Case study about predicting who is at risk for opioid abuse and overdose.
Data Science

Advancing the fight against opioid abuse and overdose

February 11, 2021
Predicting opioid abuse and overdose helps patients at high risk.
Lead Principal Data Scientist

“In 2015, the amount of opioids prescribed was enough for every American to be medicated around the clock for three weeks.”

The ability to forecast future risk and earlier interventions offer healthcare organizations the real possibility to reduce the toll of opioid abuse and overdose.

Opioid abuse is a public health crisis that knows no boundaries. From 1999 to 2017, more than 702,000 people died from a drug overdose. Nearly 68 percent involved a prescription or illicit opioid. What’s more, an estimated 40 percent of opioid overdose deaths involved a prescription opioid.

With that said, while alarming, there should be no surprise to learn 21 to 29 percent of patients prescribed opioids for chronic pain misuse them, and CDC research suggests people can become addicted to opioids in as few as five days. The data is profound and it is urgent for the healthcare community to rally around this issue – to identify who is at risk of misuse and overdose before they receive a prescription. In fact, for many conditions there is limited evidence of opioid effectiveness, and identifying who is at risk and who should receive an alternative treatment could save lives without sacrificing care and pain management.

Statistics from the CDC on opioid abuse

This is why the Geneia Data Intelligence Lab (GDI Lab), which develops innovative and powerful data models to lower healthcare costs and improve health outcomes, chose to tackle this critical issue. In this case, the GDI Lab’s data scientists used a national dataset of insurance claims for medical encounters and pharmacy prescriptions to create an opioid abuse and overdose predictive model*+ that, for those who have received at least one opioid prescription, accurately identified 80 to 88 percent of the people likely to have an opioid abuse diagnosis or an overdose event in the next six months.

The model was evaluated on several independent datasets, with more than 100 variables considered. Ultimately, 22 variables were found to be the most predictive of opioid abuse and overdose. Demographic characteristics and number of opioid fills were important variables correlated to the model’s predictions.

Without a doubt, this is progress in the fight against opioid addiction. The ability to forecast future risk and earlier interventions offer healthcare organizations the real possibility to reduce the toll of opioid abuse and overdose to individuals and families across the country.

To learn more about this model and how you can improve the health of your most vulnerable patients while mitigating future costs, download the Predicting Opioid Abuse and Overdose Case Study.

*Certain data used in this study were supplied by International Business Machines Corporation. Any analysis, interpretation, or conclusion based on these data is solely that of the authors and not International Business Machines Corporation.

+Predictive models, by their very nature, contain certain assumptions. This is not an attempt to practice medicine or provide specific medical advice, and it should not be used to make a diagnosis or to replace or overrule a qualified healthcare provider's judgment.