Risk Factors for Post-Operative Sepsis and Septic Shock in Patients Undergoing Emergency Surgery

Patients and Methods

Before the initiation of this project, approval was obtained from the University of California, Irvine (UCI) Institutional Review Board to analyze the data collected in the UCI Trauma Database.

This study utilized The American College of Surgeons' National Surgical Quality Improvement Program (ACS-NSQIP) dataset between 2012–2015. This dataset contains prospectively gathered clinical data, including pre-operative, intra-operative, and 30-day post-operative data for patients undergoing surgical procedures in both in-patient and out-patient settings. The database excludes patients under 18 years of age, as well as transplant and trauma patients. The study included all emergency abdominal operations, including those involving the peritoneum and omentum, biliary tract, liver, pancreas, colon and rectum, esophagus, liver, small intestine, and stomach. Specific operations were identified by use of the Current Procedural Terminology (CPT) codes. Please see Appendix A for a full list of CPT codes. We included all cases designated by the NSQIP as emergency, which were identified using the “EMERGNCY” variable in the NSQIP PUF file. The NSQIP reports an operation as an “emergency case” if it is designated as such by the surgeon or anesthesiologist, referring to a case “performed within a short interval of time between patient diagnosis or the onset of related pre-operative symptomatology,” implying that the patient's well-being and outcome might be threatened by unnecessary delay and the patient's status could deteriorate unpredictably or rapidly.

The 2012–2015 NSQIP participant user file was queried for patient demographics, selected pre-operative characteristics, common pre-operative co-morbidities, and elective versus emergency cases. All elective cases were excluded. Patients designated as having S/SS were identified. Those having severe sepsis and septic shock are grouped together by NSQIP into the “septic shock” category, whereas patients having sepsis are designated in the “sepsis” category. The detailed NSQIP definitions of sepsis and septic shock are presented in Table 1. The primary endpoints were the incidence rate and risk factors predictive of sepsis or septic shock.

Two separate groups were generated from patients who met the inclusion criteria: Those without a septic complication and those with either S or SS. A descriptive model was then constructed modeling the probability of developing S/SS in the entire study population based on key patient and procedure characteristics. A multiple logistic regression was performed, including the relevant co-variables. The descriptive associations with individual characteristics were quantified in terms of adjusted odds ratios (ORs). These were reported along with 95% confidence intervals (Cis) and the corresponding p values. P < 0.05 was considered statistically significant. Associations were based on logistic regression for post-operative complications. Robust standard errors were used to guard against model misspecification. P values were not adjusted for multiple comparisons. When performing comparisons based on procedure type, procedures on the colon and rectum were taken to be the reference category, as these operations were the most numerous. The overall descriptive power of the model was quantified by the C-statistic. Data management was completed using SAS (SAS Institute, Cary, NC). Statistical analysis was performed using R Statistical software [15].

Results

The 2012–2015 NSQIP dataset contains information on 122,281 ES patients who met the inclusion criteria. Of these, 2,399 (2%) developed post-operative S/SS (60% S; 40% SS). Table 2 provides the descriptive statistics for the entire study population. Sixty-seven percent of the patients in the S/SS group were white, and 54% were male. Compared with patients without septic complications, patients developing S/SS were significantly older (S/SS 64.2 years; no sepsis 48 years; p < 0.05), more were partially or totally dependent (S/SS 10.4%; no sepsis 2.1%; p < 0.05), and they had a higher mean pre-operative serum creatinine concentration (S/SS 1.3 mg/dL; sepsis 0.9 mg/dL; p < 0.05). More patients in the S/SS group received a pre-operative blood transfusion (S/SS 8.8%; no sepsis 1.7%; p < 0.05). Patients in the S/SS group had more co-morbid conditions and had a longer operative time (S/SS 121.9 minutes; no sepsis 68.5 minutes; p < 0.05). The most common category of procedure in patients in the S/SS group (49.8%) was the colon and rectum. The overall 30-day mortality rate among ES patients was 1.4%, whereas the rate in the S/SS group was 18.3%, and the rate among patients with no septic complications was 1% (OR 22.0; CI 19.5–25.0; p < 0.0001).

