This editorial refers to ‘The relation between volume and outcome of coronary interventions: a systematic review and meta-analysis’†, by P.N. Post et al., on page 1985
In the past 30 years, a large number of studies predominantly performed in the USA have investigated the relationship between short-term outcomes and provider volume for a wide variety of medical conditions and procedures, in particular percutaneous coronary interventions (PCI) and coronary artery bypass grafting (CABG) surgery.1–3
With the use of adequate statistical methods, Post et al.4 have performed a meta-analysis of 10 PCI studies, including 1 322 342 patients in 1746 hospitals, and of 10 CABG studies, including 1 754 777 patients in 2391 hospitals. They found that patients undergoing PCI or CABG in high-volume hospitals performing >600 cases per year exhibit significantly lower periprocedural mortality than those treated at lower volume hospitals.
This meta-analysis raises several important questions. Does the volume–outcome relationship persist after adjusting for patient case mix? Does the association vary as a function of patient age and predicted risk? To what extent is this association affected by patient clustering at specific centres? How do site variance-related issues affect the association? Is the association between hospital volume and outcome influenced by individual physician volume? What are the potential health policy implications of using hospital volume as a quality indicator?
Impact of volume on outcome of CABG surgery
With few exceptions,5,6 studies have indicated that hospitals and surgeons with higher volumes have lower mortality rates for CABG operations.7–10 The relative risk reduction for CABG-related mortality was 9% in the present meta-analysis and was shown to remain valid in the most recent studies. Results from 30 New York State centres noted a significant decrease in operative mortality in high-volume compared with low-volume hospitals, a difference that was maintained after risk adjustment.8 Although this state-wide study was among the most complete analyses based on clinical data, only a small proportion (<3%) had their CABG performed at low-volume hospitals due to New York State restrictions. There appears to be a substantial relationship between the volume of activity of the individual surgeon and outcome.8,11 However, annual surgeon volume is only one measure of surgeon experience. The cumulative surgeon volume occurring over many years could be a more appropriate metric. Another open issue is whether annual surgeon volume is equally important in determining quality of care for surgeons with many years of experience vs. less experienced surgeons. Zacharias et al.6 showed that, when the surgical team applies similar patient care standards and clinical pathways and emphasizes a team approach to cardiac surgery care, it is possible to achieve similar operative CABG outcomes at centres with low-volume or high-volume cardiac surgery programmes.
Impact of volume on PCI outcome
The relative risk reduction for PCI-related mortality was 13% in the present meta-analysis and was not attenuated over time. The advent of coronary stents has reduced the incidence of PCI-related complications, including the risk of emergency CABG. As a result, procedural outcomes of elective procedures have been levelled out. In a Californian registry study carried out between 1984 and 1996, the disparity in outcomes between low- and high-volume centres narrowed, and complication rates improved in all hospitals.12 Therefore, for procedures with low short-term mortality such as PCI, additional outcome measures, including the need for same-day CABG surgery, same-stay CABG surgery, and readmission within a short period of time for complications related to the index admission, should be considered. In a population-based study, examining the data from New York's PCI Reporting system in 1998–2000, Hannan et al.11 showed that for hospital volumes <400 and operator volumes <75, the respective odds of mortality, same-day CABG surgery, and same-stay CABG surgery were 5.9, 4.0, and 3.9 times the odds for hospital volumes of ≥400 and operator volumes of ≥75.
The volume–outcome relationship appears even more meaningful for high-risk and emergency PCI. Vakili et al.13 reported on the global experience across an entire US state, including rural and urban, teaching and non-teaching, not-for-profit and for-profit hospitals. The average mortality rate for patients treated with primary PCI for acute myocardial infarction (AMI) was ∼5-fold higher than in patients undergoing elective PCI. The AMI patient is more often experiencing active ischaemia, manifested by chest pain, arrhythmias, and haemodynamic instability. High-volume centres tended to administer life-saving reperfusion therapies faster than low-volume centres. The difference was larger for patients who underwent primary PCI than for those who received fibrinolytic therapy.14 Similar findings were reported in two recent studies from Europe. Spaulding et al.15 found a strong inverse relationship between hospital PCI volume and in-hospital mortality after emergency procedures. Zahn et al.,16 reporting on behalf of the ALKK registry, have demonstrated, in a study not included in the present meta-analysis, a significant inverse relationship between hospital mortality (from 4.41 to 2.78%, P = 0.004) and quartiles of PCI volume. Similar findings were reported from the multicentric German Cypher registry.17 Therefore, tolerance of low-volume thresholds for angioplasty centres with the purpose of providing primary PCI should not be recommended, even in underserved areas.
Cross-talk between volumes of CABG and PCI procedures
Programmes that perform a relatively large number of PCI procedures and a low volume of CABG procedures might tend to perform CABG procedures on a more complex patient population. Carey et al.18 showed that a statistically significant effect of PCI volume on CABG mortality occurs in lower-volume hospitals when the PCI/CABG ratio exceeds 2.0. Compared with high-volume hospitals, low-volume hospitals tended to operate on patients at higher risk and under more emergent conditions. Reasons for those differences may include adverse selection, variance in clinical coding among hospitals, and differential threshold for surgery due to altered cluster experience and/or institutional financial pressure. More straightforward cases are referred for PCI, and patients with characteristics not defined in risk models, such as poor vessel quality or diffuse disease, tend to be referred for surgery.
