|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 1 (January 1), 2007: pp. 91-96 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2006.07.2454 Volume-Based Referral for Cancer Surgery: Informing the Debate
From the Department of Urology, University of Michigan; and Michigan Surgical Collaborative for Outcomes Research and Evaluation, Ann Arbor, MI Address reprint requests to Brent K. Hollenbeck, MD, MS, 1500 E Medical Center Dr, TC3875-0330, Ann Arbor, MI 48109-0330; e-mail: bhollen{at}umich.edu
Purpose: Mounting evidence suggests a relationship between hospital volume and outcomes after major cancer surgery; however, the absolute benefits of volume-based referral on a national basis are unclear. Patients and Methods: Data from the Nationwide Inpatient Sample were used to measure the likelihood of operative mortality and a prolonged length of stay (LOS) after six cancer surgeries (prostatectomy, cystectomy, esophagectomy, pancreatectomy, pneumonectomy, and liver resection) between 1993 and 2003. Using sampling weights, the adjusted likelihood of the outcomes was used to calculate the number of lives saved (or prolonged LOS avoided) in the United States. Results: The magnitude of the volume–operative mortality effect varied from an adjusted odds ratio (OR) of 1.3 (95% CI, 0.8 to 2.3) for cystectomy to 4.9 (95% CI, 2.4 to 10.1) for pancreatectomy. After accounting for varying rates of procedure utilization, the lives saved per 100 surgeries regionalized ranged from 0.2 (95% CI, 0.12 to 0.24 lives saved) for prostatectomy to 9.2 (95% CI, 6.7 to 10.4 lives saved) for pancreatectomy. The volume–prolonged LOS effect varied from an adjusted OR of 0.9 (95% CI, 0.5 to 1.6) for liver resection to 4.8 (95% CI, 3.5 to 6.7) for prostatectomy. After accounting for procedure use, the number of prolonged hospitalizations avoided ranged from –1.7 (95% CI, –11.3 to 3.6 hospitalizations) to 14.3 (95% CI, 12.9 to 15.4 hospitalizations) per 100 surgeries regionalized for liver resection and prostatectomy, respectively. Conclusion: For patients undergoing major cancer surgery, the benefits of volume-based referral depend on the interplay between procedure utilization, the magnitude of effect, and the outcome chosen.
Underlying concerns regarding the quality of health care in the United States have been bolstered by empirical data.1 Amid this quality gap, increasing health care costs2 have energized payers to implement programs to motivate global improvements in the quality of care delivered.3,4 Although variation in outcomes is undeniable,5 surgical disciplines have been insulated from quality improvement efforts because of the absence of robust performance measures. This reality has prompted stakeholders to develop expedient measures of surgical quality, including hospital volume.6-11 For this reason, volume-based referral has been advocated as a means to improve quality and save lives for some high-risk procedures.12,13 Prior work has imparted considerable support for such a policy for cancer care.11 Generally, a large body of evidence supports volume-based referral for many technically complex cancer surgeries, including esophagectomy,6,7,14 pancreatectomy,7,14 cystectomy,7,15 pneumonectomy,7,9 liver resection,6,16 and prostatectomy.17 However, the advantages of volume-based referral are commonly defined in relative terms (odds ratios [OR] or relative risk [RR]), and the absolute benefits (lives saved) remain largely unknown. Operative mortality occurs rarely for most procedures and is more frequent for uncommon procedures; hence, the absolute number of lives saved by regionalization may be few despite a large effect size. Alternatively, other benchmarks, such as length of stay (LOS), likely exhibit substantial variation by hospital volume but may lack sufficient leverage to motivate change in referral patterns. For these reasons, the impact of volume-based referral for cancer surgery on global public health is unclear. To better understand its potential health benefits, we quantify the impact of volume-based referral in absolute terms, on a national level, for patients undergoing cancer surgery. We explore the benefits of regionalization of cancer procedures that span the spectrum of utilization from relatively rare (eg, esophagectomy, pancreatectomy, liver resection, and cystectomy) to common (eg, prostatectomy and pneumonectomy) using outcome metrics (operative mortality and prolonged LOS) that exhibit a similar dichotomy.
