Our results also offer evidence of ASPD prediction from the aggression subtypes, direct aggression in particular. In all models, direct aggression ratings significantly predicted ASPD at levels similar to those of the total aggression models, with similar AUC values as well. This may indicate direct aggression as a specific subtype closely associated with ASPD, however, these results may also reflect the higher proportion of direct aggression items comprising the aggression scale. Regarding indirect aggression, while our power is perhaps low due to the indirect aggression score coming from only two questions, ratings from at least one informant (self, age 14), significantly predicted ASPD. These aggression subtype findings were consistent whether the two aggression subtypes were modeled simultaneously or as residual scores separately. Of note, while the odds ratios were similar between the two models, the AUC values were consistently higher when the subtypes were modeled simultaneously. Although the two subtypes are correlated, perhaps unique differences between the subtypes convey additional value in improving the strength of the prediction when simultaneously modeled.
