Use Case 2 (UC2) Aim 4: Identifying Use of Cannabis-Derived Products (CDP) from Free-Text Notes

This project was aim 4 of the overall UC2 project “Empirical Application of the Sentinel EHR and Claims Data Partner Network to Enhance ARIA Sufficient Inferential requests and Atypical Descriptive Requests.”

The FDA understands that there is increasing interest in the potential utility of cannabis for a variety of medical conditions, and, as consequence, a need to strengthen surveillance on the potential adverse health effects from cannabis use in the American public. While the FDA has not approved any marketing application for cannabis for the treatment of any disease or condition, it has approved Epidiolex (cannabidiol (CBD)), a cannabis-derived drug product and three prescription synthetic cannabis-related drug products, Marinol® (dronabinol), Syndros® (dronabinol), and Cesamet® (nabilone). 

At the federal level, the Agriculture Improvement Act of 2018 legalized hemp, defined as cannabis (Cannabis sativa L.) and derivatives of cannabis with low concentrations of the psychoactive compound delta-9-tetrahydrocannabinol (THC). Since then, the CBD and cannabinoid hemp product market has substantially increased in volume and complexity, spanning thousands of products. The CBD market alone was estimated to be $4.6 billion in 2021 and forecasts predict the market will quadruple by 2026. Also, an increasing number of state, local, tribal, and territorial bodies have enacted laws allowing medical or both medical and nonmedical adult use of cannabis.

Despite increased patient usage of cannabis and non-FDA-approved cannabinoid products (CP) in recent years, CP are not currently captured in claims data. Instead, current FDA safety surveillance and signal detection capabilities for cannabis and non-FDA-approved CP rely largely on passive surveillance systems and surveys in large part because use of these products is not documented in structured healthcare data. However, unstructured patient-reported data within electronic health records (EHRs) may capture non-FDA-approved CP use. 

As part of its mission, the FDA Sentinel Innovation Center (IC) developed and implemented methods aimed at better utilizing EHR data for use in Sentinel. The purpose of this project was to use natural language processing to extract information on cannabis and CP exposure from free-text notes. Vanderbilt University Medical Center (VUMC) is one of the four lead hubs in the Sentinel IC Development Network that provides ready access to claims linked to structured and unstructured EHR data. By using the VUMC EHR-claims linked data covering 1.7 million lives, IC investigators tested and evaluated methods to improve FDA surveillance of cannabis and non-FDA-approved CP.