Do-not-resuscitate (DNR) orders instruct medical personnel to forego cardiopulmonary resuscitation (CPR)—including chest compressions, artificial respirations, cardiac defibrillations, and intravenous medications—in the event of cardiopulmonary arrest.1,2 In general, DNR status serves as a surrogate marker for profound illness,3–5 advanced age,6,7 and impending death.3,4,8,9
Importantly, a DNR order does not preclude medical care. Patients with a DNR order may consent to a myriad of surgical procedures, from palliative procedures (ie, lysis of adhesions to relieve a bowel obstruction) to potentially life-extending procedures. In one survey of DNR patients, 83.3% would agree to palliative procedures or procedures unrelated to the primary diagnosis.10 Previous studies have suggested that at least 15% of patients with confirmed DNR status undergo surgery.11,12 However, there exists limited information on the types of procedures performed on DNR patients. Unfortunately, there is also a lack of rigorous evidence regarding the outcomes of palliative care interventions in surgical patients.13 This pleads the question: while the “do-not-resuscitate” is not equivalent to “do-not-treat,” when is the decision to pursue surgery appropriate for a DNR patient?
Enforcement of DNR orders during the perioperative period is another ongoing source of controversy among both anesthesiologists and surgeons, because of the inherent resuscitative measures such as intubation, employed in even routine surgery.14–19 Both the American Society of Anesthesiologists (2001, revised 2008 and 2013) and the American College of Surgeons (1994, revised 2014) now advocate for “required reconsideration” of existing DNR orders, wherein patients or their designated proxy determine the extent of resuscitative measures to be employed during surgery and the immediate postoperative period.20,21 This practice ties into a broader phenomenon of patient-centered shared decision-making in the perioperative period, which allows providers and patients to come to a common decision regarding appropriateness of care and patient preferences for degree of intervention.46 However, although these conversations are increasingly enforced as standard of care, the effect on choice of therapy in the perioperative period has not yet been characterized.
Surgical outcomes for patients with a DNR status have largely been assessed through retrospective observational studies querying institutional or national databases across varied patient populations. These studies have been limited by focus on a narrow subset of surgical procedures, small cohorts, and older data. Nevertheless, DNR status has been found to be an independent predictor of mortality in trauma, emergent bowel obstruction, vascular, and cardiothoracic surgery,22–28 although increased risk of postoperative complications has been less consistent.27–29 Mortality and morbidity could differ in DNR patients depending on the clinical management or even surgery urgency.28a Specifically, there is significant concern for a “failure-to-rescue” phenomenon, in which patients with a DNR status receive less aggressive treatment overall despite the fact that their wishes only forbid strict resuscitation. This concern stems from a number of studies suggesting that DNR patients may receive lower-quality or less aggressive care,29–34 although this finding is not ubiquitous.35 An alternate explanation is that increased mortality reflects patient or designated proxy choice to limit excessive care.6 Ultimately, no studies have succeeded in differentiating these potential causes of mortality, nor have any been able to draw conclusions about postoperative events aside from mortality.
To deliver optimal perioperative care for this population, we must understand what current practice entails as well as the subsequent outcomes, including adherence to goals of care. This study aimed to characterize preoperative characteristics and patterns of care in surgical patients with a DNR status, as well as 30-day postoperative outcomes and predictors of postoperative mortality. In addition, we assessed for the presence of a “failure to rescue” phenomenon, wherein DNR patients would suffer excessive postoperative complications relative to baseline status because of less aggressive treatment. We hypothesized that DNR patients would demonstrate increased postoperative mortality with no concomitant increase in morbidity relative to non-DNR patients undergoing the same procedures, arguing against the “failure to rescue” phenomenon. Furthermore, we hypothesized that DNR patients would demonstrate decreased incidence of cardiopulmonary resuscitative interventions. Together, this would suggest that physicians have honored the DNR status by providing all necessary therapies except for resuscitative measures. The results will crucially inform both patients and physicians in the shared decision-making process and further aid in determining the appropriateness of surgical intervention.
Data were collected by the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database for 2,820,370 cases from the Participant Use Data File for 2007 to 2013. ACS-NSQIP is a multi-institutional program representing more than 400 academic and private hospitals across the United States, wherein participating hospitals voluntarily report surgical quality outcome measures with the aim of improving quality of care on both an institutional and national level. The ACS routinely audits contributing institutions and excludes data from sites where the 30-day follow-up rate is <80% or the interrater reliability disagreement rate exceeds 5%.
