Novel Frailty Assessment Based on Multidimensional Physical Frailty Parameters Using Unsupervised Clustering in...

Background: Frailty impacts the prognosis of respiratory diseases but lacks standardized evaluation criteria. This pilot study aimed to develop a frailty assessment method using unsupervised clustering of various physical function tests. Methods: Clinical data, handgrip strength (HS), lower limb strength (LLS), the 6 min walk test (6 min WT), the 5 m walk test (5 m WT), body composition, such as skeletal muscle mass index (SMI), whole-body phase angle (WBPhA), and pulmonary function variables were measured. Frailty status was evaluated in three groups (frail, pre-frail, and robust) using the J-CHS and Kihon Checklist. Unsupervised hierarchical clustering was performed, followed by dimensionality reduction using Principal Component Analysis. Results: Ninety-eight patients and healthy volunteers (70 males, 28 females; mean age, 57.5 years) were divided into four clusters, ranging from robust to pronounced frailty. On the 2-principal component plane, data points formed clusters across the four regions. The biplot showed variables aggregating in two directions: one including %FEV₁, FEV₁%, 6 min WT, and 5 m WT speed (exercise tolerance), and the other including HS, LLS, SMI, and WBPhA (physical elements). Tracking 39 participants (mean, 636 days later) showed cluster shifts that were broadly reproducible, although the small follow-up sample warrants cautious interpretation. Conclusions: As an exploratory, hypothesis-generating pilot study with a small single-center sample, this novel frailty model may offer a more granular assessment to help guide management; however, external validation in larger cohorts is required before clinical application.