Objectives: To compare the ability of Frailty Phenotype (FP), FRAIL and Frailty Index (FI) to predict adverse outcomes.

Design: A prospective cohort study.

Setting: A senior community in Beijing, China.

Participants: A total of 188 older adults aged 65 years or older (mean age 84.0 ± 4.4 years, 58.5% female).

Measurements: Frailty was evaluated by FP, FRAIL and FI. The agreement between scales was assessed by Cohen kappa coefficient. The predictive value of the three scales for adverse outcomes during one-year follow-up period were analyzed using decision curve analysis(DCA) and receiver operating characteristic curve (ROC) analysis.

Results: Frailty ranged from 25% (FRAIL) to 42.6% (FI). The agreement between scales was moderate to good (Cohen's kappa coefficient 0.44~0.61). DCA showed though the curves of the scales overlapped across all relevant risk thresholds, clinical treating had a higher net benefit than "treat all" and "treat none" when risk of unplanned hospital visits ≥30%, risk of functional decline or falls ≥15%. The three scales had similar predictive value for unplanned hospital visits (area under ROC, AUC 0.63, 0.64 and 0.69). FRAIL and FI had similar predictive value for functional decline (AUC 0.63,0.65). FI had predictive value for falls (AUC 0.65).

Conclusions: All three scales showed clinical utility but FRAIL may be best in practice because it is simple. Multidimensional measures of frailty are better than unidimensional for prediction of adverse outcomes among older adults.