Computer-adaptive balance testing improves discrimination between community-dwelling elderly fallers and nonfallers.

Area of Science:

• Gerontology
• Rehabilitation Science
• Psychometrics

Background:

• Traditional balance assessments like the Berg Balance Scale (BBS), Performance-Oriented Mobility Assessment (POMA), and dynamic gait index (DGI) are fixed-form and may have limitations.
• Developing a more efficient and accurate balance assessment tool is crucial for evaluating older adults.

Purpose of the Study:

• To develop an item response theory-based computer adaptive test (CAT) for balance assessment.
• To compare the psychometric performance of the developed balance CAT against individual traditional balance measures.

Main Methods:

• Secondary analysis of two existing datasets from community-dwelling older adults (N=187, age ≥65).
• Items from BBS, POMA, and DGI were used to create an initial 38-item bank, calibrated using the Rasch partial credit model.
• Computer adaptive test (CAT) simulations were performed to select an optimal 10-item CAT, evaluating its accuracy, reliability, and effects.

Main Results:

• A 23-item bank was finalized, leading to a 10-item CAT with high association (r=.97) and good reliability (.78).
• The CAT demonstrated better reliability in lower to mid-balance ranges and a significantly lower ceiling effect (11.2%) compared to POMA (40.1%) and DGI (40.3%).
• The balance CAT was the only measure to significantly differentiate between fallers and non-fallers (P=.007) and was the strongest predictor of self-reported function.

Conclusions:

• The developed balance CAT exhibits excellent accuracy, reliability, and validity, outperforming traditional fixed-form measures.
• The CAT's ability to distinguish between fallers and non-fallers highlights its clinical utility for fall risk assessment.
• Further prospective studies in at-risk populations and development of an expanded item bank are recommended.