James Nazroo Alan Marshall, Kris Mekli and Neil Pendleton neil.pendleton@manchester.ac.uk

1. Measuring FrailtyA comparison of the frailty phenotype and frailty index for the prediction of all-cause mortality James Nazroo Alan Marshall, Kris Mekli and Neil Pendleton neil.pendleton@manchester.ac.uk

2. Background • Specific definitions and models of frailty are contested • Broad agreement that frailty is a non-specific state reflecting age-related declines in multiple systems, which lead to adverse outcomes (mortality, hospitalisation) • Two common approaches to characterise frailty: • Frailty index (Rockwood and colleagues) • Frailty phenotype (Fried and colleagues) • Compare the two approaches and, in particular, their success in predicting all cause mortality • And how well do they perform compared with other markers of risk?

3. The English Longitudinal Study of Ageing(www.ifs.org.uk/elsa) • A panel study of people aged 50 and older, recently finished our sixth wave of data collection, with additional wave 0 data available • Sample at wave 1 (2002) was approximately 11,400 people born before 1st March 1952 who were in the private household sector. Drawn from Health Survey for England (wave 0). • Face to face interview every two years since 2002, with a biomedical assessment carried out by a nurse every four years. • Those incapable of doing the interview have a proxy interview. • End of life interviews are carried out with the partners or carers of people who died after wave 1. • Detailed content on: demographics, health, performance, biomarkers, wellbeing, economics, housing, employment, social relationships, social civic and cultural participation, life history. • Sister study to HRS, SHARE, KLOSA, CHARLS, etc.

4. Frailty Index (FI) • Based on accumulation of ‘deficits’ (from 30 items) • Activities of Daily Living • Cognitive function • Chronic diseases • CVD • Depression/mental health • Poor eyesight/hearing • Falls, fractures and joint replacements • 0-1 scale for each component • Calculate the proportion of deficits held (so 0-1 scale) • Can be divided into three categories • Robust (0-0.12) • Pre-frail (0.13-0.21) • Frail (>0.21)

5. Frailty Phenotype (FP): Fried et al. 2001 • Aim: to establish a standardized definition of frailty • 5 items: • Sarcopenia: unintentional weight loss, > 8% bodyweight • Exhaustion: had both ‘everything they did was an effort’ and ‘could not get going much of the time’, during the past week • Low physical activity: no work and no other physical activities • Slowness: timed walk in the slowest 20% of population • Weakness: grip strength in the weakest 20% of population • Outcome • Robust phenotype: positive for 0 item • Pre-frail phenotype: positive for 1-2 items • Frail phenotype: positive for 3-5 items

6. Analysis • Those aged sixty or older (walking speed) • Cumulative distribution of Frailty Index score for each Frailty Phenotype category (‘robust’, ‘pre-frail’ and ‘frail’) • Kaplan-Meier survival plots • Cox proportional hazard models • Test association between Frailty Index and Frailty Phenotype (at wave 2) and all cause mortality • Unadjusted model and then adjusted for demographics and then key risk factors (education, wealth, bmi, smoking) • Compare goodness of fit of models, and compare fit with models using self-assessed health and wealth

7. Frailty index: distribution (wave 1) 0.21 used as a cut-off for dichotomous frailty variable 30% frail in wave 1

8. Descriptives P value compares the survived and died groups (t-test and chi squared test)

9. Frailty Index cumulative distribution by Frailty Phenotype Females Males Robust Robust Cumulative distribution Pre-frail Pre-frail Frail Frail Frailty index Frailty index

10. Kaplan-Meier survival estimates - Males Frailty Index Frailty Phenotype Blue = Robust group. Red = Pre-frail group. Green = Frail group

11. Kaplan-Meier survival estimates - females Frailty Index Frailty Phenotype Blue = Robust group. Red = Pre-frail group. Green = Frail group

12. Cox survival model - results Model 1 controls for age and sex. Model 2 controls for age, sex, education, wealth, smoking, couple

13. Cox survival model: Health and wealth Model 1 controls for age and sex. Model 2 controls for age, sex, education, (wealth), smoking, couple

14. Conclusions • Prediction of mortality over an eight year period for older people (aged 60+) living in the community: • Comparable results for Frailty Index and Frailty Phenotype • Comparable results for self-assessed health and wealth • Choice of measure might reflect the particular setting • Frailty Phenotype advantageous in clinical setting: detailed longitudinal ‘diagnostic’ measure • Frailty index useful in community environment: checklist approach