Current methods used to identify Medicare patients at high risk of overdosing on prescription opioids target many people who are not really at high risk, according to a team of researchers who found that over 90% of patients have little to no risk of overdosing.

“The ability to identify such risk groups has important implications for policymakers and insurers who currently target interventions based on less accurate measures,” said lead author Wei-Hsuan “Jenny” Lo-Ciganic, PhD, a professor of pharmaceutical outcomes and policy at the University of Florida, who reported her findings in JAMA Network Open.  

Lo-Ciganic and her colleagues at the University of Pittsburgh, Carnegie Mellon University and University of Utah studied health data on over half a million Medicare beneficiaries who filled one or more prescriptions for opioids between 2011 and 2015. The researchers identified which patients overdosed and then used machine-learning algorithms to analyzed their demographics and health records.

Put another way, out of 100,000 Medicare patients in the low risk group, six would have an overdose; while there would be 1,770 overdoses in a high risk group of the same size.

Not surprisingly, the computer models found that high doses of opioids and a prior history of substance abuse significantly raise the risk of an overdose. So does a person’s age, disability status and whether they are co-prescribed benzodiazepines. Patients who live in certain states (Florida, Kentucky or New Jersey) are also at higher risk.

Top 10 Predictors of Opioid Overdose

1. Total MME (morphine milligram equivalent)
2. History of substance or alcohol abuse
3. Average daily MME
4. Age
5. Disability status
6. Number of opioid refills
7. Resident state
8. Type of opioid
9. Number of benzodiazepine refills
10. Drug use disorders  

The study found that the machine-learning algorithms the researchers developed performed well in predicting overdose risk and in identifying patients with a low risk. Machine learning is an alternative analytic approach to handling complex interactions in large data.  It can discover hidden patterns and generate predictions in clinical settings. Based on their findings, the researchers concluded that their approach outperformed other methods for identifying risk used by the Centers for Medicare and Medicaid Services.

“Machine-learning models that use administrative data appear to be a valuable and feasible tool for identifying more accurately and efficiently individuals at high risk of opioid overdose,” says Walid Gellad, MD, a professor of medicine at the University of Pittsburgh and senior author on the study. “Although they are not perfect, these models allow interventions to be targeted to the small number of individuals who are at much greater risk.”