Quality improvement in cardiac care has made considerable progress over the past 30 years. During that period, there has been the development of multi-institutional databases to monitor outcomes following cardiothoracic surgery. These databases initially began using only volume and unadjusted operative (30-day) mortality as outcome criteria. There has been a progressive increase in their sophistication, with the building of risk models based on preoperative variables, which accurately predict the risk of adverse outcomes. Other outcomes have been added including risk-adjusted mortality and morbidity; efficiency outcomes such as length of stay, quality of life, functional health status, neuropsychological outcomes; and long-term outcomes.
The multi-institutional databases that have been leaders in this effort have been the Department of Veterans Affairs (VA) National Cardiac database monitored by the VA Cardiac Surgery Consultants Committee, the Society of Thoracic Surgeons (STS) National Database, the Northern New England (NNE) Database, and the New York State Database. The challenges have been implementing the use of data to improve processes of care and, thus, improve outcomes. This is frequently done at the state level and may involve round-robin visits with the learning of "best practices." All of these databases have demonstrated a statistically significant reduction in risk-adjusted mortality over the past decade. This has occurred in spite of patients being older and of higher risk. Continued challenges include measuring of other outcomes, continuing this trend in improvement, and the costs associated with these databases. Others areas of monitoring in cardiothoracic surgery include congenital heart disease and general thoracic surgery. The models used for quality improvement in cardiac surgery should be easily adaptable for other areas of surgery for both monitoring of quality and improvement of care.
For more than 30 years, individual cardiothoracic surgeons, groups, and hospitals have monitored their volume and mortality statistics in an effort to continuously improve the quality of care that they render. The first large, multi-institutional monitoring of cardiac surgery quality began with the VA in 1971 with the establishment of the Cardiac Surgery Advisory Group. The VA and Congress wanted to have continual surveillance of the quality of care of cardiac surgery in the VA system and used volume and unadjusted operative mortality as measures of that quality. Takaro et al1 published an analysis of the VA experience from 1975 to 1984 and noted that the annual volume of cardiopulmonary bypass surgery increased from 3074 in 1975 to 6455 in 1984. During that period operative mortality declined from 8.3% to 4.7%. Mortality for valve procedures declined from 10.9% to 5.9% and for coronary bypass procedures from 4.7% to 3.6%. They noted that the extent of the patients' disease accounted for most of the operative mortality, but problems related to the adequacy of myocardial protection and to surgical technique were also important factors.1
On March 12, 1986, the Department of Health and Human Services Health Care Financing Administration released a list of hospitals that had mortality rates for Medicare patients that were in excess of predicted mortality rates for those hospitals. They were forced to release these data because of requests through the Freedom of Information Act, although these data were originally to be used for state peer review organizations as a quality improvement tool. These mortality rates were only grossly adjusted for patient risk factors, particularly compared with present day standards. In addition, some diagnostic categories were grouped together without appropriate adjustment. The release of these data prompted the appointment of an Ad Hoc Committee on Risk Factors for Coronary Artery Bypass Surgery by the STS. This committee strongly advocated sophisticated risk-adjusted methods, noting the experience and methods of the Collaborative Study in Coronary Artery Surgery.2 The action by the Health Care Financing Administration served as a stimulus for the STS to initiate its own cardiac surgical database with risk-adjusted methods.
As noted, the Collaborative Study in Coronary Artery Surgery developed a mutlivariate discriminant analysis of both the clinical and angiographic predictors of operative mortality for coronary artery surgery.3 In 1987, Sethi et al4 reported on the clinical, hemodynamic, and angiographic predictors of operative mortality in patients undergoing single valve replacement based on the VA Cooperative Study on Valvular Heart Disease.
