SCIP indicates Surgical Care Improvement Project; VASQIP, Veteran Affairs Surgical Quality Improvement Program.
Unadjusted (A) and adjusted (B) odds ratios (ORs) for the association between timing of prophylactic antibiotics and surgical site infection, with P values representing the significance of the association.Solid line indicates the OR estimate for surgical site infection; shaded area, 95% CI; dashed vertical line, incision time; and dashed horizontal line, an OR estimate of 1.0. Because spline fits for timing in the adjusted overall and vascular models were nearly linear (df < 1.1), final models considered timing as a linear effect.
eTable. Odds Ratio and Significance of Generalized Additive Models for Assessing Surgical Site Infection.
eFigure. For the Whole Cohort, Unadjusted Odds Ratios (A) and Adjusted Odds Ratios (B) for the Association between Timing of Prophylactic Antibiotics and Surgical Site Infection (SSI)
Hawn MT, Richman JS, Vick CC, Deierhoi RJ, Graham LA, Henderson WG, Itani KMF. Timing of Surgical Antibiotic Prophylaxis and the Risk of Surgical Site Infection. JAMA Surg. 2013;148(7):649-657. doi:10.1001/jamasurg.2013.134
Copyright 2013 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Timing of prophylactic antibiotic administration for surgical procedures is a nationally mandated and publicly reported quality metric sponsored by the Centers for Medicare and Medicaid Services Surgical Care Improvement Project. Numerous studies have failed to demonstrate that adherence to the Surgical Care Improvement Project prophylactic antibiotic timely administration measure is associated with decreased surgical site infection (SSI).
To determine whether prophylactic antibiotic timing is associated with SSI occurrence.
Retrospective cohort study using national Veterans Affairs patient-level data on prophylactic antibiotic timing for orthopedic, colorectal, vascular, and gynecologic procedures from 2005 through 2009.
National Veterans Affairs Surgical Care Improvement Project data from 112 Veterans Affairs hospitals and matched Veterans Affairs Surgical Quality Improvement Program data.
Patients undergoing hip or knee arthroplasty, colorectal surgical procedures, arterial vascular surgical procedures, and hysterectomy.
Timing of prophylactic antibiotic administration with respect to surgical incision time.
Main Outcomes and Measures
Data for prophylactic antibiotic agent, prophylactic antibiotic timing with respect to surgical incision, and patient and procedure risk variables were assessed for their relationship with the occurrence of a composite superficial or deep incisional SSI within 30 days after the procedure. Nonlinear generalized additive models were used to examine the association between antibiotic timing and SSI.
Of the 32 459 operations, prophylactic antibiotics were administered at a median of 28 minutes (interquartile range, 17-39 minutes) prior to surgical incision, and 1497 cases (4.6%) developed an SSI. Compared with procedures with antibiotic administration within 60 minutes prior to incision, higher SSI rates were observed for timing more than 60 minutes prior to incision (unadjusted odds ratio [OR] = 1.34; 95% CI, 1.08-1.66) but not after incision (unadjusted OR = 1.26; 95% CI, 0.92-1.72). In unadjusted generalized additive models, we observed a significant nonlinear relationship between prophylactic antibiotic timing and SSI when considering timing as a continuous variable (P = .01). In generalized additive models adjusted for patient, procedure, and antibiotic variables, no significant association between prophylactic antibiotic timing and SSI was observed. Vancomycin hydrochloride was associated with higher SSI occurrence for orthopedic procedures (adjusted OR = 1.75; 95% CI, 1.16-2.65). Cefazolin sodium and quinolone in combination with an anaerobic agent were associated with fewer SSI events (cefazolin: adjusted OR = 0.49; 95% CI, 0.34-0.71; quinolone: adjusted OR = 0.55; 95% CI, 0.35-0.87) for colorectal procedures.
Conclusions and Relevance
The SSI risk varies by patient and procedure factors as well as antibiotic properties but is not significantly associated with prophylactic antibiotic timing. While adherence to the timely prophylactic antibiotic measure is not bad care, there is little evidence to suggest that it is better care.
Preventing surgical site infection (SSI) is a national priority and a major focus of the Centers for Medicare and Medicaid Services Surgical Care Improvement Project (SCIP).1 The efficacy of prophylactic antibiotics in reducing SSI for major surgical procedures has been proven in clinical trials. However, the SCIP implemented timing standards for appropriate antibiotic prophylaxis in an effort to reduce SSI. The SCIP measure assesses compliance for prophylactic antibiotic administration within 60 minutes prior to surgical incision, not whether an antibiotic is administered.
