Hospital Volume and Patient Outcomes in Major Cancer Surgery: Title and subTitle BreakA Catalyst for Quality Assessment and Concentration of Cancer Services

What can physicians do today about the cancer burden? The managed care industry has principally focused on early detection of breast, cervical, and colorectal cancers. Cancer professional organizations have pushed to expand access to research trials and experimental therapy. The role of genetic screening is also hotly debated.

However, relatively little attention has been given to the actual care provided to patients with cancer. In other words, does it matter where patients live, what type of insurance coverage they have, what their physician's specialty is, and at what type of medical center they are being treated or what its volume of cases is for a specific cancer? With the prominent exception of factors associated with breast-conserving surgery and local radiation compared with mastectomy for early stage breast cancer, the answer is the US health care system does not know and has not even tried to look (B.E.H. and T. J.S., unpublished data, 1998).

The study by Begg et al¹ in this issue of THE JOURNAL contributes important data on factors associated with outcomes for patients undergoing major cancer surgery. In this retrospective, population-based cohort study that uses Surveillance, Epidemiology, and End Results (SEER) data from 1984 through 1993, the authors show that the 30-day mortality for elderly patients undergoing complex surgical procedures—including pancreatectomy, esophagectomy, liver resection, and pelvic exenteration—was associated with the procedural volume at a specific hospital. Higher-volume centers had lower mortality before and after adjustments for case mix and patient factors. These results are consistent with prior work, provide insights into the quality of surgical cancer care, and should lead to explicit steps to concentrate where complex high-risk cancer surgery is performed.

Several essential data elements are needed to identify superior or inferior cancer care. First, it is necessary to know when disease is initially identified and to determine its extent. Cancer is the only chronic disease, other than HIV, for which the incidence and stage are registered. The National Cancer Institute's SEER program and the National Cancer Database of the American College of Surgeons collect data on type, incidence, stage, and therapy. However, because these data are hospital based, they are not readily available to insurers, clinicians, or patients.

Second, the processes of care must be defined. Obtaining this information is difficult because cancer registries are hospital based, yet evaluation and treatment are increasingly ambulatory events. The processes of care data are reflected principally in administrative claims. However, due to the uncommon nature of most cancers, only the largest insurers can potentially describe patterns of care for specific malignancies. Only recently have efforts been made to combine registry data with claims to examine patterns of care and quality indicators.^{2 - 3}

Third, significant variation in processes and outcomes of cancer care must be identified. Substantial variation in processes of care for breast, ovarian, lung, and prostate cancer have been associated with different geographic regions, hospitals, and providers.^{4 - 7} What is seldom recognized is that 5-year survival rates have been found to differ significantly for many types of cancers among SEER locations, especially for younger patients with local stage disease.⁸

Fourth, the relationship between the process of care and outcomes must be established. Only for select high-risk surgery is 30-day mortality an important outcome. For breast, colorectal, and prostate cancer, diseases that together account for 40% of all cancers,⁹ 5-year recurrence free or survivals are the dominant end points. For these conditions, surgical technique (process) will be reflected in local recurrence rates (outcome). The surgeon's role in coordinating care (process) will be reflected in the longer-term survival rates. In Canada and Britain, increased surgeon procedure volume leads to fewer local recurrences in colon cancer¹⁰ and better survival in breast cancer.¹¹ Currently, the process-long-term outcome linkage is only well established for adjuvant therapies in breast cancer where regionalization of (Canadian) care confirms that clinical trial results can be translated into community effectiveness.¹² Minimal attention has focused on process-outcome associations in other important phases of cancer, care such as symptom management and end-of-life care.¹³

The one area in which the process-outcome relationship has been explored is the "process volume" by physician or hospital and outcome of cancer care. In the United States, the concentration of pediatric cancer care at select medical centers may, in part, explain the major progress in this field.¹⁴ Restricting bone marrow transplantation to select centers is supported by strong evidence of better outcomes at higher-volume centers.¹⁵ However, such examples of concentration of care appear to be more of a reflection of an elastic supply-demand curve than a national policy. For cancer surgery no planned or informal arrangement has occurred.

The "higher procedural volume, better outcome" relationship has an extensive literature in cardiovascular disease.^{16 - 17} For cancer surgery, similar literature extends back to the 1960s but includes only 5 large US studies in the last 15 years.^{18 - 22} The technical skill of the surgical team (process) may be reflected in the perioperative morbidity and mortality (outcome). The evidence supporting the hypothesis that greater hospital volume improves outcome is most compelling for lung and pancreatic surgery.^{19 - 23} In these studies, patients treated at the lowest-volume hospitals, generally defined as the bottom quartile, represented 65% to 80% of all hospitals performing surgery. The 30-day risk-adjusted mortality at these low-volume hospitals was 40% to 80% higher than at high-volume centers. For example, the risk-adjusted mortalities following pancreatic resection was 12% at low-volume centers vs 4% at high-volume centers.²⁰ These studies evaluated all patients and used hospital discharge summaries that included the International Classification of Diseases, Ninth Revision, Clinical Modification coding but not the cancer staging. In contrast, the study by Begg et al¹ is the first to use cancer registry data linked with insurance claims for exploring the proportion of cases and the volume-outcome relationship using a population-based approach.

The volume-outcome association reported by Begg et al¹ was striking and did not have any volume threshold. This is not surprising given the low absolute case numbers for these complex surgical procedures, even at the highest-volume centers. Begg et al¹ also demonstrate how rarely elderly patients underwent curative-intent major surgery during this 10-year period. Less than 2% of patients with lung cancer and 4% with pancreatic cancer underwent total excisions. For pancreatic, esophageal, gastric, and lung cancers, the majority of new patients are diagnosed at a stage at which resection is not beneficial. However, other patients actually may be surgical candidates but are never referred to high-volume centers. In part, this may be due to reluctance to refer patients based on the high complication rates at low-volume hospitals.

Whether the volume-outcome relationship observed for the select procedures in the study by Begg et al¹ extends to other cancers or management of advanced stages of disease has been evaluated minimally in North America and should not be inferred automatically. Current data can only make an evidence-based statement for high-risk cancer surgery but is likely to hold, by inference, for centers performing high volumes of other technically challenging procedures. In general, hospital type has been consistently found to be less important than its case volume.

The study by Begg et al,¹ in combination with these prior studies, should be a catalyst for developing mechanisms to encourage low-volume centers to discontinue performing these procedures.²³ Patients who are potential candidates for complex surgical procedures for cancer treatment should be referred to "centers of excellence" based on hospital performance, rather than other factors, such as personal relationships or convenience. Although this approach may be debated by some hospitals and physicians, the evidence suggests that patients will benefit. High-risk surgery is only a small component of the multiple dimensions of cancer care. Additional research attention and resources are needed on the measurement of processes and audit of outcomes as necessary steps for quality improvement in all elements of cancer care.