In this large, national registry of PCI procedures, we evaluated the association between a hospital’s proportion of inappropriate PCI in non-acute settings – as defined by the AUC – and traditional performance measures of processes of care and post-procedural outcomes. In 203,561 PCIs from 779 hospitals, we found no relationship between hospital tertiles of inappropriate PCIs and in-hospital mortality, periprocedural bleeding, or medical therapy at discharge. These findings suggest that PCI appropriateness, relative to processes of care and post-procedural outcomes, measures a different aspect of PCI quality. Furthermore, the large hospital-level variation in the proportion of inappropriate PCIs for non-acute indications suggests that there are significant differences in the quality of patient selection for PCI across facilities that are unrelated to how well the procedure is performed. Therefore, measurement of PCI appropriateness and post-procedural outcomes are equally important to informing PCI quality.
Although AUC have been developed for a number of diagnostic and therapeutic procedures in cardiovascular medicine,4, 13–15
to our knowledge, this is the first study to assess the relationship between facility-level procedural appropriateness and traditional metrics of procedural quality. Although appropriateness assessment, processes of care, and post-procedural outcomes are all quality measures for PCI, the systems required to improve quality in these domains are likely very different (). Hospital systems to ensure proper patient selection likely include decision-making tools and interventions prior to patient arrival in the cardiac catheterization laboratory. Among patients being considered for non-acute PCI, this may include ensuring an adequate assessment of ischemic risk, a trial of robust anti-anginal medications prior to PCI, since medications alone may alleviate patients’ angina,16
and the avoidance of revascularization in asymptomatic patients. Furthermore, proper patient selection to avoid inappropriate PCI may be disincentivized by monetary reimbursement, referral structures and the expectation of colleagues, or abundance of catheterization facilities and interventional cardiologists.17
Systems to support high-quality patient selection are likely unrelated to systems that ensure minimization of procedural complications and promote high-quality post-procedural care. Procedural systems may include optimization of bleeding avoidance strategies (e.g. radial access site and bivalrudin), renal protective measures for chronic kidney disease, and development of care pathways to improve adherence to guideline-directed medications.18–23
Conceptual Framework for Systems and Measurement of High-Quality PCI
In this context, our study has several important implications. First, the lack of association between hospital level appropriateness classification and PCI outcomes indicates that measures of appropriateness alone are inadequate in determining PCI quality, as they do not describe hospitals with higher or lower rates of procedural complications. In fact, both hospitals with high and low proportions of PCIs classified as inappropriate perform the procedures with relatively low mortality and bleeding rates. Second, the considerable hospital-level variation in inappropriate PCIs suggests there exists a substantial opportunity to explore upstream PCI quality metrics to ensure patients are expected to benefit from the procedure. Notably, among PCIs performed at hospitals in the highest-tertile of inappropriate PCIs, nearly 25% were for asymptomatic patients where there is no expectation for clinical benefit.24–26
Similarly, 75% of patients undergoing PCI at hospitals in the highest-tertile were not on maximal anti-anginal therapy prior to the procedure, precluding the opportunity for adequate medical therapy to cost-effectively control patients’ symptoms.16, 24, 25, 27
Thus, a higher proportion of patients undergoing PCI at hospitals in the highest-tertile of inappropriate PCI are exposed to the clinical risk of PCI without reasonable expectation of greater benefit as compared with more conservative management strategies. If these inappropriate PCIs represent unnecessary procedures, then their identification represent an opportunity to improve PCI quality by reducing unnecessary complications and the resource utilization associated with these procedures.
One particular area of concern when evaluating AUC involves the issue of missing stress test data. We therefore conducted extensive sensitivity analyses to explore the impact of missing stress test data on classification of procedural appropriateness on our findings. Importantly, the association between hospital tertile of PCI classified as inappropriate and post-procedural processes of care and outcomes was not meaningfully influenced by assumed stress test results. However, the categorization of hospital tertile changed depending on assumptions about missing stress test data. Given the implications of missing stress test data on the site level assessment of PCI appropriateness, reducing site level variation in PCI performed without adequate documentation is an important corollary goal to reducing hospital variation in inappropriate PCI. In the interim, site-to-site comparisons of PCI appropriateness must account for the distribution of PCI without documentation of preprocedural stress testing to ensure equitable conclusions.
Strengths of our analysis include the large number of participating facilities and non-acute PCIs from a nationwide registry. However, our findings should be considered in the context of the following limitations. First, participation in NCDR is often voluntary and observed results may not reflect non-NCDR PCI hospitals. However, analysis from a statewide quality improvement program that includes non-NCDR hospitals suggests similarity of PCI appropriateness across NCDR participation status.28
Second, there are limitations in the application of the AUC for Coronary Revascularization, most notably due to missing results for non-invasive stress testing. However, our sensitivity analyses that assumed the highest and lowest risk for missing stress tests did not alter our conclusions. Additional limitations in the application of AUC have been described,29
however it is unclear these limitations importantly influence the assessment of patient selection for PCI across broad practice settings. Third, our study does not address the potential association between hospital PCI appropriateness and long-term outcomes. In addition to in-hospital complications, PCI incurs long-term risk such as bleeding related to dual anti-platelet therapy, acute thrombosis, and the need for repeat revascularization. It is possible that patient factors not accounted for in the AUC, but associated with increased risk of these long-term complications (e.g. prior bleeding event), are less frequently considered at facilities that also perform more inappropriate PCI. As a result, facilities performing more inappropriate PCI may have higher long-term complication rates. Finally, although our risk-adjusted analyses considered key variables identified from contemporary models developed and validated within NCDR, residual confounding is possible given the observational nature of our study.
In conclusion, in this large national registry, we found significant variation in the hospital proportion of non-acute PCIs classified as inappropriate. The hospital proportion of inappropriate PCI was not associated with other measures of PCI quality, including in-hospital mortality, periprocedural bleeding, and medication treatment after PCI. Our findings suggest that PCI appropriateness measures unique and important information that complements traditional PCI metrics to more fully inform quality. Additionally, these findings suggest hospitals with low rates of PCI complications do not necessarily provide high-quality PCI in settings where suboptimal patient selection results in more frequent use of PCI for inappropriate clinical indications. Hospital-based systems are needed to both ensure proper patient selection to maximize anticipated procedural benefit and to minimize post-procedural complications.