Identifying and quantifying exposures involving counterfeit opioid analgesic products

Area of Science:

• Public Health
• Toxicology
• Data Science

Background:

• Counterfeit opioid drugs pose a significant threat to public health surveillance systems.
• Accurate data on counterfeit opioid products is crucial for understanding and responding to the opioid epidemic.

Purpose of the Study:

• To develop and validate a novel machine learning approach to identify and quantify exposures involving suspected counterfeit opioid products.
• To analyze trends in suspected counterfeit opioid exposures using United States poison center data.

Main Methods:

• An ecological study utilized data from the Researched Abuse, Diversion and Addiction Related Surveillance System (RADARS) between 2009 and 2021.
• A machine learning natural language processing (NLP) model was developed to analyze narrative case notes from poison center calls.

Main Results:

• The NLP model demonstrated high sensitivity (92%) and specificity (73%) in detecting suspected counterfeit-involved exposures.
• While only 2.1% of calls were predicted as counterfeit-involved from 2009-2021, a significant exponential increase was observed, with 23.7% of opioid analgesic exposures being suspected counterfeit-involved in 2021.
• A 7-fold increase in estimated suspected counterfeit exposures occurred between 2009 and 2021.

Conclusions:

• Machine learning NLP is a feasible and reliable method for identifying suspected counterfeit opioid exposures in poison center data.
• Suspected counterfeit opioids have substantially influenced intentional abuse and misuse rates of opioid analgesics in recent years.