AAST 2014 PLENARY PAPER
Impact of specific postoperative complications on the outcomes of emergency general surgery patients Christopher Cameron McCoy, MD, Brian R. Englum, MD, Jeffrey E. Keenan, MD, Steven N. Vaslef, MD, PhD, Mark L. Shapiro, MD, and John E. Scarborough, MD, Durham, North Carolina
The relative contribution of specific postoperative complications on mortality after emergency operations has not been previously described. Identifying specific contributors to postoperative mortality following acute care surgery will allow for significant improvement in the care of these patients. METHODS: Patients from the 2005 to 2011 American College of Surgeons’ National Surgical Quality Improvement Program database who underwent emergency operation by a general surgeon for one of seven diagnoses (gallbladder disease, gastroduodenal ulcer disease, intestinal ischemia, intestinal obstruction, intestinal perforation, diverticulitis, and abdominal wall hernia) were analyzed. Postoperative complications (pneumonia, myocardial infarction, incisional surgical site infection, organ/space surgical site infection, thromboembolic process, urinary tract infection, stroke, or major bleeding) were chosen based on surgical outcome measures monitored by national quality improvement initiatives and regulatory bodies. Regression techniques were used to determine the independent association between these complications and 30-day mortality, after adjustment for an array of patient- and procedure-related variables. RESULTS: Emergency operations accounted for 14.6% of the approximately 1.2 million general surgery procedures that are included in American College of Surgeons’ National Surgical Quality Improvement Program but for 53.5% of the 19,094 postoperative deaths. A total of 43,429 emergency general surgery patients were analyzed. Incisional surgical site infection had the highest incidence (6.7%). The second most common complication was pneumonia (5.7%). Stroke, major bleeding, myocardial infarction, and pneumonia exhibited the strongest associations with postoperative death. CONCLUSION: Given its disproportionate contribution to surgical mortality, emergency surgery represents an ideal focus for quality improvement. Of the potential postoperative targets for quality improvement, pneumonia, myocardial infarction, stroke, and major bleeding have the strongest associations with subsequent mortality. Since pneumonia is both relatively common after emergency surgery and strongly associated with postoperative death, it should receive priority as a target for surgical quality improvement initiatives. (J Trauma Acute Care Surg. 2015;78: 912Y919. Copyright * 2015 Wolters Kluwer Health, Inc. All rights reserved.) LEVEL OF EVIDENCE: Prognostic and epidemiologic study, level III. KEY WORDS: Acute care surgery; quality improvement; pneumonia; complications; outcomes. BACKGROUND:
T
he association between emergency operation and postoperative morbidity and mortality has been extensively documented in the literature. Recent data from the Michigan Surgical Quality Collaborative demonstrate that emergency procedures accounted for only 11% of the total number of surgical cases included in that regional data registry but for 28% of the postoperative complications and 47% of the postoperative deaths.1 A separate analysis of more than 400,000 general surgical procedures performed at 198 hospitals participating in the American College of Surgeons’ National Surgical Quality Improvement Program (ACS-NSQIP) demonstrated a 1.26-fold
Submitted: September 9, 2014, Revised: February 6, 2015, Accepted: February 9, 2015. From the Division of Trauma and Critical Care, Department of Surgery, Duke University, Durham, North Carolina. This study was presented at the 73rd annual meeting of the American Association for the Surgery of Trauma, September 9Y13, 2014, in Philadelphia, Pennsylvania. Address for reprints: Christopher Cameron McCoy, MD, Division of Trauma and Critical Care, Department of Surgery, Duke University, Durham, NC, DUMC 3443, Durham, NC 27710; email: [email protected]. DOI: 10.1097/TA.0000000000000611
increase in the risk of major postoperative morbidity and a 1.39-fold increase in the risk of postoperative death in patients who required emergency as opposed to elective operation.2 Given their disproportionate contribution to overall surgical morbidity and mortality, emergency surgery patients represent an ideal target for surgical quality improvement initiatives. Although the link between emergency operation and postoperative morbidity has been extensively described, little is known about the frequency with which specific complications occur after emergency surgery or the relative contribution that these complications make to subsequent mortality. Such information will be critical for determining whether improvements in the outcomes of emergency surgery patients can be achieved simply by a more aggressive application of existing quality initiatives to these patients or instead whether novel initiatives will be required for this patient population. The objectives of our study were therefore to (1) determine the incidence with which specific postoperative complications occurred in patients from the ACS-NSQIP database who underwent emergency general surgery and (2) determine the independent association between each of these complications and J Trauma Acute Care Surg Volume 78, Number 5
912
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
subsequent mortality after adjustment for a comprehensive array of patient- and procedure-related factors.
PATIENTS AND METHODS The 2005 to 2011 ACS-NSQIP Participant Use Data Files were used for our analysis. The data contained within this data source is prospectively collected by ACS-trained on-site surgical clinical reviewers and has been shown through interrater reliability audits to be highly accurate.3 Of the total number of procedures included in that data source, the percentage that were performed as an emergency and by a general surgeon were determined. Similarly, of the total number of 30-day postoperative deaths included in the data source, the percentage that followed emergency operation by a general surgeon were also determined. Subsequent analysis included all patients who underwent emergency operation by a general surgeon for an acute general surgical diagnosis. The diagnoses included for analysis (intestinal
obstruction, gallbladder disease, abdominal wall hernia, diverticular disease, gastroduodenal ulcer disease, intestinal ischemia, and intestinal perforation) were selected because they were the seven most frequent such diagnoses in the ACSNSQIP database with the exception of acute appendicitis. Patients were selected based on the DRG International Classification of DiseaseV9th Rev. (ICD-9) code that was associated with their index operation. Patients were excluded from our study if they (1) underwent another operation in the 30 days before their index procedure, (2) had any condition that might preclude a desire for aggressive management of postoperative complications (including the presence of a preoperative do-notresuscitate order, disseminated cancer, a tumor of the central nervous system, chemotherapy in the 30 days before index operation, or radiotherapy in the 90 days before index operation), or (3) had missing information for any of the potential predictor variables included in our analysis except race/ ethnicity (for which an ‘‘unknown’’ classification was created)
TABLE 1. Patient- and Procedure-Related Characteristics Considered as Potential Predictors of Primary 30-Day Outcomes After Emergency General Surgery Characteristic
Components if Composite Variable V
Patient age Patient sex Race/ethnicity Cerebrovascular disease Chronic cardiac disease Ascites Moderate ethanol use Diabetes mellitus Hypertension Chronic steroid use Bleeding disorder Decreased function Peripheral vascular disease End-stage renal disease Acute cognitive dysfunction Acute renal failure Acute cardiac disease Congestive heart failure Acute pulmonary disease Preoperative sepsis
V V Previous stroke with or without residual neurologic deficit, and/or previous transient ischemic attack Previous coronary artery bypass graft and/or coronary stenting procedure V V V V V V Quadriparesis, paraparesis, hemiplegia, and/or nonindependent functional health status Rest pain/gangrene and/or previous lower extremity revascularization procedure V Preoperative coma 9 24 h and/or impaired sensorium V Angina in previous 30 d and/or MI in previous 6 mo V Preoperative pneumonia and/or need for mechanical ventilation V
Preoperative serum albumin Preoperative infected wound ASA physical status classification Preoperative length of stay Incisional wound classification
V V V V V
Operative time Total work relative value units
V V
Classification for Multivariate Logistic Regression Models Each 10-y interval increase in age beyond 16Y39 y (ordered categorical variable) Male or female White (reference), black, Hispanic, other, or unknown No or yes No or yes No or yes No or yes None (reference), nonYinsulin requiring, insulin requiring No or yes No or yes No or yes No or yes No or yes No or yes No or yes No or yes No or yes No or yes No or yes None (reference), systemic inflammatory response syndrome, sepsis, septic shock Normal (Q3.5 g/dL), low (G3.5 g/dL), or missing No or yes 1Y3 or 4Y5 0Y1 d or 91 d Clean/clean-contaminated (reference), contaminated, dirty/infected Numeric Numeric
ASA, American Society of Anesthesiologists.
