Machine Learning Used to Identify Undiagnosed PTSD in Civilian Patients in Analysis Group Study
October 17, 2022
Analysis Group researchers were part of the team publishing a study demonstrating how machine learning could be used to help identify undiagnosed post-traumatic stress disorder (PTSD). While veterans are systemically screened for PTSD in order to be connected with resources, civilian patients typically are not unless they mention trauma to a health care provider. Machine learning tools that can screen routinely collected clinical information to identify patients at risk for undiagnosed PTSD offer the potential to head off adverse health outcomes and lower health care resource utilization for these patients.
In “Identifying Individuals with Undiagnosed Post-Traumatic Stress Disorder in a Large United States Civilian Population – A Machine Learning Approach,” the study authors developed a machine learning model to screen non–trauma-based features in data from over 2 million civilian patients. The data cohort included patients diagnosed with PTSD as well as those without a diagnosis.
Managing Principal Annie Guérin, Vice President Martin Cloutier, and Manager Patrick Gagnon‑Sanschagrin worked with researchers from Otsuka Pharmaceutical Development & Commercialization, Inc., Lundbeck LLC, the Tuscaloosa Veterans Affairs Medical Center, and the University of Alabama Heersink School of Medicine. The machine learning model they developed analyzed demographic characteristics, clinical comorbidities, symptoms and complications potentially related to PTSD, treatments received, and use of emergency department services. Using these, the algorithm was able to distil the top seven predictive characteristics of likely PTSD. The study authors suggest that, lacking targeted PTSD treatment, undiagnosed patients’ long-term outcomes could include poorer quality of life and substantially higher health care costs, as patients’ symptoms are treated rather than a root diagnosis. In conclusion, the authors note that the study results could contribute to a “simple, accessible clinical screening tool” that does not rely on disclosure of trauma, but rather on readily available information.
The study was funded by Otsuka Pharmaceutical Development & Commercialization, Inc., and appeared in BMC Psychiatry in October 2022.