Reliability, Validity, and Responsiveness of a 3-Item Barthel for Characterizing the Physical Function of Patients Hospitalized for Acute Stroke

Background and Purpose: The analysis of outcomes requires that therapists possess measures that are reliable, valid, and responsive. To be employed in acute care, the measures also need to be practicable. The purpose of this study was to determine if a 3-item subset of the Barthel Index (BI[3]) had these characteristics when applied to patients with stroke. Methods: Patients hospitalized for ischemic stroke (n = 279) were interviewed regarding their prestroke and poststroke function using the 10-item Barthel Index (BI[10]). Cronbach's alpha was used to examine the internal consistency of the 3 items (bed-chair transfers, mobility, and bladder incontinence) of the BI(3). Spearman correlations (rs) were employed to describe the concurrent validity of the BI(3) relative to the BI(10). Spearman correlations and multiple regression were to characterize the predictive validity of the BI(3) for 3 outcome measures. Effect size (for prestroke to poststroke changes) was used as an indicator of responsiveness. Results: The items of the BI(3) were reliable (alpha = .83) and correlated highly (rs = .95) with the BI(10).The BI(3) and BI(10) correlated significantly with hospital length of stay (-.35 & -.35), discharge home (.59 & .63), and 3-month BI(10) (.46 & .50). Linear regression corroborated these findings but showed that age added to the explanation of length of stay. The correlations with these outcomes were not significantly better for the BI(10) than for the BI(3). Age added to the prediction of length of stay provided by admission BI(3) and BI(10) scores. The effect sizes were 2.79 and 3.19 for the BI(3) and BI(10), respectively. Conclusions: The statistical properties of the BI(3) along with its brevity support its use for characterizing physical functioning in patients hospitalized for acute ischemic stroke.