For many surgical procedures, outcomes are improved when the operation is performed by high-volume surgeons and at high-volume hospitals.1–3 Although this phenomenon has been demonstrated for a wide variety of procedures, the influence of volume is most pronounced for high risk procedures associated with a significant risk of complications.2,3 For ovarian cancer, women who undergo surgery at high-volume centers are more likely to receive guideline recommended care and have improved survival.4–9
Recognition of the volume-outcomes paradigm has led to efforts to concentrate care for complex surgical procedures to high-volume hospitals and surgeons.10 Much of these efforts have focused on developing specialized centers and on public reporting of hospital volume to help inform patient choice. To further advance these efforts, minimum procedural volume standards that hospitals and surgeons must achieve to continue to perform a given procedure have been proposed for some nongynecologic surgeries.11,12 These standards are meant to restrict low-volume hospitals and surgeons from performing a procedure.11–14 To date, minimum-volume standards have been largely untested and the effect of implementing these standards remains unknown for most procedures.15
Given the association between procedural volume and outcomes for ovarian cancer, there is a strong rationale to limit care to high-volume centers. We used nationwide data to model the potential effects of implementing hospital-level minimum-volume standards for ovarian cancer. Specifically, we examined how enacting minimum-volume standards affects survival and access to care for women with ovarian cancer.
The National Cancer Database was used for analysis. The National Cancer Database is a hospital-based registry developed by the American College of Surgeons and American Cancer Society that collects data on patients who receive their cancer diagnosis or treatment at more than 1,500 Commission on Cancer–affiliated hospitals across the United States.16–18 The National Cancer Database currently captures approximately 70% of all incident cancer cases.17 Data elements include patient socio-demographics, tumor characteristics, first course of treatment before disease progression or recurrence, follow-up and survival.17,18 Regular audits are performed to guarantee data integrity and completeness reported to the National Cancer Database. This study was deemed exempt by the Columbia University Institutional Review Board.
We identified women with invasive ovarian cancer diagnosed from 2005 to 2015 who had ovarian cancer as their first cancer diagnosis that was confirmed histologically. After patient selection, we identified all hospitals that treated at least 1 patient during the study period. The primary focus of the study was to determine how a hospital's prior year surgical volume was associated with outcomes during the subsequent year. As such, the cohort was further limited to hospitals that treated at least one patient with newly diagnosed ovarian cancer in the prior year. Each hospital year was treated as a separate unit of analysis. For example, if a given hospital treated patients and had a prior year volume for five different years, this would represent five separate observations.
Patients' demographic data included age at diagnosis (younger than 40, 40–49, 50–59, 60–69, 70–79, at least 80 years), race and ethnicity (white, black, Hispanic, other, unknown), year of diagnosis, and insurance status (private, Medicaid, Medicare, uninsured, other governmental, or unknown). Patients' socio-economic statuses were measured by median household income and percentage of adults who did not graduate from high school in a patient's zip code area from census tract survey data. Median household income was classified as less than $38,000, $38,000–$47,999, $48,000–$62,999, more than $63,000, or unknown; and percentage of adults who did not graduate from high school was classified a 21% or more, 13–20%, 7.0–12.9%, less than 7%, or unknown. Patients' residential locations were estimated by matching state and county code to rural-urban continuum codes from the United States Department of Agriculture Economic Research Service, and classified as metropolitan, urban, rural, and unknown. Comorbidity was measured using the Deyo classification of the Charlson comorbidity score, and grouped as 0, 1, or 2 or greater.19 Tumor characteristics included tumor stage (INOS [not otherwise specified], IA, IB, IC, IINOS, IIA, IIB, IIINOS, IIIA, IIIB, IIIC, IV, unknown), histology (serous, mucinous, endometrioid, clear cell, transitional cell, NOS), and grade (well, moderate, poorly, unknown).
Hospital characteristics included facility region (eastern, Midwest, south, west, unknown) and facility type as defined by the American Cancer Society's Commission on Cancer Accreditation program criteria, classified as academic centers, community centers or comprehensive community cancer centers, and integrated network cancer program.16
Patient demographic, tumor characteristics and hospital factors are presented descriptively based on hospital prior year volume quartile and were compared using χ2 tests. The objectives of our study were to estimate the association between a hospital's prior year volume and overall mortality at 60 days, 1 year, 2 years, and 5 years (from the date of diagnosis) in patients treated during the current year. The mortality cohorts were restricted to patients who either died within the specified time frames (60 days, 1 year, 2 years, and 5 years after diagnosis) or who were alive with follow-up.
