Methods: We conducted a retrospective analysis of all hospitalized patients diagnosed with CBSI at a large tertiary care hospital between 2002 and 2015. Data included demographics, comorbidities, laboratory parameters, and outcomes. Univariate and multivariable logistic regression analyses were used to build the predictive model.
Results: Of 1,873 cases of CBSI, 47 patients were identified to have CE. The most commonly isolated species were C. albicans (59.6%) followed by C. parapsilosis (16.2%). On univariate analysis, preexisting valvular disease (7.95, 95% CI [3.16, 20.02]) was associated with a higher risk of CE (p<0.05). Factors such as isolation of C. glabrata (0.17, 95% CI [0.04, 0.68]), hematologic malignancy (0.09, 95% CI [0.01, 0.68]), and total parenteral nutrition (TPN) (0.40, 95% CI [0.17, 0.95]) were all associated with a lower risk of CE. In multivariable modeling, the factors of valvular disease (5.05, 95% CI [1.77, 14.43]), isolation of C. glabrata (0.19, 95% CI [0.05, 0.80]), hematologic malignancy (0.09, 95% CI [0.01, 0.66]), and total parenteral feeding (0.43, 95% CI [0.17, 1.09]) remained significant. The final model had a C-statistic of 0.82. The crude 90-day mortality for CE was 48.9%, similar to the overall CBSI mortality of 42.1%.
Conclusion: In a population of patients with CBSI, previous valvular disease was the only factor associated with a greater risk of development of CE. Use of TPN, hematologic malignancy, and isolation of C. glabrata were protective factors. A predictive model may reduce the need for expensive and sometimes invasive diagnostic imaging such as trans-esophageal echocardiography, as a subset of patients may be at low enough risk for CE not to warrant them.
K. S. Foong,
A. S. Salazar, None
K. Hsueh, None
C. Lin, None
W. Powderly, None
A. Spec, None