This Article Abstract Full Text (PDF) CME: Take the course for this article: Course on Unanticipated Day of Surgery Deaths Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Bishop, M. J. Articles by Domino, K. B. Search for Related Content PubMed PubMed Citation Articles by Bishop, M. J. Articles by Domino, K. B. Related Collections Critical Care Complications Outcomes Patient Safety

---

PATIENT SAFETY

Factors Associated with Unanticipated Day of Surgery Deaths in Department of Veterans Affairs Hospitals

Michael J. Bishop, MD^*||, Jennifer E. Souders, MD, Cecilia M. Peterson, MSPH, William G. Henderson, PhD, and Karen B. Domino, MD, MPH^||

From the *Department of Veterans Affairs Central Office Anesthesia Service, Washington, DC; The Puget Sound Veterans Health Care System, Seattle, Washington; The Colorado Health Outcomes Program of the University of Colorado Health Sciences Center, Denver, Colorado; and ||The Department of Anesthesiology at the University of Washington School of Medicine, Seattle, Washington.

Address correspondence to Michael J. Bishop, MD, 1660 S Columbian Way, Seattle, WA 98108. Address e-mail to michael. bishop{at}va.gov.

Abstract

BACKGROUND: Patients of ASA physical status 1, 2, and 3 undergoing elective surgery do not have underlying conditions that are a constant threat to life, and hence should not be expected to be at significant risk for death on the day of surgery.

METHODS: We analyzed 815,077 ASA physical status 1, 2, and 3 elective surgery patients in the Department of Veterans Affairs National Surgical Quality Improvement Program database to identify patients who died on the day of surgery. We then attempted to identify factors predictive of unexpected death and to identify potential areas for improvement in care. A subset of the cases underwent individual chart review as well to identify areas for improvement in anesthesia care.

RESULTS: Of the total patients, 0.08% died on the day of surgery. The strongest predictive factor by multiple variable regression was the type of surgery, with aortic surgery resulting in an odds ratio of 13.67, (95% CI 9.76–19.17). Other factors predictive of death were identified by multiple variable regressions and included low albumin, existence of dyspnea, and elevated bilirubin or creatinine. Chart reviews of 88 of the deaths found that opportunities for improved anesthesia care were present in 13 of the 88. We estimated that a death that might have been prevented by improved anesthesia care occurred in approximately 1/13,900 cases. Myocardial infarction and hemorrhage were frequently identified factors. An unexpected factor was that the period between the conclusion of surgery and the final transfer of care in recovery was a time when many of the deaths occurred.

CONCLUSIONS: We conclude that, although patient and surgical factors lead to the vast majority of deaths on the day of surgery, there are identifiable areas for reducing the incidence of such deaths by improvements in anesthesia care.

Since 1991, the Department of Veteran Affairs has monitored risk factors and 30-day mortality and morbidity for a selected portion of cases to assess surgical outcomes in its hospitals.1 Cases monitored included a sample of up to 36 consecutive eligible cases per hospital within an 8-day cycle with each cycle starting on a different day of the week, including only cases undergoing general or neuraxial anesthesia, as well as all carotid endarterectomies and hernia surgeries. Within the Department of Veterans Affairs hospital system, overall mortality for noncardiac surgery done under general or neuraxial anesthesia has declined from 3.1% to 2.1% during the years 1991 to 2003.2 This represents the 30-day mortality for all cases from all causes. However, the program, now entitled the National Surgical Quality Improvement Program (NSQIP), has not previously focused on deaths on the day of surgery. We sought to identify mortality that might have been avoided by improvements in anesthesia care.

In order to identify deaths that might be prevented by better perioperative anesthesia care, we studied the deaths occurring on the day of surgery in relatively low risk patients: elective noncardiac surgery during the years 1992–2001 for ASA physical status 1–3 individuals. We also performed chart reviews on a subset of deaths from that group.

We hypothesized that, by studying deaths on the day of surgery in patients without life-threatening comorbidity, we would find a high proportion of cases with identifiable opportunities for improvement in anesthesia care. We also hypothesized that use of the extensive database concerning preoperative risk factors would help to identify the patients at greatest risk for unexpected death.

