Background As ovarian cancer treatment shifts to provide more complex aspects of care at high-volume centers, almost a quarter of patients, many of whom reside in rural counties, will not have access to those centers or receive guideline-based care.
Objective To explore the association between proximity of residential zip code to a high-volume cancer center with mortality and survival for patients with ovarian cancer.
Methods The National Cancer Database was queried for cases of newly diagnosed ovarian cancer between January 2004 and December 2015. Our predictor of interest was distance traveled for treatment. Our primary outcomes were 30-day mortality, 90-day mortality, and overall survival. The effect of treatment on survival was analyzed with the Kaplan-Meier method. Multiple logistic regression for binary outcomes and Cox proportional hazards regression for overall survival were used to assess the effect of distance on outcome, controlling for potential confounding variables.
Results A total of 115 540 patients were included. There was no statistically significant difference in 30- or 90-day mortality among any of the travel distance categories. A statistically significant decrease in 30-day re-admission was found among patients who lived further away from the treating facility. A total of 105 529 patients were available for survival analysis, and survival curves significantly differed between distance strata (p<0.0001). The adjusted regression models demonstrated increased long-term mortality in patients who lived farther away from the treating facility after controlling for potential confounding.
Conclusion Although 30- and 90-day mortality do not differ by travel distance, worse survival is observed among women living >50 miles from a high-volume treatment facility. With a national policy shift toward centralization of complex care, a better understanding of the impact of distance on survival in patients with ovarian cancer is crucial. Our findings inform the practice of healthcare delivery, especially in rural settings.
- ovarian cancer
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The 30- and 90-day mortality from ovarian cancer do not differ by travel distance.
Long-term mortality is increased in patients who live farther away from the treating facility.
Worse survival is observed among women living >50 miles from a high-volume treatment facility.
Geographic analyses of health disparities are crucial for ensuring that our system of cancer care delivery provides equitable access. Studies demonstrating improved survival for women with ovarian cancer when treated by a gynecologic oncologist have resulted in changes to national guidelines; now emphasizing the importance of evaluation by a gynecologic oncologist prior to initiation of therapy.1 Access to guideline-adherent care remains an obstacle, especially for those residing in more rural counties. It is estimated that 25% of women with ovarian cancer receive their treatment at very low-volume hospitals and do not receive care consistent with best practice guidelines.2 Despite efforts aimed at centralizing more complex aspects of care to high-volume centers, many patients face a worse survival because of travel distance.2–4
Geographic variation in treatment of gynecologic cancers is also based on specialist availability. The demand for visits to oncologists is predicted to increase dramatically, especially in rural/underserved locations. This rise is driven largely by the expected doubling of the number of Americans older than 65 and an increase in those diagnosed with, or surviving, cancer; yet among gynecologic oncologists, only 13% work in rural settings.5 A better understanding of the impact of distance and population density on mortality is crucial, and robust data are currently lacking for gynecologic malignancies. The aim of our study is to test the association between travel distance to a high-volume Commission on Cancer center with mortality and survival for women with newly diagnosed ovarian cancer.
We conducted a query of the National Cancer Database, a publicly available database that is jointly sponsored by the American College of Surgeons and the American Cancer Society. The data are reported by all Commission on Cancer accredited hospitals, represent more than 70% of newly diagnosed cancer cases nationwide from over 1500 such hospitals, and encompass more than 34 million historical records. These data are retrospective and de-identified, thus ensuring patient confidentiality. Approval to conduct this study was obtained from the Maine Medical Center institutional review board. The requirement for informed patient consent was waived (IRB study #1212611–1).
We identified adult women with a new primary diagnosis of ovarian cancer (all stages) using the National Cancer Database topography code C56.9, according to the International Classification of Diseases, 10th edition, between January 2004 and December 2015.6 The predictor of interest was travel distance, defined as <25 miles, 25–50 miles, >50–100 miles, and >100 miles. Potential confounding variables included age, race, education level of area of residence, metropolitan status of residence, stage, surgical approach, and facility volume. Demographic, socioeconomic, and clinical variables were categorized according to the Participant User File Data Dictionary provided by the National Cancer Database.7 Identification of low- and high-volume centers was based on criteria set in prior literature.8 The primary outcome of interest was overall survival, defined as months from cancer diagnosis to confirmed death or date of last contact. Secondary outcomes included 30-day re-admission, 30-day mortality, and 90-day mortality.
Categorical variables were described using proportions; continuous variables were described using means and SD. Group differences by travel distance were evaluated using chi square tests for categorical variables and analysis of variance for continuous variables. The Kaplan-Meier method was used to calculate survival curves. We used multiple logistic regression to assess the effect of travel time on outcome, adjusting for potential confounding variables, including age, race, education, stage, rurality, surgical approach, and volume. Similarly, Cox proportional hazard models were used to calculate adjusted hazard ratios. Confounding was evaluated by assessing for a change in OR or HR of 10% after inclusion of the potential confounder. Data analysis was performed using Statistical Analysis Software (SAS) EG7.1.