Certified surgical clinical reviewers prospectively collect 30-day postoperative data for randomly assigned patients (to ensure accurate sampling), in accordance to strict protocol. The database is deidentified and meets the criteria of the Health and Insurance Portability and Accountability Act for the protection of personal information. Institutional Review Board (Brigham and Women’s Hospital, Boston, MA) approval was obtained for analysis of the data and was exempted from the consent requirement because of the deidentified nature of the data. The database contains 306 Health and Insurance Portability and Accountability Act compliant variables, including preoperative risk factors, intraoperative variables, and 30-day postoperative mortality and morbidity outcomes for patients undergoing major surgical procedures in both the inpatient and outpatient setting.
Of note, patients below age 18 are not included in the database and all patients greater than 90 years old are recorded as “90+.” Only major surgeries are included, which are defined as (1) under general, spinal, or epidural anesthesia; or (2) regardless of anesthetic choice, any carotid endarterectomy, inguinal herniorrhaphy, parathyroidectomy, thyroidectomy, breast lumpectomy, and endovascular abdominal aortic aneurysm repair. Trauma, transplant, and concurrent cases are excluded.
This article adheres to the applicable EQUATOR guidelines for analysis and presentation of retrospective data.
Using 2007 to 2013 data from the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP), we performed a retrospective analysis on patients with a DNR status undergoing the most frequent procedures performed in this population, including elective and emergent procedures.
ACS NSQIP defines DNR status as follows: If the patient has had a Do-Not-Resuscitate (DNR) order written in the physician’s order sheet of the patient’s chart and it has been signed or cosigned by an attending physician. There must be an active DNR order at the time the patient is going to the operating room. However, if the DNR order, as defined above, was rescinded immediately before surgery to operate on the patient, enter “YES.”
Using this variable, we identified 9854 DNR patients who underwent surgery as recorded in ACS-NSQIP from 2007 to 2013 (Figure). We then identified the principal operative procedure current procedural terminology (CPT) code for each case. To assess the most representative procedures performed on patients with a DNR status, we decided to investigate only those procedures with ≥50 DNR patients represented. This was corroborated by review of the CPT codes, which showed predominantly emergent surgeries not related to the primary illness that were expected to relieve symptoms rather than extend life or reduce risk. Other, often more extreme, procedures may be rarely performed but do not represent the usual management for DNR patients. This resulted in a cohort of 5629 patients with a confirmed DNR status. These patients were then matched to 443,000 patients with non-DNR status having the same operations based on the principal operative CPT code. All patients undergoing the included surgical procedures were included in the matching process.
Patient demographic data were collected for age, age group (<50, 50–64, 65–79, >80 years old), ASA physical status (PS) classification, sex, race, Hispanic ethnicity, transfer status, height, weight, body mass index, functional status before surgery, and estimated probability of mortality/morbidity. Preoperative comorbidity data were collected for diabetes mellitus (insulin and noninsulin dependent); smoking status within 1 year before admission; pack-years of smoking; current alcohol use (defined as >2 drinks/d before admission); long-term steroid use; weight loss >10% in the past 6 months; chemotherapy for malignancy <30 days before surgery; and radiotherapy for malignancy in the past 90 days. Several additional pulmonary, cardiovascular, hepatobiliary, renal, neurologic, hematological, and infectious comorbidities were collected as defined by the ACS-NSQIP 2013 user guide. Operative data were collected for emergent versus nonemergent procedures; elective versus nonelective procedures; previous operation within 30 days; highest level of resident surgeon present (by postgraduate year); operative wound classification; significant intraoperative event (transfusion, cardiopulmonary resuscitation, myocardial infarction); case duration; surgical subspecialty classification (cardiac, general, gynecological, neurosurgery, orthopedics, otolaryngology [ENT], plastics, thoracic, urology, vascular, and other/unspecified); and work relative value unit (0–19, 20–39, 40–59, 60–79, >80). Although it is true that surgical residents do not practice independently, we decided to include resident level as a variable of interest regarding patient outcomes because intraoperative resident participation has been associated with slightly higher morbidity, but slightly lower mortality rates across a breadth of procedures.36,37 Facilities and geographic regions have been deidentified, and thus we could not acquire those data.
