Multimorbidity Trajectories Across Three National Ageing Cohorts: Early Branching States and Persistent Cardiometabolic-Musculoskeletal Profiles.

Background/

Objectives

Disease count alone does not show which multimorbidity combinations diversify or persist. We examined longitudinal changes in clinically recognisable multimorbidity profiles across three national ageing cohorts.

Methods

Harmonised data from the China Health and Retirement Longitudinal Study (CHARLS), the English Longitudinal Study of Ageing (ELSA), and the U.S. Health and Retirement Study (HRS) were analysed. Eight physician-diagnosed chronic conditions were encoded as binary states, and wave-to-wave transitions (four windows in CHARLS and ELSA; five in HRS) were assessed within each cohort. State-level measures characterised accumulation, branching, persistence, and stabilisation sensitivity, supplemented by sensitivity analyses, BranchScore decomposition, prevalence-adjusted enrichment, and a disease-count-preserving permutation null model.

Results

The analysis included 17,142 CHARLS, 10,272 ELSA, and 22,034 HRS participants, with baseline multimorbidity of 23.7%, 41.0%, and 58.5%, respectively. Transitions are concentrated around common profiles. One- and two-condition states (hypertension, diabetes, heart disease, chronic lung disease, psychiatric or emotional disorders) showed faster accumulation and greater branching; later persistent states were dominated by cardiometabolic-musculoskeletal combinations, particularly hypertension-heart disease-arthritis and hypertension-diabetes-arthritis. Targeted stabilisation produced modest perturbations exceeding random benchmarks. Count-preserving null models showed that early branching was largely structural, whereas selected lock-in states exceeded null expectations.

Conclusions

Across three ageing cohorts, multimorbidity trajectories reflected disease composition as well as count. Because branching was strongly influenced by disease-count geometry, early branching states should be interpreted as structural cohort-level features of the cumulative framework rather than inherently predictive clinical entities. Selected cardiometabolic-musculoskeletal profiles were more persistent and may help frame integrated long-term management; patient-level prediction requires outcome-based studies.