896

Problems Inherent in Studying Anesthesia-Related Risk

Several issues related to studying anesthesia-related risk can affect the findings. For example, multiple definitions of perioperative mortality exist. In particular, the time frame in which a death can be attributed to the surgery and delivery of anesthesia varies. Because in-hospital stays have shortened, many events related to surgery may occur after discharge. This is particularly the case with ambulatory or short-stay surgery, for which recovery from anesthesia occurs at home. As we continue to monitor our patients for safety, it will be important to extend follow-up care beyond the traditional hospital setting, and many recent studies have included such an approach.

A major problem in any study of risk is the actual rate of complications in the population of interest. There are multiple sources of data for studying perioperative risk. Although some of the original studies used data from only one or a small group of institutions, such approaches may not be practical in the current era. As demonstrated throughout this chapter, most individuals believe that risk related to anesthesia has decreased over time, although this premise has been questioned.^[19] For example, the rate of anesthesia-related mortality described in the Confidential Enquiry into Perioperative Deaths (CEPOD) of 1987 was 1 in 185,000 patients, compared with the 1 in 2680 cases reported by Beecher and Todd about 30 years earlier.^{[20] [21]} Considering the current rate, any study would have to be enormous to detect anesthesia-related mortality. Such a study would require information from a large number of sites or data covering multiple years from a single institution.^[22] Considering concerns in the United States regarding confidentiality and legal liability, a multicenter study of risk in unselected patients would be difficult to undertake. Such problems may not be inherent in other countries.

Another issue is the effect of the act of studying outcome. As in any study of risk, the actual rate of complication decreases as a result of increased observation. This is frequently observed as a lower than expected rate of complications in the placebo arm of a randomized, controlled trial.

An alternative approach is to identify bad outcomes and study them for patterns of errors. For example, Cheney and colleagues^[23] developed the American Society of Anesthesiologists (ASA) Closed Claims Study (ACCS). By obtaining the records of major events that led to legal litigation, they were able to identify factors that contributed to bad outcomes. With this methodology, selected morbidities that lead to litigation can be identified. The limitation of this methodology is that the actual rates of complications are not known; only the number of closed legal claims is identified. Cases that do not result in litigation are not included in the database.

Several attempts have been made to establish large epidemiologic databases. One example of such an approach has been the work of Mangano and the Multicenter Study of Perioperative Ischemia group with regard to cardiac surgery. This group used their database to evaluate issues such as the rate and importance of atrial fibrillation after cardiac surgery and the use of aspirin.^{[24] [25]} Other approaches include the development of cardiac surgery databases by the Society of Thoracic Surgeons, the U.S. Veterans Administration, and the New England Collaborative Project.^{[26] [27] [28] [29] [30]} These databases are used to define risk factors for poor outcome, to benchmark local to national complication rates, and as educational tools. Such databases include little or no information regarding anesthesia practice. They include a predominance of patients from academic or major medical centers, whereas smaller or community hospitals may have very different patterns of care. Although these databases may provide extremely important information to improve care, the ability to generalize results to the nonacademic centers is unknown.

This final issue of the institutional site of the surgery may have other implications related to the pattern of care at a particular hospital. Although studies frequently categorize risk as related to anesthesia, surgery, or patient disease, postoperative care may have a profound impact on risk. For example, the risk of pulmonary embolism may be related to nursing care and the frequency of patient ambulation after surgery.^[31] The presence of an intensivist who makes daily rounds and higher nurse staffing ratios may also affect outcome.^{[32] [33] [34]} Variations in the total care provided by a hospital, including the anesthesiologists and surgeons, may significantly affect outcomes.

One advance in determining local complication rates is the development of hospital information systems. The use of computer databases in the assessment of risk began with Marx and colleagues^[35] in the 1970s. These information systems now are almost a requirement of survival for the hospital. The ability to query these systems will make large-scale studies of institutional risk much less burdensome in the future. Reich and coworkers^[36] used one such system to identify intraoperative hemodynamic abnormalities, including pulmonary hypertension, hypotension during cardiopulmonary bypass, and pulmonary diastolic hypertension after cardiopulmonary bypass, as independent predictors of mortality, stroke, and perioperative myocardial infarction beyond the effects of other preoperative risk factors.

When extremely large sample sizes are needed, administrative databases may be among the most cost-effective approaches to this issue. Examples of administrative databases include Medicare claims files, private insurance company claims, and hospital electronic records. These databases include a small number of data points for an extremely large number of subjects. For example, the Medicare database includes financial data, disease codes (International Classification of Diseases, 9th revision [ICD-9]), and procedure codes (Current Procedural Terminology [CPT]) for each patient. It also includes information regarding location of care and provider type. The Medicare claims file is being extensively used to provide benchmarks for rates of mortality and major complications after coronary bypass surgery.^[37] Hospitals can compare their rates with those of neighboring and competing hospitals and may use these data as markers for quality of hospital care.^{[38] [39]} Silber and colleagues^[40] took an alternative approach to the use of Medicare databases and evaluated perioperative factors that influence failure

897