Are there subgroup differences in the accuracy of 'screening' questions for mood and anxiety disorder diagnostic interviews?

OBJECTIVE

To examine the impact of potential measurement bias (i.e., differential item functioning [DIF]) across sex, age, employment, location, and substance use disorders on the screening properties of epidemiological surveys that utilise screening questions when estimating prevalence of mood and anxiety disorders.

METHODS

Data comprised of 15,893 respondents who completed the 2020-2022 Australian National Survey of Mental Health and Wellbeing. Questions from the screening module of the Composite International Diagnostic Interview 3.0 were analysed using confirmatory factor analysis and DIF across subgroups of interest. Sensitivity, specificity, and classification rate were derived and compared across models that did and did not adjust for significant levels of DIF.

RESULTS

Sources of DIF were identified across the items was due to age and sex at birth with relatively fewer items displaying DIF across employment, location, and substance use disorders. In terms of screening, the absolute differences in sensitivity and specificity between the DIF-free and DIF models ranged from 0.001 to 0.091.

CONCLUSIONS

The current study found some evidence of DIF in the screening questions used to evaluate mental health disorder prevalence. However, the overall influence of DIF on screening into at least one mood and anxiety disorder module was found to be minimal.