Properties of Proxy-Derived Modified Rankin Scale Assessment

Abstract

Cognitive or communication issues may preclude direct modified Rankin Scale interview, necessitating interview with a suitable surrogate. The clinimetric properties of this proxy modified Rankin Scale assessment have not been described. To describe reliability of proxy-derived modified Rankin Scale and compare with traditional direct patient interview. Researchers assessed consenting stroke inpatients and their proxies using a nonstructured modified Rankin Scale approach. Paired interviewers (trained in modified Rankin Scale) performed independent and blinded modified Rankin Scale assessment of patients and appropriate proxies. Interobserver variability and agreement between patient and proxy modified Rankin Scale were described using kappa statistics (k, 95% confidence interval) and percentage agreement. Ninety-seven stroke survivors were assessed. Proxies were family members (n = 29), nurses (n = 50), or physiotherapists (n = 25). Median modified Rankin Scale from both patient and proxies was 3 [interquartile range (IQR): 2-4]. Reliability for patient modified Rankin Scale interview was weighted kappa = 0·70 (95% confidence interval: 0·30-1·00). Reliability for proxy modified Rankin Scale weighted kappa = 0·62 (95% confidence interval: 0·34-0·90). Subgroup analysis of various proxy information sources were as follows: family weighted kappa = 0·61; nurse weighted kappa = 0·58; therapist weighted kappa = 0·58. There was disagreement between patient-derived modified Rankin Scale and corresponding proxy modified Rankin Scale weighted kappa = 0·64 (95% CI: 0·42-0·86). There is potential for substantial interobserver variability in proxy modified Rankin Scale and validity of certain proxy assessments is questionable. Direct modified Rankin Scale interview is preferred.