Multidimensional social determinants of health profiles and incident severe disability and recovery among older Chinese adults: a longitudinal CHARLS study

Social disadvantages often coexist, but conventional analyses usually examine single factors or scores. We examined whether multidimensional social determinants of health (SDOH) profiles were associated with incident severe disability among initially independent older Chinese adults; recovery was a secondary, exploratory outcome. We analysed 2011–2018 China Health and Retirement Longitudinal Study data. Participants were aged at least 60 years, had no baseline activities of daily living (ADL) or instrumental ADL (IADL) limitation, and completed at least one follow-up. Thirteen binary SDOH indicators entered a sampling-weighted Bernoulli finite-mixture model using full-information maximum likelihood. Severe disability was any observed episode of at least three limitations across 11 ADL/IADL items. Survey-weighted logistic and Fine–Gray-equivalent models estimated age- and sex-adjusted associations; death competed with severe disability. Weighted multistate recovery estimates were exploratory. Among 2,585 participants, four profiles were identified: relative advantage (weighted proportion 27.1%), rural/low-education (15.0%), rural/environmental deprivation (49.3%), and social isolation/deprivation (8.6%). Overall, 773 participants experienced severe disability. Relative to relative advantage, adjusted odds ratios (ORs) were 1.60 (95% confidence interval [CI] 1.04–2.47), 2.05 (1.49–2.84), and 1.44 (0.91–2.30), respectively. Weighted subdistribution hazard ratios (sHRs) were 1.34 (0.95–1.89), 1.64 (1.26–2.14), and 1.24 (0.85–1.80). Exploratory five-year mild-to-robust probabilities ranged from 0.341 to 0.561, but model-based CIs were wide and transition intensities varied over time. SDOH profiles were associated with incident severe disability. They describe combinations of social conditions, not effects of individual determinants. Specific supports require evaluation in designs suited to causal questions.