Measuring Patient-Reported Outcomes Adaptively: Multidimensionality Matters!

Area of Science:

• Psychometrics
• Health Measurement
• Psychological Assessment

Background:

• Increasing use of unidimensional (UCAT) and multidimensional computerized adaptive testing (MCAT) in psychological and health measurement.
• Need to evaluate the added value of MCAT over multiple UCATs for assessing multiple dimensions.

Purpose of the Study:

• To assess the incremental value and efficiency of using MCAT versus separate UCATs for each dimension in health measurement.

Main Methods:

• Simulations based on empirical health measurement data with 4 dimensions and strong inter-correlations (.77-.87).
• Utilized the multidimensional graded response model (GRM) for item parameter and variance-covariance matrix estimation.
• Evaluated both variable- and fixed-length test conditions, analyzing CAT length, RMSE, and bias.

Main Results:

• MCAT demonstrated superior efficiency (shorter test length) and accuracy (lower RMSE) compared to UCATs.
• Variable-length UCATs with an SE < .387 termination rule exhibited the highest absolute average bias.
• Variable-length MCATs achieved test lengths 20-25% shorter than aggregated UCATs.

Conclusions:

• MCAT provides clear advantages in efficiency and accuracy for health measurement settings.
• MCAT is a more effective approach than employing separate UCATs for multidimensional assessments.