The STS Cardiac Surgery National Database: An Update

Transcription

1 ORIGINAL ARTICLES The STS Cardiac Surgery National Database: An Update Richard E. Clark, MD For the Committee for the National Database for Thoracic Surgery Since inception in 1990, The Society of Thoracic Surgeons (STS) National Cardiac Database has grown rapidly. More than 1,500 surgeons working in 706 hospitals have contributed more than one half million patient records. Geographic distribution of those participating is proportional to the number of centers performing heart surgery. The STS system is in use in all 49 states where centers are operating. There has been a significant decrease in length of stay for most patients having heart operations and a modest fall in coronary artery bypass grafting operative mortality from 3.7% to 3.3% over the past 3 years. Coronary artery bypass grafting case mix also is changing nationally as evidenced by a decline of 17% in the best-risk cases and concomitant increases in those with predicted risks of 5% to 10% and greater. New uses for local data in addition to self assessment and quality assurance include development of critical clinical pathways, support for managed-care group applications, and regional use. Minnesota has established a statewide STS system and Florida is soon to follow. The key to acceptance has been a peer-reviewed risk-stratification system that continues to be refined each year. Finally, a major effort will be made this year to increase the participation of general thoracic surgeons, particularly with respect to lung cancer. (Ann Thorac Surg 1995;59: ) he rationale, goals, and progress of The Society of T Thoracic Surgeons' (STS) National Database (NDB) have been published periodically [1-6]. A 2-year status report was presented to the membership at its Twentyninth Annual Meeting in San Antonio, Texas [4]. The historical aspects are summarized succinctly in Dr Richard P. Anderson's editorial [7], which prefaces three articles: one giving the first 2-year report [4], the second on the risk stratification system [5], and the third on the impact of use of the internal mammary artery [6]. The NDB Committee thanks the Council and Program Committee for the opportunity to inform the STS members and the public of our continued progress in this, the second biannual report. The purpose of this report is twofold: first to update previous data on growth, trends, and risk stratification, and second to detail the new uses of the national cardiac database for individual practices and regional applications in the managed care setting. Presented at the Thirty-first Annual Meeting of The Society of Thoracic Surgeons, Palm Springs, CA, Jan 30-Feb 1, The past and present members of the Committee are Richard E. Clark, MD (Chairman), Anthony J. Acinapura, MD, Richard P. Anderson, MD, Douglas M. Behrandt, MD*, Fred H. Edwards, ME), L. Penfield Faber, MD*, T. Bruce Ferguson, Jr, MD, Frederick L. Grover, MD, William R. E. Jamieson, MD, Forrest L. Junod, MD*, John E. Mayer, Jr, MD, Constatine Mavroudis, MD, Richard B. McElvein, MD*, Joseph I. Miller Jr, MD, Walter E. Pae, Jr, MD, Victor Parsonnet, MD*, Richard M. Peters, MD, Valerie A. Rusch, MD, Quentin R. Stiles, MD*, Victor F. Trastek, MD, Daniel J. Ullyot, MD, and Donald C. Watson, MD (*retired from Committee). Address reprint requests to Dr Clark, Cardiovascular and Pulmonary Research Center, Allegheny-Singer Research Institute, 320 E North Ave, 8th Floor, ST, Pittsburgh, PA Updates Growth In the past 4 years, the number of surgeons having the STS system available has increased nearly thirteenfold and now numbers 1,528 (Fig 1). The number of hospitals in which the surgeons work and use the data has increased more than twentyfold since 1990 (Fig 2), which represents a majority of the 900 to 1,100 hospitals performing cardiac surgery in the United States. Importantly, the total number of patient records since 1980 has grown to more than 520,000, with most of the increase in coronary artery bypass