Evaluating Frailty Index Integrity: Insights from an International Network Study

Deficit-accumulation Frailty Indices (FIs) have rapidly been integrated into health systems worldwide to quantify the state of reduced physiologic reserve to recover from a health insult in the aging population. Common data models (CDMs) have further enabled international, interinstitutional applications of FIs. However, the validity of such applications is unknown. We conducted an international network study comparing estimates of frailty from two electronic health record (EHR)-based FIs: United States (US)-based Veterans Affairs FI (VAFI) and United Kingdom (UK)-based electronic FI (eFI) across 5 research and clinical databases. US: <i>All of Us</i> (n=159,721), IQVIA Pharmetrics+ (n=5,099,557); UK: IMRD-THIN (n=3,036,003), IMRD-EMIS (n=832,455), and BioBank (n=207,202). In US databases, VAFI identified higher proportions of frailty ([VAFI; eFI] <i>All of US</i>: 10.3%; 2.5%, Pharmetrics+: 9.6%; 2%) while in UK databases, the eFI identified higher proportions (IMRD-THIN: <0.003%; 0.08%, IQVIA-EMIS: 0.09%; 0.4%, UK BioBank: 0.03%; 0.1%). Additional manipulations (alternative lookback periods, FI variations) were examined. FIs are likely dependent on their development context, such as local coding behaviors and incentives, limiting their external validity despite CDM harmonization. We suggest caution in the application of FIs outside of their development context and recommend further instrument development before more widespread use.