Page 1

White Paper

Interpreting the Volume–Outcome Relationship in the Context of Cancer Care

National Cancer Policy Board

BACKGROUND

A higher volume of care translates into improved short-term outcomes for certain complex treatments for cancer. As much as a threefold increase in deaths following esophagectomy and pancreatectomy in lower- as compared to higher-volume hospitals has, for example, been reported in the health services research literature. These findings prompted the National Cancer Policy Board (board) to recommend in its 1999 report Ensuring Quality Cancer Care that cancer care is optimally delivered in systems of care that

Ensure that patients undergoing procedures that are technically difficult to perform and have been associated with higher mortality in lower-volume settings receive care at facilities with extensive experience (i.e., high-volume facilities). Examples of such procedures include removal of all or part of the esophagus, surgery for pancreatic cancer, removal of pelvic organs, and complex chemotherapy regimens.

Although evidence of the relationship between higher volume and better outcomes is strong and consistent for certain relatively uncommon procedures, the board did not have evidence to support a broader application of its recommendation. Many questions arose in board deliberations regarding the nature of the relationship and the processes of care that might explain it. Furthermore, the board recognized potential difficulties in implementing policies to concentrate care into higher-volume settings and decided that such issues had to be explored further.

On May 11, 2000, the Institute of Medicine (IOM, 2000a) held a workshop to bring together experts to:



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 1
Page 1 White Paper Interpreting the Volume–Outcome Relationship in the Context of Cancer Care National Cancer Policy Board BACKGROUND A higher volume of care translates into improved short-term outcomes for certain complex treatments for cancer. As much as a threefold increase in deaths following esophagectomy and pancreatectomy in lower- as compared to higher-volume hospitals has, for example, been reported in the health services research literature. These findings prompted the National Cancer Policy Board (board) to recommend in its 1999 report Ensuring Quality Cancer Care that cancer care is optimally delivered in systems of care that Ensure that patients undergoing procedures that are technically difficult to perform and have been associated with higher mortality in lower-volume settings receive care at facilities with extensive experience (i.e., high-volume facilities). Examples of such procedures include removal of all or part of the esophagus, surgery for pancreatic cancer, removal of pelvic organs, and complex chemotherapy regimens. Although evidence of the relationship between higher volume and better outcomes is strong and consistent for certain relatively uncommon procedures, the board did not have evidence to support a broader application of its recommendation. Many questions arose in board deliberations regarding the nature of the relationship and the processes of care that might explain it. Furthermore, the board recognized potential difficulties in implementing policies to concentrate care into higher-volume settings and decided that such issues had to be explored further. On May 11, 2000, the Institute of Medicine (IOM, 2000a) held a workshop to bring together experts to:

OCR for page 1
Page 2 1. review evidence of the relationship between volume of services and health-related outcomes for cancer and other conditions; 2. discuss methodological issues related to the interpretation of the association between volume and outcome; 3. assess the applicability of volume as an indicator of quality of care; and 4. identify research needed to better understand the volume—outcome relationship and its application to quality improvement. The workshop was structured around presentations of two commissioned papers: 1. “How Is Volume Related to Quality in Health Care? A Systematic Review of the Research Literature,” by Ethan A. Halm, Clara Lee, and Mark R. Chassin; and 2. “When and How Should Purchasers Seek to Selectively Refer Patients to High-Quality Hospitals?” by R. Adams Dudley, Richard Y. Bae, Kirsten L. Johansen, and Arnold Milstein. The workshop was jointly sponsored by IOM's Committee on Quality of Health Care in America and the National Cancer Policy Board, with financial support from the Agency for Healthcare Research and Quality (AHRQ). The board discussed volume-related policy issues at their October 2000 board meeting with three participants of the IOM workshop, Colin Begg, R. Adams Dudley, and Edward Hannan. This White Paper briefly summarizes the findings from the workshop and presents board recommendations for implementing volume-based policies relevant to cancer care. THE VOLUME–OUTCOME RELATIONSHIP IN THE CONTEXT OF HEALTH CARE QUALITY MEASUREMENT The National Cancer Policy Board concluded in its 1999 report Ensuring Quality Cancer Care that, based on the best available evidence, some individuals with cancer do not receive care known to be effective for their conditions. The magnitude of the problem is not known, but the board believes it is substantial. Evidence points to underuse of some interventions known to be effective (e.g., radiation therapy, adjuvant chemotherapy after surgery), overuse of interventions for which evidence supports alternative interventions (e.g., mastectomy versus breast conserving surgery), and misuse of effective interventions (e.g., administering inappropriate doses of chemotherapy). Despite compelling evidence of quality problems, it is difficult for individual consumers, health care purchasers, and others to make informed choices about cancer care, in part because the data needed to provide quality information specific to a particular physician or hospital are generally not available (IOM, 2000b). To ascertain