graft (CABG) patients accrued in the past 3 years (1992 through 1994) (Fig 3). The question of geographic misrepresentation, prominent in the first 2 years, is now moot as shown in Figure 4, where the distribution pattern follows the general population pattern. The STS NDB is represented in every state except Wyoming, where there are no open heart surgical centers. The distribution and volume of surgical groups and volume are representative of the majority of the STS members who work outside of large academic tertiary referral centers. Despite the fact that many of these large centers had their own cardiac databases in 1990, 112 teaching institutions, of which 54 are university medical centers and 13 are Veterans Affairs hospitals, have now joined the STS database system. An important addition to the database was the launching of the congenital heart surgery component after 2 years of development under the leadership of Dr Constantine Mavroudis. The rationale and approach used are discussed in the subcommittee's report published recently in The Annals of Thoracic Surgery [8]. Forty institutions currently have this system in place by The Society of Thoracic Surgeons /95/$ (95)00288-V

2 Ann Thorac Surg CLARK ;59: STS CARDIAC SURGERY DATABASE " 1400 C b'~ oo Z Years Fig 1. The number of surgeons participating in The Society of Thoracic Surgeons" National Cardiac Database for the years 1990 through The numbers joining each year were 320, 365, 420, and 293, respectively. The number of surgical groups represented by these surgeons is 517 in 1994, with a nearly constant gain of approximately 120 groups per year. Trends Illustrations of trends of complete clinical data are too comprehensive for this short report. Over the past 4 years, the average age for CABG operations (64 years) has changed little, as has the percent of female patients (28%). Also relatively constant are the percentages of elective operations (80%) and reoperations (10%), and the ejection fraction (0.51) before operation. Significant trends are now evident over the past 5 years in postoperative length of stay, which reached 7.5 days in 1994 for first-operation elective CABG (Fig 5), whereas the preoperative length of stay only recently has declined to 2.5 days (Fig 6). The second major trend, which began in 1991, is a continued decline in the perioperative ~ Fig 3. The number of patient records in thousands in the Society of Thoracic Surgeons" National Cardiac Database for the years 1990 through There has been a twelvefold increase since inception. mortality of all patients having CABG operations. Since 1988, there has been a near-linear further decline of 0.13% per year, reaching 3.3% in 1994 (Fig 7). These data are in keeping with those of Hannan and colleagues [9] for New York State, where the mortality rate decreased from 3.5% to 2.8% during the 1989 to 1992 interval. Risk Stratification Methodology that can adjust for different case mixes of various practices has been the keystone of the adult cardiac database. The use of Bayesian theory was controversial at inception but has received acceptance by STS members, hospitals, various third-party payors, and regulatory agencies. The details of the algorithm were published by Edwards and associates [5]. Essentially, it predicts the probability of a future event based on the characteristics of past events. This is accomplished by constructing a matrix of conditional probabilities relating 6 I ~ 00 ~ 200 o Years Fig 2. The number of hospitals in which The Society of Thoracic Surgeons' National Cardiac Database participating surgeons practice for the years 1990 through This represents 65% to 80% of the United States hospitals in which cardiac surgery is performed. Fig 4. The distribution of The Society of Thoracic Surgeons' (STS) National Cardiac Database participating hospitals in the United States and Canada in December There are no cardiac centers in Wyoming. (Map of the United States Creative Teaching Press, Inc. Reprinted with permission.)