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
913
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
and preoperative serum albumin (for which an ‘‘unknown classification’’ was created). The primary outcome variables for our analysis were the incidences of 30-day postoperative end-organ dysfunction and mortality as well as the length of postoperative hospitalization (among those patients who survived to discharge). For this study, we defined a patient as having sustained end-organ dysfunction if they developed any of the following complications within 30 days of their index operation: need for postoperative mechanical ventilation greater than 48 hours, postoperative coma, septic shock, and/or postoperative renal insufficiency (with or without need for hemodialysis). The definition used by ACSNSQIP for septic shock is ‘‘Iclinical signs and symptoms of SIRS or sepsis AND documented organ and/or circulatory dysfunction.’’4 The primary predictor variables for our analysis were the presence or absence of each of the following complications in the 30 days following index operation: incisional surgical site infection (SSI) (including superficial and/or deep incisional SSI), organ/space SSI, urinary tract infection (UTI), thromboembolism (including deep venous thrombosis and/or pulmonary embolism), pneumonia, myocardial infarction (MI), stroke, and major bleeding (defined as the need for Q5 U of packed red blood cells within 72 hours of the completion of the index operation). Additional potential predictor variables included an array of patient- and procedure-related characteristics that are described and defined in Table 1. To determine the independent association between our primary predictor variables and 30-day postoperative end-organ dysfunction and mortality, nonparsimonious multivariate logistic regression models were developed for each outcome, which included all other known patient- and procedure-related factors as potential predictor variables including operative diagnosis. Odds ratios and 95% confidence limits for these models are presented. A nonparsimonious linear regression model was created using similar methods to determine the independent association between each postoperative complication and postoperative length of hospitalization among those patients in our study population who survived to hospital discharge. In this model, the logarithmic transformation of postoperative length of hospital stay was used as the dependent variable, and patients who experienced postoperative death were excluded. A coefficient and 95% confidence intervals (CIs) for this model are presented. To determine the relative associations of each of our postoperative complications and 30-day mortality, we calculated the survival time for patients who sustained one of the postoperative complications of interest in our study. Survival time was defined as either the interval between the date of development of a complication and the date of subsequent death (for patients who died within 30 days of their procedure) or the interval between the date of development of a complication and the 30th postoperative day (which is the extent of follow-up for patients who are included in the NSQIP). These survival times were then used to construct an unadjusted Kaplan-Meier 30-day survival curve for each of the specific postoperative complications included in our study and were also used as the dependent variable in a nonparsimonious Cox proportional hazards model to determine the independent association between each complication and subsequent mortality. Patients who did not sustain any of the 914
postoperative complications of our study or who sustained more than one such complication were excluded from this analysis. Patients who sustained one postoperative complication but in whom the timing of that complication was unknown were also
TABLE 2. Patient and Preoperative Characteristics of 43,429 Patients Undergoing Emergency General Surgery* Demographic Characteristics Age, mean (SD), y Female Race/ethnicity White Black Hispanic Other Unknown Chronic comorbid conditions Cerebrovascular disease Chronic cardiac disease Ascites Ethanol Use Diabetes mellitus None NonYinsulin requiring Insulin requiring Hypertension Chronic steroid use Bleeding disorder Decreased function Peripheral vascular disease End-stage renal disease Acute comorbid conditions Acute cognitive dysfunction Acute cardiac disease Acute pulmonary disease Acute renal failure Congestive heart failure Preoperative sepsis classification None Systemic inflammatory response syndrome Sepsis Septic shock Preoperative serum albumin Normal Low Missing Infected wound Other characteristics ASA physical status classification Q 4 Preoperative length of hospitalization 9 1 d Incisional wound classification Clean/clean-contaminated Contaminated Dirty/infected
59.3 (18.2) 24,472 (56.4%) 29,517 (68.0%) 5,191 (12.0%) 2,894 (6.7%) 1,430 (3.3%) 4,397 (10.1%) 3,117 (7.2%) 4,561 (10.5%) 2,422 (5.6%) 1,782 (4.1%) 37,172 (85.6%) 3,801 (8.8%) 2,456 (5.7%) 21,065 (48.5%) 2,336 (5.4%) 4,468 (10.3%) 7,749 (17.8%) 1,010 (2.3%) 1,073 (2.5%) 1,644 (3.8%) 945 (2.2%) 2,188 (5.0%) 1,078 (2.5%) 925 (2.1%) 26,974 (62.1%) 10,285 (23.7%) 3,789 (8.7%) 2,381 (5.5%) 20,782 (47.8%) 12,239 (28.2%) 10,408 (24.0%) 1,240 (2.9%) 8,157 (18.8%) 8,774 (20.2%) 25,279 (58.2%) 8,353 (19.2%) 9,797 (22.6%)
*Expressed as n (%) of overall study sample unless otherwise indicated. ASA, American Society of Anesthesiologists.
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
TABLE 3. Diagnosis-Specific 30-Day Postoperative Incidence of Specific Postoperative Complications in Patients Undergoing Emergency General Surgery Postoperative Complication Incisional SSI
Pneumonia Major Bleeding
UTI
Operative Diagnosis
Organ/Space SSI
Deep Venous Thrombosis/ Pulmonary Embolism
MI
Stroke
Complications, n (%)
Obstruction (n = 10,652) 977 (9.2) 715 (6.7) Gallbladder disease 183 (1.8) 140 (1.4) (n = 10,015) Hernia (n = 9,995) 605 (6.1) 347 (3.5) Diverticular disease 383 (10.9) 233\ (6.6) (n = 3,518) Gastroduodenal ulcer 215 (6.7) 334 (10.4) (n = 3,213) Intestinal ischemia 218 (7.2) 403 (13.3) (n = 3,041) Intestinal perforation 293 (10.4) 300 (10.7) (n = 2,818) All diagnoses 2,887 (6.7) 2,490 (5.7) (n = 43,429)
266 (2.5) 72 (0.7)
458 (4.3) 84 (0.8)
300 (2.8) 102 (1.0)
296 (2.8) 69 (0.7)
104 (1.0) 23 (0.2)
40 (0.4) 19 (0.2)
124 (1.2) 171 (4.9)
209 (2.1) 122 (3.5)
135 (1.4) 158 (4.5)
150 (1.5) 133 (3.8)
50 (0.5) 48 (1.4)
25 (0.3) 22 (0.6)
224 (7.0)
105 (3.3)
131 (4.1)
131 (4.1)
50 (1.6)
20 (0.6)
313 (10.3)
166 (5.5)
134 (4.4)
133 (4.4)
77 (2.5)
62 (2.0)
187 (6.6)
118 (4.2)
216 (7.7)
140 (5.0)
48 (1.7)
30 (1.1)
1,366 (3.2)
1,267 (2.9)
1,195 (2.8)
1,059 (2.4)
excluded. All statistical analyses were performed using Stata version 11.0 (STATA Corp., College Station, TX).
RESULTS Emergency surgery patients accounted for only 175,480 (14.6%) of the 1,198,287 operations in ACS-NSQIP that were performed by a general surgeon but for 10,207 (53.5%) of the 19,094 deaths that occurred within 30 days of operation. Of the 43,886 patients included in the 2005 to 2011 ACS-NSQIP database who underwent emergency operation for one of the seven general surgery diagnoses included in our study, 457 were excluded because they lacked information for one or more of the potential predictor variables used in our analysis. Patient
401 (0.9) 219 (0.5)
and procedure characteristics of the remaining 43,429 patients who were included for subsequent analysis are demonstrated in Table 2. Table 3 shows the 30-day postoperative incidences of specific complications in our study population, stratified by operative diagnosis. Incisional SSI was the most frequent complication after emergency general surgery, occurring in 2,877 (6.7%) patients. The second most common complication was pneumonia, which occurred in 2,499 patients (5.7%). MI and stroke were the least common complications in our study population, occurring in 401 patients (0.9%) and 219 patients (0.5%), respectively. The overall incidence of end-organ dysfunction in our study was 5,650 (13.0%). Figure 1 shows the incidence of
Figure 1. Incidence of complication-associated end organ dysfunction. EOD, end organ dysfunction; SSI, surgical site infection; UTI, urinary tract infection; MI, myocardial infarction. *Incidence; adjusted odds ratio (confidence interval); p-value. Odds ratio adjusted for known patient- and procedure-related variables described in Table 1. Area under ROC (receiver operating characteristic) curve = 0.9207. Hosmer-Lemeshow statistic = 290.67 (p G 0.001). * 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
915
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
end-organ dysfunction stratified by the specific type of complication sustained by patients postoperatively. Of the complications considered in our study, only incisional SSI did not portend a significantly increased risk of developing concomitant endorgan dysfunction. In contrast, 66.5% of the patients who developed postoperative pneumonia exhibited evidence of endorgan dysfunction (adjusted odds of end-organ dysfunction in setting of pneumonia, 8.88; 95% CI, 7.96Y9.90; p G 0.001). The overall 30-day mortality rate in our analysis was 6.8%. Figure 2 demonstrates the number of patients with each of the specific complication types who did or did not survive to their 30th postoperative day. The largest absolute number of postoperative deaths was demonstrated for patients with postoperative pneumonia, while the complication that was associated with the highest rate of 30-day postoperative mortality was postoperative stroke. To determine the adjusted associations between each of the specific complications and subsequent mortality, we next determined the survival time for each patient who sustained one of the eight specific postoperative complications considered in our study (with survival time being defined as the interval between the dates of complication and death [in patients who died] or between the date of complication and the 30th postoperative day [in patients who survived]). Patients who did not sustain any of the eight complications or patients who sustained more than one of these complications (n = 1,904) were excluded from this analysis, as were patients in whom the date of complication was unknown (n = 236), yielding a sample of 6,488 patients for our adjusted survival analysis. Figure 3 demonstrates the Kaplan-Meier 30-day survival curves for patients stratified by type of postoperative complication, and Table 4 shows the adjusted proportional hazards of postoperative death for each of the specific complications that we examined, using incisional SSI as the reference complication. Stroke, major bleeding, MI, and pneumonia exhibited the strongest associations with death, although patients who sustained UTI, organ/space SSI, or thromboembolism also demonstrated an increased risk of death when compared with patients who sustained incisional SSI.
Figure 3. Kaplan-Meier 30-day survival stratified by type of postoperative complication. SSI, surgical site infection; UTI, urinary tract infection; MI, myocardial infarction.
The median length of postoperative hospitalization among those patients in our study sample who survived to hospital discharge was 2 days (interquartile range, 5Y9 days). A linear regression model was used to determine the association between specific complications and postoperative length of hospital stay. The results of this model are demonstrated in Figure 4. Although all of the complications considered in our analysis were associated with a significant increase in the length of postoperative hospitalization, patients who sustained postoperative stroke or pneumonia required the longest postoperative hospital stays (medians 21 and 19 days; interquartile ranges, 0.24Y0.42 and 0.43Y0.48 days, respectively).
DISCUSSION Because preoperative risk factor modification is generally not feasible in patients who require emergency surgical intervention, efforts to improve the postoperative outcomes of such patients will necessarily require focus on either the prevention of postoperative morbidity or on the timely recognition and TABLE 4. Adjusted Proportional Hazard of Death After Experiencing Postoperative Complication After Emergency General Surgery Complication
Figure 2. Incidence of complication-associated mortality within 30 days. SSI, surgical site infection; URI, urinary tract infection; MI, myocardial infarction. *Incidence. 916
Adjusted Hazard Ratio*
95% CI
p
4.29 5.18 1.56 2.03 2.52 5.13 8.33
3.10Y5.94 3.69Y7.27 1.00Y2.43 1.31Y3.16 1.68Y3.77 3.33Y7.89 5.25Y13.2
G0.001 G0.001 0.05 0.002 G0.001 G0.001 G0.001
Pneumonia Major bleeding UTI Organ/space SSI Thromboembolism MI Stroke
*Adjusted proportional hazard of postoperative death after adjusting for patient- and procedure-related variables and operative diagnosis.
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
Figure 4. Median complication-associated postoperative length of stay. SSI, surgical site infection; UTI, urinary tract infection; MI, myocardial infarction. *Median postoperative length of stay; adjusted beta coefficient (confidence interval); p-value. Analysis based upon logarithmic transformation of postoperative length of stay. Adjusted R-squared = 0.6533.
management of complications that do occur.1,2,5 Although previous studies have examined the importance of postoperative morbidity as a determinant of subsequent surgical mortality, the current study is the first to our knowledge to describe the incidence of specific complications that occur after emergency general surgery and to determine the associations between these complications and subsequent death.5Y8 Of the complications that we included for analysis, postoperative pneumonia stands alone as being both relatively common after emergency general surgery and having a strong association with postoperative endorgan dysfunction and death. Other index complications that exhibited strong associations with both end-organ dysfunction and death (e.g., postoperative MI, stroke, and major bleeding) were much less common than pneumonia. In contrast, incisional SSI was the only complication that occurred more frequently than pneumonia, although this complication was associated with the fewest number of subsequent deaths of all of the index complications that we assessed. Taken together, the findings of our study suggest that postoperative pneumonia should receive priority as a target of initiatives that are designed to improve the outcomes of patients who require emergency general surgery. Existing quality initiatives that apply to emergency general surgery patients, most notably the Surgical Care Improvement Project (SCIP), do not currently include process measures targeting the prevention of postoperative pneumonia.9 The frequency with which this complication occurs and its significant impact on postoperative mortality and health care resource use mandate that a concerted effort be made by health care policymakers to codify such measures for emergency surgery patients. The framework for the development of process measures that are specific for the prevention of nosocomial pneumonia already exists.10,11 Perhaps, best known are ‘‘Wake Up and Breathe’’ protocols, which encourage the daily coupling of spontaneous awakening trials and spontaneous breathing trials in mechanically ventilated patients to reduce the incidence of
ventilator-associated pneumonia.10 In addition, Kazaure et al.11 recently reported the successful implementation of a pneumonia prevention program at their tertiary care center, which focused on nonintubated surgical patients who required only intermediatelevel care. This program, which included basic steps such as incentive spirometer use, aggressive ambulation, and proper head-of-bed position, was credited with reducing the incidence of postoperative pneumonia among ward patients at their center by 43.6% during a 5-year period. Although our analysis cannot predict whether such measures would be effective in reducing the incidence of postoperative pneumonia in the emergency general surgery population, our results suggest that there would be a potential for substantial clinical benefit and health care resource savings if such measures were successful. Our study demonstrates the significant impact that pneumonia has on outcomes and health care resource use after emergency surgery and thus provides rationale for prioritizing this complication as a target for quality improvement. Less clear is whether those processes of care that focus on the prevention of incisional SSI hold relevance to emergency general surgery patients. In our study, this complication does not increase the risk of postoperative end-organ dysfunction or death after emergency general surgery and has a relatively small effect on the length of postoperative hospitalization. Other studies have demonstrated that SSI accounts for a lower attributable fraction of total resource use than pneumonia after major surgery.12 Although process measures that target the prevention of postoperative SSI were originally developed for use in the elective surgery population, they are nevertheless also currently applied to patients who require emergency operation. However, there is a growing realization that hospital performance in elective general surgery operations does not necessarily correlate with performance in emergency operation. Given that the financial resources available for ensuring adherence to quality improvement process measures
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
917
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
are finite and compliance with SCIP measures is increasingly being tied to reimbursement by third-party payers, it would thus be reasonable to determine whether it is appropriate to apply each and every such measure to patients who require emergency general surgery.2,13,14 Our study has several important limitations. First, our definition of end-organ dysfunction is subjective and was constructed a priori based on those postoperative conditions that are tracked by ACS-NSQIP. Other possible definitions of endorgan dysfunction, although not explored in the current study, would therefore be reasonable. Second, it is likely that the associations between index complications and subsequent outcomes that we describe are significantly impacted by other variables that were not considered as confounders by our analysis (such as preoperative transfer status, etc.). Third, ACSNSQIP contains only patients who actually received operation for an emergency surgical condition. The incidence of specific complications in patients who are managed nonoperatively for such conditions is not known, and the results of our analysis therefore do not necessarily apply to such patients. Fourth, our description of the association between index complications and end-organ dysfunction cannot be assumed to imply causality. We do not know whether the index complications actually caused subsequent end-organ dysfunction or instead whether those complications occurred as a result of end-organ dysfunction. Finally, we do not know what impact existing performance improvement process measures may have already had on the incidence and impact of specific index complications that were included for analysis in our study. It is possible that without SCIP, the incidence and deleterious effect of complications such as incisional SSI might have been much greater than has been demonstrated in our study. However, the weight of available evidence does not suggest that SCIP has had a dramatic impact on the incidence of those complications that it is designed to prevent.14 Despite these limitations, our analysis of a large cohort of patients undergoing emergency general surgery demonstrates postoperative pneumonia to be a relatively important contributor to subsequent patient outcomes. Future effort should be made to develop and implement process measures for the prevention of this complication in patients who require emergency operation. Whether other process measures currently targeted by SCIP should continue to be applied to emergency operations should be a subject of debate. AUTHORSHIP Each author has significantly contributed to the completion of this study.