In 1987, the VA Cardiac Surgery Advisory Group, now called the VA Cardiac Surgery Consultants Committee, was also concerned with the use of unadjusted operative mortality and volume as major quality indicators for the approximately 45 VA cardiac surgery programs. Using the logistic regression method developed by the Collaborative Study in Coronary Artery Surgery investigators, under the direction of Karl E. Hammermeister, MD, the VA developed a risk model for coronary bypass and for valve and other surgical procedures. The first report of these risk variables was presented to the STS in 1989.5 The method included capturing data on all cardiac surgery procedures performed using a single-page data sheet including clinical risk, cardiac catheterization, operative, and outcome variables. Risk variables predictive for mortality for patients undergoing coronary artery bypass grafting were determined to be age, prior heart surgery, priority of surgery, pulmonary rales, New York Heart Functional Class, peripheral vascular disease, current diuretic use, and chronic obstructive pulmonary disease. The most common predictors of operative mortality for patients undergoing valve and other cardiac operations were age, priority of surgery, peripheral vascular disease, and great vessel repair.
Since 1987, the VA has continued to monitor mortality and morbidity using risk-adjusted methods, and some measures of efficiency such as length of stay. These data are analyzed every 6 months for each cardiac center, recalculating the risk model set developed using a logistic regression approach (with a test-retest method based on the most recent 3-year period of data). Table 1 lists the variables that are significant predictors of operative mortality for coronary artery bypass graft and for valve and other procedures. The risk-adjusted outcome on resource consumption statistics is also summarized and presented at semiannual meetings of the VA Cardiac Surgery Consultants Committee. The blinded, confidential reports are also transmitted to each of the following personnel: cardiac surgical program director, chief of surgery at each VA medical center, each VA hospital medical director and chief of staff, and the regional directors. Although this is a top down–driven program with a mandate to monitor quality, the feedback of data and consultation services supplied by the VA Cardiac Surgery Consultants Committee creates more of a continuous quality improvement environment rather than a quality control monitoring and enforcement program. At this point, more that 90 000 patients have been enrolled in the VA database. There has been a statistically significant reduction in risk-adjusted operative mortality over the past 10 years.6 Although there are many factors that have likely contributed to the improved outcomes such as better myocardial protection and surgical techniques, the feedback of data to the individual programs and also the proactive reviews performed by the VA Cardiac Surgery Consultants Committee as outside consultants has also very probably made a contribution.
In 1989, the STS similarly developed a voluntary risk-adjusted database in an effort to help members assess the quality of care in their individual practices and affiliated hospitals, with the goal of having accurate data with sophisticated risk-adjustment techniques. Under the guidance and leadership of Richard Clark, MD, one objective was to counterbalance the release of unadjusted or poorly adjusted and often inaccurate and misleading mortality data compiled by other organizations and agencies.7 This database initially used the Bayes theorem as its risk-adjusted method.8 Approximately 5 years ago logistic regression modeling was begun as the primary risk-adjustment tool with all risk models recalculated back to 1990 using this more contemporaneous approach.9
Data for more than 1.6 million cardiac surgical patients have been entered into the STS database and have demonstrated a significant reduction in operative mortality over the past 10 years. The STS uses 217 core fields and 255 extended fields. There are 176 variables for all cases and an additional 8 for coronary artery bypass graft, 6 for valve replacement, 10 for prior intervention, 16 for minimally invasive surgery, 11 for other cardiac procedures, 4 for other noncardiac procedures, and 29 for complications.10 The variables that are significant predictors of operative mortality for coronary artery bypass surgery in the STS model are listed in Table 2. Risk factors for mortality following a valve replacement and valve and coronary artery bypass graft are listed in Table 3.11,12
One of the major differences between the VA and the STS databases, in addition to one being mandatory and the other being voluntary, respectively, is the fact that only 1% of the VA population is female and 29% of the STS population is female. In addition, the VA serves a mostly uninsured population that generally has a significantly lower baseline quality of life as measured by the 36-Item Short-Form Health Survey.13
A prime example of a successful regional quality improvement program in cardiac surgery is the NNE Cardiovascular Disease Study Group. This NNE group initially reported on its study of in-hospital mortality following coronary artery bypass grafting performed between 1987 and 1989 at the 5 medical centers in Maine, New Hampshire, and Vermont. The study included 3055 patients and was risk adjusted. After adjusting for the preoperative patient variables, there were statistically significant differences among medical centers and among surgeons.14 The NNE group has used their data for internal benchmarking and peer review to improve the quality of care.
