Computerized Adaptive Testing for the Berg Balance Scale Improves Measurement Efficiency Without Compromising...

Area of Science:

• Rehabilitation Medicine
• Biostatistics
• Neuroscience

Background:

• The Berg Balance Scale (BBS) is a common tool for assessing balance in stroke patients.
• Evaluating the psychometric properties of the BBS and the efficiency of its computerized adaptive testing (CAT) version is crucial for clinical practice.

Purpose of the Study:

• To confirm the measurement properties and unidimensionality of the BBS using item response theory (IRT) in individuals with subacute and chronic stroke.
• To assess the precision and efficiency of the BBS computerized adaptive testing (CAT) for balance assessment.

Main Methods:

• A principal component analysis (PCA) of residuals and Rasch analysis were used to evaluate BBS unidimensionality and model fit.
• Item parameters were generated for CAT calibration, and simulations examined measurement precision with varying item numbers and a precision-based stopping rule (0.5 logit SE).
• Data from 519 ambulatory individuals with subacute and chronic stroke were analyzed.

Main Results:

• The BBS demonstrated sufficient unidimensionality, supporting the use of a rating scale model for measurement.
• BBS CAT achieved high correlation with actual balance ability (r=0.952) using a precision-based stopping rule, administering an average of 5.43 items.
• Measurement precision (SE < 0.48 logits) was achieved with 5 or more items, and maximum precision (SE=0.40 logits) with all 14 items.

Conclusions:

• The BBS is a valid and unidimensional measure suitable for the rating scale model.
• BBS CAT is an efficient and reliable method for assessing balance in ambulatory stroke survivors, comparable to the full BBS.
• Further research should focus on enhancing the interpretability of BBS CAT measures for clinical decision-making.