DISCLOSURE The authors declare no conflicts of interest.
REFERENCES 1. Smith M, Hussain A, Xiao J, Scheidler W, Reddy H, Olugbade K Jr, Cummings D, Terjimanian M, Krapohl G, Waits SA, et al. The importance of improving the quality of emergency surgery for a regional quality collaborative. Ann Surg. 2013;247(4):596Y602. 2. Ingraham AM, Cohen ME, Raval MV, Ko CY, Nathens AB. Comparison of hospital performance in emergency versus elective general
918
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
surgery operations at 198 hospitals. J Am Coll Surg. 2011;212: 20Y28.e1. Shiloach M, Frencher SK Jr, Steeger JE, Rowell KS, Bartzokis K, Tomeh MG, Richards KE, Ko CY, Hall BL. Toward robust information: data quality and inter-rater reliability in the American College of Surgeons National Surgical Quality Improvement Program. J Am Coll Surg. 2010;210:6Y16. User Guide for the 2011 ACS NSQIP Participant Use Data File. American College of Surgeons National Surgical Quality Improvement Program; 2012. Available at: http://site.acsnsqip.org/wp-content/uploads/2012/03/ 2011-User-Guide_Final.pdf. Sheetz KH, Krell RW, Englesbe MJ, Birkmeyer JD, Campbell DA Jr, Ghaferi AA. The importance of the first complication: understanding failure to rescue after emergent surgery in the elderly. J Am Coll Surg. 2014;219(3):365Y370. Khuri SF, Henderson WG, DePalma RG, Mosca C, Healey NA, Kumbhani DJ. Determinants of long-term survival after major surgery and the adverse effect of postoperative complications. Ann Surg. 2005;242(3):326Y341. Silber JH, Rosenbaum PR, Trudeau ME, Chen W, Zhang X, Kelz RR, Mosher RE, Even-Shoshan O. Changes in prognosis after the first postoperative complication. Med Care. 2005;43(2):122Y131. Bentrem DJ, Cohen ME, Hynes DM, Ko CY, Bilimoria KY. Identification of specific quality improvement opportunities for the elderly undergoing gastrointestinal surgery. Arch Surg. 2009;144(11):1013Y1020. Bratzler DW, Hunt DR. The surgical infection prevention and surgical care improvement projects: national initiatives to improve outcomes for patients having surgery. Clin Infect Dis. 2006;43(3):322Y330. Girard TD, Kress JP, Fuchs BD, Thomason JW, Schweickert WD, Pun BT, Taichman DB, Dunn JG, Pohlman AS, Kinniry PA, et al. Efficacy and safety of a paired sedation and ventilator weaning protocol for mechanically ventilated patients in intensive care (Awakening and Breathing Controlled trial): a randomized controlled trial. Lancet. 2008;371:126Y134. Kazaure HS, Martin M, Yoon JK, Wren SM. Long-term results of a postoperative pneumonia prevention program for the inpatient surgical ward. JAMA Surg. 2014;149(9):914Y918. Dimick JB, Pronovost PJ, Cowan JA, Lipsett PA. Complications and costs after high-risk surgery: where we should focus quality improvement initiatives. J Am Coll Surg .2003;196:671Y678. Ingraham AM, Cohen ME, Bilimoria KY, Raval MV, Ko CY, Nathens AB, Hall BL. Comparison of 30-day outcomes after emergency general surgery procedures: potential for targeted improvement. Surgery. 2010;148:217Y238. Weston A, Caldera K, Doron S. Surgical care improvement project in the value-based purchasing era: more harm than good? Clin Infect Dis. 2013;56:424Y427.
DISCUSSION Dr. David Harrington (Providence, Rhode Island): I would like to thank and congratulate Dr. McCoy and his fellow authors on a provocative manuscript and thank them for getting the manuscript well in advance. I will get right to the questions. First, you excluded 13% of your original study population due to missing critical data. That is not unusual or unreasonable. But I think you owe the audience a description of this population to make sure that it is reasonably similar to your included dataset. This would go some way to convince us that there has not been a confounding influence in your conclusions. Did you analyze this population or did you consider any imputation techniques to fill in this missing data? One-half of your data are appendectomies that, not surprisingly, have a much lower rate of complications. Did the overwhelming presence of such a low acuity of disease obscure conclusions you could have drawn with a higher acuity population? Did you, perhaps, do a subgroup analysis with appendectomies excluded? * 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
Third, you purposely did not explore the characteristics of premorbid disease or present any data into your analysis because you wanted, as you stated, to look at just the things that you could change V the postoperative complications. Okay, but if you used information present on admission you might be able to clarify some of the complications as preventable and others much more difficult, if nearly impossible, to prevent. Did you have any data on the pre-existing conditions or risk profile of the patients who developed complications? Some complications might be due to pre-existing conditions and some due to less-than-idea care. This would be a very interesting area for exploration. Lastly, I liked that you looked at mortality at 30 days and then also mortality that occurred after 7 days and the 7-to30-day mortality to eliminate potential survivor bias. It was, therefore, comforting to see that pneumonia and myocardial infarctions stayed significant. But how do you explain that both the 30-day and the 7-to-30 day mortality analysis was shown that surgical site infection was actually protective and actually in a statistically significant manner? Your odds ratio were .4 and .7 for surgical site infection. Does this suggest something we do not yet understand about surgical site infection? Or does this call into question your analysis? Or to end somewhat provocatively, does this suggest that we should all start spitting in our wounds so that we will then derive the protective effect of surgical site infection? I, again, congratulate Dr. McCoy for an interesting manuscript and look forward to his closing comments. Dr. Elliott Haut (Baltimore, Maryland): So I’ve got just one quick question about timing that might go along with your statements of correlation versus causation. You haven’t said ‘‘causation’’, you’ve said ‘‘correlation,’’ which I think is right. It’s really a chicken or egg. Which comes first, the complication or the end-organ dysfunction? Is it just that these patients with end-organ dysfunction stay in the ICU a really long time, develop DVT, get screened for DVT, get a pneumonia because they are on the ventilator, et cetera? Or does the complication come first, leading to the endorgan dysfunction? I’d like to know if you have data on that question and if you can provide it. Dr. Lewis Kaplan (Philadelphia, Pennsylvania): I enjoyed your presentation. I have one question that relates to definitions. NSQIP defines renal insufficiency, a term that has fallen out of favor, as a creatinine that exceeds two. Recognizing that that does not fit the RIFLE, AKIN or KDIGO metrics, how do you think using any of those
would impact your data? And how would that change what is the most fruitful area of prevention for the AAST in which to engage? Dr. Christopher C. McCoy (Durham, North Carolina): Thank you for these interesting questions. First to address Dr. Harrington’s question with regard to the exclusion of 13% of our original study population due to missing data. Originally we identified out of the 1.2 million total cases 175,000 emergency surgery cases that were performed by a general surgeon. We then isolated 100,000 that fit these eight diagnoses. Of those 100,000, 13,000 were excluded for missing predictor variable data. These excluded patients shared the eight diagnoses of interest and did not demonstrate any significant differences with regard to patient characteristics. With regard to the prevalence of appendectomy in this study population, we utilized a generalized estimating equation to compensate for the perceived over-representation of appendectomy in this dataset. We have discussed excluding appendectomy and repeating our analysis as well as applying our study techniques to major elective surgeries to look at the relationship between post-operative complications and mortality. With regard to pre-existing conditions and risk profiles of patients who developed complications, we adjusted for a wide array of acute and chronic comorbid conditions in this population in the regression analysis. It is also potentially important to consider some of these postoperative complications as surrogates for other events. These would include pneumonia associated with prolonged mechanical ventilation, as well as thromboembolic complications and urinary tract infection associated with prolonged use of indwelling catheters. A prospective study specifically designed to elucidate the root causes of these complications could address this issue. With regard to the relative risk of surgical site infection and its association with mortality, we do not believe that this data supports a protective effect from developing certain postoperative complications versus others. All five studied complications were associated with mortality rates that were equal to or greater than the 3% rate of 30-day post-operative mortality in the entire 87,000 patient study population. With regard to end-organ dysfunction, we utilized an a priori definition that included post-operative mechanical ventilation for greater than 48 hours, post-operative renal insufficiency and post-operative coma. We could easily go back and alter this definition to include other terms for the data dictionary and refine the analysis to test for robustness of our conclusions related to end-organ dysfunction. Thank you.
