Cardiac arrest is a well-established risk of anesthesia, surgery, and interventional procedures. One recent report has documented improved survival rates among patients who undergo cardiopulmonary resuscitation in the hospital,1 but this report aggregated all hospitalized patients, including perioperative patients. Other studies of cardiac arrest in patients undergoing anesthesia have examined incidence and risk factors for these events,2–4 some drawing on large databases.3,5 A recent study that queried the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) database concluded that the in-hospital risk for cardiac arrest in patients undergoing selected elective surgeries was approximately 1 in 200.6 This estimate is substantially lower than the percent of surgical patients who die before hospital discharge (4 per 100) reported in a recent pan-European paper,7 but substantially higher than the anecdotal experience of most anesthesiologists viewing just intraoperative and postanesthesia care unit (PACU) events. A single-center experience published in 2014 documented mortality of 7 per 10,000 patients within 24 hours of surgery during a 10-year period.8 Another single-center study of emergency response pages in a large academic operating room (OR) suite found a rate of 1 in 1400 cases.9 We sought to establish the risk for cardiac arrest and death during the immediate perioperative period in a large, multicenter, and contemporaneous sample.
The National Anesthesia Clinical Outcomes Registry (NACOR) is a potentially useful tool to describe cardiac arrest in periprocedural patients. Although NACOR lacks details about patient comorbidities and hospitalizations, it is rich in reliable information about anesthetics and the procedures for which the anesthetics were administered. Surgically oriented databases overlook patients undergoing procedures in radiology or the gastrointestinal suite, who are a growing population of patients in the United States and internationally. NACOR captures the portion of these patients, presumably sicker or more complex, who receive care from an anesthesiologist.
During the immediate periprocedural period, the patient is intensively monitored by an anesthesia team, with a provider-to-patient ratio in excess of 1. Using NACOR data, we performed a retrospective analysis on a purposive sample of patients who experienced cardiac arrest during the immediate periprocedural period. Our goal was to identify factors associated with the occurrence of cardiac arrest during this time.
NACOR collects administrative and clinical information from every case performed in participating anesthesia practices. Data are harvested in periodic reports from electronic systems (billing, clinical, and quality) and mapped directly into the NACOR schema. Data are authenticated through a variety of automated routines and through periodic “outlier” reviews by Anesthesia Quality Institute personnel. Data files to NACOR are heterogeneous in content, ranging from a minimal dataset of about 20 elements per case up to full reporting from anesthesia information management systems, which can include every medication dose, every vital sign recorded, and structured data from the entire course of the anesthesia procedure.10,11 Now, >200 practices participate in NACOR, representing 8000 anesthesiologists in 1800 procedural facilities. Most have contributed electronic records dating to January 1, 2010. Approximately 15% of cases come from practices that routinely record clinical quality indicators for every case, including the occurrence of “cardiac arrest.” This field is completed at the time of discharge from care by an anesthesiologist.
Figure 1 demonstrates how the current study dataset was constructed from NACOR as a whole by selecting those practices and cases in which the outcome of interest (perioperative cardiac arrest) is reliably reported. Although this includes only 15% of all cases in NACOR, the resulting population of 1.69 million patients, from 27 practices, remains substantial.
Because NACOR absorbs data from multiple different health care information technology platforms, variation in coding standards must be carefully identified and managed when constructing research datasets. An example can be seen in Figure 1, in which the exclusion of patients presenting to the OR for organ procurement after the declaration of death must be excluded using 2 different filters: The first filter eliminates cases based on coding as ASA physical status (PS) VI, a common, but not universal, convention. The second filter eliminates cases based on the Current Procedural Terminology code for organ harvest surgery and was added when review of common cases in the final dataset revealed this category. Of note, in the 900 cases removed in this way, the self-reported outcome of “patient mortality” was only sporadically present, presumably because most providers would not think to enter this as an adverse outcome in such a case.
The definition of perioperative cardiac arrest is heterogeneous across NACOR participants. Although there are precise definitions in the Standard Nomenclature for Medicine terminology set and in the recommendations of the American Society of Anesthesiologists’ Committee on Performance and Outcome Measures, it is unlikely that every practitioner is familiar with these. However, the definition of cardiac arrest across the practices reporting this measure to NACOR is consistent, and aggregation to identify a national benchmark for quality reporting is reasonable. The definition in practice includes any unexpected episode of cardiopulmonary resuscitation (open or closed chest) in the OR or stage 1 or 2 PACU.
Anesthesia quality reporting tools capture only the most severe adverse outcome when >1 is present. Episodes of bradycardia that respond to atropine or glycopyrrolate are generally not reported as cardiac arrest in the OR. If the patient has a cardiac arrest and dies, the reported code is “death” rather than cardiac arrest even though the patient experienced both events. Therefore, we have inferred that all patients who died also experienced a cardiac arrest and have included them in the study dataset. This allows us to calculate a crude rate of survival from periprocedural cardiac arrest that can be compared with data from other sources.
