In Table 2 further analyses were then carried out to more directly evaluate how items from Table 1 that were significantly different across the combined Groups 1-3 versus Group 4 performed when each of Groups 1, 2 and 3 was evaluated for differences from Group 4. This approach used a series of backward elimination logistic regressions to identify the odds ratios (ORs) for optimal combinations of items for each group that best differentiated it from the no diagnosis Group 4. In Table 2, the regression predicting Group 3 (AUDs only) from among Group 3 plus the no diagnosis Group 4 identified six items including: a lower LR per drink; a higher score on an externalizing questionnaire; two alcohol use characteristics; one drug use item; and lower T1 education. Five of these six items also related to the combined diagnosis Group 1 compared to Group 4, with the sixth characteristic similar to Group 3 in that it involved an externalizing questionnaire score. The regression relating to Group 2 (SUD only) from among Groups 2 plus 4 had contributions from three items:
