Predicting Mortality and Adverse Outcomes: Comparing the Frailty Index to General Prognostic Indices
Sandra M. Shi,Ellen P. McCarthy,Susan L. Mitchell,Dae Hyun Kim
DOI: https://doi.org/10.1007/s11606-020-05700-w
IF: 5.7
2020-01-01
Journal of General Internal Medicine
Abstract:Background Mortality prediction models are useful to guide clinical decision-making based on prognosis. The frailty index, which allows prognostication and personalized care planning, has not been directly compared with validated prognostic models. Objective To compare the discrimination of mortality, disability, falls, and hospitalization between a frailty index and validated prognostic indices. Design Secondary Analysis of the National Health and Aging Trends Study. Participants Seven thousand thirty-three Medicare beneficiaries 65 years or older. Measurements We measured a deficit-accumulation frailty index, Schonberg index, and Lee index at the 2011 baseline assessment. Primary outcome was mortality at 5 years. Secondary outcomes were decline in activities of daily living (ADL), decline in instrumental activities of daily living (IADL), fall, and hospitalization at 1 year. We used C-statistics to compare discrimination between indices, adjusting for age and sex. Results The study population included 4146 (44.8%) with age ≥ 75 years, with a median frailty index of 0.15 (interquartile range 0.09–0.25). A total of 1385 participants died (14.7%) and 2386 (35.2%) were lost to follow-up. Frailty, Schonberg, and Lee indices predicted mortality similarly: C-statistics (95% confidence interval) were 0.78 (0.77–0.80) for frailty index; 0.79 (0.78–0.81) for Schonberg index; and 0.78 (0.77–0.80) for Lee index. The frailty index had higher C-statistics for decline in ADL function (frailty index, 0.80 [0.78–0.83]; Schonberg, 0.74 [0.72–0.76]; Lee, 0.74 [0.71–0.77]) and falls (frailty index, 0.66 [0.65–0.68]; Schonberg, 0.61 [0.58–0.63]; Lee, 0.61 [0.59–0.63]). C-statistics were similar for decline in IADL function (frailty index, 0.61 [0.59–0.63]; Schonberg, 0.60 [0.59–0.62]; Lee, 0.60 [0.58–0.62]) and hospitalizations (frailty index, 0.68 [0.66–0.70]; Schonberg, 0.68 [0.66–0.69]; Lee, 0.65 [0.63–0.67]). Conclusions A deficit-accumulation frailty index performs as well as prognostic indices for mortality prediction, and better predicts ADL disability and falls in community-dwelling older adults. Frailty assessment offers a unifying approach to risk stratification for key health outcomes relevant to older adults.