Life expectancy estimates based on comorbidities and frailty to inform preventive care

Background: Long-term prognostication is important to inform preventive care in older adults. Existing prediction indices incorporate age and comorbidities. Frailty is another important factor in prognostication. In this project, we aimed at developing life expectancy estimates that incorporate both comorbidities and frailty. Methods: In this retrospective cohort study, we used data from a 5% sample of Medicare beneficiaries with and without history of cancer from Surveillance, Epidemiology, and End Results (SEER) cancer registry areas. We included adults aged 66–95 years who were continuously enrolled in fee-for-service Medicare for ≥1 year from 1998 to 2014. Participants were followed for survival until 12/31/2015, death, or disenrollment. Comorbidity (none, low/medium, high) and frailty categories (low, high) were defined using established methods for claims. We estimated 5- and 10-year survival probabilities and median life expectancies by age, sex, comorbidities, and frailty. Results: The study included 479,646 individuals (4,128,316 person-years), of whom most were women (58.7%). Frailty scores varied widely among participants in the same comorbidity category. In Cox models, both comorbidities and frailty were independent predictors of mortality. Individuals with high comorbidities (HR, 3.24; 95% CI, 3.20–3.28) and low/medium comorbidities (HR, 1.36; 95% CI, 1.34–1.39) had higher risks of death than those with no comorbidities. Compared to low frailty, high frailty was associated with higher risk of death (HR, 1.55; 95% CI, 1.52–1.58). Frailty affected life expectancy estimates in ways relevant to preventive care (i.e., distinguishing <10-year versus >10-year life expectancy) in multiple subgroups. Conclusion: Incorporating both comorbidities and frailty may be important in estimating long-term life expectancies of older adults. Our life expectancy tables can aid clinicians' prognostication and inform simulation models and population health management.