Abstract
<title>Abstract</title> <p>Background Systematic screening in the community enables the early detection of frailty, a reversible pre-disability condition that can be treated, forestalling the transition to dependency. Few studies have systematically compared the added predictive ability of different frailty-screening instruments beyond age, sex and multimorbidity to predict mortality or dependency within the same population of community-dwelling older adults. Our aim was to compare the ability of seven commonly used frailty instruments to predict the risk of mortality or developing basic activities of daily living (BADL) dependency within three-years among community-dwelling older adults. Methods The study cohort consisted of 779 non-dependent community-dwelling Spaniards aged ≥ 70 years. We developed logistic regression models to compare the discriminatory performance of seven frailty instruments -Fried´s phenotype, frailty index (FI), FRAIL scale, FTS3, FTS5, SPPB, and gait speed- in predicting death or the onset of dependency in BADL over a three-year follow-up. Discriminatory ability was assessed using the c-statistic, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Results The continuous FI showed the highest c-statistic (0.808; 95%CI: 0.761–0.849), followed closely by the continuous FRAIL scale (0.798; 95%CI: 0.754–0.844). The continuous FI achieved the second-highest NRI (0.470; 95%CI: 0.277–0.662), while the continuous FTS5 yielded the highest IDI (0.029; 95%CI: 0.011;0.046). Continuous measures added granularity to risk prediction. Conclusion According to our findings, the continuous electronic FI, the FRAIL scale, and the FTS5, provided greatest added discriminatory values. SPPB and gait speed with traditional cut-offs are not useful for this purpose.</p>