Coumadin

Cardiac surgery patient operative outcomes have been the most tracked, scored, analyzed, and assessed procedures in the history of medicine. Cardiac surgeons are the most subjectively and objectively "scored" subspecialists in clinical medicine. In 1989, the Society of Thoracic Surgeons (STS) began a voluntary database that to this day is the most effective, accurate and statistically rigorous effort to record, analyze, and relate outcomes to quality improvement. u More than 2 million surgical procedures from over 70% of the hospitals performing heart surgery are archived in the STS database. Approximately 200,000 patients (with 2000 data points/patient) are entered each year nationally.

The most frequent operation within that database is the coronary artery bypass procedure (CABG), which has allowed a tracking of expected outcomes in patients with specific risk profiles that is matched in accuracy and quality by no other database or analytic procedure; 23 critical pre-operative variables are used to create algorithms to estimate a patient's risk of death or morbidity categorically for a specific cardiac surgical operation based upon the expected outcomes of all similar patients previously enrolled in the database, weighting the most recent enrollees higher than older historical data input. This provides MDs and patients a realistic assessment of risk. Stratification into up to 10 different risk categories on a continuous scale allows more valid comparison of outcomes amongst programs as compared to national statistical outcomes. It also has allowed tracking national trends in patient characteristics over time. For example, over the past 4 years patients undergoing CABG are older, heavier, have a greater incidence of diabetes and pre-existing diagnoses of cerebral vascular and peripheral vascular disease versus patients prior to the year 2000. As a consequence, the increased preoperative severity of medical illnesses and co-morbidities is responsible for elevations in intraoperative and postoperative morbidities (e.g. increased requirement for blood transfusion, longer postoperative ventilator times and more hospital readmissions within 30 days after discharge). Despite the increased morbidity, the mortality rates have remained stable or declined. The overall operative mortality from 1995-1999 for CABG was 3.3% declining to 2.6% in the time period 2000-2004.3 Low-risk patients (with a predicted mortality of less than 2%) sustained a mortality rate of 0.79% and patients categorized as high mortality risk (i.e. greater than 2% risk of death) sustained perioperative mortalities at 5.37%. Because of the extreme low frequency of death in the low-risk patients, it is, for all practical purposes, impossible for individual cardiologists to assess significant differences among surgical providers for low risk patients. High-risk patients are so variable that without strict application of criteria, valid observed-to-expected death rates cannot be discerned casually and again because of low frequency of the various diagnostic categories denominators can vary widely from year-to-year. O/E ratio = to 1.0 means that for the specific risk profile the expected mortality is being realized. We became STS database participants on the authors arrival in Providence in 1996. The Lifespan Hospitals and their surgeons have consistently had O/E's at or below 1.0 (and way below 1.0 ratios in some time periods) for virtually all major categories of surgeries.

Databases are at their best when assessing programs with large volume subsets. Other methods, especially those using administrative data, are prone to egregious errors, are notoriously subject to biased reporting and are totally inadequate for assessing quality.' These include all "analyses" based on the administrative data sets reported by hospitals for the Medicare subset as available under the Freedom of Information Act. A small cottage industry of internet-based "rating services" markets to the consumer the bogus concept that the quality of a cardiac surgical program can be ranked in a 1-5 star order, like hotels. The data used for these proprietary-rating systems are fundamentally flawed and ill-suited for outcomes research (business administration data and hospital coding for Medicare reimbursement). Their algorithms are rarely published, and the results are highly subject to bias and modeling manipulations. Most importantly, morphing such data into "ratings" does not denote quality. Statisticians, epidemiologists, and cardiac disease specialists have consistently condemned these ratings as unsuitable for making quality comparisons.3-10

In studies on entry data quality, the STS found its database quality to be extraordinarily high with discrepancies less than 5% in audited fields when on-site "real time" quality control was performed by a Division of Cardiac Surgery. The relational database has allowed comparison of mortality and other post-operative outcomes such as hours on the ventilator, etc. to be used for quality improvement programs in hospitals as well as tracking key quality indicators such as sternal wound infections, leg wound infections, strokes, etc. Interestingly the post-operative length of stay while gradually decreasing around the country to a minimum of 6.3 days following CABG (similar to RI) seems to be rising as a consequence of the higher age and elevated co-morbidities in patients referred for a surgical revascularizations. STS data have also been used to compare outcomes of CABG in large datasets for comparison to percutaneous coronary interventions or medicaJ management of coronary disease including, stabile angina, or stress-induced ischemia. The Medicine, Angioplasty, or Surgery study (Mass-II) [Journal of American College of Cardiology] reported on the key summation outcome: the likelihood of survival free of cardiac mortality, unstable angina, need for revascularization or myocardial infarction among the three treatment groups. The best event-free survival was with CABG at 94%; the second was medical therapy at 89%; the worst outcome was with percutaneous intervention at 76% (P

