Associations of Healthcare Affordability, Availability, and Accessibility with Quality Treatment Metrics in Patients with Ovarian Cancer

In the United States (US), ovarian cancer survival has improved steadily over the past few decades; five-year survival for patients with invasive ovarian cancer increased from 27% in 1990–1994 to 37% in 2010–2014 (1). However, although the five-year survival for White ovarian cancer patients increased from 35% to 47% between 1975–77 and 2008–2014, the rate for Black patients declined from 42% to 39% in the same period (2). Lack of access to quality treatment contributes to the observed racial disparities in survival (3–6). In fact, some studies have shown that receipt of guideline adherent ovarian cancer treatment is associated with equivalent survival benefit across racial groups (3, 5, 7). On the contrary, other studies have shown that survival disparities remain even upon receipt of guideline adherent treatment (4, 8), highlighting the complexity of fully characterizing racial and ethnic disparities in ovarian cancer outcomes. Nevertheless, lack of access to treatment is a key driver of survival disparities. Indeed, important racial differences have been reported in indicators of high quality treatment, including receipt of ovarian cancer surgery (9, 10), care by gynecologic oncologists (11–13), and receipt of guideline-recommended chemotherapy (8, 14). For example, treatment by a gynecologic oncologist increases the likelihood of guideline adherent treatment and is associated with improved patient survival (11, 15, 16). Strikingly, only about a third of patients with ovarian cancer receive appropriate surgical care (17), highlighting an urgent need to better characterize measures of healthcare access (HCA) driving treatment quality in diverse patient groups.

Although HCA is fundamental to receipt of quality treatment, it has been inconsistently and narrowly defined in the literature to date. Penchansky and Thomas (18) proposed five dimensions of HCA: Affordability (ability to pay for healthcare), availability (type, quality, and quantity of healthcare resources), accessibility (location of healthcare resources relative to patient), accommodation (organization of healthcare resources relative to patient preferences), and acceptability (patient experience, and quality of patient–provider interaction). Although certain aspects, for example, affordability, have been well characterized in the scientific literature (7, 12, 19–21), others remain largely unexamined. Three of these dimensions: affordability, availability, and accessibility are measurable in administrative claims databases, providing a unique opportunity to examine how they independently and jointly affect cancer treatment outcomes. Thus, the purpose of this study is to examine measures of affordability, availability, and accessibility among non-Hispanic (NH)-Black, Hispanic and NH-White patients with ovarian cancer in relation to two treatment quality metrics: receipt of primary debulking ovarian cancer surgery performed by a gynecologic oncologist and initiation of guideline-recommended systemic therapy.

This was a retrospective cohort study of Surveillance, Epidemiology, and End Results (SEER)–Medicare patients of NH-Black, Hispanic or NH-White race and ethnicity ages 65+ diagnosed with a primary ovarian cancer from 2008 to 2015 (Fig. 1). The SEER-Medicare database combines cancer registry data from 12 US States with linked Medicare claims. Patients with cancers originating from the fallopian tubes or peritoneum were not included in this study due to very small numbers (N < 11). Patients were excluded if they were diagnosed at autopsy/death, or if their ovarian cancer was not their first or second primary tumor in the SEER registry. Patients were also required to have: (i) at least 12 months of continuous enrollment in Medicare fee-for-service parts A and B before diagnosis; (ii) at least one Medicare inpatient, outpatient, or carrier claim with a diagnosis code for ovarian cancer (ICD-9-CM and ICD-10-CM diagnosis codes 183.0 or C569) within two months of the SEER diagnosis; and (iii) continuous fee-for-service Medicare enrollment in the 12 months following their diagnosis date, or until death, whichever came first.

We examined patient demographic and clinical characteristics from SEER data, including race and ethnicity (NH-White, NH-Black, or Hispanic), age at diagnosis, sex, cancer stage at diagnosis (stages I–IV or other), histology at diagnosis (type I epithelial, type II epithelial, or other), marital status (married or not married), geographic region of residence (Midwest, Northeast, South, West, or not available), and residence in a metropolitan area. We used validated coding algorithms to assess patient comorbidities and to calculate the patient's Charlson Comorbidity Index score in the 12 months before ovarian cancer diagnosis using diagnosis codes (Supplementary Table S1) from inpatient, outpatient, and carrier Medicare claims files (22, 23).

