Substantial and appropriate national attention has been focused over the preceding decade on examining outcomes and quality performance in cardiac catheterization laboratories performing percutaneous coronary interventions (PCI). Previous publications have in general found an inverse relation between volume and quality outcomes.1, 2 In response to these data, policymakers such as the Leapfrog Group, a national employer coalition that provides insurance for more than 35 million Americans, has asked insurance companies that cover its beneficiaries to refer patients with PCI to high-volume hospitals.
More recently, with the introduction and increased utilization of intracoronary stenting and the introduction of platelet glycoprotein IIb/IIIa inhibitors, the data have been somewhat less clear about the volume-outcome relation, but most reports have found a persistent albeit smaller correlation between increased catheterization laboratory volume and higher quality.3 The majority of these reports have assessed quality by using mortality and same day coronary artery bypass grafting (CABG) rates without considering other adverse procedure-related outcomes, process measures of care, or other performance indicators.
In this issue of the Journal, Brown, in his article “Analysis of the institutional volume-outcome relation for balloon angioplasty and stenting in the stent era in California,” offers additional evidence on the volume-quality relation in PCI.4 By using 1997 data from the California Office of Statewide Health Planning and Development administrative database, Brown found the continued presence of a volume-quality relation in stented patients with PCI: a 1.2% difference in absolute mortality rates between low- and high-volume hospitals.
There are, however, significant limitations of Brown's analysis that are only partially acknowledged in his report. PCI remains a constantly evolving technology. In this 1997 study, only 57% of all patients had intracoronary stenting. Although Brown provides no adjunct therapy data, platelet glycoprotein IIb/IIIa inhibitors were just being introduced into widespread clinical use in 1997. By the year 2000, however, the American College of Cardiology–National Cardiovascular Data Registry (ACC-NCDR) reported that stenting accounted for more than 77% of PCI procedures5, 6 and that more than 55% of patients received platelet glycoprotein IIb/IIIa inhibitors as adjunct PCI therapy.7
Another significant limitation of Brown's study is that no risk adjustment was performed to correct for differences in patient populations between low- and high-volume centers. This would be unimportant if the measurable risk factors appeared to distribute equally among hospitals by volume, but they do not. Compared with high-volume hospitals, low-volume hospitals had 9% more patients whose PCI was performed in the setting of acute myocardial infarction (AMI, 34% versus 25%). This fact alone could explain much of the mortality rate difference found between hospital groups. The ACC-NCDR mortality rate for PCI associated with hospital diagnosis of AMI was 5.2%, whereas PCI for stable or unstable angina has an associated mortality rate of only 1.1% and 1.0%, respectively.6 Although Brown offers important information on the volume-outcome relation in the era of stenting, his findings should be considered in the context of other similar studies. For instance, initial analysis of data from 1998 to 2001 from the ACC-NCDR assessing more than 250,000 PCIs performed at 225 sites showed a smaller difference (1.85% versus 1.3%) in risk-adjusted mortality rates between low- and high-volume centers.10 In addition, low-volume centers had slightly higher rates, after risk adjustment, of urgent CABG after PCI: 1.70% versus 1.45%.8
Some of the difference in findings between Brown and the ACC-NCDR analysis may reflect the data sources used (administrative claims data versus a specialized clinical database set up for the purpose of evaluating PCI outcomes). Claims data generally do not include the clinical detail needed to perform adequate risk adjustment, and even some of the variables that are included (such as AMI) may be miscoded either through simple mistakes or in response to reimbursement incentives.
Even if Brown had attempted risk adjustment, that might not have been sufficient to convince all doubters about a volume-outcome relation, as the science of risk adjustment is still in evolution. Recently, there has been increasing focus on the use of hierarchical modeling for analyzing data such as this to minimize false inferences that traditional statistical techniques risk by ignoring the structure of the data.9, 10 Hierarchical statistical modeling helps account for “data clusters” of patients, “random effects,” and issues related to variations in results from individual hospitals affecting that hospital group. On the basis of individual hospital outcome variability, different conclusions could be reached by using hierarchical modeling. Application of hierarchical statistical modeling to the ACC-NCDR data will clearly be of further interest.
