Optimizing Oxford Shoulder Scores with computerized adaptive testing reduces redundancy while maintaining precision

Area of Science:

• Orthopedics
• Medical Informatics
• Psychometrics

Background:

• The Oxford Shoulder Score (OSS) is a standard 12-item questionnaire for assessing shoulder surgery outcomes.
• Patient-reported outcome measures (PROMs) are crucial for evaluating surgical effectiveness.
• There is a need for more efficient and tailored assessment tools to reduce patient burden.

Purpose of the Study:

• To evaluate the feasibility and accuracy of Computerized Adaptive Testing (CAT) for the Oxford Shoulder Score (OSS).
• To determine if CAT can provide a shortened, individually tailored questionnaire while maintaining the reliability of the full OSS.
• To assess the impact of CAT on patient compliance and reduce assessment time.

Main Methods:

• Utilized a large dataset of 16,238 preoperative OSS from the National Joint Registry (NJR).
• Established Item Response Theory (IRT) assumptions: unidimensionality, monotonicity, and local independence.
• Simulated CAT with stopping criteria at standard error (SE) < 0.32 and < 0.45, comparing results to the full 12-item OSS.

Main Results:

• Confirmatory factor analysis indicated satisfactory unidimensionality (RSMSR=0.06) but mixed fit for other indices (CFI=0.85, TLI=0.82).
• Monotonicity (H=0.482) and local independence were generally supported.
• CAT simulations required a median of 2-3 items (16-25% of the full questionnaire) to achieve target reliability levels.

Conclusions:

• Item Response Theory (IRT) calibration enables an efficient, shortened CAT version of the OSS.
• CAT maintains the accuracy and reliability of the full OSS, offering a promising approach for outcome assessment.
• This shortened CAT method can alleviate patient time burden and potentially enhance compliance with shoulder surgery outcome measures.