Validity and reliability of a performance evaluation tool based on the modified Barthel Index for stroke patients

BACKGROUND: The Barthel Index (BI) is a measure of independence in activities of daily living (ADL). In the modified Barthel Index (MBI), a five-point system replaced the original two or three or four point rating system. Based on this modified measure, the performance evaluation tool MBI (PET-MBI) was developed in Japan. Although the reliability and validity of PET-MBI have been verified for older people, the use of this tool in stroke patients has not been evaluated. This study investigated the validity and reliability of PET-MBI for stroke patients. METHODS: Ten raters independently determined the BI and PET-MBI scores of stroke patients by direct observation. These patients’ ADL were videotaped, and 10 other raters then evaluated the videos privately and assigned PET-MBI scores twice, one month apart. The criterion-related validity of the PET-MBI against the BI was evaluated using the correlation coefficients for their total scores. Furthermore, to assess inter- and intra-rater reliabilities from the results of the first and second sessions, Fleiss’ intraclass correlation coefficients (ICCs) were calculated for the total scores, with the lower limits of the 95% confidence interval (95%CI), along with weighted kappa (κ(w)) coefficients for agreement in individual tasks of this evaluation tool. ICC and κ(w) coefficients of 0.81–1.00 were considered to be “almost perfect” agreement. RESULTS: The mean age of the 30 patients (23 men, 7 women) was 71.9 (standard deviation 10.5) years. One patient had diplegia, 14 had right hemiplegia, and 15 had left hemiplegia. For the total scores obtained by direct evaluation, Pearson’s and Spearman’s correlation coefficients of the BI versus the PET-MBI were both 0.95 (lower limit of the 95%CI, 0.90). The ICC representing inter-rater reliability for the first session was 0.99 (lower limit of the 95%CI, 0.98]. For intra-rater reliability, the mean value of the ICCs was 0.99 (range, 0.99–1.00). For individual tasks of the PET-MBI, inter-rater κ(w) coefficients for the first session ranged from 0.77 to 0.94, with intra-rater κ(w) coefficients from 0.85 to 0.96. CONCLUSIONS: PET-MBI showed strong criterion-related validity against the BI, with high reliabilities. This scoring system may become a convenient tool allowing anyone to assess ADL.