The selection of the 60-minute window for prophylactic antibiotic administration was primarily based on 2 types of evidence: pharmacokinetics of the antibiotic agents and 1 large single-institution cohort study analyzing the association between timing of antibiotic administration and subsequent SSI.1,2 However, in the cohort study of 2847 patients, only 44 (1.5%) developed an SSI. Furthermore, the only significant associations between SSI and prophylactic antibiotic timing were for the categories of antibiotic administration more than 120 minutes prior to incision or more than 180 minutes after incision. Subsequent studies examining the relationship between timing of prophylactic antibiotic administration and SSI have failed to demonstrate the superiority of the 60-minute preincision window.3- 9 Moreover, all studies to date have assessed prophylactic antibiotic administration in arbitrary timing categories rather than as a continuous variable.
Quiz Ref IDProphylactic antibiotic timing is a nationally mandated quality metric used for public reporting and performance pay initiatives based on limited evidence to support the effectiveness in reducing SSI. We undertook this study to assess the relationship of timing as a continuous variable and postoperative SSI in a large, national, contemporary surgical cohort to determine whether timing is significantly associated with SSI occurrence.
We conducted a retrospective cohort study of surgical procedures meeting criteria for the SCIP measure from 2005 through 2009 to examine the relationship between prophylactic antibiotic timing and SSI in the Veterans Affairs (VA) system. National VA SCIP data from 112 VA hospitals with information on antibiotic name and time of administration were matched to VA Surgical Quality Improvement Program (VASQIP) data containing information on time of surgical incision and the occurrence of SSI. The association of the time in minutes between prophylactic antibiotic administration and surgical incision with subsequent SSI was assessed. The study protocol was reviewed and approved by the local VA institutional review board with waiver of informed consent.
Information on antibiotic agent and timing was obtained from the VA Office of Information and Analytics External Peer Review Program. The VA contracts with the West Virginia Medical Institute to collect VA hospital SCIP measures according to guidelines set forth by The Joint Commission and the Centers for Medicare and Medicaid Services.10
Patient, procedure, and outcome variables were obtained from matched records in the VASQIP data. The VASQIP assesses risk-adjusted 30-day postoperative morbidity and mortality data within the VA health care system.3,4 The methods of the noncardiac VASQIP have been previously published.11 The VASQIP collects demographic characteristics, preoperative risk data, operative data, and 30-day postoperative mortality and morbidity outcomes on a sample of patients undergoing major surgery in the VA health care system. Clinical nurse reviewers trained in clinical medicine and quality assurance complete in-depth training on the data collection procedures and detailed definition of each variable.12
Quiz Ref IDThe SCIP population for the prophylactic antibiotic measures includes patients undergoing 5 types of major surgical procedures: (1) cardiac surgery, (2) hip or knee arthroplasty, (3) colorectal surgery, (4) arterial vascular surgery, and (5) hysterectomy.10 Cardiac cases were excluded because the VASQIP only assesses mediastinitis as an SSI outcome for cardiac procedures.
The independent variable of interest was timing of antibiotic administration in minutes relative to surgical incision time. If more than 1 antibiotic was administered, the prophylactic antibiotic administered closest to but before incision was used. The dependent variable of interest was the occurrence of a superficial or deep incisional SSI within 30 days postoperatively as reported by the VASQIP, following the Centers for Disease Control and Prevention definition.13 Superficial and deep incisional SSIs were combined to create a composite SSI outcome variable, and sensitivity analyses were performed with organ-space SSI included in the outcome.
Cases with a single prophylactic antibiotic were classified by name. Cases with a combination of 2 antibiotics were classified as follows: (1) cefazolin sodium plus metronidazole; (2) clindamycin plus gram-negative coverage (aminoglycoside [n = 67] or aztreonam [n = 3]); (3) vancomycin hydrochloride plus other SCIP-approved antibiotic (cefazolin [n = 1227], cefoxitin sodium [n = 59], quinolone [n = 36], or ampicillin sodium/sulbactam sodium [n = 4]); and (4) quinolone plus anaerobic coverage (metronidazole [n = 284] or clindamycin [n = 50]). Cases receiving any non–SCIP-approved antibiotics or combinations were excluded from primary analyses; however, sensitivity analyses were performed including these cases.
