Factors other than the Duke criteria associated with infective endocarditis among injection drug users

Background: Modified Duke criteria were applied to consecutive injection drug users (IDUs) who were admitted to an inner-city hospital with a clinical suspicion of infective endocarditis, and the presence of any other clinical variables that were predictive of the presence of infective endocarditis was determined. Methods: Clinical data on consecutive IDUs who were hospitalized over 15 months in Vancouver were collected. Data included the admission history, and findings on physical examination and on initial laboratory investigations. Each subject's course in hospital was followed until discharge or death during the index hospitalization. Follow-up data collected included culture results, the interpretation of the echocardiogram and the discharge diagnosis. The modified Duke criteria were used for the diagnosis of infective endocarditis (definite, possible or rejected). Multiple logistic regression was used to determine what clinical variables (exclusive of the Duke criteria) available within 48 hours of presentation were independent predictors of infective endocarditis. Results: One hundred IDUs were enrolled. Fifty-one were female, and 58 were HIV-positive. Twenty-three met the modified Duke criteria for definite infective endocarditis, and 25 had possible infective endocarditis. IDUs with definite infective endocarditis were more commonly noted to have evidence of vascular phenomena (arterial embolism, septic pulmonary infarction, mycotic aneurysm, intracranial hemorrhage or Janeway lesions) (6 [26%]) than those who had possible endocarditis (1 [4%]). Those with definite infective endocarditis more often had multiple opacities on chest radiography (56% v. < 12%), and fewer had an obvious source of infection (52% v. 72% and 81% of possible and rejected infective endocarditis, respectively). Among febrile IDUs, definite endocarditis was highly associated with having no obvious source of infection (odds ratio 3.1 [95% confidence interval 1.1-8.7]) compared with febrile IDUs with an obvious source of infection. In similarly compared groups, the presence of hematuria, proteinuria or pyuria was also predictive of definite endocarditis (odds ratio 2.9 [95% CI 1.1-8.6]). Conclusions: Among IDUs, the modified Duke criteria are useful for classifying cases with definite infective endocarditis and rejecting cases without infective endocarditis. The classification of possible infective endocarditis is suitable for this population.