Rasch analysis of DSM-5 alcohol use disorder in a large inpatient sample

In the present study, we extend previous psychometric evaluations of the DSM-5 alcohol use disorder (AUD) criteria using Rasch analysis in a large clinical sample.

Adult patients with AUD (N = 1101) completed the DSM-5 AUD symptom checklist as part of routine clinical assessment upon admission to an inpatient substance use disorder treatment program. We conducted Rasch analysis of responses to the 11 criteria, examining model fit and item severities. We also examined whether there was evidence of differential item functioning based on sex and age.

Results supported the unidimensionality and local independence of the criteria, although some items were a suboptimal fit to the Rasch model. In particular, across all indicators, hazardous use exhibited misfit with model expectations. Additionally, the range of item severities did not span the full range of problem severity within this clinical sample, with many patients at the high end of the severity continuum and no items to differentiate them. There was evidence of differential item functioning by both sex and age, but effect size indices suggested that differences were unlikely to be clinically meaningful.

The present study supports the unidimensionality of the DSM-5 AUD diagnosis, but the misfit of certain items to the Rasch model and the narrow range of item severities suggest that more granular distinctions in AUD may be limited in high-severity samples. The results also suggest that the assumptions of interval-level measurement may not hold in clinical populations.