Factors influencing rate of Barthel Index change in hospital following stroke

Background and purpose: Randomized controlled trials of specific interventi ons in stroke rehabilitation are few. In using this study design to compare different rehabilitation interventions or different intensity of those int erventions, measuring disability status at a fixed time point is one outcom e option. In order for this approach to be valid, factors that might indepe ndently affect the speed of disability change have to be matched at baselin e. We sought to investigate the impact of different factors on rate of disa bility change following stroke. Methods: A prospective hospital-based study of consecutive patients admitte d to each of three general hospitals in Wellington, New Zealand with acute stroke. Patients were assessed using the Barthel Index (BI, scored 0-20) wi thin a few days of stroke and fortnightly until hospital discharge, Strokes were classified using the Oxfordshire Community Stroke Project classificat ion. Main outcome measure: Barthel Index rate of change (BRC) computed from the difference between the first and last hospital BI scores divided by the tim e in weeks between these assessments. Results: Of 104 subjects with two or more disability assessments in hospita l, BRC was a mean 1.26 units per week (95% confidence interval (CI) 0.97, 1 .56). The only factors significantly associated with BRC on univariate anal ysis were stroke type, pre-stroke Modified Rankin Scale (MRS) and pre-strok e London Handicap Score. In a general linear model analysis of covariance, stroke type (p=0.015) and pre-stroke MRS (p<0.001) remained significant. Th e mean BRC for lacunar infarcts (1.76 units per week, 95% CI 1.22, 2.30) wa s over three times that of total anterior circulation infarcts (0.48 units per week, 95% CI 0.20, 0.76). Conclusions: Future randomized controlled trials of rehabilitation interven tions for stroke in hospitalized patients need to allow for pre-stroke leve l of dependence and stroke type in the study design, either excluding parti cular patients or ensuring even randomization of these key variables if spe ed of disability change is to be a valid outcome variable.