Using Item Response Theory Models in Scaling Severity Scores in Alcohol Research-A Tutorial

Area of Science:

• Psychometrics
• Quantitative Psychology
• Statistical Modeling

Background:

• Questionnaires are standard tools for assessing drug and alcohol dependence.
• Traditional sum scores often lack psychometric validation, leading to unclear reliability and sufficiency.
• Sum scores do not weight items by severity or account for measurement error, potentially underestimating true effects.

Purpose of the Study:

• To introduce item response theory (IRT) models for analyzing categorical questionnaire data in psychometrics.
• To explain IRT concepts like model parameters, fit, and group comparisons in accessible language.
• To demonstrate how IRT models can evaluate assumptions underlying traditional sum scores.

Main Methods:

• Application of item response theory models for categorical data analysis.
• Explanation of common IRT models, their parameters, and model fit assessment.
• Utilizing real-world data from a Danish alcohol and drug consumption survey.

Main Results:

• Item response theory models provide a robust alternative to sum scores for questionnaire analysis.
• These models account for measurement error and allow for differential item weighting.
• IRT enables more reliable estimation of underlying constructs and accurate group comparisons.

Conclusions:

• Item response theory models offer superior psychometric properties for assessing drug and alcohol dependence compared to sum scores.
• The study provides practical guidance and reproducible code (STATA, R) for applying IRT.
• Recommendations for reporting guidelines, pitfalls, and limitations in IRT analysis are discussed.