Abstract Background Candida bloodstream infections (CBSI) continue to be associated with high mortality, despite changes in antifungal treatment and diagnostics. Methods All patients age 18 or greater with a first episode of CBSI by blood culture from 1/2002 to 1/2015, admitted to Barnes-Jewish Hospital, a tertiary referral hospital in St. Louis, MO, were included. We collected data on demographics, comorbidities, laboratory values, vital signs, indwelling devices, and medical treatments of interest from the electronic medical record. We analyzed the potential predictor variables using univariate logistic regression. Variables associated with mortality were considered for model inclusion. The final model was built using multivariable binary logistic regression. A predictive equation was created, and a receiver–operator curve (ROC) was calculated to determine the appropriate cut-off points and c-statistic. Results Of the 1873 episodes of CBSI identified, 789 (42%) resulted in death in 90 days. The variables included in this model were age (40–49: OR 0.463, 95% CI 0.291–0.736; 50–69: 0.542, 0.342–0.860; ≥70: 0.560, 0.400–0.785); history of CAD (1.616, 1.171–2.230), chronic liver disease (2.247, 1.327–3.806); maximum heart rate (1.496, 1.126–1.989) and temperature (0.537, 0.408–0.708); AST (1.817, 1.343–2.459) and platelet count (1.563, 1.178–2.073); the presence of ventilator (1.847, 1.321–2.582), urinary catheter (1.365, 1.008–1.847), two or more central lines (1.658, 1.020–2.694); removal of lines after positive culture (0.259, 0.181–0.370); ophthalmology consult during admission (0.441, 0.329–0.592); thoracentesis/chest tube (3.827, 1.550–9.448); diagnosis of secondary malignancy (2.131, 1.488–3.053); whether antimetabolites (2.119, 1.353–3.318), dapsone (4.507, 1.450–14.012), linezolid (1.605, 1.059–2.435), quinolones (1.384, 0.998–1.920) were ordered 90 days before positive culture. An ROC curve was calculated with an internal c-statistic of 0.806. Conclusion We created a risk predictive model for 90-day mortality in patients with CBSI, with 81% probability of predicting mortality. This model can lead to development of point-of-care applications to aid decision-making regarding escalation/de-escalation of care. Disclosures W. Powderly, Merck: Grant Investigator and Scientific Advisor, Consulting fee and Research grant Gilead: Scientific Advisor, Consulting fee Astellas: Grant Investigator, Research grant A. Spec, Astellas: Grant Investigator, Grant recipient