To identify multimorbidity phenotypes in middle-aged and older adults, examine their long-term health trajectories, and develop a predictive model for functional dependency.

This longitudinal analysis used five waves(2011-2020) of the China Health and Retirement Longitudinal Study(CHARLS) and included 298 participants with complete follow-up. Phenotype subgroups were derived by Kmeans clustering based on baseline disease count. Generalized Estimating Equations(GEE) were applied to assess health differences and trajectories across phenotypes. A care-needs prediction model was developed by integrating prior health status(2018) and pre-2018 dynamic slopes.

Three multimorbidity burden phenotypes were identified: high, moderate, and low. A higher baseline disease count was independently associated with increased risks of functional dependency(ADL: OR=2.398; IADL: OR=2.346), depression, and pain(all P<0.001). The high-burden phenotype showed faster rates of decline in cognitive function, pain accumulation, and life satisfaction(all P<0.01). The prediction model, incorporating prior functional status, depressive symptoms, phenotype, and rate of pain progression, demonstrated good discrimination(AUC for ADL dependency=0.807, 95%CI: 0.740-0.874; AUC for IADL dependency=0.765, 95%CI: 0.678-0.851).

Multimorbidity phenotypes are associated with distinct long-term health trajectories. A model combining static phenotypes and dynamic health indicators can help identify individuals at high risk of functional dependency, supporting targeted prevention and care planning.