Patient-level covariates known to predict the occurrence of SSI, including demographic characteristics, lifestyle variables (eg, tobacco and alcohol use), and cardiovascular, pulmonary, renal, hepatobiliary, nutritional, and immune comorbidities were obtained from the VASQIP.14 Surgery characteristics considered in our analyses included surgical specialty (orthopedic, colorectal, vascular, and gynecologic), emergent operation, American Society of Anesthesiologists status, wound classification (clean, clean/contaminated, contaminated, or infected), and duration of the operation (incision to surgery end time). To account for the complexity of the operation, the resource-based relative value units for the Current Procedural Terminology code of the performed operation were used.
Unadjusted testing for association between patient characteristics, prophylactic antibiotic timing intervals, and SSI was done using χ2 and Wilcoxon rank sum tests. The relationship between timing and SSI rates was modeled considering time as a nonlinear effect using generalized additive models (GAMs). Age, operative duration, and relative value units were also included as continuous nonlinear terms in GAMs. Separate GAM fits modeled the effect of timing and continuous covariates using both penalized thin plate regression splines and locally weighted scatterplot smoothing; results were similar and the spline models are presented. When the estimated df of the adaptive spline was less than 1.1 during model building, the final model considered the effect to be linear. The estimated df for timing modeled as a regression spline ranged from less than 1.1 to 7.5. In overall models, timing was ultimately treated as linear in the adjusted model (spline df < 1.1), while the df for the overall unadjusted model was 6.4. Approximate P values for spline terms are derived from analysis of variance χ2 statistics using the estimated df. To minimize the influence of outliers on the partial effect plots, the 51 observations (0.2%) with antibiotic administration more than 180 minutes prior to incision time were excluded from final modeling. All models were fit with the entire cohort adjusting for surgical specialty and repeated stratified by specialty. Full models included adjustment for patient and operative characteristics known to be predictive of SSI.14 Congestive heart failure and renal insufficiency variables were excluded from the models as there were too few outcomes to provide reliable estimates. To examine the relative contribution of each variable to the adjusted models, we calculated the analysis of variance χ2 statistic for each variable minus its df (estimated for spline terms) following the method recommended by Harrell15 so that larger values indicated greater contribution to the model. The statistical threshold for significance was set at P = .05 for a 2-tailed test.
We performed sensitivity analyses to evaluate whether results were changed by including the following: (1) cases with prophylactic antibiotic deemed inappropriate by SCIP guidelines; (2) VA facilities as random effects; and (3) inclusion of the timely discontinuation of prophylactic antibiotic within 24 hours of surgery end time (SCIP-INF-3 measure). Of the 32 459 surgical procedures, 26 327 (81.1%) had information on the SCIP-INF-3 measure, of which 22 753 (86.4%) were adherent. All analyses and figure preparations were completed using SAS/STAT statistical software version 9.2 (SAS Institute, Inc) and R statistical software version 2.15.1 (R Foundation).16 The GAMs used the R packages GAM and MGCV for locally weighted scatterplot smoothing and spline models, respectively (R software version 1.06.2; R Foundation).17
A total of 36 947 noncardiac SCIP cases with information on antibiotic name, infusion route, time of administration, and matched VASQIP records were identified (Figure 1). The following cases were excluded: (1) prophylactic antibiotic administered more than 8 hours prior to incision (n = 20); (2) prophylactic antibiotic administered after surgery end time (n = 689); (3) non–SCIP-approved prophylactic antibiotic (n = 2398); and (4) SCIP-approved prophylactic antibiotic inappropriate for procedure (n = 1381). The final 2 exclusions were applied to limit the analyses to antibiotics with appropriate coverage for the surgical procedure.
Quiz Ref IDOf the 32 459 cases performed at 112 VA hospitals included in the study, 1497 (4.6%) developed an SSI within 30 days of surgery. The demographic characteristics, comorbidities, and procedure characteristics of the study population stratified by prophylactic antibiotic timing intervals are shown in Table 1. Patients with comorbid conditions (P < .05) and those with an emergent, colorectal, or vascular procedure, a contaminated/dirty wound classification, and longer operations were more likely to have antibiotic administered more than 60 minutes prior to or after incision (P < .001). Higher SSI rates were observed with administration more than 60 minutes prior to incision (odds ratio [OR] = 1.34; 95% CI, 1.08-1.66) but not after incision (OR = 1.26; 95% CI, 0.92-1.72) compared with administration within 60 minutes prior to incision. There was a significant relationship between prophylactic antibiotic timing and SSI in an unadjusted GAM (P = .003) (Figure 2A). Significantly higher unadjusted odds of SSI were observed both within and outside the 60-minute preincision window as represented where the 95% CI does not cross 1.0.
