Customize your JAMA Network experience by selecting one or more topics from the list below.
Elliott IA, Chan C, Russell TA, et al. Distinction of Risk Factors for Superficial vs Organ-Space Surgical Site Infections After Pancreatic Surgery. JAMA Surg. 2017;152(11):1023–1029. doi:10.1001/jamasurg.2017.2155
Can risk factors for surgical site infections after pancreatic surgery be modified?
In a retrospective analysis of the prospectively maintained National Surgical Quality Improvement Program–Hepatopancreaticobiliary Collaborative database, risk factors for superficial and organ-space surgical site infections after pancreatic surgery were found to differ. Preoperative biliary stenting was an independent risk factor for superficial infections, whereas soft gland texture was an independent risk factor for organ-space infections, which were also found to heavily overlap with pancreatic fistulae.
Reporting of surgical site infections after pancreatic surgery should distinguish superficial vs organ-space infections, which will likely require different preventative strategies.
Surgical site infection (SSI) rates are increasingly used as a quality metric. However, risk factors for SSI in pancreatic surgery remain undefined.
To stratify superficial and organ-space SSIs after pancreatectomy and investigate their modifiable risk factors.
Design, Setting, and Participants
This retrospective analysis included 201 patients undergoing pancreatic surgery at a university-based tertiary referral center from July 1, 2013, through June 30, 2015, and 10 371 patients from National Surgical Quality Improvement Program–Hepatopancreaticobiliary (NSQIP-HPB) Collaborative sites from January 1, 2014, through December 31, 2015.
Main Outcomes and Measures
Superficial, deep-incisional, and organ-space SSIs, as defined by NSQIP.
Among the 201 patients treated at the single center (108 men [53.7%] and 93 women [46.3%]; median age, 48.6 years [IQR, 41.4-57.3 years]), 58 had any SSI (28.9%); 28 (13.9%), superficial SSI; 8 (4%), deep-incisional SSI; and 24 (11.9%), organ-space SSI. Independent risk factors for superficial SSI were preoperative biliary stenting (odds ratio [OR], 4.81; 95% CI, 1.25-18.56; P = .02) and use of immunosuppressive corticosteroids (OR, 13.42; 95% CI, 1.64-109.72; P = .02), whereas soft gland texture was the only risk factor for organ-space SSI (OR, 4.45; 95% CI, 1.35-14.66; P = .01). Most patients with organ-space infections also had grades B/C fistulae (15 of 24 [62.5%] vs 4 of 143 [2.8%] in patients with no SSI; P < .001). Organ/space but not superficial SSI was associated with an increased rate of sepsis (7 of 24 [29.2%] vs 4 of 143 [2.8%]; P < .001) and prolonged length of hospital stay (12 vs 8 days; P = .04). Among patients in the NSQIP-HPB Collaborative, 2057 (19.8%) had any SSI; 719 (6.9%), superficial SSI; 207 (2%), deep-incisional SSI; and 1287 (12.4%), organ-space SSI. Preoperative biliary stenting was confirmed as an independent risk factor for superficial SSI (OR, 2.07; 95% CI, 1.58-2.71; P < .001). In this larger data set, soft gland texture was an independent risk factor for superficial SSI (OR, 1.45; 95% CI, 1.14-1.85; P = .002) but was more strongly and significantly associated with organ-space SSI (OR, 2.32; 95% CI, 1.88-2.85; P < .001).
Conclusions and Relevance
Preoperative biliary stenting and coriticosteroid use increase superficial SSI, even in patients receiving perioperative piperacillin-tazobactam. Additional measures, including extended broad-spectrum perioperative antibiotic treatment, should be considered in these patients. Organ/space SSIs appear to be related to pancreatic fistulae, which are not modifiable. Reporting these different subtypes as a single, overall rate may be misleading.
Surgical outcomes are increasingly tracked, reported, and used as quality metrics.1,2 Surgical site infections (SSIs) are a prominent example of a complication that has been deemed to be preventable and consequently tied to reimbursement by the Centers for Medicare & Medicaid Services and other health care payors.2,3 Such policies have created increased incentives to decrease SSI rates. To this end, the American Hepato-Pancreato-Biliary Association Clinical Trials Committee has proposed a prospective randomized clinical trial for patients undergoing pancreaticoduodenectomy to investigate combined pipercillin and tazobactam vs cephalosporins for perioperative prophylaxis against infection (Henry A. Pitt, MD; Michael I. D’Angelica, MD; email communication; August 8, 2016).