We identified 194 828 women with newly diagnosed ovarian cancer. Excluding those with non-epithelial histology (n=18 845), those who received all care at a hospital other than the treatment facility of interest or no surgery at the treatment center (n=44 080), those with non-invasive or stage 0 disease (n=13 956), and missing variables (n=2407), the final cohort included 115 540 patients with invasive epithelial ovarian cancer. Of the final cohort, 77 749 (67%) patients lived less than 25 miles, 17 714 (15%) lived between 25 and 50 miles, 12 817 (11%) lived >50–100 miles, and 7260 (6%) patients lived >100 miles from the treating facility (Table 1). The majority of patients were white (n=101 027, 88%), non-Hispanic (n=108 044, 94%), insured (n=108 411, 94%), lived in an area where at least 80% of people completed high school (n=97 848, 85%), lived in a metropolitan area (n=95 418, 83%), and received primary cytoreductive surgery via an open approach (n=48 589, 57%). Using the American Joint Commission on Cancer (AJCC) staging system, 32 209 (28%), 11 935 (10%), 52 713 (46%), and 18 683 (16%) of patients were diagnosed with stage I, II, III, and IV respectively.9
The percentage of Caucasian patients increased with increasing travel distance – 86%, 93%, 93%, and 92% for <25, 25–50, >50–100, and >100 miles, respectively (Table 1). The percentage of patients treated with open surgery (laparotomy) increased with travel distance; 80%, 81%, 83%, and 85% for <25, 25–50, >50–100, and >100 miles, respectively. The percentage of patients treated with neoadjuvant chemotherapy also increased with travel distance; 17%, 19%, 21%, and 25% for <25, 25–50, >50–100, and >100 miles, respectively. The proportion of patients who waited more than 42 days to begin treatment was 3.5% for patients living closest to the reporting hospital and 4.5% for those living furthest away. Of the 10 717 facilities that reported data, 17% had <9 cases per year, 27% had between 9 and 20 cases per year, 27% had between 21 and 34 cases per year, and 30% had greater than or equal to 35 cases per year. Overall, patients who traveled least were more likely to attend low-volume hospitals and less likely to attend academic medical centers (Table 2). There was no difference in 30- or 90-day mortality among any of the travel distance categories even after controlling for potential confounding variables (Figure 1). There was, however, a significantly lower 30-day re-admission rate among patients who lived further away from the treatment facility than for those who lived less than 25 miles from the treatment facility: <25 miles reference category; 25–50, >50–100, and >100 miles from the treatment facility OR=0.85 (95% CI 0.76 to 0.96); OR=0.70 (95% CI 0.66 to 0.88); OR=0.54 (95% CI 0.44 to 0.65), respectively.
A total of 105 529 patients were available for Kaplan-Meier survival analysis after eliminating cases with invalid survival time. Survival curves differed between distance strata (p<0.0001, Figure 2). Cox regression models (Table 3) showed attenuation of the HR for distance strata. Long-term mortality hazard ratios and confidence intervals for distance from unadjusted models were 1.08 (95% CI 1.06 to 1.11), 1.21 (95% CI 1.17 to 1.24), and 1.20 (95% CI 1.16 to 1.25) for patients living 25–50, >50–100, and more than 100 miles away, respectively, compared with patients living within 25 miles of the treatment facility. Adjusting for age, race, ZIP code level education, stage, rurality of residence, surgical approach, and volume attenuated these estimates: compared with those who lived within 25 miles, HR was 0.98 (95% CI 0.94 to 1.03) for patients who lived 25–50 miles away, HR 1.07 (95% CI 1.01 to 1.13) for patients who lived >50–100 miles away, and HR 1.09 (95% CI 1.02 to 1.16) for patients who lived more than 100 miles away (Table 3). In the full model, all covariates were significantly associated with mortality except education. However, due to missing data, the sample size for these fully adjusted models, while large, was substantially smaller than those data used in the unadjusted survival curves and the bivariate models.