The observed/expected ratio of mortality was calculated for each hospital at each year. The expected morbidity rate was estimated as the mean predicted rate using a multivariable Poisson regression model adjusting for age, race, year of treatment, insurance, comorbidity, household income, education level, location, tumor stage, histology, and grade.
An observed/expected ratio of less than 1 indicates that the mortality rate of a hospital was lower than expected, and an observed/expected ratio of 1 or greater indicates that the mortality rate is the same as or higher than predicted.13 Separate observed/expected mortality ratios were calculated for each hospital annually. The association between a hospital’s prior year volume and observed/expected mortality ratio was examined by fitting Poisson regression models based on the methods of generalized estimating equations with unstructured variance matrix to account for a repeated observed/expected ratio within the same hospital across the years. Robust score tests were used to evaluate the curvature (nonlinear relationship) between a hospital's prior year volume and observed/expected mortality ratios and the overall significance of the curves. The predicted observed/expected mortality ratios of all hospitals during a given year were plotted against a hospital's prior year volume. In our analysis, the robust score test indicated a linear association between hospital prior year volume and observed/expected mortality ratio. Restricted cubic spline transformations with 3 knots (at 5th, 50th, and 95th) were also explored, and the results from restricted cubic spline transformations were similar to the linear association.
We modeled the effect of minimum-volume standards based on a number of theoretic volume cutpoints. We modeled the effect of a prior year hospital volume of one, two, three and five cases. At each theoretical cutpoint, we determined the number of patients and hospitals in the current year that would fall above and below the given cutpoint applied to prior year volume. We determined the mean observed/expected ratio for hospitals above and below the cutpoints and determined the number of hospitals with observed/expected ratios less than 1 or at least 1 across the cutpoints. We report descriptively the effect of minimum-volume cutpoints for hospitals during 1 year (2015).
We determined the number needed to treat for volume cutpoints at given time periods to reduce one death. We first estimated the absolute risk adjusted mortality reduction as the absolute difference in the observed/expected ratio between hospitals at or below and those above a given volume threshold multiplied by the overall mortality at a given time point. The number needed to treat then was estimated as 100 divided by the absolute risk adjusted mortality reduction. The number needed to treat was estimated for volume thresholds of one, two, three and five cases in the prior year. To assess the potential bias from missing values for some covariates we performed a sensitivity analysis using multiple imputation.20 Using the fully conditional specification method, we imputed all missing values for stage, grade, race, income level, education, urban or rural location, facility location and facility type. All analyses were conducted using SAS version 9.4 (SAS Institute Inc, Cary, North Carolina). All hypothesis testing was two-sided and a P-value of <0.05 was considered statistically significant.
A total of 136,196 patients treated at 1,321 hospitals were identified. When stratified by patients into quartiles, low-volume hospitals treated fewer than three patients per year, medium low-volume hospitals treated three to five patients per year, medium high-volume hospitals treated 6 to 15 patients per year, and high-volume centers treated more than 15 patients per year (Table 1). Patients at low-volume centers were older, more often white, were more commonly Medicare recipients, residents of urban areas, and had greater comorbidity (P<.001 for all). Women at low-volume hospitals more frequently had nonserous histologies and more commonly had stage IV tumors (P<.001 for both). Low-volume hospitals were more often community cancer centers, located in the Midwest and urban facilities (P<.001 for all).
There was a statistically significant decline in the observed/expected ratio with increasing prior year hospital volume for 60-day (P=.004), 1-year (P<.001), 2-year (P<.001) and 5-year (P=.008) mortality (Fig. 1). Figure 2 displays the individual hospital observed/expected observations based on prior year volume. The number of hospitals caring for women with ovarian cancer ranged from 1,014 to 1,056 per year, with 11,463–12,948 patients per year treated (Table 2). The median prior year hospital volume was six cases each year, except for 2013 in which the median volume was seven. The lowest 25th volume percentile was two to three cases each year, and the lowest 10th percentile by volume was one to two cases per year.
