Predictive performance of four frailty screening tools in community-dwelling elderly

Bienvenu Bongue, Aurélie Buisson, Caroline Dupre, François Beland, Régis Gonthier, Émilie Crawford-Achour

This study compares the performance of four frailty screening tools in predicting relevant adverse outcome (disability, institutionalization and mortality) in community-dwelling elderly. Our study involved a secondary analysis of data from the FréLE cohort study. We focused on the following four frailty screening tools: the abbreviated Comprehensive Geriatric Assessment (aCGA), the Groningen Frailty Indicator (GFI), the Vulnerable Elders Survey-13 (VES-13) and the Fried scale. We used the Barberger-Gateau scale to assess disability. For comparison, we determined the capacity of these tools to predict the occurrence of disability, institutionalization or death using the receiver operating characteristic (ROC) curve. We also determined the threshold at which an optimal balance between sensitivity and specificity was reached. Odds ratios (ORs) were calculated to compare the risk of adverse outcome in the frail versus non-frail groups. In total, 1643 participants were included in the mortality analyses; 1224 participants were included in the analyses of the other outcomes (74.5% of the original sample). The mean age was 77.7 years, and 48.1% of the participants were women. The prevalence of frailty in this sample ranged from 15.0% (Fried) to 52.2% (VES-13). According to the Barberger-Gateau scale, 643 (52.5%) participants were fully independent; 392 (32.0%) were mildly disabled; 118 (9.6%) were moderately disabled; and 71 (5.8%) were severely disabled. The tool with the greatest sensitivity for predicting the occurrence of disability, mortality and institutionalization was VES-13, which showed sensitivities of 91.0%, 89.7% and 92.3%, respectively. The values for the area under the curve (AUC) of the four screening tools at the proposed cut-off points ranged from 0.63 to 0.75. The odds (univariate and multivariate analysis) of developing a disability were significantly greater among the elderly identified as being frail by all four tools. The multivariate analyses showed that the VES-13 may predict the occurrence of disability, mortality and institutionalization. However, the AUC analysis showed that even this tool did not have good discriminatory ability. These findings suggest that despite the high number of frailty screening tools described in the literature, there is still a need for a screening tool with high predictive performance.