Is Onsite Surgery Backup Necessary for Percutaneous Coronary Interventions?

The number of percutaneous coronary intervention (PCI) procedures performed each year has increased substantially, particularly since the introduction of coronary stents. The current increase is being fueled by the shift from thrombolysis to the performance of primary PCI for treatment of acute myocardial infarction (MI). Also, with the availability of drug-eluting stents, patients who previously would not have been eligible for PCI are now candidates for this intervention. The drive to treat patients with acute ST-elevation MI (STEMI) in a timely manner poses the question of whether many more moderate-sized and often suburban hospitals, which in most cases do not have cardiac surgery onsite, should provide primary PCI for patients with STEMI. On the surface, it seems unnecessary to provide surgical backup onsite in such programs because the need for emergency surgery for failed PCI for STEMI is small (1% or less).^{1 - 2} It would seem adequate to have an emergency plan for patient transfer to a tertiary cardiac hospital in the unlikely event that surgery was needed. Earlier smaller series of patients treated with primary PCI at centers without cardiac surgical backup have shown that this procedure can be done at such facilities and the outcomes are good and comparable to results in centers with onsite surgery.^{3 - 11}

In this issue of JAMA, Wennberg and colleagues¹² raise serious questions about this new strategy. Using the Medicare Provider Analysis and Review database of patients treated from 1999 through 2001, the authors found a 29% higher mortality rate in patients having PCI procedures in hospitals without surgical backup. Clearly, if these findings are true and representative, it is necessary to reconsider and perhaps curtail expanding the availability of PCI.

As witnessed by the increasing availability of many tertiary diagnostic and therapeutic procedures, such as CT scanning, magnetic resonance imaging, cancer treatment centers, complex orthopedic surgery, and even neurosurgery at more and more hospitals, patients’ first choice is to be hospitalized in local hospitals close to their families and homes. Patients prefer to be treated by and receive the advice of their primary care physician and generally prefer to be treated in a hospital that has access to their medical records so that potentially better decision making takes place and unnecessary duplicate testing is avoided. These considerations are often lost when patients are admitted to a more remote tertiary center, where medical records often are not readily available and communication with the patient’s physician is often after the fact, if it occurs at all.

Most hospital administrators also encourage providing tertiary services at their local hospital. Under the current health care reimbursement plans, tertiary services such as PCI are necessary to keep the hospital bottom line healthy and to offset often inadequately reimbursed admissions of general medicine patients.^{13 - 14} Local physicians also have a vested interest, not only financial but in terms of convenience and time to care for the patient in the facility where they spend most of their time.

However, patients want the best possible outcomes at the lowest possible cost. Unknowingly, patients may assume that the quality of service is equal among all hospitals offering a service. Also, with the exception of the few outcome measurements for cardiac diseases and procedures, surgical infection rates, and hospital-acquired pneumonia, there are few indicators to measure hospital quality. Unlike many patients, most physicians who undergo a potentially disabling or life-threatening procedure would, if given the choice, select an experienced (ie, high-volume) operator and a high-volume medical center because at least for many operative procedures, higher volumes have been shown to be associated with better outcomes.^{12 ,15 - 17}

The study by Wennberg et al does not prove that the proliferation of cardiac programs without surgical backup must stop, but their findings are provocative and demonstrate the need for more data. By design, data from an administrative registry such as used in this study have the potential of being inadequate to adjust for important patient differences that could affect the findings. One possible hint that this may be present in this analysis was that patients who were admitted to hospitals without onsite surgery and who received the procedure were more likely to have acute MI (51% vs 29%) than those admitted to hospitals with onsite surgical hospitals. Higher mortality would be expected in this subset of patients receiving elective PCI procedures and hence possibly could explain the worse results.