The NNE group instituted regular round-robin site visits for feedback on outcome data and training in continuous quality improvement techniques for their centers and participating physicians. O'Connor and his colleagues in the NNE group15 compared the observed with expected hospital mortality ratios prior to intervening with the above site visit and quality improvement methods and afterwards. Following the intervention, there was a 24% reduction in deaths that was statistically significant. This reduction was consistent across patient subgroups and was related to the timing of the interventions. They, therefore, concluded that a multi-institutional regional model for improvement in cardiac surgical care was feasible and effective.15
The New York State Department of Health began collecting prospective clinical data on all patients undergoing cardiac surgery in 1989. These data are reported to the Cardiac Surgery Reporting System and the program is overseen by the Cardiac Advisory Committee of New York State. The major purposes of the registry are to provide information to hospitals to aid them in assessing and improving their quality and appropriateness of care, to assist the New York State Department of Health in quality improvement activities, and to provide consumers with information that could aid them in selecting providers of cardiac surgery. In the process of doing this, New York State has developed a sophisticated risk-adjustment model under the direction of Edward L. Hannan, PhD. In their initial report in 1990,16 they found 4 of 28 hospitals to have a significantly higher mortality rate than expected using 95% confidence intervals. The authors note that subsequent site visits and medical record reviews to these centers confirmed quality of care problems among those cases that resulted in death. In a follow-up report in 1994, Hannan et al17 noted that the actual mortality decreased from 3.52% in 1989 to 2.78% in 1992 and that there was an increase in severity of patient disease. This resulted in an even greater decrease in risk-adjusted mortality of 41%, from 4.17% in 1989 to 2.45% in 1992. They concluded that the collection and dissemination of mortality data played a significant role in this decline in mortality and that these types of quality improvement programs should be generalized to other procedures and conditions. Of further interest from New York State's efforts is the fact that all groups of providers demonstrated large reductions in risk-adjusted mortalities, but the groups that had the highest initial mortalities had the most improvement. Also the volume of surgical operations for the various groups did not substantially change in that 4-year period.18
The STS has strongly encouraged state activities as the best method of taking data from the database back to the local surgeons and their colleagues to implement improving processes and, thus, the quality of care. The first state to implement this on a voluntary basis was Minnesota, which in 1993 organized the Minnesota STS and the Minnesota Cardiac Surgery Database.19 This database was organized largely in response to a third-party payer demand for data about practice protocols and patient outcomes and was an effort to develop one tool to satisfy those needs for the third-party carriers, but also as a quality improvement tool for the 46 cardiothoracic surgeons and more than 14 institutions. This group established round-robin site visits with 2 purposes—to learn from each other and to validate data completeness and accuracy between institutions. The analysis is performed for the STS National Database by the Duke Clinical Research Institute, Durham, NC, and reports are generated for review quarterly. Data remain confidential because the hospitals are blinded to each other's results.
