Thursday, April 10, 2025
4/10/2025
Leveraging Machine Learning Approaches to Understand Mechanisms of Exposure Therapy in Real-World Settings - Project Summary/Abstract Exposure therapy is the most effective treatment available for obsessive compulsive disorder (OCD), yet up to 50% of patients do not recover. Progress has stalled because we do not fully understand how or why exposure exerts its effects. Studies that test mechanistic theories of exposure (based on habituation or inhibitory learning models) have yielded mixed findings for both self-report and physiological mechanisms, with a near exclusive focus on group-level effects in tightly controlled settings. While exposure likely works differently for different people, research on which mechanisms are most important for which individuals has been absent. As a result, clinicians are left to navigate mixed findings from different theoretical models, leading to variability in how they conduct exposure and greater treatment failures among patients suffering with OCD. The primary objective of this proposal is to determine which target mechanisms are most critical to engage in real-world exposure sessions to produce treatment response. Adult participants (N = 400) with OCD receiving exposure therapy from two sites (McLean Hospital, San Diego State University) across the continuum of care (outpatient, partial hospital, residential) will complete baseline clinical and demographic measures as well as weekly symptom reports. We will measure exposure mechanisms across three levels of analysis (self-report, observer-rated behavior, physiology) during each exposure session. Mechanisms assessed will include a broad range of variables based on both habituation and inhibitory learning models of exposure. For self-report and observer- rated mechanisms, we will use the Exposure Feedback Form (EFF), created and piloted by the study team. For physiological mechanisms, we will measure skin conductance response, heart rate, and heart rate variability via a wearable device. Leveraging machine learning analyses, we will develop a predictive model of patient outcomes (OCD symptoms) that incorporates both mechanistic data and baseline demographic and clinical characteristics. At the group level, we will test the hypotheses that incorporation of mechanistic variables will improve predictive performance over baseline variables alone, and that greater habituation and inhibitory learning will predict superior clinical outcomes. At the individual level, we will test the hypotheses that lower baseline depression and greater anxiety sensitivity will predict stronger habituation-outcome relationships, whereas lower baseline self-efficacy and sexual minority identity will predict stronger inhibitory learning-outcome relationships. This will be the first study to predict mechanism-outcome relationships for exposure therapy based on individual characteristics. Consistent with NIMH’s Strategic Objective 3, this proposal aims to identify therapeutic targets for personalized interventions by utilizing innovative machine learning approaches with scalable methods. Future studies will translate insights gained from this project to test implementation of personalized medicine approaches to exposure therapy, with the ultimate goal of improving care for the many patients who do not remit following exposure for OCD.
HHS’ Tracking Accountability in Government Grants System (TAGGS) website provides access to detailed descriptions of grants, loans, aggregated direct payments and other types of financial assistance awarded by the HHS Operating Divisions (OPDIVs) and the Office of the Secretary Staff Divisions (STAFFDIVs).