The relationships between antibiotic class, timing, and SSI stratified by surgical procedure type are shown in Table 2. The median time between prophylactic antibiotic administration and surgery was 28 minutes (interquartile range, 17-39 minutes). In unadjusted analysis of antibiotic choice and SSI risk, several SCIP-approved antibiotics had observed differences in SSI. For orthopedic procedures, vancomycin alone was associated with a higher SSI occurrence (OR = 1.75; 95% CI, 1.24-2.46; cefazolin as reference group). For colorectal procedures, compared with cefoxitin, cefazolin plus metronidazole (OR = 0.48; 95% CI, 0.33-0.70) and quinolone plus anaerobic agent (OR = 0.57; 95% CI 0.38-0.85) were associated with lower SSI occurrence.
The relative contributions of GAM covariates to SSI risk as calculated by subtracting the df from the χ2 value are shown in Table 3. In the overall model adjusted for surgical specialty, prophylactic antibiotic timing ranked 15th of the 16 variables. Among the 3 specialty-stratified models, antibiotic timing ranked 5th (orthopedic), 14th (vascular), and 11th (colorectal) and was not statistically significant in any model. The most informative variables by model type were surgical specialty (overall model), current smoker (orthopedic model), diabetes (vascular model), and operative duration (colorectal model).
The choice of prophylactic antibiotic for orthopedic and colorectal procedures was associated with SSI. Vancomycin was associated with higher SSI occurrence for orthopedic procedures (adjusted OR = 1.75; 95% CI, 1.16-2.65), whereas cefazolin or quinolone in combination with an anaerobic agent were associated with fewer SSI events (cefazolin: adjusted OR = 0.49; 95% CI, 0.34-0.71; quinolone: adjusted OR = 0.55; 95% CI, 0.35-0.87) for colorectal procedures. We substituted half-life in minutes for antibiotic agent in the model, and this did not explain the differences in observed effectiveness of the agents. The ORs, 95% CIs, and P values for all model terms are included in the eTable in Supplement.
The partial effect plots of the association between the antibiotic timing and SSI from the adjusted GAM are shown in Figure 2B. Quiz Ref IDThere was no significant association between timing and SSI for the overall or specialty-stratified models for orthopedic, colorectal, and vascular procedures. In adjusted overall and vascular GAMs, the estimated df for the splined timing were both less than 1.1, so their corresponding final models considered timing as a linear effect.
Sensitivity analyses including the 3779 cases with either a non–SCIP-approved antibiotic or a SCIP-approved antibiotic that was inappropriate for the procedure (Figure 1) did not demonstrate an association between prophylactic antibiotic timing and SSI (partial effect plots are shown in the eFigure in Supplement). The inclusion of VA facilities as random effects, organ-space infection in the composite SSI outcome, and the SCIP-INF-3 measure of timely discontinuation of antibiotics in the models did not significantly alter the association of antibiotic timing and SSI.
To our knowledge, this is the largest study to date to evaluate the relationship between prophylactic antibiotic timing and subsequent SSI for major surgical procedures and the only study to assess timing as a continuous variable. Quiz Ref IDWe did not observe a significant relationship between timing and SSI in this contemporary national VA SCIP cohort, but we did observe a significant association between choice of antibiotic and SSI for orthopedic and colorectal procedures.
The empirical superiority of the 60-minute timing metric has yet to be substantiated. The evidence for this metric was based on antibiotic pharmacokinetics, including studies of antibiotic tissue concentration during surgery and a large cohort study.1,2,18,19 The study by Classen et al2 assessed the relationship between prophylactic antibiotic timing and SSI in a broad group of patients undergoing elective surgery and has several strengths, including prospective in-hospital assessment of SSI occurrence. However, of the 2847 patients in their cohort, only 44 (1.5%) developed an SSI. While they observed the lowest SSI rate in the group who received prophylactic antibiotics within 2 hours prior to incision, that rate was not significantly different from the group who received antibiotics in the perioperative period (≥3 hours after incision). The only significant differences were for the category of early, defined by antibiotics given 2 to 24 hours prior to surgical incision, and postoperative, defined by antibiotics given 3 to 24 hours after incision.