3 1378 CLARK Ann Thorac Surg STS CARDIAC SURGERY DATABASE 1995;59: "v.,3'411, % 7.0% 6.0% e~ lo 9 5.0% 4.4% % 4.0%j.,.8% ~375~x ~?/~ 3.7% 3.5% % gh 8 3.0%.... o ,339 6,761 6,251 I I I ,369 10,339 14,872 24,997 41, ,001 Fig 5. Annual postoperative length of stay (LOS) for first operation, elective coronary artery bypass grafting. There has been an approximate 25% decline in 5 years. (Data for 1994 are first half of the year only. Patients missing dates are censored.) (SE - standard error of the mean.) the frequency with which a given risk factor is seen in association with a specific outcome, eg, survival or death. Using a training-set]test-set approach, the first version of the prediction program was derived from nearly 60,000 CABG patients for the years 1984 to The system used 23 variables, most of the yes/no variety for presence of a particular co-morbid factor. The fit of the predictedto-observed mortality was good and the equation was applied to patients operated on in the years 1991 through Variances became evident over time, with a separation of the observed and expected mortality curves. A reassessment of the last 3 years of CABG data was undertaken in It was found that there were sufficient differences in the distribution of preoperative patient characteristics between years to require yearly algo ~, h el l.s i I I ,339 6,76i 6,251 8, ~39 14,872 24, ,911 53,090 39,001 Fig 6. Annual preoperative length of stay (LOS) for first operation, elective coronary artery bypass grafting. The decline was insignificant until 1992, when external and internal forces resulted in an approximate 15 /o decrease. (Data for 1994 are first half of the year only. Patients missing dates are censored.) (SE = standard error qf the mean.) 2.0% --!.0% - - i 0.0% I I I I ,645 10,370 10,011 13,072 15,525 20,031 37,487 r ,007 80,661 65,704 Fig 7. Annual average mortality for patients (n = 325, 438) undergoing coronary artery bypass grafting, including all reoperations and elective, urgent, emergent, and salvage categories. A gradual decline is noted for the past 5 years and reached 3.3% in (Data for 1994 are first half of the year only.) rithms to portray the expected risk accurately. Consequently, users of the STS system received new programs for prediction in the second half of 1994 after their data were harvested. The new system predicts 1994 and 1995 mortality based on the 1993 algorithm. Updates will occur yearly. The case mix for the past 10 years (1985 through 1994) has been examined using the risk prediction system. The data show a decline of 17% in the best-risk patients (0% to 2.5%), little change in the next-to-lowest risk patients (2.5% to 5.0%), and increases in the frequency of all higher risk categories (>5.0% to 10.0%, >10% to 20%, >20% to 30%, and >30% to 50%) over the decade. New Uses of the STS National Cardiac Database Quality Assurance Two of the major motives in establishing the STS NDB were to provide each STS member with a convenient, private, self-owned system to assess his or her performance and to make comparisons with averaged national data that have been risk stratified using the same variables and the same algorithm. Clearly, in smaller volume practices, a few catheterization laboratory calamities can drastically alter a monthly or yearly mortality/morbidity report. The STS risk stratification system has put into perspective the incidence of mortality and various morbidities. It also has enabled the user to self-judge to make improvements where necessary and reinforce continued practice of a specific technique or care routine. Nearly all members of the STS NDB system use it regularly for local quality assurance reports. Thus, a primary goal of the program has been met. Not envisioned at the onset was the development of "critical clinical pathways," a branching system of dos and don'ts to avoid unnecessary tests and procedures to control costs. In many instances and locations, especially in teaching hospitals, these have proved useful for straightforward, elective, very low risk operations. For

4 Ann Thorac Surg CLARK ;59: STS CARDIAC SURGERY DATABASE cardiac surgery, and CABG operations specifically, the pathways are more complex because there is a wide spectrum of clinical presentations and pathophysiology in this elderly population. Several clinical pathways based on broad categories of risk have proved more useful than one pathway alone. In determining these categories, the STS risk stratification method has been useful and effective, and has provided the basis for the desired flexibility of assessment and treatment plans. Finally, the gathering storm of managed care that was on the horizon in 1990 in most areas has enveloped nearly every sector of this country in The individual practitioners and small groups were particularly vulnerable to the requirements of large, managed care systems unless they had risk-stratified data. The letter received by many members of the STS from U.S. Healthcare in April 1994 is a case in point. The letter suggested three mandates for membership: (1) current, clear, concise, accurate data, (2) use of a risk assessment system, and (3) volume of more than 150 CABG cases per year. The STS system was suggested as appropriate in that letter. The members using the system have found that they could meet the requirements