METHODS

NSQIP Database
The NSQIP database prospectively collects preoperative, intraoperative, and postoperative information on major operations within the Department of Veterans Affairs health care system.3,4 Major operations are defined as surgery performed under general anesthesia or neuraxial block. Carotid endarterectomies and inguinal herniorraphies are included, even when performed under local anesthesia. Cardiac surgical procedures are excluded from this database. The data are collected by specially trained nurse reviewers, and the data are transmitted to a centralized Veterans Administration data coordinating center. The data elements have been previously described.3,4 The data analyzed for this study were collected for the years 1992–2001. The outcome assessed was death on the day of the surgery.

We used this prospectively collected database of the NSQIP to identify the factors most likely to be associated with perioperative death. Data on some variables (the presence of angina or a recent myocardial infarction, hypertension, esophageal varices, history of peripheral vascular disease, rest pain/gangrene, and absent peripheral pulses) were collected only from 1992 to 1995 resulting in a smaller sample size for those variables. These were not included in the multiple variable regression analyses. The period 1992–1995 represented the pilot phase of NSQIP and, during the next stage, variables that did not reach statistical significance for the overall mortality level were removed from the collection protocol (although some have since been reinstated for use in subanalyses such as the current analysis). A few other variables (chemotherapy, radiotherapy, or presence of sepsis) were added to the database later in the NSQIP and, thus, also have reduced sample sizes. Most variables were very complete, except for the preoperative laboratory values, which were sometimes not ordered. Missing laboratory values were determined according to the methods of Buck5 and using a SAS macro developed by Roberts and Capalbo.6 This macro uses complete patient data to develop regression equations of each variable or combinations of variables that have missing values regressed on the other variables in the dataset. Once these regression equations are developed, the variables that have complete data for a given patient are input into the equations to impute the values of the variables with missing data for that patient.

In-depth chart reviews for deaths were performed to assess specific causes of death and to identify opportunities for improvement in anesthetic care. Charts were requested for all deaths of elective surgeries that met the NSQIP guidelines above and were ASA physical status 1–3 for the years 1995–2001. The focused reviews were conducted for these years because of difficulty retrieving original charts for earlier years. The instrument for chart reviews was adapted from the data instruments of the American Society of Anesthesiologists Closed Claims Project and the Pediatric Outcomes of Cardiac Arrest registry using techniques developed as part of the Closed Claims Project.7,8 Each of the charts was reviewed by two of the three anesthesiologist reviewers (KBD, MJB, and JES).

Each reviewer classified the contribution of anesthesia, surgery, or patient's underlying condition to the patient's death. Anesthesia care was then classified as less than appropriate, appropriate, or impossible to judge. Finally, each reviewer answered the question of whether any of three changes in care would have prevented the death: better preoperative evaluation, better preoperative preparation, or better monitoring. For each of the three choices, the reviewer answered either no, yes, or impossible to judge. After the second review of the chart, the two reviewers discussed discrepancies to reach an agreement on whether anesthesia played a major role. In two cases in which the two reviewers did not fully agree about whether anesthesia played a role, the third reviewer also reviewed the chart to reach a consensus.

Statistical Methods
We compared day-of-surgery death rates for patient baseline characteristics (demographic variables, preoperative comorbidities, and preoperative laboratory values), operative variables, and postoperative complications using the ² test for categorical variables and the t-test for independent samples for continuous variables.

A step-up multiple logistic regression analysis to identify predictors of day-of-surgery and 30-day postoperative death was used, with criteria for entry into or exit from the model of 0.05. Two models were developed, one for day-of-surgery death and another for 30-day death. Forty-five preoperative variables were allowed to be included in the models. Most of the variables were dichotomous and coded as 0 = absence of the comorbidity and 1 = presence of the comorbidity. A few variables (ASA class, diabetes, functional status, dyspnea, and category of surgery) were polychotomous, and were entered as class variables. Two variables (age and serum albumin level) were entered as continuous variables. Since there were 646 day-of-surgery deaths and 10,201 30-day deaths, the rule of thumb of 10 or more events per independent variable was satisfied to prevent overfitting of the models. Interactions, tests of linearity or co-linearity, and validation through split-sampling or bootstrapping were not done since the purpose of the analysis was just to get an idea of the important preoperative predictors of postoperative death, and not to develop a prediction equation. Validation of NSQIP models has been done in previous studies.1,9 Discrimination and goodness of fit were assessed by the c-index and the Hosmer-Lemeshow test.

RESULTS

Of 815,077 cases in the National Surgical Quality Improvement database as of September 30, 2001, 646 (0.08%) resulted in death on the day of surgery.

