Prediction of Adverse Events Risk in Patients with Comorbid Post- Traumatic Stress Disorder and Alcohol Use Disorder...

Area of Science:

• Computational psychiatry
• Clinical informatics
• Machine learning in healthcare

Background:

• Accurate prediction of high-risk events in patients with mental disorders is critical.
• Previous work developed DeepBiomarker using electronic medical records (EMR) for suicide-related event (SRE) risk in post-traumatic stress disorder (PTSD) patients.

Approach:

• The DeepBiomarker2 model was applied, integrating multimodal data including lab tests, medications, comorbidities, and social determinants of health.
• EMRs from 5,565 patients with PTSD and alcohol use disorder (AUD) were analyzed for risks of adverse events (opioid use disorder, SREs, depression, death).

Key Points:

• DeepBiomarker2 achieved an AUROC of 0.94 in predicting adverse events within 3 months for PTSD + AUD patients.
• Identified 11 medications (piroxicam, vilazodone, dronabinol, etc.) with potential to reduce adverse event risk.

Conclusions:

• DeepBiomarker2 demonstrates high accuracy in predicting multiple adverse event risks.
• The model can identify potential risk and beneficial factors, supporting personalized interventions for diverse populations.