Surgical site infections (SSIs) are healthcare-associated infections (HAIs) that contribute significantly to adverse clinical outcomes and increased healthcare costs.Reference Shepard, Ward and Milstone 1 Reported rates vary according to surgical procedure, spanning <1% for clean proceduresReference Agency 2 to 30% for some colorectal procedures.Reference Edwards, Peterson, Andrus, Dudeck, Pollock and Horan 3 – Reference Smith, Bohl and McElearney 6 Standardized surveillance has long been recognized as a minimum and necessary requirement for effectual prevention strategies,Reference Geubbels, Nagelkerke, Mintjes-De Groot, Vandenbroucke-Grauls, Grobbee and De Boer 7 – Reference Diaz-Agero Perez, Robustillo Rodela, Pita Lopez, Lopez Fresnena and Monge Jodra 9 and diminishing SSI rates have been noted following the implementation of surveillance programs.Reference Brandt, Sohr, Behnke, Daschner, Ruden and Gastmeier 10 Surveillance data have recently been used to evaluate appropriateness of surgical antimicrobial prophylaxis in the United States.Reference Berrios-Torres, Yi and Bratzler 11
In Australia, national guidelines for surgical antibiotic prophylaxis are widely available. 12 However, evaluation of pathogen-specific data concerning infections complicating the range of surgical procedures has not been performed to determine whether guidelines are concordant with local epidemiology.
In 2002, the Victorian Healthcare-Associated Infection Surveillance System (VICNISS) commenced monitoring of SSIs.Reference Russo, Bull and Bennett 13 , Reference Friedman, Russo, Bull, Richards and Kelly 14 The objectives of this study were (1) to examine rates of SSIs according to operative procedure groups and (2) to analyze trends over time for SSI rates, pathogens responsible for SSIs, and antimicrobial susceptibility of pathogens responsible for SSIs in Victorian hospitals for the period October 1, 2002, to June 30, 2013.
The VICNISS program was established to monitor a range of HAI processes and outcomes in Victorian hospitals, including SSIs.Reference Russo, Bull and Bennett 15 The SSI surveillance module is based on the National Healthcare Safety Network (NHSN), Centers for Disease Control and Prevention (CDC). 16 , Reference Horan, Culver, Gaynes, Jarvis, Edwards and Reid 17 For the current study, data collected between October 1, 2002, and June 30, 2013, were analyzed.
During the study period, all large Victorian public hospitals participated in the VICNISS program. Hospitals were required to monitor at least 2 surgical procedure groups for each 12-month period, with continuous surveillance mandated for coronary artery bypass graft (CABG) procedures and for hip/knee arthroplasty procedures if >50 procedures were performed annually. Submission of data for other procedures and participation by private hospitals was voluntary.
Consistent with NHSN/CDC criteria, 16 SSIs were classified as superficial, deep, or organ/space. Microbiologically confirmed SSIs associated with the following operative procedure groups were analyzed: (1) CABG with both chest and donor site incisions or CABG with chest incision only, (2) Caesarean section, (3) hip and knee prostheses, (4) colorectal surgery, (5) other abdominal surgery (eg, gastric surgery, cholecystectomy, herniorrhaphy, appendectomy, or small bowel surgery), (6) cardiac surgery (non-CABG), (7) hysterectomy (abdominal or vaginal), and (8) vascular surgery.
Patients were monitored for signs of SSI until discharge from the acute-care facility. If patients were readmitted to the hospital with SSI within 12 months of the procedure, these infections were recorded. SSI rates were calculated by dividing the number of SSIs by the number of specific operative procedures and multiplying the results by 100. Risk-indexing was performed according to NHSN methodology.Reference Horan, Culver, Gaynes, Jarvis, Edwards and Reid 17 , Reference Friedman, Bull, Russo, Gurrin and Richards 18
Data Collection and Validation
Data were collected by infection prevention staff affiliated with participating healthcare facilities. All were trained regarding case definitions and surveillance methods by VICNISS Coordinating Centre staff. When a pathogen was isolated, the organism and antimicrobial susceptibilities were recorded using reported antimicrobial breakpoints (ie, susceptible, resistant, or intermediate susceptibility). All data were submitted via a standardized data collection tool, with electronic data submission available from 2010. Prospective monitoring was performed using patient records and microbiology and radiology results.
