Predictive performance of 7 frailty instruments for short-term disability, falls and hospitalization among Chinese community-dwelling older adults: A prospective cohort study

Abstract

BackgroundFrailty becomes a great challenge with population aging. The proactive identification of frailty is considered as a rational solution in the community. Previous studies found that frailty instruments had insufficient predictive accuracy for adverse outcomes, but they mainly focused on long-term outcomes and constructed frailty instruments based on available data not original forms. The predictive performance of original frailty instruments for short-term outcomes in community-dwelling older adults remains unknown. ObjectiveTo examine the predictive performance of seven frailty instruments in their original forms for 1-year incident outcomes among community-dwelling older adults. DesignA prospective cohort study. SettingsA total of 22 communities were selected by a stratified sampling method from one Chinese city. ParticipantsA total of 749 older adults aged ≥ 60 years (mean age of 69.2 years, 69.8% female) were followed up after 1 year. MethodsBaseline frailty was assessed by three purely physical dimensional instruments (i.e. Frailty Phenotype, the Study of Osteoporotic Fracture and FRAIL Scale) and four multidimensional instruments (i.e. Frailty Index, Groningen Frailty Indicator, Tilburg Frailty Indicator and Comprehensive Frailty Assessment Instrument), respectively. Outcomes included incident disability, falls, hospitalization and the combined outcome at 1-year follow-up. The receiver operating characteristic curves were plotted to assess the predictive performance of frailty instruments. ResultsThe areas under the curves of seven frailty instruments in predicting incident outcomes ranged from 0.55 [95% confidence interval (CI): 0.51–0.60] to 0.67 (95% CI: 0.61–0.72), with high specificity (72.3–99.2%) and low sensitivity (4.0–49.6%). Four multidimensional instruments had much higher sensitivity (20.9–49.6% versus 4.0–11.7%) than three purely physical dimensional instruments. Overall, the Frailty Index was more accurate than some instruments in predicting incident outcomes, while several self-report instruments had comparable predictive accuracy to the Frailty Index for all (FRAIL Scale) or some (Groningen Frailty Indicator and Tilburg Frailty Indicator) of the incident outcomes. ConclusionsAll frailty instruments have inadequate predictive accuracy for short-term outcomes among community-dwelling older adults. The Frailty Index roughly performs better but self-report instruments are comparable to the Frailty Index for all or some of the outcomes. An accurate frailty instrument needs to be developed, and the simple self-report instruments could be used temporarily as practical and efficient tools in primary care.