OCR for page 1
Page 3 whether practitioners are providing appropriate radiation and adjuvant chemotherapy after surgery, for example, one would have to assemble data from hospitals, outpatient settings, and possibly patients themselves (e.g., to ascertain treatment preferences). Such data may be examined as part of a health services research project or within a specific care system, but they are generally not available regionally or nationally. In the absence of good data on processes of care, data about outcomes (e.g., mortality, functional status) that include risk adjustment using detailed clinical data (usually available only in the medical chart or specialized databases) provide the best measurement of quality of care. Such risk-adjusted outcomes data are, however, not generally available to assess the quality of cancer care because of the time and expense associated with gathering and interpreting clinical data. When data on processes and outcomes of care are not available, alternative indicators may be used to ascertain quality. Health services researchers have assessed whether the site at which care is delivered is predictive of outcomes by examining associations between aspects of the organization and delivery of cancer care and health outcomes. Available evidence is insufficient to say that cancer care is better or worse when offered by specialized compared to generalist facilities or providers, or in managed care versus fee-for-service environments (IOM, 1999). Evidence is compelling, however, for a strong positive association between the volume of certain types of cancer care and better outcomes. Assessments of the volume–outcome relationship have tended to focus on surgical interventions because hospital data are generally available to study surgical procedures and their associated short-term mortality. It is more difficult to study the relationship between volume and outcomes for other types of interventions (e.g., chemotherapy, medical management) because there are insufficient sources of data on care administered outside of hospitals and on specific processes of care (IOM, 2000b). Likewise, there are virtually no widely available sources of information on longer-term outcomes of care such as quality of life and functional status. What follows is a summary of the literature on cancer surgery and the volume–outcome relationship. EVIDENCE OF A VOLUME–OUTCOME RELATIONSHIP FOR CANCER INTERVENTIONS As part of their synthesis of the literature on the volume–outcome relationship, Dr. Halm and colleagues reviewed 20 population-based studies of surgical interventions for cancer (the section of the literature review pertaining to cancer is included in Appendix A). The studies varied in their definition of high and low volume, whether hospital or physician volume (or both) were assessed, and how differences in characteristics of patients in high- and low-volume hospitals were taken into consideration in the analysis (i.e., case-mix adjustment). Despite differences in study design and methods, there is consistency in the published

OCR for page 1
Page 4 results—a higher-volume–better-outcome association was observed in all but three of the studies reviewed (these three studies showed no volume–outcome association). Almost all studies focused on short-term postoperative mortality (i.e., either in-hospital or 30-day mortality) and surgical complications, but a few investigators extended the time to follow-up. Birkmeyer and colleagues, for example, found a fourfold increase in in-hospital mortality rates following pancreaticoduodenectomy performed for Medicare patients in low- as compared to high-volume hospitals (16 versus 4 percent) (Birkmeyer et al., 1999a). A follow-up study of these patients showed that the significant volume-related mortality advantage persisted at 3 years post-surgery (37 versus 25 percent in low- versus high-volume hospitals) (Birkmeyer et al., 1999b). The volume–outcome relationship appears to be particularly strong for certain low-frequency, high-risk surgical procedures such as surgery for cancer of the pancreas and esophagus. For these procedures, rates of short-term mortality are generally at least two to three times greater in low- versus high-volume hospitals. Operative mortality rates by volume for four high-risk, cancer-related surgical procedures performed among Medicare beneficiaries are shown in Figure 1 (Begg et al., 1998). Although there is a statistically significant trend confirming improved outcomes with higher volume, providers in the intermediate-volume group are sometimes indistinguishable from either low- or high-volume providers. ~ enlarge ~ FIGURE 1 Impact of hospital volume on operative mortality for major cancer surgery among Medicare beneficiaries. SOURCE: Begg et al., 1998.