However, appropriate risk adjustment of patients and identification of viable prevention strategies requires a precise definition of SSI and a comprehensive understanding of risk factors. Previous literature on patients undergoing pancreatic surgery has yielded conflicting rates of and risk factors for SSI likely owing to small sample size and/or inconsistent outcome definitions.4-9 Furthermore, how classic risk factors for postoperative wound infections, such as diabetes, obesity, nutritional status, and operative time, affect patients undergoing pancreatectomy is unknown. These patients also have a higher prevalence of other potentially relevant comorbidities, such as malignant neoplasms, exposure to chemotherapy, and weight loss.
The National Surgical Quality Improvement Program (NSQIP) is a powerful resource that enables comparison of patient outcomes across participating institutions by standardizing definitions of clinical variables and postoperative complications. Since 2014, the NSQIP-Hepatopancreaticobiliary (HPB) Collaborative has been aggregating pancreatectomy-specific data, enabling examination of variables and outcomes that are important for patients undergoing pancreatic surgery, such as preoperative biliary stenting, pathologic characteristics of specimens, postoperative fistulae, and delayed gastric emptying.
The definitions used by NSQIP can differ from those previously used in the literature, thus limiting comparison of NSQIP data with previous studies. For example, NSQIP codes 3 SSI types (superficial, deep-incisional, and organ-space), but reports typically include 1 aggregate SSI rate. We suspect that the NSQIP definitions of superficial and deep-incisional SSIs probably reflect what was previously termed a wound infection in prior studies of patients undergoing pancreatectomy, whereas the classification of organ-space infection may have significant overlap with that of pancreatic fistula. We therefore hypothesized that these SSI subtypes might have distinct risk factors and different clinical outcomes. The aim of the present study was to elucidate these distinct risk factors for the SSI subtypes and examine the association of SSIs with other clinical outcomes to inform potential quality-improvement strategies.
We obtained NSQIP data for 201 patients undergoing pancreaticoduodenectomy (Current Procedural Terminology [CPT] codes 48150, 48152, 48153, or 48154) or distal pancreatectomy (CPT codes 48140, 48145, or 48146) from July 1, 2013, through June 30, 2015, at UCLA (University of California, Los Angeles). These data are collected by American College of Surgeons NSQIP–certified surgical clinical reviewers at each participating hospital. When NSQIP data were incomplete, medical records were reviewed manually. Standard practice at UCLA includes routine administration of piperacillin-tazobactam for perioperative prophylaxis during pancreatic surgery (or tigecycline in cases of penicillin allergy). In addition to NSQIP variables, we collected wound culture results when available. This study was performed in compliance with the Health Insurance Portability and Accountability Act and approved by the institutional review board of UCLA, which waived the need for informed consent for use of deidentified retrospective data.
To compare outcomes at UCLA with those from the larger NSQIP-HPB Collaborative, general NSQIP and NSQIP-HPB public use files were obtained from January 1, 2014, through December 31, 2015. With use of the same CPT codes as indicated in the previous paragraph, a total of 13 773 cases were abstracted from the NSQIP public use files and 11 219 from the NSQIP-HPB public use files. Cases were matched by NSQIP case identification number. A total of 3335 cases did not have a corresponding entry in both public use files and were therefore excluded. Among the remaining 10 438 cases, 67 were removed because basic data were incomplete (length of stay in 13 cases and designation of malignant histologic findings in 54 cases). After NSQIP and NSQIP-HPB data were merged, 10 371 patients remained for analysis.
Preoperative, intraoperative, and postoperative variables were classified according to NSQIP definitions. NSQIP classifies the following 3 types of SSI: superficial-incisional (involves only the skin or subcutaneous tissue of the surgical incision), deep-incisional (involves deep soft tissues; eg, fascial and muscle layers), and organ-space (any part of the anatomy, organs, or spaces other than the incision that was opened or manipulated during an operation). Pancreatic fistulae are defined by NSQIP-HPB as persistent drainage of amylase-rich fluid requiring a drain to be continued longer than 7 days, percutaneous drainage, or reoperation. These were imputed to International Study Group on Pancreatic Fistula classifications of grade A, B, or C fistula, respectively.10 Length of stay was calculated from the date of operation to the date of discharge. Readmission was defined as any readmission within 30 days of surgery.