ACS-NSQIP provides outcomes data for postoperative morbidity and mortality up to 30 days after surgery. The database provides standard definitions for outcomes of interest that have been widely adapted by clinical practices and are often binary (present or no complication). However, definitions of specific outcomes may be inconsistent between sites. For our study, clinical outcomes data were collected for occurrence of 1 or more postoperative complications; destination on discharge (home, skilled care, or rehab); total length of stay (LOS); reoperation; and death within 30 days of surgery. For the 30-day mortality, we used the variable “YRDEATH,” which defines the year in which the patient died. However, all surgical patients within ACS-NSQIP are only followed for 30 days. If the patient is alive at 30 days, the variable is marked as “−99.” We thus defined patient mortality at 30 days as the sum of all cases where the value of the “YRDEATH” variable was not equal to “−99.” We did not consider death as a competing risk event in our analysis given the nonuniform reporting of time-to-event for both postoperative mortality and morbidity in ACS-NSQIP, including aberrant data extending beyond the 30-day reporting limit.
We reported all major postoperative complications included in the ACS NSQIP Participant Use Data File, including superficial and deep wound infections, organ/space infections, wound dehiscence, pneumonia, unplanned intubation, failure to wean after 48 hours, progressive renal insufficiency, acute renal failure, urinary tract infection, stroke, cardiac arrest requiring CPR, myocardial infarction, transfusion requirement, deep venous thrombosis requiring therapy, pulmonary embolism, sepsis, and septic shock. All outcomes were reported as percentages, with the numerator defined as the absolute count reporting a given outcome, and the denominator defined as the total number of cases reporting any outcome for that variable.
R Project for Statistical Computing (R Studio, version 3.1.2; R Foundation for Statistical Computing, Vienna, Austria) was used to perform all statistical analysis. For all demographic, comorbidity, and operative characteristics, a univariable logistic regression model was fitted to assess the association of each variable with DNR status. Of note, the database does not report any postoperative outcomes (including death) occurring more than 30 days after surgery. For our primary analysis, we applied a multivariable logistic regression model accounting for preoperative variance to better assess the association between DNR status and postoperative outcomes. Variables in the model included all preoperative and operative characteristics univariably associated with DNR status with statistical significance, defined as P < .05. A reference group was chosen for each characteristic and is noted in the table. For the logistic regression, odds ratios (ORs) were reported with associated 95% confidence intervals (CIs). OR not including 1.00 in the 95% CI were considered statistically significant.
To assess the association specifically between LOS and DNR status, a Cox proportional hazard model was fitted, incorporating the demographic and comorbidity covariables as described above. The model was right censored with death as a competing event.
Our initial sample size of 448,629 was developed using a similar methodology to Kazaure et al41 based on principal operative CPT codes. No formal power calculation was performed after the set of included CPT codes was generated. Power calculations were performed using G*Power (version 3.1; Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany).38 We performed calculations using a z test with an α of .002381 (obtained by utilizing a Bonferroni adjustment on an α of .05), power of 0.9, sample size of 9854 for the DNR group and 438,775 for the control group, and accounting for all predictors in the multivariable model. We found that had 90% power to detect an absolute difference in proportions of 0.031 = 3.1% and a relative difference of 25.4%, using a baseline risk of 0.024.
The significance level for each hypothesis was .05. Adjustment for multiple testing of each outcome was made using Bonferroni correction. The significance levels for the analyses in Tables 4 and 5 were thus .0024 and .0019, respectively.
The most common procedures performed on patients with a DNR status are listed in Table 1. A cohort containing all surgical procedures performed >50 times on patients with a DNR status was then used to create a control cohort, resulting in a control cohort of 443,555 non-DNR patients (Figure).
Table 2 displays the patient-related demographics and comorbidities for both DNR and non-DNR patients undergoing these procedures, which allowed us to assess potential confounding variables to be included in our multivariable analysis. Patients with a DNR status tend to be older (mean age of 79.1 ± 12.1 years vs 56.0 ± 18.9 years) and are more likely to be female (62.8% vs 50.7%). DNR patients have a globally increased comorbidity burden compared with non-DNR patients, with increased incidence in preoperative pneumonia (4.9% vs 0.7%), congestive heart failure (8.2% vs 1.2%), prior myocardial infarction (4.3% vs 0.9%), prior wound infection (22.6% vs 5.0%), and preoperative transfusion (5.1% vs 1.0%). This difference in comorbidity burden was reflected in the ASA PS classification, which was most often 3 or 4 for DNR patients, compared with 2 or 3 for non-DNR patients.
