This study, presented by Bareis et al., leverages machine learning to identify primary behavioral-health phenotypes among adults with schizophrenia and to evaluate how those phenotypes relate to psychotropic medication use and clinical outcomes. Because 70% of people with schizophrenia in the U.S. receive healthcare through Medicaid, the authors analyzed Medicaid claims from 2010 (N = 249,006 adults aged 18–64) and applied Latent Dirichlet Allocation to ICD-9-CM diagnostic codes. After mapping psychotropic medications using FDA National Drug Codes, they identified five behavioral-health phenotypes: substance use (10.9%), mania–mixed mood (16.8%), depression (12.2%), conduct disorder–developmental delay (15.3%), and anxiety–paranoid (13.6%), along with a schizophrenia-only group (31.2%). Notably, 68.8% of participants had more than one co-occurring behavioral-health diagnosis. Medication patterns varied substantially by phenotype, with over 87% of the cohort receiving any psychotropic medication and ~46% receiving mood stabilizers, concentrated in the mania phenotype. Outcomes also differed: ~16% of the cohort had a psychiatric inpatient admission and ~21% had an ED visit within the following year. The authors conclude that phenotype membership correlates with distinct prescribing patterns and outcomes, underscoring the importance of stratified pharmaco-epidemiologic research.