“Inappropriate” PCIs in the U.S.: Appropriate Use Criteria, Referrals to the Cath Lab, and the Nuances of Intervention

In July 2011, the Journal of the American Medical Association published “Appropriateness of Percutaneous Coronary Intervention,” a look at 500,000 PCIs at 1,000 hospitals from 2009-10.¹ Data was gathered from the American College of Cardiology’s National Cardiovascular Data Registry (CathPCI). Senior author Dr. Paul Chan reflects on the study and where we go from here.

The study had good news.

In discussions with various news reporters and writers, I have found some people think it is good news and some think it is bad news. It is probably a little bit of both.

We were interested in getting a more comprehensive sense of the use of angioplasty in the United States. There have been prior studies that have looked at whether PCIs were appropriate or not, but many of those studies used the authors’ own appropriateness assessments. What we wanted to leverage this time was a national set of appropriate use guidelines from organizations such as the American College of Cardiology (ACC) and the Society for Cardiac Angiography and Interventions (SCAI).² We set forth to match percutaneous coronary intervention (PCI) procedures from a large national registry, the NCDR’s CathPCI Registry, to indications from the appropriate use criteria. The rationale for doing so was to get a better sense of contemporary rates of “appropriate,” “uncertain,” and “inappropriate” PCI use, and to better understand how we can improve, moving forward. To date, no other subspecialty in medicine has been as self-reflecting and self-critical as cardiology in looking at how we can do better and improve quality of care.

We found a difference in the performance of acute and nonacute procedures. By an acute procedure, we mean procedures done for myocardial infarction (MI) or unstable angina with high-risk features. Acute procedures, at least in the CathPCI Registry, comprised 70% of the procedures that we were able to map to the appropriate use criteria. The vast majority of acute procedures were considered appropriate, with very few (just over 1%) being considered inappropriate. That’s probably nothing to be surprised about, because we know that PCI in these settings can improve survival and prevent MI. In the elective, non-acute setting, the landscape is different. The vast majority of cardiologists would probably agree that the major benefit of PCI in the non-acute setting is symptom relief and improvement in quality of life.  Of the procedures we could map in the non-acute elective setting, we found about half were appropriate, while 38% were of uncertain appropriateness, and 12% were inappropriate.

One of the things that is challenging about the appropriate use criteria is the nomenclature that is used. The three categories of “appropriate,” “uncertain,” and “inappropriate” can often be confusing, because they each have lay connotations. In the appropriate use criteria, what we mean by an appropriate procedure is one where in which PCI is expected to confer a definite or a likely benefit to patients. An uncertain procedure suggests that PCI may confer a possible benefit to patients, but the weight of the evidence is not as substantive to support a definitive benefit. An inappropriate procedure means there is unlikely to be benefit from PCI in those circumstances.

There were over 1,000 hospitals that submitted data during the time period that we were looking at (2009-10). One of the most important findings from our study was that, in the elective setting, the inappropriate rates at the hospital level varied a great deal. We were not surprised to find variation, but we were surprised at the wide variation in the rates of inappropriate procedures across hospitals. Some hospitals did very well. In the elective setting, a quarter of hospitals had rates of inappropriate cases at 6% or lower. However, another quarter had 16.7%, or 1 out of 6 elective procedures, that were inappropriate. In some hospitals, the inappropriate rate for elective PCIs was even much higher than 16.7%. We believe that the true value of the appropriate use criteria is in the hospital-level analysis — to examine the degree of variation across hospitals. We hope people will look at whether or not their hospital is an outlier compared to the national average for inappropriate PCI, and specifically, whether or not their hospital is a high inappropriate outlier. If it is, then that hospital’s leadership may need to pause and reflect on what practice patterns are occurring that might lead to many of its facility’s inappropriate cases. We have some theories as to why that may be occurring, but it’s hard to know without looking at it on a case-by-case basis.

The JAMA article notes that “it’s likely that clinician factors are responsible for many of the inappropriate procedures.”

There are at least two major factors at play. First, interventional cardiologists themselves do not have a consensus as to what is appropriate and inappropriate. In 2008, we gave an appropriate use survey to cardiologists in 10 institutions ranging from Yale and Washington University (some of the big academic institutions), to community hospitals.³ Half the respondents were interventionalists and half were not. We asked them to rate the same clinical scenarios as the appropriate use criteria technical panel. This survey was done before the appropriate use document came out in 2009,² so, in effect, the responses by the cardiologists were ‘blinded’ to the appropriate use criteria ratings. We found that respondents often couldn’t agree, even amongst themselves, whether or not a procedure that was rated as appropriate, uncertain, or inappropriate by the technical panel should be appropriate, uncertain or inappropriate. The rates of agreement of individual cardiologists with the technical panel for the 68 scenarios that we surveyed ranged anywhere from the Kappa statistic 0.05 (which meant that cardiologists almost never agreed with the appropriate use criteria ratings) to 0.78. There was no cardiologist who had full agreement with the technical panel. Even for “appropriate” ratings, we found that there was non-agreement among the cardiologist group. So, we essentially found that, among cardiologists who are doing these procedures, there is not consensus as to what would constitute an inappropriate, uncertain, or appropriate procedure, which certainly might lead to some variation in practice.