The NACOR database had 11,478,920 cases on September 1, 2013, of which 1,691,472 were from practices that routinely reported cardiac arrest as a quality outcome. A total of 951 cardiac arrests were reported during OR and PACU care. Of these, there were 396 survivors and 555 deaths (58.4% mortality). These results are the reported data from 49 facilities, 27 practices, and 408 providers (Table 1). One hundred percent of these practices reported at least 1 cardiac arrest.
Figure 2 illustrates the distribution of incidents by age. For patients with age <1 year, the incidence of cardiac arrest is highest (0.35%). After the age of 1 year, mortality decreases, and then there is an increase in number of arrests and incidence by age, up to the age group 66 to 79 years. Patients aged 80 years or older had fewer arrests numerically, but the rate of arrest in this age group was higher than patients aged 66 to 79 years (0.10% and 0.08%, respectively). Between the ages of 50 and 79 years, 55.9% of arrests occurred. Figure 3 illustrates rates by gender. Mortality was twice as high in males compared with females (0.045% vs 0.024%). By ASA PS class, most arrests (62.5%) were in individuals with ASA PS class III or IV. This is illustrated in Figure 4. Not surprisingly, the highest cardiac arrest rate (7.2%) was in patients with ASA PS V, >10 times the mortality of the next highest class (ASA PS IV, 0.39%).
Most arrests (89.1%) occurred in cases classified as general anesthesia. Figure 5 shows the relative distribution of arrests by anesthesia type. General anesthesia cases had higher absolute values (847) and a higher incidence (.075%) of cardiac arrest. Table 2 describes the arrest data by anatomic region and type of surgery. Intracranial procedures (0.35%) had the highest incidence of arrest, whereas intra-abdominal procedures (15.75% of all arrests) were the largest subgroup. According to the procedure codes, “Anesthesia for intraperitoneal procedures in upper abdomen including laparoscopy; not otherwise specified” (108, 77% mortality) and “Anesthesia for intraperitoneal procedures in lower abdomen including laparoscopy; not otherwise specified” (71, 83% mortality) were the most common procedure types associated with cardiac arrest. Mortality was higher for “Anesthesia for intracranial procedures; craniotomy or craniectomy for evacuation of hematoma” (13/14, 93%) and all open intracranial procedures (25/30 arrests in all categories, 83%). We were unable to differentiate between scheduled cases and trauma or emergency cases with the data available.
Our data suggest that cardiac arrest and death are less common during perioperative anesthesia care than during the subsequent course of hospitalization of a surgical patient. Kazaure et al.,6 using NSQIP data, noted an overall risk of 1 in 200 for in-hospital cardiac arrest in patients undergoing surgery. Our data suggest that a small fraction of these arrests takes place in the OR. In our NACOR dataset, approximately 1 in 1800 patients undergoing surgery had a cardiac arrest in the immediate perioperative period, compared with approximately 1 in 1400 from the NSQIP data.6 In the latter, 86% of cardiac arrests occurred postoperatively; how many occurred during the first 24 hours of postoperative care versus later in the hospital course of a patient was not reported. Ellis et al.8 documented mortality of 7 per 10,000 occurring within 24 hours of surgery in a single center from 1999 to 2009, which would imply a cardiac arrest rate of at least 1 in 1400.
Weingarten et al. documented a 1 in 1400 rate of emergency page activations in operative cases at the Mayo Clinic, with a higher rate in children aged <1 year. The difference between this figure and the rate of cardiac arrests in NACOR may be due to different case mix—the Mayo Clinic is a tertiary facility with a likely higher proportion of seriously ill patients—or to inclusion of some noncardiac arrest emergency calls in the Mayo data. It is interesting that each of these sources identifies intraoperative emergencies at about the same order of magnitude and with about the same distribution of patient age.
There are several factors that could challenge these findings. Because the Anesthesia Quality Institute is a voluntary reporting database, it may not represent the broad population of patients. Although cardiac arrest and death, unlike many variables that might be reported, are sufficiently distinct that misclassification is less problematic, these events may occur more or less frequently in this sample of practices. It is possible that the 49 academic and private facilities reporting these cases could have an overall outcome that is significantly different from the rest of the country, based on the rigor of their quality management programs, their earlier participation in NACOR, or other unmeasured confounders.
There are interesting trends in the data worth investigating in the future. Patients classified ASA PS III or IV constitute >60% of the cardiac arrests in the database; similarly, more than half of the arrests occurred in patients aged between 50 and 79 years. The highest incidences of arrest were in patients with ASA PS IV or V and in patients aged <1 year and >80 years, concordant with common clinical intuition. The NSQIP data analyzed by Kazaure et al. are consistent with these findings. In that study, >73% of cardiac arrests were in patients between 55 and 85 years, and >80% of arrests occurred in patients with ASA PS III and IV.
The differences in risk and outcome related to gender and age are remarkable here. Not only did males have a higher overall risk, but also that risk increased with age from 50 to 79 years (Fig. 6). The greatly increased risk for cardiac arrest in patients <1 year is also noteworthy and not yet explained. In the NSQIP data, which do not include intracranial procedures, cardiac, vascular, and thoracic surgical specialties were associated with the highest incidences of cardiac arrest, although general surgical cases constituted the largest number of arrests (more than half of the total). Our data do not specify surgical specialty, but intracranial surgeries had the highest overall incidence of cardiac arrest (0.30%).
