Agreement between standard and self-reported assessments of physical frailty syndrome and its components in a registry of community-dwelling older adults

Abstract

Abstract Background The ability to identify frail older adults using a self-reported version of the physical frailty phenotype (PFP) that has been validated with the standard PFP could facilitate physical frailty detection in clinical settings. Methods We collected data from volunteers (N = 182), ages 65 years and older, in an aging research registry in Baltimore, Maryland. Measurements included: standard PFP (walking speed, grip strength, weight loss, activity, exhaustion); and self-reported questions about walking and handgrip strength. We compared objectively-measured gait speed and grip strength to self-reported questions using Cohen’s Kappa and diagnostic accuracy tests. We used these measures to compare the standard PFP with self-reported versions of the PFP, focusing on a dichotomized identification of frail versus pre- or non-frail participants. Results Self-reported slowness had fair-to-moderate agreement (Kappa(k) = 0.34–0.56) with measured slowness; self-reported and objective weakness had slight-to-borderline-fair agreement (k = 0.10–0.21). Combining three self-reported slowness questions had highest sensitivity (81%) and negative predictive value (NPV; 91%). For weakness, three questions combined had highest sensitivity (72%), while all combinations had comparable NPV. Follow-up questions on level of difficulty led to minimal changes in agreement and decreased sensitivity. Substituting subjective for objective measures in our PFP model dichotomized by frail versus non/pre-frail, we found substantial (k = 0.76–0.78) agreement between standard and self-reported PFPs. We found highest sensitivity (86.4%) and NPV (98.7%) when comparing the dichotomized standard PFP to a self-reported version combining all slowness and weakness questions. Substitutions in a three-level model (frail, vs pre-frail, vs. non-frail) resulted in fair-to-moderate agreement (k = 0.33–0.50) with the standard PFP. Conclusions Our results show potential utility as well as challenges of using certain self-reported questions in a modified frailty phenotype. A self-reported PFP with high agreement to the standard phenotype could be a valuable frailty screening assessment in clinical settings.