Does method of frailty measurement affect prevalence or associations with health-related outcomes in those affected by stroke?

Talk Code: 
Katie Gallacher
Peter Hanlon, Terry Quinn, Jenni K Burton, Frances S Mair
Author institutions: 
Institute of Health and Wellbeing, University of Glasgow Institute of Cardiovascular and Medical Sciences, University of Glasgow


There has been recent interest in the measurement of frailty to enable identification of those at risk of adverse health-related outcomes. There is a scarcity of research on frailty after stroke, with no consensus on method of measurement. This study aims to examine the prevalence of frailty in stroke survivors and associations with health-related outcomes using differing methods of measurement.


A pooled cohort was created of participants with self-reported stroke aged >50 years from three datasets: the USA Health and Retirement Survey (HRS), the Survey for Health, Ageing and Retirement in Europe (SHARE) and the English Longitudinal Study of Ageing (ELSA). Frailty was assessed at baseline using: 1) frailty phenotype (low grip strength, unintentional weight loss, slow walking speed, exhaustion, low physical activity; 1-2 = pre-frail, 3 or more = frail); 2) the frailty index (a cumulative total of age-related deficits, divided by the total number of possible deficits; <0.12 = robust , 0.12-0.24 = mild, 0.24-0.36 = moderate, >0.36 = severe frailty); 3) the frailty index with an additional six cognitive measures (e.g. immediate and delayed recall, naming current date). Outcomes were assessed at the first data collection point following baseline (approximately 2 years follow-up) and included mortality/hospital admission in the preceding 2 years, and recurrent stroke. Logistic regression models were adjusted for age and sex.


Pooled cohort data included 12,422 participants with stroke (HRS: n=3164; SHARE: n=8179; ELSA: n=1079). Mean age (sd): 71.9 (10.8), 50.4% female. Frailty phenotype: robust 1925 (28.8%); pre-frail 3406 (50.9%); frail 1355 (20.3%). Frailty index without cognitive variables: robust 1454 (13.7%), mild 3049 (28.7%), moderate 2352 (22.2%), severe 3758 (35.4%). Frailty index with cognitive variables: robust 1447 (14.6%); mild 3107 (31.3%); moderate 2064 (20.8%); severe 3320 (33.4%). When measuring frailty with the frailty index (without cognitive measures), stroke survivors with moderate frailty had nearly triple the mortality risk compared to those who were robust (RR 2.80; CI 1.9-4.2) and stroke survivors with severe frailty had six times the mortality risk (6.2;4.4-9.1). Compared to robust stroke survivors, risk of hospital admission was double in the moderate frailty group (2.3;1.9-2.9) and triple in the severe frailty group (3.1; 2.52-3.84). There were no associations found between frailty and recurrent stroke, even when comparing those with severe frailty to those who were robust (0.9; 0.66-1.24). Results for health-related outcomes were similar when using all three methods of frailty measurement.


Use of the frailty index found a higher prevalence of post-stroke frailty than the frailty phenotype, however adding cognitive variables to the frailty index did not alter findings. All measurements gave similar results when examining frailty and health-related outcomes. Building an international consensus and harmonising measures across registries/trials would facilitate future comparative research.

Submitted by: 
Katie Gallacher
Funding acknowledgement: 
The Stroke Association TSA LECT 2017/01, MRC Clinical Research Training Fellowship