Environmental except exposure was assumed to be constant between two screenings. Hence, patients who acquired P. aeruginosa had several environmental pressure profiles (including patient colonization pressure and tap water colonization pressure) allowing a comparison with patients who did not acquire P. aeruginosa.Statistical analysisQuantitative variables were compared using the Student’s t-test or Wilcoxon test according to the distribution of data. Qualitative variables were compared using the Chi2 or Fisher’s exact test. A marginal logistic regression model accounting for repeated measurements [20] was used to assess the relationship between environment, antibiotic pressure and P. aeruginosa acquisition each day, and the results were expressed as odds ratios (OR) and 95% confidence intervals (CI).
Univariate analysis of P. aeruginosa acquisition included: (i) fixed variables for patient characteristics at admission; (ii) longitudinal data on patient/tap water colonization pressures, as described above, on the cumulative number of days since admission with a nasogastric tube (which was selected to represent invasive devices as it is strongly associated with the use of other invasive devices in our clinical practice) or with antibiotics classified as active or inactive against P. aeruginosa. Selection of the environmental exposure index (previous or cumulated colonization pressure) was based on Akaike criteria [21]: patient/tap water colonization pressure on the previous day was finally introduced in the multivariate analysis.
Quantitative data were analyzed as categorical variables when the log-linearity assumption was not followed. All factors with a P-value < 0.20 in univariate analysis were selected for multivariate analysis. In multivariate analysis, the factors related to patient/tap water colonization pressures, that is, "patients on the same ward", "tap water from the ICU", "tap water from the shared rooms" or antibiotics were first introduced together and forced in the model. Because wards are included in the ICU, only the most significant index among colonization pressure onto the ward or the ICU was selected for analysis purpose. Other factors were then introduced in a stepwise manner to control for confounding. According to our main objective, the final model looked for interactions Cilengitide between each of the three patient/tap water colonization pressures and antibiotic variables. A P-value of <0.05 was considered significant. Data were recorded prospectively with Epidata (3.1; Odense, Denmark). The model was fitted using the GENMOD procedure on SAS software (SAS Institute, Inc., Cary, NC, USA).