Predictive Validity of Two Physical Frailty Phenotype Specifications Developed for Investigation of Frailty Pathways in Older People

Background: For investigating the relationship of frailty with physical, psychological, and social conditions on pathways, the frailty definition should ideally exclude these conditions. Based on the frailty phenotype, 2 candidate physical frailty specifications or instruments with 3 indicators, namely slowness, weakness, and exhaustion, and 4 indicators with addition of weight loss were previously developed, and had their construct and concurrent validity demonstrated. Objective: This study seeks to evaluate the predictive validity of 2 candidate physical frailty specifications with respect to death, functional impairment, and poor quality of life in older people. Methods: For 4,368 respondents aged 65-89 years from wave 2 of the English Longitudinal Study of Ageing, confirmatory factor analysis is performed for these 2 physical frailty specifications to obtain unique factor scores for each respondent. Prediction of death, basic and instrumental activities of daily living (BADL and IADL) difficulty, and poor quality of life (reverse of Control, Autonomy, Self-realization, and Pleasure [19 items] or CASP-19) 2 years later by factor scores for these 2 specifications is evaluated using standardized coefficients, c-statistics, and r2 values from regression analyses. Their performance is compared with those of alternative specifications with 3 (slowness, weakness, and weight loss) and 5 indicators (slowness, weakness, exhaustion, weight loss, and low physical activity), and Frailty Index (FI). Results: For the 2 candidate specifications, an increase of 1 standard deviation (SD) predicts 50-57% increase in odds of death, 0.10-15 SD increase in change in BADL or IADL difficulty, and poor quality of life at 2 years. They predict these outcomes as well or better than the alternative specification with 3 indicators, but marginally worse than that with 5 indicators. Compared with FI, they predict death and poor quality of life similarly, but perform worse for functional impairment. Minor differences are observed across gender. Conclusion: Reasonable predictive validity of 2 candidate physical frailty specifications based on the frailty phenotype with 3 and 4 indicators is demonstrated for death, functional impairment, and poor quality of life. These findings offer evidence to support their suitability for employment in investigating frailty pathways in older people.