Leveraging Electronic Health Records to Advance the Study of Gene-by-Environment Interactions in Psychiatry

Abstract

Leveraging real-world clinical biobanks to investigate the relationships between genetic and environmental risk factors for mental illness may help direct clinical screening efforts and evaluate genetic effects across environmental contexts. In this thesis, we focused on sexual trauma (ST) exposure, a known mental health risk factor that can be documented in electronic health records (EHRs). We developed an approach to extract ST documentation from the EHR, identified clinical diagnoses and settings associated with ST, and quantified the contributions of ST and polygenic scores (PGS) to risk for schizophrenia, bipolar disorder (BD), and major depressive disorder (MDD). First, we conducted a search using relevant keywords and phrases to identify disclosures of ST in the Vanderbilt University Medical Center (VUMC) EHR. In a phenome-wide association study, ST disclosures significantly associated with 465 clinical traits, including 90% of the psychiatric diagnoses tested. Next, we examined the clinical settings of ST documentation, finding that documentation occurred primarily in psychiatry (38.4% of notes), primary care (14.6%), and obstetrics/gynecology (14.5%) settings. Finally, in a meta-analysis across VUMC and Mass General Brigham biobanks, we studied the influence of ST on psychiatric risk in the context of genetics. ST exposure and PGS jointly explained 3.8-8.8% of phenotypic variance across traits (schizophrenia, BD, and MDD) and analysis sites. Further, a PGS-by-ST interaction analysis revealed that schizophrenia and BD PGS were more strongly associated with their respective diagnoses in individuals without ST documentation. These associations were robust to confounding by comorbid psychiatric conditions, substance use disorders, housing instability, and the clinical setting of ST documentation (psychiatry clinics). Taken together, these studies emphasize the public health impact of ST, illustrate the importance of considering settings of ST documentation, and quantify the interplay between ST and genetics in psychiatric disease. The results of the PGS-by-ST interaction analysis suggest that genetic predisposition to psychiatric illness may be less impactful in the presence of a severe environmental risk factor. More broadly, this work demonstrates that EHR-linked biobanks can be valuable resources for studying interactions between genetics and the environment in psychiatry.