Northern California Institute for Research and Education San Francisco United States
We identified a cohort of OEFOIFOND Veterans who had a diagnosis of PTS after the end oftheir last deployment N308,556. This PTS cohort had a mean age of 35.9 years SD8.8years, was 89.6 male, and was comprised of those who were 57.5 white, 13.9 black, 11.4 Hispanic, 3.9 other and 13.9 unknown. We identified 5.7 million psychotherapy visits madeby these Veterans in the study period and linked these to approximately 8 millionpsychotherapy notes. We completed a standardized annotation guide specifying each category ofpsychotherapy to be identified and then four study team members annotated a set of 650 notes.We used a bag of words and support vector machine SVM approach to generate a classifier for irrelevant notes, which had a classification accuracy of 88 . This allowed us to removeirrelevant documents and generate an enriched pool of 4 million psychotherapy notes to beused for ongoing rounds of annotation. Also during this period, we identified 30,884 10 Veterans in our PTS cohort with three or more PCLs within an 18 week period, and,separately, 101,761 33 of PTS cohort members of the PTS cohort with two or more suicide screenings during the post-deployment period.