Prestroke Modified Rankin Stroke Scale Has Moderate Interobserver Reliability and Validity in an Acute Stroke Setting

Abstract

Background and Purpose— The modified Rankin Scale (mRS) is the recommended functional outcome assessment in stroke trials. Utility of mRS may be limited by interobserver variability. prestroke function, described using mRS, is often used as trial entry criterion. We assessed the reliability and validity of prestroke mRS in acute stroke. Methods— We present two complementary analyses of the properties of prestroke mRS: (1) Paired interviewers (trained in mRS) performed independently a blinded assessment of mRS and prestroke mRS. Interobserver variability was described using percentage agreement and weighted (kw) κ statistics with 95% confidence interval (95% CI). Validity was assessed by comparing prestroke mRS with other markers of function (comorbidity; medication count; need for carers). (2) We further assessed validity using a larger retrospective dataset. We compared prestroke mRS with Charlson comorbidity index (CCI) and the Rockwood frailty index. Rank correlation coefficient or Fisher exact test were used as appropriate. Results— Paired interviewers assessed 74 stroke survivors. Median standard mRS was 4 (interquartile range [IQR], 2–4), median prestroke mRS was 1 (IQR, 0–3; range, 0–4). Reliability for standard mRS interview was 56% agreement, kw=0.55 (95% CI, 0.39–0.71). Reliability for prestroke mRS was 70%, kw=0.70 (95% CI, 0.53–0.87). The retrospective dataset described 231 subjects. In this data set, Spearman Rho for prestroke mRS and frailty index was J. 0.82 (95% CI, 0.78–0.86); CCI 0.50 (95% CI, 0.40–0.59); patient age 0.45 (95% CI, 0.34–0.54); medication count 0.28 (95% CI, 0.15–0.40). There was no association between need for carers and prestroke mRS (p=0.10). Conclusions— Interobserver reliability of prestroke mRS is limited but comparable with standard mRS. Poor correlation of prestroke mRS with certain markers of function suggests limited validity. Our data would suggest that relying on mRS alone may be a suboptimal measure of prestroke function and could potentially bias trial samples.