Reporting and Statistical Analysis
Multilevel mixed-effects Poisson regression was used to model counts of pathogen-specific SSIs for procedure categories where the aggregate number of procedures per time period formed the exposure. In the absence of explicit data regarding individual hospital characteristics (eg, case mix), the hospital identifier was modeled as a random effect. Effect size was quantified as the risk ratio (RR). A linear relationship between SSI count and time variables was assumed. The validity of this assumption was confirmed by comparing a base linear Poisson model with alternate nonlinear quadratic and cubic relationships using analysis of residual plots, a likelihood ratio test, and assessment of adjusted coefficient of determination (R2). In each instance, residuals were minimized and R2 was maximized in the linear model compared to either quadratic or cubic models.
Analysis was confined to pathogen and procedure categories for which sufficient annual data were available. Given a limited number of events for some pathogens, detailed analysis was restricted to the 6 most highly ranked SSI pathogens. In all instances, P<0.05 was considered significant. Analysis was undertaken using Stata version 13 (StataCorp, College Station, Texas).
No patient-identifying data were captured for the purposes of the study, and all hospital-level data were deidentified. This study was a quality-assurance audit without direct impact upon patient care; therefore, an ethics review was not required.
Between October 1, 2002, and June 30, 2013, a total of 183,625 procedures (164,721 patients) were monitored and 5,123 SSIs were reported, corresponding to an overall crude SSI rate of 2.8 per 100 procedures. Of these, 3,318 (64.8%) were microbiologically confirmed SSIs. A total of 81 healthcare facilities participated in surveillance activities. Of patients with infection, 2,942 (57.4%) were female and 2,164 (42.2%) were male (patient sex was not recorded in 17 instances). Median (IQR) age at surgery was 63 years (range, 40–74 years).
Trends in SSI Rates
Across all procedures and SSI categories, every 1-year increase across the observation period was associated with a 9% decrease in the risk of SSI (RR 0.91, 95% confidence interval [CI] 0.90–0.92). Of the 5,123 SSIs, 2738 (53.5%) were superficial, 1,210 (23.6%) were deep incisional, and 1,174 (22.9%) were organ/space infections, corresponding to 1.49, 0.66, and 0.64 infections per 100 procedures, respectively. Significant reductions in SSI were observed for all infection categories: every 1-year increase across the observation period was associated with an 11% decrease in the risk of superficial SSI (RR, 0.89; 95% CI, 0.88–0.90), a 9% decrease in the risk of deep SSI (RR, 0.91; 95% CI, 0.90–0.93), and a 5% decrease in the risk of organ/space SSI (RR, 0.95; 95% CI, 0.93–0.97).
Crude SSI rates (per 100 procedures) were highest for vascular surgical procedures (11.79; 95% CI, 8.70–15.64) and lowest for hysterectomies (2.01; 95% CI, 1.68–2.39). Colorectal procedures were associated with a crude SSI rate of 8.67 per 100 procedures (95% CI 8.06–9.31). Table 1 provides the number of monitored procedures and procedure-specific SSI rates by risk index. For procedures in which continuous surveillance was performed (CABG and orthopedic procedures), the majority were classified as risk index category 1. SSI rates for risk index category 2 were ~50% higher than rates for risk index category 1.
NOTE. CABG, coronary artery bypass graft; CI, confidence interval.
a SSI risk index calculated as sum of the following factors (if present at time of operation): (1) operation lasting more than the duration cut point hours, where the duration cut point is the approximate 75th percentile of the duration of surgery in minutes for the operative procedure, (2) contaminated (class 3) or dirty/infected (class 4) wound class, and (3) American Society of Anaesthesiologists physical status classification of 3, 4, or 5. Risk index reduced by 1 unit if laparoscopic procedure performed.Reference Horan, Culver, Gaynes, Jarvis, Edwards and Reid 17
b Includes “chest and donor” and “chest only.”
c Continuous surveillance performed at participating hospitals during the study period.
d Periodic surveillance performed by selected hospitals during the study period.
e Includes gastric surgery, cholecystectomy, herniorrhaphy, appendectomy, and small bowel surgery.
f Includes abdominal and vaginal hysterectomies.
g Includes femoropopliteal and femorotibial bypass, abdominal aortic aneurysm repair, and vascular surgery not otherwise specified.
h Overall (crude) data.