Other studies have shown that the 60-minute window is not exclusively the most effective timing window and that infection rates among intervals within the 60-minute window are not equivalent.5- 9,20,21 Some studies suggest that antibiotics administered 0 to 30 minutes prior to incision are associated with higher SSI rates compared with antibiotics given 30 to 60 minutes before incision.6,20 Conversely, another multicenter study of 4472 patients undergoing cardiac surgery, hysterectomy, or hip or knee arthroplasty reported an infection rate of 1.6% when the antibiotic was administered within 30 minutes prior to incision compared with 2.4% when administered between 31 and 60 minutes prior to incision (not statistically significant). Compared with the 60-minute preincision window, they observed higher SSI rates when antibiotics were administered after incision (OR = 2.20; 95% CI, 1.03-4.66) but not prior to 60 minutes.8 Our partial effect plots confirm that giving antibiotics closer to the time of incision appears to be an equally effective practice. The variability in antibiotic timing importance among these reports and our study confirms that the relationship between timing and SSI is highly dependent on the selection of timing intervals, the antibiotic agent, and the surgical population studied.
We observed significant differences in risk-adjusted SSI rates for several SCIP-approved antibiotics. Vancomycin as a sole agent was associated with higher SSI rates. It is unclear whether the selection of vancomycin is an indicator of patients at higher risk for SSI; however, the higher rate was observed when vancomycin was used alone but not in combination with another SCIP-approved antibiotic. This is in contrast to a prior study that found higher rates of SSI when vancomycin was used in combination rather than as a single agent.8 We also observed a protective effect with cefazolin and quinolone in combination with anaerobic agents for colorectal surgery. Similar findings of decreased infections with these combinations as well as ertapenem have been reported for a large multicenter study of colorectal surgery.22 Furthermore, a randomized controlled trial comparing ertapenem and cefotetan reported lower rates of SSI for ertapenem compared with cefotetan (adjusted OR = 0.41; 95% CI, 0.28-0.61).23 Studies are needed to determine whether the lack of antibiotic redosing or bacterial resistance patterns led to differences in antibiotic effectiveness.
Our study has several strengths, including assessing timing as a continuous variable, detailed patient- and procedure-level data, and assessment of SSI out to 30 days after the surgical procedure.24 Furthermore, the antibiotic timing and SSI outcome data were assessed by 2 independent programs, limiting the chance for bias.
The study sample primarily comprises older male patients, limiting the generalizability to women, and we were not able to assess cardiac procedures. Intraoperative redosing of antibiotics has been shown to be important for longer procedures,5 and we were not able to assess whether redosing occurred. Our cohort is large in absolute terms, with the highest SSI rates observed for antibiotics administered more than 120 minutes prior to incision; however, this group was a small proportion of the cohort. Given current high adherence to the timely antibiotic measure, it is likely that all observational studies will share this limitation.8
Timing of prophylactic antibiotic administration is not significantly associated with SSI occurrence. While adherence to the timely prophylactic antibiotic measure is not bad care, there is little evidence to suggest that it is better care. Future endeavors for reducing SSI should robustly correlate with improved outcomes and include studies to refine recommended antibiotic choice and redosing.
Corresponding Author: Mary T. Hawn, MD, MPH, Section of Gastrointestinal Surgery, Department of Surgery, University of Alabama at Birmingham, 1922 Seventh Ave S, KB 428, Birmingham, AL 35294 (email@example.com).
Accepted for Publication: December 15, 2012.
Published Online: March 20, 2013. doi:10.1001/jamasurg.2013.134.
Author Contributions: Dr Richman and Ms Vick had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hawn, Vick, Henderson, and Itani.
Acquisition of data: Vick, Deierhoi, Graham, and Henderson.
Analysis and interpretation of data: All authors.
Drafting of the manuscript: Hawn, Richman, Vick, and Deierhoi.
Critical revision of the manuscript for important intellectual content: Hawn, Richman, Vick, Graham, Henderson, and Itani.
Statistical analysis: Richman, Vick, Deierhoi, Graham, and Henderson.
Obtained funding: Hawn and Vick.
Administrative, technical, and material support: Vick, Deierhoi, Graham, and Itani.
Conflict of Interest Disclosures: Drs Hawn, Richman, and Itani and Mss Vick and Deierhoi are VA employees. Dr Itani has received research support from Merck, Cubist, and Trius Therapeutics and has been a consultant to Forest Pharmaceuticals.
Funding/Support: This study is funded by grant PPO 10-296 from the VA Health Services Research and Development Service. Dr Richman is supported by a Career Development Award from the VA Health Services Research and Development Service.
Role of the Sponsors: The VA Office of Research and Development had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation or review of the manuscript. The VA National Surgery Office and Office of Information and Analytics approved the manuscript for adherence to the data use agreements.
Disclaimer: The opinions expressed are those of the authors and not necessarily those of the Department of Veterans Affairs or the US government.
Additional Contributions: The VA Surgical Quality Data Use Group acted as scientific advisors and provided critical review of data use and analysis presented in this article.