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
919
Impact of specific postoperative complications on the outcomes of emergency general surgery patients Christopher Cameron McCoy, MD, Brian R. Englum, MD, Jeffrey E. Keenan, MD, Steven N. Vaslef, MD, PhD, Mark L. Shapiro, MD, and John E. Scarborough, MD, Durham, North Carolina
The relative contribution of specific postoperative complications on mortality after emergency operations has not been previously described. Identifying specific contributors to postoperative mortality following acute care surgery will allow for significant improvement in the care of these patients. METHODS: Patients from the 2005 to 2011 American College of Surgeons’ National Surgical Quality Improvement Program database who underwent emergency operation by a general surgeon for one of seven diagnoses (gallbladder disease, gastroduodenal ulcer disease, intestinal ischemia, intestinal obstruction, intestinal perforation, diverticulitis, and abdominal wall hernia) were analyzed. Postoperative complications (pneumonia, myocardial infarction, incisional surgical site infection, organ/space surgical site infection, thromboembolic process, urinary tract infection, stroke, or major bleeding) were chosen based on surgical outcome measures monitored by national quality improvement initiatives and regulatory bodies. Regression techniques were used to determine the independent association between these complications and 30-day mortality, after adjustment for an array of patient- and procedure-related variables. RESULTS: Emergency operations accounted for 14.6% of the approximately 1.2 million general surgery procedures that are included in American College of Surgeons’ National Surgical Quality Improvement Program but for 53.5% of the 19,094 postoperative deaths. A total of 43,429 emergency general surgery patients were analyzed. Incisional surgical site infection had the highest incidence (6.7%). The second most common complication was pneumonia (5.7%). Stroke, major bleeding, myocardial infarction, and pneumonia exhibited the strongest associations with postoperative death. CONCLUSION: Given its disproportionate contribution to surgical mortality, emergency surgery represents an ideal focus for quality improvement. Of the potential postoperative targets for quality improvement, pneumonia, myocardial infarction, stroke, and major bleeding have the strongest associations with subsequent mortality. Since pneumonia is both relatively common after emergency surgery and strongly associated with postoperative death, it should receive priority as a target for surgical quality improvement initiatives. (J Trauma Acute Care Surg. 2015;78: 912Y919. Copyright * 2015 Wolters Kluwer Health, Inc. All rights reserved.) LEVEL OF EVIDENCE: Prognostic and epidemiologic study, level III. KEY WORDS: Acute care surgery; quality improvement; pneumonia; complications; outcomes. BACKGROUND:
T
he association between emergency operation and postoperative morbidity and mortality has been extensively documented in the literature. Recent data from the Michigan Surgical Quality Collaborative demonstrate that emergency procedures accounted for only 11% of the total number of surgical cases included in that regional data registry but for 28% of the postoperative complications and 47% of the postoperative deaths.1 A separate analysis of more than 400,000 general surgical procedures performed at 198 hospitals participating in the American College of Surgeons’ National Surgical Quality Improvement Program (ACS-NSQIP) demonstrated a 1.26-fold
Submitted: September 9, 2014, Revised: February 6, 2015, Accepted: February 9, 2015. From the Division of Trauma and Critical Care, Department of Surgery, Duke University, Durham, North Carolina. This study was presented at the 73rd annual meeting of the American Association for the Surgery of Trauma, September 9Y13, 2014, in Philadelphia, Pennsylvania. Address for reprints: Christopher Cameron McCoy, MD, Division of Trauma and Critical Care, Department of Surgery, Duke University, Durham, NC, DUMC 3443, Durham, NC 27710; email: [email protected]. DOI: 10.1097/TA.0000000000000611
increase in the risk of major postoperative morbidity and a 1.39-fold increase in the risk of postoperative death in patients who required emergency as opposed to elective operation.2 Given their disproportionate contribution to overall surgical morbidity and mortality, emergency surgery patients represent an ideal target for surgical quality improvement initiatives. Although the link between emergency operation and postoperative morbidity has been extensively described, little is known about the frequency with which specific complications occur after emergency surgery or the relative contribution that these complications make to subsequent mortality. Such information will be critical for determining whether improvements in the outcomes of emergency surgery patients can be achieved simply by a more aggressive application of existing quality initiatives to these patients or instead whether novel initiatives will be required for this patient population. The objectives of our study were therefore to (1) determine the incidence with which specific postoperative complications occurred in patients from the ACS-NSQIP database who underwent emergency general surgery and (2) determine the independent association between each of these complications and J Trauma Acute Care Surg Volume 78, Number 5
912
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
subsequent mortality after adjustment for a comprehensive array of patient- and procedure-related factors.
PATIENTS AND METHODS The 2005 to 2011 ACS-NSQIP Participant Use Data Files were used for our analysis. The data contained within this data source is prospectively collected by ACS-trained on-site surgical clinical reviewers and has been shown through interrater reliability audits to be highly accurate.3 Of the total number of procedures included in that data source, the percentage that were performed as an emergency and by a general surgeon were determined. Similarly, of the total number of 30-day postoperative deaths included in the data source, the percentage that followed emergency operation by a general surgeon were also determined. Subsequent analysis included all patients who underwent emergency operation by a general surgeon for an acute general surgical diagnosis. The diagnoses included for analysis (intestinal
obstruction, gallbladder disease, abdominal wall hernia, diverticular disease, gastroduodenal ulcer disease, intestinal ischemia, and intestinal perforation) were selected because they were the seven most frequent such diagnoses in the ACSNSQIP database with the exception of acute appendicitis. Patients were selected based on the DRG International Classification of DiseaseV9th Rev. (ICD-9) code that was associated with their index operation. Patients were excluded from our study if they (1) underwent another operation in the 30 days before their index procedure, (2) had any condition that might preclude a desire for aggressive management of postoperative complications (including the presence of a preoperative do-notresuscitate order, disseminated cancer, a tumor of the central nervous system, chemotherapy in the 30 days before index operation, or radiotherapy in the 90 days before index operation), or (3) had missing information for any of the potential predictor variables included in our analysis except race/ ethnicity (for which an ‘‘unknown’’ classification was created)
TABLE 1. Patient- and Procedure-Related Characteristics Considered as Potential Predictors of Primary 30-Day Outcomes After Emergency General Surgery Characteristic
Components if Composite Variable V
Patient age Patient sex Race/ethnicity Cerebrovascular disease Chronic cardiac disease Ascites Moderate ethanol use Diabetes mellitus Hypertension Chronic steroid use Bleeding disorder Decreased function Peripheral vascular disease End-stage renal disease Acute cognitive dysfunction Acute renal failure Acute cardiac disease Congestive heart failure Acute pulmonary disease Preoperative sepsis
V V Previous stroke with or without residual neurologic deficit, and/or previous transient ischemic attack Previous coronary artery bypass graft and/or coronary stenting procedure V V V V V V Quadriparesis, paraparesis, hemiplegia, and/or nonindependent functional health status Rest pain/gangrene and/or previous lower extremity revascularization procedure V Preoperative coma 9 24 h and/or impaired sensorium V Angina in previous 30 d and/or MI in previous 6 mo V Preoperative pneumonia and/or need for mechanical ventilation V
Preoperative serum albumin Preoperative infected wound ASA physical status classification Preoperative length of stay Incisional wound classification
V V V V V
Operative time Total work relative value units
V V
Classification for Multivariate Logistic Regression Models Each 10-y interval increase in age beyond 16Y39 y (ordered categorical variable) Male or female White (reference), black, Hispanic, other, or unknown No or yes No or yes No or yes No or yes None (reference), nonYinsulin requiring, insulin requiring No or yes No or yes No or yes No or yes No or yes No or yes No or yes No or yes No or yes No or yes No or yes None (reference), systemic inflammatory response syndrome, sepsis, septic shock Normal (Q3.5 g/dL), low (G3.5 g/dL), or missing No or yes 1Y3 or 4Y5 0Y1 d or 91 d Clean/clean-contaminated (reference), contaminated, dirty/infected Numeric Numeric
ASA, American Society of Anesthesiologists.
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
913
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
and preoperative serum albumin (for which an ‘‘unknown classification’’ was created). The primary outcome variables for our analysis were the incidences of 30-day postoperative end-organ dysfunction and mortality as well as the length of postoperative hospitalization (among those patients who survived to discharge). For this study, we defined a patient as having sustained end-organ dysfunction if they developed any of the following complications within 30 days of their index operation: need for postoperative mechanical ventilation greater than 48 hours, postoperative coma, septic shock, and/or postoperative renal insufficiency (with or without need for hemodialysis). The definition used by ACSNSQIP for septic shock is ‘‘Iclinical signs and symptoms of SIRS or sepsis AND documented organ and/or circulatory dysfunction.’’4 The primary predictor variables for our analysis were the presence or absence of each of the following complications in the 30 days following index operation: incisional surgical site infection (SSI) (including superficial and/or deep incisional SSI), organ/space SSI, urinary tract infection (UTI), thromboembolism (including deep venous thrombosis and/or pulmonary embolism), pneumonia, myocardial infarction (MI), stroke, and major bleeding (defined as the need for Q5 U of packed red blood cells within 72 hours of the completion of the index operation). Additional potential predictor variables included an array of patient- and procedure-related characteristics that are described and defined in Table 1. To determine the independent association between our primary predictor variables and 30-day postoperative end-organ dysfunction and mortality, nonparsimonious multivariate logistic regression models were developed for each outcome, which included all other known patient- and procedure-related factors as potential predictor variables including operative diagnosis. Odds ratios and 95% confidence limits for these models are presented. A nonparsimonious linear regression model was created using similar methods to determine the independent association between each postoperative complication and postoperative length of hospitalization among those patients in our study population who survived to hospital discharge. In this model, the logarithmic transformation of postoperative length of hospital stay was used as the dependent variable, and patients who experienced postoperative death were excluded. A coefficient and 95% confidence intervals (CIs) for this model are presented. To determine the relative associations of each of our postoperative complications and 30-day mortality, we calculated the survival time for patients who sustained one of the postoperative complications of interest in our study. Survival time was defined as either the interval between the date of development of a complication and the date of subsequent death (for patients who died within 30 days of their procedure) or the interval between the date of development of a complication and the 30th postoperative day (which is the extent of follow-up for patients who are included in the NSQIP). These survival times were then used to construct an unadjusted Kaplan-Meier 30-day survival curve for each of the specific postoperative complications included in our study and were also used as the dependent variable in a nonparsimonious Cox proportional hazards model to determine the independent association between each complication and subsequent mortality. Patients who did not sustain any of the 914
postoperative complications of our study or who sustained more than one such complication were excluded from this analysis. Patients who sustained one postoperative complication but in whom the timing of that complication was unknown were also
TABLE 2. Patient and Preoperative Characteristics of 43,429 Patients Undergoing Emergency General Surgery* Demographic Characteristics Age, mean (SD), y Female Race/ethnicity White Black Hispanic Other Unknown Chronic comorbid conditions Cerebrovascular disease Chronic cardiac disease Ascites Ethanol Use Diabetes mellitus None NonYinsulin requiring Insulin requiring Hypertension Chronic steroid use Bleeding disorder Decreased function Peripheral vascular disease End-stage renal disease Acute comorbid conditions Acute cognitive dysfunction Acute cardiac disease Acute pulmonary disease Acute renal failure Congestive heart failure Preoperative sepsis classification None Systemic inflammatory response syndrome Sepsis Septic shock Preoperative serum albumin Normal Low Missing Infected wound Other characteristics ASA physical status classification Q 4 Preoperative length of hospitalization 9 1 d Incisional wound classification Clean/clean-contaminated Contaminated Dirty/infected
59.3 (18.2) 24,472 (56.4%) 29,517 (68.0%) 5,191 (12.0%) 2,894 (6.7%) 1,430 (3.3%) 4,397 (10.1%) 3,117 (7.2%) 4,561 (10.5%) 2,422 (5.6%) 1,782 (4.1%) 37,172 (85.6%) 3,801 (8.8%) 2,456 (5.7%) 21,065 (48.5%) 2,336 (5.4%) 4,468 (10.3%) 7,749 (17.8%) 1,010 (2.3%) 1,073 (2.5%) 1,644 (3.8%) 945 (2.2%) 2,188 (5.0%) 1,078 (2.5%) 925 (2.1%) 26,974 (62.1%) 10,285 (23.7%) 3,789 (8.7%) 2,381 (5.5%) 20,782 (47.8%) 12,239 (28.2%) 10,408 (24.0%) 1,240 (2.9%) 8,157 (18.8%) 8,774 (20.2%) 25,279 (58.2%) 8,353 (19.2%) 9,797 (22.6%)
*Expressed as n (%) of overall study sample unless otherwise indicated. ASA, American Society of Anesthesiologists.
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
TABLE 3. Diagnosis-Specific 30-Day Postoperative Incidence of Specific Postoperative Complications in Patients Undergoing Emergency General Surgery Postoperative Complication Incisional SSI
Pneumonia Major Bleeding
UTI
Operative Diagnosis
Organ/Space SSI
Deep Venous Thrombosis/ Pulmonary Embolism
MI
Stroke
Complications, n (%)
Obstruction (n = 10,652) 977 (9.2) 715 (6.7) Gallbladder disease 183 (1.8) 140 (1.4) (n = 10,015) Hernia (n = 9,995) 605 (6.1) 347 (3.5) Diverticular disease 383 (10.9) 233\ (6.6) (n = 3,518) Gastroduodenal ulcer 215 (6.7) 334 (10.4) (n = 3,213) Intestinal ischemia 218 (7.2) 403 (13.3) (n = 3,041) Intestinal perforation 293 (10.4) 300 (10.7) (n = 2,818) All diagnoses 2,887 (6.7) 2,490 (5.7) (n = 43,429)
266 (2.5) 72 (0.7)
458 (4.3) 84 (0.8)
300 (2.8) 102 (1.0)
296 (2.8) 69 (0.7)
104 (1.0) 23 (0.2)
40 (0.4) 19 (0.2)
124 (1.2) 171 (4.9)
209 (2.1) 122 (3.5)
135 (1.4) 158 (4.5)
150 (1.5) 133 (3.8)
50 (0.5) 48 (1.4)
25 (0.3) 22 (0.6)
224 (7.0)
105 (3.3)
131 (4.1)
131 (4.1)
50 (1.6)
20 (0.6)
313 (10.3)
166 (5.5)
134 (4.4)
133 (4.4)
77 (2.5)
62 (2.0)
187 (6.6)
118 (4.2)
216 (7.7)
140 (5.0)
48 (1.7)
30 (1.1)
1,366 (3.2)
1,267 (2.9)
1,195 (2.8)
1,059 (2.4)
excluded. All statistical analyses were performed using Stata version 11.0 (STATA Corp., College Station, TX).
RESULTS Emergency surgery patients accounted for only 175,480 (14.6%) of the 1,198,287 operations in ACS-NSQIP that were performed by a general surgeon but for 10,207 (53.5%) of the 19,094 deaths that occurred within 30 days of operation. Of the 43,886 patients included in the 2005 to 2011 ACS-NSQIP database who underwent emergency operation for one of the seven general surgery diagnoses included in our study, 457 were excluded because they lacked information for one or more of the potential predictor variables used in our analysis. Patient
401 (0.9) 219 (0.5)
and procedure characteristics of the remaining 43,429 patients who were included for subsequent analysis are demonstrated in Table 2. Table 3 shows the 30-day postoperative incidences of specific complications in our study population, stratified by operative diagnosis. Incisional SSI was the most frequent complication after emergency general surgery, occurring in 2,877 (6.7%) patients. The second most common complication was pneumonia, which occurred in 2,499 patients (5.7%). MI and stroke were the least common complications in our study population, occurring in 401 patients (0.9%) and 219 patients (0.5%), respectively. The overall incidence of end-organ dysfunction in our study was 5,650 (13.0%). Figure 1 shows the incidence of
Figure 1. Incidence of complication-associated end organ dysfunction. EOD, end organ dysfunction; SSI, surgical site infection; UTI, urinary tract infection; MI, myocardial infarction. *Incidence; adjusted odds ratio (confidence interval); p-value. Odds ratio adjusted for known patient- and procedure-related variables described in Table 1. Area under ROC (receiver operating characteristic) curve = 0.9207. Hosmer-Lemeshow statistic = 290.67 (p G 0.001). * 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
915
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
end-organ dysfunction stratified by the specific type of complication sustained by patients postoperatively. Of the complications considered in our study, only incisional SSI did not portend a significantly increased risk of developing concomitant endorgan dysfunction. In contrast, 66.5% of the patients who developed postoperative pneumonia exhibited evidence of endorgan dysfunction (adjusted odds of end-organ dysfunction in setting of pneumonia, 8.88; 95% CI, 7.96Y9.90; p G 0.001). The overall 30-day mortality rate in our analysis was 6.8%. Figure 2 demonstrates the number of patients with each of the specific complication types who did or did not survive to their 30th postoperative day. The largest absolute number of postoperative deaths was demonstrated for patients with postoperative pneumonia, while the complication that was associated with the highest rate of 30-day postoperative mortality was postoperative stroke. To determine the adjusted associations between each of the specific complications and subsequent mortality, we next determined the survival time for each patient who sustained one of the eight specific postoperative complications considered in our study (with survival time being defined as the interval between the dates of complication and death [in patients who died] or between the date of complication and the 30th postoperative day [in patients who survived]). Patients who did not sustain any of the eight complications or patients who sustained more than one of these complications (n = 1,904) were excluded from this analysis, as were patients in whom the date of complication was unknown (n = 236), yielding a sample of 6,488 patients for our adjusted survival analysis. Figure 3 demonstrates the Kaplan-Meier 30-day survival curves for patients stratified by type of postoperative complication, and Table 4 shows the adjusted proportional hazards of postoperative death for each of the specific complications that we examined, using incisional SSI as the reference complication. Stroke, major bleeding, MI, and pneumonia exhibited the strongest associations with death, although patients who sustained UTI, organ/space SSI, or thromboembolism also demonstrated an increased risk of death when compared with patients who sustained incisional SSI.
