Matt Harmon

186 Chapter nine The effect of prophylactic antibiotics on the incidence of pneumonia and bacteremia Within the scope of this sub-study, five centers used SDD, seven centers used prophylactic cephalosporin treatment, one center used amoxicillin/clavulanic acid and two used a β -lactam that was not further specified (supplemental table 5 and 6 respectively show the site-specific prophylaxis and total antibiotic use in patients, supplemental table 7 shows the distribution of pathogens per center). We analyzed the incidence of infections in centers using antibiotic prophylaxis compared to centers that do not. Baseline characteristics between groups can be seen in supplemental table 8. In univariate analysis, patients in centers using prophylactic antibiotics had a lower incidence of both pneumonia (141 (41%) vs. 190 (54%), p=0.001) and bacteremia (4 (1%) vs. 24 (7%), p=0.0001) compared to patients in centers not receiving antibiotic prophylaxis (table 2). In the initial logistic regression model, prophylactic antibiotics was associated to a lower risk of pneumonia in the first 14 days of ICU-admission (adjusted odds ratio (aOR) 0.59, 95% confidence interval (CI) 0.43-0.79, p = 0.0005). After correcting for confounders, the relationship between prophylaxis and pneumonia remained significant (aOR 0.64, 95%CI (0.46-0.90), p = 0.01) (table 3). To adjust for potential clustering within centers, an additional model was run correcting for a potential cluster effect. In this model, the risk of acquiring pneumonia in the first 14 days of ICU admission further declines with prophylactic use of antibiotics, but statistical significance is lost. (aOR 0.55, 95%CI (0.20 – 1.47), p = 0.22) As the exact time of death was not noted and we wanted to account for the effect of mortality on outcome, we made the assumption that death within 3 days of ICU admission after cardiac arrest is likely not due to infection. The model was repeated with classification of all patients with early deaths as having an infection and as not having an infection. There was no difference between these models (Supplemental table 9 and figure 1). Discussion Incidence of nosocomial infections following cardiac arrest was high in this study, primarily driven by pneumonia. Nosocomial infection was most often due to gram-negative bacteria, although Staphylococcus aureus was the most commonly cultured micro-organism. Target temperature did not impact the distribution of pathogens in terms of gram-negative or gram-positive organisms. Antibiotic prophylaxis was significantly associated with a lower incidence of pneumonia and bacteremia in cardiac arrest patients in a model uncorrected for center effect.