Data from the Healthcare Cost and Utilization Project Nationwide Inpatient Sample were abstracted for the years 1993 through 2003. Patients undergoing one of six procedures (prostatectomy, cystectomy, pneumonectomy or lobectomy, pancreatectomy, esophagectomy, and liver resection) were identified using the procedural terminology based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9) codes. This population was then limited to the subset of patients undergoing their procedure for a cancer diagnosis (prostate, bladder, lung, pancreas, esophagus, and liver cancer) using the ICD-9 codes listed in Appendix Table A1 (online only). Secondary ICD-9 diagnosis codes were abstracted to enumerate comorbid conditions according to the methods described by Elixhauser et al.18 Demographic information (including age, race, sex, admission acuity, and insurance type) was also abstracted. Annual hospital volume was our primary exposure variable. A separate hospital volume variable was developed uniquely for each of the six procedures. Within each year, hospital volume was measured and sorted into deciles separately for each procedure. These deciles were ranked from low (bottom decile) to high (top decile) and then aggregated across the years. Thus, the actual threshold for a high-volume hospital varied from year to year but was representative of the volume status of a hospital relative to others performing the same procedure in the United States within the same year. For the purpose of analysis, hospital volume was used as a discrete measure, with high-volume hospitals (top decile) serving as the reference group.
The primary outcome was operative mortality, which was defined as an intraoperative death or death during the course of hospitalization after cancer surgery. The secondary outcome was prolonged LOS. This was determined uniquely for each procedure and was defined as the group of patients whose LOS was greater than the 90th percentile within each year of the study. As with hospital volume, the threshold for prolonged LOS for each procedure varied from year to year but was reflective of practice styles within a given year for each procedure. The bivariate relationship between each of the outcomes and the categoric covariates was assessed using the Rao-Scott For each procedure, unconditional logistic regression models were used to estimate the adjusted attributable risk as a result of hospital volume, comparing low-volume with high-volume hospitals. Discrete indicators were included in the model for each volume decile, with the exception of the high-volume group, which served as the reference category. Separate models were fit for operative mortality and prolonged LOS, and the models were adjusted for age, race (white, African American, Hispanic, other, or missing), sex, admission acuity (elective v nonelective admission), type of insurance (private, Medicare, or other), and comorbidities (based on the methodology described by Elixhauser et al18). SUDAAN software (version 9.0.0; RTI International, Research Triangle Park, NC) was used to fit the models to account for the survey design of the Nationwide Inpatient Sample and the potential clustering of outcomes within a hospital.19 Survey weights were used in all models so that the results would reflect all hospital discharges in the United States occurring during the study period. Rare factors causing model convergence problems and those found to have strong linear tendencies in multicollinearity diagnostics were removed as warranted. C-indices were measured for each of the adjusted models to provide insight into model discrimination qualities. The C-index represents the proportion of pairs in which the predicted probability of an outcome is higher for patients who had an outcome (ie, operative mortality or prolonged LOS).20 Next, percent attributable risk (PAR) was estimated from the adjusted logistic models using the formula based on RR proposed by MacMahon and Pugh21: PAR = (RR – 1)/RR. RR was estimated from the adjusted ORs of the adjusted SUDAAN models using the method described by Zhang and Yu.22 The 95% CI for PAR was estimated in a similar fashion using the values of the 95% CI for the OR derived from the model. To estimate the public health benefit of regionalization of patients treated at very low–to very high–volume hospitals, we calculated rates of lives saved and of avoiding a prolonged hospitalization per 100,000 US population using estimates from the US Census Bureau. To provide a policy context, we estimated the average lives saved and prolonged LOS avoided per 100 surgeries regionalized. For this measure, the numerator was the average annual improvement in outcome (if a procedure at a low-volume hospital had been performed at a high-volume hospital) multiplied by 100. The denominator was the average number of surgeries performed annually at the low-volume hospital. In accordance with the Code of Federal Regulations Title 45, Subpart A, Section 46.101, Paragraph b, Subparagraph 4, institutional review board approval was waived for this study. All testing was two tailed and carried out at the 5% significance level.