Our study demonstrates that women traveling longer distances to access treatment for ovarian cancer do not face higher rates of hospital re-admission or worse short-term outcomes. They do, however, face worse long-term survival outcomes. This is a critical finding because for women with ovarian cancer, undergoing surgery at a high-volume treatment facility is associated with a higher likelihood of receiving guideline-adherent care and improved survival.10 A growing body of literature has sought to identify barriers to receiving guideline-adherent care; most focused on socioeconomic status and race. Rural healthcare and geographic studies have historically not garnered the attention of researchers until recently. The importance of receiving care in a high-volume center is not specific to gynecologic cancers, and a majority of patients are willing to travel further to achieve survival benefit.11 12 Gynecologic oncologists, though, are not evenly distributed across the United States, and this distribution has been associated with worse ovarian cancer mortality.13
Our data support the findings of prior reports—namely, that while patients who travel long distances for treatment do not seem to have worse short-term surgical outcomes, they do have worse long-term outcomes. Our findings demonstrate that increased travel distance significantly reduces overall survival in patients who travel >50–100 miles to their treatment facility. This is even more pronounced in women traveling >100 miles to their treatment facility. One possible explanation for reduced overall survival in patients who travel long distances to their facility is poor continuity of post-operative care or lack of access to appropriate follow-up care locally. For instance, patients may not have increased mortality in the immediate post-operative period, but over a longer period of time, may have increased mortality due to lack of access to local survivorship resources and follow-up care. Additionally, differences in treatment seeking behaviors among patients who travel long distances may also offer a possible explanation for the reduced survival.
One may propose that women traveling further to access care at high-volume centers may be more severely ill at the time of diagnosis, perhaps because of overall poor access to quality medical care, thus contributing to their worse overall survival. Yet, short-term outcomes, as demonstrated above, do not appear to differ by distance traveled. Interestingly, increased travel distance was also not associated with a higher 30-day re-admission rate. This is a difficult finding to further elucidate, in part due to the nature of a large database analysis, which inherently lacks granular data. One potential explanation is that women are receiving adjuvant care at a local, non-index hospital where data would not be added to the National Cancer Database. In other disease sites, for instance, patients living further from the facility where they underwent surgery were more likely to be re-admitted to a non-index hospital.14–16 We did not find this, but do feel that this topic warrants further investigation. Another interesting finding of our analysis is that patients who traveled the least received care at low-volume hospitals and that distance does not appear to be a barrier to patients in rural areas, or >50 miles from a high-volume center, having access to a specialist. Those patients who traveled further, though, were more likely to receive open surgery and neoadjuvant chemotherapy. This may reflect two practices: (1) patients are receiving upfront cytoreduction, generally accomplished via laparotomy or (2) patients who reside further away from a Commission on Cancer center or gynecologic oncologists may present later in their disease course, at which time neoadjuvant chemotherapy is a more appropriate treatment option. This latter scenario warrants further investigation. A third explanation is that practice patterns may be evolving. The use of neoadjuvant chemotherapy versus cytoreduction, or primary debulking surgery, has been greatly debated in the gynecologic oncology literature.17 A prior National Cancer Database analysis analyzed patterns of care between 2003 and 2011 and showed that patients receiving neoadjuvant chemotherapy were older, had stage IV disease, more co-morbid conditions, and were treated in high-volume facilities.18 The study also showed an increased trend toward use of neoadjuvant chemotherapy in patients with stage IIIC disease. Between 2003–2004 and 2009–2011 cohorts, there was an almost twofold increase in the use of neoadjuvant chemotherapy. Although this analysis did include geographic region of care and neighborhood characteristics, it did not stratify by distance to treatment facility.
Our analysis also shows an overall increase in the percentage of patients receiving neoadjuvant chemotherapy (17–25%), with those traveling further being more likely to receive it. Likewise, regionalization of healthcare may have unintended consequences, including potential delay in diagnosis and/or potential delay in treatment. Our analysis confirmed the latter, with a small percentage of patients traveling further starting treatment more than 42 days after diagnosis.
Although this study represents a large sample size, as approximately 70% of new cancer cases are reported to the National Cancer Database, the dataset may not be representative of all hospitals. It might also not represent various models of care. For instance, we included women who had at least some of their care at the reporting facility, including surgery, but women who had all of their care or their primary surgery at a non-reporting hospital were excluded. Our study has other limitations. As with any large database analysis, one loses the granularity of data and so some information, such as exact amount of residual disease or an indication of the radicality or complexity of surgery, is not included. The National Cancer Database permits evaluation only of all-cause mortality, not cancer-specific mortality. This analysis is also subject to the limitations of any retrospective analysis—namely, selection bias. In addition, missing data for several variables, including the long-term survival outcome, led to loss of sample size for the multiple regression models.
Our work highlights the impact of distance traveled to treatment facility on mortality and overall survival. Many women live in rural counties without access to a gynecologic oncologist. Some may choose not to travel, despite a demonstrated survival benefit.11 As complex cancer care becomes more centralized, and fewer gynecologic oncologists practice in rural settings, these patients may face worsening health inequities. Innovative strategies, including both a change in policy and a change in healthcare delivery, will need to be employed to overcome these geographic disparities.
Contributors All authors contributed to to the study design, analysis, and writing of the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data are available in a public, open access repository. The National Cancer Database (NCDB) is a publicly available deidentified dataset.