Rapid cognitive decline, one-year institutional admission and one-year mortality: analysis of the ability to predict and inter-tool agreement of four validated clinical frailty indexes in the SAFEs cohort.

objectivesTo evaluate the predictive ability of four clinical frailty indexes as regards one-year rapid cognitive decline (RCD - defined as the loss of at least 3 points on the MMSE score), and one-year institutional admission (IA) and mortality respectively; and to measure their agreement for identifying groups at risk of these severe outcomes.

designOne-year follow-up and multicentre study of old patients participating in the SAFEs cohort study.

settingNine university hospitals in France.

participants1,306 patients aged 75 or older (mean age 85±6 years; 65% female) hospitalized in medical divisions through an Emergency department. MEASUREMENTS: Four frailty indexes (Winograd; Rockwood; Donini; and Schoevaerdts) reflecting the multidimensionality of the frailty concept, using an ordinal scoring system able to discriminate different grades of frailty, and constructed based on the accumulation of identified deficits after comprehensive geriatric assessment conducted during the first week of hospital stay, were used to categorize participants into three different grades of frailty: G1 - not frail; G2 - moderately frail; and G3 - severely frail. Comparisons between groups were performed using Fisher's exact test. Agreement between indexes was evaluated using Cohen's Kappa coefficient.

resultsAll patients were classified as frail by at least one of the four indexes. The Winograd and Rockwood indexes mainly classified subjects as G2 (85% and 96%), and the Donini and Schoevaerdts indexes mainly as G3 (71% and 67%). Among the SAFEs cohort population, 250, 1047 and 1,306 subjects were eligible for analyses of predictability for RCD, 1-year IA and 1-year mortality respectively. At 1 year, 84 subjects (34%) experienced RCD, 377 (36%) were admitted into an institutional setting, and 445 (34%) had died. With the Rockwood index, all subjects who experienced RCD were classified in G2; and in G2 and G3 when the Donini and Schoevaerdts indexes were used. No significant difference was found between frailty grade and RCD, whereas frailty grade was significantly associated with an increased risk of IA and death, whatever the frailty index considered. Agreement between the different indexes of frailty was poor with Kappa coefficients ranging from -0.02 to 0.15.

conclusionThese findings confirm the poor clinimetric properties of these current indexes to measure frailty, underlining the fact that further work is needed to develop a better and more widely-accepted definition of frailty and therefore a better understanding of its pathophysiology.