Objective: To assess the utility of postoperative antibiotic exposure as an indicator of postoperative infection after coronary artery bypass graft surgery.
Design: We determined an optimal antibiotic exposure threshold by creating receiver operating characteristic curves.
Setting: Tertiary healthcare institution (United States); national sample (Israel).
Patients: 5,887 patients undergoing coronary artery bypass graft surgery.
Results: Postoperative antibiotic exposure with at least 9 days between the first and last dates of antibiotic administration, excluding the first postoperative day, had a sensitivity of 95% (261/276) and specificity of 85% (3,944/4,628) for identifying surgical-site infection, using as a gold standard surgical-site infections identified by conventional prospective surveillance or extrapolated from review of a sample of medical records. In contrast, using the same gold standard for surgical-site infections, the sensitivity of routine prospective surveillance alone was only 60%. The predictive value positive of the defined antibiotic exposure was 28% (261/945) for surgical-site infection and 60% (563/945) for any nosocomial infection. In the Israeli cohort, the sensitivity was 87% (74/85) and the specificity was 82% (735/898).
Conclusion: Antibiotic exposure of sufficient duration and timing was more sensitive than conventional methods in detecting nosocomial infection and required substantially less effort to collect. Although the predictive value positive for surgical-site infection was only moderate, the majority of individuals identified this way had a nosocomial infection.