Figure 1. Mortality rates of select complications stratified by timing with respect to the day of cardiopulmonary resuscitation (CPR). ARF indicates acute renal failure; ARI, acute renal insufficiency; and MI, myocardial infarction.
Figure 2. Outcomes (≤30 postoperative days) of patients who experienced cardiopulmonary resuscitation (CPR) by surgical specialty, American College of Surgeons–National Surgical Quality Improvement Program (2005-2010).
Figure 3. Multivariate analysis of factors associated with 30-day mortality among patients who experienced cardiopulmonary resuscitation, American College of Surgeons–National Surgical Quality Improvement Program (2005-2010). Multivariate regression model adjusted for more than 30 risk factors. Referents are as follows: age was analyzed as a continuous variable; do-not-resuscitate (DNR) status, no preexisting DNR order; American Society of Anesthesiologists (ASA) class, 2 or less; surgical specialty, general surgery; arrest type, intraoperative arrest; all other referents, not present. COPD indicates chronic obstructive pulmonary disease. Error bars indicate 95% CIs.
Customize your JAMA Network experience by selecting one or more topics from the list below.
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(1):14–21. doi:10.1001/jamasurg.2013.671
Author Affiliations: Department of General Surgery, Stanford University, Palo Alto, California (Dr Kazaure); and Department of Surgery, Yale University School of Medicine, New Haven, Connecticut (Drs Roman, Rosenthal, and Sosa).
Objectives To describe the incidence, characteristics, and outcomes of surgical patients who experience cardiac arrest requiring cardiopulmonary resuscitation (CPR).
Design Retrospective cohort study.
Setting American College of Surgeons–National Surgical Quality Improvement Program (ACS-NSQIP), 2005-2010.
Main Outcome Measures Incidence of CPR, complications, mortality, and survival to hospital discharge at 30 days or less after surgery.
Results A total of 6382 nontrauma patients (mean age, 68 years) underwent CPR; 85.9% of events occurred postoperatively, of which 49.8% occurred within 5 days after surgery. Overall incidence of CPR was 1 in 203 surgical cases but varied by specialty (1 in 33 for cardiac surgery vs 1 in 258 for general surgery). The mortality rates varied by specialty (45.0%-74.5%) and were associated with comorbidity burden (58.7% for no comorbidity, 63.1% for 1 comorbidity, and 72.8% for ≥2 comorbidities; P < .001). A total of 77.6% of CPR patients experienced a complication; approximately 75.0% occurred before or on the day of CPR, and septicemia (26.7%), ventilator dependence (22.1%), significant bleeding (13.9%), and renal impairment (11.9%) were the most common. The overall 30-day mortality was 71.6%. Survival to discharge in 30 postoperative days or less was 19.2%; 9.2% of CPR patients were alive but hospitalized at postoperative day 30. Older age, a preexisting do-not-resuscitate order, renal impairment, disseminated cancer, preoperative sepsis, and postoperative arrest were among the factors independently associated with worse survival.
Conclusions One in 203 surgical patients undergoes CPR, and more than 70.0% of patients die in 30 postoperative days or less. Complications commonly precede arrest; prevention or aggressive treatment of these complications may potentially prevent CPR and improve outcomes. These data could aid discussions regarding advance directives among surgical patients.
Cardiac arrest requiring cardiopulmonary resuscitation (CPR) is a major public health problem1,2; 1 event occurs every 90 seconds.3 An estimated 200 000 to 750 000 cardiac arrests occur annually among hospitalized patients.2,4,5 Not all cardiac arrests have the same origin6: out-of-hospital arrest usually results from an acute onset of cardiac arrhythmia,7,8 whereas up to 14% of in-hospital arrests are preceded by complications, such as hypotension, metabolic or electrolyte disturbances, and respiratory insufficiency, and are potentially preventable or modifiable.1,2,9-11
Before the introduction of closed-chest cardiac massage in 1960,12 results of resuscitation after cardiac arrest were reported mainly in surgical patient populations and surgical journals. At that time, the only method of resuscitation available was internal cardiac massage through a thoracotomy performed in the operating room. Introduction of closed-chest CPR redirected scholarly attention to patients who experienced cardiac arrest outside the operating room.13 Currently, most studies1,2,5,14,15 of CPR outcomes are based on data derived from the medical population. The most commonly reported outcome of in-hospital CPR is the rate of survival to hospital discharge, which has remained at 15% to 18% for the past 3 decades.