Comparison of Death Rate for Patient Characteristics
Table 1 compares postoperative death rates for patient demographic characteristics. Male patients had a higher death rate compared to female patients (0.08% vs 0.03%, P = 0.0002). Patients who died on the day of surgery had a mean age 7 yr older than the survivors (67.6 vs 60.0, P < 0.0001, Table 2—continuous variables).

Table 3 compares death rates for patients with and without preoperative comorbidities. Patients who were in a coma or were ventilator-dependent had the most extreme and statistically significant differences in the death rate between patients with and without these conditions. However, these cases were very few in number and would not have been classified as ASA 1, 2, or 3 by any of the authors. Patients with the following conditions also had large (threefold to sixfold) and statistically significant differences between those with and without the conditions: ASA class 3; dependent functional status; do not resuscitate order; history of congestive heart failure, myocardial infarction, or angina; impaired sensorium; ascites; dyspnea; pneumonia; acute renal failure or undergoing dialysis; disseminated cancer; weight loss >10%; bleeding disorder; preoperative transfusion; or preoperative sepsis.

Table 5 compares the postoperative death rates for variables related to the surgery itself. Comparing different types of operations, aortic, brain, and thoracic operations had the highest rates (Fig. 1), although most deaths occurred in bone and abdominal cases due to their greater frequency in the sample (Fig. 2). Aortic surgery had the highest likelihood of death on the day of surgery relative to the likelihood of death by day 30 (Fig. 3), with more than 30% of the deaths within 30 days actually occurring on the first day.

View larger version (9K): [in this window] [in a new window]   Figure 1. Risk of death per 10,000 cases by type of surgery. Carotid and infra-inguinal vascular procedures are included in "Other vasc.".

View larger version (9K): [in this window] [in a new window]   Figure 2. Total numbers of deaths by type of surgery.

View larger version (9K): [in this window] [in a new window]   Figure 3. The relative risk of dying on day 1 as a fraction of the total deaths in the first 30 days after surgery.

Table 2 presents data for the continuous variables. Patients who died on the day of surgery had older mean age, longer operative times, lower preoperative albumin levels, and higher number of pack-years of tobacco smoking.

Finally, Table 6 presents death rates in patients who experienced or did not experience certain postoperative complications. Patients who experienced one or more postoperative complications had a much higher death rate compared to those who did not (0.73% vs 0.02%, P < 0.0001). Postoperative complications that were most highly associated with death included cardiac arrest, myocardial infarction, coma, unplanned intubation, pulmonary embolism (PE), acute renal failure, and postoperative bleeding. Postoperative complications that were not associated with higher death rates included pneumonia, progressive renal insufficiency, urinary tract infection, superficial and deep wound infection, dehiscence, and prolonged ileus.

Predictive Model of Postoperative Death
Table 7 presents the multivariable predictive model for postoperative death. The table gives the step of the logistic regression analysis at which the variable entered the model, the odds ratio for the variable, and the 95% confidence interval for the odds ratio. The most important predictor variable was type of surgery. Aortic (OR = 13.67), brain (OR = 5.91), and thoracic (OR = 3.62) operations had the highest risk of mortality on the day of surgery. Other preoperative variables that were strong predictors of death included serum albumin, dyspnea, bilirubin, creatinine, age, BUN, and ASA class. There were 17 independent predictors in the model at the 0.05 level of statistical significance. The model had a moderate c-index of 0.690. The Hosmer-Lemeshow goodness-of-fit test had a P value of 0.21, indicating good calibration of the model. Also presented in Table 7 is the prediction model of 30-day postoperative death, which has been used for more than a decade in the NSQIP for quality improvement assessment. All 17 variables in the prediction model for perioperative death are in the prediction model for 30-day death, but in different rank orders. Six variables are among the top 10 variables in both models: type of surgery, albumin, age, BUN, ASA class, and weight loss >10%. Dyspnea, bilirubin, and creatinine tend to be more important in predicting perioperative death compared to 30-day death, while disseminated cancer, functional status, ascites, and high white blood cells tend to be more important in predicting 30-day death compared to perioperative death. The c-index for the model predicting 30-day death is higher (0.860) than the c-index for predicting perioperative death (0.690).