Applying even a low-volume cutoff would eliminate a substantial number of hospitals. For example, in 2015, using a minimum-volume cutpoint of one case in the prior year would eliminate 144 (13.6%) hospitals that treated 2.6% of the patients. A cutpoint of three cases in the prior year would eliminate 364 (34.5%) hospitals that treated 7.7% of the patients. A cutpoint of five would eliminate 510 (48.3%) hospitals caring for 12.6% of the patients.
Using a volume cutoff of one case in the prior year, the mean observed/expected ratio for 60-day mortality was 1.14 for hospitals with a prior year volume of 1 case per year vs an observed/expected ratio of 1.05 for those with a volume greater than 1 (Table 3). Among the centers with a volume of one, 22.7% had an observed/expected ratio for 60-day mortality of at least 1 (indicating worse than expected performance) and 77.3% had an observed/expected ratio of less than 1. The mean observed/expected ratio for 1-year mortality was 1.06 at hospitals with a volume of 1 vs 1.07 at those with higher volume; the observed/expected ratio for 5-year mortality was 1.08 and 1.05 for the hospital groups, respectively.
Similarly, with a volume cutoff of three cases in the prior year, the observed/expected ratio for 60-day mortality for the low-volume centers was 1.08 vs 1.05 at higher volume centers. The corresponding observed/expected ratios for 1-year mortality were 1.08 vs 1.06 and 1.07 vs 1.04 for 5-year mortality at the low (three cases or fewer per year) compared with higher volume centers, respectively. For all of the metrics, 32.2–32.9% of hospitals with a prior year volume of three cases or fewer per year had an observed/expected ratio of less than 1, suggesting better than predicted performance.
Implementing a minimum-volume standard of one case in the prior year would result in one fewer deaths for every 198 patients at 60 days, for every 613 patients at 1 year, and for every 62 patients at 5 years (Table 4). Similarly, implementation of a volume threshold of three cases in the prior year would result in one fewer death for every 594 patients at 60 days, for every 307 patients at 1 year, and for every 62 patients at 5 years.
These data suggest that a large number of hospitals have a low annual surgical volume of patients with ovarian cancer. Implementing minimum-volume standards for hospitals would result in restricting care at a significant number of hospitals. Further, despite the fact that the outcomes are in general inferior at low-volume centers, a significant number of low-volume hospitals deliver outcomes that are better than predicted.
The association between higher volume and improved outcomes provides a strong rationale for minimum-volume standards for ovarian cancer. Numerous studies have found that outcomes are superior for women treated by high-volume surgeons and centers and when treatment is provided by specialists with expertise in gynecologic oncology.4–9 One report of nearly 46,000 ovarian cancer patients found that patients treated at the lowest volume hospitals had a 14% increase in mortality.6 Treatment at higher volume hospitals has been consistently associated with delivery of guideline-based care which likely contributes to the improved outcomes at these centers.5,6,21,22 Similarly, we noted that overall mortality rates were higher at low compared with high-volume hospitals.
Based on the association between surgical volume and outcomes, there has been a concentration of care for many high-risk surgical procedures to a smaller number of surgeons and centers that has in turn led to improved outcomes for some procedures.10 Although many of these efforts have been the result of public reporting of data, more recently there has been interest in developing more formal minimum-volume standards for hospitals and physicians.11,12 For ovarian cancer, there has been some concentration of care to a higher volume hospitals and specialized physicians. Prior work for Scandinavia suggests that concentration of care to high-volume centers can not only improve guideline adherence, but also improve survival.23
Although implementing minimum hospital-level volume standards for ovarian cancer is appealing, our data highlight a number of practical concerns. First, we found that a large number of hospitals have a very low annual volume of patients who undergo surgery for ovarian cancer. Implementing even a very low minimum-volume standard would dramatically alter where care for ovarian cancer could be rendered. For example, a minimum-volume standard of only four or more cases in the prior year would have meant that in 2015, 35% of the hospitals in the United States that cared for women with ovarian cancer could no longer treat such patients. Proposed hospital-level standards have ranged from 20 cases per year for pancreatic and esophageal cancer surgery to 50 cases per year for hip and knee arthroplasty.24 Such a minimum-volume standard for ovarian cancer would result in a significant concentration of care to the highest volume centers. Such regionalization of care is often unpopular with patients and may limit access to care for residents of rural regions and minority and underserved populations.25–28
Second, although on average lower volume centers have inferior outcomes, a relatively large number of low-volume hospitals have better than expected survival. Using the theoretical cutoff of three cases in the prior year, 51% of the low-volume hospitals had a 2-year mortality that was lower than predicted based on their case mix, and more than three quarters of the low-volume centers had a 60-day mortality rate that was better than expected. An arbitrary volume standard may be unnecessarily punitive for lower volume centers with good outcomes. Incorporation of other metrics, such as adherence to process measures, may be more appropriate for ovarian cancer.1,29 A prior study of hospital-level care for ovarian cancer found that outcomes were significantly better at low-volume centers that rigorously adhered to evidence-based treatment guidelines for ovarian cancer.1 Alternatively, instead of an arbitrary cutpoint for all hospitals, a hospital's prior performance and outcomes data could be used as a metric.