Second, when we consider how to address and lower rates of inappropriate PCI, we need to remind ourselves that the interventionalists are not the only players needing to be proactive, because frequently they are on the ‘distal’ end of the referral base. Interventionalists often have not met the patients on whom they are doing procedures, because these patients are referred by non-interventionalists. A certain proportion of the inappropriate cases may be patients who are being cared for by very well intentioned invasive and non-invasive cardiologists. For example, you may have a patient who undergoes a screening coronary calcium study or a high sensitivity C-reative protein (CRP) test, because he just turned 50 and has coexisting hypertension and diabetes. The patient is asymptomatic, but went to his primary for a physical. The primary then referred him to cardiology for risk stratification. The patient is screened with a non-invasive stress test. Perhaps he had abnormalities on the stress test and is eventually referred to the cath lab. When the interventionalist does the cath, they see a 70% mid-circumflex stenosis and do an intervention, because the referring doctor had sent this patient to them. This typical scenario shows a disconnect in how patients present, are evaluated, and then are intervened upon in the cath lab. It may explain why we are seeing some proportion of these inappropriate cases in the cath lab. It would help explain how is it possible that an asymptomatic patient winds up in the cath lab. We see it all the time, because a patient had a stress test that showed either intermediate or some minor ischemia, and somebody wanted to be safe and thorough, and sent the patient for a cath. Therefore, if we are to be serious about addressing inappropriate rates, we need to rethink where change needs to happen. Change partly needs to happen in the cath lab, but it also needs to happen at the primary care and the non-invasive cardiology level. Who is sending these referrals, and if patients are getting stress tests, are they appropriately getting stress tests? For an elevated CRP or an elevated coronary calcium score, the appropriate thing to do is probably to put the patient on a statin and a baby aspirin, have them lose 20 lbs, and stop smoking. But do they need a stress test? There is no evidence, per se, to support the use of a stress test in these circumstances if they are not having symptoms.

So intervention in the case example you provided, with a 70% stenosis, is considered inappropriate?

Yes, if it is a one-vessel disease, and not a proximal left anterior descending or left main coronary artery stenosis, it is inappropriate if the patient is asymptomatic, and has either a low risk or intermediate risk stress test.

How much can technology like intravascular ultrasound (IVUS) or fractional flow reserve (FFR) alleviate some of these questions?

FFR is used in borderline lesions, but in this study, we are not dealing with borderline lesions. They are 70% stenoses or greater. We do have some indications for 50-60% lesions where FFR or IVUS are used. But the vast majority of these procedures, and certainly in the vast majority of inappropriate procedures, there was no use of FFR or IVUS at all. Whether or not there is a role for FFR and IVUS for a patient with a ≥ 70% stenosis and a stress test result is unclear.

What about cost savings?

We are certainly not talking about rationing for the sake of rationing in this paper. We are really addressing whether or not we are providing benefit to patients if we are going to do a procedure. It is certainly important from a patient’s perspective if the procedure incurs costs yet does not lead to a clinical benefit. When we do these procedures, we do need to keep in mind whether we are providing benefit to patients. The appropriate use criteria address circumstances when patients are not likely to receive any benefit at all. 

Also, keep in mind that the goal of the appropriate use criteria is not to have a zero percent inappropriate rate. The people who produced the appropriate use criteria recognized early on, and state this explicitly in the document, that the appropriate use criteria cannot capture the myriad of nuanced indications a physician is confronted with in routine practice. There are going to be extenuating circumstances and we try to address this in our discussion. There are times when an otherwise inappropriate case might be okay to do, and it might be more uncertain in terms of benefit to patients. However, this does not explain the huge variation in rates across hospitals. It is unlikely that Hospital A, which has an inappropriate rate of 29% compared to Hospital B, which has an inappropriate rate of 8%, nearly a 4x rate, will have 4x the number of extenuating circumstances. It is unlikely that the patient population and coronary anatomy differs that much in a random sampling of people who present to hospitals. It’s just not conceivable. We are not comparing Florida with South Dakota. These variations exist within the same state across different hospitals.

What has been the interventional community’s reaction?

Interactions both in my own institution and certainly outside my own institution have been fascinating. It reminds me of going through the four stages of grief. There is denial. You ask a room of 100 interventionalists, “How many of you have ever seen an inappropriate interventional procedure performed on someone?” Almost everyone will raise their hand. You ask, “How many of you have ever performed an inappropriate procedure?” The brave ones will raise their hands. At first, there was denial and anger about the paper — how dare anyone look at my own practice when I think I am providing the best care I can? Yet, as I have talked with groups around the country, there is a gradual acceptance of the merits of taking a hard look at appropriate use.

