Temperature box plot (median, 25%-75%, range) of the lowest intraoperative temperature for those without and with surgical site infections. Min indicates minimum; Max, maximum.
Walz JM, Paterson CA, Seligowski JM, Heard SO. Surgical Site Infection Following Bowel SurgeryA Retrospective Analysis of 1446 Patients. Arch Surg. 2006;141(10):1014-1018. doi:10.1001/archsurg.141.10.1014
We sought to determine whether the administration of preoperative antibiotics, intraoperative transfusion of blood products, and intraoperative hypothermia has any impact on the incidence of postoperative surgical site infections (SSIs) in a heterogeneous patient population undergoing bowel surgery.
From September through December 2002, data on 1472 patients undergoing bowel surgery at 31 academic medical centers in the United States were collected.
Patients were included in the analysis if they were older than 17 years of age and underwent any surgery involving the small bowel, colon, or rectum.
Main Outcome Measure
Postoperative SSI. Variables that might affect the risk for developing SSIs were analyzed using multivariate logistic regression analysis.
Perioperative transfusion (P = .04; odds ratio, 1.64), and the presence of any infection at the time of surgery (P = .05; odds ratio, 2.46) were independent risk factors for SSI. Patients with a lower intraoperative temperature nadir had a lower risk for SSI (P = .05; odds ratio, 1.33), although this difference is not clinically significant (35.8°C ± 0.8°C vs 36.0°C ± 0.9°C, P<.05). There was a trend toward statistical significance for wound class when added to the multivariate model (P = .09; odds ratio, 1.41). The administration of antibiotics within 120 minutes prior to incision or within 120 minutes prior to and 120 minutes after incision had no effect on SSIs in this patient population.
This study validates perioperative transfusion as an independent risk factor for SSI. The lack of effectiveness of perioperative antibiotic prophylaxis is surprising because it is discordant with the previous literature, and this finding needs further evaluation.
Surgical site infections (SSIs) represent the second most common type of nosocomial infection (20%) and are a major source of morbidity.1 The associated increase in treatment cost is estimated to be around $2000 to $4500 per case, and the postoperative length of stay is extended by 7 to 10 days.2- 4 Kirkland et al5 showed that the development of SSI results in a 225% increase in total direct costs per patient after laparotomy and a 77% increase after colon surgery. According to the National Nosocomial Infections Surveillance System Report,6 the incidence of SSI following bowel surgery ranged from 1% to 13.5% (median) for the period from January 1992 through June 2003. Most studies investigating the incidence and risk factors for SSI in bowel surgery have focused on a certain anatomic region within the gastrointestinal tract, (eg, gastroduodenal or colorectal surgery7) while other studies have analyzed mixed patient populations with respect to their wound class, ie, clean and clean-contaminated wounds.8,9 Surgical procedures for which there is currently no controversy over the need for antimicrobial prophylaxis to reduce SSIs include cardiac and noncardiac surgery like vascular surgery, general abdominal colorectal surgery, abdominal and vaginal hysterectomy, and surgery for trauma patients who sustained penetrating abdominal wounds.10,11 Other interventions that have been associated with a decreased risk for SSI include the avoidance of intraoperative hypothermia12 and the appropriate timing of antimicrobial prophylaxis prior to incision.13 Although the efficacy of preoperative mechanical bowel preparation to reduce the incidence of SSI has recently been questioned,14 perioperative transfusion has frequently been shown to increase the risk for SSI.15 We sought to investigate the relative contribution of these variables in combination to SSI following bowel surgery (small bowel, colon, and rectum) in a cohort of 1472 patients.
From September through December 2002, data on patients (24-50 patients per site) undergoing bowel surgery from 31 academic medical centers were collected (University HealthSystem Consortium Clinical Data Base). Patients were included in the analysis if they were older than 17 years of age and underwent any surgery involving the small bowel or colon (diagnosis-related groups 148 and 149 with and without complications). Patients were excluded if their hospital length of stay was beyond 3 standard deviations from the median. A total of 1472 patients met eligibility criteria. Wounds were stratified into clean-contaminated, contaminated, and dirty/infected based on the classification system developed by the National Academy of Sciences and the National Research Council.16 The circulating nurse in the operating room documented wound class electronically after verification by the attending surgeon. According to the Centers for Disease Control and Prevention, SSI was defined as an infection related to the operative procedure that occurred at or near the surgical incision within 30 days of the operative procedure. Clinical criteria for the diagnosis of SSI included any of the following: a purulent exudate draining from the surgical site, a fluid culture positive for organisms obtained from a surgical site that was closed primarily, the surgeon's diagnosis of infection, or a surgical site that required reopening. Surgical site infections were divided into superficial infection, deep incisional infection, organ/space infection, and wound disruption.17 All types of SSIs were included in the study. Antibiotic prophylaxis and the attempt to maintain intraoperative normothermia were standard in all participating medical centers; however, choice of antibiotics, use of heating devices (fluid warmers, convective heating blankets), and means of assessing intraoperative patient temperature were not standardized. Perioperative transfusion was defined as transfusion with packed red blood cells intraoperatively and up to 48 hours postoperatively.
