Application of an Electronic Frailty Index to Identify High‐Risk Older Adults Using Electronic Health Record Data

ABSTRACT Background Measurement of frailty is limited in clinical practice. Existing electronic frailty indices (eFIs) are derived from routine primary care encounters, with near‐complete health condition capture. We aimed to develop an eFI from routinely collected clinical data and evaluate its performance in older adults without complete health condition capture. Methods Using Electronic Health Record (EHR) data from an integrated regional health system, we created a cohort of patients who were ≥ 60 years on January 1, 2017 with two outpatient encounters in 3 years prior or one outpatient encounter in 2 years prior. We developed an eFI based on 31 age‐related deficits identified using diagnostic and procedure codes. Frailty status was categorized as robust (eFI < 0.1), prefrail (0.1–0.2), frail (0.2–0.3), and very frail (> 0.3). We estimated cumulative incidence of mortality, acute care visits and readmissions by frailty, and fit Cox proportional hazards models. We repeated analyses in a sub‐cohort of patients who receive primary care in the system. Results Among 518,449 patients, 43% were male with a mean age of 72 years; 73% were robust, 16% were pre‐frail, 7% were frail, and 4% were very frail. Very frail older adults had a significantly higher risk for mortality (HR: 4.1, 95% CI: 4.0–4.3), acute care visits (HR: 5.5, 95% CI: 5.4–5.6), and 90‐day readmissions (HR: 2.1, 95% CI: 2.1–2.2) than robust older adults. In a primary care sub‐cohort, while prevalence of deficits was higher, associations with outcomes were similar. Conclusions This eFI identified older adults at increased risk for adverse health outcomes even when data from routine primary care visits were not available. This tool can be integrated into EHRs for frailty assessment at scale.