DSM-IV criteria endorsement patterns in alcohol dependence: relationship to severity.

BACKGROUND: In DSM-IV, the diagnostic threshold for alcohol dependence (AD) is met when a patient presents with at least 3 of 7 criteria. We have computed the predictive value for each individual DSM-IV AD criterion, and examined subtypes of AD criteria endorsement patterns and their associated severity indicators for community-dwelling AD individuals.
METHODS: We utilized data from the 2001 to 2002 National Epidemiological Survey on Alcohol and Related Conditions (NESARC). Positive predictive values (PPV) for DSM-IV AD were computed for each of the individual criteria. Patterns of criteria endorsements were identified by latent class analysis (LCA). Sociodemographic status, age of onset and duration of AD, patterns of drinking, and drinking treatment history, were conditional on DSM criteria endorsement clusters, as indicators of the respondents' clinical severity.
RESULTS: At the individual criterion level, the single criterion with the greatest PPV was D7--"Activities given up" with approximately 95% of drinking individuals who endorsed this DSM criterion correctly diagnosed as having DSM-IV AD. In addition to D7, only D5--"Physical/Psychological problems", and D6--"Time spent" had a PPV for AD substantially >50%. The LCA of AD endorsement patterns yielded a 6-cluster solution. The most common response pattern (34.5% of those with AD) was endorsement of 5 criteria: D1--"Quit/Control," D2--"Larger/Longer," D3--"Tolerance," D4--"Withdrawal," and D5--"Physical/Psychological problems." The most severe cluster (14%) was comprised of those who were likely to endorse 7/7 criteria. Cluster 1 (8.3%) did not include an endorsement of withdrawal, despite a heavy pattern of alcohol consumption. Unmarried status was associated with more severe criteria endorsement patterns.
CONCLUSIONS: The present findings indicate a Guttman-like scaling of endorsement which yielded associations with severity for some of the concurrent indicators included in the analysis. However, severity measures did not always increase with DSM-IV AD criterion endorsement counts. Although endorsement of 6/7 or 7/7 criteria was associated with greater severity across a variety of indicators, fewer criteria were randomly associated with these measures. These data do not support the use of AD symptom counts as a phenotypic dependent variable. At least 2 extant diagnostic criteria showed relatively low PPV for AD, indicating a need for further assessment of these criteria with new symptoms or re-wording of the current symptom items.