Variables that might be associated with SSI were selected based on previous studies and analyzed using univariate logistic regression analysis. Variables were included in a multivariate model if the P value was less than .25 in the univariate analysis.18 Statistical significance in the multivariate model was defined as a P value less than .05. Models were performed with Statistica version 6 (StatSoft, Tulsa, Okla).
At the end of the study period, 1472 cases were included in the database. Of those, 26 cases with lengths of stay ranging from 38 to 93 days were excluded from the database by the University HealthSystem steering committee to achieve a more homogenous patient population. Wound categories were not documented in 42 of 1446 patients included in the analysis. The median patient age was 57 years with a range of 18 to 96 years. Demographics and patient characteristics are presented in Table 1. The rate of SSI for all wound categories combined was 8.7%. For patients with clean-contaminated wounds, the rate of SSI was 7.9%; for those with contaminated or dirty/infected wounds, the rates were 12.0% and 20.4%, respectively.
Comorbidities and risk factors that were identified in this patient cohort are listed in Table 2. Variables that were potentially associated with SSI and analyzed in the univariate model are presented in Table 3. Some of the factors that did not reach statistical significance in the univariate analysis were the use of supplemental oxygen in the postanesthesia care unit, American Society of Anesthesiologists physical status score, classification of surgery (elective vs emergent), the type of anesthesia (regional, general, or combined regional and general), and the administration of a preoperative bowel preparation. The remaining variables excluded from further analysis were as follows: age; sex; obesity; immunosuppression; and a history of cirrhosis, smoking, alcoholism, diabetes mellitus, or renal failure. Logistic regression with SSI as the dependent variable incorporated perioperative transfusion, the intraoperative temperature nadir, presence of any infection at the time of surgery, wound class, surgical time, and perioperative administration of antibiotics as independent variables. Factors that were independently associated with an increased risk for SSI in the multivariate analysis were perioperative transfusion of packed red blood cells and presence of any infection at the time of surgery (Table 4). Patients who had a lower intraoperative temperature nadir had a lower risk for SSI (Figure).
The duration of surgery (≤4 h vs >4 h) had no impact on the rate of SSI. Furthermore, the perioperative administration of antibiotics (within 120 minutes prior and 120 minutes after the incision) was not independently associated with a decreased risk for SSI (Table 4). This was true even when the statistical analysis was limited to administration of perioperative antibiotics either 120 or 30 minutes prior to incision (data not shown). There was a trend toward statistical significance in the multivariate analysis when wound class (clean-contaminated vs contaminated or infected) was factored in to the model (P = .09).
To determine whether the anatomic site of surgery has an influence on the risk for postoperative SSI, patients were separated into those who underwent large bowel procedures and those who had small bowel procedures. The multivariate analysis as described earlier was then repeated and was not statistically significant for small bowel procedures. For large bowel surgery, factors that significantly increased the risk for SSI were wound class (odds ratio, 1.61; confidence interval, 1.01-2.58), and perioperative blood transfusion (odds ratio, 1.82; confidence interval, 1.04-3.13).
The major findings of this study are that perioperative blood transfusion, a higher intraoperative temperature nadir, and presence of any infection at the time of surgery were associated with a risk for postoperative SSI. The fact that a higher intraoperative temperature nadir was associated with higher incidence of postoperative SSI is surprising considering that a number of recent trials have demonstrated that avoidance of intraoperative hypothermia reduces the incidence of postoperative wound infection.12,19 Although the difference between the lowest temperatures in the patients with and without wound infections is statistically significant, the clinical difference is negligible. A major difference between this analysis and previous studies is that patients with the use of immunosuppressive drugs; a recent history of fever; or evidence of infection, malnutrition, and bowel obstruction were not excluded from the database. In addition, previous studies that have examined this issue have used surgical wound infection as the outcome variable rather than SSI (which includes organ/space infection and wound disruption). Approximately 25% of the hospitals in this study routinely used convection warmers; however, specific data regarding the use of these devices were not collected. Because the difference in temperatures between the groups is so narrow and several prospective studies have shown the use of such devices is not associated with an increased risk for wound contamination,20- 22 it is unlikely that the use of these warmers caused an increase in SSI.
