‘Gaming’ the AUC for Revascularization? Study Suggests Not

PCIs for acute MI increased, but there was no uptick for unstable angina, an indication with considerable latitude for upcoding.

‘Gaming’ the AUC for Revascularization? Study Suggests Not

Concerns that physicians may be “upcoding” stable ischemic heart disease patients to an acute indication in order to meet appropriate use criteria (AUC) that would justify an intervention are unwarranted, according to a large analysis of Medicare beneficiaries and patients in the Veterans Affairs (VA) system.

Between 2009 and 2013, there was a 2.0% per year increase in the number of PCIs coded for an acute indication in Medicare patients who underwent coronary revascularization. The increase in procedures, however, was driven by PCIs for acute MI and not unstable angina. In the VA system, where hospitals aren’t graded by their adherence to the AUC, there was a 1.2% increase per year in the number of PCIs coded for acute coronary syndrome. Again, the increase was driven by an uptick in PCI for acute MI and not unstable angina.  

“I think it does give reassurances,” said lead investigator Elias Dayoub, MD (University of Pennsylvania Perelman School of Medicine, Philadelphia). “There had been some concern that there might be some degree of upcoding in order to meet the appropriate use criteria for PCI, particularly upcoding stable angina patients to unstable angina. Based on what we looked at, particularly being able to compare it to the [VA] population where there shouldn’t be a strong incentive for that behavior to occur, we didn’t see any differences in the way PCI was coded before and after the appropriate use criteria.”

The new study was published March 15, 2021, in Circulation: Cardiovascular Quality and Outcomes.

Potential to Game AUC

The AUC were developed by the American College of Cardiology and American Heart Association (ACC/AHA) in 2009 to curb overuse of PCI in patients with stable ischemic heart disease. Starting in 2011, the ACC’s National Cardiovascular Data Registry began providing reports to hospitals participating in the CathPCI Registry about how well they are adhering to the AUC.

Some of the earliest studies showed that the AUC led to a decrease in the total number of PCIs performed in the US, while others showed there was an improvement in the number of PCIs performed for appropriate indications, said Dayoub. Although they were initially lauded as a success, other data hinted that the AUC might lead to changes in coding, specifically upcoding from stable ischemic heart disease to unstable angina, rather than changes in physician behavior. In 2018, for example, a study by Rishi Wadhera, MD, and Robert Yeh, MD (both Beth Israel Deaconess Medical Center, Boston, MA), showed that after the release of the AUC for coronary revascularization, there was an increase in the number of outpatient PCIs classified as acute, mainly coded for unstable angina, in New York, Michigan, and Florida.  

“The thought was, perhaps, that we weren’t doing more appropriate PCIs but that maybe clinicians were changing the coding to make it look like they were doing more appropriate procedures,” said Dayoub.

With that in mind, the researchers evaluated temporal trends in PCI indication around the time of the AUC rollout. They used the VA healthcare system as a comparator because physicians and hospitals wouldn’t have the same incentive of altering coding to reflect adherence to the AUC. Using administrative healthcare data representing 5% of Medicare fee-for-service beneficiaries, researchers identified 87,464 PCIs performed between 2009 and 2013. Additionally, they evaluated 30,251 PCIs performed in the VA healthcare system.

In the Medicare cohort, the percentage of PCIs performed for acute MI increased from 31.9% to 41.0% and the percentage performed for unstable angina declined from 12.6% to 10.5% between 2009 and 2013. In the VA cohort, the percentage of PCIs for acute MI increased from 26.5% to 34.3% and those for unstable angina declined from 15.7% to 12.3% during the same time period.

Regarding coded indications for inpatient PCIs among the Medicare population, the proportion of procedures for ACS increased from 65.6% to 73.6% between 2009 and 2013. In the VA system, the proportion of inpatient PCIs coded for ACS increased from 64.8% to 69.2%. In both the Medicare and VA cohorts, the proportion of inpatient PCIs for acute MI increased while those coded for unstable angina declined. A similar pattern was observed among those treated as outpatients. The proportion of outpatient PCIs with an ACS indication increased from 20.9% to 28.3% in the Medicare cohort and from 24.8% to 29.6% in the VA patients. Again, in both cohorts, PCIs coded for unstable angina in the outpatient setting declined from 2009 to 2013, while those coded for acute MI increased.