Such statistical analyses and comparison of local outcomes with very large regional and national datasets have allowed a degree of precision in assessing outcomes that is unparalleled in the history of medicine. At least one operation (CABG) has a fairly consistent population subset, has distinct variations in techniques which can be identified and correlated to outcome and due to the large denominator allows (at least in some cases) comparison of the outcomes of one hospital or program to another. Comparisons of surgeons to each other or to normative values is difficult without large numbers of similar patients. But over time a general picture can emerge. What has been shown to be likely true is that low volume surgeons in low volume programs do not have outcomes quite as good as low volume surgeons in high volume programs who perform essentially as well as high volume surgeons in high volume programs.3 While perhaps not intuitively obvious why, the high volume issue actually becomes important mostly for highrisk esoteric cases (e.g. transplant, arterial switch, etc.). Data suggest that an adult cardiac surgical program (i.e. a hospital) should perform ≥ 600 cases/ year. That would give individual surgeons the best opportunity to perform well and essentially guarantee patients better-than-average results at that hospital.

Interestingly, the CHSS (Congenital Heart Surgeons Society) and the STS databases for pediatric congenital heart disease, as well as analogous European databases, have demonstrated that surgeon and hospital case volumes are not a predictor of complex congenital surgeries patient survival" while lesion, surgical complexity and patient-based factors drive outcomes. One explanation has been proffered - that congenital heart surgeons are a select and perhaps elite subgroup of their training "class" and spend more years in training both as fellows and as a junior staff "apprentices." Thus the more experience, more maturity, and greater technical expertise combined with a specialty that includes a vast array of procedures, usually technically and time pressured which demands surgeons of competency results in outcomes data suggesting that for both adult and pediatric procedures, surgeon-based variability is neutralized to a great extent (i.e. either by filtering out lesser qualified trainees or by delivering higher quality training, virtually all of these surgeons, are good to excellent).

Assessment of an individual surgeon's results is tricky - analogous to assessing the hitting ability of a baseball player. In the major leagues it is difficult to assess individual bunting ability statistically because bunts are rare. But, for hitting, a starting player typically has 3-4 at bats per game; over the course of a season of 160 games he will have 300-500/ outcomes/year (N=4000/for a 10-year career). These are the numbers needed for statistically valid comparisons. Probably no sport is more statistically driven than baseball and has fascinated statisticians (analogous to cardiac surgery). The batting average norm across professional baseball continues to be approximately 275 (i.e. out of 1000 at bat 275 times a hit will be accomplished). But is a 300 BA better than a 290? In many cases, no (e.g. if the 290 player always hits homeruns and the 300 hitter only singles). Many fans believe in streaks - runs of good hitting and bad hitting. When rigorous statistical methods are applied, the evidence for true streakiness is minimal and can be demonstrated only in a few players since data have been archived (essentially since 1904). 12 at bats with either zero hits or with 12 consecutive hits is a meaningless observation in valuing the offensive ability of a hitter; similarly 12 sequential operations (or even 100 of different types) cannot provide an adequate database on which to assess a single surgeon as "better" although it can identify outcomes that are truly horrible (i.e. 2 standard deviations off the norm). The analogy to baseball raises even trickier problems: the value of a given operation (or a given at bat) depends (under the best of circumstances) upon the level of risk for mortality, (i.e. strikeout, the degree of difficulty to hit a specific pitcher) and the operation being attempted (i.e. is the batter trying to get a home run or just a hit?). The first problem is modeled in the STS algorithms by assessing the degree of risk that the individual patient exhibits based upon preoperative characteristics. The second issue has been harder to manage; "Olympic scoring"12 does address this, for example a simple swan dive may have a degree of difficulty of three where as a 2 ½ somersault-gainer with a half twist may have a degree of difficulty of 7. Accomplishing a perfect 10 on each gives the diver a score of 30 for the first and a score of 70 for the second. But cardiac operations which may have a better outcome for the patient but which are technically more challenging to perform are not perceived differently. Thus, when limiting outcome assessment even to current STS type of measurements, there is a bias or pressure built in for the surgeon to perform the simplest operation, not necessarily the best operation, as he is scored on patient based risk factors over which he has no control and the diagnosis (e.g. CAD or aortic valve) but not the actual operation performed. To carry the baseball analogy further - this is like limiting hitting assessment to batting average only, while ignoring other measures such as slugging %, on base %, etc. Thus, a heart valve transplant which is more technically demanding than the implantation of mechanical prosthesis is perceived by the patient (and often the cardiology community) as the same operation (i.e. valve replacement) but the outcome for the patient with the mechanical valve may be less salubrious due to mandatory Coumadin therapy and anti-coagulation related life threatening complications which approach 100% over 10-15 years; whereas the more surgically challenging valve transplant procedure (which does not need anticoagulation) may give perfect hemodynamics and unload the heart to the point that left ventricular muscle mass returns to normal and the patient lives a normal life being able to undertake all physical activities including those that have risk of bleeding. Even STS scoring does not account for such higher "value" options or Olympic scoring "difficulty" nor are batting averages weighted differently when the batter faces Nolan Ryan versus a new rookie just up from minors. The Congenital Heart Surgeon's Society in the United States and sister organizations around the world have attempted to develop scoring systems for degree of difficulty for operations (e.g. RACHS-1, Aristotle). Such methods are in their infancy but will be a great step forward in the measurement of outcomes based not only on patient-based profiles of risk but also upon surgical intended quality and the inherent technical difficulty of various therapeutic procedures.13,14 Of course, this means quantitating "value", which will become even more important in Era IV (vida infra).

There is another other statistical anomaly, mathematically inherent to virtually all such assessments. Since at the low end of risk, mortality still has to be a positive number (there can be no negative mortality rates) then there are obligatory mathematical consequences of overestimating risk; similarly at the other end of the curve (high risk- high mortality), one cannot have more than 100% mortality, so the regression curve tends to underestimate risk. If one surgeon has a low risk patient mix, the numbers may overestimate quality of outcomes when compared to a surgeon whose case mix profile includes more highrisk patients. Thus an analyst must carefully analyze the data before reaching comparative conclusions.

This discussion highlights the difficulty inherent in quality management and outcome assessment. Fewer "routine" or "bread and butter" type procedures (e.g. simple, straight-forward 3-5 vessel coronary vessel bypass grafting) will be performed. In addition, the standard median sternotomy (on cardiopulmonary bypass) procedure, which has been refined since the 1960s to an extraordinarily low level of mortality and morbidity, may be minimized as enthusiasm builds for off-bypass, minimally invasive and less traumatic procedures. The challenge to the profession will be to evolve these technological and surgical approaches in such a way that the outcomes are not compromised. Additionally, with the advent of stenting and percutaneous procedures and thus the lack of referral for second opinions on the relative value of surgical options, patients are being treated with "simpler" but perhaps less effective modalities.15 Simpler or less invasive are not the same as safer or more effective. Patients referred for surgery are now more complicated, sicker, are more variable in their presentations and have sustained many more previous procedures, thereby increasing their overall risk profile for the surgery. This trend will continue. New strategies will have to be developed to minimize the surgical risks, which likely will require increased amounts of technology, team surgery and advanced methods that will include biological adjuvants along with the anatomical interventions. For example, coronary bypass surgery may be combined with treatment of ventricular scars by replacing the dead tissue with living functioning myocardial implants and the reduction of electrical arrhythmias by MAZE-like management of the electrical substrates. Gene therapy will likely be cell-based and delivered during surgery to specific areas in the heart necessary to reduce the progression of artherosclerosis, to improve heart muscle function, or valve function. To reduce the downstream consequences of long standing heart disease such as pulmonary valvular regurgitation complicated by pulmonary hypertension, pulmonary valves (tissue engineered or native) may be seeded with genetically modified cells that secrete pulmonary vasodilating proteins directly into the pulmonary circulation. To a great extent, each of these operations will be a hybrid of surgical virtuosity, technical sophistication, and advanced biomedical and cell/tissue engineering methods. Achieving the expertise to manage these aspects of cardiac surgery of the future will require very different training than has been the case for cardiac surgeons over the past 40 years. 16 Sorting out competing strategies for the benefit of specific patients will require dispassionate and statistically rigorous assessments.

REFERENCES

1. Welke KF, Ferguson TB, et al. Ann Thorac Surg. 2004;77:1137-9.

2. Herbert MA, Prince SL, et al. Ann Thorac Surg 2004;77:1960-5.

3. Jones RH. J Amer Coll Card 2005;45:1517-28.

4. Torchiana DF, Meyer GS. J Thorac Cardiovas Surg 2005;129:1223-5.

5. Carey JS, Danielsen B, et al. J Thorac Cardiovasc Surg. 2005;129:1276-82.

6. Mack MJ, Herbert M, et al. J Thorac Cardiovasc Surg. 2005;129:1309-17.

7. Iezzoni LI. Ann Intern Med 1997;127:666-74..

8. Hannan EL, Kilburn H, et al. Med Care 1992;30:892-907

9. Dranove D, Kessler D, et al. J Polit Econ. 2003;111:555-89.

10. Hueb W, Soares PR, Gersh BJ. J Am Coll Cardiol 2004;43:1743-51.

11. Karamlou T, McCrindle BW, et al. CHSS Data Center Report, September 2005

12. Bennett JA. Curveball, baseball statistics and the role of chance in the game. Springer-Verlag, New York, 2001

13. Omar A, J Thorac Cardiovasc. Surg. 2004;77:2232-7.

14. Nilsson, Algotsson. J Thorac Cardiovasc. Surg 2004;77:12350 -40

15. Califf RM, "Stenting or Surgery" J Am Col Cardiol. 2005;46:5890 -91.

16. Kincaid EH, Atala A, Kon ND. CTSNet January 2005.

RICHARD A. HOPKINS, MD

Copyright Rhode Island Medical Society Jan 2006
Provided by ProQuest Information and Learning Company. All rights Reserved