Table 3 displays the patient-related surgical characteristics for both DNR and non-DNR patients undergoing these procedures. As above, this analysis allowed us to assess potential confounding variables to be included in our multivariable analysis. DNR patients were less likely to undergo elective procedures (14.5% vs 58.3%) and more likely to undergo emergent procedures (38.3% vs 25.1%) compared with non-DNR patients. DNR patients underwent orthopedic (23.0% vs 5.0%) and vascular (21.7% vs 14.5%) surgeries significantly more often than non-DNR patients. Mean anesthesia time was minimally longer (145.5 ± 74.1 minutes for DNR patients versus 140.9 ± 81.1 minutes). Mean operative time was minimally shorter, with 87.2 ± 60.0 minutes for DNR patients compared with 91.4 ± 67.6 minutes for non-DNR patients.
Table 4 demonstrates the 30-day postoperative outcomes for DNR compared with non-DNR patients before risk adjustment. DNR patients had significantly greater rates of all postoperative morbidities except for organ/space (nonincisional) surgical site infections. The most common adverse event was death, occurring in 24.4% of DNR patients compared with 2.4% of non-DNR patients (OR, 13.12 [12.31–13.98]; P < .001), followed by transfusion (12.6% vs 3.1%, OR 4.50 [4.15–4.88]; P < .001). Mean LOS was increased in the DNR cohort at 13.1 ± 14.5 vs 6.7 ± 11.6 days, corresponding to a hazard ratio of 3.72 (P < .001).
Because of the overrepresentation of female patients in the DNR cohort, Supplemental Digital Content, Table 1, http://links.lww.com/AA/B659, demonstrates the 30-day postoperative outcomes for DNR patients by sex. There was no significant difference in mortality (OR, 1.2 [1.06–1.36]; P = .01), but male patients were more likely to suffer pneumonia (OR, 1.71 [1.38–2.11]; P < .001) and septic shock (OR 1.35 [1.06–1.71], P < .001) but less likely to suffer urinary tract infections (OR, 0.65 [0.51–0.83]; P < .001) and undergo transfusions intraoperatively/postoperatively (OR, 0.75 [0.63–0.89]; P < .001).
Table 5 demonstrates the risk-adjusted 30-day postoperative outcomes for procedure-matched DNR compared with non-DNR patients, accounting for all significant demographic, comorbidity, and surgical characteristics. Each of these results reflects a separate multivariable model. DNR patients had significantly increased odds of death (OR, 2.54 [2.29–2.82]; P < .001) and discharge to skilled care (OR, 5.44 [4.21–7.03]; P < .001) or rehabilitation (OR, 2.55 [1.81–3.59]; P < .001). In addition, DNR patients had an increased total LOS with a hazard ratio of 2.16 (P < .001). Patients with a DNR status also had significantly reduced rates of unplanned intubation (OR, 0.51 [0.41–0.62]; P < .001), failure to wean (OR, 0.66 [0.56–0.76]; P < .001), and cardiac arrest requiring CPR (OR, 0.33 [0.22–0.50]; P < .001). After multiple comparisons, patients with a DNR status did not have a significantly different rate of transfusion intraoperatively/postoperatively (OR, 1.20 [1.06–1.35]; P < .004).
This study characterized patient demographics, patterns of surgical care, and postoperative outcomes including resuscitative interventions in a contemporary cohort of DNR patients. In total, 5629 cases performed on DNR patients and 443,555 cases performed on procedure-matched non-DNR patients were collected for analysis. When adjusting for preoperative factors, DNR patients were still found to have increased incidence of postoperative mortality, but not postoperative morbidity at 30 days. In addition, cardiopulmonary resuscitative measures occurred less frequently in the DNR cohort.
We report that the most common procedures performed on DNR patients fell into the main categories of intra-abdominal/bowel (51.6%), lower extremity orthopedic (22.3%), lower extremity amputation (16.0%), extremity vascular (6.3%), and major vascular (3.8%). In review of the precise CPT codes, the list represents predominantly emergent procedures such as extremity amputation, relief of small-bowel obstruction, femoral and hip fracture repair, cholecystectomy, appendectomy, and thrombectomy. Importantly, the most common procedures performed on DNR patients did not feature life-extending (ie, gastrostomy tube or dialysis catheter placement) or risk reduction procedures. This is consistent with goals of care centered on immediate symptom relief, rather than attempting to prolong life with aggressive surgical management of a primary illness such as metastatic cancer. Of note, significantly fewer DNR patients underwent procedures that were considered elective compared with the non-DNR group (14.5% vs 58.3%; P < .001). This may reflect adherence to goals of care, wherein unnecessary interventions (including elective surgery) are limited in end-of-life care. However, this variable is difficult to interpret because ACS-NSQIP labels all emergent/urgent surgeries as “nonelective,” in addition to all surgeries performed on patients arriving from an inpatient setting or emergency ward.
