Modeling multimorbidity among older adults in India using holdout-validated logistic regression

Population ageing in India is accompanied by a rising prevalence of multiple chronic conditions, posing significant public-health challenges. This study estimates the prevalence and determinants of multimorbidity among older adults in India and evaluates the predictive performance of a holdout-validated logistic regression model. Data from Wave 1 of the Longitudinal Ageing Study in India (LASI, 2017–2018) were analyzed for adults aged ≥ 45 years (n = 57,417). A cross-sectional analytical design was applied. The dataset was split (70% training, 30% testing) for holdout validation. Associations between multimorbidity (≥ 2 chronic conditions) and socio-demographic, behavioral, and functional variables were examined using χ² tests and binary logistic regression. Model performance was assessed through receiver-operating-characteristic (ROC) analysis, area under the curve (AUC), and Z-tests comparing classification accuracy between training and test datasets. The prevalence of multimorbidity was 18%. Significant predictors included age, gender, marital status, education, work status, religion, caste, residence, region, MPCE quintile, tobacco use, depression, sleep quality, physical activity, IADL, ADL, life satisfaction, and self-rated health. The logistic model achieved an AUC of 0.754 (95% CI 0.748–0.759), with balanced sensitivity (68.6%) and specificity (68.9%) at the optimal cut-off (p = 0.187). Z-test results (|Z| < 1.96) indicated no significant difference in prediction accuracy between training and test datasets, confirming model stability. Multimorbidity affects nearly one-fifth of Indian older adults, highlighting the need for risk-stratified, evidence-based interventions. The validated logistic model demonstrates robust internal reliability and can aid early detection and planning of geriatric health strategies in India.