Clinical Utility of Patient-Reported Outcome Measurement Information System Domain Scales

Abstract

Background: Patient-reported outcome measures are increasingly being utilized in clinical care and research to evaluate outcomes following stroke. To optimize the clinical utility of these measures, we aimed to quantify meaningful change by establishing minimal important differences (MIDs), or responder definitions, for 4 domains affected in ischemic and hemorrhagic stroke patients.

Methods and results: We performed a retrospective cohort study of stroke patients seen in the Cleveland Clinic cerebrovascular center between September 2, 2012 and November 7, 2017. Four Patient-Reported Outcome Measurement Information System (PROMIS) scales were completed within 1 month poststroke and again at 6 months. MIDs were estimated using an anchor-based approach based on a global impression of change question and supported using 3 distribution-based methods. Cumulative distribution functions assessed responder thresholds. MIDs were additionally derived across sex, race, and varying levels of severity as defined by the modified Rankin Score and baseline PROMIS score. During the study period, 337 incident stroke patients completed at least 1 PROMIS domain scale at both time points (average age 61±14, 56% female). Estimates from the 4 methods were triangulated to provide a MID range across PROMIS domain: 2.5 to 6.5 T-score points for physical function and fatigue, 2.5 to 7.5 for social role satisfaction, and 3.0 to 8.0 for anxiety. Cumulative distribution functions plots identified between 30% and 40% of patients as having meaningful improvement based on the anchor-based estimates across all 4 domains. Meaningful change thresholds remained consistent across categories of sex and race. Anchor-based MIDs increased with increasing severity, whereas distribution-based MIDs remained consistent across severity levels.

Conclusions: Our study is the first to evaluate interpretability of changes in PROMIS scores for stroke survivors. Future studies can utilize these thresholds to identify responders of stroke interventions. Based on our estimated MID ranges, researchers and clinicians can choose a responder threshold for comparing change in domain score at the group level, individual level, or by severity.

Keywords: anxiety; female; patient-reported outcome measures; research; stroke.