It is apparent from the comparison across sites that there are several differences in both the discrimination and severity estimates for different criterion. The bottom of Table 4 shows the values of discrimination and severity for the consumption variables. In general the inclusion of a consumption measure item along with other DSM items appears to have accomplished the stated goal of including an additional item that is consistent with other items of alcohol use disorders and that also taps the mildest range of the alcohol use disorder continuum. Nevertheless, the usual 5+/4+ is too low in discrimination and severity (also with very high Bayesian Information Criterion [BIC] values) and while slightly lower BICs were found in the model using the drunkenness variable, the BIC value was still too high, possibly a result of multidimensionality introduced by including the item. These conclusions apply for all sites. The two best candidates to tap the lower level of severity of the continuum while keeping acceptable BICs scores were 12+ drinks monthly/any for men/women, and 5+ drinks weekly/monthly for men/women. The variable 5+ drinks
