Determining the role of discrimination in clinical presentation and treatment response among sexual minority people with OCD: A machine learning approach - Sexual minority (SM; e.g., gay, lesbian, bisexual) individuals are at increased risk for psychopathology compared to heterosexual people. These disparities are especially striking for obsessive-compulsive disorder (OCD), with past 12-month diagnosis rates being 9 times higher for SM compared to heterosexual people. Further, SM people represent 18% of patients receiving OCD treatment, which is significantly higher than the percentage of SM people in the general population (4-7%). Prior research has demonstrated that SM people enter OCD treatment with more severe OCD and comorbidities (e.g., anxiety, depression) than heterosexual people. In general, the mental health inequities affecting SM people have been attributed to their unique experiences of stress, such as discrimination. Despite accumulating evidence of sexual orientation disparities in OCD diagnosis and severity, there are significant gaps in our understanding of clinical presentation, treatment outcomes, and the impact of minority stress. To address these gaps, the current study will recruit 103 SM adults in partial hospital/residential treatment for OCD. As part of routine clinical monitoring and an established clinical research program, participants will complete measures of demographic characteristics, general risk factors (e.g., emotion dysregulation, distress intolerance), sexual minority stress (e.g., discrimination, internalized stigma), treatment variables (e.g., credibility, expectancy), and OCD severity (primary outcome) upon admission. At discharge, they will re-complete measures of OCD severity and quality of life (secondary outcome). These data will be used to address two specific aims: (1) Determine correlates of OCD severity at baseline for SM people with OCD, and (2) Predict clinical outcomes at discharge for SM people with OCD. Given the number of potential risk factors for OCD severity among SM individuals and the lack of prior research in this area, machine learning provides an ideal method to understand risk in this population as it can: 1) handle large numbers of predictors, 2) include predictors that are highly correlated, which general risk factors often are, 3) be optimized to select the subset of variables that maximize predictive performance, and 4) provide interpretable coefficients that lend themselves well to prediction in real-world settings. As such, findings can be used to enhance evidence-based interventions by identifying specific treatment targets for SM people with OCD, consistent with a recent Notice of Special Interest (NOT-OD-22- 032), emphasizing the need for research to reduce mental health inequities among SM populations.
