Functional impairment profiles and long-term care implications among community health service-managed older adults with disability in China: a latent profile and network analysis study

Abstract

Objective To identify functional impairment profiles among older adults with disability managed within a community health service catchment in China and examine classification-adjusted associations between living arrangement and profile membership. Methods This cross-sectional study included 361 older adults with disability from 13 communities in Kunming, China, surveyed from January to March 2026. Functional ability was assessed using the Specification for Ability Assessment of Older Adults (GB/T 42195–2022). Latent profile analysis used self-care ability, basic mobility, and mental state as indicators; a four-domain model including sensory perception and social participation was tested for sensitivity. Sequential R3STEP models examined living arrangement while accounting for classification uncertainty and progressively adjusting for demographic, socioeconomic, family-resource, and health-related variables. Exploratory network analysis compared the four functional domains between living-arrangement groups. Results Four profiles were identified: severe multidomain impairment (28.3%), impaired mental state with relatively preserved physical function (12.7%), moderate functional impairment (53.5%), and relatively preserved function (5.5%). The four-domain model showed 92.5% classification agreement (Cohen's κ = 0.882). In the fully adjusted R3STEP model, compared with institutional living, home living was associated with greater odds of membership in the moderate functional impairment profile rather than the severe multidomain impairment profile (aOR = 7.046, 95% CI: 2.827–17.562; P < 0.001) and the relatively preserved function profile (aOR = 15.408, 95% CI: 2.652–89.528; P = 0.002), although the latter estimate was imprecise. No significant association was found for the impaired mental state with relatively preserved physical function profile (aOR = 0.570, 95% CI: 0.131–2.476; P = 0.453). Estimates were consistent across sequential models. Network analysis showed differences in overall structure and global expected influence, but not global strength; only the self-care ability–mental state edge differed after Holm correction. Conclusion Functional impairment was heterogeneous. Home living was associated with membership in the moderate functional impairment and relatively preserved function profiles after adjustment for measured covariates. These cross-sectional associations do not establish causal effects because residual confounding and care-selection effects remain possible. Profile-based assessment may complement existing grading by identifying differentiated care needs.