Finally, although outcomes for ovarian cancer are superior at high-volume centers, the magnitude of this differential in outcomes is relatively small. For example, implementing a minimum-volume standard of two cases in the prior year (absolute difference in observed/expected ratios for 60-day mortality of 0.06 for low vs higher volume centers) would result in one fewer deaths for every 297 patients treated.
Although this study benefits from a large sample size, we acknowledge a number of important limitations. First, although the National Cancer Database captures approximately 70% of newly diagnosed cancer cases in the United States, the dataset may not be representative of all hospitals. Second, our primary analysis focused on mortality. For ovarian cancer, perioperative morbidity and complications are strongly associated with the aggressiveness of the surgical effort and prior studies have shown that complications are often higher at high-volume hospitals.30,31 We therefore believe that mortality is the most important metric and we analyzed short- and long-term survival. Third, our study focused on hospital volume. Physician volume and specialty also play a role in ovarian cancer outcomes and clearly merit further study. Fourth, we examined the association between the hospital that rendered initial treatment and outcomes. Women with ovarian cancer often receive numerous therapies over the course of their disease and thus many factors aside from the treating hospital influence outcomes. To mitigate this bias, we examined short- and long-term outcomes. In the cohort stage and grade were missing in a sizable number of patients. Although we report 5-year survival rates, a priori the objectives of our analysis were on 60-day and 2-year survival.
Developing strategies to centralize care for the surgical treatment of ovarian cancer is clearly challenging. Although prior work has clearly demonstrated that regionalized care for ovarian cancer can improve outcomes, our findings highlight some of the practical difficulties with centralization. Given that a large number of hospitals will likely be affected by even modest minimum-volume standards, any proposed strategies for regionalized care should result in clinically meaningful improvements in outcomes. Further, careful consideration must be given to minimize the burden to rural communities and underserved populations. Prior work already demonstrates significant gaps in access to gynecologic oncology care in many areas in the United States32 Given the large variability of outcomes among low-volume hospitals, other quality improvement initiatives may achieve a better balance of improving outcomes while ensuring all patients have access to care.
1. Wright JD, Chen L, Hou JY, Burke WM, Tergas AI, Ananth CV, et al. Association of hospital volume and quality of care with survival for ovarian cancer. Obstet Gynecol 2017;130:545–53.
2. Birkmeyer JD, Siewers AE, Finlayson EV, Stukel TA, Lucas FL, Batista I, et al. Hospital volume and surgical mortality in the United States. N Engl J Med 2002;346:1128–37.
3. Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. Surgeon volume and operative mortality in the United States. N Engl J Med 2003;349:2117–27.
4. Bristow RE, Chang J, Ziogas A, Randall LM, Anton-Culver H. High-volume ovarian cancer care: survival impact and disparities in access for advanced-stage disease. Gynecol Oncol 2014;132:403–10.
5. Cliby WA, Powell MA, Al-Hammadi N, Chen L, Philip Miller J, Roland PY, et al. Ovarian cancer in the United States: contemporary patterns of care associated with improved survival. Gynecol Oncol 2015;136:11–7.
6. Bristow RE, Palis BE, Chi DS, Cliby WA. The National Cancer Database report on advanced-stage epithelial ovarian cancer: impact of hospital surgical case volume on overall survival and surgical treatment paradigm. Gynecol Oncol 2010;118:262–7.
7. Bristow RE, Zahurak ML, Diaz-Montes TP, Giuntoli RL, Armstrong DK. Impact of surgeon and hospital ovarian cancer surgical case volume on in-hospital mortality and related short-term outcomes. Gynecol Oncol 2009;115:334–8.
