Answers are grounded in this paper and cite their sources.
Background Low birth weight (LBW), remains a major determinant to infant mortality and morbidity, serves as a critical indicator of maternal health, nutrition, and socio-economic conditions. Early detection enables targeted interventions in vulnerable rural settings. Aims This study applied artificial neural network (ANN)-based models across four domains. Model 1- socio-demographic and economic factors, Model 2 - prenatal conditions and service utilization factors, Model 3- maternal complications and morbidity factors and Model 4- maternal nutritional factors- to identify determinants for early risk prediction of LBW among neonates in rural eastern India. Methods Cross-sectional data from 352 pregnant women, encompassing prenatal to postnatal predictive factors, used to develop four ANN predictive models. Predictor importance was estimated using sensitivity analysis in SPSS v22 (®IBM Inc) Neural Networks. Model performance was evaluated using area under the curve (AUC), root mean squared error (RMSE) and F1 score. As very few infants met World Health Organization (WHO)-defined very low birth weight (VLBW) threshold (<1.5 kg), a study-defined VLBW subgroup (<2.2 kg) was adopted solely for analytical stratification. Results Model 3 showed best performance, with highest AUC (0.756), and lowest RMSE (0.394). Model 2 demonstrated moderate performance (AUC of 0.716) with F1 scores of 0.677 (study- defined VLBW) and 0.675 (LBW). Conclusions Maternal age, hemoglobin level and pre-pregnancy body mass index (BMI) emerged as key predictors of LBW. ANN models demonstrated discriminatory performance within internal validation framework, offering exploratory insights for early risk prediction. However, external validation using independent datasets is required before broader applicability can be established.
No takes yet. Share an insight, caveat, or question.
Ghosh et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: