Detecting Possible Opioid Abuse or Dependence via Mathematical Modeling

Background: The misuse of prescription drugs, opioids in particular, is a significant public health problem. In this study we attempt to identify “stealth” opioid misusers, who are individuals that have characteristics of opioid misusers, but have not received an opioid misuse diagnosis, via a combination of mental health and other predictor variables. Methods: The dataset was obtained from the Thompson Reuters MarketScan Commercial Claims and Encounters database, and contains 2,841,793 individuals who received an opioid prescription between 2000 and 2008, of which 2,913 also had an opioid misuse diagnosis within the two-year period following the initial opioid prescription. Twenty-nine logistic regression models were calculated to test seven expanded definitions (and combinations thereof) of diagnosed and undiagnosed opioid misusers. The seven definitions include such indicators as filling opioid prescriptions at multiple pharmacies within 24 hours, significant dose escalation, or other unspecified drug dependence diagnosis.

Results: In the context of key variables that predict opioid misuse (e.g., demographic characteristics, other mental health and substance misuse diagnoses, and healthcare utilization), two of the stealth misuse definitions were approximately 90% concordant in identifying actual misusers. The first definition, which operationalized stealth misuse as being prescribed a long acting opioid and two or more short acting opioids, identified 90,740 individuals as actual or stealth misusers. The second definition included all stealth misuse indicators and identified 92,889 actual or stealth misusers.

Discussion: Expanded definitions of stealth misuse may help identify those at increased risk for opioid misuse. Implications exist for opioid misuse intervention and prevention efforts.