8. Vernooij F, Heintz AP, Coebergh JW, Massuger LF, Witteveen PO, van der Graaf Y. Specialized and high-volume care leads to better outcomes of ovarian cancer treatment in The Netherlands. Gynecol Oncol 2009;112:455–61.
9. Vernooij F, Heintz AP, Witteveen PO, van der Heiden-van der Loo M, Coebergh JW, van der Graaf Y. Specialized care and survival of ovarian cancer patients in The Netherlands: nationwide cohort study. J Natl Cancer Inst 2008;100:399–406.
10. Finks JF, Osborne NH, Birkmeyer JD. Trends in hospital volume and operative mortality for high-risk surgery. N Engl J Med 2011;364:2128–37.
12. Birkmeyer JD, Finlayson EV, Birkmeyer CM. Volume standards for high-risk surgical procedures: potential benefits of the Leapfrog initiative. Surgery 2001;130:415–22.
13. Ruiz MP, Chen L, Hou JY, Tergas AI, St Clair CM, Ananth CV, et al. Effect of minimum-volume standards on patient outcomes and surgical practice patterns for hysterectomy. Obstet Gynecol 2018;132:1229–37.
14. Adam MA, Thomas S, Youngwirth L, Hyslop T, Reed SD, Scheri RP, et al. Is there a minimum number of thyroidectomies a surgeon should perform to optimize patient outcomes? Ann Surg 2017;265:402–7.
16. The National Cancer Database. Available at: http://www.facs.org/cancer/ncdb/index.html
. Retrieved March 10, 2012.
17. Bilimoria KY, Stewart AK, Winchester DP, Ko CY. The National Cancer Database: a powerful initiative to improve cancer care in the United States. Ann Surg Oncol 2008;15:683–90.
18. Boffa DJ, Rosen JE, Mallin K, Loomis A, Gay G, Palis B, et al. Using the national cancer database for outcomes research: a review. JAMA Oncol 2017;3:1722–8.
19. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613–9.
20. Liu Y, De A. Multiple imputation by fully conditional specification for dealing with missing data in a large epidemiologic study. Int J Stat Med Res 2015;4:287–95.
21. Bristow RE, Chang J, Ziogas A, Anton-Culver H. Adherence to treatment guidelines for ovarian cancer as a measure of quality care. Obstet Gynecol 2013;121:1226–34.
22. Bristow RE, Chang J, Ziogas A, Campos B, Chavez LR, Anton-Culver H. Impact of national cancer Institute comprehensive cancer centers on ovarian cancer treatment and survival. J Am Coll Surg 2015;220:940–50.
23. Dahm-Kahler P, Palmqvist C, Staf C, Holmberg E, Johannesson L. Centralized primary care of advanced ovarian cancer improves complete cytoreduction and survival—a population-based cohort study. Gynecol Oncol 2016;142:211–6.
24. Low volume hospitals: what to ask. Available at: https://www.usnews.com/news/articles/2015/05/18/low-volume-hospitals-what-to-ask
. Retrieved April 24, 2018.
26. Symer MM, Abelson JS, Yeo HL. Barriers to regionalized surgical care: public perspective survey and geospatial analysis. Ann Surg;2019;269:73–78.
27. Finlayson SR, Birkmeyer JD, Tosteson AN, Nease RF Jr. Patient preferences for location of care: implications for regionalization. Med Care 1999;37:204–9.
28. Alvino DML, Chang DC, Adler JT, Noorbakhsh A, Jin G, Mullen JT. How far are patients willing to travel for gastrectomy? Ann Surg 2017;265:1172–7.
29. Auerbach AD, Hilton JF, Maselli J, Pekow PS, Rothberg MB, Lindenauer PK. Shop for quality or volume? Volume, quality, and outcomes of coronary artery bypass surgery. Ann Intern Med 2009;150:696–704.
30. Uppal S, Spencer RJ, Rice LW, Del Carmen MG, Reynolds RK, Griggs JJ. Hospital readmission as a poor measure of quality in ovarian cancer surgery. Obstet Gynecol 2018;132:126–36.
31. Wright JD, Herzog TJ, Siddiq Z, Arend R, Neugut AI, Burke WM, et al. Failure to rescue as a source of variation in hospital mortality for ovarian cancer. J Clin Oncol 2012;30:3976–82.
32. Shalowitz DI, Vinograd AM, Giuntoli RL II. Geographic access to gynecologic cancer care in the United States. Gynecol Oncol 2015;138:115–20.