These data do not imply that cardiac arrest is a negligible problem for perioperative clinicians. One hundred percent of reporting anesthesia practices, almost one-third of facilities and more than one-fifth of reporting providers, reported at least 1 cardiac arrest in the 3.5 years of data analyzed. These percentages suggest that although the risk may be low for individual patients, cardiac arrest is a regular occurrence at the practice level. Wide confidence intervals, as seen in Figures 3–5, imply that we still lack a precise estimate of arrest in subsets of our population. Training in resuscitation and ongoing efforts to prevent and mitigate the effects of cardiac arrest are likely to be beneficial, including the emerging use of crisis manuals and checklists in many practices.
The potential strengths of our analysis include novel data from a large database and easily captured end points. As a database analysis, the results are subject to potential biases in reporting and assessment. The results should be considered hypothesis forming. Longer-term outcomes, especially survival to hospital discharge, are not known. Future studies should more closely examine outcomes in anesthesia patients <1 year old and associations with type of facility, type of surgery, and training of providers. The difference in outcomes over all ages based on patient gender should be similarly scrutinized. Future explorations of age and gender differences in cardiac arrest risk and features of anesthesia and surgery that most affect these risks could add to the already favorable safety profile of anesthetic care.
Name: Mark E. Nunnally, MD, FCCM.
Contribution: This author helped conceive this study and helped write the manuscript.
Attestation: Mark E. Nunnally approved the final manuscript.
Name: Michael F. O’Connor, MD, FCCM.
Contribution: This author helped conceive this study and write the manuscript.
Attestation: Michael F. O’Connor approved the final manuscript.
Name: Hubert Kordylewski, PhD.
Contribution: This author retrieved the data and generated the figures with Dr. Dutton.
Attestation: Hubert Kordylewski approved the final manuscript.
Name: Benjamin Westlake, BS.
Contribution: This author retrieved the data and generated the figures with Dr. Dutton and generated the confidence intervals in the incidence data.
Attestation: Benjamin Westlake approved the final manuscript.
Name: Richard P. Dutton, MD, MBA.
Contribution: This author helped conceive this study and coordinated the retrieval of the data from NACOR. He generated the confidence intervals in the incidence data and helped write the manuscript.
Attestation: Richard P. Dutton approved the final manuscript.
This manuscript was handled by: Sorin J. Brull, MD, FCARCSI (Hon.).
1. Girotra S, Nallamothu BK, Spertus JA, Li Y, Krumholz HM, Chan PSAmerican Heart Association Get with the Guidelines–Resuscitation Investigators. . Trends in survival after in-hospital cardiac arrest. N Engl J Med. 2012;367:1912–20
2. Caplan RA, Posner KL, Ward RJ, Cheney FW. Adverse respiratory events in anesthesia: a closed claims analysis. Anesthesiology. 1990;72:828–33
3. Biboulet P, Aubas P, Dubourdieu J, Rubenovitch J, Capdevila X, d’Athis F. Fatal and non fatal cardiac arrests related to anesthesia. Can J Anesth. 2001;48:326–32
4. Ramachandran SK, Mhyre J, Kheterpal S, Christensen RE, Tallman K, Morris M, Chan PSAmerican 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;119:1322–39
5. Lagasse RS. Anesthesia safety: model or myth? A review of the published literature and analysis of current original data. Anesthesiology. 2002;97:1609–17
6. 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
7. Pearse RM, Moreno RP, Bauer P, Pelosi P, Metnitz P, Spies C, Vallet B, Vincent JL, Hoeft A, Rhodes AEuropean Surgical Outcomes Study (EuSOS) Group for the Trials Groups of the European Society of Intensive Care Medicine and the European Society of Anaesthesiology. . Mortality after surgery in Europe: a 7 day cohort study. Lancet. 2012;380:1059–65
8. Ellis SJ, Newland MC, Simonson JA, Peters KR, Romberger DJ, Mercer DW, Tinker JH, Harter RL, Kindscher JD, Qiu F, Lisco SJ. Anesthesia-related cardiac arrest. Anesthesiology. 2014;120:829–38
9. Weingarten TN, Abenstein JP, Dutton CH, Kohn MA, Lee EA, Mullenbach TE, Narr BJ, Schroeder DR, Sprung J. Characteristics of emergency pages using a computer-based anesthesiology paging system in children and adults undergoing procedures at a tertiary care medical center. Anesth Analg. 2013;116:904–10
10. Dutton RP, Dukatz A. Quality improvement using automated data sources: the anesthesia quality institute. Anesthesiol Clin. 2011;29:439–54
11. Grissom TE, DuKatz A, Kordylewski H, Dutton RP. Bring out your data: the evolution of the National Anesthesia Clinical Outcomes Registry. Int J Comput Models Algorithms Med. 2011;2:51–9