Electronic frailty index based on comprehensive geriatric assessment in primary care: feasibility, validity, and reliability

BACKGROUND: To aid frailty management in primary care we developed an electronic Frialty Index based on a Comprehensive Geriatric Assessment (eFI-CGA). Here we report the feasibility, validity, and reliability of the eFI-CGA when conducted in primary care. METHODS: To assay feasibility, we measured assessment completion rate and time required. We tested validity by analyzing the content and characteristics of the eFI-CGA, its associations with age and the Clinical Frailty Scale (CFS), and prediction of three-year adverse outcomes. Reliability was evaluated by having primary care providers and geriatricians independently assess patients at baseline and at six months, yielding inter- and intra-rater estimates, as well as validity compared with a reference standard. RESULTS: Community-dwelling older adults from 27 clinics were enrolled (n = 281; 65% females). The eFI-CGA showed a four-factor structure, including function/mobility, psychological/social, cognition, and physical/medical problems. The eFI-CGA and CFS scores were moderately correlated: r = 0.65 (95% confidence intervals (CI): 0.56 − 0.72). The age − eFI-CGA correlation (r) was modest (r = 0.22; CI: 0.10 − 0.34). Each 0.01 eFI-CGA increment was associated with a Hazard Ratio (HR) of 1.04 for three-year mortality (CI: 1.02 − 1.06). The eFI-CGA discriminated individuals with an adverse outcome with 75% accuracy (CI: 68%−82%). The eFI-CGA conducted by geriatricians and primary care providers showed comparable characteristics and predictive capacities. Inter-rater intraclass correlation coefficients (ICCs) were high: 0.91 (CI: 0.84 − 0.94) baseline) and 0.87 (CI: 0.80 − 0.92) follow-up. Intra-rater ICCs were excellent 0.95 (CI: 0.92 − 0.97) for geriatricians and 0.93 (CI: 0.89 − 0.96) for primary care providers. CONCLUSIONS: Frailty assessment using the eFI-CGA is feasible, valid, and reliable in primary care settings. Larger studies are needed to evaluate its generalizability.