The complex sternal site infection rate (1.3%) in our study is comparable with findings from the HAI surveillance system in Norway (1.1%, 2,440 CABG procedures)  and the NHSN system in the US (1.2%, 133,503 procedures) . Characteristics of patients and CABG procedures were similar across these three study settings in terms of compositions of gender, age and wound classification, and proportions of emergency procedures. However, a higher proportion of patients with ASA score of 4/5 (73%) was reported by the NHSN system, as compared to those from the Norwegian (27%) and our data (45%), indicating a higher severity of underlying disease among patients undergoing CABG surgery captured in the NHSN system. Furthermore, the complex sternal site infections reported by the NHSN excluded cases detected during post-discharge surveillance (cases detected at readmission were included), due to concern of inconsistency in post-discharge case finding across 293 participating hospitals. In contrast, both in-hospital and post-discharge detected infection cases were included in Norwegian and our data, where post-discharge SSI surveillance data were collected from a relatively small number of participating hospitals (six and three respectively).
Our data indicate the importance of Gram-negative bacteria as causative agents for SSIs following CABG procedures. As observed in three participating hospitals, proportions of all SSIs (superficial and complex at both the sternal and harvest sites) caused by Gram-negative organisms increased from 38% in 2003 to 62% in 2009, despite a reduction to 42% in 2012. Harrington and colleagues  reported that Gram-negative bacilli were isolated from 18% of SSIs after CABG surgery, based on surveillance data collected during 1998–2001 in Victoria, Australia, while the corresponding figure based on our surveillance data over 2003–2012 was 44%. Further analysis of pathogens causing complex sternal site infections in our study showed Gram-negative bacteria were responsible for 27% of these serious infections, which is consistent with recently published NHSN data (Gram-negative organisms accounting for 34% of pathogens in complex SSIs after CABG surgery) . An earlier study conducted in the US during 2000–2004 found 11% of sternal SSIs were caused by Gram-negative aerobes . The apparently increasing proportion of CABG surgical site infections caused by Gram-negative organisms has important implications for clinical practice. Widely used guidelines (for example, the Sanford Guide to Antimicrobial Therapy) recommend cefazolin, cefuroxime or vancomycin as perioperative prophylaxis for cardiovascular surgery . The current Australian Therapeutic Guidelines recommend three options for antibiotic prophylaxis in cardiac surgery: cefazolin alone, a combination of flucloxacillin and gentamicin, or a combination of vancomycin and gentamicin . Psedomonas aeruginosa is intrinsically resistant to first generation cephalosporins. Other major Gram-negative organisms reported in our study (e.g. Enterobacter spp. and Serratia marcescens) belong to a genus whose inducible beta-lactamase production is a common property . Given the increased prevalence of Gram-negative organisms associated with SSIs, we believe that cardiac surgeons should consult with their microbiologists to discuss local antibiograms prior to selecting surgical prophylactic antibiotics for patients undergoing CABG procedures.
Patient-based active surveillance data on CABG procedures provide an opportunity to investigate the long term trend in SSI rates. An increase in complex sternal site infection rates over 2003–2012 can be partially explained by variation in case mix of patients. There was a substantial increase in the proportion of patients with ASA score of 4/5 (indicating severe underlying disease) from 18% to 71% over the last decade. This might reflect the changing landscape of coronary intervention characterised by more patients with coronary artery disease undergoing treatment by percutaneous revascularisation (e.g. balloon angioplasty and stenting), while surgical revascularisation (CABG) is reserved for patients with more comorbid conditions and more severe underlying disease . This study provides evidence on the increased risk of surgical site infections driven by increasing severity of illness among the patient population electing for CABG surgery.
The relatively poor discriminatory ability of the NNIS risk index for CABG surgical site infections has been widely reported across various study settings [11, 14, 19, 25–27]. The underlying reason is that nearly all patients undergoing CABG surgery would have an ASA score ≥ 3 and have wound classification of clean or clean-contaminated, leaving the NNIS risk index differentiating patients just based on the procedure duration (top quartile vs. the rest). In this study, we re-categorised the ASA score into three groups (1/2, 3, and 4/5) to differentiate host susceptibility to infection for epidemiological purposes. Use of alternative risk scores (such as the EuroSCORE  and the Admission-specific Chronic Disease Scores ) has been associated with improved predictive performance for SSIs following CABG procedures. However, construction of these risk scores requires extensive and complex clinical data; their application to routine SSI surveillance data is subject to advancements of the underlying surveillance systems.
There is increasing interest in developing risk adjustment models for appropriate comparison of SSI rates following CABG surgery as part of publicly available hospital performance metrics. A recently published risk adjustment model based on the NHSN surveillance data included five predictors (ASA score, procedure duration, medical school affiliation, and interaction of age and sex) , achieving predictive performance of 0.62 (c-index) for complex sternal site infections after CABG surgery. A similar study (from Victoria, Australia) reported a predictive performance of 0.64 (c-index) for a model containing two predictors: diabetes, and body mass index (BMI > 35) . In our study, a logistic regression model with two independent predictors (ASA score; absence of documentation of antibiotic prophylaxis) reported a c-index of 0.59 in predicting complex sternal site infections following CABG surgery. It appears that inclusion of a common set of patient/procedural factors such as ASA score, procedure duration, diabetes status, obesity (BMI) and antibiotic prophylaxis status in a risk adjustment model would enhance our understanding of the extent to which these factors contribute to and predict the SSIs after CABG surgery. However, none of the models mentioned above was able to test this full set of potentially important predictors, due to lack of data on one or more factors in their respective HAI surveillance systems. As not reaching a c-index of 0.7, considered to be indicative of acceptable predictive performance, those risk adjustment models may not be sufficient for appropriate comparison of infection rates across hospitals.
There were some limitations associated with our HAI surveillance data. First, the standardised surveillance data on CABG surgical site infections were collected from a total of three public hospitals which had capacity to provide CABG surgery for all public patients in Queensland. However, patients undergoing CABG at private hospitals (accounting for approximately 40% of the total patients)  were excluded from surveillance, thus limiting our ability to generalise findings to the patient population in Queensland as a whole. Second, there was concern regarding inconsistency in detecting SSIs (particularly the superficial infections) at post-discharge surveillance, due to inter-hospital variation in terms of completeness of follow-ups and potentially unreliable diagnosis of infections based on patient self-reporting . This can be partly addressed by focusing on complex sternal site infections, which are more likely to be reported by patients at post-discharge surveillance and diagnosed by clinicians. Our data indicated 41% of complex sternal site infections were detected following discharge. In addition, complex sternal site infections have a greater consequence for patients. Third, procedure duration was not included as a potential risk factor for CABG surgical site infections due to concern with inconsistency in collecting this information, thus limiting our ability to assess its role in predicting surgical site infections.
With transition of the eICAT to a more sophisticated Multiprac© surveillance system in Queensland Health, which features automatic linkages to multiple patient demographic and clinical databases, additional data on relevant risk factors (e.g. diabetes status, isolated CABG procedures or in conjunction with valve replacements) and infection outcomes will be routinely collected. There is scope to further refine our logistic regression model for risk adjustment, and to assist ongoing monitoring, public reporting, and improving surgical site infection control practice in hospitals.