Most studies of CPR among surgical patients were conducted more than a decade ago16 and were single-institution studies that either excluded or were restricted to cardiac surgery patients.13,16-19 Their focus was often intraoperative arrests examined from the anesthesia perspective.13,17,18 As such, the overall incidence, characteristics, and outcomes of surgical patients who receive CPR are largely unknown.
Using a multi-institutional database, we performed a retrospective analysis of surgical patients who received intraoperative and/or postoperative CPR to determine the incidence of CPR, patient characteristics, and 30-day outcomes. Our aim was to identify factors associated with 30-day mortality that could be targeted to improve patient outcomes.
Quiz Ref IDWe used data collected from up to 250 US hospitals participating in the American College of Surgeons–National Surgical Quality Improvement Program (ACS-NSQIP) that were included in the Participant Use Data File (2005-2010). In ACS-NSQIP, cardiac arrest requiring CPR is defined as the “absence of cardiac rhythm or the presence of chaotic cardiac rhythm that results in loss of consciousness requiring the initiation of any component of basic and/or advanced cardiac life support within 30 days of the operation. Patients with automatic implantable cardioverter defibrillators that fire but the patient has no loss of consciousness should be excluded.” Through medical record review and patient follow-up, trained nurses prospectively collected data regarding preoperative and operative characteristics, as well as 30-day postoperative outcomes of surgical, nontrauma patients captured in the database.
Overall incidence of CPR was calculated as the ratio of (intraoperative and postoperative) CPR occurrences to the total number of surgical cases in the entire ACS-NSQIP Participant Use Data File (2005-2010). The ACS-NSQIP collects data for more than 40 comorbidities. On the basis of a calculated mean of 1.6 comorbidities per surgical case in the database, CPR incidence also was stratified by the presence of 0, 1, or 2 or more comorbidities. The database includes a variable that specifies the primary specialty of the attending surgeon. Incidence of CPR by specialty of the attending surgeon was calculated as the ratio of CPR occurrence in each specialty to the total number of cases for that specialty in the database. Primary operative procedures were abstracted using the Current Procedural Terminology variable included in the ACS-NSQIP.
Demographic characteristics included age, sex, race, preexisting do-not-resuscitate (DNR) orders instituted within the month before surgery, residence, and functional status before surgery. Clinical characteristics included emergency vs nonemergency surgery, American Society of Anesthesiologists (ASA) classification, and operative wound classification as defined by the primary surgeon.
Preexisting conditions included hypertension requiring medication; diabetes mellitus; ventilator dependence; history of chronic obstructive pulmonary disease; current pneumonia; newly diagnosed or worsening congestive heart failure 30 days or less after surgery; myocardial infarction 6 months before surgery; history of cardiac surgery or percutaneous coronary intervention; peripheral vascular disease, including revascularization and rest pain; ascites 30 days or less before surgery; acute renal failure 24 hours or less before surgery; dialysis dependence (≤2 weeks before surgery); history of paralysis (hemiplegia, paraplegia, or quadriplegia); impaired sensorium 48 hours or less before surgery; non–drug-induced preoperative coma; disseminated cancer; presence of an open wound before surgery; preoperative sepsis; and preoperative blood transfusion of 5 U or more in the 72 hours before surgery.
Clinical outcomes of interest were occurrence of 1 or more postoperative complications, survival to hospital discharge, and mortality at 30 days or less after surgery. Complications included (re)intubation, prolonged ventilator use (failure to wean from ventilator at >48 hours), pneumonia, venous thromboembolism (pulmonary embolism, deep venous thrombosis, or thrombophlebitis), myocardial infarction, stroke, coma lasting longer than 24 hours, wound infections (superficial, deep, and organ space) and dehiscence, severe sepsis or shock, bleeding requiring 5 U or more of blood, acute renal failure or insufficiency, and urinary tract infection. The timing of the occurrence of complication(s) was categorized into 4 occurrence intervals: (1) more than 1 day before CPR, (2) 1 day before CPR, (3) on the day of CPR, and (4) 1 or more days after CPR.
