Methods
Setting
The study was performed at Children's Mercy Hospitals and Clinics (CMH), a freestanding 317-bed tertiary care children's hospital located in Kansas City, Missouri. The study was approved by the CMH Institutional Review Board.
Study Design
Chart review was conducted for all patients with a blood culture positive for fungal growth from January 1, 2000, through December 31, 2010. Patients more than 18 years old at insertion and those who had central lines placed either as an outpatient or at a facility other than CMH were excluded. Central lines were identified as any CVC and were categorized as permanent or temporary. Permanent catheters were defined as tunneled or implanted CVCs, while temporary catheters were defined as peripherally inserted central catheters or subclavian, femoral, and internal jugular catheters. Patients who had midline catheters and neonates with umbilical venous lines were excluded. Those who had multiple blood cultures positive for yeast were included if the subsequent infection occurred (a) more than 8 weeks after the initial episode, (b) involved a different central line than the original, and/or (c) occurred during a hospitalization distinct from the initial inpatient stay.
Control subjects were selected via a random number generator from the CMH central line insertion database, a comprehensive database of all patients with central catheters placed at our institution since January 1, 2000. Prior to control selection from the database, all subjects with Candida CLABSI were removed (as the database contained all positive subjects as well). An underlying diagnosis of bacteremia did not preclude individuals from serving as controls. Each year was sampled independently to provide a representative sample. We planned to evaluate 3 controls for each Candida CLABSI case. Eligible controls met the same criteria as positive subjects except that they did not have a positive culture. Those potential controls with more than 1 central line placed during an inpatient hospital stay were eligible to be counted as controls only once per inpatient hospitalization, although all selected controls were eligible for future inclusion as such within a given year (provided a new central line had been placed during a separate inpatient hospital encounter). Controls from a previous year were also eligible to be counted as controls in subsequent years (again, provided that a new central line had been placed during a separate inpatient hospitalization).
Characteristic variable data were obtained from positive subjects starting 30 days prior to the identification of the Candida CLABSI. Corresponding data for controls were obtained over the time period from 30 days prior to and including the date of CVC removal or hospital discharge, whichever occurred first. Data collected included underlying diagnoses, type(s) of CVCs present, and whether a white blood cell count and platelet count had been performed prior to Candida CLABSI diagnosis. Data on length of antibiotics as well as receipt (and, if so, the corresponding duration) of parenteral nutrition, systemic corticosteroids, and blood transfusions prior to diagnosis were also obtained. Presence of indwelling foreign bodies, including endotracheal tubes, foley catheters, gastrostomy tubes, ileostomy or colostomy stomas, and neurosurgically implanted shunts was confirmed. Results from echocardiography, ultrasound of lines, cerebrospinal fluid evaluation, osteoarticular imaging, eye examination and other computed tomography imaging were evaluated. Finally, we collected the survival and sequelae outcomes for each subject.
Data Analysis
Bivariate analysis was used to compare demographic and underlying comorbidity data between positive subjects and controls and to determine associations between potential risk factors and development of Candida CLABSI. Categorical variables were compared using the Fisher exact or χ test, while continuous variables were compared using the t test for independent samples. All variables with P ≤ .05 were included in the multivariate model.
Logistic regression was then used to find the best collection of variables to predict the development of Candida CLABSI. Demographic, underlying comorbidity variables, and year of central line insertion were controlled for by forcing them to enter and stay in the logistic regression model. The remaining risk factors that were significant in the bivariate analysis were included in the model using stepwise selection. Temporal trends of causative fungal pathogens and incidence of complications were analyzed using descriptive statistics, with frequencies and percentages calculated when applicable. All statistical calculations were performed using SPSS, version 18 (PASW Statistics).