In the difference-in-difference modelling, the proportion of PCIs coded for an acute indication increased by 1.8% more in the Medicare cohort relative to the VA group but this was not statistically significant (P = 0.21). Also, the change in PCIs coded for an acute indication between the Medicare and VA cohorts before and after AUC reporting was not significantly different when stratified by the inpatient versus outpatient setting.     

“What we ultimately saw was that there was no significant change between the two groups,” said Dayoub. “Overall, acute indications for PCI were increasing over time, but when we stratified that further, unstable angina, which is a bit of a ‘gray zone’ where it’s up to the clinician to decide, decreased over time.”

In other words, the publication of AUC grading did not appear to lead to an increase in the number of patients being “switched” from a stable ischemic heart disease indication to unstable angina, he said. The increase in PCI for acute MI is likely attributable to the introduction of high-sensitivity cardiac troponin testing, added Dayoub.

“Clever” Study 

David Brown, MD (Washington University School of Medicine, St. Louis, MO), called the new analysis clever, particularly in its use of the VA cohort as a comparator arm. Like Dayoub, Brown said the data, specifically the absence of an increase in the number of patients undergoing PCI for unstable angina, is a positive finding. As to why these data differ from the 2018 analysis by Wadhera and Yeh, Brown said it might be attributable to different patient or provider populations.

Dayoub noted that the previous study included patients from New York State, where the proportion of outpatient PCIs for ACS increased 14-fold after the introduction of the AUC. New York State had previously announced a plan not to reimburse PCIs deemed inappropriate—they never followed through with tying reimbursement to the AUC—and this might explain the different findings between studies.  

One aspect of the analysis that puzzled Brown, however, is the high percentage of outpatients with a diagnosis of acute MI. In the Medicare cohort, 14.7% of patients with acute MI were treated as outpatients in 2013, as were 20.5% of patients in the VA system. In the private sector, turning over patients quickly—in less than 48 hours—might make more sense given the financial incentives, but it’s confusing to see this high rate in the VA. While some patients might be fast-tracked, it’s typically not a Medicare or the VA/Medicare type of patient that goes home so quickly. Brown stressed that he’s not suggesting there’s upcoding going on, but simply that the high rate of outpatient PCI for acute MI is perplexing.

To TCTMD, Dayoub explained that “outpatient” is typically defined by the “two-midnight rule” and that contemporary studies within the same time period have shown that a similar proportion of patients undergoing primary PCI were discharged in 2 days or less. He also noted that the amount of time patients spend in hospital has declined over time, particularly with the increasing use of radial access, and this might explain why there has been an increase in acute MI patients treated as an outpatient. In focusing on the shorter outpatient stays, where some have suspected there might be the potential for upcoding because these patients appeared well enough to go home, they still didn’t observe any increase in coding for unstable angina, he said.

Michael O’Riordan is the Associate Managing Editor for TCTMD and a Senior Journalist. He completed his undergraduate degrees at Queen’s…

Read Full Bio
  • Dayoub and Brown report no conflicts of interest.



Michael Rinaldi

2 years ago
This study is flawed. You don't need to code unstable to get appropriate. Class 3 chronic stable angina is almost always appropriate. A study that analyses for "upcoding" based on "unstable" coding trends misses this critical nuance. Administrative data is almost always terminally flawed and uninterpretable. This is particularly true when there are incentives to code a specific way either for outcomes reporting or reimbursement. Some flawed data, many argue, is better than no data at all. This is wrong. Dirty data says whatever we want it to say and is misleading to the point it is never a reflection of reality. What is disappointing is that we all know this but there are so many people who make a career analyzing and publishing this stuff that they end up misleading the public