Of all microbiologically confirmed SSIs, 1,174 (35.4%) were associated with orthopedic procedures, 827 (24.9%) were associated with CABG, 414 (12.5%) were associated with colorectal procedures, 131 (3.9%) were associated with other abdominal surgeries, 490 (14.8%) were associated with Caesarean sections, 87 (2.6%) were associated with other cardiac procedures, 66 (2.0%) were associated with hysterectomies, and 31 (0.9%) were associated with vascular surgery.
Trends in Pathogens Responsible for SSIs
Table 2 demonstrates the relative frequency (ranking) of infecting pathogens according to surgical procedure group, based upon the 10 most frequent SSI pathogens for all procedures. Across all procedures, S. aureus was identified as the most frequent pathogen; it was responsible for 46.5% of microbiologically confirmed SSIs.
NOTE. CABG, coronary artery bypass graft; CNS, coagulase-negative staphylococci.
a Includes Pseudomonas aeruginosa and Pseudomonas spp. not elsewhere classified.
b Zero SSIs recorded.
c Equal ranking with other pathogen(s) for procedure group.
d Not ranked in top 10.
e Includes Enterococcus faecalis, E. faecium and Enterococcus spp. not elsewhere classified.
f Includes Enterobacter cloacae, E. aerogenes and Enterobacter spp. not elsewhere classified.
g Includes Klebsiella pneumoniae and K. oxytoca.
For SSIs following CABG and orthopedic procedures, Pseudomonas spp. were the second most frequently reported pathogen, and coagulase-negative staphylococci (CNS) were ranked third. Coagulase-negative staphylococci were responsible for 5.9% and 8.9% of microbiologically-confirmed SSIs following CABG orthopedic procedures, respectively.
Of pathogens responsible for SSI following colorectal procedures, 28.7% were S. aureus, and 46.2% of tested isolates were methicillin-resistant S. aureus (MRSA). Of pathogens responsible for SSI following abdominal procedures, 42.0% were S. aureus, and 23.4% of tested isolates were MRSA.
Modeled annual time trends for pathogen-specific SSIs are provided in Table 3. Across all procedures, every 1-year increase was associated with significant reductions in risk of S. aureus, Pseudomonas spp., and E. coli infections (risk reductions of 12%, 7%, and 11%, respectively). Whereas annual risks for SSIs due to the majority of pathogens diminished during the study period, CABG procedures were associated with a 14% annual risk increase for SSIs due to CNS.
NOTE. CABG, coronary artery bypass graft; CNS, coagulase-negative staphylococci; RR, risk ratio; 95% CI, 95% confidence interval.
a Analysis confined to the 6 most highly-ranked SSI pathogens, excluding indicated instances where insufficient data were available for time-trend analysis.
b Includes Pseudomonas aeruginosa and Pseudomonas spp. not elsewhere classified.
c Includes Enterococcus faecalis, E. faecium and Enterococcus spp. not elsewhere classified.
d Includes Enterobacter cloacae, E. aerogenes and Enterobacter spp. not elsewhere classified.
Trends in Antimicrobial Resistance of Pathogens Responsible for SSIs
Table 4 summarizes the proportion of infecting S. aureus, Pseudomonas spp., E. coli, and Enterococcus spp. isolates with antibiotic resistance. Of all S. aureus isolates tested for vancomycin resistance (n=1,131), no resistant or intermediate-resistant isolates were reported. Of all isolates tested for methicillin resistance, 31.6% were confirmed to be MRSA. Across all procedures, an 18% average annual risk reduction in SSI due to MRSA was observed.
NOTE. RR, risk ratio; 95% CI, 95% confidence interval.
a Insufficient data available for time-trend analysis.
b Includes Pseudomonas aeruginosa and Pseudomonas spp. not classified elsewhere.
c Includes Enterococcus faecalis, E. faecium and Enterococcus spp. not classified elsewhere.