In addition to methodological issues related to risk adjustment, the clinical reality of changing practice patterns in the management of acute coronary syndromes (ACS) needs to be taken into account when considering using hospital PCI volume for any regulatory, reimbursement, or referral purpose. For example, we increasingly recognize the importance of timely intervention in ACS, and policies that would shift patients away from low volume centers to high volume centers might increase the time required to achieve revascularization. Furthermore, as the evidence base supporting the use of primary PCI rather than thrombolysis in AMI grows, so does the strength of the clinical rationale to equip more rather than fewer centers to do PCI, even hospitals without onsite surgical standby.11, 12
Finally, there are policy considerations to be made when choosing PCI performance measures. Central among these is that setting up a PCI minimum volume criterion risks creating incentives for hospitals and physicians to perform PCIs on patients in whom they may only marginally be indicated.
For all these reasons, we do not conclude from Brown's findings or those of the recent ACC-NCDR analysis that volume is a preferred quality indicator in PCI. Rather, the important message from the volume-outcome literature is that performance varies so much that even a crude indicator such as volume can pick up differences. This strongly suggests that we need to put in place, in all cardiac catheterization laboratories, systems for performance evaluation and reporting. Risk-adjusted mortality rates, although clearly important, are also an inadequate yardstick for quality assessment as mortality rates are low and falling, and confidence intervals on mortality estimates for low-volume centers are wide. This means that it will be more feasible to identify differences in processes of care—such as “door-to-balloon time” for primary PCI and appropriate use of adjunctive therapies—than mortality rates. Data about processes also may help hospitals with higher-than-expected mortality rates to understand how to improve outcomes. Therefore, we believe a performance evaluation system should collect data about patient demographics, noncardiac risk factors, and the clinical parameters of their coronary disease as well as process measures (such as compliance with ACC/AHA Clinical Guidelines) and adverse outcomes and report this data to a national database. The database should feed back to hospitals risk-adjusted mortality and morbidity rates. When calculating these outcome measures for low volume centers, 3 to 4 years of rolling data could be used to reduce the width of confidence intervals and help hospitals and other stakeholders better understand each institution's performance.
Fortunately, such a national database already exists for PCI, the ACC-NCDR. Comparing an individual hospital's outcomes and process measure compliance rates with ACC-NCDR benchmarks would give the hospital, regulatory agencies, and other stakeholders a much better assessment of performance in the catheterization laboratory than using volume criteria as a quality surrogate. With its ACC-NCDR hospital-specific performance report, a hospital could assess its systems of care to identify opportunities to improve. The application of the ACC/SCAI Catheterization Lab CQI (continuing quality improvement) Toolkit, available in the fall of 2003, will help catheterization laboratories implement positive changes to promote success and quality. Developed jointly by the ACC and the Society for Cardiac Angiography and Interventions(SCAI), CathKit focuses on quality issues and patient outcomes through improvement of systems and processes rather than punishment of individuals. CathKit contains CQI methodology with tutorials for implementation of CQI projects along with up-to-date information on guidelines, regulatory challenges, benchmarks, and training standards. The Leapfrog Group has asked all hospitals performing PCI to join the ACC-NCDR as an alternative to its prior policy of asking insurance companies to direct patients to high-volume PCI centers.
At present, there still remains significant variation in performance among our nation's catheterization laboratories. A minimum volume criterion as a surrogate for quality always was an inadequate measuring stick but may have been appropriate when the volume-outcome relation was stronger, when it was less clear that PCI was preferable to thrombolysis, and in the absence of an alternative approach to providing valid performance assessments. However, more meaningful quality measures—risk-adjusted outcomes with the use of models built on a national database and comparisons of outcome and process performance to national norms—now are available. Furthermore, policymakers such as the Leapfrog Group have expressed their preference for these more valid indicators. Participation in such databases may also allow some providers to identify previously unrecognized opportunities to improve their own management of their cardiac care system through blame-free performance evaluation and feedback. For all these reasons, physicians and hospitals should move aggressively to report the preferred quality measures rather than refusing to participate and leaving policymakers with no alternative but to continue to use volume as a surrogate for quality. Doing so will help patients and clinicians.
FULL TEXT
