Ageing in the Field: Interpretable Chronological Age Prediction from Functional and Physiological Markers in LASI Wave 1

Abstract

Population ageing in India presents an urgent need for scalable, interpretable and field-compatible approaches to understand age-related functional and physiological variation. Chronological age is universally available, but it does not fully capture differences in health status, mobility, strength, metabolic condition and functional capacity among individuals of the same age. This study developed a reproducible chronological age-prediction framework from the Longitudinal Ageing Study in India. Data from 4,473 participants were integrated across Individual, Biomarker, Dried Blood Spot and Spirometry datasets. Base predictors included sex, maximum grip strength, walking speed, pulse rate, mean sleep problem score and body mass index, while the extended predictor set additionally included blood-pressure, hemoglobin, HbA1c, C-reactive protein, FVC and FEV1 measures. Four analytical datasets were constructed using complete-case and median-imputed strategies. Linear Regression, Random Forest and Gradient Boosting models were evaluated using five-fold cross-validation and an 80:20 internal holdout test design, supported by leakage audit, baseline comparison and residual diagnostics. The Base Complete-Case Linear Regression model achieved the strongest performance, with Test R² = 0.86, RMSE = 4.09 years, MAE = 3.25 years and MAPE = 6.08%. Base Complete-Case Random Forest and Gradient Boosting followed closely, with Test R² values of 0.85 and 0.84, respectively. Qualitatively, the findings indicate that simple field-compatible indicators captured substantial chronological age-related variation, whereas extended biomarker and spirometry variables did not improve prediction under the present missing-data conditions. The framework offers a transparent research-oriented approach for ageing studies in India, requiring longitudinal and external validation before clinical or public-health application.