OCR for page 1
Page 5 For other procedures or conditions under review, the volume effect was not as great or as consistent. For common cancer-related surgical procedures (e.g., surgery for colorectal cancer), some studies show no effect, whereas others show statistically significant, but relatively small, effects. In a recent study by Hannan, for example, adjusted mortality rates were from 2 to 7 percentage points higher for low- compared to high-volume hospitals performing colectomy, lobectomy, and gastrectomy (Hannan, in press). In the few studies in which the effects of both surgeon and hospital volume have been assessed, only hospital volume is consistently related to better outcomes. INTERPRETING THE VOLUME–OUTCOME RELATIONSHIP Volume is recognized as an imperfect correlate of quality. Volume per se does not result in good outcomes in health care but is instead a proxy measure for other factors that affect care. These factors might include physician skill, experienced interdisciplinary teams, or well-organized care processes. However, with few exceptions, the literature does not shed light on the structures or processes of care that underlie the apparent relationship. A strength of volume as an indicator of quality of care is that it is relatively easy to obtain from available administrative databases. For procedures performed infrequently (e.g., esophagectomy), it is very difficult to measure quality directly (i.e., using physician- or hospital-specific outcomes data) because of the instability of small numbers (e.g., a few deaths can greatly influence annual rates); thus proxies such as volume may have to suffice. Here, one could base selective referral on minimum volume standards. For conditions in which higher volume has been shown to improve outcomes and for which caseloads are large enough to support outcomes measurement, it is also feasible to assess quality based on both outcomes data and volume. The use of a combination of data sources is preferable to using any one source alone. Assessing multiple indicators of quality is also preferable to relying on any one indicator alone (IOM, 2000a). Coupling volume data with clinical data could lead to the identification of processes and structures of care that distinguish high- and low-volume providers and that predict outcomes. The relative contributions of volume, process, and structure of care can be assessed only when comprehensive data are collected systematically as part of a special registry. In New York and New Jersey (and soon in California), for example, statewide clinical databases are available for cardiac surgery that allow analyses of outcomes by both individual surgeons and hospitals. Cancer registries are available for surveillance purposes, but they usually lack sufficient clinical information for quality-of-care studies (IOM, 2000b). Volume, when used as an indicator of quality, can be imprecise. Even though, in the aggregate, high- compared to low-volume providers have better outcomes, there is some variation so that not all high-volume providers have

OCR for page 1
Page 6 better outcomes and not all low-volume providers have worse outcomes. Consequently, the quality of care offered by any particular provider cannot be predicted accurately with information on volume alone. Furthermore, most volume studies to date have focused on short-term outcomes and on mortality. Whether outcomes such as quality of life or functional status improve with higher volume is not known. Despite its apparent value as a quality indicator, especially for low-frequency care, there are a number of unresolved issues that make volume difficult to operationalize in the context of health care quality improvement programs: When volume effects have been noted, it is unclear where along the volume continuum a threshold exists, above which outcomes are better but do not continue to improve with further volume increases. Studies generally do not illuminate how experience with procedures that are closely related to the procedure under study affect outcomes (e.g., should esophagectomies performed for indications other than cancer “count” toward volume?). It is likely that effects of physician and hospital volume combine or interact. The relative contributions of physician and hospital volume to outcomes, however, have been examined in only a few studies. Once high volume is attained, does it have to be sustained, or can lower volumes be adequate to maintain good performance? POTENTIAL IMPACT OF POLICIES TO CONCENTRATE CANCER CARE IN HIGH-VOLUME HOSPITALS There are roughly 5,000 community hospitals in the United States and virtually all of them provide at least some cancer care (AHA, 2000). Cancer-related surgeries for which the relation between volume and outcome appears to be strongest are performed infrequently: in 1997, there were an estimated 2,011 cancer-related esophagectomies and 3,832 pancreatectomies performed in the United States (Table 1). Relatively few hospitals would likely be affected by policies involving these uncommon procedures because no more than one-quarter of hospitals perform such surgeries. If a volume–outcome effect were established for more common procedures (e.g., gastrectomy, lobectomy, colectomy), a larger share of hospitals (up to 80 percent) would likely be affected by volume-based policies, such as selective referral programs. Selective referral programs might be difficult to implement for infrequently performed procedures because of the limited number of hospitals that have high volumes of these procedures: in 1997, an estimated 37 hospitals nationwide performed seven or more esophagectomies and 85 hospitals had this volume of pancreatectomies. The

