In-hospital cardiac arrest is a serious complication associated with high mortality, affecting more than 200,000 adults in the United States per year, with fewer than 20% surviving to hospital discharge.1 Among patients with cardiac arrest, surgical patients have characteristics distinguishing them from medical patients.2 Studies of cardiac arrest in surgical patients have tended to be older, single-center studies focusing on anesthetic management,3 but recently, there has been a broader understanding of the incidence and risk factors of cardiac arrest and mortality in the surgical population.2–5 Fortunately, the incidence of cardiac arrest in surgical patients is low, with estimates ranging from 7 per 10,000 noncardiac surgeries for intraoperative cardiac arrest5 to 6 per 10,000 anesthetics for arrests in the operating room or postanesthesia care unit (PACU).4 Unfortunately, the mortality associated with cardiac arrest is high, with reported rates of 63% for intraoperative cardiac arrests,5 and 58% for operating room/PACU cardiac arrests.4
While patient risk factors are important determinants of cardiac arrest and mortality, the development of postoperative complications may be a critical factor affecting outcomes after cardiac arrest, and other postoperative complications often precede cardiac arrest in surgical patients.3 In exploratory analyses, we identified potential interaction effects between cardiac arrest and other complications, indicating that they were not independent in their effects on mortality,6 but the specific relationships between complications and cardiac arrest have not been fully investigated. The prevention of postoperative complications represents an area where outcomes may be improved for those with cardiac arrest.7
Patients developing cardiac arrest are an extremely sick group of patients and closer examination could reveal characteristics that differentiate among those in this group. We evaluated general surgery patients experiencing cardiac arrest after postoperative day (POD) #0, up to POD #30, using the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP), hypothesizing that major postoperative complications occurring before cardiac arrest would be significantly associated with mortality. Our findings will enable further study to identify specific aspects in which the care of patients with cardiac arrest might be improved.
This study was not subject to review by the local institutional review board because it did not require access to protected health information. This is a retrospective, observational cohort study of general surgery in patients experiencing cardiac arrest within 30 days of surgery using data from the 2012–2013 ACS-NSQIP participant use files. The ACS-NSQIPa is a validated, prospectively collected national data set aimed at improving surgical quality and outcomes,8 with data collected from 374 sites in 2012 and 435 sites in 2013. Detailed descriptions of case inclusion criteria, systematic sampling process, quality measures, and variable descriptions are available from ACS-NSQIP.9 Patients are followed for 30 days after surgery, including postdischarge, by the surgical clinical reviewer at each participating site to monitor postoperative outcomes. This article adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational studies.10
Of the 1,195,825 initial records in the data set, 349,027 patients were classified as being inpatient and general surgery (Supplemental Digital Content 1, Figure 1, http://links.lww.com/AA/C5). Of these, 1730 patients experienced cardiac arrest, defined as “the absence of cardiac rhythm or presence of chaotic cardiac rhythm which results in a cardiac arrest requiring the initiation of CPR (cardiopulmonary resuscitation), which includes chest compressions. Patients are included who are in a pulseless ventricular tachycardia or ventricular fibrillation in which defibrillation is performed and pulseless electrical activity arrests requiring chest compressions. Patients with automatic implantable cardioverter defibrillator that fire but the patient has no loss of consciousness should be excluded.”
Because the data set does not allow for the determination of previous complications if cardiac arrest occurred on POD #0 (discussed more in detail below), the main analysis cohort consisted of the 1352 patients experiencing cardiac arrest after POD #0.
Demographic and Operative Variables
Procedures were classified using the Clinical Classifications Software for Services and Procedures (Agency for Health care Research and Quality, Rockville, MD) based on the primary Current Procedural Terminology (American Medical Association, Chicago, IL) code.
Body mass index (kg/m2) was calculated from height and weight data and categorized as: <25, ≥25 and <30, ≥30, or missing. The estimated glomerular filtration rate (mL/min/1.73 m2) was calculated using the Chronic Kidney Disease Epidemiology Collaboration formula incorporating creatinine, sex, age, and race,11 and categorized into groups corresponding to the stages of chronic kidney disease12: <30, ≥30 and <60, ≥60 and <90, ≥90, or missing. History of cancer was determined as described previously.13 Continuous variables were grouped into deciles and plotted against the log (hazard ratio [HR]) for mortality after cardiac arrest to visualize the relationships. Based on these analyses, age and operative time were entered as continuous variables, and hematocrit was categorized as ≤29, >29 and ≤35, >35 and ≤41, >41, or missing. Other variables were collected directly from the data set.