Statistical analysis was performed comparing patients with no SSI and those with any SSI (superficial, deep-incisional, and/or organ-space) and those with superficial or organ-space SSIs specifically. The overall number of deep-incisional SSIs was too small to analyze for independent risk factors. Patient characteristics and operative and postoperative variables were compared using the χ2 test or Fisher exact test for binary categorical variables and comparison of medians and Mann-Whitney tests for continuous variables after all continuous variables were determined to be nonnormally distributed by Shapiro-Wilk test. To identify independent risk factors for SSI subtypes in our population, we entered clinically and statistically significant variables into a multivariable logistic regression model. Where data were missing in NSQIP-HPB and not amenable to imputation, sensitivity analyses were completed to confirm that inclusion did not significantly alter the model. To address the association of deep-incisional with the other SSI types, sensitivity analyses were performed by combining the outcome of deep-incisional with superficial-incisional and organ-space SSIs and comparing resultant logistic regression models. Hypothesis tests were 2-sided. Analyses were conducted using SPSS statistical software (version 24.0; IBM Corp). Statistical significance was defined as P ≤ .05.
Among the 201 patients who underwent pancreatectomy at UCLA (108 men [53.7%] and 93 women [46.3%]; median age, 48.6 years [IQR, 41.4-57.3 years]), 58 (28.9%) had any SSI; 28 (13.9%), superficial SSI; 8 (4%), deep-incisional SSI; and 24 (11.9%), organ-space SSI. Only 4 patients (2%) had concurrent superficial and organ-space SSIs.
We examined risk factors for SSI overall (superficial, deep-incisional, and/or organ-space SSI) and those associated with superficial or organ-space SSI. Demographic characteristics, comorbidities, and preoperative and intraoperative factors for patients who developed no SSI vs any SSI, superficial SSI, or organ-space SSI are shown in Table 1. In univariable analysis of overall SSIs, preoperative biliary stenting (29 of 58 [50%] vs 43 of 143 [30.1%]; P = .008) and malignant histologic findings (48 of 58 [82.8%] vs 98 of 143 [68.5%]; P = .04) were significant. However, the risk factors specifically associated with superficial or organ-space infections were different; long-term corticosteroid use (4 of 28 [14.3%] vs 4 of 143 [2.8%]; P = .03) and preoperative biliary stenting (17 of 28 [60.7%] vs 43 of 143 [30.1%]; P = .002) were associated with superficial SSIs, whereas only soft gland texture (15 of 19 [78.9%] vs 47 of 100 [47%]; P = .01) was associated with organ-space SSIs.
Multivariable analysis, shown in Table 2, confirmed these distinct risk factors for superficial vs organ-space SSIs; long-term corticosteroid use (odds ratio [OR], 13.42; 95% CI, 1.64-109.72; P = .02) and preoperative biliary stenting (OR, 4.81; 95% CI, 1.25-18.56; P = .02) were independent risk factors for superficial SSI, whereas only soft gland texture (OR, 4.45; 95% CI, 1.35-14.66; P = .01) was an independent risk factor for organ-space SSI. Because data on gland texture were available for only 142 patients, we repeated the multivariable logistic regression with and without gland texture and found no change in which risk factors were significant. Sensitivity analysis by combination of deep-incisional with superficial-incisional and organ-space SSIs did not significantly alter the results of the multivariable logistic regression model.
We then examined the association of SSIs with other postoperative outcomes, shown in Table 3. Overall SSI was associated with an increased rate of sepsis (10 of 58 [17.2%] vs 4 of 143 [2.8%]; P = .001), but this appeared to be driven by an increased rate of sepsis among patients with organ-space SSI (7 of 24 [29.2%] vs 4 of 143 [2.8%]; P < .001). Organ-space SSI was also associated with increased postoperative length of stay (12 vs 8 days; P = .04).
Soft gland texture has been previously established as a risk factor for pancreatic fistula; therefore, we were not surprised to find a high concordance of pancreatic fistula with organ-space infection, as also shown in Table 3.11,12 For example, although superficial SSI was associated with an increased rate of grade B/C fistulae (4 of 28 [14.3%] vs 4 of 143 [2.8%]; P = .03), the concordance with organ-space SSI was stronger, because most of those patients had a grade B/C fistula (15 of 24 [62.5%] vs 4 of 143 [2.8%]; P < .001).
