Item Response Theory Approaches to Harmonization and Research Synthesis

Area of Science:

• Health Economics
• Psychometrics
• Biostatistics

Background:

• Harmonizing diverse health outcome metrics is crucial for research synthesis and economic evaluations.
• Existing methods often struggle to equate scales measuring the same underlying construct.
• Item-level data harmonization is needed for comprehensive health technology assessments.

Purpose of the Study:

• To describe the application of multidimensional item response theory (IRT) for equating different measurement scales at the item level.
• To demonstrate how multidimensional IRT can predict responses for unadministered scales, facilitating economic evaluations.
• To present a general framework for harmonizing research instruments using multidimensional IRT.

Main Methods:

• Overview of multidimensional item response theory (IRT), focusing on the bi-factor model.
• Equating underlying true scores of multiple scales measuring the same latent variable.
• Predicting item responses from one scale to another using multidimensional IRT.

Main Results:

• Multidimensional IRT accurately predicted EQ-5D descriptive system and preference index scores from SF-12 data.
• This approach enabled economic evaluations using data not directly suitable for such analyses.
• Multidimensional IRT outperformed traditional regression methods in harmonizing research instruments.

Conclusions:

• Multidimensional IRT provides a robust framework for harmonizing health outcome metrics at the item level.
• This methodology enhances the utility of existing datasets for economic evaluations and research synthesis.
• The described approach offers a valuable tool for improving the comparability of health intervention research.