Document Type



Media is loading

Publication Date



Presentation: 5:52

Poster attached as supplemental file below


Major depression (MDD) and bipolar disorder (BD), impact approximately 28 million individuals in the United States. The two conditions are often mistaken for one another due to significant overlap in symptomology, resulting in misdiagnosis and delays in appropriate treatment selection. The application of artificial intelligence (AI) has been proposed to address these challenges; however, it is unclear how capable such methods are. In order to investigate AI machine learning methods’ capabilities to improve pharmacotherapy treatment selection for individuals with BD and MDD, a rapid systematic review of the PubMed database was conducted. 182 articles published between 2018 and 2022 were screened, yielding an analytic sample of 15 studies that met criteria for analysis. All studies aimed to develop predictive models to improve treatment selection either by forecasting the degree of treatment response (n=8) or by identifying meaningful subgroups based upon patient similarities (n=7). Data domains assessed for selection of model predictors included genetic, clinical, and sociodemographic. Studies reported achieving proof of concept for AI machine learning capabilities, but no model was ready for clinical deployment. Future studies should focus on replication, validation, and larger, more diverse datasets to enable models to have greater generalizable and predictive power.