Back Extensor Strength as a Potential Marker of Frailty Using Propensity Score Matching and Machine Learning

Abstract

This study assessed the potential of back extensor strength as an alternative marker of frailty. A total of 560 farmers were included. Computed tomography scans measured fat and muscle mass volumes at the mid-L4 vertebral level. Back extensor strength was measured in a seated posture. Multivariate linear regression was used to analyze the associations between back extensor strength and trunk muscle/fat compositions. The participants were divided into two groups based on back extensor strength. Propensity score matching, multivariate logistic regression, and Extreme Gradient Boosting (XGBoost) were employed to evaluate the relationship between Fried’s frailty criteria and back extensor strength. Back extensor strength exhibited positive associations with abdominal muscle volume (r = 1.12) as well as back muscle volume (r = 0.89) (p < 0.05). Back extensor strength was linked to more frail status, such as reduced grip strength, walking speed, and frequent self-reported exhaustion. Multivariate logistic regression indicated that back extensor strength was associated with higher frail status (OR = 0.990), and XGBoost analysis identified back extensor strength as the most important predictor (gain = 0.502) for frailty. The prediction models using grip strength produced similar results (OR = 0.869, gain = 0.482). These findings suggested the potential of back extensor strength as an alternative frailty marker.