As expected, patients with a DNR status were older and more likely to have comorbidities and functional impairment compared with procedure-matched non-DNR patients. In our study, we further report that postoperative mortality was independently associated with several preoperative characteristics, including ASA PS classification III–V, nonindependent functional status, orthopedic surgery, and age >65 years. Ultimately, this implies a “poor substrate” effect, wherein a DNR patient with poorer health at baseline will be more likely to die in the 30-day postoperative period.
Interestingly, female patients accounted for 62.8% of the DNR surgical patients in our study. This overrepresentation of female patients in the DNR cohort is consistent with the literature on the epidemiology of DNR orders.5,39,40 In addition, it is known that women outnumber men in the geriatric surgical population, which may also contribute to the preponderance of women with DNR orders.41 Women are also more likely to be prone to chronic diseases wherein DNR orders may be appropriate, whereas men are more likely to die suddenly of causes such as myocardial infarction, homicide, and suicide.
We then quantified the incidence of postoperative mortality and morbidity in our patient cohort. As previously mentioned, the association of DNR status to postoperative mortality and morbidity has been well characterized in the literature.22–28 Kazaure et al42 determined that DNR status was associated with increased LOS, higher complication rate, and mortality across a variety of surgical procedures from 2005 to 2008, and that DNR status remained an independent predictor of mortality after risk adjustment (OR 2.2). Similarly, we report a global increase in the incidence of postoperative complications in DNR patients compared with non-DNR patients using univariable regression analysis. After risk adjustment, we found that DNR status remained significantly associated with mortality (OR, 2.54 [2.29–2.82]; P < .001), but the incidence of most other postoperative adverse events was not significantly different from non-DNR patients. Although DNR patients were more likely to have an increased LOS and require discharge to nonhome locations even after risk adjustment, we hypothesize that this could reflect preoperative variables signifying baseline illness that we were not able to account for due to the limited preoperative variables available within ACS-NSCIP.
Together, these findings suggest that mortality in DNR patients is not due to a “failure to rescue” phenomenon. If this were the case, we would expect to see an excess of postoperative complication burden relative to baseline status; however, the increased risks of postoperative morbidity on univariable regression disappear with multivariable regression accounting for all relevant preoperative comorbidities. Instead, we propose that increased incidence of mortality in this cohort reflects proper adherence to a DNR order in the setting of critical illness. In accordance with this theory, DNR patients had reduced rates of both unplanned reintubation (OR, 0.51 [0.41–0.62]; P < .001) and cardiopulmonary resuscitation (OR, 0.33 [0.22–0.50]; P < .001) as well as no difference in rate of transfusion (OR, 1.2 [1.06–1.35]; P < .004) in the risk-adjusted analysis. While ACS-NSQIP does not report other treatment variables, this suggests that cardiopulmonary resuscitative measures were held accordingly but other treatment strategies remained in place. In other words, “do-not-resuscitate” was not managed as “do-not-treat,” and patient wishes against life-sustaining interventions were honored in end-of-life care. As for the DNR patients who did receive resuscitative measures, it is possible these patients may have had altered or revoked DNR orders in the intraoperative or postoperative period. Unfortunately, the exact cause and circumstances of death in each patient are not recorded in ACS-NSQIP. Thus, although our data are suggestive of appropriate management for patients with DNR status, the exact reason for increased postoperative mortality cannot be elucidated. Further investigation is warranted to continue to characterize the exact circumstances of death in DNR patients undergoing surgical procedures.
We also report that patients with a DNR status had increased total LOS relative to non-DNR patients with 13.1 ± 14.5 days versus 6.7 ± 11.6 days (P < .001). The mean cost per inpatient day for nonprofit hospitals across the United States was estimated to be $2289 in 2013.43 Given the increased average LOS for DNR versus non-DNR patients, this suggests that the cost of hospitalization increases by at least $14,649.60. Furthermore, this calculation does not take into account the potential additional costs of the intensive care unit, where we anticipate many DNR patients would be kept because of numerous comorbidities and emergent surgery, as well as nonresuscitative interventions for postoperative complications. All would impose an even greater cost toward DNR patients and their families than our basic estimate.