Demographic data regarding patients undergoing each of the six procedures are listed in Table 1. Generally, the patient population was youngest for liver resection and oldest for cystectomy. Age, admission acuity, and insurance type were uniformly associated with operative mortality for each of the cancer operations. Conversely, race, admission acuity, and insurance type were uniformly associated with a prolonged LOS.
Differences in operative mortality at low- and high-volume hospitals are listed in Table 2. For all of the procedures studied, high-volume hospitals had lower operative mortality rates. Generally, mortality declined over time for all of the cancer procedures. Among low-volume hospitals, the largest absolute decrease was observed for pancreatectomy (17.1% in 1993 to 9.8% in 2003; P < .001 for trend), and the largest relative decrease was observed for prostatectomy (0.64% in 1993 to 0.08% in 2003; P < .0001 for trend). Of note, the largest absolute difference in mortality rates at low- and high-volume hospitals was noted for pancreatectomy (11.8% v 1.7%, respectively) and esophagectomy (14.7% v 4.8%, respectively). After adjusting for differences in patient characteristics, the likelihood of operative mortality at low-volume hospitals was greatest for pancreatectomy (adjusted OR, 4.9; 95% CI, 2.4 to 10.1) and prostatectomy (adjusted OR, 3.8; 95% CI, 1.8 to 7.9).
Absolute and relative differences in prolonged LOS rates are listed in Table 3. With the exception of cystectomy for bladder cancer and liver resection for liver cancer, patients treated at high-volume hospitals were less likely to have a prolonged LOS compared with patients treated at low-volume hospitals after accounting for differences in patient characteristics. The most dramatic difference was observed in the prostatectomy cohort, where 19.0% of patients at low-volume hospitals had a prolonged LOS compared with only 3.9% of patients at high-volume hospitals. After adjusting for patient differences, those treated at low-volume hospitals were 4.8 times more likely to have a prolonged LOS than patients treated at high-volume hospitals (adjusted OR, 4.8; 95% CI, 3.5 to 6.7).
To illustrate the impact of regionalizing care from low- to high-volume hospitals, we calculated the adjusted PAR of low hospital volume on each outcome and estimated the number of lives saved and prolonged hospitalizations avoided (Table 4). The largest absolute benefit in lives saved would be realized if patients undergoing pneumonectomy at low-volume hospitals were regionalized to high-volume hospitals (635 lives saved, or 58 lives saved per year on average). However, the greatest mortality risk attributable to a hospital's low volume was noted for patients undergoing pancreatectomy and prostatectomy (77.6% and 73.5%, respectively). When standardized to US population-based rates, regionalization of these cancer patients from low- to high-volume centers would have saved between 0.03 and 0.23 persons per 100,000 over the course of the study.
In terms of avoiding a prolonged LOS, we noted a dramatic effect of low hospital volume in the prostatectomy population. If patients undergoing prostatectomy at low-volume hospitals were regionalized to high-volume hospitals, we estimate that 10,209 patients would have avoided a prolonged hospitalization, or 928 patients annually on average. To place these findings into context, we estimated the lives saved and prolonged hospital stays avoided if surgeries performed at low-volume centers had been performed at high-volume hospitals. For every 100 patients regionalized from low-volume to high-volume hospitals, we estimated that 6.3, 7.3, and 9.2 patients undergoing liver resection, esophagectomy, and pancreatectomy, respectively, would have been saved (Fig 1). Similarly, a prolonged hospitalization would be avoided in 14.3 men undergoing prostatectomy (Fig 2).