We then performed the same aforementioned analyses on the 2014-2015 NSQIP-HPB collaborative data set. Among 10 371 patients undergoing pancreatic resections, 2057 (19.8%) had any SSI; 719 (6.9%), superficial-incisional SSI; 207 (2%), deep-incisional SSI; and 1287 (12.4%), organ-space SSI. Similar to the results from the UCLA data, 136 patients (1.3%) had concurrent superficial and organ-space SSIs.
Univariable analysis of demographic characteristics, comorbidities, and preoperative and intraoperative variables in this multi-institutional data set with substantially greater power revealed additional risk factors for various types of SSI, as shown in Table 4. In multivariable logistic regression analysis, shown in Table 5, independent risk factors for SSI overall were body mass index (calculated as weight in kilograms divided by height in meters squared) of greater than 30 (OR, 1.42; 95% CI, 1.19-1.69; P < .001), preoperative biliary stenting (OR, 1.40; 95% CI, 1.17-1.67; P < .001), prolonged operative time (OR, 1.35; 95% CI, 1.13-1.60; P = .001), Whipple operation (OR, 1.95; 95% CI, 1.53-2.47; P < .001), and soft gland texture (OR, 1.88; 95% CI, 1.59-2.21; P < .001). When superficial and organ-space SSIs were modeled separately, some distinct risk factors were again identified. Of note, preoperative biliary stenting was an independent risk factor for only superficial SSI (OR, 2.07; 95% CI, 1.58-2.17; P < .001). Although soft gland texture was an independent risk factor for superficial SSI (OR, 1.45; 95% CI, 1.14-1.85; P = .002), it was more significantly associated with organ-space SSI (OR, 2.32; 95% CI, 1.88-2.85; P < .001). Sensitivity analysis by combination of deep-incisional with superficial-incisional and organ-space SSIs did not significantly alter the results of the multivariable logistic regression model.
We performed an analysis of risk factors for SSI in patients undergoing pancreatic surgery at our institution and then validated these risk factors using the large, multi-institutional NSQIP-HPB Collaborative data set. Our data suggest that reporting a single, overall SSI rate is misleading, because this rate appears to consist of the following 2 distinct clinical entities: superficial SSIs, which correspond to what is commonly thought of as a wound infection, and organ-space SSIs, which correspond to pancreatic leaks. Because these entities have distinct risk factors and distinct clinical consequences, we believe that separate reporting of them and design of targeted preventative strategies for each are important.
Deep-incisional SSIs, as defined by NSQIP, are rare after pancreatic surgery. We believe that these likely represent a more severe wound infection than superficial-incisional SSIs, and these could be reported together. In fact, when we performed sensitivity analysis by grouping deep-incisional with superficial or with organ-space SSIs, we found no significant change in the results of the models likely owing to the low incidence of deep-incisional SSIs.
In examining superficial-incisional SSIs, we identified preoperative biliary stenting as an independent risk factor at our institution and in the NSQIP-HPB Collaborative. This finding is consistent with previous studies that have shown that patients with biliary stents are at higher risk for postoperative infectious complications.5,13 In light of these data, routine preoperative biliary stenting is not performed and is reserved for patients whose serum bilirubin levels are markedly elevated and those who are symptomatic or have evidence of cholestatic liver dysfunction. Although further minimizing the use of stents in patients with clinically significant obstruction may not be possible, broadening standard antibiotic prophylaxis from third-generation cephalosporins to pipercillin-tazobactam has been suggested in this subgroup, and a randomized clinical trial to investigate this change has been proposed. However, at our institution, we have been routinely using pipercillin-tazobactam since 2008 for perioperative prophylaxis; thus, this modification is unlikely to reduce the influence of biliary stents as an independent risk factor for superficial SSIs.6 Instead, we favor extended broad-spectrum prophylaxis regimens for these high-risk patients, as has been suggested by others in the literature.7,14 A trial comparing cessation of broad-spectrum antibiotic therapy (eg, pipercillin-tazobactam) within 24 hours of surgery and continuation for a period of 5 days might be more useful. We had intended to perform an in-depth analysis of the pathogens causing superficial SSIs at our institution and their antibiogram; however, culture data were available for only 7 of 28 patients. Most of these findings (5 of 7) were enteric bacteria, supporting the need for coverage of gram-negative bacteria, but we hesitate to overgeneralize from this small number of cases. We observed a higher rate of overall SSI at our institution than in the NSQIP-HPB Collaborative, driven mostly by a higher rate of superficial SSI. However, we did not find any clear explanation for this rate in our analysis, because the rate of biliary stenting in both cohorts was comparable (72 of 201 [35.8%] vs 3441 of 10 371 [33.1%]).
