Developing a Genetic Algorithm-Based Frailty Index for China Health and Retirement Longitudinal Study

Abstract

Objective

To develop a frailty index (FI) for predicting mortality and falls using The China Health and Retirement Longitudinal Study (CHARLS) data over 9 years.

Methods

We analyzed the 2011–2020 waves of CHARLS, employing a genetic algorithm (GA) for optimization. The outcomes focused on 9-year mortality and 2-year falls. Validation analyses included descriptive characteristics, concurrent correlation, predictive performance, calibration, and clinical utility assessments.

Results

The study included 6805 participants aged over 60 with a mean age of 68.4 years. The GA-FI, comprising 10 deficits, showed improved performance in all comparisons despite a modest AUC of 0.658 for predicting 9-year mortality. Although GA-FI improved falls prediction, together with other frailty measures the AUCs were consistently below 0.6, indicating challenges in predicting 2-year falls.

Discussion

The GA-FI is a valuable frailty measure in future CHARLS studies, and the identified deficits may guide frailty interventions and health education initiatives.