Chart Reviews
Medical records of the 191 patients who died in the period occurring during the years 1995–2001 were requested from 55 hospitals. Ninety-three charts were sent from 27 hospitals, but 5 of those charts contained insufficient information to assess the causes of death. Twenty-eight hospitals did not respond to the request. ASA classification was ASA 2 for 27 of the charts (29%) and ASA 3 for 66 (71%). There were no day of surgery deaths in ASA 1 patients during the 6 yr selected for chart reviews, and only 3 such deaths during the 10 yr span analyzed statistically. Day of surgery deaths of ASA 1 patients accounted for <0.5% of acute deaths and occurred approximately once per 13,000 cases during the duration of the study.

A significant role for improvement in anesthetic care was identified in 13 of 88 deaths, and a possible role for improvement in another 14 of the cases, for a total of 31% of deaths for which improved care might have affected outcome. In 61 deaths, the reviewers felt that changes in anesthesia care were unlikely to have made a difference in outcome. The kappa statistic for agreement between reviewers as to whether anesthesia played a role in the death was 0.84 (CI, 0.74–0.94).

Uncontrolled surgical hemorrhage was the leading cause of death, occurring in 21 patients. The leading causes of nonsurgical death were acute myocardial infarction in 20 patients and presumed PE in 12 patients (one of the myocardial infarctions was subsequent to massive hemorrhage). Fourteen percent of the deaths reviewed occurred during or after surgery for an abdominal aortic aneurysm.

Sixteen of the patient deaths occurred during the interval between the end of surgery and the first 30 min in the recovery room or intensive care unit (ICU). Of these deaths, eight were thought likely to have resulted from hypovolemia despite an absence of massive hemorrhage, with severe hypotension or arrest on turning or moving the patient. Four patients had respiratory deaths in this period, including loss of airway, a respiratory arrest during transport to the ICU, a presumed PE in a patient who was several days out from a lower leg fracture, and a case of severe bronchospasm.

The events that led to deaths for which anesthesia care appeared to play a major role are listed in Table 8. One of these cases was a tension pneumothorax from a subclavian line that was actually placed by a surgeon in the operating room (OR) at the end of a case. Since this was a preventable death that would often be part of the perioperative anesthetic care, and monitoring was provided by the anesthesia provider, it was included with the anesthesia-related deaths. Similarly, deaths related to apparent respiratory arrest in perioperative patients were included as a death where perioperative anesthesia care played a role because of the possible relationship to narcotic analgesia. These two deaths included one patient receiving epidural morphine postoperatively and one patient using patient-controlled analgesia.

DISCUSSION

The significant findings of this study include: 1) The type of surgical case is the major predictor of unexpected death; 2) ASA physical status is highly predictive of unexpected death; 3) The immediate postsurgery period (transport and arrival in recovery or ICU) tends to be a time of unexpected deaths; and 4) Myocardial infarction is a significant cause of unexpected death.

In support of our hypothesis, we did find that changes in perioperative management by the anesthesia providers might have improved outcome in 31% of deaths. However, contrary to our hypothesis, we found almost no incidents of unequivocal anesthetic mishaps. Clear-cut proximal causes for the deaths were uncommon, with the most obvious one being an unrecognized tension pneumothorax. A retrospective analysis of closed anesthesiology-related claims published in 1991 found a significant number of adverse outcomes due to acute respiratory events, such as a loss of a patent airway for mask ventilation or inability to intubate the trachea.10 A more recent publication from the same database found that such respiratory accidents have declined and are now similar in magnitude to cardiovascular events.11 Fortunately, we found that anesthetic accidents resulting in acute death to be extremely rare with just one such incident in 815,000 cases. Our finding that two deaths occurred from respiratory arrest while patients were receiving narcotics via an epidural or via patient-controlled analgesia corresponds with the report of an increasing number of such adverse outcomes noted by the American Society of Anesthesiologists Closed Claims Project.12

From the large database portion of this study, we conclude that the major risk factor that predicted death in patients undergoing noncardiac surgery was undergoing aortic surgery. A recent study of predictors of postoperative cardiac adverse events after general and vascular surgery similarly found that operation type and urgency and ASA physical status classification were important predictors of cardiac events within the first 30 days but did not identify how often these deaths occurred in the immediate perioperative period.13 In that study, cardiac events accounted for 30% of the deaths within 30 days and accounted for 23% of the deaths in our chart review study. Although the surgery to be performed may not be a modifiable risk factor, minimization of cardiac risk certainly seems to be a focus for potential improvement.

