Outcome-specific thresholds and clinically relevant changes in frailty: evidence from a population-based longitudinal cohort

Abstract

Background: The usefulness of assessing Frailty Phenotype is hampered by the limitations of classical tools in defining outcome-specific risk trajectories, stablishing its severity, monitoring longitudinal changes and quantifying the changes needed to modify the risk. Objective: To assess the risk along the score range, defining outcome-specific thresholds, slopes of risk and degrees of severity and to determine Minimal Clinically Important Differences (MCIDs) through numerical-score based tools: Frailty Trait Scale (FTS). Design, setting and subjects: Prospective community-dwelling cohort of people ≥65 years. Methods: Frailty was determined at baseline (Wave 1) and after 5.02 years (Wave 2) using the FTS (n = 1811) and its shorter 5-items form (FTS-5; n = 1597). Primary outcomes assessed at Wave 2 were worsening disability (median follow-up 5.02 years), hospitalisation (3.3) and mortality (5.4). For MCIDs analyses, time-to-event outcomes assessed at Wave 3 were 2.99, 4.18 and 6.77 years, respectively. Cox proportional hazards models and logistic regression were used.

Results: A continuous dose-response association was observed between frailty scores and all outcomes, with distinct thresholds, slopes and saturation points. Threshold for disability risk was 25/100 FTS (or 10/50 for FTS5); 40/100 (15/50 FTS5) for hospitalisation and 50/100 (25/50 FTS5) for mortality. The sequence and intensity of these risks allowed to define five different categories of severity. MCID analyses showed increases in frailty scores associated with higher risk of all outcomes, whereas reductions protected against hospitalisation [FTS ≥ 10.3: HR 0.74 (95% CI, 0.56-0.99); FTS5 ≥ 6.5: 0.74 (0.55-0.996)] and mortality [FTS ≥ 4.5: HR 0.65 (95% CI, 0.45-0.95); FTS5 ≥ 1.0: 0.65 (0.45-0.93)]. Conclusions: Both FTS and FTS5 show characteristics that make them useful in clinical settings, overcoming some of the pitfalls of classical tools.