Bivariate analyses of categorical variables were performed using a 2-tailed χ2 test. A multivariate stepwise logistic regression model was generated to determine risk factors for 30-day postoperative mortality among CPR patients; P < .10 was used as the initial entry criterion. Odds ratios with 95% CIs were calculated. P < .05 was considered statistically significant.
Data analyses and management were performed using SPSS statistical software for Windows, version 19.0 (SPSS, Inc). Definitions of variables used in this study are available from the ACS-NSQIP Participant Use Data File user guide.20 The ACS-NSQIP Participant Use Data File is a public database with deidentified data; therefore, this study was granted exemption by our institutional review board.
Quiz Ref IDAmong more than 1.3 million surgical cases captured in the ACS-NSQIP data set, 6382 patients experienced cardiac arrest requiring CPR. The overall incidence was 1 event per 203 cases. The incidence of CPR among patients with 0, 1, or 2 or more comorbidities was 1 in 2174, 1 in 699, and 1 in 95, respectively. Do-not-resuscitate orders were present in 0.6% of all cases in the ACS-NSQIP Participant Use Data File; 1.9% of patients who received CPR had a DNR order.
Quiz Ref IDThe overall 30-day mortality in the ACS-NSQIP data set was 1.7%; patients who received CPR had a mortality of 71.6% (P < .001). The mortality on the day of CPR was 51.5%; 1 day after CPR, it was 59.4%. Mortality in the interval between the day of CPR and postoperative day 30 increased by 39%. The overall rate of survival to discharge at 30 postoperative days or less was 19.2%; 9.2% of patients who received CPR were alive but hospitalized at postoperative day 30.
The mean age of the CPR patients was 68.2 years (range, 17-90 years), which was higher than the mean age of patients in the entire ACS-NSQIP (55.5 years; P < .001). As indicated in Table 1, CPR patients were more often male and assigned an ASA classification of 4 to 5. A total of 39.5% of CPR patients had systemic sepsis preoperatively (Table 2). The mean number of comorbidities was approximately 5, which was higher than the approximate mean of 2 comorbidities for the entire ACS-NSQIP data set (P < .001); 8.8% of CPR patients had 1 preexisting condition, and 2.8% had none. Mortality rates of CPR patients increased with more comorbidities (58.7%, 63.1%, and 72.8% for patients with 0, 1, and ≥2 conditions captured in the data set; P < .001).
Quiz Ref IDApproximately 86.0% of cardiac arrests occurred postoperatively. The overall incidences of postoperative arrest and intraoperative arrest were 0.42% and 0.07%, respectively. A total of 240 patients (3.8%) sustained more than 1 arrest. Nearly 9.0% of patients who experienced an intraoperative arrest had a postoperative arrest. Approximately half of postoperative arrests occurred within 5 days after surgery. The mortality for intraoperative arrest was 61.1%, and for postoperative arrest, 73.1% (P < .001). Compared with patients who experienced postoperative arrest, those who experienced an intraoperative event had a higher rate of survival to hospital discharge (17.0% vs 33.7%; P < .001).
A total of 77.6% of CPR patients sustained a complication. A total of 13 467 postoperative complications other than cardiac arrest occurred among these patients; 75.2% of complications occurred before or on the day of arrest. As indicated in Table 3, the most common postoperative complications among CPR patients included respiratory distress, septicemia, renal dysfunction, and bleeding that required 5 U or more of blood. When categorized by timing of complications, mortality rates were worse when complications occurred before, rather than after, CPR (Figure 1).
The incidence of CPR in the ACS-NSQIP database varied by surgical specialty (Table 4). The rates of preexisting DNR orders in the database varied also, ranging from 0.1% among gynecologic surgery patients to 1.3% among vascular surgery patients.