Ultimately, these data inform the continued evolution of patient-centered shared decision-making for surgical patients with a DNR status, particularly in end-of-life care. As perioperative physicians, anesthesiologists in particular play a key role in this process, helping patients to elucidate their goals of care and directly enforcing the patient’s wishes for degree of intervention in the operating suite and increasingly in intensive care units. Furthermore, although increased mortality appears to result from appropriate restriction of life-saving interventions, determination of the exact cause of increased mortality will require rigorous assessment of not only postoperative management, but also intraoperative management for patients with a DNR status. It is possible that differences in intraoperative management, as determined by patient’s wishes preoperatively, may influence apparent postoperative outcomes as well.
Limitations of this study are largely centered on the nature of ACS-NSQIP and the retrospective design of the study. All data reported to ACS-NSQIP must be contributed voluntarily by participating practices with no compensation for contribution, implying that there may be variability in the depth, frequency, and accuracy of reporting by institution. However, ACS excludes all data from sites where the 30-day follow-up rate is <80% or the interrater reliability disagreement rate exceeds 5%. This robust audit procedure has apparently paid off, with audit results suggesting that ACS-NSQIP data have been reliable from their inception with reliability improving each year, and identify outcomes more consistently than other similar databases.44,45 Because ACS-NSQIP limits its data collection to defined 30-day outcomes, we cannot assess any complications that occur after this time length. Similarly, there are limited data available on intraoperative events that could lead directly to postoperative complications. In addition, because of its deidentified nature, there is no information available in ACS-NSQIP regarding the socioeconomic status of patients as well as site-specific or regional characteristics. There is also limited information regarding the timing, duration, and interventions required for postoperative complications, which prevents an in-depth economic analysis of the cost of surgery in DNR patients. We did not consider death as a competing risk in our analysis given the nonuniform reporting of time-to-event for both postoperative mortality and morbidity. We acknowledge that this could generate bias, wherein DNR patients may appear to develop fewer postoperative morbidities because of the increased risk of mortality in this population. Finally, there is no indication as to the duration of DNR status preoperatively, time and conditions of a “held” DNR, and when (or even if) DNR status was reinstituted postoperatively. The rigor of the “required reconsideration” discussion preoperatively and the accuracy of information entered in the chart is also unknown. We do know, however, that ACS-NSQIP bases its DNR variable on the most accurate data in the chart at the time of surgery. Despite these limitations, the large size of the data across a breadth of practice environments suggests that our results should be broadly generalizable and reflective of national outcomes for surgical patients with a DNR status.
We characterized the patterns of surgical care and report 30-day postoperative outcomes, including the incidence of resuscitative interventions, in a large, risk-adjusted, contemporary cohort of DNR surgical patients. The most common surgeries performed on DNR patients were classified as emergent and centered on immediate symptom relief. Preoperative comorbidities were independently associated with mortality. However, even after adjustment for these factors, DNR patients were found to have increased incidence of postoperative mortality. Incidence of postoperative morbidity was increased, but ultimately found to be no more frequent than expected after significant risk adjustment. In addition, cardiopulmonary resuscitative measures including CPR and unplanned intubation were less common in the DNR cohort, suggesting that these interventions were held accordingly without exclusion of other treatment options. Together, these findings suggest that increased mortality is the result of adherence to goals of care rather than a “failure to rescue” phenomenon. However, given the increased mortality in the DNR cohort as well as high financial burden for postoperative complications and extended hospitalization, ongoing study is warranted regarding the appropriateness of care in this population.
Name: Elisa C. Walsh, BS.
Contribution: This author helped design and perform the data analysis, interpret the data, and draft the manuscript.
Name: Ethan Y. Brovman, MD.
Contribution: This author helped conceive the study, design and perform the data analysis, interpret the data, and draft the manuscript.
Name: Angela M. Bader, MD, MPH.
Contribution: This author interpret the data and revise the manuscript.
Name: Richard D. Urman, MD, MBA.
Contribution: This author helped conceive the study, design and perform the data analysis, interpret the data, and revise the manuscript.
This manuscript was handled by: Tong J. Gan, MD.