OCR for page 1
Page 7 TABLE 1 Distribution of Selected Cancer-Related Surgical Procedures by Hospital Volume, United States, 1997 Procedure a Number of Discharges   Hospitals Performing at Least One Procedure     Definition of Lower Volume b Percentage of Discharges from Lower-Volume Hospitals Number Percentage Estimate 95% CI   Estimate 95% CI Estimate 95% CI Number Estimate 95% CI Total   5,113 100.0 Esophagectomy 2,011 1,626–2,396   869 772–966 17.0 15.1–18.9 <3 41.9 32.3–51.4 Pancreatectomy 3,832 2,808–4,857     1,252 1,141–1,363 24.5 22.3–26.7 <3 30.2 21.1–39.3 Gastrectomy 10,592 9,744–11,440     2,526 2,410–2,642 49.4 47.1–51.7 <5 34.5 30.1–38.9 Lobectomy 27,763 25,056–30,470     2,443 2,334–2,552 47.8 45.6–49.9 <10 20.1 17.0–23.2 Colectomy 86,676 82,518–90,834     4,165 4,053–4,277 81.5 79.3–83.6 <20 23.2 21.0–25.5 NOTE: CI = confidence interval; ICD-9-CM = International Classification of Diseases, Ninth Edition, Clinical Modifications. aEsophagectomy procedures defined as ICD-9-CM diagnostic codes = 150.x; procedure codes = 42.4x, 42.5x, 42.6x. Pancreatectomy defined as ICD-9-CM diagnostic codes = 157.x; procedure codes = 52.51, 52.53, 52.59, 52.6, 52.7. Gastrectomy defined as ICD-9-CM diagnostic codes = 151.x; procedure codes = 43.5–43.99. Lobectomy of lung defined as ICD-9-CM diagnostic codes = 162.2–162.9; procedure codes = 32.4. Colectomy defined as ICD-9-CM diagnostic codes = 153.x; procedure codes = 45.73, 45.74, 45.75, 45.76. SOURCE: Special tabulations: NCPB staff; Healthcare Cost and Utilization Project (HCUP), 1997 National Impatient Sample, Release 6, Agency for Healthcare Research and Quality, 1999.

OCR for page 1
Page 8 potential impact of volume-based policies on outcomes for these two procedures appears to be substantial because a relatively large share of discharges are from very low-volume hospitals (i.e., in 1997, an estimated 42 percent of esophagectomies and 30 percent of pancreatectomies were performed in hospitals with fewer than three procedures per year [Table 1]). Relatively few cancer-related esophagectomies and pancreatectomies (5 and 4 percent of discharges for these procedures, respectively) are performed in hospitals located in nonmetropolitan areas according to the 1997 Healthcare Cost and Utilization Project analyses. In a study of the potential impact in California of selective referral to high-volume hospitals, Dudley and colleagues (2000) estimated that in 1997, 27 deaths associated with esophagectomy and pancreatectomy performed in low-volume hospitals could have been averted with care in a high-volume hospital. NATIONAL CANCER POLICY BOARD RECOMMENDATIONS It is often difficult to judge when to implement policies based on research findings. The board considered four criteria to assess the strength of the evidence on the volume–outcome relationship and its adoption as a criterion for referral: 1. The relationship must be plausible and logical. 2. The observed trend must be consistent in available studies. 3. The size of the outcome difference must be substantial and clinically significant, and must meet stringent statistical criteria. 4. The effect must be confirmed in multiple studies. The board concluded that these criteria are met for two procedures included in the literature review—surgery for cancer of the pancreas and esophagus. The board chose to limit its recommendation to these two surgical procedures because of the size of the relationship and the consistency of the findings in the literature. Although the board found evidence regarding other procedures compelling (e.g., removal of pelvic organs, complex chemotherapy), it concluded that initial applications of cancer-related volume measures in quality assurance and improvement programs should be limited to those areas in which the body of evidence is robust. Furthermore, the board concluded that when research confirms a volume–outcome link, information should be disseminated to the public to support health care decision-making. In making such a disclosure, however, the limitations of the data and how to interpret them must be clear for their intended audience. The board proposes the following two recommendations to incorporate well-validated volume measures into quality assurance and improvement programs and to support further research on the volume–outcome relationship and its value in improving the quality of cancer care.