The appropriate use criteria are not perfect. They will not always reflect the absolute number of inappropriate cases. All the hospitals now on the NCDR are getting quarterly reports with an actual line item list of patients who are deemed to have had an inappropriate PCI. Hospitals are encouraged to go back and look at what is happening in their own groups. In our practice, we have done this as well. People are now recognizing that there are cases that, even for the most stalwart of interventionalists, are not necessarily appropriate. When people are presented with the rates at the hospital or operator level, it does create a defensive response at first, and that is natural. However, when I talk with the interventionalists, I remind them that this is not only their problem. This is a problem of how we conceptualize risk stratification in cardiology. We need to engage not only interventional but also non-interventional people in order to think about how we refer patients to cath labs. At our institution, we are piloting a new pathway through some of the electronic medical records (EMRs). When a person refers a patient to our cath lab for a diagnostic and possible therapeutic intervention, they need to state explicitly the patient’s symptom burden and as part of the order set, the ischemia level. Referring physicians now have to make explicit what is going on with their care of the patient. We are hoping it will prompt people to do a double take. The physician will say to himself or herself, ‘Wait a minute, she has Class I symptoms and is low-risk. Should I really be taking this patient to the cath lab?’ Within our EMR, we are also building a way to tell the physician ordering an elective cath, based on the symptoms, the medical therapy and the stress test results, what the likelihood of an appropriate, inappropriate, or uncertain procedure would be based on the coronary anatomy once the patient gets to the cath lab. That can also be taken back to the interventionalist when they do the procedure. We are certainly not asking interventionalists to say the appropriate use criteria are the ten commandments and be robotic. The document, however, should serve as a guide for all to reflect on the appropriateness of the procedure, depending on the coronary anatomy.

These are some of the first steps, along with a quarterly review of the NCDR CathPCI report, going case by case over all of the inappropriate cases. We need to have free discussions and ask: can we make an improvement? Can we do better? We are trying to pilot some of these things in our own institution to test their usefulness.

What has been the impact of the ACC-NCDR in general on the cardiology community?

The creation of these registries was, in many ways, one of the most important steps in looking at and performing outcomes research in cardiology. It is not just for PCI and appropriate use. There have been previous studies out of the CathPCI Registry looking at bleeding rates and the difference in bleeding rates⁴, complication rates from PCI⁵, and misdosing of glycoprotein IIb/IIIa inhibitors or other agents in patients with renal disease (published in JAMA⁶). There was an article in late August in Circulation celebrating the advances we have made in door-to-balloon (D2B) times.⁷ A decade or two ago, we thought there was no way you could get below 90 minutes. The vast majority of institutions had mean D2B times of 110- to 120-minutes. There are times that the NCDR will come out with articles that can be provocative and can make some of its membership concerned or worried. In this case, if we change PCI volume, does it change revenues to physician groups or to hospitals, and what does that mean long-term?

Still, the NCDR and the group effort behind it has fostered the expectation that we keep moving the quality yardstick one step further, with the goal of improving the care that we provide to our patients. Ultimately, that is really what it is about — ensuring safety and the delivery of the highest quality care we as a profession can provide. Most hospitals that participate in the NCDR value the comparative numbers that their participation provides them. It allows them to get a good handle on how they are doing compared to other hospitals nationwide.

Dr. Paul Chan can be contacted at paulchan.mahi@gmail.com.

Learn more about the NCDR CathPCI Registry at https://www.ncdr.com.

References

1. Chan PS, Patel MR, Klein LW, et al. Appropriateness of percutaneous coronary intervention. JAMA 2011;306(1):53-61.
2. ACCF/SCAI/STS/AATS/AHA/ASNC 2009 appropriateness criteria for coronary revascularization. J Am Coll Cardiol 2009; 53:530-553, doi:10.1016/j.jacc.2008.10.005
3. Chan PS, Brindis RG, Cohen DJ, et al. Concordance of physician ratings with the appropriate use criteria for coronary revascularization. J Am Coll Cardiol 2010;57(14):1546-1553,  doi:10.1016/j.jacc.2010.10.050
4. Marso S, Amin A, House J, et al. Association between use of bleeding avoidance strategies and risk of periprocedural bleeding among patients undergoing percutaneous coronary intervention. JAMA 2010;303(21); 2156–2164.
5. Roe MT, Messenger JC, Weintraub WS, et al. Treatments, trends, and outcomes of acute myocardial infarction and percutaneous coronary intervention. J Am Coll Cardiol 2010; 56:254–263, doi:10.1016/j.jacc.2010.05.008
6. Tsai TT, Maddox TM, Roe MT, et al. Renally cleared anti-thrombotic agents among dialysis patients receiving percutaneous coronary interventions: insights from the ACC-NCDR Cath-PCI Registry. JAMA 2009;302(22):2458–2464, doi:10.1001/jama.2009.1800
7. Krumholtz HM, Herrin J, Miller LE, et al. Improvements in door-to-balloon time in the United States, 2005 to 2010. Circulation 2011;124:1038–1045, doi:10.1161/CIRCULATIONAHA.111.044107