Predictive modeling of adverse mental health outcomes following acute hospitalizations for serious illness: A machine learning approach with natural language processing

This study addresses predicting risk of post-traumatic stress disorder and other mental health conditions following hospitalization for life-threatening conditions.
We will use an existing, retrospective cohort of 408,377 patients hospitalized with severe illness, including sepsis, to develop and validate predictive models of mental health diagnoses using only electronic health record data, including features extracted with natural language processing from symptoms recorded in clinician notes.

0