Should Sensory Impairment Be Considered in Frailty Assessment? A Study in the GAZEL Cohort.

Linard, M; Herr, M; Aegerter, P; Czernichow, S; Goldberg, M; Zins, M; Ankri, J

OBJECTIVES The assessment of sensory difficulties is sometimes included in the screening of frailty in ageing population. This study aimed to compare the prevalence of frailty and associated risk of adverse outcomes depending on whether sensory difficulties participated in the definition of frailty. DESIGN Prospective cohort study - GAZEL cohort. SETTING France. PARTICIPANTS The 13,128 subjects who completed a questionnaire in 2012. MEASUREMENTS According to the Strawbridge questionnaire, subjects were considered frail if they reported difficulties in two domains or more among physical, nutritive, cognitive and sensory domains. The risk of adverse health outcomes was assessed by using logistic regression models (hospitalisations, onset of difficulty in performing movements of everyday life) and multivariate Cox proportional hazards models (mortality). RESULTS Mean age was 66.8 +/- 3.4 years and 73.8% were males. The prevalence of frailty varied from 4.4 to 14.2% depending on whether the sensory domain was excluded or included. During follow-up, 182 deaths (1.4%), 479 hospitalisations (3.6%) and 703 cases of new disability (8.0%) were observed. Both definitions of frailty predicted the onset of difficulties to perform everyday movements, with 2 to 3-fold increase in the risk. The inclusion of the sensory domain in the definition made frailty predictive of hospitalisations (Odds Ratio 1.31 [1.01-1.70]) but the association with mortality was only observed when sensory difficulties were ignored (Hazard Ratio 2.28 [1.32-3.92]). CONCLUSION The inclusion of a sensory domain into a frailty screening instrument has a major impact in terms of prevalence and modifies the risk profile associated with frailty. In order to develop the use of frailty screening instruments in clinical practice, further researches will need to carefully evaluate the impact on risk prediction of the different domains involved.