Although minimizing cardiac risk is, of course, a laudable goal, the mechanisms to do so often remain elusive. A survey of experts presented with hypothetical cases of patients with coronary artery disease for elective vascular surgery found that opinions on whether to intervene with revascularization preoperatively remain controversial.14 However, randomized studies involving coronary revascularization do not support the idea that outcomes are improved by such preoperative intervention.15,16 Similarly, the efficacy of initiating β-blockade for the control of heart rate and prevention of ischemia remains highly controversial.17 Our study provides reasons for continued investigations for preventing myocardial ischemia and death in the perioperative period, and it is beyond the scope of this paper to provide recommendations in this area.

The predictive value of BUN for death on the day of surgery was surprisingly greater than its predictive value for 30-day death overall. Sudden cardiac death is the most common cause of mortality in patients with chronic renal failure and may often occur shortly after dialysis.18 QT intervals have been shown to vary greatly in the peridialysis period, potentially contributing to ventricular arrhythmias. Although most practitioners agree that dialysis before surgery is desirable to minimize fluid overload and hyperkalemia, immediate preoperative hemodialysis may itself be a risk. Thus, while this greatly increased mortality is obviously of concern, this retrospective look cannot define the best measures to take to decrease this risk.

There are several limitations of our study. A patient enters the NSQIP database only if a surgery has actually been performed. Hence, anesthetic disasters at the time of induction that resulted in death may have been missed if the surgery itself was cancelled. Adverse anesthetic outcomes also could have been missed if they resulted in bad outcomes beyond the operative day. For example, a difficult intubation that resulted in prolonged hypoxia might have been recognized in time that the patient survived a short time but eventually died or had significant long-term damage. Since we selected our cases based on death on the day of surgery, we would not have reviewed such cases. A prior study of deaths in the perioperative period found that the proportion of patients who remained comatose was small compared to those who died (16 times as many patients died as remained comatose), but we have no way to know the proportion in our database.19 As noted in the Methods section, we did not have data on some variables after 1995, most notably for the presence of angina or a recent myocardial infarction. These variables were deleted from NSQIP collection at that time because they did not have a large impact on the 30-day model for mortality for all patients, which is the primary monitoring role for NSQIP. However, collection of these variables has recently been reinstituted with the thought they may be more significant in certain subpopulations or in analysis of outcomes other than 30-day deaths.

We did not do any validation of our models reported in Table 7. However, similar NSQIP models have been validated using split-sample methods in previous publications and have been demonstrated to have good validation.1,9 For the all operations, the NSQIP predictive model for mortality, split-sample methods found that the average degradation in the c-index from the learning to test samples was 0.003 for a c-index of 0.89.1

The decision to focus on ASA physical status 1, 2, and 3 patients was made on the basis that ASA physical status 4 patients had a preexisting condition that made death a much more likely event and that anesthetic mishaps would also be potentially more easily identifiable with fewer underlying conditions. However, the assignment of ASA status was made by the individual practitioners at the time of surgery and could have varied substantially among providers and across medical centers. As an example, one might challenge whether patients with known renal failure should ever have been classified as ASA 3. The presence of a few patients in coma or receiving mechanical ventilation surprised the authors, none of whom would consider such patients ASA 1, 2, or 3. However, such cases were few in number (patients with preexisting coma accounted for <1% of the deaths). Our best guess is that these cases may have been patients who were ASA 1, 2 or 3 upon admission to the hospital, but before the actual surgery, and were not reclassified on the day of surgery.

Another major limitation of our study is that, of the 191 charts we requested for review, we were able to obtain only 93 and 5 of those contained insufficient data for review. We considered the possibility that charts were selectively withheld. Charts are stored in warehouses off-site after death of the patient, and we had to rely on the good will of the local medical records departments in many hospitals to locate the chart and send it to us for review. We made the assumption that the 88 charts actually reviewed did not differ from the charts not received. However, there did not seem to be any selective withholding of charts; facilities either sent all of their charts or did not respond.

