The Prognostic Accuracy of Frailty and Vulnerability Screening for Older Adults in the Emergency Department: A Systematic Review and Meta-analysis.

Objective: Frailty and vulnerability are associated with increased morbidity and mortality in older adults, yet the optimal screening tool for predicting adverse outcomes in the emergency department (ED) remains unclear. Our question is: Which frailty or vulnerability screening instrument has the highest prognostic accuracy for adverse outcomes in older adults visiting the ED?

Methods: We included observational studies involving patients aged more than or equal to 60 years presenting to the ED that applied frailty or vulnerability instruments and reported sensitivity, specificity, and area under the curve (AUC). We searched MEDLINE, EMBASE, Cochrane Library, CINAHL, and CNKI through January 2025. Study quality was assessed using an updated the Quality in Prognosis Studies tool. Two investigators independently screened studies. Meta-analysis was conducted for instruments with consistent cutoffs reported in more than or equal to 4 studies.

Results: Fifty-seven studies (125,412 patients) assessed 39 instruments with varied cutoffs. The Identification of Seniors at Risk and Clinical Frailty Scale were studied most. Most tools demonstrated high sensitivity but low specificity. For 30-day mortality, pooled Identification of Seniors at Risk estimates were as follows: sensitivity 92% (95% confidence interval, 84 to 96), specificity 37% (26 to 50), likelihood ratio (LR)+ 1.47 (1.25 to 1.79), LR- 0.24 (0.11 to 0.42), and AUC 0.84 (0.60-0.89). Clinical Frailty Scale (cut-off of more than or equal to 5) showed sensitivity of 81% (62 to 91), specificity 71% (54 to 83), LR+ 2.80 (1.96 to 3.82), LR- 0.29 (0.15 to 0.48), and AUC 0.82 (0.77 to 0.85).

Conclusion: Current screening instruments fail to identify older individuals who will experience adverse outcomes. Our findings suggest that EDs should reconsider relying on existing screening tools as standalone prognostic instruments and explore incorporating additional domains to improve accuracy.