The state of California has 2 parallel data collection systems in an effort to improve quality.20 The California Coronary Artery Bypass Graft Mortality Reporting Program (CCMRP), a cooperative venture between the State of California Office of Statewide Health Planning and Development and the private group, Pacific Business Group on Health, was initiated in 1995. Although this is reportedly a voluntary reporting system, there is considerable external pressure to join since it is publicly reported and those that do not volunteer are publicly listed as nonparticipants. Recently the state of California passed a law mandating public reporting of hospital- and surgeon-specific mortality outcomes for coronary artery bypass procedures. Many of the surgical groups and hospitals are also participating in the STS National Database and the California STS Database. The STS National Database and Duke Clinical Research Institute provide the California STS with state-specific reports with blinded hospital risk-adjusted mortality rates. State members are using these reports not only for quality assurance in their own institutions, but also to compare the quality of their performance with their peer group. They are also using the STS National Database reports as a way of checking on the accuracy of the CCMRP report, which has just been released to the public. Approximately 5 to 6 years ago the STS leadership urged the State of California Office of Statewide Health Planning and Development/Pacific Business Group on Health/CCMRP to use STS national data and to serve as the auditors of those data to ensure a high degree of accuracy and consistency in completeness. This offer was turned down by the CCMPR that preferred to continue with this parallel, yet costly, endeavor with much less statistical power than the large STS National Database. They are, however, using many of the STS's variables and definitions, and the STS and CCMRP are endeavoring to cooperate with each other to ensure the highest level of accuracy in the California reporting. It would, however, be most efficient if the 2 could work together using the STS's National Database with the state supporting auditors to assess the quality of the data entry. That level of cooperation would most likely result in more current and accurate reports that are statistically more valid and meaningful and with more "buy-in" from the surgeons in the process.
Colorado, Iowa, Alabama, and Minnesota are participating in a clinical research study funded by the Agency for Healthcare Research and Quality to evaluate the effectiveness of feedback of data from the STS's databases on quality improvement in these states. Incorporated into this Agency for Healthcare Research and Quality grant is education on certain processes of care with follow-up of utilization of these processes and their potential effect on outcomes. This study involves the production by the STS National Database–Duke Clinical Research Institute of state-specific reports with each of the hospitals blinded for the various preoperative risk factors and outcome variables. These data are reviewed in regular state meetings and discussed. There are round-robin site visits where a multidisciplinary cardiac surgical team from one hospital visits another sharing ideas for best practices and processes of care. In addition, data managers from the different hospitals visit each other and audit each other's data to check on data completeness and consistency in definition and interpretation.10
New Jersey and Massachusetts are using the STS's National Database for state reports and for monitoring of quality. Similarly, the STS would like to explore areas for collaboration with the NNE Database and the New York State Database in the future.
Most of the activity described herein involves the measurement of operative mortality and morbidity. Other areas that need to be explored include measurement of quality of life, functional health status, longitudinal or long-term follow-up, cost and efficiency, and the outcome measurements of different processes of care. Some work had been accomplished in these areas, but much remains to be done. Rumsfeld and his colleagues21 in a review of VA data, found that the physical component of quality of life as measured by the preoperative 36-Item Short-Form Health Survey was a statistically significant risk factor for 6-month risk-adjusted mortality following coronary artery bypass surgery. Specifically, a 10-point lower 36-Item Short-Form Health Survey physical component summary score had an odds ratio of 1.39 for predicting mortality. This indicates that a significant lower quality of life can lead up to an almost 40% increased risk of death following coronary artery bypass holding all other factors constant. Both the STS's and the VA's databases are measuring preoperative, postoperative, and total lengths of stay as a surrogate for the measurement of cost and efficiency. This is in the process of being risk adjusted and then each hospital will be better able to compare where they stand with their peers.
There are several studies that have been generated from both the VA's and STS's databases on the effect of varying processes on outcomes. Two examples of these are the studies on the use of the internal mammary arteries,22,23 which revealed a significant reduction in risk-adjusted operative mortality when the mammary artery was used to perform at least one of the coronary artery bypasses (Figure 1).
Recently, both the VA's and the STS's database committees have tried to respond to the increasing array of new technology by measuring the effect of such technology on the outcomes. An example of this is the measurement of mortality and morbidity following off-pump coronary artery bypass compared with on-pump coronary artery bypass.24,25 The use of off-pump resulted in an approximately 40% reduction in risk-adjusted operative mortality in the VA's and 20% the STS's systems (Figure 2). These data must be interpreted cautiously, however, because both of these studies are from observational databases and it is well known that there is considerable case selection and the risk-adjusted method may not completely control for the case selection bias that has likely occurred. Nevertheless, the difference is of such magnitude that it is compelling data to justify a prospective randomized study. Such a study has just been funded by the VA Cooperative Studies Program and is scheduled to start with patient accrual in early 2002.
