Lasso and XGBoost‐Enabled Prediction Models for Sensory Dysfunction, Biological Age, and APOE Genotype in Cognitive Decline Risk Assessment

Comprehensive Summary

This study examines how sensory dysfunction, biological age, and APOE genotype interact to predict cognitive decline in adults aged 50 and older. Using data from 13,222 participants in the Health and Retirement Study, the researchers evaluated vision, hearing, and taste impairments alongside a biomarker‑based biological age index reflecting cardiovascular, kidney, and metabolic health. Cognitive performance was measured using a validated 27‑point scale, with the lowest quartile classified as cognitive decline. Machine‑learning methods, specifically Lasso regression for feature selection and XGBoost for predictive modeling, were applied to identify the most influential risk factors. The results show that sensory impairments and elevated biological age are strong predictors of cognitive decline, and individuals carrying the APOE‑ε4 allele exhibit the highest vulnerability. The machine‑learning models outperformed traditional logistic regression, achieving improved accuracy and discrimination when sensory measures were included.

Outcomes and Implications

These findings are important because they suggest that simple, everyday health indicators, like changes in hearing or vision, can serve as early warning signs for cognitive decline. By integrating sensory assessments with biological age and genetic information, healthcare providers could identify at‑risk adults earlier and intervene sooner. This approach may improve screening strategies and help guide personalized prevention efforts in aging populations.