Our definition of potential for improved anesthetic management was also a very broad, and depended on the agreement of two anesthesiologists from a tertiary care medical center. The time period extended from preoperative evaluation through the end of anesthesia care in the recovery area. Thus, a patient with known coronary disease undergoing a vascular procedure and not receiving β-blockers preoperatively was considered a minor opportunity for improvement. This remains controversial, and one could certainly argue that this criterion should not have been included.20,21

A major limitation of peer review of appropriateness of care includes bias based on knowledge of the outcome and the relatively poor agreement between two reviewers in most studies.22,23 Our high degree of agreement results, in part, from the fact that in the majority of cases, it was easy to identify that there was no opportunity for improvement in anesthetic care. Reviewer agreement has been shown to be improved by discussion among reviewers, a practice which we incorporated into our reviews.24 We also brought in a third reviewer in cases of disagreement. Furthermore, unlike a malpractice case, in which a "yes or no" answer might be required, we were seeking a sense of whether we were faced with a clear mishap versus a case in which there was a possibility that anesthesia care might have improved outcome. We were primarily looking for the clear mishaps and beyond that areas in which we might focus improvement activities, and thus, could tolerate the lack of a definitive answer.

Our study differed from prior studies of perioperative death because it focused on patients not thought to have life-threatening conditions (ASA 4 and 5 patients). A study by Newland et al. of cardiac arrest in the OR found that 68% of arrests were ASA 4 patients, who were excluded in our study.25 In their study, as in ours, the adverse events clearly attributable to anesthesia management were few, but the number for which anesthesia care was contributory was noted to be about twice the number for which anesthesia was the primary cause. However, as Lagasse in an editorial in Anesthesiology pointed out, perioperative death (within 2 days in his analysis) usually does not involve intraoperative cardiac arrest.26 This agrees with our case reviews, which found that the period immediately after transfer out of the OR was a very high-risk period. Much more important for improving care may be "failure-to-rescue" (deaths that occur after a complication) and ultimately may provide the best opportunities for improving care.27 Reviews of ASA 4 patients are more likely to produce such events, but the role of the patient disease versus appropriateness of care is likely to be harder to define. However, anesthetic management errors may have more profound effects in patients with greater comorbidity. Our chart reviews did not include such patients, and we acknowledge that the importance of anesthesia care in unanticipated deaths in such patients could actually be greater than in healthier patients. The exclusion of ASA 4 and 5 patients from the database analysis also may have biased the results for predictive factors for unanticipated deaths, although ASA 5 deaths probably can never be classified as "unanticipated." Thus, these results can only be applied to ASA 1, 2, and 3 patients.

Based on the chart review, we calculated that cases in which improved care was warranted occurred in 1 of 28,900 cases and that, in 1 of 13,900 cases, there was possible or definite opportunities for improved care that might have avoided a death. This is very close to the figure of 1:13,000 suggested by Lagasse in a comprehensive review of anesthesia safety and of his own data from Montefiore Hospital.28 His data included all cases, not just ASA 1–3. We agree with Lagasse that clear cases of mortality due to anesthesia mishaps are uncommon but that opportunities for limiting mortality by improved care appear to be far more frequent.

Perhaps the most unexpected finding was that patients are at high risk during the period of transfer of care from the OR to the postoperative arena. We suspect this may be a period of diminished vigilance and plan to use these data to emphasize the need for close monitoring during transfer of patients who undergo surgery requiring significant fluid administration. Our study also supports the continuing need for developing evidence-based strategies for minimizing perioperative myocardial infarction. This continues to be an area of active investigation with studies of the utility of perioperative markers of myocardial ischemia an area of current interest.29

Footnotes

Accepted for publication June 17, 2008.

Supported by the Epidemiology Research and Information Center of the Department of Veterans Affairs.

Reprints will not be available from the author.