One of the most important and dramatic effects of all of these cardiac surgical databases and quality improvement initiatives has been the uniform demonstration of improvement in risk-adjusted outcomes over the period that they have been implemented. This has been seen in the VA's, STS's, New York State's, and NNE group's efforts. In isolated coronary artery bypass grafting, the unadjusted operative mortality in the VA decreased from 4.3% in 1989 to 2.7% in 2000, the observed-to-expected mortality ratio over that period has decreased from 1.0 to 0.82 (P = .008) (Figure 3). The STS has also seen a decrease in operative mortality from 3.8% in 1991 to 2.7% in 2000 and a decrease in the observed-to-expected mortality ratio from 1.5 to 0.9 (P<.001) (Figure 4).6
There are many factors that have contributed to the improvement in outcomes. One of these may well be the collection of risk-adjusted outcome data and the feedback of data to the cardiac surgical team calling their attention to areas in need of improvement. For example, there has been an increase in the use of the internal mammary artery as a conduit since the publication of the STS's and VA's reports22,23 showing the improved survival with its use. There are many other factors that have no doubt contributed, such as improved myocardial protection and improvements in technology, examples being off-pump coronary artery bypass surgery, better anesthesia, and critical care monitoring techniques. The bottom line, however, is that whatever the reason, the operation is safer despite the fact that the patients are higher risk, as demonstrated by all of the databases.
One of the major challenges for multi-institutional databases, which are necessary components of quality improvement programs, are the new Health Insurance Probability and Accountability Act of 1996 regulations regarding confidentiality. To perform any longitudinal follow-up, some form of patient identifier will be necessary and to improve efficiency and early feedback of data, the use of the Internet would be desirable. It is hopeful that these regulations will not be so burdensome as to impede progress with data collection, which can be used to improve quality. It is important that all of the medical and surgical specialties work together to educate government officials regarding the importance of reasonable legislation, rules, and regulations that protect patient confidentiality but do not go so far as to impair our quality improvement tools. The STS is in the process of incorporating patient consent for encrypted core data collection and transmission of patient data for quality improvement and longitudinal follow-up. However, the patient consent process may lead to some patients refusing, which could bias data and have a negative effect on the accuracy and reliability of the data for use in quality improvement.
Another challenge is maintaining the high quality of data submitted. New York State audits data as do the NNE group and several of the STS's regional groups including Minnesota, Colorado, Iowa, and Alabama that perform round-robin site visits to audit data completeness and accuracy. The VA has the added advantage of having several parallel databases that can be used to verify the accuracy of its national cardiac surgery database. Outside of the states mentioned earlier, the STS relies largely on data software checks and checks for completeness and outliers for various risk factors and outcome variables in the Duke Clinical Research Institute quality checks. The STS has developed thresholds for mortality and morbidity both high and low, which produce queries to the local groups when they fall out of range. A major effort has been made in the education of data managers and surgeons with regard to definition variables and the importance of data completeness and quality. This has resulted in a progressively higher completion rate and consistency in the interpretation of definitions.
Other challenges are to encourage surgeons and other health care providers on the cardiac surgical team to use these data and to turn them into action for improving quality. This is always a major challenge and the STS and the VA continue to try to improve feedback to the local providers in an effort to call their attention to the data in the hopes that it will be used constructively. The effectiveness of this feedback reporting concept is a major hypothesis being tested by the Agency for Healthcare Research and Quality grant to the STS.
Cardiac surgery databases are also a valuable tool for affecting health care policy and relationships with government, the insurance and health maintenance organization industry, and large purchasers of health insurance. For example, both the VA's and STS's databases have found that there is a broad range of cardiac surgery volumes that can be performed with excellent outcomes. A minimum threshold for volume of surgery is unlikely to produce actual outcomes that are lower than expected.26,27 This lack of a documented volume relationship has led both the STS and the VA Cardiac Surgery Consultants Committee to make a strong case to government, third-party carriers, and state regulatory agencies to evaluate cardiac surgical programs on their outcome performance rather than on the volume of surgical cases. Measuring outcomes such as postoperative length of stay and relating the current cardiac surgical database to cost databases, could lead to decreasing costs that will be a benefit not only to the local providers and hospitals but also to the government and third-party carriers.