Association or causal relationship?
The literature distinguishes between two hypotheses as the potential explanations for the inverse volume–outcome relationship. The first is that ‘practice makes perfect’. The assumption is that a greater volume of patients should allow the ‘operative’ team to develop greater skills and judgement in the management of complex clinical conditions and therefore achieve better outcomes. The second is called ‘selective referral patterns’: ‘operating’ teams with superior outcomes attract a greater number of patients. These hypotheses have different policy implications. Under the first hypothesis, and in an environment where cost containment is important, centralization in the least costly institution would be a reasonable strategy; institutions would logically improve outcome with further experience and accomplish this at a lower cost. If the second hypothesis holds, then this strategy would not be reasonable, and centralization should be in those institutions with better outcomes. Both hypotheses are not mutually exclusive, as a more accomplished team will enjoy selective referral and demonstrate a greater improvement in outcomes as a result of the larger practice than would a less accomplished team.
In the past few years, there have been advocates in several countries for changing the landscape of health care delivery based on what is known about volume–outcome relationship. It was reasoned that a significant number of procedure-related deaths may be avoided if a policy of ‘regionalization’ was adopted to avoid low-volume hospitals. Such procedure volume-driven policy guidelines may have a number of drawbacks: low-volume hospitals with good outcomes are handled in the same way as those with poor quality of care, similarly higher quality of care is assumed at all high-volume hospitals, and decisions do not account for potentially substantial physician–volume confounding effects. Enforcing a strict regionalization policy may cause significant disruption in the process of care delivery, with thousands of patients being referred away from low-volume to high-volume hospitals. Therefore, it has been recommended that such referrals be limited to high-risk patients. Risk of revascularization procedures can be estimated from a number of risk scores, of which the EuroSCORE for estimation of CABG-associated risk appears very robust.19 High-risk PCI procedures pertain both to the acute clinical presentations, i.e. cardiogenic shock, emergency or primary PCI cases, and to a number of anatomic subsets, among which are PCI of unprotected distal left main, chronic total occlusions, and complex multivessel disease.
Volume as a surrogate for quality
Using hospital procedural volume as a quality indicator is attractive and widely accepted. This structural characteristic is readily available from administrative data, requires no complex adjustment techniques, is easily interpretable by the public, and is consistent with the common belief that ‘practice makes perfect’. Establishing a sound link with outcome requires, however, the availability of detailed clinical data as well as appropriate risk adjustment. Collection of clinical data sets for the purpose of public reporting has raised concerns about upcoding of risk factors, avoidance of performing interventions on high-risk patients, or lowering of indication thresholds in order to inflate procedural volumes. At the same time, it was shown that programmes introduced in New York and other US states have been associated with reductions in procedural deaths, in part because monitoring of comparative outcomes encourages quality improvement efforts. Peterson et al.9 concluded that hospital procedural volumes are best considered as a surrogate metric for quality in a setting where other more direct process and outcome assessments are not available. It seems reasonable to support the continued growth of national clinical databases, which are capable not only of tracking risk-adjusted care patterns and outcomes, but also of improving them.
Recent cross-sectional analysis of Medicare claims
Ross et al.20 recently investigated the potential relationship between hospital volume and 30-day mortality for three common medical conditions, based on nearly 3.5 million hospitalization records. They found that Medicare beneficiaries who were hospitalized between 2004 and 2006 in acute care hospitals in the USA for AMI, heart failure, or pneumonia had a decrease in the rate of death if they were admitted to a hospital that handled a large condition-specific volume of patients every year.20 However, the relationship between volume and decreased mortality was attenuated at greater volumes, and there was a threshold for each condition above which an increase in hospital volume was no longer associated with lower mortality. Moreover, once the annual volume reached 100 cases, the curve representing the association between volume and risk-adjusted mortality began to flatten, suggesting that the benefit of an increased volume of patients at a hospital would be most pronounced at low-volume hospitals and would be attenuated as the hospital's volume increased.
Given the complexity of factors that contribute to patient outcome after revascularization procedures, this recent study is consistent with previous evidence that the best performance is obtained through an optimal interaction between human and material resources, experience and routine, process and process management, case-mix and volume load.
Data collection and reporting mechanisms are important and therefore they should be accurate and timely. The focus on short-term mortality should be extended to long-term outcomes and patient's symptoms, as reflected by their functional status and quality of life, particularly in light of the growing emphasis on the appropriate use of procedures. We need to understand better how the information is used by physicians, hospitals, the public, purchasers, payers, and referring doctors. Compliance with guidelines on hospital volumes should be strongly encouraged by national and international cardiology and cardiac surgery societies, and their implementation monitored by local regulatory boards. While the availability and accuracy of clinical quality metrics are constantly improved, it is appropriate to rely on procedural volumes, given the indisputable evidence that low volume means worse outcome. For example, the Leapfrog group recommends contracting with hospitals having annual volumes of at least 400 procedures a year for PCI, and >450 for CABG. Among other national or international agencies, the American College of Cardiology and the American Heart Association both recommend that PCI be performed by experienced operators in high-volume hospitals, namely >400 PCI procedures per year per hospital and at least 75 procedures per year for operators.21
Conflict of interest: none declared.