Objective-To analyse hospital acquired infective endocarditis cases with respect to age, sex, clinical, laboratory, and echocardiographic features, predisposition, complications, surgery, mortality, and diagnostic criteria. Design-Prospective cohort study. Setting-Teaching hospital. Patients-A series of 200 patients with infective endocarditis presenting over 11 years, 168 with native valve infective endocarditis, of whom 22 acquired this infection in hospital. Results-22 (14%) of the 168 cases of native valve infection were hospital acquired. The most common pathogens were staphylococci (77%). Two thirds of patients had no cardiac predisposition; one third had end stage renal disease. The most common source of infection was vascular access sites (73%). Eleven patients died. In 11 cases, infective endocarditis was proven pathologically (six at necropsy, five during surgery) and analysis of these showed that 45% were classed as probable by the Beth Israel criteria, 73% as definite by the Duke criteria, and 91% as definite by our suggested modifications of the Duke criteria. Figures for the 11 cases not proven pathologically were 27%, 73%, and 91%, respectively. Five of the 22 cases (22%) were rejected by the Beth Israel criteria but none were rejected by the Duke criteria with or without our modifications. Conclusions-Hospital acquired infective endocarditis is diYcult to diagnose. The Duke criteria have improved diagnostic sensitivity and our modifications have improved it further. Mortality is high but has been reduced by surgery. This serious infection could, in many cases, be prevented by improved care of intravascular lines and prompt removal when obviously infected. (Heart 1998;79:442-447)