Abstract / Summary
ObjectiveFrailty has become an important health concern in older adults in the context of global population aging. The electronic frailty index (eFI), derived from routinely collected electronic health record data, has been increasingly used to assess frailty in clinical practice and research. However, no meta-analysis has systematically synthesized the association between eFI-defined frailty and mortality in older adults.MethodsWe searched PubMed, Embase, Web of Science, Cochrane Library, ProQuest, Ovid MEDLINE Complete, CINAHL, and Scopus for studies examining the association between eFI-defined frailty and mortality in older adults, from database inception to April 1, 2025. Studies reporting hazard ratios (HRs), odds ratios (ORs), or relative risks (RRs) with 95% confidence intervals (CIs) were included. Risk of bias was assessed using the Newcastle-Ottawa Scale (NOS). Meta-analysis was performed in Stata 17.0, and heterogeneity was assessed using the I2 statistic.ResultsTwelve studies comprising 13 cohorts were included, because one study contributed two analytical cohorts. Among these, 11 cohorts provided HR-based estimates and 2 cohort provided OR-based estimates. HR-based and OR-based estimates were analyzed separately. In the HR-based meta-analysis, eFI-defined frailty severity was associated with higher mortality risk. Compared with the non-frail group, the pooled HRs were 1.81 for mild frailty, 2.95 for moderate frailty, and 4.41for severe frailty. Substantial heterogeneity was observed across the main analyses. Subgroup analyses stratified by follow-up duration, eFI cut-off definition, study setting, and disease category suggested that the magnitude of association varied across study characteristics.ConclusionsThis meta-analysis provides quantitative evidence of a graded association between eFI-defined frailty severity and mortality in older adults. These findings suggest that the eFI may be useful in frailty screening and risk stratification in geriatric care, although the substantial between-study heterogeneity warrants cautious interpretation and the results should not be interpreted as direct evidence of predictive performance.Systematic review registrationhttps://www.crd.york.ac.uk/PROSPERO/, identifier CRD42025632191.