1. Loertscher L, Reed DA, Bannon MP, Mueller PS. Cardiopulmonary resuscitation and do-not-resuscitate orders: a guide for clinicians. Am J Med. 2010;123:4–9.
2. Burns JP, Edwards J, Johnson J, Cassem NH, Truog RD. Do-not-resuscitate order after 25 years. Crit Care Med. 2003;31:1543–1550.
3. Hakim RB, Teno JM, Harrell FE Jr, et al. Factors associated with do-not-resuscitate orders: patients’ preferences, prognoses, and physicians’ judgments. SUPPORT Investigators. Study to understand prognoses and preferences for outcomes and risks of treatment. Ann Intern Med. 1996;125:284–293.
4. Wenger NS, Pearson ML, Desmond KA, Brook RH, Kahn KL. Outcomes of patients with do-not-resuscitate orders. Toward an understanding of what do-not-resuscitate orders mean and how they affect patients. Arch Intern Med. 1995;155:2063–2068.
5. Wenger NS, Pearson ML, Desmond KA, et al. Epidemiology of do-not-resuscitate orders. Disparity by age, diagnosis, gender, race, and functional impairment. Arch Intern Med. 1995;155:2056–2062.
6. Adams SD, Cotton BA, Wade CE, et al. Do not resuscitate status, not age, affects outcomes after injury: an evaluation of 15,227 consecutive trauma patients. J Trauma Acute Care Surg. 2013;74:1327–1330.
7. Cherniack EP. Increasing use of DNR orders in the elderly worldwide: whose choice is it? J Med Ethics. 2002;28:303–307.
8. Morrell ED, Brown BP, Qi R, Drabiak K, Helft PR. The do-not-resuscitate order: associations with advance directives, physician specialty and documentation of discussion 15 years after the Patient Self-Determination Act. J Med Ethics. 2008;34:642–647.
9. Shepardson LB, Youngner SJ, Speroff T, Rosenthal GE. Increased risk of death in patients with do-not-resuscitate orders. Med Care. 1999;37:727–737.
10. Clemency MV, Thompson NJ. Do not resuscitate orders in the perioperative period: patient perspectives. Anesth Analg. 1997;84:859–864.
11. La Puma J, Silverstein MD, Stocking CB, Roland D, Siegler M. Life-sustaining treatment. A prospective study of patients with DNR orders in a teaching hospital. Arch Intern Med. 1988;148:2193–2198.
12. Scott TH, Gavrin JR. Palliative surgery in the do-not-resuscitate patient: ethics and practical suggestions for management. Anesthesiol Clin. 2012;30:1–12.
13. Lilley EJ, Khan KT, Johnston FM, et al. Palliative care interventions for surgical patients: a systematic review. JAMA Surg. 2016:151:172–183.
14. Truog RD. “Do-not-resuscitate” orders during anesthesia and surgery. Anesthesiology. 1991;74:606–608.
15. Ewanchuk M, Brindley PG. Perioperative do-not-resuscitate orders—doing ‘nothing’ when ‘something’ can be done. Crit Care. 2006;10:219.
16. Truog RD, Waisel DB, Burns JP. Do-not-resuscitate orders in the surgical setting. Lancet. 2005;365:733–735.
17. Waisel DB, Burns JP, Johnson JA, Hardart GE, Truog RD. Guidelines for perioperative do-not-resuscitate policies. J Clin Anesth. 2002;14:467–473.
18. Cohen CB, Cohen PJ. Do-not-resuscitate orders in the operating room. N Engl J Med. 1991;325:1879–1882.
19. Guarisco KK. Managing do-not-resuscitate orders in the perianesthesia period. J Perianesth Nurs. 2004;19:300–307.
20. American College of Surgeons. Statement on advance directives by patients: “do not resuscitate” in the operating room. Bull Am Coll Surg. 2014;99:42–43.
22. Siracuse JJ, Jones DW, Meltzer EC, et al. Impact of “do not resuscitate” status on the outcome of major vascular surgical procedures. Ann Vasc Surg. 2015;29:1339–1345.
23. Jawa RS, Shapiro MJ, McCormack JE, Huang EC, Rutigliano DN, Vosswinkel JA. Preadmission do not resuscitate advanced directive is associated with adverse outcomes following acute traumatic injury. Am J Surg. 2015;210:814–821.
24. Speicher PJ, Lagoo-Deenadayalan SA, Galanos AN, Pappas TN, Scarborough JE. Expectations and outcomes in geriatric patients with do-not-resuscitate orders undergoing emergency surgical management of bowel obstruction. JAMA Surg. 2013;148:23–28.
