Mapping the mRS Scores Into the EQ-5D-5L in Patients With Ischemic Stroke.

Abstract

Mapping concerns the cross-walking of one health-related quality of life instrument to another. This study aimed to translate the modified Rankin Scale (mRS) score at 3 months after ischemic stroke into a 5-level version of the EQ-5D-based utility values, which may be used for the derivation of quality-adjusted life years for cost-effectiveness analyses. Data from 3 Dutch multicenter phase III clinical trials in ischemic stroke were pooled (MR CLEAN-NO-IV, MR CLEAN-MED, and MR CLEAN-LATE). The mRS score and 5-level version of the EQ-5D were assessed by telephone interview at 3 months poststroke. The correlation (Spearman coefficient) between the mRS score and each 5-level version of the EQ-5D domain was assessed. Various direct and indirect mapping algorithms were explored. Model performance (predictive ability) was assessed using bootstrapping in the full testing set (mRS score 0-5) and in patients with independent (0-2) and dependent (3-5) mRS scores separately. Results were validated in an external validation set derived from a cross-sectional Dutch study. Final utility estimates per mRS score were assessed in the pooled trial and external validation data set. The 5-level version of the EQ-5D dimensions self-care, daily activities, and mobility demonstrated the strongest correlation to the mRS score (<i>ρ</i>=0.46-0.70). Correlation coefficients of the pain and anxiety dimensions were relatively poor (<i>ρ</i>=0.37-0.45). Indirect mapping with a multinomial logit model was identified as the preferred mapping algorithm. In the pooled data set, the median (interquartile range) age was 71 (62-79), and 56% of patients were male. Mean (SD) fitted utility values per mRS score were mRS score 0: 0.947 (0.004); mRS score 1: 0.860 (0.009); mRS score 2: 0.758 (0.013); mRS score 3: 0.597 (0.017); mRS score 4: 0.352 (0.017); and mRS score 5: 0.147 (0.017). The mean utility weights provided in this study may be used for the direct estimation of quality-adjusted life years from mRS scores. Furthermore, the multinomial logit model may be used to derive utility values from new data sets with different country-specific utility tariffs. Our results are ideally suited for application in cost-effectiveness analyses.