Item response theory and structural equation modelling for ordinal data: Describing the relationship between...

Area of Science:

• Health outcomes research
• Psychometrics
• Biostatistics

Background:

• Health assessment questionnaires often yield ordered categorical data.
• Analyzing such data requires specialized statistical approaches.
• Community health settings present unique challenges for data analysis.

Purpose of the Study:

• To compare item response theory (IRT) and structural equation modeling (SEM) for ordinal health data.
• To investigate the relationship between quality-of-life (KIDSCREEN) and participation (Life-H) in children with cerebral palsy.
• To discuss practical issues in fitting and assessing model fit for ordinal data.

Main Methods:

• Application of both IRT and SEM to ordinal rating scales.
• Analysis of data from the SPARCLE project, focusing on children with cerebral palsy.
• Exploration of model fitting challenges, including correlation matrices and model fit assessment.

Main Results:

• IRT models facilitate assessment of measurement instrument properties like conditional independence.
• SEM provides a more convenient framework for analyzing multidimensional latent traits, as seen in the SPARCLE data.
• The study highlights the utility of both methods while favoring SEM for complex, multidimensional health data.

Conclusions:

• Both IRT and SEM are valuable tools for analyzing ordinal health data.
• SEM offers advantages for complex, multidimensional constructs common in pediatric health research.
• Careful consideration of practical modeling issues is crucial for robust analysis of health assessment data.