Each patient was assigned a primary cancer treatment physician and a primary treatment hospital based on the provider and facility listed on the plurality of cancer claims and the plurality of inpatient and outpatient claims, respectively. This is described in more detail in the Supplementary Methods. Physician specialties were determined from Medicare claims files using Health Care Financing Administration (HCFA) specialty codes (17).

Measures of healthcare affordability included dual enrollment in Medicaid, census tract-level measures of socioeconomic status (SES), and county-level health insurance coverage. A patient's dual Medicaid enrollment status in the 12 months before ovarian cancer diagnosis was sourced from the SEER-Medicare dataset, as were the following SES indicators of the patient's residential census tract at the time of diagnosis drawn from data from the US Census Bureau's American Community Survey: percentage of Black residents, percentage of adults 25+ with less than a high school education, and percentage of households with incomes below the poverty level. Measures of census tract SES were categorized into quartiles and presented as binary variables (highest quartile vs. lower three quartiles). More details on the included affordability measures are presented in the Supplementary Methods.

A patient's residence in a metropolitan or rural area at the time of diagnosis was drawn from SEER data. Distance to main hospital facility was calculated using a straight-line distance approach from the center point of the patient's residential zip code to the center of the hospital facility's zip code (24).

Healthcare availability metrics for the patient's county and healthcare referral regions (HRR) were linked to SEER-Medicare data using year of diagnosis, county and state Federal Information Processing Standards codes, and patient zip codes from the Area Healthcare Resources File and the Dartmouth Atlas Project. More detail on this linkage and the included measures are included in the Supplementary Methods. The National Cancer Institute (NCI) hospital file was used to determine hospital facility-associated availability metrics, including the hospital's ownership (Non-profit/Proprietary/Government), affiliation with a medical school (Yes/No), NCI Cancer Center designation (Yes/No), critical access status (Yes/No), and number of beds (<100/100–200/200+) in the year of the patient's cancer diagnosis. If the hospital's information was missing in a calendar year, the information was imputed as the highest availability value for the hospital recorded in the study time period. Ovarian cancer surgical volume for Medicare beneficiaries for each facility was calculated per year by summing the number of Medicare claims among all SEER-Medicare patients with ovarian cancer in the calendar year with a Current Procedural Terminology (CPT) code for any ovarian surgical procedure (Supplementary Table S2), allowing 1 surgical claim per patient per day.

Among patients with available tumor stage, grade, and histology information who were not classified as having borderline epithelial tumors, receipt of ovarian cancer debulking surgery performed by a gynecologic oncologist was assessed. Ovarian cancer primary debulking surgery performed by a gynecologic oncology specialist in the two months before/six months following a patient's diagnosis date were identified in Medicare claims using CPT codes for debulking surgery (Supplementary Table S3), and physician identifiers and their HCFA specialty codes. The 2013 National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines for Ovarian Cancer specify recommended treatment based on tumor stage, grade, and histology (ref. 25; Supplementary Figs. S1 and S2). Patients were considered to have initiated a guideline-recommended systemic therapy if they had at least one Medicare claim with a CPT code or National Drug Code for administration of one of the recommended systemic therapies for the patient's stage, grade, and histology in the 12 months following the patient's diagnosis date (Supplementary Table S4).

Healthcare affordability, accessibility, and availability metrics were calculated for the full cohort and stratified by patient race and ethnicity. Continuous variables were described using mean (SD) or median (Q1 and Q3). Categorical variables were described using percentages. To describe the associations between race and receipt of quality care and between measures of HCA and quality care, and to assess whether inclusion of HCA measures attenuate observed associations between race and treatment receipt, we modeled associations between race and treatment receipt, race + each individual HCA dimension measures and treatment receipt, and race + all HCA dimension measures and treatment receipt. All models were adjusted for patient clinical covariates. Univariable and multivariable-adjusted modified Poisson regression with robust error estimation (26) were used to assess the relative risk (RR) associations between patient race and ethnicity, HCA dimension variables, and two indicators of access to quality cancer care: (i) receiving debulking surgery performed by a gynecologic oncologist and (ii) initiating a guideline recommended systemic therapy. Because multiple variables were expected to be collinear as they are measures of the same dimensions of HCA, collinearity of model variables was assessed using the variance inflation factor method; variables were excluded from the model if they had a variance inflation factor of >10. Models with adjustment for the following covariable sets were run: (i) patient race and ethnicity and clinical characteristics; (ii) patient race and ethnicity, clinical characteristics, and measures of affordability; (iii) patient race and ethnicity, clinical characteristics, and measures of availability; (iv) patient race and ethnicity, clinical characteristics, and measures of accessibility; and (v) patient race and ethnicity, clinical characteristics, and measures of affordability, accessibility, and availability. Variables excluded due to collinearity were type of hospital ownership and most county-level measures of healthcare availability (due to collinearity with HRR-level measures of availability). To determine whether inclusion of non-epithelial ovarian tumors was influencing the results by introducing heterogeneity, a sensitivity analysis limited to patients with epithelial ovarian tumors was conducted. All analyses were conducted using SAS 9.4.

