ED-delivered tool feasible for predicting PTSD, major depression after trauma

A question set feasibly administrable in the ED predicted 3-month PTSD or major depressive episode among individuals who experienced a motor vehicle collision, according to an analysis of a longitudinal study in JAMA Psychiatry.

“PTSD, by definition, occurs in the aftermath of a traumatic event, and many people who go on to develop PTSD after traumas present at EDs in the immediate aftermath of the trauma but before the onset of PTSD,” Ronald C. Kessler, PhD, McNeil family professor in the department of health care policy at Harvard Medical School, told Healio Psychiatry. “It’s impossible to implement potentially expensive preventive interventions with all the millions of trauma patients who come to EDs each year. An easy-to-implement screening method is needed to determine which patients are at high PTSD risk to target preventive interventions, which is the unmet need we addressed in this research.”

Kessler and colleagues aimed to develop and validate a prediction tool based on ED reports following a motor vehicle collision to predict PTSD or major depressive episode at 3 months. They analyzed data of 1,003 patients (median age, 34.5 years; weighted 67.9% women) included in the Advancing Understanding of Recovery After Trauma study, which investigated adverse posttraumatic neuropsychiatric sequalae among individuals who visited 28 U.S. urban EDs immediately following a traumatic experience, with enrollment beginning on Sept. 25, 2017. Participants completed 3-month evaluations by Jan. 31, 2020, with each having received a baseline ED evaluation in addition to follow-up self-report surveys 2 weeks, 8 weeks and 3 months later.

The researchers predicted 3-month PTSD or major depressive episode from baseline information using an ensemble machine learning method. Further, they used the PTSD Checklist for DSM-5 to evaluate PTSD, as well as the Patient Reported Outcomes Measurement Information System Depression Short-Form 8b to evaluate major depressive episode.

Results showed 274 patients (weighted 26.6%) fulfilled criteria for PTSD or major depressive episode at 3 months. The ensemble machine learning model restricted to 30 predictors in a training sample exhibited good prediction accuracy and calibration in an independent test sample. Kessler and colleagues noted patients in the top 30% of predicted risk represented 65% of all 3-month PTSD or major depressive episode; the mean positive predictive value among these patients was 58.2%.

The researchers observed good consistency for the model across regions of the U.S. regarding both area under the curve and calibration. Symptoms of anxiety sensitivity and depressive disposition, psychological distress in the 30 days prior to motor vehicle collision and peritraumatic psychosomatic symptoms served as the most important predictors in terms of Shapley Additive Explanations values.

“The findings could help mental health treatment providers target high-risk patients for preventive interventions in the immediate aftermath of trauma,” Kessler said. “This could prevent an enormous amount of trauma-related psychopathology by intervening at a time before PTSD has a chance to occur, become chronic and lead to the wide range of secondary physical and psychiatric problems we know it to predict.”