Matt Harmon

181 Chapter nine To study the relationship between prophylactic antibiotic use on the risk of infection while accounting for mortality in the course following cardiac arrest, a linear mixed logistic regression model was used, with the participating hospitals as a random intercept. ‘A priori, age, SOFA score day 1, asthma/COPD, time from cardiac arrest to ROSC, and immunosuppression (defined as either an immunosuppressive condition (i.e. HIV, hematologic or other malignancy, alcoholism or IV drug abuse), or the use immunosuppressive medication were selected as confounders. The variables were sequentially added to the model to analyze the relationship between prophylactic antibiotics and pneumonia in the first 14 days of ICU admission. Covariates remained in the model when the Likelihood Ratio test was significant, p ≤ 0.05. The Likelihood-Ratio test compares nested models for which the parameters of one model are a subset of the second model. Collinearity diagnostics were performed using the variance inflation factor to check for variable independence. In this model, mortality is not accounted for, as the exact time of death was not known. We made the assumption that mortality in the first 3 days of ICU admission is unlikely to be affected by prophylactic antibiotic use, and we consider the mortality in the first 3 days of ICU admission as a competing risk. Patients who died within 3 days of ICU admission could have been part of the outcome with or without an infection. To estimate the robustness of the model we repeated the initial model with inclusion of patients who died in the first 3 days as having an infection and subsequently also repeated the model including these patients as not having an infection and compared the standardized beta values for the three models. A p-value less than 0.05 was considered significant. All analyses were performed in R (version 3.1.1). Results Patients Of the 950 patients randomized in the TTM-trial in 36 centers, data on microbiology results was available for 696 patients in 23 centers. In this cohort, the mean age was 64 (12) and 568 (82%) were male (table 1 shows all baseline characteristics of patients). Of the 696 patients, 186 patients (27%) developed a confirmed infection (either confirmed pneumonia or bacteremia). Within the confirmed infection group, 158 (89%) patients had pneumonia with confirmed microbiological culture and 28 (11%) patients had bacteremia. There were no differences in baseline characteristics between patients with a confirmed infection compared to patients without a confirmed infection (patients with suspected pneumonia and no infection).