Exposures and Outcomes
For the primary analysis, the exposure variable was the development of 1 or more of the following 7 postoperative complications before the day of cardiac arrest, as defined by the data set (Supplemental Digital Content 2, Table 1, http://links.lww.com/AA/C6): (1) acute kidney injury; (2) acute respiratory failure; (3) bleeding transfusions; (4) deep vein thrombosis/pulmonary embolus; (5) myocardial infarction; (6) sepsis/septic shock; and (7) stroke. These complications were chosen based on previous work demonstrating their associations with postoperative mortality in general surgery patients.6 For planned secondary analyses, each individual complication was analyzed separately.
The primary outcome variable was mortality within 30 days of the surgical procedure. The data set identifies the POD in which complications, cardiac arrest, and death occur.
Differences in the proportions or means of preoperative characteristics and comorbidities between patients with and without previous complications were compared with the t test, the χ2 test, or Fisher exact test as appropriate. Differences in survival patterns between those with and without previous complications were analyzed using the Kaplan-Meier estimator14 and the log-rank test.15 Cox proportional hazard modeling was used to estimate the relative hazard for mortality after cardiac arrest, adjusting for confounders, comparing those with and without previous complications. Adjusted models included procedure class and variables associated with mortality in a previous study of postoperative cardiac arrest (age, chronic obstructive pulmonary disease [COPD], sepsis, cancer, and preoperative renal failure).3 In addition, we used stepwise multivariable Cox modeling to identify additional significant covariates. In the stepwise analysis, the previously mentioned variables were forced into the model while the following variables were subject to the stepwise criteria: female, white, body mass index, emergency, diabetic, mechanical ventilation, dyspnea, functionally dependent, current smoker, congestive heart failure, hypertension, ascites, steroid use, bleeding disorder, preoperative transfusion, estimated glomerular filtration rate, hematocrit, and log (operative time). A significance level of 0.1 was required to enter the model and a level of 0.2 was required to remain in the model.
Statistical analyses were performed using SAS Software version 9.4 (SAS Institute, Cary, NC) and GraphPad Prism 6.07 (GraphPad Software, Inc, La Jolla, CA). In the primary analysis of the associations between previous complications and mortality, statistical significance was determined with a P value <.05. Analysis of individual complications, 7 in total, were subjected to Bonferroni correction to adjust for multiple comparisons16 (P < .05/7 = 0.007).
A formal analysis plan was not filed with the institutional review board because this was an exempt study. The major post hoc change to the analysis involved the definition of previous complication. Initially, a previous complication was defined as one that occurred before the day of cardiac arrest or on the day of cardiac arrest. However, it became apparent that for complications occurring on the same day as cardiac arrest, we were unable to determine which event occurred first, so complications occurring on the same day as cardiac arrest were no longer classified as a previous complication. As a result of this change, patients with cardiac arrest on POD #0 were excluded because a complication could not occur before POD #0. We include post hoc analyses of mortality comparing patients with cardiac arrest on POD #0 versus after POD #0.
Because this is a retrospective, observational study, our sample size is limited by the available data. We estimated the sample size required to detect a relative hazard of 1.25 with α = .05 and power = .90, assuming equal sample sizes in each group (previous complication versus no previous complication). Based on previous literature, we assumed a cumulative mortality of 70% in the no previous complication group.3 Other assumptions include a baseline event rate of 0.24/d in the no previous complication group, a censoring rate of 0.10/d, and an average follow-up of 30 days.17,18 Based on these parameters, the required sample size is 1159, so the actual sample size in the study (N = 1352) is adequate to detect a clinically meaningful difference in mortality between those with and without complications before cardiac arrest.
Baseline Characteristics of Patients Developing Cardiac Arrest
Of the 1352 patients with cardiac arrest after POD #0, there were 746 (55%) patients who developed at least 1 complication before cardiac arrest (Table 1). Compared to those without previous complications, those with previous complications were more likely to be American Society of Anesthesiologists physical status IV–V, mechanically ventilated, have an emergent procedure, COPD, ascites, steroid use, preoperative sepsis/septic shock, preoperative transfusions, and a lower preoperative hematocrit. There were no significant differences in factors such as age, sex, and cancer diagnoses.