OCR for page 1
Page 9 Recommendation 1: When a large and significant volume–outcome relationship is established firmly by the literature through consistent findings in multiple studies (i.e., esophagectomy, pancreatectomy), volume should be incorporated as a quality indicator into ongoing quality-of-care programs and initiatives. Examples of such applications include the following: 1. Public and private health care purchasers' use of quality indicators (e.g., selective referral programs, consumer education); 2. Health plans' and providers' internal quality assurance monitoring; 3. Quality assurance organizations' surveillance activities: The Health Care Financing Administration's Peer Review Organizations' (PROs') assessment of the quality of care for Medicare beneficiaries Joint Commission on Accreditation of Healthcare Organizations' surveys of hospital and other health care organizations National Committee for Quality Assurance reporting of quality indicators for managed care health plans 4. Professional societies' assessments of patterns of care (e.g., American College of Surgeons' Commission on Cancer [ACSCoC]) 5. Consumer groups' campaigns to educate the public on quality-of-care issues. Well-validated quality measures can be applied in a variety of settings. Health insurance purchasers could use findings from research on the volume–outcome relationship to stipulate “evidence-based referrals” in contracts with health plans. The Pacific Business Group on Health (PBGH), a large purchasing coalition, is negotiating with health plans with which it contracts to increase the proportion of patients with selected conditions who are treated at high-volume hospitals (e.g., individuals with esophageal cancer in need of esophagectomies are to be referred to hospitals performing at least seven such procedures each year). PBGH also provides condition-specific volume data for all California hospitals on its consumer website ( www.healthscope.org) along with guidance on how to interpret the data. Employers could also make information about the relationship between volume and outcome available to employees directly (e.g., on a company intranet site) and could encourage employees to choose hospitals and providers based on available evidence. Similarly, health plans could direct members to high-volume providers. Information about the relationship between volume and outcome could be provided more broadly through public websites, via advocacy groups, or as part of widely distributed quality report cards. In New York State, for example,

OCR for page 1
Page 10 information on the volume of cardiovascular and other procedures performed by individual surgeons and by hospitals is available through the Center for Medical Consumers, a nonprofit advocacy organization ( www.medicalconsumers.org). Information about volume can be applied without public disclosure, for example, within systems of care for quality improvement programs. HCFA, for example, could use volume data to target interventions of its PROs that operate at the state level to ensure the quality of care for Medicare beneficiaries. Low-volume providers may withdraw voluntarily to avoid scrutiny, or they may be motivated to achieve minimal volume standards. Recommendation 2: Federal and private research sponsors such as the National Cancer Institute, the Agency for Healthcare Research and Quality, health care purchasers, health plans, and provider groups, through public–private partnerships, should support program evaluation and research projects to: (1) elucidate the nature of the volume–outcome relationship and its application to quality improvement, and (2) monitor the implementation (and effects) of volume-based policies. Much remains to be known about the relationship between volume and outcomes in the context of cancer care. Although a number of databases exist with which to assess the relationship, they have not been used extensively to assess cancer care, nor have evaluations been planned of ongoing efforts to integrate volume-based measures into quality improvement programs. A wide-ranging research agenda—from policy research to basic methodological research—is necessary to better understand the relationship between volume and outcome and how best to implement policies to improve care. Elucidating the Nature ofthe Volume–Outcome Relationship and Its Application to Quality Improvement Research is needed to determine the range of cancer care for which a volume–outcome relationship exists. This could be accomplished through a systematic and comprehensive examination of the relationship for both surgical and nonsurgical interventions, using existing data resources (e.g., AHRQ's HCUP database, state hospital discharge files, cancer registries, ACS-CoC's and the American Cancer Society's National Cancer Data Base). Such research will help determine the need for condition- or procedure-specific, prospective, population-based clinical databases and registries. Clinical databases and registries may be needed for more common cancer-related interventions to examine factors that mediate volume–outcome relationships.