Figure 3. Kaplan-Meier 30-day survival stratified by type of postoperative complication. SSI, surgical site infection; UTI, urinary tract infection; MI, myocardial infarction.
The median length of postoperative hospitalization among those patients in our study sample who survived to hospital discharge was 2 days (interquartile range, 5Y9 days). A linear regression model was used to determine the association between specific complications and postoperative length of hospital stay. The results of this model are demonstrated in Figure 4. Although all of the complications considered in our analysis were associated with a significant increase in the length of postoperative hospitalization, patients who sustained postoperative stroke or pneumonia required the longest postoperative hospital stays (medians 21 and 19 days; interquartile ranges, 0.24Y0.42 and 0.43Y0.48 days, respectively).
DISCUSSION Because preoperative risk factor modification is generally not feasible in patients who require emergency surgical intervention, efforts to improve the postoperative outcomes of such patients will necessarily require focus on either the prevention of postoperative morbidity or on the timely recognition and TABLE 4. Adjusted Proportional Hazard of Death After Experiencing Postoperative Complication After Emergency General Surgery Complication
Figure 2. Incidence of complication-associated mortality within 30 days. SSI, surgical site infection; URI, urinary tract infection; MI, myocardial infarction. *Incidence. 916
Adjusted Hazard Ratio*
95% CI
p
4.29 5.18 1.56 2.03 2.52 5.13 8.33
3.10Y5.94 3.69Y7.27 1.00Y2.43 1.31Y3.16 1.68Y3.77 3.33Y7.89 5.25Y13.2
G0.001 G0.001 0.05 0.002 G0.001 G0.001 G0.001
Pneumonia Major bleeding UTI Organ/space SSI Thromboembolism MI Stroke
*Adjusted proportional hazard of postoperative death after adjusting for patient- and procedure-related variables and operative diagnosis.
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
Figure 4. Median complication-associated postoperative length of stay. SSI, surgical site infection; UTI, urinary tract infection; MI, myocardial infarction. *Median postoperative length of stay; adjusted beta coefficient (confidence interval); p-value. Analysis based upon logarithmic transformation of postoperative length of stay. Adjusted R-squared = 0.6533.
management of complications that do occur.1,2,5 Although previous studies have examined the importance of postoperative morbidity as a determinant of subsequent surgical mortality, the current study is the first to our knowledge to describe the incidence of specific complications that occur after emergency general surgery and to determine the associations between these complications and subsequent death.5Y8 Of the complications that we included for analysis, postoperative pneumonia stands alone as being both relatively common after emergency general surgery and having a strong association with postoperative endorgan dysfunction and death. Other index complications that exhibited strong associations with both end-organ dysfunction and death (e.g., postoperative MI, stroke, and major bleeding) were much less common than pneumonia. In contrast, incisional SSI was the only complication that occurred more frequently than pneumonia, although this complication was associated with the fewest number of subsequent deaths of all of the index complications that we assessed. Taken together, the findings of our study suggest that postoperative pneumonia should receive priority as a target of initiatives that are designed to improve the outcomes of patients who require emergency general surgery. Existing quality initiatives that apply to emergency general surgery patients, most notably the Surgical Care Improvement Project (SCIP), do not currently include process measures targeting the prevention of postoperative pneumonia.9 The frequency with which this complication occurs and its significant impact on postoperative mortality and health care resource use mandate that a concerted effort be made by health care policymakers to codify such measures for emergency surgery patients. The framework for the development of process measures that are specific for the prevention of nosocomial pneumonia already exists.10,11 Perhaps, best known are ‘‘Wake Up and Breathe’’ protocols, which encourage the daily coupling of spontaneous awakening trials and spontaneous breathing trials in mechanically ventilated patients to reduce the incidence of
ventilator-associated pneumonia.10 In addition, Kazaure et al.11 recently reported the successful implementation of a pneumonia prevention program at their tertiary care center, which focused on nonintubated surgical patients who required only intermediatelevel care. This program, which included basic steps such as incentive spirometer use, aggressive ambulation, and proper head-of-bed position, was credited with reducing the incidence of postoperative pneumonia among ward patients at their center by 43.6% during a 5-year period. Although our analysis cannot predict whether such measures would be effective in reducing the incidence of postoperative pneumonia in the emergency general surgery population, our results suggest that there would be a potential for substantial clinical benefit and health care resource savings if such measures were successful. Our study demonstrates the significant impact that pneumonia has on outcomes and health care resource use after emergency surgery and thus provides rationale for prioritizing this complication as a target for quality improvement. Less clear is whether those processes of care that focus on the prevention of incisional SSI hold relevance to emergency general surgery patients. In our study, this complication does not increase the risk of postoperative end-organ dysfunction or death after emergency general surgery and has a relatively small effect on the length of postoperative hospitalization. Other studies have demonstrated that SSI accounts for a lower attributable fraction of total resource use than pneumonia after major surgery.12 Although process measures that target the prevention of postoperative SSI were originally developed for use in the elective surgery population, they are nevertheless also currently applied to patients who require emergency operation. However, there is a growing realization that hospital performance in elective general surgery operations does not necessarily correlate with performance in emergency operation. Given that the financial resources available for ensuring adherence to quality improvement process measures
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
917
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
are finite and compliance with SCIP measures is increasingly being tied to reimbursement by third-party payers, it would thus be reasonable to determine whether it is appropriate to apply each and every such measure to patients who require emergency general surgery.2,13,14 Our study has several important limitations. First, our definition of end-organ dysfunction is subjective and was constructed a priori based on those postoperative conditions that are tracked by ACS-NSQIP. Other possible definitions of endorgan dysfunction, although not explored in the current study, would therefore be reasonable. Second, it is likely that the associations between index complications and subsequent outcomes that we describe are significantly impacted by other variables that were not considered as confounders by our analysis (such as preoperative transfer status, etc.). Third, ACSNSQIP contains only patients who actually received operation for an emergency surgical condition. The incidence of specific complications in patients who are managed nonoperatively for such conditions is not known, and the results of our analysis therefore do not necessarily apply to such patients. Fourth, our description of the association between index complications and end-organ dysfunction cannot be assumed to imply causality. We do not know whether the index complications actually caused subsequent end-organ dysfunction or instead whether those complications occurred as a result of end-organ dysfunction. Finally, we do not know what impact existing performance improvement process measures may have already had on the incidence and impact of specific index complications that were included for analysis in our study. It is possible that without SCIP, the incidence and deleterious effect of complications such as incisional SSI might have been much greater than has been demonstrated in our study. However, the weight of available evidence does not suggest that SCIP has had a dramatic impact on the incidence of those complications that it is designed to prevent.14 Despite these limitations, our analysis of a large cohort of patients undergoing emergency general surgery demonstrates postoperative pneumonia to be a relatively important contributor to subsequent patient outcomes. Future effort should be made to develop and implement process measures for the prevention of this complication in patients who require emergency operation. Whether other process measures currently targeted by SCIP should continue to be applied to emergency operations should be a subject of debate. AUTHORSHIP Each author has significantly contributed to the completion of this study.
DISCLOSURE The authors declare no conflicts of interest.