Low rates of ciprofloxacin, gentamicin, and imipenem resistance were reported in Pseudomonas isolates responsible for SSIs (0.49%, 0.48%, and 0.84%, respectively). Ceftazidime resistance was reported in 7.7% of Pseudomonas isolates, but no significant trend over time was demonstrated.
Of E. coli isolates responsible for SSIs, gentamicin resistance was reported in 7.6% of these cases, ciprofloxacin resistance was reported in 6.1%, and ceftriaxone resistance was reported in 15.2% of tested isolates. The average risk of SSIs due to ceftriaxone-resistant E. coli increased by 37% annually during the study period.
Of Enterococcus isolates responsible for SSIs, vancomycin resistance was reported in 27.5% of these cases. A nonsignificant trend toward increased vancomycin resistance in Enterococcus isolates was observed during the study period.
Our findings represent the first detailed evaluation of SSI rates and pathogens responsible for SSIs in Australian patients. S. aureus was the most frequent infecting organism, and a diminishing rate of infection was observed for the majority of pathogens over the 11-year study period. These findings are consistent with international reports that suggest S. aureus to be the most frequent infecting organism across a range of surgical procedures,Reference Berrios-Torres, Yi and Bratzler 11 , Reference Takesue, Watanabe and Hanaki 19 even in the context of reductions in SSI rates for CABGs and orthopedic procedures over the last decade.Reference Alasmari, Tleyjeh and Riaz 20 , Reference Skramm, Saltyte Benth and Bukholm 21
Contrary to time trends for the majority of pathogens, we observed a significantly increasing incidence of infections due to CNS following CABG procedures and a notable proportion of orthopedic infections (8.9%) due to CNS. The reasons for these trends are not clear, but some have proposed that infection prevention measures targeting S. aureus may not be as effective for reducing risk for CNS SSI and that additional measures (eg, environmental decontamination) may be required to reduce CNS SSI risk.Reference Yavuz, Tarcin and Ada 22
Antimicrobial-resistant (AMR) pathogens responsible for SSIs displayed significant trends over time. We observed a reduction in infections due to MRSA, recognizing that widespread reductions in bloodstream infections due to S. aureus and MRSAReference Worth, Spelman, Bull and Richards 23 and low rates of MRSA nonsurgical wound infection in smaller hospitalsReference Bennett, Bull and Dunt 24 have previously been reported in Victoria. The factors contributing to a reduction in MRSA SSIs are not clear and require further study. We also observed an increasing incidence of ceftriaxone-resistance in E. coli isolates responsible for SSIs. AMR Gram-negative infections, including those due to extended-spectrum β-lactamase-producing (ESBL) organisms, have increasingly been reported elsewhere over the past decade,Reference Takesue, Watanabe and Hanaki 19 and our findings are consistent with this trend. In contrast, we observed low numbers of carbapenem-resistant isolates responsible for infection, with imipenem resistance reported for only one Pseudomonas isolate. This finding indicates that carbapenem-resistant Enterobacteriacae (CRE) have not contributed significantly to SSI trends to date in Victoria and is quite different from findings reported in JapanReference Takesue, Watanabe and Hanaki 19 and the United States,Reference Sievert, Ricks and Edwards 25 where CRE have emerged as significant pathogens. The observed low rate of ciprofloxacin resistance in E. coli isolates responsible for SSIs is consistent with restricted use of fluoroquinolone agents in Australia.Reference Cheng, Turnidge, Collignon, Looke, Barton and Gottlieb 26 Ongoing targeted surveillance is required in Victoria to ensure that emerging resistant pathogens are detected in a timely manner. A trend toward an increased proportion of Enterococci with vancomycin resistance (VRE) was observed during the study period, likely a reflection of VRE emerging as endemic in many Victorian centers.Reference Worth, Slavin, Vankerckhoven, Goossens, Grabsch and Thursky 27 , Reference Johnson, Ballard and Grabsch 28
