Validation of an Automated Brief Frailty Screening Tool Using Interactive Voice Response in Community-Dwelling Older Adults.

BACKGROUND: Frailty is a multidimensional condition linked to adverse outcomes in older adults, yet most screening tools rely on in-person assessment, limiting scalability. Automated approaches are needed for large-scale screening. PARTICIPANTS AND SETTING: We studied community-dwelling older adults enrolled in the Aging Study of PyeongChang Rural Area (ASPRA), South Korea. Cohort 1 included 154 participants (mean age 77.4 years, 77.3% women) with comprehensive geriatric assessments; Cohort 2 included 1532 participants (mean age 74.9 years, 53.8% women). An additional External Cohort of 86 adults aged ≥ 90 years (mean age 92.1 years, 59.3% women) was recruited across five urban and rural sites to evaluate generalizability. METHODS: We developed a rule-based frailty classification algorithm implemented on an Interactive Voice Response (IVR) platform. Participants were contacted by automated telephone calls and categorized as robust, pre-frail, frail, or severely frail. Validity was assessed against the 45-item Frailty Index (FI) and other geriatric outcomes. Discriminative ability was compared with the Clinical Frailty Scale (CFS) using area under the curve (AUC) analyses. RESULTS: IVR-based frailty classification showed strong, graded associations with FI, disability, gait speed, SPPB, and health-related quality of life (all p for trend < 0.01). In Cohort 1, the AUC for detecting frailty (FI ≥ 0.25) was 0.77 (95% CI, 0.69-0.85), comparable to the CFS (AUC 0.76, 95% CI, 0.67-0.85; p = 0.87). In Cohort 2, CFS scores increased consistently across IVR frailty categories in all age and gender strata (p for trend < 0.001). In the External Cohort, the FI also increased progressively across IVR categories (p for trend = 0.006), demonstrating external validity. CONCLUSIONS: A fully automated IVR system is a valid and feasible tool for frailty screening in community-dwelling older adults. Requiring no internet or human intervention, it offers a scalable solution for population-level frailty surveillance, especially in resource-limited settings.