of managed care entities for cardiac surgical data in a timely manner. No adverse comments have been received. Verbal and written communication asking for help has been received by both the Chairman and particularly the database manager, Summit Medical, Inc, for those without the STS system. Below are four vignettes of cardiac surgical practices that were saved from exclusion from managed care systems when they were able to extract recent data rapidly from office and hospital records, put these into a local personal computer with an STS program, and provide the required collated summaries to various managed care systems. VIGNETTE 1. A large midwest hospital with an active cardiovascular surgery program in a highly competitive town that is moving quickly into a managed care environment received a notice from a large health maintenance organization (HMO) network directing them to submit risk-stratified data on mortality, complications, and length of stay by surgeon within 30 days. The communiqu6 from the HMO further informed the institution that its failure to provide the information would make them ineligible to participate in their network, thereby eliminating 50% of their cardiac patient referrals from their hospital system. The surgeons from the hospital signed up immediately for the STS program and mobilized a team of nurses, surgeons, and technicians to enter data on 500 cases into the STS software system. With assistance from the Summit Medical Clinical Services department, the team was able to prepare the data for presentation to the HMO. To the astonishment of the HMO, the institution complied with all the requests of the HMO, and more, and continued to function as part of the HMO network. VIGNETTE 2. A large eastern cardiovascular surgical center was notified by the State Health Commissioner that its operative mortality was excessive and that failure to explain could jeopardize the ongoing operation of the surgical program. The surgical groups on staff had been participating in the STS-NDB for several years. An indepth analysis by surgeons using risk stratification uncovered two surgeons with patients whose operative risk was clearly two to three times that of the group and whose raw mortality was twice that of the group. Using risk stratification methods of the STS, the rebuttal to the State Commissioner of Health noted that the high mortality rate of the institution was due to well-defined and well-identified high-risk patients. The extensive data presentation was well received by the Commissioner of Health, and the institution now markets its program as one that effectively accommodates high-risk patients. VIGNETTE 3. A small-to-moderate size cardiovascular surgical program in a southern hospital was under investigation to be decertified as a qualified open heart center because the two-man group had performed only 135 procedures in the previous year and therefore did not qualify under the payer's "hurdle" requirement of 200 cases a year. The group had been participating in the STS-NDB program for 2 years and was able to compile data on 260 cases showing that their actual mortality was slightly less than the predicted operative mortality for each year, as well as for the 2-year period. Also, their results were consistent with national findings for similar risk populations. Their formal response was further augmented with data on complications and low mortality rates with emergent and redo cases, as well as reduced length of stay. The payer was satisfied with the clinical outcomes and maintained the program in its network. VIGNETTE 4. A moderate-size, private, practice-based cardiac surgery group in the midwest has used the STS database program to prepare a formal preferred provider organization proposal to two major industries in their community. In this proposal the group profiled outcomes of all its patients from the previous 3 years and prepared a separate section on the patients employed by these two large corporations. It was their ability to categorize the patients prospectively by risk levels that convinced the two companies of the practice's attention to detail and analysis of outcomes. The ensuing contract resulted in an additional 148 patient referrals from these businesses during the first year of the contract. Regional Use of the STS System Minnesota has long had some form of managed care. With the rapid increase and evolution of these entities, several major payors in Minnesota focused on cardiac catheterization, percutaneous transluminal coronary angioplasty procedures, and CABG operations. Each wanted to use a separate reporting form of patient preoperative and postoperative outcome variables, and each had different definitions of key variables. One payor was going to require payment by the surgeon of $25.00 for each claim submitted. Additionally, this payor was going to require data concerning all patients operated on

