Validation of an Algorithm to Identify Infective Endocarditis in People Who Inject Drugs.

Abstract

Infective endocarditis is associated with high morbidity and mortality. Currently, there is concern that the incidence of infective endocarditis associated with people who inject drugs (PWID) is increasing. However, it is difficult to monitor population-wide trends in PWID-associated infective endocarditis, as there is no International Statistical Classification of Diseases, 10th Revision (ICD-10) code for injection drug use. To address this barrier, we sought to develop a validated algorithm using ICD-10 discharge diagnosis codes. We constructed a cohort of patients whose hospital discharge diagnosis included infective endocarditis. We reviewed 100 patients with incident infective endocarditis from 2014 to 2016 for their infective endocarditis and injection drug use status. We calculated the operating characteristics for algorithms constructed using permutations of ICD-10 codes associated with injection drug use. We repeated this analysis in a cohort of 100 patients with incident infective endocarditis from 2009 to 2011 to examine the temporal stability of the operating characteristics of each algorithm. We found that a combination of hepatitis C virus, drug use, and mental/behavioral disorder codes yielded the highest sensitivity (93%) and positive predictive value (83%) of the algorithms analyzed. We have described the first algorithm, validated against chart review data, for identifying PWID-associated infective endocarditis cases using ICD-10 codes. The high sensitivity and positive predictive value indicate that this algorithm can be used for surveillance and research with confidence. This algorithm will enable researchers to examine epidemiological trends in PWID-associated infective endocarditis.