The STS serves as an example of a professional society working with government on follow-up of specific devices after they have been released by the Food and Drug Administration. The STS with its partnering with the Food and Drug Administration is following up on transmyocardial revascularization after market release. In addition, there is an opportunity for partnering with industry to determine outcomes related to their products across a representative group of medical centers. All of these activities should ultimately lead to better patient care.
For many years, the STS has been developing a general thoracic surgical database. Initially, there was a simplified form during the early years of the database. More recently, because of the efforts of Mark Orringer, MD, David Harpole, MD, and Bill Putnam, MD, and their committee members, a relatively simplistic data form has been developed for patients generally having several thoracic procedures and includes preoperative risk factors, operative-specific data, and outcomes including mortality and morbidity and length of stay (http://www.sts.org). This has been placed at beta-test sites. Inexpensive software is being developed internally and is due to be released in 2002, allowing members to store their own data on desktop computers. Following completion of software development and further experience with beta testing, data will then be transmitted to the Duke Clinical Research Institute on either an annual or semiannual basis for analysis so that members will be able to compare their experience and outcomes in general thoracic surgery with the national cohort.
In a similar fashion, the Congenital Database Subcommittee under the leadership of C. Mavroudis, MD, has developed a data dictionary with complete description of the definitions for the multiple variations of congenital defects.28 This has been an international effort sponsored by the STS. There is a simple core data set being collected, but also software has been developed for a complex data set for higher volume centers.
As noted earlier, there has been considerable activity with evolution and development of quality improvement activities in the cardiothoracic surgical community using multi-institutional databases, allowing surgeons to compare their results to local, regional, and national peer groups. Most commonly reported in a blinded, confidential manner, these data are regularly distributed to the surgeons and their teams allowing them to identify areas for improvement. In addition to risk-adjusted morbidity and mortality outcomes, various processes are measured and the utilization of these processes compared with the national peer group. These databases have also allowed for the implementation of multiple research studies resulting in publication and dissemination of information that, very likely, has led to improved quality of care.
The challenges that remain ahead are to collect long-term outcome data and data related to quality of life, disease-specific symptoms and their relief, freedom from further revascularization procedures, and cost. The opportunity for collaboration with industry, government, and third-party carriers is considerable—it is hoped that by partnering with outside agencies such as these even further improvements in quality of care can be achieved. These quality improvement activities in cardiothoracic surgery can and should be applied to other specialties of surgery.
Corresponding author: Frederick L. Grover, MD, Department of Surgery, Campus Box 305, University of Colorado Health Science Center, 4200 E Ninth Ave, Room 5622, Denver, CO 80262 (e-mail: Frederick.Grover@uchsc.edu).
PN Quality control for cardiac surgery in the Veterans Administration. Ann Thorac Surg.
1986;4237- 44Google ScholarCrossref
GG Report of the Ad Hoc Committee on Risk Factors for Coronary Artery Bypass Surgery. Ann Thorac Surg.
1988;45348- 349Google ScholarCrossref
et al. Multivariate discriminant analysis of the clinical and angiographic predictors of operative mortality from the Collaborative Study in Coronary Artery Surgery (CASS). J Thorac Cardiovasc Surg.
1980;80876- 887Google Scholar
et al. for the Veterans Administration Cooperative Study on Valvular Heart Disease, Clinical, hemodynamic, and angiographic predictors of operative mortality in patients undergoing single valve replacement. J Thorac Cardiovasc Surg.
1987;93884- 897Google Scholar
C Initial report of the Veterans Administration Preoperative Risk Assessment Study for Cardiac Surgery. Ann Thorac Surg.