The C-statistic for the model was 0.84 (95% CI 0.83–0.84), indicating a relatively good fit. Risk factors for S or SS were American Society of Anesthesiologists Physical Status (ASA PS) class 2 or higher (OR 2.57; 95% CI 2.19–3.02; p < 0.0001), totally dependent (OR 2.00; 95% CI 1.38–2.83; p = 0.00021) or partially dependent (OR 1.62; 95% CI 1.35–2.00; p < 0.0001) functional status, and male gender (OR 1.31; 95% CI 1.18–1.45; p < 0.0001). Patients with a history of chronic obstructive pulmonary disease, metastatic cancer, steroid use, serum creatinine >1.2 mg/dL, and albumin <3 g/dL had a higher risk of developing S/SS. Compared with colorectal procedures, patients undergoing pancreatic (OR 2.33; 95% CI 1.40–3.87; p = 0.00108) and small intestine (OR 1.27; 95% CI 1.12–1.44; p = 0.00015) surgery were more likely to develop S/SS. Patients undergoing biliary procedures (OR 0.38, 95% CI 0.30–0.48; p < 0.0001) were less likely to develop septic complications (Table 3).

The three most common emergency pancreatic, small intestine, and biliary procedures are listed in Table 4. Additional information on emergency cases included in the study cohort based on procedure type and study group are provided in Appendix B (S/SS) and Appendix C (No complication). Interestingly, the presence of diabetes (OR = 0.95; 95% CI 0.79–1.13; p = 0.54) did not correlate with the risk of developing S/SS. The adjusted odds of developing S/SS were higher for patients with increased age (OR 1.02; 95% CI 1.013–1.021; p < 0.0001) and those with longer procedure duration (OR 1.05; 95% CI 1.041–1.051; p < 0.0001).

Discussion

We found that ASA PS class, partially or totally dependent functional status, and male gender were associated with a higher risk of S/SS. Also, compared with colorectal procedures, patients undergoing emergency pancreatic and small intestine operations were at higher risk. The mortality rate among ES patients developing S/SS was 18.3%, much higher than the rate in non-septic ES patients.

The incidence of S/SS in our study population was 2%. This is lower than the 5% incidence of both S and SS in ES patients reported by Moore et al. in a retrospective NSQIP-based study that included data from 121 academic and community-based hospitals [9]. The difference may reflect the recent advances in early recognition with sepsis screening tools and treatment of sepsis in ES patients [16]. Alternatively, it may be explained by the increased accuracy of data within NSQIP over time, leading to vigilance in the coding of post-operative complications. Specifically, in 2011, a new variable was added to the database that aimed to differentiate complications as conditions present at the time of surgery, thereby identifying patients who had developed S/SS prior to their operation, not as a complication of the operation per se. Therefore, prior NSQIP-based studies may have overestimated the incidence of post-operative sepsis among ES patients. Our study suggests that the rate of post-operative S/SS after ES may be closer to that of elective surgery, which was previously quoted to be around 2% and 1.2%, respectively [9].

Predictors of post-operative sepsis have been studied previously. In addition to advanced age and non-white race, in a study of severe sepsis in 93 patients undergoing major surgery, Mokart et al. showed that male gender, Charlson Co-Morbidity Index, and ASA PS class >2 were associated with severe sepsis [17]. In addition, our study confirmed that male gender increases the risk of septic complications in patients undergoing ES. Prior studies have shown that blocking testosterone receptors in males is associated with improved immune function, whereas estradiol prevents suppression of the immune response [18, 19]. Additionally, epidemiologic studies have shown that female gender is associated with a lower risk of post-operative sepsis [20]. We also found that ASA PS >2 and non-independent functional status are associated with development of S/SS among ES patients. Both are likely markers of poor overall health and increased patient frailty and correlate with the overall rate of post-operative complications [21]. Understanding factors such as these associated with a higher incidence of S/SS may lead to more appropriate patient counseling and informed consent.

Several surgical risk assessment tools have been developed, including the Surgical Risk Scale (SRS), Physiologic and Operative Severity Score for the enumeration of Mortality and Morbidity (POSSUM), and the widely utilized ACS-NSQIP Surgical Risk Calculator (SRC) [22–24]. The NSQIP SRC already includes gender, functional status, and ASA PS class in its risk prediction algorithm.

Despite the multitude of surgical risk prediction models, to our knowledge, only the Emergency Surgery Score (ESS) predicts outcomes among ES patients. This tool uses 22 independent predictors of death to assign a score ranging from 0–29 [12]. The ESS has been shown to predict post-operative complications, including septic complications, in ES patients, [10]. It already utilizes functional status as one of the co-variables, and gender was included in multiple regression models when deriving the score. We suggest adding ASA PS class to the ESS to improve the predictive ability of the scoring system with respect to septic complications among ES patients.

The procedure-specific risk of post-operative sepsis has been studied also. An NIS-based study of 6.5 million patients undergoing elective surgery found that among abdominal procedures, patients undergoing pancreatic, gastric, and small bowel surgery were at greatest risk for the development of post-operative sepsis [13]. As in the study of elective surgery patients by Vogel et al., patients undergoing ES on the small intestine and pancreas in our study were at greater risk of septic complications.

