Measurement Properties of the Barthel Index in Geriatric Rehabilitation

Objective: The Barthel index (BI) is a widely used observer-based instrument to measure physical function. Our objective is to assess the structural validity, reliability, and interpretability of the BI in the geriatric rehabilitation setting.

Design: Two studies were performed. First, a prospective cohort study was performed in which the attending nurses completed the BI at admittance and discharge (n = 207). At discharge, patients rated their change in physical function on a 5-point Likert rating scale. To assess the internal structure of the BI, a confirmatory factor analysis was performed. Unidimensionality was defined by comparative fit index and Tucker-Lewis index of >0.95, and root mean square error of approximation of <0.06. To evaluate interpretability, floor/ceiling effects and the minimal important change (MIC) were assessed. Predictive modeling was used to calculate the MIC. The MIC was defined as going home and minimal patient-reported improvement defined as slightly or much improved physical function, which served as anchors to obtain a clinical- and patient-based MIC. A second group of 37 geriatric rehabilitation patients were repeatedly assessed by 2 attending nurses to assess reliability of the BI. The intraclass correlation coefficient, standard error of measurement, and smallest detectable change were calculated.

Setting and participants: Patients receiving inpatient geriatric rehabilitation admitted to 11 Dutch nursing homes (n = 244).

Results: Confirmatory factor analysis showed partly acceptable fit of a unidimensional model (comparative fit index 0.96, Tucker-Lewis index 0.95, and root mean square error of approximation 0.12). The clinical-based MIC was 3.1 [95% confidence interval (CI) 2.0-4.2] and the patient-based MIC was 3.6 (95% CI 2.8-4.3). The intraclass correlation coefficient was 0.96 (95% CI 0.93-0.98). The standard error of measurement and smallest detectable change were 1.1 and 3.0 points, respectively.

Conclusions/implications: The structural validity, reliability, and interpretability of the BI are considered sufficient for measuring and interpreting changes in physical function of geriatric rehabilitation patients.