25. Aziz H, Branco BC, Braun J, et al. The influence of do-not-resuscitate status on the outcomes of patients undergoing emergency vascular operations. J Vasc Surg. 2015;61:1538–1542.
26. Maxwell BG, Lobato RL, Cason MB, Wong JK. Perioperative morbidity and mortality of cardiothoracic surgery in patients with a do-not-resuscitate order. PeerJ. 2014;2:e245.
27. Matsushima K, Schaefer EW, Won EJ, Armen SB. The outcome of trauma patients with do-not-resuscitate orders. J Surg Res. 2016;200:631–636.
28. Saager L, Kurz A, Deogaonkar A, et al. Pre-existing do-not-resuscitate orders are not associated with increased postoperative morbidity at 30 days in surgical patients. Crit Care Med. 2011;39:1036–1041.
28a. Beverly A, Brovman EY, Urman RD. Comparison of postoperative outcomes in elderly patients with a do-not-resuscitate order undergoing elective and nonelective hip surgery. Geriatr Orthop Surg Rehabil. 2017;8:78–86.
29. Scarborough JE, Pappas TN, Bennett KM, Lagoo-Deenadayalan S. Failure-to-pursue rescue: explaining excess mortality in elderly emergency general surgical patients with preexisting “do-not-resuscitate” orders. Ann Surg. 2012;256:453–461.
30. Chen JL, Sosnov J, Lessard D, Goldberg RJ. Impact of do-not-resuscitation orders on quality of care performance measures in patients hospitalized with acute heart failure. Am Heart J. 2008;156:78–84.
31. Beach MC, Morrison RS. The effect of do-not-resuscitate orders on physician decision-making. J Am Geriatr Soc. 2002;50:2057–2061.
32. Scarborough JE, Pappas TN. The effect of do-not-resuscitate status on postoperative mortality in the elderly following emergency surgery. Adv Surg. 2013;47:213–225.
33. Henneman EA, Baird B, Bellamy PE, Faber LL, Oye RK. Effect of do-not-resuscitate orders on the nursing care of critically ill patients. Am J Crit Care. 1994;3:467–472.
34. Silvennoinen K, Meretoja A, Strbian D, Putaala J, Kaste M, Tatlisumak T. Do-not-resuscitate (DNR) orders in patients with intracerebral hemorrhage. Int J Stroke. 2014;9:53–58.
35. Powell ES, Sauser K, Cheema N, et al. Severe sepsis in do-not-resuscitate patients: intervention and mortality rates. J Emerg Med. 2013;44:742–749.
36. Raval MV, Wang X, Cohen ME, et al. The influence of resident involvement on surgical outcomes. J Am Coll Surg. 2011;212:889–898.
37. Rosenthal GE, Harper DL, Quinn LM, Cooper GS. Severity-adjusted mortality and length of stay in teaching and nonteaching hospitals. Results of a regional study. JAMA. 1997;278:485–490.
38. Faul F, Erdfelder E, Buchner A, Lang AG. Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behav Res Methods. 2009;41:1149–1160.
39. Nakagawa K, Vento MA, Seto TB, et al. Sex differences in the use of early do-not-resuscitate orders after intracerebral hemorrhage. Stroke. 2013;44:3229–3231.
40. Salottolo K, Offner PJ, Orlando A, et al. The epidemiology of do-not-resuscitate orders in patients with trauma: a community level one trauma center observational experience. Scand J Trauma Resusc Emerg Med. 2015;23:9.
41. Deiner S, Westlake B, Dutton RP. Patterns of surgical care and complications in elderly adults. J Am Geriatr Soc. 2014;62:829–835.
42. Kazaure H, Roman S, Sosa JA. High mortality in surgical patients with do-not-resuscitate orders: analysis of 8256 patients. Arch Surg. 2011;146:922–928.
44. Shiloach M, Frencher SK Jr, Steeger JE, et al. 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;210:6–16.
45. Steinberg SM, Popa MR, Michalek JA, Bethel MJ, Ellison EC. Comparison of risk adjustment methodologies in surgical quality improvement. Surgery. 2008;144:662–667.
46. Hickey TR, Cooper Z, Urman RD, Hepner DL, Bader AM. An Agenda for Improving Perioperative Code Status Discussion. A A Case Rep. 2016;6:411–415.