5 1380 CLARK Ann Thorac Surg STS CARDIAC SURGERY DATABASE 1995;59: by the specific surgeon regardless of health insurance carrier. Both the cardiologists and the surgeons quickly organized, the latter led by Dr Kit V. Arom of Minneapolis. Despite intense competition for patients, the 14 centers for heart surgery agreed to a united front in proposing the STS system to all payers. This provided the surgeons, the managed care entities, and the state regulatory agency with three fundamental tools: (1) common format, (2) common definitions, and (3) common risk stratification system. The surgeons formed a state cardiac surgical database to submit pooled, averaged, risk-stratified data as encoded anonymous center-specific reports. The key and underlying commonality was that surgeons of all 14 centers were members of the STS-NDB. Negotiations are underway in Florida for a replication of the Minnesota model, and discussions to develop statewide systems based on the STS system are ongoing in Connecticut, Rhode Island, and Michigan. Even the state of Pennsylvania, where the HealthCare Cost Containment Council controls statewide reporting of CABG outcomes, has begun to inquire about use of the STS system because of its acceptance by most of the surgeons in the state and its risk stratification system. These events demonstrate that the direction chosen 6 years ago by the STS-NDB Committee and the STS Counsel to design a nonacademic system applicable to the practicing community surgeon, with emphasis on preoperative risk factors and outcomes end points, was a wise one. Future Prospects The NDB Committee and software developer continue to work on risk stratification methodology. A project has begun with Lincoln Laboratories of the Massachusetts Institution of Technology to use techniques derived from its work on artificial intelligence and strategy prediction for the Department of Defense and other intelligence organizations. The system is called "neural networks," which is thought to emulate the methods by which humans reason and make go/no-go decisions. Such a system, if implemented in the STS database, might provide greater predictive accuracy, especially for the highrisk patient. A second project involves the use of our current Bayesian system to predict complications and length of stay from the preoperative risk factors and demographics. This ability has great portent for both patient and fiscal management of high-cost high-gain operations. A third project involves augmenting the capabilities of the currently available bedside pocket calculator to predict local operative risk and comparing this with the calculated average national risk for CABG patients. Predictive risk analysis for all cardiac adult operations will be developed in 1995, which will provide not only estimated mortality but probabilities of major complications. A fourth project seeks to decrease database costs by use of "pen" computing. The data coordinator, perfusionist, intensive care unit nurse, and others will have a notebook computer with a touch screen pen. No data collection forms are used. Data are entered directly at the locale (floor, operating room, intensive care unit) and, after collection, are sent by radiofrequency transmission to a receiver at the file server, which enters the data directly. By eliminating long forms and, in essence, networking multiple data sites by persons involved in care, the cost of acquiring data can be reduced drastically. Finally, major emphasis will be placed on further development of a wider array of modules for various diseases of the lungs, pleura, mediastinum, diaphragm, and chest wall. It is the Committee's hope that it can provide general thoracic surgeons with the same tools as those used by their cardiac surgical colleagues. References 1. Clark RE. It is time for a National Cardiothoracic Surgical Data Base. Ann Thorac Surg 1989;48: Clark RE. The STS National Database: alive, well and growing. Ann Thorac Surg 1991;52:5. 3. Clark RE. Report of the first presentation of the National Database. Ann Thorac Surg 1991;52: Clark RE. The Society of Thoracic Surgeons National Database status report. Ann Thorac Surg 1994;57: Edwards FH, Clark RE, Schwartz M. Coronary artery bypass grafting: The Society of Thoracic Surgeons National Database experience. Ann Thorac Surg 1994;57: Edwards FH, Clark RE, Schwartz M. Impact of internal mammary conduits on operative mortality in coronary revascularization. Ann Thorac Surg 1994;57: Anderson RP. First publication from The Society of Thoracic Surgeons National Database. Ann Thorac Surg 1994;57: Mavroudis C (for The Society of Thoracic Surgeons' Ad Hoc Committee to Develop a National Database for Thoracic Surgery). The Society of Surgeons' National Congenital Heart Surgery Database. Ann Thorac Surg 1995;59: Hannan EL, Sui AL, Kumar D, Kilburn H Jr, Chassin MR. The decline in coronary artery bypass graft surgery mortality in New York. JAMA 1995;273: DISCUSSION DR SIDNEY LEVITSKY (Boston, MA): Doctor Clark and the STS Committee for the National Database for Thoracic Surgery are to be commended for providing us with the STS Cardiac Surgery National Database. Doctor Clark's report documents the exponential growth in the registry, its representation of the entire national effort in cardiac surgery, and recent trends demonstrating a decrease in length of stay and operative mortality for patients undergoing coronary revascularization. He also has provided us with anecdotal material suggesting new uses for the database, such as quality assurance, the development of critical clinical pathways designed to decrease length of stay, thus increasing hospital profitability, and clinical practice response to managed care initiatives regarding outcomes.