Motivations for volume-based referral depend largely on perspective. At what point should cancer surgery be regionalized? Presumably, hospital volume is a proxy for processes of care that result in better patient outcomes after major cancer surgery.6,14 From a patient perspective, each life is equally valuable, and saving the most lives should be the focus of referral. Alternatively, from a policy perspective, any volume-based referral initiative must be measured against potential trade-offs (eg, cost and travel time)23 and practicality. These data illustrate that the benefits of regionalizing cancer surgery depend on the interplay between the annual utilization of the procedure, the magnitude of the hospital volume effect, and the choice of outcome measure used to frame the debate. If the goal is to intervene for those at greatest risk (eg, the public health perspective), the solution would be to regionalize all high-risk patients (as advocated by the Leapfrog Group and others).24 For example, although nearly three quarters of deaths after prostatectomy and pancreatectomy are directly related to low hospital volume (as measured by their PAR), regionalizing patients would yield markedly different results. Because of the low risk of mortality among prostatectomy patients (0.2%), volume-based referral would result in only 142 lives saved compared with 597 lives saved if a similar initiative were undertaken for pancreatectomy. Thus, volume-based referral of only the high-risk procedures of esophagectomy, pancreatectomy, and liver resection would translate into approximately 31, 54, and 17 lives saved per year, respectively. At most, this corresponds to a public health benefit of 0.23 lives saved per 100,000. Conversely, smoking cessation and cholesterol reduction among those at risk would translate into approximately 120.2 and 48.6 lives saved per 100,000, respectively.25 Although a rough estimate, these data provide a context in which to interpret the relative public health impact of volume-based referral initiatives. Alternatively, if the goal was to save as many lives as possible irrespective of cost (eg, the patient perspective), then our data suggest that a fraction of all major cancer patients should be moved from lower to higher volume hospitals; on average, this would translate into 1,977 lives saved over the course of the study or 180 lives saved per year on average. Over the study period, concentration of pneumonectomy and pancreatectomy to high-volume hospitals would result in the greatest number lives saved (635 and 597 lives saved, respectively). However, for practical reasons, regionalization of a fraction of patients may not be possible. In other words, to achieve efficiency in terms of lives saved (the policy perspective), the greatest "bang for the buck" would be to focus on uncommon procedures with a high risk of mortality attributable to low volume (eg, pancreatectomy, esophagectomy, and liver resection). Indeed, the efficiency of such an initiative is borne out by the examining the relationship between the number of lives saved by regionalization and the annual utilization of each procedure. Despite similar absolute benefits in lives saved, volume-based referral of pancreatectomy and pneumonectomy would have different efficiencies. Volume-based referral of pneumonectomy, a relatively common procedure, would result in 1.4 lives saved for every 100 patients regionalized compared with 9.2 lives saved for pancreatectomy. For this reason, volume-based referral of common procedures may be impractical. Should the focus of referral remain on saving lives, where the benefits are likely to affect a relative few? Or, should the focus be on improving other, more common outcomes that may have the potential to impact many more lives? Our findings support the notion that the projected benefits of procedure-specific regionalization may vary substantially based on the outcome considered. Indeed, a shift in focus from operative mortality to avoidance of a prolonged LOS imply that the benefits of regionalizing prostatectomy far outweigh those realized by mandating referral of higher risk procedures, such as pancreatectomy or esophagectomy. Accordingly, although the policy lever for regionalization of high-risk, uncommon cancer operations may be best articulated in terms of lives saved, the benefits