Abstract
BACKGROUND: As the population ages it is important to identify frailty, a powerful predictor of morbidity and mortality, and often an important unmeasured confounder. We sought to develop a frailty index in the Physician's Health Study (PHS) and estimate the association with mortality. METHODS: Prospective cohort study. Annual questionnaire assessed mood, function and health status. Two frailty scores were compared - cumulative deficit frailty index (PHS FI) and modified Study of Osteoporotic Fracture (mSOF) frailty score. Endpoints committee confirmed mortality. RESULTS: 12,180 male physicians ≥60 years were analyzed. Mean(SD) follow-up was 10(3) years, 2168 deaths occurred. PHS FI identified 4412 (36%) physicians robust, 5305 (44%) pre-frail, and 2463 (20%) frail, while mSOF identified 7323 (61%) robust, 3505 (29%) pre-frail and 1215 (10%) frail. Age-standardized rate of death was lower among subjects identified as robust using the PHS FI, 11/1000 person-years (PY) (95% Confidence Interval (CI): 9.5-11.9) compared to 14/1000PY (95% CI: 13.5-15.4) using mSOF [P-difference <0.001]. In the prefrail group, death rates were 16/1000PY in PHS FI and 21/1000PY in mSOF, [P-difference <0.001]. There was no difference in age-adjusted mortality rates in the frail group according to each definition (35 vs 33/1000PY). Survival analysis showed an increased risk of mortality in each frailty category using either definition, (log-rank p<0.001). CONCLUSION: The PHS FI outperformed mSOF in identifying risk of death particularly in robust and pre-frail categories. Similar indices can be created in existing datasets to identify frail individuals and where appropriate account for frailty, an often unmeasured confounder.