The SEER-Medicare database is owned and managed by the National Cancer Institute. Information on how to obtain these data is available here: https://healthcaredelivery.cancer.gov/seermedicare/obtain/.

The study cohort included 9,726 patients with ovarian cancer diagnosed from 2008 to 2015: 672 (7.0%) NH-Black, 580 (6.0%) Hispanic, and 8,474 (87%) NH-White (Table 1). The mean age at diagnosis for NH-Black patients was slightly lower than for NH-White patients [NH-Black Mean (SD) = 76.4 (6.9); NH-White Mean (SD) = 77.6 (7.4)]. NH-Black and Hispanic patients had higher total comorbidity burdens at diagnosis than NH-White patients (median score of 3 and 3 vs. 2). NH-Black and Hispanic patients were also more likely to be diagnosed with Stage IV disease (40% and 34% vs. 32%). Hispanic patients were more likely to be diagnosed with unknown stage disease compared with NH-White and NH-Black patients (16% vs. 13% for both NH-White and NH-Black patients).

About 42% of NH-Black patients and 49% of Hispanic patients were dual-enrolled in Medicaid and Medicaid in the year before diagnosis compared with 11% of NH-White patients (Table 2). NH-Black and Hispanic patients were also twice as likely as NH-White patients to live in a census tract in the highest quartile of adult residents without a high school degree, and in the highest quartile of households living below the poverty line.

NH-Black and Hispanic patients were more likely to live in a metropolitan area than NH-White patients (90% and 92% vs. 83%; Table 2). NH-Black patients lived in counties with higher per capita number of hospitals, hospitals with American College of Surgeons cancer programs, hospital acute care beds, primary care physicians, and obstetrician–gynecologist physicians than NH-White patients, whereas Hispanic patients had lower availability in many of these metrics. NH-Black patients were more likely to be seen at a hospital with a medical school affiliation than NH-White patients (61% vs. 50%; Table 3), whereas only a quarter of Hispanic patients were seen at a medical school-affiliated hospital. NH-White patients were seen at facilities with fewer numbers of beds on average than NH-Black and Hispanic patients and were more likely to be seen at a hospital designated as a critical access facility. However, on measures of the quality of available healthcare in the HRR (Table 2), NH-Black patients did slightly worse than NH-White and Hispanic patients, living in regions with HRRs that had higher rates on average of discharges for ambulatory sensitive conditions (65 vs. 58 per 1000) and slightly higher rates of 30-day hospital readmission rates (16.1% vs. 15.5%).

A total of 8,155 patients in the cohort had complete tumor stage, grade, and histology information and met the study eligibility criteria. Of these, 2,343 (28.7%) underwent debulking surgery that was performed by a gynecologic oncologist in the two months before/six months following their diagnosis date; 21% of NH-Black patients compared with 29% of NH-White patients and 28% of Hispanic patients (Table 3). In adjusted Poisson regression analyses (Table 4), NH-Black patients were less likely than NH-White patients [RR, 0.83; 95% confidence interval (CI), 0.70–0.98] to receive debulking surgery by a gynecologic oncologist after adjusting for clinical characteristics; however, after adjustment for measures of affordability, the racial disparity became attenuated (RR, 0.91; 95% CI, 0.77–1.08). In fully adjusted models, dual enrollment in Medicaid remained associated with lower likelihood of debulking surgery performed by a gynecologic oncologist (RR, 0.86; 95% CI, 0.76–0.97) compared with patients with Medicare only, while receiving treatment at a facility in the highest quartile of ovarian cancer patient surgical volume was associated with higher likelihood (RR, 1.12; 95% CI, 1.04–1.21). Patients living in HRRs with higher rates of Medicare beneficiaries being seen regularly by a primary care doctor were also slightly more likely to receive this procedure (RR, 1.02; 95% CI, 1.01–1.03). A sensitivity analysis limited to patients with epithelial ovarian tumors only (N = 8,086) resulted in similar findings (Supplementary Table S5).

