Anesthesia, Perioperative Mortality, and Predictors of Adverse Outcomes

John H. Eichhorn

Zaki-Udin Hassan

CASE SUMMARY

A 39-year-old, 5 ft 5 in., 380-lb woman (body mass index = 63) presented to a recently opened bariatric surgery center at a 200-bed exurban community hospital for laparoscopic gastric bypass surgery. She had type 2 diabetes and dyspnea, with palpitations from minor effort but no cardiac diagnosis. A nonsmoker, she slept most nights sitting in a 60-degree position in a lounge chair but never had a sleep study. Her internist and surgeon cleared her for anesthesia. The surgeon had been recently recruited from fellowship training.

PREOPERATIVE FINDINGS: Blood pressure 155/95, heart rate 90, respiratory rate 22, temperature 37°C, and oxygen saturation as measured by pulse oximetry (SpO₂) 91% while she breathed room air, rising to 95% on O₂ 2 L per minute through nasal prongs. Her electrocardiogram showed normal sinus rhythm at 96 bpm with left ventricular hypertrophy and nonspecific ST-T wave changes. A right subclavian double-lumen catheter was inserted on the second pass. General anesthesia was conducted with midazolam, fentanyl, etomidate, rocuronium, and desflurane in O₂, with additional intermittent boluses of fentanyl. The surgeon requested limitation of crystalloid infusion. Peak airway pressures exceeded 45 cm H₂O. With positive end-expiratory pressure of 7 cm H₂O, SpO₂ stayed in the low 90s. At the end of the case, the surgeon urged that the patient be extubated immediately. When she was placed in a semisitting position, she met the usual criteria and was extubated in the operating room at 5:15 PM.

She received no pain medication in the postanesthesia care unit and was stable, with SpO₂ 91% while she breathed O₂ 3 L per minute through nasal prongs. When the postanesthesia care unit closed at 7:00 PM, she was transported in a semisitting position to a standard single room on the floor, with orders for sliding scale insulin and morphine through patient-controlled analgesia. At 10:00 PM, her vital signs (including spot-check SpO₂) were unchanged, and she was alone, dozing (with snoring) and watching television. At next contact, almost 3 hours later (12:45 AM), the patient was found dead and did not respond to extensive resuscitation. The surgeon dictated into the medical record that this was an “anesthesia death,” which was reported to the family, the hospital quality assurance (QA) system, and the state department of health.

Has Anesthesia Outcome Improved?

Anesthesia practitioners who have practiced more than 25 years widely accept that anesthesia care is safer now than, for example, in 1980—at least regarding catastrophic adverse outcomes and probably for complications in general. Traditionally, because surgical anesthesia care is facilitative rather than therapeutic, a good outcome is measured in terms of the absence of complications or adverse outcomes. Avoidance of preventable complications that lead to adverse patient outcomes has been the focus of the organized patient safety movement that began in the United States in the early 1980s and spread worldwide.¹

Many reasons can be cited as to why anesthesia care has become safer. Utilization of sensitive electronic monitors, such as the capnograph and pulse oximeter, extends the human senses of the physician. Application of monitoring standards²,³,⁴ is a dramatic example of anesthesia patient safety improvement, particularly regarding intraoperative anesthesia catastrophes that lead to cardiac arrest, permanent brain damage, or death. Other important components contributing to improved anesthesia patient safety include:

Improved training of practitioners (better students entering training, better teachers, a longer training period, improved/expanded textbooks and journals, teaching with simulation)
Better communication of safety information and recommendations through multiple professional organizations and the Anesthesia Patient Safety Foundation
More extensive and focused research into safety and human factors
More and improved anesthetic and ancillary medications
Other equipment and technologies, such as fiberoptics for airway management; and the development of practice checklists, protocols, and algorithms from various sources

However, some counterbalancing forces have tended to increase the risk of complications and adverse patient outcomes. Surgical procedures are expanded in scope, complexity, and length, and older patients and patients who are more ill are considered potential surgical candidates than in previous decades. The combination of these factors has increased the aggregate intensity of the challenge involved in providing safe anesthesia care. Further, the demand for anesthesia care in many locations outside the traditional hospital or surgicenter operating room is perceived by many practitioners as a risk factor for increased complications and adverse patient outcomes.

Evidence that anesthesia care is safer now includes an important alternate definition of truth that goes beyond the traditional “p < 0.05” statistical approach of data analysis. The nonparallel construction of studies and databases use variable definitions and methods. In addition, no adjustments are made for differing settings and patient populations. Comparatively small samples are assessed in an arena where it would take huge population groups to demonstrate statistically valid changes in the rate of very rare events. Therefore, it is very difficult to draw conclusions by comparing statistical studies and reports in the literature.⁵

One approach in the United States is to consider the conclusions on this point by an integral component of the American health care system—the medical malpractice insurance industry. Malpractice insurance premiums directly reflect the losses of insurance companies from adverse patient outcomes that result in insurance claims, settlements, and court judgments. Since the end of the 1980s, malpractice insurance premiums in the United States actually decreased for many anesthesiologists and did not increase at the same rate as they did for virtually all other physicians. This observation provides significant functional evidence indicating that anesthesia care has become safer.

Insurance company actuaries are not charitable people, but they will reduce premiums, actually and relatively, only when the calculations show a decreased loss by their companies. Given the nature of the medical-legal system and associated public attitudes over the last 3 decades in the United States, the vast majority of unexpected catastrophic patient outcomes in the perioperative period should have been brought to the attention of plaintiffs’ attorneys. The decrease in insurance company losses in this time frame is a result of fewer and less serious insurance cases involving anesthesia catastrophes. Insurance premiums have increased in this decade, and the availability of malpractice insurance is problematic in certain locations. Nonetheless, the trend of proportionately less increases in premiums persist for anesthesiologists compared with those for physicians in the high-risk specialties of obstetrics and gynecology, neurosurgery, and orthopedic surgery, thereby supporting the improved safety in anesthesia practice. As an example, in 2002, the inflationadjusted malpractice insurance premium for Harvard Medical School-insured anesthesiologists was approximately one fourth of the rate in the mid-1980s.⁵ For all anesthesiologists, based on analysis of premium amounts for $1 million or $3 million policy limits, the average premium in 2006 was $19,558 compared with an average inflation-adjusted premium of $32,502 in 1985.⁶ While the overall trend toward safer anesthesia should continue, in light of the competing forces outlined in the preceding text, it is appropriate to consider anesthesia perioperative mortality and predictors of adverse outcomes.

How Is Anesthesia Perioperative Mortality Defined?

Although death is clearly definable, historically, the concept of anesthesia mortality has been confused and very poorly articulated. Various definitions of perioperative mortality have been employed since the 1950s; accordingly, the evaluation and comparison of the various reports and statistics are difficult.

▪ POSTOPERATIVE TIME FRAME

A key factor is the issue of time frame in relation to the surgery. Beyond the agreement that intraoperative deaths qualify as perioperative mortality, there is little consistency as to how long to extend the postoperative period in order for deaths to count as perioperative. Different sources have used widely differing postoperative time boundaries:

The American College of Surgeons—30 days
The American Hospital Association—no identifiable limit
The Joint Commission on Accreditation of Healthcare Organizations—within 48 hours
The U.S. Center for Medicare and Medicaid Services—no identifiable limit
The British National Health Service in their National Confidential Enquiry into Perioperative Death—within 29 days
Study groups, such as the South Australian Perioperative Mortality Committee—within 24 hours, compared with the neighboring Anesthesia Mortality Commission of Western Australia—within a 48-hour time frame

▪ CAUSE OF DEATH

Beyond the significant difficulty resulting from nonparallel definitions of the time frame that defines perioperative mortality is the even greater problem of dissecting and attributing the causation for the deaths. Some discussions
consider death from any cause as perioperative mortality, and record it as such, be it direct surgical error or complications; direct anesthesia error or complications; the underlying surgical problem (such as sepsis following a perforated bowel); any preexisting concurrent medical condition (such as coronary artery disease); any preexisting evolving medical condition (such as acute respiratory distress syndrome in a patient in the intensive care unit); or any new problem (such as a fatal postoperative pulmonary embolus).

This expansive but nonspecific approach makes the meaningful analysis of death rates virtually impossible. However, this method of statistical analysis must be acknowledged because it is the basis of information that various government agencies, regulatory bodies, insurance providers, and media are trying to collect and publish. By doing so, they may promote competition among health care providers and facilities, in addition to providing public information based on which choices in health care can be made. The specific consideration of the role of anesthesia care in perioperative mortality is extremely difficult to derive, because there is no consistency in the available information to be examined.

▪ ANESTHETIC DEATHS

Some authors have accepted all operative and postoperative deaths as anesthetic deaths. Others have attempted to define anesthesia-related deaths or very limited examples of death or catastrophe caused solely by anesthesia care. Sometimes they refer specifically to an anesthesia accident, such as an unrecognized esophageal intubation that causes an immediate hypoxemic cardiac arrest and death, as opposed to an intraoperative myocardial infarction during an anesthetic that seemed to go well but caused death days later.

Definitions of Terms

Consideration of anesthesia perioperative mortality was undertaken in an extensive review by Lagasse that also incorporated original data from one hospital system on perioperative mortality.⁷ One conclusion was a challenge to the claim that anesthesia care is much safer now than in previous decades. This challenge was opposed in an accompanying editorial5 that noted it is “notoriously difficult” to assess the contribution of anesthesia care to perioperative mortality, and even more problematic to make comparisons across time periods using small sample sizes from widely differing institutions involving deaths that anesthesia variably “solely caused,” or was “related to, contributory to, or associated with.”

Lagasse, nonetheless, attempted to assemble virtually all the data in the literature through 1997. He did not include the landmark 1954 Beecher and Todd study,⁸ in which a death rate with anesthesia as “a very important contributing factor” of 1:1,560 was reported, because it was published before the beginning of the Medline database in 1966. The various mortality statistics from the 23 studies reviewed were presented as:

PERIOPERATIVE MORTALITY: 13 results, range 1:53 (1 death per 53 cases) to 1:5417
ANESTHESIA-RELATED MORTALITY: 22 results, range 1:388 to 1:85,708 to 0 deaths
ANESTHESIA SOLELY RESPONSIBLE MORTALITY: 7 results, range 1:9,090 to 1:200,200 to 0 deaths
PREVENTABLE ANESTHETIC MORTALITY: 6 results, range 1:1,707 to 1:48,748 to 0 deaths (the 0 deaths study was of 27,184 inpatients at four Canadian hospitals⁹)

The wide variability among these results clearly illustrates the difficulties in comparing disparate sources at different times using different definitions of mortality. This problem, of course, makes it difficult to identify trends and epidemiologic patterns. Lagasse added data from his own university hospital system for two time periods: 1992 to 1994 and 1995 to 1999. The perioperative mortality was 1:332 and 1:632, respectively. These figures may seem high but they included a mortality rate of 1:4.6 for American Society of Anesthesiologists (ASA) Physical Status (PS) class V patients. Using “death within 48 hours with anesthetist contribution (human error)” as the definition of anesthesia-related mortality, the mortality rates were 1:12,641 and 1:13,322, respectively, with 0 deaths in the anesthesia-solely-responsible category for either group. These data include patients of all ASA PS classes.⁷ The vast majority of perioperative deaths was in ASA PS class V, with a few in class IV, and very few in class III. On the explanatory graph, the anesthesia-related mortality did not register for ASA PS class I patients and was barely perceptible for PS class II patients. The rate appeared to be 1:6,883 for PS class III, 1:2005 for PS class IV, and 1:715 for PS class V patients.

In the discussion, Lagasse noted that, of the 14 reported anesthesia-related deaths, 4 “were the result of major contribution from the anesthesia personnel,” yielding a rate of 1:46,118 anesthetics. Of those four, only one was in ASA PS class I or II, yielding a mortality rate of 1:126,711 for these “healthy patients”; that death was not intraoperative and, therefore, would not have met the inclusion criteria in the Eichhorn study³ of intraoperative catastrophes in 1,001,000 healthy ASA PS classes I and II patients. Again, specifically noting that there was no death considered “due solely to anesthetic management” in the 184,472 anesthetics reported, Lagasse concluded: “Therefore, our results are consistent with the Eichhorn study, which may have been the basis of the IOM claim.” That very widely publicized 1999 report from the Institute of Medicine (IOM) heralded the improvement in anesthesia patient safety, citing a decrease in “anesthesia mortality rates” from approximately 1:5,000 anesthetics at the start of the 1980s to 1:200,000 to 300,000 (in the late 1990s).¹⁰

Data from Other Countries

Other relatively recent publications, including some from other countries, report values that are not too dissimilar, particularly because the rates of extremely rare events can sometimes be dramatically altered or even quite skewed by a single occurrence. A 2001 multicenter study of
869,483 patients of all ASA PS classes in the Netherlands identified 811 patients (1:1,072) who died or suffered brain damage within 24 hours of surgery.¹¹ One hundred and nineteen deaths were determined to be “anesthesiarelated” (1:7,306), and seven were “solely attributable to anesthesia,” for a rate of 1:124,000. A comprehensive Norwegian study¹² from a single hospital of 83,844 patients of all ASA PS classes over a 5-year period revealed 111 patients with “serious problems” for which anesthesia was the major contributing factor. Forty-two patients died intraoperatively (1:1,996), and all belonged to ASA PS class IV or V, except three with uncontrollable surgical bleeding. No intraoperative deaths solely related to anesthesia were reported. One patient had a cardiac arrest from an anaphylactic reaction to a muscle relaxant and later died in the intensive care unit (making the “anesthesia mortality” 1:83,844, if that case is counted as an anesthetic death).

