Substitution of self-reported measures for objectively assessed grip strength and slow walk in the Physical Frailty Phenotype: ramifications for validity

Abstract

Background: Frailty assessment promises to identify older adults at risk for adverse consequences following stressors and target interventions to improve health outcomes. The Physical Frailty Phenotype (PFP) is a widely-studied, well validated assessment but incorporates performance-based slow walk and grip strength criteria that challenge its use in some clinical settings. Variants replacing performance-based elements with self-reported proxies have been proposed. Our study evaluated whether commonly available disability self-reports could be substituted for the performance-based criteria in the PFP while still identifying as "frail" the same subpopulations of individuals.

Methods: Parallel analyses were conducted in 3393 female and 2495 male Cardiovascular Health Study, Round 2 participants assessed in 1989-90. Candidate self-reported proxies for the phenotype's "slowness" and "weakness" criteria were evaluated for comparable prevalence and agreement by mode of measurement. For best-performing candidates: Frailty status (3 + positive criteria out of 5) was compared for prevalence and agreement between the PFP and mostly self-reported versions. Personal characteristics were compared between those adjudicated as frail by (a) only a self-reported version; (b) only the PFP; (c) both, using bivariable analyses and multinomial logistic regression.

Results: Self-reported difficulty walking ½ mile was selected as a proxy for the phenotype's slowness criterion. Two self-reported weakness proxies were examined: difficulty transferring from a bed or chair or gripping with hands, and difficulty as just defined or in lifting a 10-pound bag. Prevalences matched to within 4% between self-reported and performance-based criteria in the whole sample, but in all cases the self-reported prevalence for women exceeded that for men by 11% or more. Cross-modal agreement was moderate, with by-criterion and frailty-wide Kappa statistics of 0.55-0.60 in all cases. Frail subgroups (a), (b), (c) were independently discriminated (p < 0.05) by race, BMI, and depression in women; by age in men; and by self-reported health for both.

Conclusions: Commonly used self-reported disability items cannot be assumed to stand in for performance-based criteria in the PFP. We found subpopulations identified as frail by resultant phenotypes versus the original phenotype to systematically differ. Work to develop self-reported proxies that more closely replicate their objective phenotypic counterparts than standard disability self-reports is needed.

Keywords: Aging; Construct validation; Diagnostic accuracy; Measurement error; Physical function; Vulnerability.