Primary or rescue PCI was also more common at centers without onsite surgery (22% and 5.6%, respectively), suggesting that many PCI programs existed primarily to provide primary PCI. Moreover, the heterogeneity of outcomes for patients having primary or rescue PCI compared with elective or later procedures was dissimilar. After adjusting for baseline differences, mortality rates tended to be slightly (but not significantly) better in the hospitals without onsite surgery for primary PCI, whereas mortality for those treated after day 1 was 38% worse in the same hospitals. These dissimilar findings call into question whether patient differences rather than procedural and health care system differences actually influenced the findings. It seems unlikely that the interventional cardiologist would be a poor operator for one kind of procedure and a superior operator for the other. Primary PCI is often a more difficult procedure and in the setting without onsite surgery, case selection for patients not needing emergency treatment usually favors selecting lower-risk cases, which would tip the balance in the direction opposite from that observed. The same is true for the hospital staff and other support infrastructure—it is unclear why it was better in one situation and worse in the other.

Since the publication of several studies showing improved outcome from acute MI with PCI compared with thrombolysis, there has been increased “pressure” to expand PCI to many more hospitals to treat this condition quickly.^{18 - 19} It is probably not ideal to have these facilities performing only primary PCIs. If confined to performing only primary PCI, centers would always be relatively low volume. With few exceptions, it would be difficult to develop an experienced and dedicated team that only performs primary PCI. Also, requiring surgical backup at these centers is also likely to result in small-volume cardiac surgical programs. Diluting the already declining cardiac surgical experience to more hospitals in a geographic area will undoubtedly lead to lower surgical volumes for all and to worse outcomes. In some states, legislation mandates surgical backup for performance of PCI, and this regulation serves to increase the number of small surgical programs. It also does not make sense to develop PCI programs that will not meet the ACC/AHA (American College of Cardiology/American Heart Association) guidelines for the minimum number of cases (400 per year and 75 per operator).²⁰

Transfer of patients with acute MI from hospitals without PCI programs to those with them can be achieved with good results. The largest trials showed improved outcome for patients transferred even after median delays of 90 minutes or more compared with treatment with thrombolysis in outside hospitals.^{21 - 22} Many programs have streamlined the transfer process for patients from outlying facilities and therefore have reduced the need to establish more PCI programs. Patients require education that there is more to medical care than just convenience. If given the choice of undergoing a procedure where outcomes might be worse nearby home vs being treated at a more distant, larger, and possibly more impersonal center with potentially better results, most patients would still choose the latter if fully informed.

The findings by Wennberg et al¹² need to be replicated before policy changes are recommended. Although the Medicare database is large, the patients treated in centers without onsite surgery represented only 1% of the total. In a recent report from a large registry study, Kutcher et al²³ examined data from 278 000 patients undergoing PCI, 1% of which were performed in facilities without onsite surgical backup. The authors found an imbalance in the number of patients treated having acute MI, a strong predictor of survival, with more PCIs performed in centers without onsite surgery. There was a nonstatistically significant trend toward increased in-hospital death for patients treated in the facility without onsite surgical backup (mortality, 1.7% vs 1.3%, respectively P = .14). Mortality in patients with non-STEMI was higher in hospitals without onsite surgery but was almost identical for those patients with STEMI. When adjusted for prior histories and demographics, there was no difference in outcome in the 2 types of centers for both patient groups. There were, however, more deaths in patients referred for emergency surgery in centers with onsite surgery (9% vs 0%), possibly reflecting that patients with more futile outcomes may have been sent for surgery when it was readily available. Only 0.4% of patients treated in facilities without onsite surgical backup required emergency surgery. Every interventional cardiologist realizes there are occasional complications from the procedure that cannot be predicted and for some patients, emergency surgery might be lifesaving.

For now, the available data are not sufficient to require halting the performance of PCI in centers without surgical onsite backup. It does not make sense, however, to begin new PCI programs that cannot meet the procedural volume guidelines for PCI. For current PCI programs, it is mandatory that each center compare its results with others using national benchmark data, such as from the National Cardiovascular Data Registry, to ensure high-quality clinical outcomes.