These findings are relevant to recommendations for antimicrobial prophylaxis prior to surgical procedures. It is appropriate to target a broad spectrum of likely SSI pathogens in settings in which SSI rates are high or in settings in which SSI rates are low but catastrophic consequences would result from infection (eg, sternal wound infections following CABG procedures). Review of pathogens responsible for SSIs is necessary in the formulation of local guidelines for antibiotic prophylaxis. For example, we identified >13% of SSIs following colorectal procedures to be due to MRSA. Current national guidelines do not recommend the use of prophylactic antibiotic agents with anti-MRSA activity. 12 While the use of vancomycin prophylaxis may need to be considered, preoperative MRSA screening and decolonization may be more appropriate, given the potential for routine vancomycin administration to contribute to further increases in already rising rates of SSI due to VRE in our region. This approach is consistent with international consensus guidelines.Reference Bratzler, Dellinger and Olsen 29
Risk-adjustment is necessary for reporting SSI surveillance data,Reference Worth, Bull and Richards 30 and we applied risk-indexing according to criteria established by the NHSN/CDC (Table 1).Reference Horan, Culver, Gaynes, Jarvis, Edwards and Reid 17 These data confirm that crude SSI rates do not reliably reflect trends or enable interhospital comparison.Reference van Dishoeck, Koek, Steyerberg, van Benthem, Vos and Lingsma 31 For CABG and orthopedic procedures, although our crude SSI rates were similar to adjusted rates for risk index category 1, a large number of patients were classified as risk index category 2, where SSI rates were ~50% higher. A crude rate would not adequately convey this increase in risk. Increasingly, the standardized infection ratio (SIR) is used for adjustment of hospital-level data,Reference Rioux, Grandbastien and Astagneau 32 , Reference Gustafson 33 despite criticisms that this may be an over-simplification and not a meaningful indicator of infection prevention initiatives implemented in specialized surgical units.Reference Birnbaum, Zarate and Marfin A. 34 Looking ahead, we believe that SIR may hold merit for reporting Victorian data, provided the required data elements can be feasibly collected by participating centers.
Limitations of this study include the fact that surveillance activities were not performed continuously in all participating hospitals. Orthopedic and CABG surveillance were performed continuously, but reporting of other procedures may have occurred periodically at some centers. For example, crude SSI rates for vascular surgical procedures were high. Although this finding is comparable to rates reported elsewhere,Reference Pounds, Montes-Walters and Mayhall 35 periodic surveillance was performed for these procedures, and the potential for ascertainment bias must also be considered. Notwithstanding this limitation, our strategy encompassed all public hospitals in Victoria, and our data are very likely to reflect trends in the region. Secondly, factors potentially responsible for diminishing SSI rates were not evaluated. It is well-recognized that reductions in HAIs are observed following commencement of formal surveillance activities.Reference Gastmeier, Schwab, Sohr, Behnke and Geffers 36 For the current study, SSI risk-reduction strategies potentially contributing to time trends, such as hospital infection prevention practicesReference Bull, Wilson and Worth 37 or use of appropriate surgical antibiotic prophylaxis,Reference Bull, Russo, Friedman, Bennett, Boardman and Richards 38 were not analyzed. Thirdly, pathogen-specific data were insufficient to permit modeling by surgical risk category. A larger national dataset would allow more detailed analyses. Furthermore, recording of pathogen and antimicrobial susceptibility was reliant upon availability of laboratory reports to infection prevention staff. Our data may therefore underestimate antimicrobial resistance. Looking ahead, this process could be refined by automated extraction of laboratory data. Finally, rare but significant events were not captured (eg, molecular mechanisms of antimicrobial resistance). A national, laboratory-based surveillance system would more appropriately meet this need.
In summary, internationally accepted SSI surveillance methods have been successfully and sustainably implemented in Victoria, Australia. Over an 11-year period, a diminishing rate of SSI was observed for the majority of pathogens responsible for SSIs, although infections due to ceftriaxone-resistant E. coli increased significantly. National standardized SSI data would enable timely identification of changes in epidemiology,Reference Worth, Bull, Thorpe and Richards 39 particularly time trends in risk-adjusted data and rare but significant events.
The authors acknowledge the infection prevention consultants in participating Victorian hospitals for the collection and submission of data.
Financial support: The VICNISS Coordinating Centre is funded by the Victorian Department of Health.
Potential conflicts of interest: All authors report no conflict of interest or financial disclosures.