Machine learning models using multiple phenotypes—PGS, neurophysiological information, and behavioral data—have previously been used to identify risk for AUD development. 122 , 123 , 124 When applied to remission and recovery, these models hinted at some improvement in accuracy in predicting remission in COGA participants. 125 Specific patterns in EEG‐derived brain network functional connectivity consistently predicted remission across sex and race/ethnic groups, but the addition of a number of PGS related to personality traits, aggression, depression, and alcohol use, along with marital status, medication use, and employment status, further improved prediction, in particular among sex‐ and race/ethnicity‐defined subgroups. 125 This work highlighted the utility of incorporating a wide array of measured indicators, including polygenic, electrophysiological, and psychosocial domains, in the construction of predictive models of remission from AUD.