OCR for page 1
Page 11 Monitoring the Implementation (and Effects) of Volume-Based Policies Several concerns have been raised regarding the adoption of volume-based quality measures, so methods are needed to monitor the impact of the adoption of such measures. A major concern is that selective referral programs might run counter to patient preferences for care close to home. Access to high-volume providers might be especially difficult for residents of rural areas and for those who lack resources to travel to hospitals that are far away. For some procedures, it may never be possible to regionalize care fully because some patients may need immediate treatment or be too unstable to transfer to a higher-volume setting. Where low-volume services have been closed, patients may experience a loss of access to a range of services, not just the procedure for which a volume–outcome relationship is known. Furthermore, providers may lose the ability to appropriately manage the postsurgical complications that arise in patients who have been referred to a distant high-volume hospital, but who return home for follow-up care. There are also potential effects on area marketplace structure and competition, such as the increased market power of high-volume hospitals (e.g., prices could rise), or barriers to the entry of new competitors (i.e., it is difficult to start at high volume). Furthermore, there is a potential for unintended consequences of a selective referral program—there could be medically inappropriate admissions to boost volumes to meet cutoffs. Also unexplored is a potential decrement in quality at very high volumes as a consequence of selective referral programs. Mechanisms are needed to monitor these and other effects of volume-based policies. One resource to assess the impact of such policies is the AHRQ's Healthcare Cost and Utilization Project (HCUP). The HCUP National Inpatient Sample in 1997 included information on 7.1 million discharges from a 20 percent sample of U.S. community hospitals (1,012 hospitals in 22 states). Because a hospital's total discharges are available, the annual volume of any particular procedure can be tallied by diagnosis. Preliminary analyses of HCUP data suggest that volume-based policies for low-frequency procedures such as esophagectomy and pancreatectomy might involve relatively few hospitals, but could have great impact on outcomes because many of these procedures appear to be performed in very-low-volume hospitals. Surveillance data from HCUP (and other sources) should be scrutinized by organizations that could use the information to implement programs to target interventions to areas where care remained concentrated in lower-volume settings. Referral patterns for medical care are very difficult to change, and the efforts of a number of groups will be required to foster the concentration of selected care in higher-volume settings. HCUP includes states' hospital discharge data, a valuable data source for health services research, but the database has certain limitations (e.g., lacks detailed clinical data, is not longitudinal, provides

OCR for page 1
Page 12 no information on patient preferences). Additional sources of data will be needed to fully evaluate the impact of volume-based policies. Research is also necessary to assess consumer and provider response to volume-based quality indicators—their interest in, and interpretation of, volume as an indicator of health care quality relative to other measures. Technical issues will have to be resolved through implementation-related research. Operationalizing volume-based quality indicators will require agreement on the definition of the conditions and procedures to be included in the measure (e.g., ICD-9 codes); volume thresholds or cutpoints to identify highand low-volume hospitals, standard methods to measure and monitor hospital volume (e.g., annual or biannual measurement, effects of hospital mergers and affiliations on categorization), and appropriate reporting formats for health care consumers. REFERENCES American Hospital Association (AHA), “Fast Facts on U.S. Hospitals from Hospital Statistics,” www.aha.org/resource/newpage.html, accessed September 12, 2000 . Begg CB, Cramer LD, Hoskins WJ, Brennan MF, 1998 “Impact of Hospital Volume on Operative Mortality for Major Cancer Surgery,” JAMA 280(20): 1747–1751. Birkmeyer JD, Finlayson SR, Tosteson AN, et al., 1999a , “Effect of Hospital Volume on In-Hospital Mortality with Pancreaticoduodenectomy,” Surgery 125(3): 250–256. Birkmeyer JD, Warshaw AL, Finlayson SR, et al., 1999b , “Relationship Between Hospital Volume and Late Survival After Pancreaticoduodenectomy,” Surgery 126(2): 178–183. Dudley RA, Johansen KL, Brand R, et al., 2000 , “Selective Referral to High-Volume Hospitals: Estimating Potentially Avoidable Deaths,” JAMA 283(9): 1159–1191. Hannan EL, Radzyner M, Rubin D, et al. (Surgery, in press), “The Influence of Hospital and Surgeon Volume on Mortality for Common Cancer Procedures.” IOM, Hewitt M, Simone JV (editors), 1999 , Ensuring Quality Cancer Care, National Academy Press , Washington DC. IOM, Hewitt M, 2000a , Interpreting the Volume–Outcome Relationship in the Context of Health Care Quality: Workshop Summary, National Academy Press , Washington DC. IOM, Hewitt M, Simone JV (editors), 2000b , Enhancing Data Systems to Improve the Quality of Cancer Care, National Academy Press , Washington DC.