Be part of the knowledge.

We’re glad to see you’re enjoying ReachMD…
but how about a more personalized experience?

Register for free

Machine Learning Proves Effective for Opioid Risk Stratification

Machine Learning Proves Effective for Opioid Risk Stratification

Machine learning algorithms can offer effective risk stratification to identify individuals at elevated risk of an opioid overdose, says a new study published in JAMA Network Open this month.

A machine learning tool developed by researchers at the University of Florida, University of Pittsburgh, Carnegie Mellon University, and the University of Utah was able to use administrative data to stratify Medicare patients into risk-based subgroups, allowing providers to focus their resources on higher-risk individuals.

“Our model was effective in dividing the participants into three risk groups according to predicted risk score, with three-quarters in a low-risk group with a negligible rate of overdose, and more than 90 percent of individuals with overdose captured in the high- and medium-risk groups,” explained Jeremy C. Weiss, assistant professor of health informatics at Carnegie Mellon University's Heinz College.

With opioid overdoses claiming the lives of more than 100 people every day, the ability to leverage existing clinical and administrative data to identify potential victims can have immediate impacts on outcomes.

While many opioid-related overdoses occur due to non-medical use of narcotics, Medicare beneficiaries with opioid prescriptions are also at high risk of addiction and accidental overdose.

The team studied data from more than 560,000 fee-for-service Medicare beneficiaries to examine the impact of potential predictors of overdose, including demographic traits, health status, patterns of opioid use, and socioeconomic determinants.

The sample did not include individuals with cancer, who typically have more complex pain management routines.

“Our primary goal was risk prediction, and the secondary goal was risk stratification (i.e., identifying patient subgroups at similar overdose risk),” the researchers explained.

“We chose Medicare because of the high prevalence of prescription opioid use and the availability of national claims data and because the program will require specific interventions targeting individuals at high risk for opioid-associated morbidity.”

The group employed several different machine learning strategies to identify the method that offered the most accurate risk prediction and stratification, including deep neural networks.

Individuals identified as high-risk were between 7 and 8 times more likely than low-risk individuals to experience an overdose, the study found.

The models were most effective at identifying individuals with the lowest risk of opioid overdose.

Only 0.01 percent of individuals assigned to the low-risk group subsequently experienced an overdose. The low-risk group included more than three-quarters of all individuals studied.

Using machine learning tools to stratify patients by risk could help providers eliminate candidates for enhanced management and thereby focus their efforts on higher-risk individuals.

The machine learning strategies outperformed traditional analytics tools, highlighting the emerging importance of leveraging advanced analytics strategies for population health management and predictive analytics.

“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, associate professor of medicine at the University of Pittsburgh and senior author on the study.

The researchers acknowledged that in terms of applying the findings to the existing healthcare environment, the nature of the study was imperfect. Since the team used Medicare claims data as the basis for identifying individuals with opioid prescriptions, they could not account for people who acquired opioids through non-medical means.

Limiting the data to the fee-for-service Medicare environment also fails to account for the growing number of Medicare Advantage beneficiaries who may be receiving opioids, and may make it difficult to generalize the results to other types of patient groups.

However, there is little doubt that machine learning strategies themselves will play an increasingly important role in risk stratification and predictive analytics.

Machine learning is already being integrated into a number of health IT applications, and new research on the effectiveness of the methodology is being published nearly every day.

“Although they are not perfect, these models allow interventions to be targeted to the small number of individuals who are at much greater risk,” added Gellad, giving providers much-needed insight into how to target their resources appropriately.

As the opioid crisis continues to have a profound impact on population health, innovative analytics tools that can help providers be more proactive about pain management may be able to reduce the number of avoidable deaths from opioid-related causes.

Facebook Comments

LIVE ON REACHMD RADIOBack to live radio


Programs 8/19/19

Get a Dose of ReachMD in Your Inbox
and Practice Smarter Medicine

Stay current with the best in medical education.