A multidimensional item response theory analysis of the Seattle Angina Questionnaire: measurement precision, differential item functioning, and potential for item reduction.

Abstract

Purpose: The Seattle Angina Questionnaire (SAQ) is widely used to measure disease-specific health status in coronary artery disease, but its psychometric properties have not been evaluated using multidimensional item response theory (MIRT). We evaluated the SAQ's dimensional structure, measurement precision, differential item functioning (DIF), and short form performance in a population-based cohort.

Method: SAQ responses collected within two weeks of coronary angiography from 6,665 patients in the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) registry were analyzed using a confirmatory five-factor multidimensional graded response model. Item fit, local dependence, and conditional standard errors were evaluated. DIF by sex and age was examined using a two-step anchored multigroup approach. Short forms were derived based on discrimination, test information, and DIF magnitude.

Result: The confirmatory five-factor model demonstrated good global fit (RMSEA = 0.031, CFI = 0.987, TLI = 0.982), supporting three distinct Physical Limitation dimensions: Basic, Moderate, and Strenuous. Empirical reliability ranged from 0.748 (Basic Physical Limitation) to 0.900 (Moderate Physical Limitation), with precision sufficient for individual-level interpretation in Moderate and Strenuous Physical Limitation. Moderate sex-related DIF was noted on Basic and Moderate Physical Limitation items, with females scoring lower than males. Age-related DIF was negligible to small. The 14-item short form preserved all five factors with theta correlations of 0.964 with the 18-item version.

Conclusion: MIRT revealed three distinct Physical Limitation dimensions, enabling targeted functional monitoring. Precision gradients and sex-related DIF inform interpretation, and the 14-item short form is supported as the preferred version.