A pilot study to develop a prediction instrument for endocarditis in injection drug users admitted with fever

Objective: Seeking to evaluate the feasibility of a prediction instrument for endocarditis in febrile injection drug users (IDUs), we determined (1) the frequency percentage of IDUs admitted with fever diagnosed with endocarditis and (2) whether individual or combinations of emergency department (ED) clinical criteria (patient history, physical examination findings, and laboratory tests) are associated with endocarditis in IDUs admitted to rule out endocarditis. Methods: The ED and inpatient charts of all IDUs with a diagnosis of rule out endocarditis admitted at 3 urban hospitals in 2006 were reviewed. Screening performance of individual criteria was determined, and the most sensitive combination of criteria was derived by classification tree analysis. Results: Of 236 IDUs admitted with fever, 20 (8.5%) were diagnosed with endocarditis. Lack of skin infection, tachycardia, hyponatremia, pneumonia on chest radiograph, history of endocarditis, thrombocytopenia, and heart murmur had the best screening performance. The classification tree-derived best criteria combination of tachycardia, lack of skin infection, and cardiac murmur had a sensitivity of 100% (95% confidence interval, 84%-100%) and negative predictive value of 100% (95% confidence interval, 88%-100%). Conclusions: Using ED clinical criteria, a multicenter prospective study to develop an instrument for endocarditis prediction in febrile IDUs is feasible, with an estimated target enrollment of 588 patients.