The designation of organ-space SSI after pancreatic surgery overlaps with and in most cases is attributable to concurrent pancreatic fistulae. This finding may not be attributable to fistulae causing organ-space infections but instead to most clinically significant fistulae meeting the definition for organ-space SSI and essentially representing the same clinical entity. A considerable amount of literature has explored interventions to decrease the rate of pancreatic fistula after pancreatectomy, and to date, no consensus exists on validated, evidenced-based measures.12,15 Thus, our data suggest that the rate of organ-space SSIs, as defined by NSQIP, is not modifiable.
NSQIP is a unique resource for evaluating perioperative risk factors and outcomes. By aggregating data from a number of institutions, it enables analysis of more specific subsets of patients and outcomes with a higher statistical power than could be achieved at a single institution and without the well-documented pitfalls of using large administrative billing data sets.16 However, previous studies in pancreatic surgery highlighted the limitations of the general NSQIP data in capturing relevant procedure-specific variables, which led to the creation of the NSQIP-HPB Collaborative.17 As the scope of NSQIP data evolves with procedure-specific modules, the utility of this resource will continue to expand, allowing reevaluation of risk factors and outcomes previously studied in smaller series.
As with any large database, NSQIP is limited by the accuracy and completeness of the data provided by individual institutions. For example, in the present study, data on body mass index, albumin level, preoperative biliary stenting, and gland texture were unavailable for 33%, 10%, 7%, and 39%, respectively, of the patients included, and these variables were not amenable to imputation. Because of the large number of cases available, analysis after exclusion of missing data was still high-powered, but this approach assumes that these data are missing at random, which is a potential source of bias. However, the NSQIP-HPB Collaborative was more sensitive for identifying certain risk factors, as shown by the statistical significance reached by prolonged operative time for all types of SSI that was not detected in our institutional data. We believe that simultaneous analysis of our institutional data, in which missing variables could be completed by medical record review, alongside analysis of the NSQIP-HPB Collaborative overcomes some of the limitations of either approach alone.
Preoperative biliary stenting increases the rate of superficial SSI, even in patients receiving piperacillin-tazobactam. Additional measures, including extended perioperative antibiotic therapy, should be considered in these patients. Organ-space SSIs appear to be related to pancreatic fistulae, which are not modifiable. Reporting the different subtypes of SSIs as a single, overall rate may be misleading.
Corresponding Author: Timothy R. Donahue, MD, Department of Surgery, David Geffen School of Medicine, UCLA, 10833 Le Conte Ave, Room 72-215, Center for Health Sciences, PO Box 956904, Los Angeles, CA 90095 (firstname.lastname@example.org).
Accepted for Publication: April 9, 2017.
Published Online: July 12, 2017. doi:10.1001/jamasurg.2017.2155
Author Contributions: Drs Elliott and Donahue 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: Elliott, Chan, Russell, Hines, Reber, Donahue.
Acquisition, analysis, or interpretation of data: Elliott, Chan, Russell, Dann, Williams, Damato, Chung, Girgis, Reber, Donahue.
Drafting of the manuscript: Elliott, Chan, Reber, Donahue.
Critical revision of the manuscript for important intellectual content: Elliott, Russell, Dann, Williams, Damato, Chung, Girgis, Hines, Reber, Donahue.
Statistical analysis: Elliott, Chan, Russell, Williams.
Obtained funding: Donahue.
Administrative, technical, or material support: Elliott, Russell, Dann, Williams, Damato, Chung, Girgis, Donahue.
Study supervision: Girgis, Hines, Reber, Donahue.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grant T32 NIH-NIDDK 41301 from the National Institutes of Health (Drs Elliott and Dann) and by the UCLA H. H. Lee Research Program (Dr Elliott).
Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The hospitals participating in the American College of Surgeons National Surgical Quality Improvement Program are the source of the data used herein; the data have not been verified, and the hospitals are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors.
Meeting Presentation: This study was presented at the 88th Annual Meeting of the Pacific Coast Surgical Association; February 20, 2017; Indian Wells, California.