6 Ann Thorac Surg CLARK ;59: STS CARDIAC SURGERY DATABASE As noted in the recent editorial by Dr Anderson [1], this database is an unprecedented voluntary effort by American cardiothoracic surgeons to combine their observations for the purpose of self-assessment and the advancement of knowledge. The next question is whether the promises made by this database have been met. In reality, the question is whether the output is good data that accurately reflects our work and clinical outcomes. For this critical review I have been assisted by my chief resident at the New England Deaconess Hospital, Dr Christian Campos, who is completing his PhD in epidemiology and is one of the principal investigators in the POSH study. The first question is, does the database accurately reflect the population at risk and the variety of surgical groups performing this service? This was an initial issue as large institutional results overshadowed data from small practices. Although it is assumed that eventually the larger volume of patients from numerous small practices will eliminate this bias, the authors have not provided us with any statistical methodology to validate this assumption of population homogeneity. The second and probably most important issue is the quality and accuracy of data entered into a voluntary database. Doctor Fred Edwards, in a recent presentation at the National Symposium on Using Outcomes Data to Improve Clinical Practice: Building on Models from Cardiac Surgery [2], has discussed these problems. The major question is: are all patients at a given institution being entered into the database? Should the patient undergoing closed chest massage, being transported from the angioplasty suite to the operating room, who has an eventual poor outcome, represent a salvage case whose mortality is not entered into the database? Or should the rules of the database mandate that all patients undergoing coronary artery bypass grafting, regardless of circumstances, be entered? Whereas the Health Care Financing Administration in New York Pennsylvania, and the Veterans Affairs databases have the force of law to guarantee accurate data output, the STS database only has volunteerism and good faith. Perhaps unannounced site visits for external audits should be employed, although I doubt whether this methodology is feasible with a voluntary system. Alternatively, internal audits could be performed at institutions where legally mandated government data are available for comparison. The third issue is data completeness, clinical inconsistencies, and "gaming." Are all data fields completed, and what are the cutoff criteria for registering a patient into the database? Perhaps Dr Clark will provide us with rejection criteria or the minimal numbers for each entry risk factor that allow the calculation of conditional probabilities for the risk model. Finally, in these days of intense institutional competition for managed care, "gaming" enters into the equation. Because most of us believe that we operate on the "sick patients" and our competition operates on healthy patients, some surgical groups have gone to extraordinary means to upgrade severity or to rapidly discharge sick patients, such as those with postoperative strokes, to subacute hospital-rehabilitation facilities to avoid having the eventual operative mortality attributed to their institution. Doctor Clark should inform us as to how the Database Committee deals with these issues. Nevertheless, the STS database is still in its adolescence and is constantly being refined as exemplified by the recent adaptation of an annual revision of the Bayesian conditional probability model based on changing population and severity of illness. This effort by the STS to report our operative outcomes deserves the respect and admiration of the entire community of cardiothoracic surgeons and offers promise for self-regulation, which has been the cornerstone of our profession. DR BENSON R. WILCOX (Chapel Hill, NC): The STS is an organization of a thousand committees. Most of us serve on or even chair these committees without fanfare or expectation of reward or recognition, just grateful for the opportunity to be of service, and such is the case with Dr Clark. However, The Society on rare occasions takes special notice of individuals who have made an unusual contribution, well beyond the call of duty, and such is the case with Dr Clark. Dick Clark is literally the father of the STS database, it having been his brainchild more than a decade ago. He has nurtured it into maturity and is now passing the responsibility for its care to a new chairman, Dr Fred Grover, after next year. The Society would like to recognize Dr Clark's tremendous contributions to our welfare with a tangible award that we hope you will enjoy, Dr Clark over the years to come, and with just as heartfelt a recognition, a warm round of applause. References 1. Anderson RP. First publication from The Society of Thoracic Surgeons National Database. Ann Thorac Surg 1994;57: Edwards FH, Clark RE, Schwartz M. Practical considerations in the management of large multiinstitutional databases. Ann Thorac Surg 1994;58: