Thomas R. McLean, MD, JD, Surgical Service, Eastern Kansas Health Care System, 4101 South Fourth Street Trafficway, Leavenworth, KS 66048; e-mail: email@example.com
Surgeons allege that fewer coronary artery bypass graft (CABG) operations are being performed because cardiologists are performing more percutaneous coronary intervention (PCI) procedures for monetary gain. The authors reviewed publicly accessible databases from New York State, California, Canada, and Europe for 1997 and 2003. In New York State, the number of CABGs performed fell by one-third. Patient demographics, however, remained unchanged. In contrast, while the number of PCIs substantially increased, the percentage of 3-vessel PCIs remained unchanged. In California, the rate of CABG production (per million population) fell by one-quarter, but it increased in Canada and Europe. In contrast, the rate of PCI production increased in all locations. The performance of fewer CABG procedures in America cannot be entirely explained by cardiologists' increasingly self-referring patients with coronary artery disease for PCI.
The introduction of percutaneous coronary intervention (PCI) resulted in more complex high-risk cases of coronary artery disease (CAD) being referred for cardiac surgery.1 The subsequent addition of bare metal stents (BMSs) to the PCI armamentarium, which largely eliminated the complication of acute coronary occlusion, correlated positively with a rise in the number of PCI cases and a fall in the number of coronary artery bypass graft (CABG) procedures.2 Presently, many cardiologists are extrapolating preliminary data for drug-eluting stent (DES)-assisted PCI to conclude that PCI “ought to be considered equivalent to, if not superior to, bypass surgery.”3 Cardiac surgeons believe that the optimism associated with stent-assisted PCI is misplaced and allege that cardiologists are performing PCIs in many patients with 2- and 3-vessel CAD for their own monetary gain.4
Another, less discussed explanation exists for the falling number of CABG procedures. Until recently, the United States was the only country in the world that published adverse surgeon-specific outcomes. Known as “reputational incentives,” this practice is designed to force low-quality providers from the market.5 When first introduced by New York State 2 decades ago, reputational incentives seemed to have had limited market impact.6 More recent data from New York State, however, suggest that reputational incentives do force marginal cardiac surgeons out of the market.7 So, perhaps fewer CABG operations are being performed because cardiac surgeons, wishing to avoid being listed at the bottom of public outcome data, are taking on fewer complex high-risk CABG operations. This study was undertaken to determine whether reputational incentives may be impacting the clinical decision making of cardiac surgeons.
Publicly available data from the New York State Department of Public Health8 concerning CABG (Cardiac Surgery Reporting System) and PCI (Percutaneous Coronary Intervention Reporting System) procedures performed in 1997 and 2003 were reviewed. New York's CABG data collection system is considered the gold standard for data concerning cardiac services. New York's registry for PCI, although not as well-publicized as the older CABG registry, collects and publishes analogous data concerning catheter-based intervention for CAD. The years 2003 and 1997 were selected because they allow for the most recent meaningful comparisons of CABG and PCI in several health care systems. The 6-year period of this study corresponds to the peak years of the use of BMS-assisted PCI.
Data concerning the rates of CABG procedures and PCIs in New York were contrasted with similar publicly available data from health care systems in California, Canada, and the European Union. Statistics for coronary intervention in California were obtained from the Office of Statewide Health Planning and Development. Unlike New York State's health care system, in which cardiac services are heavily regulated, in California cardiac services are subject to much less state regulation. As a consequence, in New York 78% of CABG procedures are performed at cardiac centers that do more than 500 CABG operations per year, while in California only 27% of CABG cases are performed at high-volume centers.9 The financial incentives provided by the federal government are virtually the same in New York and California, however.
Statistics for coronary intervention in Canada were obtained from previously published data.10 Like that of New York State, the Canadian system for cardiovascular services is highly regulated. In recent years, however, the Canadian provinces have liberalized the number of cardiac services that are to be reimbursed to help decrease waiting times.11 Statistics for coronary intervention in the European Union were obtained from a registry kept by the British Heart Foundation.12 The nations of the European Union, like the various states in America, have varying economic strengths and have varying degrees of control over their health care systems.13
Assumed in this analysis is that the evidence-based rules for managing CAD are similar worldwide. Briefly, CAD is managed according to the number of vessels that have hemodynamically significant stenotic regions. Patients with 1-vessel disease are managed medically or with PCI. In the BMS- and DES-assisted PCI era, however, controversy exists over how multi-vessel (ie, 2- and 3-vessel) CAD should be managed. This controversy arises primarily because studies of PCI and CABG have variable periods of follow-up after the index procedure (CABG or PCI). Thus, with some exceptions, in patients with multivessel CAD, PCI and CABG can be substituted for each other even though the individual risks of these procedures are not identical.14 Finally, although patients who have suboptimal outcomes from PCI are considered to have an indication for a CABG operation, cardiac surgeons are not required to operate on any patient with CAD if, in their professional opinion, the patient would not benefit from operation.
Data for the patients who underwent CABG operations in New York State in 1997 and 2003 are summarized in Table I. Between 1997 and 2003, the number of CABG operations performed fell by about one-third (20,220 CABG operations in 1997 vs 14,692 CABG operations in 2003). During both time periods, isolated CABG procedures (performed without concomitant operations on the aortic valves or heart muscle) accounted for approximately three-quarters of the total number of adult cardiac surgery volume (73.2% in 1997 vs 75.2% in 2003). Patient demographics were also similar during both time periods: female sex (28.7% in 1997 vs 28.4% in 2003), need for urgent/emergent surgery (2.1% in 1997 vs 1.8% in 2003), and ejection fraction <30% (7.1% in 1997 vs 7.0 in 2003). The number of redo CABG operations fell 50% from 1997 to 2003 (1335 or 6.4% of all cases in 1997 vs 632 or 4.3% of all cases in 2003).
Table I. A Comparison of CABG Procedures Performed in New York State in 1997 and 2003
The corresponding data for the PCIs performed in New York are summarized in Table II. In contrast with the falling number of CABG procedures, the number of PCI procedures performed increased by about 75% during the study period (29,516 PCIs in 1997 vs 50,046 PCIs in 2003). Like patients undergoing the CABG operation, the demographics of patients undergoing PCI were similar in 1997 and 2003: female sex (32.5% in 1997 vs 32.9% in 2003), need for urgent/emergent PCI (1.1% in 1997 vs 0.09% in 2003), ejection fraction <30% (3.6% in 1997 vs 4.3 in 2003), and 3-vessel PCI (15.4% in 1997 vs 14.5% in 2003).
Table II. A Comparison of PCIs Performed in New York State in 1997 and 2003
No. of procedures
Ejection fraction <30%
No. of vessels
Abbreviations: NP, data not provided by source; PCI, percutaneous coronary intervention.
Data for the number of CABG operations performed in each of the 4 health care delivery systems are summarized in Figure 1. The largest drop in the rate of CABG operations (per million population) occurred in New York State (1064 in 1997 vs 773 in 2003), followed by California (957 in 1997 vs 760 in 2003). On the other hand, the number of CABG operations performed increased in both the Canadian system (1000 in 1997 vs 1100 in 2003) and the European system (291 in 1997 vs 386 in 2003).
Data for the number of PCI procedures performed in each of the 4 health care delivery systems are summarized in Figure 2. In contrast with the experience with the CABG operations, the number of PCI procedures (per million population) increased in all 4 systems: New York (1553 in 1997 vs 2635 in 2003), California (1264 in 1997 vs 1700 in 2003), Canada (700 in 1997 vs 1400 in 2003), and the European Union (650 in 1997 vs 1034 in 2003).
Incentives for the Overprescription of PCI. The introduction of PCI to the cardiologists' armamentarium in the 1980s resulted in a fall in number of CABG operations performed.1 Presently, for most patterns of clinically significant CAD, stent-assisted PCI can be substituted for CABG, although controversy surrounds the long-term outcome of DES-assisted PCI.15 Because of this ability to substitute PCI for CABG, many cardiac surgeons have alleged that cardiologists are performing more PCI procedures and referring fewer patients for cardiac surgery.16 Cardiac surgeons have further opined that the cardiologists have (wrongfully) increased their own invasive treatment of CAD to compensate for a fall in their real income.17 Yet virtually nothing objective has been written to support this opinion.
Analysis of the relationship between the performance of PCI and CABG is complicated by the phenomenal market growth for cardiac services. In the current study, the total volume of cardiac services (PCI and CABG cases) performed in New York State increased 30% from 49,736 in 1997 to 64,739 in 2003. This growth rate, which is far in excess of the 1% to 2% population growth in New York State during this period, appears to be entirely due to growth in the PCI sector. During the 7 years of this study, not only did the number of PCI procedures performed increase by 75% but also the number of CABG operations fell by one-third.
Cardiologists, on the other hand, have attributed the growth of the PCI sector to the addition of stents to PCI procedures. In the view of cardiologists, stenting has made it possible for more patients with 2- and 3-vessel CAD to be treated with PCI rather than CABG operation. Yet the growth of the PCI sector cannot be entirely explained by cardiologists' self-referral patterns, for two reasons. First, hospitals give cardiologists incentives to refer patients to cardiac surgeons.18 Today, cardiologists are using more stents per patients than ever before19 to improve the long-term cost-effectiveness of stent-assisted PCI.20 Hospitals, however, are reimbursed per PCI procedure and not per number of stents used. The cardiologists' current stent/PCI utilization makes PCI a universal money-losing procedure for hospitals.2 The downward pressure that hospitals place on cardiologists not to overprescribe stent-assisted PCI is a likely reason that utilization studies from the United States21,22 and Europe23 have demonstrated that cardiologists are appropriately referring patients for PCI and CABG operations.
The second reason that growth in the PCI sector cannot be entirely explained by the self-referral patterns of cardiologists concerns the volume of CABG procedures performed in the United States and other countries. Based on patient demographic shifts observed after PCI first became available, if the fall in the number of CABG operations being performed today was solely due to greater utilization of stent-assisted PCI, one would expect the number of CABG cases to fall proportionately to the gains in PCI procedures and that those CABG cases that were performed would be more complex.
Neither of these postulates could be confirmed here. During the period of this study, the number of PCIs performed in New York State increased by 20,530. Had all of these new PCI cases occurred because the cardiologists elected to perform PCI rather than refer the patient for a CABG operation, the growth in PCI procedures would have entirely erased New York State's 1997 market of 20,220 CABG cases. Nor did the CABG operations performed in 2003 appear to be more complex. Compared with patients who underwent CABG in 1997, patients who underwent CABG in 2003 were almost identical in terms of percentage of women, isolated CABG, and requirement of an urgent/emergent operation. In fact, one could argue that the CABG cases that were performed in 2003 were less complex. During the 7 years of this study, the number of redo CABG operations in New York State fell 50% from 1335 to 632.
These data suggest that although 5528 fewer CABG cases were performed in New York State, growth occurred in this sector. One source for new CABG case referrals to cardiac surgeons is the PCI procedure itself. Recall that failure of PCI to correct a patient's CAD is an indication for surgery. Although the PCI failure rate is low, with so many more patients receiving PCI, the number of CABG operations performed for early PCI failure (0–2 days) could have stimulated growth in the CABG sector. In this study, early failure of PCI resulted in 125 more patients being referred for surgery in 2003 than in 1997. Late failures of PCI (>2 days) may have also increased the number of referrals; however, the magnitude of growth of the CABG sector because of late failure of PCI cannot be determined from the available pubic information.
In addition, evidence from other health care markets demonstrates that the CABG sector of cardiac services can grow in parallel with the PCI sector. During the period of this study, Canada simultaneously increased its per capita rate of CABG operations by 10%, while doubling the number of PCI procedures performed. Similarly, between 1997 and 2003, the European Union increased its per capita performance of CABG operations by 32%, while increasing the number of PCI procedures performed by 50%.
These observations raise an interesting question: what economic incentives allowed the number of CABG operations performed to increase in Canada and the European Union but to fall in New York State? An answer to this question is suggested by the data from California. Arguably, the laissez-faire market for cardiac services in California is substantially different from the regulated one of New York State. Yet, market shifts in cardiac services observed in California (ie, increasing numbers of PCI procedures/falling numbers of CABG operations) were similar to those noted in New York State. The different growth rates in PCI and CABG performance found in New York, California, Canada, and the European Union could be due to differences in national health care policies. Because the financial incentives in the United States are heavily influenced by Medicare reimbursement benchmarking, one might argue that the divergent finding between New York/California and Canada/Europe with regard to cardiac services is explainable by physician reimbursement. This may be true; however, a more likely explanation can be found in the use of reputational incentives in the United States.
Reputational Incentives. New York State pioneered the use of reputational incentives. Today, a number of states, including California, annually publish physicianspecific outcomes for cardiac services. Other countries, however, have been slow to adopt the use of reputational incentives. Although institution-specific outcomes data for the management of CAD have been available for several years in the European Union, only in April 2006 did the United Kingdom begin to publish surgeon-specific outcomes data for CABG cases.24 Canada, too, has demonstrated a reluctance to publicly disseminate physician-specific outcomes data for cardiac services.25
These national differences in the use of reputational incentives are important because they suggest an explanation as to why CABG volumes are moving in different directions in different health care systems. First, although public reporting of institution-specific outcomes data seems to have little effect on a hospital's market share,6 the evidence in favor of market impact for the use of reputational incentives at the provider level, once considered to be minimal, is becoming increasingly positive.
An early study of Medicare patients in New York State by Peterson and colleagues6 demonstrated that the use of physician-specific reputational incentives (transiently) stimulated patients to seek out-of-state surgeons. Patients' flight to states adjacent to New York for cardiac services proved short-lived, however, perhaps because the first of January publication of mortality rates for cardiac surgeons was forgotten by spring.
Yet, since Peterson published the paper a decade ago, the health care market has changed. The patient safety movement has made it virtually impossible for anyone who reads the newspaper not to hear, on an almost daily basis, how medical errors cause unnecessary patient deaths. Also, making physician-specific outcomes data available on the Internet means that patients who require a CABG in the summer or fall no longer need to go to the library to research their surgeons. These changes and others mean that today's patients no longer blindly accept another doctor's recommendation of a surgeon. Rather, today's patients seem to be almost routinely reviewing a surgeon's clinical report card before agreeing to undergo a surgical operation.
Thus, it is not surprising that more recent studies concerning reputational incentives have demonstrated that the publication of adverse physician-specific outcomes data does have market impact both in America and abroad.26 Jha and Epstein,7 for example, have demonstrated that there is a relationship between a surgeon's ability to work in a community and the dissemination of adverse surgeon-specific outcomes data. In their study, Jha and Epstein examined the reasons that surgeons' names disappeared from the New York State's cardiac surgery reporting system.7 These investigators found that 20% of cardiac surgeons in the quartile with the highest mortality rate ceased practicing in New York during the next 3 years. By comparison, only 5% of cardiac surgeons in the quartile with the lowest mortality rate ceased practicing in New York during the same time period.
Data from the current study compliment the work of Jha and Epstein.7 The negative publicity associated with having one's name appear in the bottom half of a cardiac surgery registry is likely to make a surgeon more reluctant to offer a CABG operation to a patient with a moderate or high risk of death. Fear of being publicly listed as a low-quality surgeon could explain (1) why fewer surgeons in 2003 were willing to take on the more technically demanding challenges of a redo CABG operation and (2) why the patient demographics associated with the risk of operation (eg, female sex, ejection fraction) were not substantially worse than corresponding parameters in 1997. In short, because reputational incentives are sufficient to motivate some surgeons to exit the market, it is reasonable to conclude that reputational incentives are also sufficient to make some cardiac surgeons who want to remain active in New York and California to select their cases with more caution.
Conversely, the lack of use of reputational incentives in Canada and in the European Union offers an explanation of the market shifts in cardiac services in these countries. In New York and California, where reputational incentives are actively employed, surgeons are given an incentive not to operate on high-risk patients. Absent the negative impact of reputational incentives, surgeons in Canada and Europe were freer to earn their living by operating on patients with a greater risk of death. More generally, absent a disincentive to operate, surgeons are given a green light to operate on any case that is referred to them. This is a formula for the overutilization of cardiac services, and it likely accounts for why the number of CABG operations and PCIs are simultaneously increasing in Canada and Europe.
Parenthetically, reputational incentives may explain why cardiac services grew faster than the population in New York State. It is possible that the superior outcomes of surgeons in the top quartile benefited from the publication of positive outcomes. Patients in states adjacent to New York with less well-developed systems for disseminating surgeon-specific outcomes may be migrating to New York State to undergo elective CABG operations from top-rated surgeons. That is, a state's use of reputational incentives is likely to act as a form of free advertising for exceptional cardiac surgeons.
Limitations. This study cannot definitively explain why cardiac services are growing faster than the population in New York State. Nor is the data contained herein sufficiently rigorous to determine whether reputational or some other incentives are responsible for the different growth rates of PCI and CABG procedures in the United States and other countries. Until more sophisticated studies can be undertaken, however, this study remains as one of the few population-based studies on the relationship between PCI and CABG utilization.
In the current era of stent-assisted PCI, the number of CABG operations being performed in New York State and California has fallen substantially. Cardiac surgeons, with few objective data, have asserted that the falling number of CABG cases is due to the greed of cardiologists. In the view of cardiac surgeons, cardiologists acting in their own self-interest convince patients with 2- and 3-vessel CAD to substitute stent-assisted PCI for CABG. While the PCI-CABG substitution theory may account for some decrease in the number of CABG operations being performed, this theory does not account for other observations including (1) why cardiac surgeons seem less willing to offer patients a redo CABG operation and (2) why the number of both CABG and PCI procedures are simultaneously increasing in Canada and the European Union. These observations suggest that factors other than the self-interest of cardiologists are impacting the number of CABG operations being performed in New York and California.
While many economic factors may influence the number of CABG cases performed, the use of reputational incentives in the United States is one factor that helps to explain the data found in this study. It is possible that fewer CABG operations are being performed in states that publish adverse surgeon-specific mortality data because surgeons in these states no longer take on moderate- to high-risk patients. Conversely, surgeons in Canada and the European Union, where reputational incentives were not in use during the study period, were freer to overprescribe the CABG operation. If reputational incentives do influence the behavior of surgeons as postulated in this article, the number of CABG cases per capita performed in Canada in the European Union over the next decade is likely to fall as these countries adopt the use of reputational incentives to control overprescription of health care services and to remove some truly bad physicians from the market.
The author wishes to acknowledge the assistance of Patrick B. McLean in the preparation of this manuscript.