There were 49 separate Clinical Classifications Software for Services and Procedures categories represented among patients with cardiac arrest. The category with the greatest number of patients was colorectal resection (36%), followed by small bowel resection (7.6%), and other hernia repair (7.1%) (Supplemental Digital Content 3, Table 2, http://links.lww.com/AA/C7).
Previous Complications and Mortality in Patients With Cardiac Arrest
There was no difference in 30-day mortality risk between cardiac arrest patients with previous complications and those without previous complications (71% [533/746] vs 70% [425/606]; P = .60 χ2 test) (Table 2). Kaplan-Meier plots of postcardiac arrest survival also demonstrated no significant differences in survival patterns between those with and without previous complications (Figure, A).
Among individual complications, transfusion (33% [445/1352]) was the most common, followed by acute respiratory failure (25% [344/1352]), and sepsis/septic shock (14% [189/1352]) (Table 2). In contingency tables, mortality was lower among those with previous acute respiratory failure than in those without previous acute respiratory failure (66% [228/344] vs 72% [730/1008]; P = .03 χ2 test). However, this result did not meet Bonferroni-adjusted criteria for statistical significance and no other individual complication was significantly associated with mortality. Kaplan-Meier plots demonstrated differences in postcardiac arrest survival patterns between those with and without previous myocardial infarction (Figure, B; P = .02 log-rank test), but this did not meet Bonferroni-adjusted criteria for significance and no other individual previous complication was significantly associated with mortality.
Cox Models for Postcardiac Arrest Mortality by Previous Complications
Compared to those without previous complications, those with a previous complication did not have significant associations in the hazard for mortality in unadjusted (HR, 1.09; 95% confidence interval [CI], 0.96–1.23; P = .21) or adjusted models (HR, 1.03; 95% CI, 0.90–1.18; P = .70) (Table 3). Confounding variables were identified from previous studies3 (COPD, preoperative sepsis/septic shock, cancer, preoperative acute renal failure/dialysis, and operative procedure) and stepwise regression (preoperative ascites and estimated glomerular filtration rate).
Among individual complications, compared to those without previous myocardial infarction, those with a previous myocardial infarction had an increased hazard for mortality in unadjusted (HR, 1.49; 99.3% CI, 0.98–2.26; P = .01) and adjusted (HR, 1.40; 99.3% CI, 0.92–2.14; P = .03) analyses, though neither result met the Bonferroni-adjusted criteria for significance. No other individual previous complication was significantly associated with mortality in unadjusted or adjusted Cox models.
Post Hoc Analyses: POD of Cardiac Arrest and Survival After Cardiac Arrest
While our main analyses excluded patients with cardiac arrest on POD #0, we included these patients in post hoc analyses. Among all patients with cardiac arrest (N = 1730; Supplemental Digital Content 1, Figure 1, http://links.lww.com/AA/C5), POD #0 was the POD with the greatest number of cardiac arrest events (Supplemental Digital Content 4, Figure 2, http://links.lww.com/AA/C8), with 378/1730 (22%) occurring on POD #0, including 28 patients with intraoperative cardiac arrest. The number of cardiac arrests declined in the subsequent days after the surgical procedure, with half of all cardiac arrest events occurring by POD #4. There were no significant differences in postarrest survival among those experiencing cardiac arrest on POD #1, POD #2, or after POD #2 (Supplemental Digital Content 5, Figure 3A, http://links.lww.com/AA/C9). When comparing postcardiac arrest survival between those with cardiac arrest on POD #0 versus after POD #0, there were significant differences in postcardiac arrest survival patterns between these 2 groups (Supplemental Digital Content 5, Figure 3B, http://links.lww.com/AA/C9).
Using a retrospective, multicenter, high-quality surgical data set, we found that in general surgery patients with cardiac arrest after POD #0, previous complications were identified in 55%, with blood transfusions and acute respiratory failure being the most common. Cardiac arrest was associated with very high mortality, with 71% of patients dying within 30 days of surgery. Overall, the development of a previous complication was not associated with an increased risk of mortality after cardiac arrest, and analyses of individual complications also did not demonstrate any previous complications that were associated with increased mortality risk. These findings suggest that knowledge of the complications that developed before postoperative cardiac arrest does not help to differentiate a patient’s risk of subsequent mortality.
A recent study by Kazaure et al,19 also using the ACS-NSQIP, found that postoperative complications commonly preceded cardiac arrest in surgical patients, but as complications were evaluated as outcomes and not risk factors, conclusions about the specific contribution of complications to mortality risk could not be drawn, though there were thoughts that aggressive management and prevention of complications might reduce the incidence of postoperative cardiac arrest.7 While our study cannot address whether aggressive management does indeed reduce the incidence of cardiac arrest, it does suggest that previous complications do not affect mortality once cardiac arrest occurs. In addition, Kazaure et al19 examined the entire ACS-NSQIP database, including all surgical specialties, and found baseline differences in the underlying patient populations. We chose to study only general surgery patients to reduce heterogeneity introduced by examining multiple surgical specialties, but note that heterogeneity also exists within general surgery patients with regards to surgical outcomes13 that necessitates accounting for the specific surgical procedure even within this category.
Much of the literature on perioperative cardiac arrest focuses on the intraoperative or immediate postoperative period up to 24 hours after the procedure,2,4,5,20 and the epidemiology of cardiac arrest events after this period is limited.3 Although it was not our original intent, by excluding patients with cardiac arrest on POD #0, our analysis focused on patients with cardiac arrest after POD #0 and factors affecting mortality in this group of patients have not been well studied. It may be thought that cardiac arrest during this time is not directly related to surgery or anesthesia and it remains to be determined if cardiac arrest and mortality are the inevitable result of the poor postoperative course of these patients or if interventions could have been performed to reduce the rate of complications and cardiac arrest events. The survival in this group is comparable to survival rates among all patients with in-hospital cardiac arrest,21 so they may be similar to those in the general medical population experiencing cardiac arrest.
Interestingly, we found that survival patterns after cardiac arrest were not affected by the POD on which cardiac arrest occurred in patients with cardiac arrest after POD #0. In contrast, post hoc analyses demonstrated better survival in patients with cardiac arrest on POD #0 compared to after POD #0, with both higher day of arrest survival (66% vs 49%) and 30-day survival (47% vs 29%). In other studies, of cardiac arrests occurring within 24 hours of surgery, those that occurred in the operating room or in the PACU were associated with better survival compared to other locations2 and survival has improved over time in patients with cardiac arrest during general anesthesia,22 with less than a quarter of events attributable to anesthetic management.23 We cannot make definitive conclusions regarding cardiac arrests occurring on POD #0 as they were not the primary focus of the analysis, but based on these findings, it appears that there are differences between cardiac arrests occurring on POD #0 versus after POD #0, and a more detailed understanding of the factors related to cardiac arrests occurring at different time points in the postoperative course is warranted.
We did not identify any individual complication that was associated with mortality when occurring before cardiac arrest, but myocardial infarction merits further discussion as the results hint at an association with mortality. Though it did not meet the statistical significance criteria when adjusted for multiple comparisons, the sample was not powered to evaluate myocardial infarction as an individual complication. Myocardial infarction occurs in 5% of patients undergoing noncardiac surgery, with most events diagnosed within 48 hours after the procedure.24 While speculative in nature at this time, we observed a temporal relationship between myocardial infarction and cardiac arrest affecting the risk of mortality as a sensitivity analysis found that myocardial infarctions occurring on the same day as or after the day of cardiac arrest were not associated with increased mortality risks (data not shown), but this finding will have to be validated in future studies.
It is likely that clinically significant perioperative myocardial ischemia is underdiagnosed as a recent study demonstrated that almost 60% of myocardial ischemic events did not meet the criteria for myocardial infarction but were nonetheless associated with increased mortality risk.25 Routine and early monitoring of postoperative troponin levels in high-risk patients may be necessary to identify patients with myocardial ischemia that may not be symptomatic.25,26 We cannot identify patients with undiagnosed myocardial infarction or myocardial injury in our sample and it is unclear what impact these conditions would have had on our analyses. Thus, the associations between previous myocardial ischemia and mortality in cardiac arrest patients need to be evaluated in a larger study and the exposure should be expanded to include patients with myocardial injury and myocardial infarction.
