Determinants of Childhood Infectious Morbidity in Indonesia: Evidence from a National Survey and Machine-Learning Prediction Models

Background Childhood infectious morbidity remains a major cause of preventable illness in low- and middle-income countries. Understanding modifiable determinants and improving risk prediction are essential to guide prevention. This study examined the association between early-life health practices and infectious morbidity among Indonesian children under five years, and compared the predictive performance of classical statistical and machine-learning models. Methods We analysed nationally representative data from 334,878 children aged 0–59 months from the 2022 Indonesian National Health Survey. Childhood infectious morbidity was defined as a composite outcome indicating at least one episode of acute respiratory infection, diarrhoea, or pneumonia during the recall period. Determinants were examined using survey-weighted logistic regression to estimate adjusted odds ratios (aORs). Predictive performance was assessed using logistic regression, Random Forest, and Gradient Boosting models. Discrimination was evaluated using the area under the receiver operating characteristic curve (AUC), and calibration using calibration plots. Subgroup analyses were conducted by age, sex, and immunisation status. Results Children aged ≥ 24 months had higher odds of infectious morbidity than those aged < 24 months (aOR 5.20; 95% CI 4.98–5.43). Male sex (aOR 1.22; 95% CI 1.20–1.24), lack of Maternal and Child Health (MCH) handbook ownership (aOR 3.27; 95% CI 3.08–3.47), absence of exclusive breastfeeding (aOR 1.12; 95% CI 1.10–1.13), and not receiving deworming treatment (aOR 0.47; 95% CI 0.44–0.49; reference = received) were also independently associated with morbidity. Immunisation status was not associated after adjustment. Predictive discrimination was modest across all models (AUC 0.605–0.612), with Gradient Boosting performing slightly better. Performance varied across subgroups and was highest among children < 24 months and those with incomplete immunisation. Conclusion Early-life health practices and indicators of health-system contact were strongly associated with infectious morbidity among Indonesian children. Machine-learning models provided only marginal gains over logistic regression. Strengthening modifiable maternal and child health practices, including breastfeeding support, MCH handbook use, and deworming coverage, may help reduce childhood infectious morbidity.