BACKGROUND: Legalization of recreational marijuana in Colorado created challenges to monitor adverse outcomes from marijuana use. Syndromic surveillance by using the Electronic Surveillance System for the Early Notification of Community-based Epidemics (ESSENCE) might serve to monitor trends in marijuana-associated emergency department (ED) visits. We sought to detect marijuana-associated ED visits by developing algorithms of specific diagnostic codes and key words and evaluated algorithm performance in ESSENCE compared with physician-performed medical record reviews.
METHODS: Algorithms representing various combinations of marijuana-specific International Classification of Diseases Tenth Revision (ICD-10) diagnostic codes or key word search terms in the chief complaint or triage note were applied to ED visits submitted through the syndromic surveillance system at 1 hospital in Colorado during June 1–September 30, 2016. We performed chart abstractions of ED medical records categorized as potentially marijuana-related by ≥1 ESSENCE algorithms to determine the case status of each patient visit. A case of an acute marijuana-associated adverse health outcome was defined as a negative health event resulting from marijuana exposure as determined by the ED physician’s clinical impression at the time of ED visit. Positive predictive values (PPVs) of each algorithm were calculated by using the medical record review as the ultimate standard.
RESULTS: Of 28,446 total ED visits during June 1–September 30, 2016 at this hospital, ESSENCE algorithms detected 281 (1%) visits as potentially marijuana-associated. Two-hundred fifty (89% of 281) medical record reviews identified 114 (46% of 250) cases of an acute marijuana-associated adverse health outcome. Median case age was 31 (range 16–69) years; 59 (52%) were male. The algorithm using only ICD-10 codes detected 69 visits; 59 met the case definition (PPV, 86%). The algorithm using only key words in the triage note detected 212 visits; 85 met the case definition (PPV, 40%). Applying an algorithm that required both ICD-10 and key words in the triage note detected 32 visits; 30 met the case definition (PPV, 94%).
CONCLUSIONS: Examining ESSENCE data by using marijuana-specific ICD-10 diagnostic codes more accurately identified ED visits for marijuana-associated adverse health outcomes, compared with key word search terms. ESSENCE as a tool for syndromic surveillance can be used to monitor adverse health effects of recreational-associated marijuana use.