Predicting Adverse Events in PTSD and Alcohol Use Disorder Patients Using Deep Learning
Author Information
Author(s): Miranda Oshin, Qi Xiguang, Brannock M. Daniel, Whitworth Ryan, Kosten Thomas R., Ryan Neal David, Haas Gretchen L., Kirisci Levent, Wang Lirong
Primary Institution: University of Pittsburgh
Hypothesis
Can integrating drug target information into deep learning models improve the prediction of adverse events in patients with comorbid PTSD and AUD?
Conclusion
The T-DeepBiomarker model effectively predicts adverse outcomes in PTSD + AUD patients and identifies candidate drugs that may reduce these risks.
Supporting Evidence
- T-DeepBiomarker achieved high predictive performance with an AUROC of 0.94 for adverse outcomes.
- Several medications were identified as potentially reducing the risk of adverse events.
- The model integrates multimodal data, enhancing prediction accuracy.
- Findings suggest a need for further investigation into identified candidate drugs.
Takeaway
This study used a computer model to help doctors predict bad outcomes for patients with PTSD and alcohol problems, and it found some medicines that might help.
Methodology
The study analyzed electronic medical records of 5565 patients with PTSD and AUD, using a deep learning model called T-DeepBiomarker to predict adverse events.
Potential Biases
Variability in PTSD diagnoses due to the use of both ICD-9 and ICD-10 criteria may have affected results.
Limitations
The study may have missed important drug-outcome associations due to the limited number of medications prescribed and did not establish causative relationships.
Participant Demographics
Patients diagnosed with both PTSD and AUD, identified through specific ICD codes.
Statistical Information
P-Value
0.006
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website