Prime Therapeutics finds different predictive models are needed to identify future high-risk prescription opioid use for new versus current opioid users

ST. PAUL, Minn., Oct. 23, 2018 /PRNewswire/ -- To help combat misuse and abuse of prescription opioids, pharmacy benefit manager Prime Therapeutics LLC (Prime) recently developed a new predictive modeling process to help better identify Medicare individuals who are at increased probability of receiving high dose opioids. This research demonstrates that separate predictive models are needed for those newly initiating an opioid versus those already using opioids.

The Centers for Disease Control and Prevention (CDC) defines high dose opioid therapy, which is associated with increased harm, as greater than or equal to 90 milligrams morphine equivalent (MME) daily dose. Prime researchers set out to predict which members could be at an increased probability to receive high dose opioids, with the hopes of intervening and preventing harm.

"We found that one predictive model does not fit all. We did not anticipate different predictive modeling techniques were needed to provide the best predictions. We found the logistic regression model performed best for current opioid users and the decision tree model was substantially better for those newly initiating opioid therapy," said Patrick Gleason, PharmD, senior director for health outcomes at Prime. "Although not surprising, different factors predicted an individual's risk for future high dose opioids. For example, among current opioid users presence of a prior high potency opioid claim and recent escalating opioid claim volume were highly predictive."

For the study, eligible members were divided into two data sets: 1) new opioid users with no opioid claim in the prior 180 days; 2) current opioid users with an opioid claim in the prior 180 days. Over 10 predictive modeling methods were used to assess both data sets. The predictive modeling process incorporated more than 170 potential predictors including pharmacy claims data, demographic information, Centers for Medicare & Medicaid (CMS) data, and unique pharmacy and prescriber factors including travel distance, prescriber specialty, and CMS controlled substance outlier status.

For new users, the decision tree model performed best and some of the important independent predictors included MME of first opioid claim and whether it is a long-acting opioid, as well as number of other medications the individual has in their claims history. For current opioid users, the logistic regression model was the most accurate and some of the most important independent predictors were prior opioid use and recent escalating use. For both models, the prescriber specialty type was an independent predictor. The highly accurate predictive models score and rank a member on their future likelihood to receive high dose opioids which will allow more targeted outreach program development with the goal being to prevent high-risk opioid use.

"This work requires data scientists to be specialized in clinical claims data and the latest predictive modeling techniques," added Gleason. "We believe we are the first in the industry using this type of machine learning to realize the necessity of separating new and current opioid users when working to predict and score individuals for future high-risk opioid use."

Prime researchers will present this gold ribbon-winning study at the Academy of Managed Care Pharmacy's (AMCP) Managed Care & Specialty Pharmacy Nexus Meeting Oct. 22-25 in Orlando.

About Prime Therapeutics
Prime Therapeutics LLC (Prime) helps people get the medicine they need to feel better and live well. Prime manages pharmacy benefits for health plans, employers, and government programs including Medicare and Medicaid. The company processes claims and offers clinical services for people with complex medical conditions. Prime serves more than 27 million people. It is collectively owned by 18 Blue Cross and Blue Shield Plans, subsidiaries or affiliates of those plans. For more information, visit www.primetherapeutics.com or follow @Prime_PBM on Twitter.

SOURCE Prime Therapeutics