A report of a huge Australian database from the mid-1990s¹³ cited “anesthesia-related deaths” at a rate of 1:62,500. A French report from the early 1990s,¹⁴ covering 101,769 anesthetics in patients of ASA PS classes I to IV (class V specifically excluded) began with an examination of intraoperative cardiac arrests. Of the 11 anesthesiarelated cardiac arrests, 3 were considered totally related to anesthesia, and 8 considered only partially related to anesthesia. Six of the 11 patients died (1 PS class II patient, 3 PS class III patients, and 2 PS class IV patients), leading to a cited “anesthesia-related mortality” of 6:101,769, or 1:16,961. If the one ASA class II patient had been removed, the rate for classes I and II would have been 1:101,769. A US study of the same type covered 72,959 anesthetics from 1989 to 1999 in a university hospital¹⁵ and focused on anesthesia-related cardiac arrests (extending to 24 hours postsurgery). The death rate from “anesthesia-attributable perioperative cardiac arrest” in the total study population was 1:18,181.

Nevertheless, one of the most quoted reports involving anesthesia mortality is the British National Confidential Enquiry into Perioperative Deaths,¹⁶ because it involved a systematic epidemiologic study of a large population. This database used a 30-day postoperative reporting period and a fairly broad set of criteria. Ultimately, an “anesthesiarelated mortality” of 1:1,351 was reported. An “anesthesia solely responsible” mortality rate of 1:185,056 was also reported, which was a remarkably low figure in 1987.

The Japanese Society of Anesthesiologists (JSA) Subcommittee on Surveillance of Anesthesia-Related Critical Incidents generated thorough statistics. From 1999 through 2002, among 3,855,384 anesthetics in JSA-certified training hospitals, 2,638 intraoperative deaths occurred (1:1,461), of which “anesthetic management was responsible” for 40, yielding an anesthesia intraoperative mortality rate of 1:96,384, which incorporated patients in all ASA PS classes.¹⁷ In the subset of ASA PS classes I and II patients, the anesthesia mortality rate was reported as 1:250,000 (within the range of Eichhorn’s reports³,⁴). A separate analysis (with the data collection period extending anesthesia deaths to 7 postoperative days) of the same database considered only 1,440,776 ASA PS class I patients having elective surgery.¹⁸ The death rate “totally attributable to anesthesia management” was 1:720,388 (two patients in that group suffered deaths for which anesthesia care was solely responsible).

Numbers Do Not Lie (Do They?)

Overall, these representative reports illustrate the scope of the type of information that is available in the literature regarding “anesthesia mortality.” Although some generalizations might be possible, they are flawed, as was previously noted, because of the wide variability in classifications, definitions, different time periods, locations, institutions, and patient populations.

One approach to this question, among many possibilities, is a rigorous attempt to isolate the anesthesia care component for consideration. This approach seeks to remove the influence of patients’ intercurrent medical conditions, the surgical issue provoking the administration of an anesthetic, and the surgical factors including misadventure during the surgery. This paradigm led to the well known (but often misquoted) statistics from the Harvard-associated teaching hospitals regarding only intraoperative catastrophes that were solely due to anesthesia care (cardiac arrest, permanent brain damage, and death).³ Of the 1,329,000 patients anesthetized from 1976 through June, 1988, 1,001,000 ASA PS classes I and II “healthy” patients were the subject of the report. A central component was the comparison of the rate of anesthesia-caused catastrophes before and after the implementation of the Harvard monitoring standards in July, 1985. Ten intraoperative catastrophes occurred out of 757,000 anesthetics (1:75,700), five of which resulted in death (1:151,400) before the implementation of the standards. In contrast, one catastrophe occurred without death in 244,000 anesthetics after implementation.

The aggregate number of five deaths solely related to anesthesia care in 1,001,000 cases was mentioned in the discussion of the report almost as an aside. Recall that the focus of the study was on the before-and-after comparison. The aggregate mortality rate for this population calculated over that period was 1:200,200. This citation in that context was offered originally as a complement to and confirmation of the British National Confidential Enquiry into Perioperative Deaths figure of 1:185,056 from 2 years earlier16 that had been questioned so vigorously. Thus was born (however unintended) the very widely quoted, improved “anesthesia mortality rate” of approximately 1:200,000 anesthetics.

Data on ASA PS classes I and II patients anesthetized at Harvard hospitals were collected for 2 more years and published⁴ as shown in Table 1.1. A principal focus of that citation was the continued, very busy, high-acuity anesthesia practices in nine diverse teaching hospitals without a single intraoperative anesthesia catastrophe that met the inclusion criteria for the study. In fact, the only poststandards, anesthesia catastrophe reported occurred in August 1985, meaning that by the end of June 1990, 467,000 consecutive anesthetics had been performed without an intraoperative catastrophe or death solely due to anesthesia care.

TABLE 1.1 Anesthesia Catastrophes and Deaths in Harvard-Affiliated Teaching Hospitals Among Healthy Patients Before and After Implementation of Monitoring Standards in July 1985

Only gold members can continue reading. Log In or Register to continue