Predictive variables for central venous catheter infection in leukemia patients

Central venous catheters (CVC) have become the mainstay for venous access in patients with malignant disease. These catheters provide access to the venous system for delivery of antibiotics, fluids, hyperalimentation, chemotherapy, and blood products. Cancer patients, who are immunosuppressed, have an increased risk for infection from these catheters. The research question for the study was as follows: to what extent do six specified environmental variables predict the development of central venous catheter infection in patients with leukemia? These environmental variables included number of blood draws and blood products, catheter dwell time, frequency of CVC dressing changes, days of hyperalimentation, and days of neutropenia.

A non-experimental retrospective design was used to assess the extent to which specified independent variables were predictive of central venous catheter infection in patients with leukemia. A questionnaire designed by the investigator was applied to 134 medical records. Discriminant function analysis was the statistical test chosen to determine if any one of the six variables would be predictive of catheter infection. However, the assumptions of equal covariance and multivariate normality for discriminant function were not met. Therefore, the Pearson chi square was used to test whether the groups (no infection versus infection) differed on the demographic variables and a t-test was done to test significant differences of the predictor variables between the two groups. The two predictor variables that were significant were catheter dwell time ($p=.004$) and frequency of family CVC dressing changes ($p=.0001$).