Limitations of our study include its retrospective nature that precludes determination of causality. Though we identified complications that temporally preceded cardiac arrest, this does not necessarily mean that the complication itself caused any changes in mortality risk. The data set does not contain some relevant clinical variables such as the cause of the cardiac arrest event, data on intraoperative management, and the location of the patient when the event occurred (PACU, intensive care unit, floor, etc). More granular data, such as coding hours, instead of days, after surgery, would have been beneficial, especially in the first 24 hours after surgery. Other factors that have been shown to affect surgical outcomes, such as hospital variation,27 Do Not Resuscitate orders,28 rapid response teams,29 and the day of week of surgery (weekday versus weekend),30 were not able to be addressed with this data set.
Observational studies are subject to certain limitations.31 Misclassification bias may exist if the reporting of previous complications was inaccurate due to issues with follow-up or charting. For example, if the time period between the development of a complication and cardiac arrest was short, the complication might not have been recognized. Our a priori hypothesis was that previous complications would be associated with increased mortality risk after cardiac arrest, but the data did not support this finding. It is plausible to think that every patient with cardiac arrest had some complication even if it was not explicitly documented, and there may be important complications affecting cardiac arrest and mortality that were not collected in the data set. There may be a lack of generalizability as the ACS-NSQIP has tended to include mainly large, academic hospitals so the results may not apply in smaller nonacademic settings.32 Residual confounding is less of a concern as our study did not identify any significant associations between the exposures and outcome.
In conclusion, we found that among general surgery patients with cardiac arrest after POD #0, mortality was extremely high and many patients experienced major postoperative complications before cardiac arrest, but these complications did not appear to increase the risk of mortality. While further study is necessary to identify measures to reduce the risk of complications and mortality after cardiac arrest, the development of a major postoperative complication before cardiac arrest does not appear to be a distinguishing factor between those likely to have better or worse outcomes after cardiac arrest.
Name: Minjae Kim, MD, MS.
Contribution: This author helped design and conduct the study, analyze the data, and prepare the article.
Name: Guohua Li, MD, DrPH.
Contribution: This author helped design the study, analyze the data, and prepare the article.
This manuscript was handled by: Richard C. Prielipp, MD.
1. Kolte D, Khera S, Aronow WS, et al. Regional variation in the incidence and outcomes of in-hospital cardiac arrest in the United States. Circulation. 2015;131:1415–1425.
2. Ramachandran SK, Mhyre J, Kheterpal S, et al.; American Heart Association’s Get With The Guidelines-Resuscitation Investigators. Predictors of survival from perioperative cardiopulmonary arrests: a retrospective analysis of 2,524 events from the Get With The Guidelines-Resuscitation registry. Anesthesiology. 2013;119:1322–1339.
3. Kazaure HS, Roman SA, Rosenthal RA, Sosa JA. Cardiac arrest among surgical patients: an analysis of incidence, patient characteristics, and outcomes in ACS-NSQIP. JAMA Surg. 2013;148:14–21.
4. Nunnally ME, O’Connor MF, Kordylewski H, Westlake B, Dutton RP. The incidence and risk factors for perioperative cardiac arrest observed in the national anesthesia clinical outcomes registry. Anesth Analg. 2015;120:364–370.
5. Goswami S, Brady JE, Jordan DA, Li G. Intraoperative cardiac arrests in adults undergoing noncardiac surgery: incidence, risk factors, and survival outcome. Anesthesiology. 2012;117:1018–1026.
6. Kim M, Li G. Two-way Interaction Effects of Perioperative Complications on 30-Day Mortality in General Surgery. World J Surg. 2017 July 28 [Epub ahead of print].
7. Zenilman ME. Cardiopulmonary resuscitation in surgical patients: it is all in the timing. JAMA Surg. 2013;148:21–22.
8. Fink AS, Campbell DA Jr, Mentzer RM Jr, et al. The National Surgical Quality Improvement Program in non-veterans administration hospitals: initial demonstration of feasibility. Ann Surg. 2002;236:344–353.
11. Levey AS, Stevens LA, Schmid CH, et al.; CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration). A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150:604–612.
12. National Kidney Foundation. K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Am J Kidney Dis. 2002;39:S1–S266.
13. Kim M, Brady JE, Li G. Variations in the risk of acute kidney injury across intraabdominal surgery procedures. Anesth Analg. 2014;119:1121–1132.
14. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457–481.
15. Peto R, Peto J. Asymptotically efficient rank invariant test procedures. J R Statist Soc A. 1972;135:185–207.
16. Armstrong RA. When to use the Bonferroni correction. Ophthalmic Physiol Opt. 2014;34:502–508.
17. Schoenfeld DA. Sample-size formula for the proportional-hazards regression model. Biometrics. 1983;39:499–503.
19. Kazaure HS, Roman SA, Sosa JA. Epidemiology and outcomes of in-hospital cardiopulmonary resuscitation in the United States, 2000-2009. Resuscitation. 2013;84:1255–1260.
20. Braz LG, Módolo NS, do Nascimento P Jr, et al. Perioperative cardiac arrest: a study of 53,718 anaesthetics over 9 yr from a Brazilian teaching hospital. Br J Anaesth. 2006;96:569–575.
21. Chan PS, Krein SL, Tang F, et al.; American Heart Association’s Get With the Guidelines–Resuscitation Investigators. Resuscitation practices associated with survival after in-hospital cardiac arrest: a nationwide survey. JAMA Cardiol. 2016;1:189–197.
22. Sprung J, Warner ME, Contreras MG, et al. Predictors of survival following cardiac arrest in patients undergoing noncardiac surgery: a study of 518,294 patients at a tertiary referral center. Anesthesiology. 2003;99:259–269.
23. Ellis SJ, Newland MC, Simonson JA, et al. Anesthesia-related cardiac arrest. Anesthesiology. 2014;120:829–838.
24. Devereaux PJ, Xavier D, Pogue J, et al.; POISE (PeriOperative ISchemic Evaluation) Investigators. Characteristics and short-term prognosis of perioperative myocardial infarction in patients undergoing noncardiac surgery: a cohort study. Ann Intern Med. 2011;154:523–528.
25. Botto F, Alonso-Coello P, Chan MT, et al.; Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Writing Group, on behalf of The Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Investigators; Appendix 1. The Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Study Investigators Writing Group; Appendix 2. The Vascular events In noncardiac Surgery patIents cOhort evaluatioN Operations Committee; Vascular events In noncardiac Surgery patIents cOhort evaluatioN VISION Study Investigators. Myocardial injury after noncardiac surgery: a large, international, prospective cohort study establishing diagnostic criteria, characteristics, predictors, and 30-day outcomes. Anesthesiology. 2014;120:564–578.
26. van Waes JA, Nathoe HM, de Graaff JC, et al.; Cardiac Health After Surgery (CHASE) Investigators. Myocardial injury after noncardiac surgery and its association with short-term mortality. Circulation. 2013;127:2264–2271.
27. Ghaferi AA, Birkmeyer JD, Dimick JB. Variation in hospital mortality associated with inpatient surgery. N Engl J Med. 2009;361:1368–1375.
28. Morrison LJ, Neumar RW, Zimmerman JL, et al.; American Heart Association Emergency Cardiovascular Care Committee, Council on Cardiopulmonary, Critical Care, Perioperative and Resuscitation, Council on Cardiovascular and Stroke Nursing, Council on Clinical Cardiology, and Council on P. Strategies for improving survival after in-hospital cardiac arrest in the United States: 2013 consensus recommendations: a consensus statement from the American Heart Association. Circulation. 2013;127:1538–1563.
29. Offner PJ, Heit J, Roberts R. Implementation of a rapid response team decreases cardiac arrest outside of the intensive care unit. J Trauma. 2007;62:1223–7
30. Glance LG, Osler T, Li Y, et al. Outcomes are worse in US patients undergoing surgery on weekends compared with weekdays. Med Care. 2016;54:608–615.
31. Hill HA, Kleinbaum DG. Bias in observational studies. Encyclopedia of Biostatistics. 2005;1.
32. Dimick JB, Osborne NH, Hall BL, Ko CY, Birkmeyer JD. Risk adjustment for comparing hospital quality with surgery: how many variables are needed? J Am Coll Surg. 2010;210:503–508.
Supplemental Digital Content
Copyright © 2017 International Anesthesia Research Society