OBJECTIVE: The objective of this study was to determine the impact of empiric antifungal therapy on mortality in hospitalized, non-neutropenic patients with bloodstream candidemia. METHODS: Retrospective cohort study to evaluate the effect of empiric therapy on mortality in patients with bloodstream candidemia. Medical and hospitalization history were collected from the medical charts of patients hospitalized from 2002 to 2004 with their first documented episode of candidemia. Patients were grouped by APACHEII scores and choice of empiric antifungal therapy (narrow spectrum: fluconazole or itraconazole; broad spectrum: amphotericin, caspofungin, or voriconazole). Impact of empiric antifungal therapy was assessed using univariate techniques and multivariate logistic regression. RESULTS: A total of 81 patients were evaluated in this study. Forty-four percent received broad spectrum, 46% received narrow spectrum, and 10% received no therapy. Overall mortality was 51%. APACHEII >20 and narrow spectrum empiric therapy were independent predictors of mortality (p=0.011). Mortality rates were similar in patients with Apache score >20 given broad (85%) or narrow spectrum (79%) empiric therapy. However, in patients with lower APACHEII scores (<20), mortality was significantly lower in patients started on broad spectrum (4%) compared to narrow spectrum (46%). After excluding patients who died while on therapy, average length of therapy was similar for both groups (11-12 days). CONCLUSIONS: Patients with candidemia with an APACHEII score <20 treated with broad spectrum antibiotics had lower mortality rates compared to those initially treated with narrow spectrum antifungal therapy.