Nosocomial candidemia has become an important infection not only because of an apparently increasing incidence but also because of its high fatality rate. We studied 28 risk factors through a matched case-control study that included 88 pairs of patients hospitalized between July 1983 and December 1986. The strongest single risk factor found in the univariate analysis was the number of prior antibiotics administered: the exposure odds ratio (OR) was 12.50 when patients who received three to five antibiotics were compared with those who received none to two antibiotics. A multiple logistic regression analysis using a conditional likelihood method was performed to evaluate several risk factors simultaneously. The final model selected by a stepwise procedure included the following variables: number of antibiotics received prior to infection (OR, 1.73 per unit increase), isolation of Candida species from sites other than blood (OR, 10.37), prior hemodialysis (OR, 18.13), and prior use of a Hickman catheter (OR, 7.23). It remains to be shown in controlled clinical trials whether limiting the number of antibiotics or instituting prophylaxis and/or early treatment for high-risk patients will reduce the incidence of nosocomial candidemia.