Quiz Ref IDPatients who experienced CPR underwent a variety of procedures. Common procedures among cardiac surgery patients included coronary artery bypass grafts (43.8%) and valve replacement or repair (37.5%); 56.0% of gynecology patients underwent hysterectomy; 42.0% of neurosurgery patients had laminectomy; 52.5% of orthopedic patients underwent femur or hip fracture repairs; 19.0% of otolaryngology patients underwent neck dissection; 33.3% of plastic surgery patients underwent abscess drainage or wound debridement; 22.9% and 20.9% of thoracic surgery patients underwent cardiac-related procedures and pneumonectomy, respectively; and 57.1% of urology patients underwent nephrectomy or prostatectomy. Aortic repairs (31.6%), bypass procedures (18.3%), and amputations (15.4%) were common among vascular surgery patients who underwent CPR. Colectomy (25.3%), small-bowel procedures (12.3%), and exploratory laparotomy (9.7%) were the most common procedures performed on general surgery patients who underwent CPR.
All outcomes of interest varied by surgical specialty (Figure 2). Given the considerable disparity in outcomes of cardiac surgery patients who received CPR vs those in other surgical specialties, a subanalysis of baseline characteristics of cardiac vs general surgery patients was performed. In terms of their noncardiac comorbidity profile, cardiac surgery patients who received CPR appeared to be healthier than general surgery patients. They had lower rates of disseminated cancer, preoperative ventilator dependence, ascites, dialysis, and preoperative sepsis (all P < .001). However, cardiac surgery patients were more likely to have cardiac comorbidities, such as a recent history of myocardial infarction, congestive heart failure, and cardiac catheterization (all P < .001). Overall, cardiac surgery patients were less likely to be assigned an ASA class of 5 (P < .001).
After adjusting for more than 30 risk factors in multivariate analysis, several baseline factors remained independently associated with mortality (Figure 3). Older age (odds ratio, 1.018; 95% CI, 1.01-1.02; P < .001) was associated with worse survival. An ASA class of 5 was most strongly associated with compromised survival.
Surgical patients who receive CPR are generally older and have more comorbidities than the average surgical patient. The overall incidence of CPR (4.9 arrests per 1000 surgical cases) found in this study is within the range of 1 to 5 arrests per 1000 admissions reported in the medical literature.5,14 The incidence of intraoperative arrest in this study (6.7 per 10 000 cases) is also within a reported range of 4.6 to 19.7 per 10 000 cases.17 Our results are consistent with published data indicating that prearrest factors, such as metastatic cancer, impaired renal function, and sepsis, have prognostic value in the evaluation of CPR outcomes.5,11,21,22 The ultimate goal of resuscitation is to improve survival with good neurologic outcomes.23 The rate of postarrest coma (3.1%) in our study is similar to the 3.6% rate found in a predominantly nonsurgical database of cardiac arrests.1
Our results suggest that cardiac arrest is a more survivable event among surgical patients than medical patients. In this study, mortality on the day of CPR was 51.5%. An analysis of more than 14 000 CPR events, of which more than 80% occurred among medical patients, revealed a 56% mortality on the day of CPR.1 Compared with the 40.6% 1-day postarrest survival in this study, Nadkarni et al23 reported a 24-hour survival of 30% for predominantly nonsurgical adult CPR patients included in a multicenter registry of cardiac arrests. The 19.2% rate of survival to hospital discharge found in this study compares favorably to the 17.5% rate found in a meta-analysis of CPR outcomes in the medical literature.11 Considering that nearly 60% of patients in this study underwent nonemergency surgery, surgical patients who receive CPR may be healthier patients at baseline than patients described in the CPR medical literature. The improved CPR outcomes of surgical patients relative to patients in the CPR medical literature could also be a reflection of the superior outcomes of intraoperative arrests than arrests outside the operating room.17
Perioperative cardiac arrest is often fatal. Because only a few patients survive, prevention is crucial. An expert panel that reviewed 118 cases of in-hospital cardiac arrest concluded that more than 60% were potentially avoidable.24 Evidence of clinical deterioration during the 8 hours before in-hospital CPR has been reported in up to 84% of cases.25 It is a significant finding that more than 1100 patients in this study who were operated on without evidence of preoperative sepsis experienced postoperative sepsis or shock, and of these, 790 died. Sepsis among surgical patients is associated with an estimated 30% to 39% mortality26; we found that the mortality rate more than doubled if sepsis occurred before arrest compared with after arrest. Overall, three-quarters of complications, including sepsis, hypotension, and acute impairment of renal function, occurred before or on the day of cardiac arrest. The timing of complications affected patient mortality; patients were more likely to die of an arrest if they had a prearrest complication. These findings indicate that a large proportion of cardiac arrests and periarrest deaths among surgical patients might be preventable if prearrest complications would be avoided or expediently treated.
