Grant: CIHR Fellowship

Background: Cannabis use disorder (CUD), or cannabis addiction, refers to the continued use of cannabis despite clinically significant impairment. Unfortunately, there is a shortage of information on CUD to guide healthcare. However, tapping into population-level data may improve CUD detection, documentation, prevention, and management. One source is health administrative data, which includes routinely collected diagnostic and procedural codes from encounters with the healthcare system. First, however, we must create and verify a "case definition" for CUD to use administrative data for CUD research. Case definitions are computer algorithms that combine diagnostic and procedural codes to identify people with CUD. In this project, we will use the latest technologies for complex data to change the current shortage of information about CUD into a wealth of new information that can be used to design new treatments, target the deployment of new services, and assess the impact of new treatments. Aims: We will validate and compare the accuracy of several CUD case definitions in Alberta administrative health data. Methods: Trained record reviewers will review 5,000 randomly selected files from a Calgary tertiary addiction centre and code them for the presence of CUD and link them to administrative health data. These records will serve as a reference standard for verifying case definitions. Two approaches for validating case definitions will be used and compared. In the traditional approach, we will make up a few case definitions and measure how sensitive and specific each is for CUD. In the second approach, we will use machine learning, in which a technique capable of learning from the data would be used to identify the best definition. Impact and Significance: Determining case definitions for substance use disorders in administrative health databases will lay the groundwork for future research.