To understand the higher risk of S/SS among patients undergoing emergency pancreatic procedures in our patient cohort, we identified the most common ES pancreatic procedures. We found that the three most common were pancreaticoduodenectomy (n = 46), distal pancreatectomy (n = 43), and pancreatic debridement for necrotizing pancreatitis (n = 17) (see Table 4). We suspect that in the case of emergency pancreatic debridement for necrotizing pancreatitis, S/SS was established prior to the procedure, as necrosectomy typically is indicated for the clinical deterioration of patients with infected necrosis. Emergency pancreaticoduodenectomy and emergency distal pancreatectomy for non-traumatic conditions is exceedingly rare, with few indications described in the literature [2,25,26]. Emergency pancreatic operations are high-risk procedures, performed when less invasive treatment approaches have failed or are unavailable. Given the rarity and complexity of these surgical procedures, often requiring long hospital stays, expert peri-operative care, and additional interventions, it is not difficult to see why this particular group of patients would be at higher risk of septic complications.

The mortality rate for non-S/SS ES patients was only 1%. This finding is consistent with the overall reduction in the mortality rate among ES patient in the past two decades [27]. Several recent improvements in the care of ES patients may be contributing to such a low mortality rate. These include early identification of patients in need of ES leading to earlier surgical consultation, the use of protocolized fluid management strategies, admission to high-level post-operative care areas, and continued involvement of senior surgical, anesthesia, and intensive care staff, among others [28]. It also appears that developing S/SS after either ES or elective surgery increases the mortality rate significantly. Vogel et al. examined trends in severe conditions from 1990–2006 (>78,000 patients) after elective surgery. The mortality rate was significantly higher among patients developing sepsis than in patients not suffering post-operative sepsis (25.88% vs. 0.81%; p < 0.0001). Similarly, we found that development of S/SS is associated with a nearly 20-fold increase in in-hospital deaths among ES patients.

Our study has several important limitations. The data are observational, and associations presented here do not represent causality. Furthermore, as we used a retrospective database, selection and input bias should be considered. Also, these analyses were conducted on a complete-case basis, and missing values may be a source of unknown bias.

Another major limitation of this study is the definitions of sepsis and septic shock used by NSQIP amidst evolving definitions utilized in other research studies and clinical practice. This may limit the broader applicability of the results. The most recent definitions of sepsis and septic shock, also known as the “Sepsis 3” definitions were created by the European Society of Intensive Care Medicine (ESICM) and Society of Critical Care Medicine (SCCM) Sepsis Redefinitions Task Force. The new definition requires the presence of “organ dysfunction,” attested to by the Sequential Organ Failure Assessment (SOFA) score, to diagnose sepsis and the need for vasopressor support to diagnose septic shock [4]. These new definitions were published in 2016, one year after our study period. The NSQIP definitions of sepsis, severe sepsis, and septic shock utilized during the study period are consistent with previous 2001 definitions established by the American College of Chest Physicians (ACCP) and SCCM International Sepsis Definitions Conference [29]. To a certain degree, this is a study of S/SS as defined by NSQIP, as opposed to the newer, more sensitive and specific Sepsis 3 definitions. We postulate that because Sepsis 3 abolishes the use of systemic inflammatory response syndrome (SIRS) and requires the use of vasopressors in its definition of septic shock, some of the patients who were in our S/SS group would move to the “no complication” group. This would make the difference in the mortality rate between the two groups even larger. In addition, given even greater potential paucity of numbers for S/SS patients in the pancreas surgery group, it is possible that the findings of the study regarding pancreatic procedures would fail to reach statistical significance.

We were unable to analyze certain factors known to be associated with post-operative sepsis, including institutional characteristics (i.e., teaching versus non-teaching hospital and hospital volume), socioeconomic factors, and operative details, such as whether source control was achieved and time to source control [3,13]. Furthermore, information regarding antibiotic type and timing, intra-operative blood loss, and peri-operative glycemic control was unavailable.

Sepsis is associated with more deaths and prolonged hospital stays, leading to higher costs of care and placing a significant burden on the healthcare system. Emergency surgery patients represent a unique population at greater risk for the development of S/SS. This study helped demonstrate factors significantly associated with the development of S/SS and demonstrated that the overall incidence appears to be lower than previously reported.

Conclusions

Development of S/SS among patients undergoing ES is associated with a substantial increase in the 30-day mortality rate. The risk of S/SS among patients undergoing ES was increased with male gender, higher ASA PS class, and partially or totally dependent functional status. We have found that patients undergoing ES procedures involving the pancreas are at higher risk than those having colorectal operations. Utilization of these findings with existing risk prediction models, such as adding ASA PS class to the ESS, may improve current risk prediction models with respect to septic complications.