In addition to having implications for clinicians, the results of this study have relevance for psychometricians. As noted previously, individuals who were positive solely for DSM-5 severe AUD had lower levels of physiological dependence, despite otherwise greater severity of AUD, than those positive solely for DSM-IV dependence. When this counter intuitive finding was explored in post-hoc analyses, the difference reflected less frequent endorsement among cases gained of sleep problems and vomiting, the mildest and most commonly endorsed withdrawal symptoms(Dawson et al., 2010; Kahler and Strong, 2006). Although these differences were marginally significant at the individual symptom level (p = .048 and .050), they resulted in a highly significant difference in the overall prevalence of physiological dependence (p <.001) for cases lost and gained. This suggests a psychometrically undesirable property of the withdrawal criterion, i.e., a tendency to be inversely related to other indicators of AUD severity when defined solely in terms of its mildest symptoms. This observation is consistent with findings reported elsewhere (Harford et al., 2009; Kahler and Strong, 2006) that the withdrawal criterion had a low discrimination score
