Computerized Adaptive Testing for the Berg Balance Scale Improves Measurement Efficiency Without Compromising Precision in People With Stroke.

Abstract

Objective: Objectives of this study were to confirm the Berg Balance Scale's (BBS) measurement properties and unidimensionality with an item response theory analysis in persons with subacute and chronic stroke and examine the precision and efficiency of computerized adaptive testing (CAT).

Methods: Data were obtained from 519 ambulatory persons with subacute and chronic stroke in 2 retrospective databases. A principal component analysis (PCA) of residuals was used to evaluate unidimensionality. BBS fit to a rating scale model versus a partial credit model was examined and item parameters were generated for CAT calibration. Person measures from all 14 items were defined as actual balance ability. BBS CAT simulations were used to examine changes in measurement precision with increasing number of items administered and a precision-based stopping rule (0.5 logit standard error (SE) threshold).

Results: A PCA of residuals supports the BBS's unidimensionality and Rasch analysis supports using the rating scale model for measurement. Maximum precision for BBS CAT was SE = 0.40 logits when administering all items. BBS CAT estimated balance ability was highly correlated with actual ability when 4 or more items were administered (r > 0.9). Precision was within 0.5 logits when 5 or more items were administered (SE < 0.48 logits). BBS CAT estimated balance ability was highly correlated with actual ability (r = 0.952) using a precision-based stopping rule. The average number of items administered with the precision-based stopping rule was 5.43.

Conclusion: The BBS is sufficiently unidimensional and the rating scale model can be used for measurement. BBS CAT is efficient and replicates the full instrument's reliability when measuring balance ability in ambulatory persons with subacute and chronic stroke. Future work should aim to enhance the interpretability of measures to facilitate clinical decision making.

Impact statement: BBS CAT provides an efficient way of measuring balance ability for individuals in stroke rehabilitation giving clinicians more time with patients.