A total of 7,702 patients were eligible for this analysis based on 2013 NCCN Guidelines. Of these, 5,213 (67.6%) initiated systemic therapy in the 12 months following diagnosis; 59% of NH-Black patients, compared with 68% of NH-White patients and 68% of Hispanic patients (Table 3). However, in Poisson regression analyses adjusted for clinical characteristics, including age at diagnosis, stage at diagnosis, histology and comorbid conditions, the racial disparity among Black patients was attenuated (RR, 0.96; 95% CI, 0.92–1.01; Table 5). Results by stage indicated that patients with stage IV cancer were the least likely to receive systemic therapy (data not shown). In models with additional adjustment for measures of HCA, the only HCA measure associated with initiation of recommended systemic therapy was Medicaid dual-enrollment; patients with dual enrollment in both Medicaid and Medicare were less likely to initiate recommended therapy compared with those enrolled in Medicare only (RR, 0.94; 95% CI, 0.92–0.97; Table 5). A sensitivity analysis limited to patients with epithelial ovarian tumors only (N = 7,670) resulted in similar findings (Supplementary Table S6)

In the first comprehensive analysis of multiple HCA dimensions among NH-Black, Hispanic, and NH-White patients with ovarian cancer in the SEER-Medicare database, we document significant disparities in measures of healthcare affordability, availability, and accessibility. Although NH-Black patients were more likely to reside in neighborhoods with greater per capita availability of healthcare resources, that is, number of hospitals, primary care physicians and obstetrician–gynecologist physicians, those healthcare resources were of lower quality, for example, facilities with higher readmission rates, compared with those in neighborhoods with NH-White patients. In addition, NH-Black patients with ovarian cancer were less likely than NH-White and Hispanic patients to receive primary tumor debulking performed by a gynecologic oncologist; this difference appeared to be driven partially by differential access to health care. NH-Black patients also had lower rates of recommended systemic therapy initiation, which appeared to be explained largely by clinical factors, including age at diagnosis, stage at diagnosis, tumor histology, and comorbid conditions.

The pattern of higher neighborhood availability of healthcare resources among Black patients compared with White patients but lower utilization of quality treatment has been previously described (9, 27–29). Indeed, multiple factors influence treatment receipt beyond the local availability of healthcare resources, including cost of care, personal preferences, values and goals, trust in the healthcare system, and patient–provider relationships (30). A consequence of broad historic and contemporary inequality and problematic interactions between health systems and local communities, many studies have documented higher rates of medical mistrust and experiences of racism among Black patients (31–34), potentially resulting in treatment delay or avoidance. Efforts to address structural racism and biases within the healthcare system, building long-lasting, trusting relationships with minority communities, and policies to enhance the quality of available healthcare resources in these communities are needed to address inequitable HCA as a key strategy in mitigating disparities in ovarian cancer outcomes.

Furthermore, consistent with our results, other studies have also documented that HCA significantly affects the quality of ovarian cancer treatment. Measures of availability and affordability, respectively, have previously been found to be independent predictors of quality ovarian cancer care (19, 35). For instance, hospital type (i.e., community cancer clinic vs. research intensive; ref. 36) and treatment in hospitals located in rural areas (37) were associated with poor quality care. We expand upon the prior literature by simultaneously examining multiple dimensions of access in the same study population to identify those most significantly associated with ovarian cancer treatment. We observed that treatment at hospitals performing higher volumes of ovarian cancer surgeries, a measure of availability, was associated with substantially increased likelihood of receiving debulking surgery performed by a gynecologic oncologist. This is also consistent with evidence that higher volume facilities often provide higher quality care to patients (38–40) due to the presence of multidisciplinary teams, specialist care, and access to therapies and clinical trials that may not be widely available at less specialized hospitals. NH-Black and Hispanic patients with ovarian cancer were less likely to be treated at high volume facilities compared with NH-White patients, consistent with previous literature (41), highlighting that strategies to equalize access to high-volume academic cancer centers may provide immediate benefit in terms of higher quality of treatment and access to novel clinical trials for this patient group.