REFERENCES 1. Smith M, Hussain A, Xiao J, Scheidler W, Reddy H, Olugbade K Jr, Cummings D, Terjimanian M, Krapohl G, Waits SA, et al. The importance of improving the quality of emergency surgery for a regional quality collaborative. Ann Surg. 2013;247(4):596Y602. 2. Ingraham AM, Cohen ME, Raval MV, Ko CY, Nathens AB. Comparison of hospital performance in emergency versus elective general
918
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
surgery operations at 198 hospitals. J Am Coll Surg. 2011;212: 20Y28.e1. Shiloach M, Frencher SK Jr, Steeger JE, Rowell KS, Bartzokis K, Tomeh MG, Richards KE, Ko CY, Hall BL. Toward robust information: data quality and inter-rater reliability in the American College of Surgeons National Surgical Quality Improvement Program. J Am Coll Surg. 2010;210:6Y16. User Guide for the 2011 ACS NSQIP Participant Use Data File. American College of Surgeons National Surgical Quality Improvement Program; 2012. Available at: http://site.acsnsqip.org/wp-content/uploads/2012/03/ 2011-User-Guide_Final.pdf. Sheetz KH, Krell RW, Englesbe MJ, Birkmeyer JD, Campbell DA Jr, Ghaferi AA. The importance of the first complication: understanding failure to rescue after emergent surgery in the elderly. J Am Coll Surg. 2014;219(3):365Y370. Khuri SF, Henderson WG, DePalma RG, Mosca C, Healey NA, Kumbhani DJ. Determinants of long-term survival after major surgery and the adverse effect of postoperative complications. Ann Surg. 2005;242(3):326Y341. Silber JH, Rosenbaum PR, Trudeau ME, Chen W, Zhang X, Kelz RR, Mosher RE, Even-Shoshan O. Changes in prognosis after the first postoperative complication. Med Care. 2005;43(2):122Y131. Bentrem DJ, Cohen ME, Hynes DM, Ko CY, Bilimoria KY. Identification of specific quality improvement opportunities for the elderly undergoing gastrointestinal surgery. Arch Surg. 2009;144(11):1013Y1020. Bratzler DW, Hunt DR. The surgical infection prevention and surgical care improvement projects: national initiatives to improve outcomes for patients having surgery. Clin Infect Dis. 2006;43(3):322Y330. Girard TD, Kress JP, Fuchs BD, Thomason JW, Schweickert WD, Pun BT, Taichman DB, Dunn JG, Pohlman AS, Kinniry PA, et al. Efficacy and safety of a paired sedation and ventilator weaning protocol for mechanically ventilated patients in intensive care (Awakening and Breathing Controlled trial): a randomized controlled trial. Lancet. 2008;371:126Y134. Kazaure HS, Martin M, Yoon JK, Wren SM. Long-term results of a postoperative pneumonia prevention program for the inpatient surgical ward. JAMA Surg. 2014;149(9):914Y918. Dimick JB, Pronovost PJ, Cowan JA, Lipsett PA. Complications and costs after high-risk surgery: where we should focus quality improvement initiatives. J Am Coll Surg .2003;196:671Y678. Ingraham AM, Cohen ME, Bilimoria KY, Raval MV, Ko CY, Nathens AB, Hall BL. Comparison of 30-day outcomes after emergency general surgery procedures: potential for targeted improvement. Surgery. 2010;148:217Y238. Weston A, Caldera K, Doron S. Surgical care improvement project in the value-based purchasing era: more harm than good? Clin Infect Dis. 2013;56:424Y427.
DISCUSSION Dr. David Harrington (Providence, Rhode Island): I would like to thank and congratulate Dr. McCoy and his fellow authors on a provocative manuscript and thank them for getting the manuscript well in advance. I will get right to the questions. First, you excluded 13% of your original study population due to missing critical data. That is not unusual or unreasonable. But I think you owe the audience a description of this population to make sure that it is reasonably similar to your included dataset. This would go some way to convince us that there has not been a confounding influence in your conclusions. Did you analyze this population or did you consider any imputation techniques to fill in this missing data? One-half of your data are appendectomies that, not surprisingly, have a much lower rate of complications. Did the overwhelming presence of such a low acuity of disease obscure conclusions you could have drawn with a higher acuity population? Did you, perhaps, do a subgroup analysis with appendectomies excluded? * 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Trauma Acute Care Surg Volume 78, Number 5
McCoy et al.
Third, you purposely did not explore the characteristics of premorbid disease or present any data into your analysis because you wanted, as you stated, to look at just the things that you could change V the postoperative complications. Okay, but if you used information present on admission you might be able to clarify some of the complications as preventable and others much more difficult, if nearly impossible, to prevent. Did you have any data on the pre-existing conditions or risk profile of the patients who developed complications? Some complications might be due to pre-existing conditions and some due to less-than-idea care. This would be a very interesting area for exploration. Lastly, I liked that you looked at mortality at 30 days and then also mortality that occurred after 7 days and the 7-to30-day mortality to eliminate potential survivor bias. It was, therefore, comforting to see that pneumonia and myocardial infarctions stayed significant. But how do you explain that both the 30-day and the 7-to-30 day mortality analysis was shown that surgical site infection was actually protective and actually in a statistically significant manner? Your odds ratio were .4 and .7 for surgical site infection. Does this suggest something we do not yet understand about surgical site infection? Or does this call into question your analysis? Or to end somewhat provocatively, does this suggest that we should all start spitting in our wounds so that we will then derive the protective effect of surgical site infection? I, again, congratulate Dr. McCoy for an interesting manuscript and look forward to his closing comments. Dr. Elliott Haut (Baltimore, Maryland): So I’ve got just one quick question about timing that might go along with your statements of correlation versus causation. You haven’t said ‘‘causation’’, you’ve said ‘‘correlation,’’ which I think is right. It’s really a chicken or egg. Which comes first, the complication or the end-organ dysfunction? Is it just that these patients with end-organ dysfunction stay in the ICU a really long time, develop DVT, get screened for DVT, get a pneumonia because they are on the ventilator, et cetera? Or does the complication come first, leading to the endorgan dysfunction? I’d like to know if you have data on that question and if you can provide it. Dr. Lewis Kaplan (Philadelphia, Pennsylvania): I enjoyed your presentation. I have one question that relates to definitions. NSQIP defines renal insufficiency, a term that has fallen out of favor, as a creatinine that exceeds two. Recognizing that that does not fit the RIFLE, AKIN or KDIGO metrics, how do you think using any of those
would impact your data? And how would that change what is the most fruitful area of prevention for the AAST in which to engage? Dr. Christopher C. McCoy (Durham, North Carolina): Thank you for these interesting questions. First to address Dr. Harrington’s question with regard to the exclusion of 13% of our original study population due to missing data. Originally we identified out of the 1.2 million total cases 175,000 emergency surgery cases that were performed by a general surgeon. We then isolated 100,000 that fit these eight diagnoses. Of those 100,000, 13,000 were excluded for missing predictor variable data. These excluded patients shared the eight diagnoses of interest and did not demonstrate any significant differences with regard to patient characteristics. With regard to the prevalence of appendectomy in this study population, we utilized a generalized estimating equation to compensate for the perceived over-representation of appendectomy in this dataset. We have discussed excluding appendectomy and repeating our analysis as well as applying our study techniques to major elective surgeries to look at the relationship between post-operative complications and mortality. With regard to pre-existing conditions and risk profiles of patients who developed complications, we adjusted for a wide array of acute and chronic comorbid conditions in this population in the regression analysis. It is also potentially important to consider some of these postoperative complications as surrogates for other events. These would include pneumonia associated with prolonged mechanical ventilation, as well as thromboembolic complications and urinary tract infection associated with prolonged use of indwelling catheters. A prospective study specifically designed to elucidate the root causes of these complications could address this issue. With regard to the relative risk of surgical site infection and its association with mortality, we do not believe that this data supports a protective effect from developing certain postoperative complications versus others. All five studied complications were associated with mortality rates that were equal to or greater than the 3% rate of 30-day post-operative mortality in the entire 87,000 patient study population. With regard to end-organ dysfunction, we utilized an a priori definition that included post-operative mechanical ventilation for greater than 48 hours, post-operative renal insufficiency and post-operative coma. We could easily go back and alter this definition to include other terms for the data dictionary and refine the analysis to test for robustness of our conclusions related to end-organ dysfunction. Thank you.
* 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
919