of volume-based referral for common procedures may be more appropriately framed in terms of reductions in resource utilization. Whether or not our findings are sufficiently provocative to motivate substantive changes in referral patterns remains a question best addressed in the political arena. The findings of this study should be interpreted with a few caveats. First, our findings should be generalized to the select cancer procedures studied, rather than to all procedures, until empirical data dictate otherwise. We acknowledge that regionalization of all procedures (the bottom 90th percentile) to high-volume hospitals would save many more lives and avoid many more prolonged hospitalizations than reported in this study. We chose the bottom decile of hospital volume as a threshold for volume-based referral across all procedures for consistency and practicality; although it may be feasible to regionalize all esophagectomies, it would be impractical to do the same for prostatectomy. Furthermore, the relationship between some procedures and outcomes may be primarily mediated at the level of the surgeon rather than at the level of the hospital. For example, surgeon volume plays a dominant role in determining mortality after cardiovascular procedures7 and in cancer control outcomes after prostatectomy.26 Finally, because this study encompasses 11 years, spontaneous regionalization of some procedures16,27 may have occurred and, therefore, attenuated the lives-saved estimates (and thus potentially biased our findings toward the null). Additionally, we do not account for differences in disease severity (eg, stage of disease), and our case-mix adjustment is limited to that which is measurable within the administrative data set. Regarding the former, such disease severity measures are typically more relevant to determining long-term outcomes and unlikely to systematically bias our findings away from the null. In terms of the latter, we used a well-accepted measure of medical comorbidity developed explicitly for ICD-9–coded data.18 Although other patient factors (eg, socioeconomic status) likely play a role in short-term outcomes, we were unable to account for patient-level differences because of changes in the coding of this variable across the years of the study. Despite these limitations, this study is among the first to quantify the benefits of volume-based referral in absolute rather than relative terms. Furthermore, these data highlight the principle that the implications of volume-based referral will vary with perspective. Processes of care underlying the volume-outcome relationship are, for the most part, unknown. Thus, at the current juncture, volume-based referral is one means of improving outcomes after cancer surgery. This study provides concrete data that can be used to guide policymakers when measuring the pros and cons of such an initiative.
Although all authors completed the disclosure declaration, the following authors or their immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors. Employment: N/A Leadership: N/A Consultant: John T. Wei, Sanofi-Synthelabo Inc Stock: N/A Honoraria: N/A Research Funds: Brent K. Hollenbeck, University of Michigan Cancer Center Testimony: N/A Other: N/A
Conception and design: Brent K. Hollenbeck, David C. Miller, David A. Taub, John T. Wei Financial support: Brent K. Hollenbeck Administrative support: Brent K. Hollenbeck Provision of study materials or patients: Brent K. Hollenbeck Collection and assembly of data: Brent K. Hollenbeck, Rodney L. Dunn, Stephanie Daignault Data analysis and interpretation: Brent K. Hollenbeck, Rodney L. Dunn, David C. Miller, Stephanie Daignault, David A. Taub Manuscript writing: Brent K. Hollenbeck, Rodney L. Dunn, David C. Miller, Stephanie Daignault, David A. Taub, John T. Wei Final approval of manuscript: Brent K. Hollenbeck, Rodney L. Dunn, David C. Miller, Stephanie Daignault, David A. Taub, John T. Wei
Supported in part by a grant from the John and Suzanne Munn Endowed Research Fund of the University of Michigan Comprehensive Cancer Center. Presented at the 100th Annual Meeting of the American Urological Association, May 21-26, 2005, San Antonio, TX. Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
