Validating a widely used measure of frailty: are all sub-components necessary? Evidence from the Whitehall II cohort study

Abstract

There is growing interest in the measurement of frailty in older age. The most widely used measure (Fried) characterizes this syndrome using five components: exhaustion, physical activity, walking speed, grip strength, and weight loss. These components overlap, raising the possibility of using fewer, and therefore making the device more time- and cost-efficient. The analytic sample was 5,169 individuals (1,419 women) from the British Whitehall II cohort study, aged 55 to 79 years in 2007-2009. Hospitalization data were accessed through English national records (mean follow-up 15.2 months). Age- and sex-adjusted Cox models showed that all components were significantly associated with hospitalization, the hazard ratios (HR) ranging from 1.18 (95 % confidence interval = 0.98, 1.41) for grip strength to 1.60 (1.35, 1.90) for usual walking speed. Some attenuation of these effects was apparent following mutual adjustment for frailty components, but the rank order of the strength of association remained unchanged. We observed a dose-response relationship between the number of frailty components and the risk for hospitalization [1 component-HR = 1.10 (0.96, 1.26); 2-HR = 1.52 (1.26, 1.83); 3-5-HR = 2.41 (1.84, 3.16), P trend <0.0001]. A concordance index used to evaluate the predictive power for hospital admissions of individual components and the full scale was modest in magnitude (range 0.57 to 0.58). Our results support the validity of the multi-component frailty measure, but the predictive performance of the measure is poor.

Fig. 1

Flow of study participants through the Whitehall II study, UK, 1989–2010

Fig. 2

Kaplan–Meier curves showing the probability of future hospitalization by frailty status, Whitehall II study, UK, 2007–2010

Fig. 3

Hazard ratios (95 % confidence interval) for the association of combinations of frailty components with later hospitalization, Whitehall II study, UK, 2007–2010. ^aResults from two models: one with different combinations included in the model (diamonds); the other one with number of frailty components included in the model (squares). All analyses were adjusted for age at baseline and sex. The reference group was those with no frailty component. ^bThree combinations were not reported owing to too few hospitalizations: weight loss and slow walking speed (n = 0), weight loss and exhaustion (n = 2), and weight loss and low grip strength (n = 3). ^cOwing to low numbers, participants with three to five frailty components were collapsed