REFERENCES

1. Khuri SF, Daley J, Henderson W, Hur K, Gibbs JO, Barbour G, Demakis J, Irvin G III, Stremple JF, Grover F, McDonald G, Passaro E Jr, Fabri PJ, Spencer J, Hammermeister K, Aust JB. Risk adjustment of the postoperative mortality rate for the comparative assessment of the quality of surgical care: results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg 1997;185:315–27[Web of Science][Medline]
2. Henderson WG, Khuri SF, Mosca C, Fink AS, Hutter MM, Neumayer LA. Comparison of Risk-adjusted 30-day postoperative mortality and morbidity in Department of Veterans Affairs hospitals and selected university medical centers: general surgical operations in men. J Am Coll Surg 2007;204:1103–14[Web of Science][Medline]
3. Khuri SF, Daley J, Henderson W, Barbour G, Lowry P, Irvin G, Gibbs J, Grover F, Hammermeister K, Stremple JF. The National Veterans Administration Surgical Risk Study: risk adjustment for the comparative assessment of the quality of surgical care. J Am Coll Surg 1995;180:519–31[Web of Science][Medline]
4. Khuri SF, Daley J, Henderson W, Hur K, Demakis J, Aust JB, Chong V, Fabri PJ, Gibbs JO, Grover F, Hammermeister K, Irvin G, McDonald G, Passaro E, Phillips L, Scamman F, Spencer J, Stremple JF. The Department of Veterans Affairs' NSQIP: the first national, validated, outcome-based, risk-adjusted, and peer-controlled program for the measurement and enhancement of the quality of surgical care. National VA Surgical Quality Improvement Program. Ann Surg 1998;228:491–507[Web of Science][Medline]
5. Buck SF. A method of estimation of missing values in multivariate data suitable for use with an electronic computer. J R Stat Soc [Ser B.22] 1960;2:302–6
6. Roberts JS, Capalbo GM. A SAS macro for estimating missing values of multivariate data. In: SAS Users Group International Twelfth Annual Conference Proceedings. Dallas, Texas, 1987; 8–11
7. Caplan RA, Posner K, Ward RJ, Cheney FW. Peer reviewer agreement for major anesthetic mishaps. QRB Qual Rev Bull 1988;14:363–8[Medline]
8. Morray JP, Geiduschek JM, Ramamoorthy C, Haberkern CM, Hackel A, Caplan RA, Domino KB, Posner K, Cheney FW. Anesthesia-related cardiac arrest in children: initial findings of the Pediatric Perioperative Cardiac Arrest (POCA) Registry. Anesthesiology 2000;93:6–14[Web of Science][Medline]
9. Daley J, Khuri SF, Henderson W, Hur K, Gibbs JO, Barbour G, Demakis J, Irvin G III, Stremple JF, Grover F, McDonald G, Passaro E Jr, Fabri PJ, Spencer J, Hammermeister K, Aust JB, Oprian C. Risk adjustment of the postoperative morbidity rate for the comparative assessment of the quality of surgical care: results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg 1997;185:328–40[Web of Science][Medline]
10. Cheney FW, Posner KL, Caplan RA. Adverse respiratory events infrequently leading to malpractice suits. A closed claims analysis. Anesthesiology 1991;75:932–9[Web of Science][Medline]
11. Cheney FW, Posner KL, Lee LA, Caplan RA, Domino KB. Trends in anesthesia-related death and brain damage: a closed claims analysis. Anesthesiology 2006;105:1081–6[Web of Science][Medline]
12. Bird M, Caplan RA, Lee LA, Stephens LS, Domino KB. Liability associated with acute pain management. Anesthesiology 2007;107:A1743 (abstract)
13. Davenport DL, Ferraris VA, Hosokawa P, Henderson WG, Khuri SF, Mentzer RM. Multivariable Predictors of postoperative cardiac adverse events after general and vascular surgery: results from the patient safety in surgery study. J Am Coll Surg 2007;204:1199–210[Web of Science][Medline]
14. Pierpont GL, Moritz TE, Goldman S, Krupski WC, Littooy F, Ward HB, McFalls EO. Disparate opinions regarding indications for coronary artery revascularization before elective vascular surgery. Current opinion on revascularization study investigators. Am J Cardiol 2004;94:1124–8[Web of Science][Medline]
15. Poldermans D, Schouten O, Vidakovic R, Bax JJ, Thomson IR, Hoeks SE, Feringa HH, Dunkelgrun M, de Jaegere P, Maat A, van Sambeek MR, Kertai MD, Boersma E. DECREASE Study Group A clinical randomized trial to evaluate the safety of a noninvasive approach in high-risk patients undergoing major vascular surgery: the DECREASE-V Pilot Study. J Am Coll Cardiol 2007;49:1763–9[Abstract/Free Full Text]
