The National Syndromic Surveillance Program, a Centers for Disease Control and Prevention (CDC) surveillance system, allows timely detection of emergency department (ED) trends by matching text and diagnosis (DX) codes to syndrome criteria. At this time, there is no CDC syndrome definition for marijuana. Accidental child exposures and over-consumption of edible products are emerging concerns. A validated marijuana syndrome will allow health departments with access to ED data to measure relative trends in marijuana-related ED visits.
METHODS:
A marijuana syndrome definition which incorporates text and DX fields was developed and evaluated with data from 14 hospitals in Adams, Arapahoe, Denver, and Douglas counties reporting to NSSP. Preliminary marijuana cases were identified based on search criteria for DX and text fields. Text criteria included terms like “marijuana” and “edible.” The DX field was searched for International Classification of Disease (ICD) 9 and 10 codes regularly used to identify marijuana-related cases.
Preliminary marijuana cases were identified based on criteria for diagnosis and text fields. These were manually reviewed and categorized according to whether the text and diagnosis codes affirmed that marijuana was involved in the circumstances of the encounter. The categories were marijuana-related, unrelated, and unclear. This assessment produced the number of marijuana-related encounters and the predictive value positive (PVP). Results were examined by search field (diagnosis, chief complaint, clinical impression, and triage notes), search term, DX code, and age. Findings were used to refine the case definition.
RESULTS:
From September – October 2015, 524 preliminary marijuana cases were identified. Of these, 94% were related to marijuana, 4% were unrelated, and 1% were unclear. Among related cases, 58% matched on a diagnosis code alone, 33% only matched on a text field, and 9% matched on diagnosis and text fields. The ICD-9 code group 305.2 (“nondependent cannabis abuse”) and text “marijuana” and “THC” returned the most related cases: 143, 109, and 48 cases respectively (including ones matching multiple criteria). The definition was refined to exclude false matches (e.g. “pot” but not “crock pot”) and remove terms with low PVP. The refined definition was tested using November data and yielded a 96% PVP.
CONCLUSIONS:
The use of text and DX criteria in the marijuana syndrome definition identified many related cases and provided a strong PVP. Applying a validated marijuana syndrome definition to syndromic surveillance or other ED data will help inform policies and monitor trends as states around the country explore and legalize retail marijuana.