Opioids Work

Our team is conducting innovative research with state, national, and community partners using unique and powerful data sets. We are developing machine learning algorithms to better predict risk for opioid overdose and other adverse opioid events. Our research aids prescribers to make informed decisions for their patients and to target resources where they are needed most. A sample of funded ongoing projects is below.

Examining the quality of opioid use disorder treatment in a Medicaid research network
PI: Julie Donohue, PhD
Funding Source: NIH/NIDA
To address the opioid crisis, many state Medicaid programs have expanded coverage and payment for treatments for opioid addiction. This project examines differences across 9 states in the quality of opioid addiction treatment delivered in Medicaid, assesses the link between treatment quality on drug overdose outcomes, and examine the impact of recent Medicaid changes on overdose outcomes.

Risk of long-term opioid use and its sequelae due to opioid prescribing for acute pain
PI: Julie Donohue, PhD
Funding Source: Benter Foundation
This project examines outcomes among patients who are prescribed opioids for pain for the first time. Analyses will include differences in treatment settings and potential long-term opioid use in patients.

Machine learning and opioid overdoses
PI: Walid Gellad, MD, MPH
Funding Source: R.K. Mellon Foundation
This project leverages data from Allegheny County’s Department of Human Services and Department of Health to develop algorithms using machine learning to predict opioid overdose.  These algorithms can be modified in the future to predict risk of other consequences of opioid addiction, such as loss of child custody, housing instability, and criminal justice contact.

Using Machine Learning to Predict Problematic Prescription Opioid Use and Opioid Overdose
PI: Walid Gellad, MD, MPH
Funding Source: NIH/NIDA
This R01 project improves on traditional methods for predicting risk of overdose from prescription opioids by applying machine learning techniques.  The study team, led by Dr. Gellad, will develop prediction algorithms to identify patients who are at risk of opioid overdose, and after testing and refining the algorithm, will compare the accuracy of an approach integrating Medicaid claims data with clinical data to an approach using claims data alone.

Recently published work in opioids