16. McFalls EO, Ward HB, Krupski WC, Goldman S, Littooy F, Eagle K, Nyman JA, Moritz T, McNabb S, Henderson WG. Veterans Affairs Cooperative Study Group on Coronary Artery Revascularization for Elective Surgery Prophylactic coronary artery revascularization for elective vascular surgery: study design. Control Clin Trials 1999;20: 297–308[Web of Science][Medline]
17. Biccard BM, Sear JW, Foëx P. Meta-analysis of the effect of heart rate achieved by perioperative beta-adrenergic blockade on cardiovascular outcomes. Br J Anaesth 2008;100:23–8[Abstract/Free Full Text]
18. Morris ST, Galiatsou E, Stewart GA, Rodger RS, Jardine AG. The Renal Unit and Department of Med and Therapeutics, Western Infirmary. QT dispersion before and after hemodialysis. J Am Soc Nephrol 1999;10:160–3[Abstract/Free Full Text]
19. Arbous MS, Grobbee DE, Van Kleef JW, deLange JJ, Spoormans HH, Touw P, Werner FM, Meursing AE. Mortality associated with anaesthesia: a qualitative analysis to identify risk factors. Anaesthesia 2001;56:1141–53[Web of Science][Medline]
20. Fleisher LA, Beckman JA, Brown KA, Calkins H, Chaikof EL, Fleischmann KE, Freeman WK, Froehlich JB, Kasper EK, Kersten JR, Riegel B, Robb JF; American College of Cardiology, American Heart Association Task Force on Practice Guidelines; Writing Committee to Update the 2002 Guidelines on Perioperative Cardiovascular Evaluation for Noncardiac Surgery; American Society of Echocardiography; American Society of Nuclear Cardiology; Heart Rhythm Society; Society of Cardiovascular Anesthesiologists; Society for Cardiovascular Angiography and Interventions; Society for Vascular Med and Biology. ACC/AHA 2006 noncardiac surgery: focused update on perioperative beta-blocker therapy—are part of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to update the 2002 Guidelines on Perioperative Cardiovascular Evaluation for Noncardiac Surgery). Anesth Analg 2007;104:15–26[Abstract/Free Full Text]
21. Flood C, Fleisher LA. Preparation of the cardiac patient for noncardiac surgery. Am Fam Physician 2007;75:656–65[Medline]
22. Caplan RA, Posner KL, Cheney FW. Effect of outcome on physician judgments of appropriateness of care. JAMA 1991;265:1957–60[Abstract/Free Full Text]
23. Posner KL, Caplan RA, Cheney FW. Variation in expert opinion in medical malpractice review. Anesthesiology 1996;85:1049–54[Web of Science][Medline]
24. Levine RD, Sugarman M, Schiller W, Weinshel S, Lehning EJ, Lagasse RS. The effect of group discussion on interrater reliability of structured peer review. Anesthesiology 1998;89:507–15[Medline]
25. Newland MC, Ellis SJ, Lydiatt CA, Peters KR, Tinker JH, Romberger DJ, Ullrich FA, Anderson JR. Anesthetic-related cardiac arrest and its mortality: a report covering anesthetics over 10 years from a US teaching hospital. Anesthesiology 2002;97:108–15[Web of Science][Medline]
26. Lagasse RS. Apples and oranges: the fruits of labor in anesthesia care. Anesthesiology 2003;99:248–9[Medline]
27. Silber JH, Kennedy SK, Even-Shoshan O, Chen W, Mosher RE, Showan AM, Longnecker DE. Anesthesiologist board certification and patient outcomes. Anesthesiology 2002;96:1044–52[Web of Science][Medline]
28. Lagasse RS, Anesthesia safety: model or myth? Anesthesiology 2002;97:1335–7[Web of Science][Medline]
29. Fleisher LA. Strategies to reduce cardiac risk in noncardiac surgery: where are we in 2005? Anesthesiology 2005;102:881–2[Web of Science][Medline]

This article has been cited by other articles:

				R. S. Lagasse The Right Stuff: Veterans Affairs National Surgical Quality Improvement Project Anesth. Analg., December 1, 2008; 107(6): 1772 - 1774. [Full Text] [PDF]

				N. L. Pace Independent Predictors from Stepwise Logistic Regression May Be Nothing More than Publishable P Values Anesth. Analg., December 1, 2008; 107(6): 1775 - 1778. [Full Text] [PDF]

				M. J. Bishop, W. G. Henderson, and K. B. Domino Regression Analysis for a Large Database Anesth. Analg., December 1, 2008; 107(6): 2090 - 2090. [Full Text] [PDF]

This Article Abstract Full Text (PDF) CME: Take the course for this article: Course on Unanticipated Day of Surgery Deaths Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Bishop, M. J. Articles by Domino, K. B. Search for Related Content PubMed PubMed Citation Articles by Bishop, M. J. Articles by Domino, K. B. Related Collections Critical Care Complications Outcomes Patient Safety