1990;5012- 26Google ScholarCrossref
et al. A Decade's experience with quality improvement in cardiac surgery using the Veterans Affairs and Society of Thoracic Surgeons National Databases. Ann Surg.
2001;234464- 474Google ScholarCrossref
M Coronary artery bypass grafting: the Society of Thoracic Surgeons National Database experience. Ann Thorac Surg.
1994;5712- 19Google ScholarCrossref
J The Society of Thoracic Surgeons National Cardiac Surgery Database: current risk assessment. Ann Thorac Surg.
1997;63903- 908Google ScholarCrossref
et al. for the Committee to Establish a National Database in Cardiothoracic Surgery, The Society of Thoracic Surgeons, The STS National Database: current changes and challenges for the new millennium. Ann Thorac Surg.
2000;69680- 691Google ScholarCrossref
et al. Prediction of operative mortality after valve replacement surgery. J Am Coll Cardiol.
2001;37885- 892Google ScholarCrossref
FH The 1996 coronary artery bypass risk model: the Society of Thoracic Surgeons Adult Cardiac National Database. Ann Thorac Surg.
1999;671205- 1208Google ScholarCrossref
et al. Self-report measures for assessing treatment outcomes in cardiac surgery patients. Med Care.
1995;33OS76- OS85Google ScholarCrossref
et al. for the the Northern New England Cardiovascular Disease Study Group, A regional prospective study of in-hospital mortality associated with coronary artery bypass grafting. JAMA.
1991;266803- 809Google ScholarCrossref
et al. for the Northern New England Cardiovascular Disease Study Group, A regional intervention to improve the hospital mortality associated with coronary bypass graft surgery. JAMA.
1996;275841- 846Google ScholarCrossref
EP Adult open heart surgery in New York State: an analysis of risk factors and hospital mortality rates. JAMA.
1990;2642768- 2774Google ScholarCrossref
MR Improving the outcomes of coronary artery bypass surgery in New York State. JAMA.
1994;271761- 766Google ScholarCrossref
MR New York State's Cardiac Surgery Reporting System: four years later. Ann Thorac Surg.
1994;581852- 1857Google ScholarCrossref
et al. Establishing and using a local/regional cardiac surgery database. Ann Thorac Surg.
1997;641245- 1249Google ScholarCrossref
BJ Statewide reporting of coronary artery surgery results: a view from California. J Thorac Cardiovasc Surg.
2001;121409- 417Google ScholarCrossref
et al. for the Participants of the Department of Veterans Affairs Cooperative Study Group on Processes, Structures, and Outcomes of Care in Cardiac Surgery, Health-related quality of life as a predictor of mortality following coronary artery bypass graft surgery. JAMA.
1999;2811298- 1303Google ScholarCrossref
KEfor the Department of Veterans Affairs Cardiac Surgeons, Impact of mammary grafts on coronary bypass operative mortality and morbidity. Ann Thorac Surg.
1994;57559- 568; discussion 568-569Google ScholarCrossref
M Impact of internal mammary artery conduits on operative mortality in coronary revascularization. Ann Thorac Surg.
1994;5727- 32Google ScholarCrossref
et al. Off-pump coronary artery bypass is associated with improved risk adjusted outcomes. Ann Thorac Surg.
2001;72114- 119Google ScholarCrossref
FL Off-pump coronary artery bypass grafting significantly decreases risk-adjusted mortality and morbidity. Ann Thorac Surg.
2001;721282- 1289Google ScholarCrossref
et al. No continuous relationship between Veterans Affairs hospital coronary artery bypass grafting surgical volume and operative mortality. Ann Thorac Surg.
1996;6117- 20Google ScholarCrossref
REfor the the Ad Hoc Committee on Cardiac Surgery Credentialing of the Society of Thoracic Surgeons, Outcome as a function of annual coronary artery bypass graft volume. Ann Thorac Surg.
1996;6121- 26Google ScholarCrossref
J Congenital Heart Surgery Nomenclature and Database Project: overview and minimum dataset. Ann Thorac Surg.
2000;69S2- S17Google ScholarCrossref