Only about 22% of people with OUD receive specific treatment, while most remain at high OD risk that is clinically under-identified. Sensitive and valid approach is critically needed to identify those individuals who are at risk for using opioids and would then be at increased risk for a trajectory of increasing use of prescription opioids, OUD severity, or opioid overdose. We have developed opioid risk (OD and OUD) prediction models that take advantage of the sequential and graph nature of electronic health records, taking advantage of a large number of EHR features from a nationwide EHR database.

References:

• Xinyu Dong, Rachel Wong, Weimin Lyu, Kayley Abell-Hart, Jianyuan Deng, Yinan Liu, Janos G. Hajagos, Richard N. Rosenthal, Chao Chen and Fusheng Wang: An Integrated LSTM-HeteroRGNN Model for Interpretable Opioid Overdose Risk Prediction. Artificial Intelligence In Medicine. Volume 135, January 2023.
  
  
• Xinyu Dong, Jianyuan Deng, Sina Rashidian, Kayley Abell-Hart, Wei Hou, Richard N Rosenthal, Mary Saltz, Joel H Saltz and Fusheng Wang: Identifying Risk of Opioid Use Disorder for Patients Taking Opioid Medications with Deep Learning. Journal of the American Medical Informatics Association (JAMIA). Volume 28, Issue 8, August 2021, Pages 1683–1693.
  
  
• Xinyu Dong, Jianyuan Deng, Wei Hou, Sina Rashidian, Kayley Abell-Hart, Richard N Rosenthal, Mary Saltz, Joel H Saltz and Fusheng Wang: Predicting Opioid Overdose Risk of Patients with Opioid Prescriptions Using Electronic Health Records Based on Temporal Deep Learning. Journal of Biomedical Informatics. 2021 Mar 9;116:103725.
  
  
• Xinyu Dong, Sina Rashidian, Yu Wang, Janos Hajagos, Mary Saltz, Joel Saltz and Fusheng Wang: Machine Learning Based Opioid Overdose Prediction Using Electronic Health Records. In Proceedings of AMIA Annual Symposium 2019, November 16-20, Washington DC, USA.