The use of allogeneic blood has been implicated with an increased risk for infection in patients following surgery and for critically ill patients in the intensive care unit.7,23 The blood that was used in the majority of these studies was not leukocyte depleted. Because leukocytes are thought to be the cause (at least in part) of transfusion-associated immunosuppression, the use of leukocyte-depleted blood should reduce the risk for infection. Indeed, several studies have shown that the use of high- efficiency leukocyte-depleting filters will reduce the risk for perioperative infections following gastrointestinal surgery.24,25 No data regarding the use of leukocyte-depleted blood were gathered in this study; consequently, we are unable to determine the extent to which blood without leukocyte depletion is implicated in SSI. Some authors suggest that the timing of perioperative transfusion and certain confounding factors have an impact on the incidence of SSI. Vamvakas and Carven26 were able to demonstrate that when severity of illness, difficulty of operation, and risk for wound infection were included in the statistical analysis, postoperative transfusion had only a marginal effect on the incidence of SSI and intraoperative transfusion had no effect on SSI.
The use of prophylactic antibiotics did not reduce the incidence of SSI in the present study. We chose to examine the effect of administration of antibiotics both 120 minutes before and 120 minutes after the incision because an analysis of antibiotic dosing just 120 minutes before incision resulted in a worse multivariate model than when the time interval of 120 minutes after incision was included. This finding is unexpected given that the evidence in support of perioperative antibiotic prophylaxis in colorectal surgery is unequivocal. In a recent publication from the Cochrane Database of Systematic Reviews, Song and Glenny27,28 concluded that antimicrobial prophylaxis is effective for the prevention of surgical wound infections after colorectal surgery; however, questions remain concerning the optimum antibiotic, timing, and duration of administration. Classen and coworkers13 prospectively studied the occurrence of surgical wound infections in 2847 patients undergoing elective clean or clean-contaminated surgical procedures at a large community hospital. They concluded that the administration of antibiotics 120 minutes prior to skin incision reduces the incidence of wound infection.13 However, their study differs from ours in several important aspects. Patients with clean wounds were included in the analysis and patients undergoing emergency surgery or with any preexisting infection were excluded from the study. The latter variable was independently associated with a higher risk for SSI in this investigation. Furthermore, previous trials on risk factors for SSI in bowel surgery were more restrictive with respect to the anatomic site of surgery criteria (eg, gastroduodenal or colorectal surgery).29- 31 This study was designed to look at best practices of individual institutions in a retrospective fashion. Therefore, no recommendations were made a priori on the timing and type of antibiotic administration prior to surgery as well as guidelines on redosing of antibiotics if the duration of surgery exceeded 120 minutes. Data on repeat antibiotic dosing were collected, but the number of patients who received a repeat dose was too small to analyze (data not shown). Taken together, these factors may explain in part the discrepancy of our findings compared with the current literature.
Several recent studies have shown that supplemental oxygen administration in the perioperative period reduces the incidence of SSI. Indeed, Greif and coworkers32 concluded that the rate of SSI in patients undergoing colorectal surgery may be reduced by as much as 50%. However, patients with a history of fever, infection, serious malnutrition, and bowel obstruction were excluded from the study. A more recent prospective randomized trial did not show a decreased incidence of SSI but did reveal potentially deleterious effects related to administration of a high fraction of inspired oxygen during the perioperative period.33 These findings confirm the result of our analysis, which did not show an association between oxygen administration in the postanesthesia care unit and the incidence of SSI.
Limitations of our study include the retrospective design and the fact that the determination of the patient's core temperature across the participating medical centers was not standardized. Coding of the wound class for a small number of patients (n = 46) was incomplete, which may have had a small effect on our calculation of the rate of SSI in this cohort. The fact that patients with extensive lengths of stay were excluded from the analysis should have minimal if any impact on the incidence of SSI in this patient population. It is uncommon for SSIs to occur beyond 2 to 3 weeks after surgery, and most infections will appear within several days to a week. After this point, the wound has most likely epithelialized and the subsequent risk for infection would be very low. Because of the retrospective design of this study, no data could be abstracted as to how the participating surgeons treated the operative wounds with respect to irrigation and wound dressings.
We demonstrate in a heterogeneous population of patients undergoing bowel surgery that perioperative blood transfusion, a higher intraoperative temperature nadir, and any preoperative infection are associated with an increased risk for SSI. Further investigations are warranted to determine optimal timing of perioperative antibiotic prophylaxis in this patient population.
Correspondence: J. Matthias Walz, MD, University of Massachusetts Memorial Medical Center, 55 Lake Ave N, Worcester, MA 01536 (firstname.lastname@example.org).
Accepted for Publication: August 15, 2005.
Author Contributions:Study concept and design: Walz, Paterson, and Heard. Acquisition of data: Paterson, Seligowski, and Heard. Analysis and interpretation of data: Walz, Paterson, and Heard. Drafting of the manuscript: Walz, Paterson, Seligowski, and Heard. Critical revision of the manuscript for important intellectual content: Walz, Paterson, and Heard. Statistical analysis: Walz, Paterson, and Heard. Administrative, technical, and material support: Walz, Seligowski, and Heard. Study supervision: Paterson and Heard.
Previous Presentation: This study was presented in part at the CHEST World Congress of the American College of Chest Physicians; October 27, 2004; Seattle, Wash.