Comparison of frailty instruments in acute coronary syndrome patients

BackgroundRecognising frailty is important to guide clinical decisions in older patients with cardiac disease. The relative strengths of different frailty instruments to predict mortality and hospitalization risk are uncertain. MethodsThe Edmonton Frail Scale (EFS), Fried Criteria (Fried), Clinical Frailty Scale (CFS), Katz score (Katz), GP Cognition test (GPCog), and Euroscore II a disease based risk score, were completed in 1174 clinically stable inpatients >70 years of age admitted with an acute coronary syndrome to 5 New Zealand hospitals. Associations with all cause mortality (n=353, 29%) during a median follow-up of 5.1 (IQR: 4.6-5.5) years and hospitalization for > 10 days in the next year (n=267, 22%) were evaluated. ResultsThere were graded associations between increasing frailty assessed by each tool and all cause mortality. For the EFS, which scores up to 17 points on different dimensions of frailty, hazard ratios for high (score 9-17, n=197) compared to low frailty (score 0-2, n=331) were 5.0 (95%CI: 3.4-7.4) for mortality, and 5.3 (3.4-8.3) for hospitalization. Discrimination for all-cause mortality according to Harrells C-index for each instrument were EFS 0.663, Euroscore II 0.654, Fried 0.648, CFS 0.640, GPCog 0.608, and Katz 0.593, P<0.001 for all. C-statistics for hospitalization >10 days were EFS 0.649, Fried 0.628, Katz 0.602, Euroscore II 0.589, CFS 0.584, and GPCog 0.552, P<0.001 for all. When combining tools integrated discrimination improvement for both mortality and hospitalization were greater for EFS than for other frailty instruments. ConclusionIn acute coronary syndrome patients aged >70 years greater frailty assessed using all tools was associated with higher mortality and hospitalization. The Edmonton Frail Scale, which provides a graded measure of severity of frailty based on information relevant to clinical care, discriminated the risk of mortality and hospitalization as well or better than other frailty instruments.