A New Dimension in the Relationship Between Procedural Volumes and Quality

Article, see p 458

Assessment of quality in the catheterization laboratory can be a tough task. Traditionally, as an outcome metric, procedural mortality has been the key variable of interest. Indeed, data from the early percutaneous coronary intervention (PCI) era suggest that more than half of the deaths after coronary angioplasty could be ascribed directly to procedure-related complications. Rapid evolution in technology and pharmacotherapy has been associated with a steady decline in mortality related to PCI.^1,2 In the current era, in-hospital mortality post-PCI for all-comers ranges from 1 to 2%, with the vast majority (>80%) occurring in patients undergoing urgent or emergent PCI, and with <10% directly attributable to a procedural complication.^3,4 It is thus less likely that additional measures undertaken to improve procedural quality will result in meaningful reductions in in-hospital mortality, and assessment of long-term outcomes has become increasingly important.

As a structural metric, volume remains an important component of quality assessment among patients undergoing PCI, similar to surgical procedures. Volume is intuitive, objective, and easy to collect and monitor. Hospital volumes, in particular, have been closely scrutinized and frequently driven policy and reimbursement decisions. For instance, 400 annual PCIs has long been held as a benchmark for PCI quality. In addition, the National Coverage Determination document by the Center for Medicare & Medicaid Services requires hospitals to maintain an annual minimum volume of 400 PCIs to be eligible for reimbursement for procedures such as transcatheter aortic valve replacement and MitraClip.⁵

Beyond the limitations of volume as a quality metric, which we have written about previously,^1,6–8 volume is also a complex metric to interpret. In their earlier work, Fanaroff, Rao, and colleagues⁹ provided a contemporary assessment of operator volumes for PCI in the United States, and the association between these volumes and in-hospital mortality. In that study, low- and intermediate-volume operators had a 16% and 5% higher odds of in-hospital mortality in comparison with high-volume operators, respectively. It is interesting to note that adjusting for important process measures such as the use of radial access significantly attenuated this association, suggesting that the contribution of process measures was fundamentally important. Another important insight from that article was the interdependence between hospital and operator volumes. The highest in-hospital mortality was observed among low-volume operators at low-volume institutions, while the lowest was observed for high-volume operators at high-volume institutions. The adjusted odds ratio for mortality for low-volume operators at high-volume hospitals was identical to that for high-volume operators at low-volume institutions.⁹

In this issue of Circulation, Fanaroff and colleagues¹⁰ now report an important follow-up analysis to their earlier article. By using linked National Cardiovascular Data Registry’s CathPCI Registry and Center for Medicare & Medicaid Services claims data, the authors assessed 723 644 PCIs performed by nearly 9000 operators in the United States between 2009 and 2014. Low-volume operators (defined as those performing <50 PCIs/y) were more likely than high-volume operators (defined as those performing >100 PCIs/y) to perform PCI for ST-segment–elevation myocardial infarction or other emergent indications. In-hospital mortality was marginally lower among high-volume operators than among low-volume operators (2.0% versus 2.4%; adjusted odds ratio, 0.82; 95% CI, 0.77–0.87), confirming their earlier findings in this linked data set.¹⁰

When considered in the context of the earlier discussed background, there are 3 important observations from this analysis that we seek to highlight. First, the authors provide longer-term follow-up data, thus allowing for a more thorough assessment of the quality of care administered during the index procedure. This is important because, in particular in the volume-outcomes space, there are few studies that extend follow-up beyond hospital discharge, particularly beyond 30 days.⁸

Second, for longer-term outcomes, this study highlights that operator volume is not a good predictor of quality, and once again brings into question its reliability as a structural metric for this purpose. In this analysis, at 30 days, there was no difference between intermediate- and high-volume operators for mortality; the difference between low- and high-volume operators was still noted, but this difference was significantly attenuated (adjusted odds ratio, 0.92). At 1 year, there was unequivocally no relationship between the 3 volume categories and risk-adjusted mortality (adjusted hazard ratio, 1.01; 95% CI, 0.97–1.05 for intermediate versus low; adjusted hazard ratio, 1.04; 95% CI, 1.00–1.08 for high versus low). Furthermore, an expanded major adverse cardiac events definition that included death, myocardial infarction, or unplanned revascularization did not show any differences based on operator volumes either (adjusted hazard ratio for major adverse cardiac events, 0.99; 95% CI, 0.96–1.01 for intermediate versus low; hazard ratio, 1.01; 95% CI, 0.99–1.04 for high versus low). This is intriguing because, although mortality could be influenced by patient-related factors even on longer-term follow-up, subsequent myocardial infarction and target lesion revascularization might be more likely to be influenced by the thoroughness of the PCI procedure itself. For instance, systematic undersizing of stents or underrecognition of issues such as edge dissections may indicate poor procedural quality/competence, and might be expected to result in higher revascularization and stent thrombosis rates.¹¹ Operator volumes, however, were unable to identify such issues.

Third, the analysis touches on yet another dimension in the volume-outcomes relationship, one of lifetime volumes or experience. These lifetime metrics seem fairly self-evident as potential indicators of operator quality, but as best as we can tell, lifetime experience has never been methodically studied like this before. By using National Provider Identifier numbers, operators were categorized into <2 years, 2 to 4 years, and >4 years of practice, defined based on the number of years that operators were submitting cases to the CathPCI Registry. In this analysis, nearly one-third of operators had been practicing for ≤4 years, with 1 in 10 PCI operators in practice for <2 years. There was an interaction between operator volumes and years of experience for in-hospital mortality. Among operators who had been practicing for at least 2 years, high-volume operators were associated with lower in-hospital PCI mortality in comparison with low-volume operators. However, as in the overall analysis, these differences were not observed on 1-year follow-up.

As can be seen from the Figure, there may be many different combinations when factoring in all 3 volume domains: annual hospital, annual operator, and lifetime operator volumes. These 3-way interactions are likely to be much more intricate and further complicate any consideration of one of them being the most definitive structural metric for quality assessment. In fact, even lifetime experience may be a nonlinear measure; for cardiac surgery, poorer outcomes have been reported for operators very early or very late in their careers.^12,13 In general, one can propose several arguments in favor of lifetime or cumulative experience for procedures. As with competitive athletes, the most obvious one is practice makes perfect. Tacit knowledge and skills accumulated over years of repetitive practice mean that experienced operators may be more adept at identifying and resolving complications than their more junior colleagues. Experienced operators are probably also more adept at patient selection (for instance, understanding when a procedure may be futile or inappropriate) and may feel less pressured to take on such cases than newer operators who may feel pressured to say yes to all cases sent their way. More seasoned operators may also understand coding and documentation nuances, and their implications on quality assessment, better.

• Download figure
• Download PowerPoint

However, some studies also suggest that older physicians who have been in practice for more years possess less factual knowledge and are less likely to adhere to appropriate standards of care.¹⁴ An underlying issue could be that physicians’ toolkits, which were created during training, are not updated regularly. Older physicians also seem less likely to adopt newly proven therapies and may be less receptive to new standards of care. A classic example of this is the slower adoption of transradial access for PCI among older operators.¹⁵

The authors should be congratulated for laying the initial groundwork for a more thorough evaluation of PCI volume as a quality metric in all its dimensions. Future studies need to build on this one and more deliberately evaluate the interplay between annual hospital, annual operator, and lifetime operator volumes, and their effects on risk-adjusted outcomes. In the quality-outcomes domain, our overreliance on procedural volumes has been a Faustian bargain; at a minimum, these complex associations need to be acknowledged and factored into future discussions about PCI volumes and quality.

Footnotes