slope. RESULTS: The prevalence of stunting, wasting, and underweight was 14.5%, 10.7%, and 17.3%, respectively. LBW was the strongest predictor in the parsimonious model for stunting (adjusted odds ratio [AOR], 4.85; 95% confidence interval [CI], 2.53-9.29), wasting (AOR, 2.88; 95% CI, 1.38-6.02), and underweight (AOR, 4.34; 95% CI, 2.30-8.19; p≤.005). A lack of formal maternal education was independently associated with stunting and underweight. The optimism-corrected AUROCs were .706, .631, and .751, with calibration slopes of .88, .75, and .90, respectively. Adding variables in the expanded model did not improve fit (likelihood-ratio p>.05) and increased optimism. Conclusion: The parsimonious model showed modest discrimination for stunting and underweight. Discrimination for wasting was inadequate. A predictive contribution from the added variables could not be demonstrated with the available number of events. These preliminary findings require external validation before community-level application.