As highlighted in our subanalysis of cardiac vs general surgery patients, variations in the incidence and outcomes of CPR by surgical specialty likely reflect differences in baseline characteristics of the patient cohorts by specialty. Differences in postoperative care also may play a role. Cardiac surgery patients often receive postoperative care in an intensive care unit (ICU); vascular and thoracic patients also are more likely to receive care in the ICU. Most studies2,5,22,27 have reported better outcomes of CPR events in the ICU compared with the general hospital. This finding might be explained by the availability of better monitoring and specialty skilled care in the ICU.27-30 Intensive monitoring, immediate availability of medical intervention, and potentially reversible causes of arrest also could explain improved survival from intraoperative arrest.
Limitations of our study largely stem from the fact that the Participant Use Data File is an administrative database; although there may be coding errors, it has been validated.31 A cause-effect relationship between prearrest complications and mortality cannot be inferred because in-depth analysis of periarrest events was not possible; intra-arrest factors known to influence resuscitation outcomes, such as promptness, duration, and quality of CPR; type of arrest rhythm; and location of the patient and monitoring status (ICU vs general hospital), are not captured in the database. Measures of severity of surgical illness and complexity of comorbidity also are not provided. Complication rates may have been underestimated because resuscitation-related injuries, such as rib and sternal fractures, pneumothorax, cervical hematoma, and perforation of abdominal organs, including liver, stomach, and spleen,32 are not captured in the ACS-NSQIP database; occurrence of such complications could affect postarrest complications or mortality risk. One study1 found that 63% of patients who survived arrest but eventually died were assigned a DNR order, and 43% had life support withdrawn in the interval between survival and death; such decisions could influence postarrest mortality rates. The ACS-NSQIP database does not capture changes in DNR status instituted during the admission under study. There is no information about hospital-level factors and outcomes beyond 30 postoperative days, so readmissions and long-term outcomes could not be analyzed. The strengths of our study include its multi-institutional nature and large sample size, both of which attenuate bias related to patient and physician characteristics.
In summary, surgical patients who experience CPR have a high mortality rate. More than three-quarters of arrest patients experienced a complication often in the prearrest period; this finding suggests that some CPR events may be preventable with timely interventions. Fastidious care is especially important for vulnerable patients who are elderly, functionally impaired, or admitted from other care facilities and who in this study had higher rates of CPR occurrence and compromised outcomes after CPR. Cardiopulmonary resuscitation has economic implications given that a third of CPR survivors continue to use in-hospital resources more than a month after surgery. Prevention, early detection of complications, and aggressive interventions may improve outcomes of CPR surgical patients. These data have implications on patient-physician discussions regarding DNR orders and the assessment of prognosis should cardiac arrest occur. Modifiable risk factors that predispose surgical patients to cardiac arrest need to be studied further because they may be specialty specific.
Correspondence: Julie A. Sosa, MD, MA, Department of Surgery, Yale University School of Medicine, 330 Cedar St, FMB 130B, PO Box 208062, New Haven, CT 06510 (email@example.com).
Accepted for Publication: March 20, 2012.
Author Contributions:Study concept and design: Kazaure, Roman, and Sosa. Acquisition of data: Kazaure. Analysis and interpretation of data: Kazaure, Roman, Rosenthal, and Sosa. Drafting of the manuscript: Kazaure, Roman, and Sosa. Critical revision of the manuscript for important intellectual content: Roman, Rosenthal, and Sosa. Statistical analysis: Kazaure and Sosa. Administrative, technical, and material support: Roman, Rosenthal, and Sosa. Study supervision: Roman and Sosa.
Conflict of Interest Disclosures: None reported.
Create a personal account or sign in to: