Machine Learning Model Predicts PTSD Following Military Deployment
One-third of US veterans flagged as high risk for PTSD by a machine learning model in a recent study accounted for 62.4 percent of cases of the condition.
Source: Getty Images
By - Researchers have developed a machine learning (ML) model that can accurately predict risk of posttraumatic stress disorder (PTSD) prior to United States military deployment, according to a study published last week in JAMA Network Open.
Military deployment carries substantial risk for life-threatening and other traumatic experiences that can lead to PTSD, and the condition can have serious adverse impacts on an individual’s life based on the severity of their symptoms.
This makes accurately predicting PTSD risk for those in the military a major research priority, as veterans are more likely to experience post-traumatic stress than civilians. To further these efforts, the researchers in this study set out to develop and validate an ML model to predict post-deployment PTSD.
The research team gathered pre-deployment assessment data from 4,771 soldiers from three US Army brigade combat teams approximately one to two months before deployment to Afghanistan. Follow-up assessments taken three to nine months after deployment were also included.
Using these data, the researchers developed multiple ML models to predict PTSD using as many as 801 predictors pulled from pre-deployment assessments. Models were then cross-validated to select an optimal model.
The final model’s performance was then evaluated using area under the receiver operating characteristics curve and expected calibration error in a different cohort.
Roughly 15.4 percent, or 746 of the veterans in the study, met the criteria for PTSD post-deployment.
The optimal model, a gradient-boosting ML algorithm, leveraged 58 core predictors to forecast PTSD risk. The model achieved an area under the curve of 0.74 and a low expected calibration error of 0.032.
The core predictors used in the model measured 17 distinct domains: stressful experiences, social network, substance use, childhood or adolescence, unit experiences, health, injuries, irritability or anger, personality, emotional problems, resilience, treatment, anxiety, attention or concentration, family history, mood, and religion.
Overall, one-third of study participants with the highest risk accounted for 62.4 percent of PTSD cases.
The researchers concluded that these findings suggest the feasibility of pre-deployment PTSD risk stratification for US veterans, an approach that could facilitate the development of early intervention and prevention strategies to improve outcomes and quality of life.
Research efforts like this one reflect a broader interest in how artificial intelligence (AI) can help enhance PTSD care.
Last year, US Defense Health Agency (DHA) shared that its PTSD Drug Treatment (PTSD-DT) Program would leverage medication adherence and digital biomarker solutions from New York-based artificial intelligence (AI) and data analytics company AiCure.
The partnership aims to optimize treatments and increase patient support by evaluating the efficacy of therapies for service members and veterans. DHA will also use AiCure’s tools to support its PTSD adaptive platform trial (APT), which will work to advance precision medicine approaches to prescribing PTSD treatments and informing future therapies.