1. McGlynn EA, Asch SM, Adams J, et al: The quality of health care delivered to adults in the United States. N Engl J Med 348:2635-2645, 2003 2. Chernew ME, Hirth RA, Cutler DM: Increased spending on health care: How much can the United States afford? Health Aff 22:15-25, 2003[Medline] 3. Baldwin FD: Where Medicare goes: The rest of the system may well follow CMS pay-for-performance example. Healthc Inform 21:24-26, 2004[Medline] 4. Birkmeyer NJ, Birkmeyer JD: Strategies for improving surgical quality: Should payers reward excellence or effort? N Engl J Med 354:864-870, 2006 5. Houghton A: Variation in outcome of surgical procedures. Br J Surg 81:653-660, 1994[Medline] 6. Begg CB, Cramer LD, Hoskins WJ, et al: Impact of hospital volume on operative mortality for major cancer surgery. JAMA 280:1747-1751, 1998 7. Birkmeyer JD, Siewers AE, Finlayson EV, et al: Hospital volume and surgical mortality in the United States. N Engl J Med 346:1128-1137, 2002 8. Schrag D, Panageas KS, Riedel E, et al: Surgeon volume compared to hospital volume as a predictor of outcome following primary colon cancer resection. J Surg Oncol 83:68-78, 2003[CrossRef][Medline] 9. Bach PB, Cramer LD, Schrag D, et al: The influence of hospital volume on survival after resection for lung cancer. N Engl J Med 345:181-188, 2001 10. Schrag D, Earle C, Xu F, et al: Associations between hospital and surgeon procedure volumes and patient outcomes after ovarian cancer resection. J Natl Cancer Inst 98:163-171, 2006 11. Hillner BE, Smith TJ, Desch CE: Hospital and physician volume or specialization and outcomes in cancer treatment: Importance in quality of cancer care. J Clin Oncol 18:2327-2340, 2000 12. Dudley RA, Johansen KL, Brand R, et al: Selective referral to high-volume hospitals: Estimating potentially avoidable deaths. JAMA 283:1159-1166, 2000 13. The LeapFrogGroup: Purchasing principles. http://www.leapfroggroup.org/for_members/what_does_it_mean/purchasing_principals 14. Finlayson EV, Goodney PP, Birkmeyer JD: Hospital volume and operative mortality in cancer surgery: A national study. Arch Surg 138:721-725, 2003 15. Konety BR, Dhawan V, Allareddy V, et al: Impact of hospital and surgeon volume on in-hospital mortality from radical cystectomy: Data from the health care utilization project. J Urol 173:1695-1700, 2005[CrossRef][Medline] 16. Dimick JB, Wainess RM, Cowan JA, et al: National trends in the use and outcomes of hepatic resection. J Am Coll Surg 199:31-38, 2004[Medline] 17. Ellison LM, Heaney JA, Birkmeyer JD: The effect of hospital volume on mortality and resource use after radical prostatectomy. J Urol 163:867-869, 2000[CrossRef][Medline] 18. Elixhauser A, Steiner C, Harris DR, et al: Comorbidity measures for use with administrative data. Med Care 36:8-27, 1998[CrossRef][Medline] 19. Panageas KS, Schrag D, Riedel E, et al: The effect of clustering of outcomes on the association of procedure volume and surgical outcomes. Ann Intern Med 139:658-665, 2003 20. Schwartz M, Ash AS: Evaluating risk-adjustment models empirically, in Iezzoni LI (ed): Risk Adjustment for Measuring Health Care Outcomes (ed 3). Chicago, IL, Health Administration Press, 2003, p 231 21. MacMahon B, Pugh TF: Epidemiology: Principles and Methods. Boston, MA, Little, Brown, and Co, 1970 22. Zhang J, Yu KF: What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 280:1690-1691, 1998 23. Birkmeyer JD, Siewers AE, Marth NJ, et al: Regionalization of high-risk surgery and implications for patient travel times. JAMA 290:2703-2708, 2003 24. Birkmeyer JD, Dimick JB: Potential benefits of the new Leapfrog standards: Effect of process and outcomes measures. Surgery 135:569-575, 2004[CrossRef][Medline] 25. Woolf SH: The need for perspective in evidence-based medicine. JAMA 282:2358-2365, 1999 26. Bianco FJ, Eastham JA, Vickers AJ, et al: Impact of the radical prostatectomy surgical technique and surgeon experience on freedom from cancer recurrence. J Clin Oncol 24:233s, 2006 (suppl, abstr 4569) 27. Hollenbeck BK, Taub DA, Miller DC, et al: The regionalization of radical cystectomy to specific medical centers. J Urol 174:1385-1389, 2005[CrossRef][Medline] Submitted May 1, 2006; accepted October 7, 2006. This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2007 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
2 test, whereas the bivariate relationship between the outcomes and the continuous covariates was tested using a simple logistic regression model and the Wald 
