Recovered but Not Reset: Longitudinal Health History and Future Vulnerability Across Three National Cohorts

Abstract Background A favorable current geriatric assessment may conceal whether an older adult has been durably preserved or has only recently recovered from multidomain vulnerability. We tested whether longitudinal health history remains prognostically informative after current health is held constant. Methods We analyzed adults aged 60 years or older in HRS, ELSA, and CHARLS using repeated mobility, instrumental activities of daily living (IADL), episodic memory, and depressive symptoms. Among observations currently multidomain-preserved, modified Poisson models compared durable preservation with recent recovery after adjustment for current raw domain values, age, sex, education, and wave. Secondary analyses examined recurrent and remote vulnerability, persistence after isolated episodes, same-domain recurrence, incident basic-ADL limitation, and history-aware prediction. Results Recent recovery was associated with higher standardized next-wave deterioration risk in HRS (54.5% vs 34.9%; RR 1.56, 95% CI 1.47–1.66), ELSA (43.6% vs 29.7%; RR 1.47, 1.36–1.59), and CHARLS (56.4% vs 40.2%; RR 1.40, 1.28–1.54). Recurrent vulnerability carried the highest risk in all cohorts (RR 1.86–2.03). In HRS and ELSA, one isolated vulnerable episode remained informative after three subsequent preserved assessments. Prior cognitive, mobility, and affective vulnerability preferentially recurred in the same domain. Earlier health also improved participant-separated out-of-sample prediction beyond current health. Conclusions Apparent recovery and durable preservation were not prognostically equivalent. Longitudinal functional, cognitive, and affective history may improve near-term geriatric risk assessment when interpreted alongside current health.