A key affordability measure, dual enrollment in Medicaid and Medicare, significantly predicted reduced likelihood of initiation of recommended systemic therapy and reduced risk of receiving debulking surgery performed by a gynecologic oncology specialist. Given that 42% of NH-Black and 49% of Hispanic patients with ovarian cancer in our cohort are dual enrolled, compared with 11% of NH-Whites, this disparity likely contributes significantly to our observed racial differences. Prior studies have documented similar associations of dual eligibility with higher rates of mortality, hospitalizations, and hospital-related mortality when compared with non-dually enrolled individuals, and specifically in the ovarian cancer population, with poor outcomes (42–44). Dually enrolled beneficiaries receive full Medicare and Medicaid benefits, including assistance with premiums. However, eligibility for Medicaid can be complex—whereas the program parameters are federally determined and cover certain mandated eligible groups such as low-income families and individuals receiving Supplemental Security Income (SSI), each state sets their own rules and may impose different eligibility requirements beyond those federally required (45). It is an imperfect measure of an individual's SES; however, it is the only SES indicator in the SEER-Medicare database that is available at the individual level. All other measures of SES are aggregate regional data. Individuals dual-enrolled in Medicaid are likely to experience other social determinant of health barriers such as poverty, housing insecurity, and residence in deprived neighborhoods (43). This is a uniquely at-risk population with significant barriers to care. Further examination of dual-enrolled patients will be necessary to understand unique barriers to treatment quality and to inform public policy regarding Medicare eligibility and requirements.

This study provides a valuable look at real-world disparities in HCA among older adults with OC, highlighting aspects of all three HCA dimensions that impact ovarian cancer outcomes and contribute to observed disparities. However, the analysis is subject to the limitations inherent to retrospective analyses of registry and claims databases. All measures of SES aside from dual enrollment in Medicaid are based on census-tract level averages and may not reflect each patient's individual financial circumstances, and not all low-income Medicare enrollees who are eligible for Medicaid are dual-enrolled; therefore, we likely underestimated the proportion of individuals who have difficulty with healthcare affordability. Administrative claims data do not provide the full clinical picture that physicians use to make treatment recommendations, nor can they capture patient preferences for treatment. In addition, as the purpose of administrative claims is billing, dates of procedures and diagnoses may not be exact and may reflect billing dates as opposed to actual dates of service; to address this, we searched Medicare claims for procedures with dates up to two months before the identified claims ovarian cancer diagnosis date. We also note that the physician specialty codes in Medicare are imperfect, and capture approximately 83%–90% of oncologists (17); this may result in an underestimate of the number of patients seeing gynecologic oncologists. However, there is no reason to believe that this misclassification would occur differentially by patient race, so it is unlikely to influence our finding of lower utilization of oncologists observed among NH-Black patients. The current SEER-Medicare database represents a less ethnically and racially diverse population compared with the total U.S. population and covers approximately 35% of the U.S. population (46), so results may not be fully representative of national trends. In addition, our study focuses on older women with ovarian cancer (65 years or older); therefore, results may not be generalizable to younger women. We also acknowledge that using the SEER-Medicare database restricts to a patient population, which, by definition, has accessed healthcare, so results may not apply to the broader ovarian cancer patient population. However, the use of registry-linked administrative claims that data provide unique opportunities to study underserved patient populations and sicker patient populations, who are often difficult to enroll into prospective cohort studies.

In conclusion, strategies to address racial differences in affordability, availability, and accessibility will be critical for addressing the utilization gap, that is, higher quantity of healthcare resources but lower receipt of quality treatment for NH-Black and Hispanic patients. These efforts will help to mitigate the persistent racial disparities in quality ovarian cancer care.

L.E. Wilson reports grants from National Cancer Institute during the conduct of the study; and grants from AstraZeneca outside the submitted work. M. Pisu reports grants from NIH during the conduct of the study. R.A. Previs reports other support from Myriad Genetics and Natera outside the submitted work. K.C. Ward reports other support from NCI SEER during the conduct of the study. M.J. Schymura reports grants from NCI/Duke University School of Medicine during the conduct of the study. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

T.F. Akinyemiju: Conceptualization, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, writing–original draft, writing–review and editing. L.E. Wilson: Data curation, formal analysis, writing–original draft, writing–review and editing. N. Diaz: Data curation, formal analysis, writing–review and editing. A. Gupta: Writing–review and editing. B. Huang: Formal analysis, methodology, writing–review and editing. M. Pisu: Writing–original draft, writing–review and editing. A. Deveaux: Writing–original draft, writing–review and editing. M. Liang: Writing–original draft, writing–review and editing. R.A. Previs: Writing–review and editing. H.A. Moss: Writing–review and editing. A. Joshi: Writing–review and editing. K.C. Ward: Writing–review and editing. M.J. Schymura: Writing–review and editing. A. Berchuck: Writing–review and editing. A.L. Potosky: Writing–review and editing.

The authors acknowledge the helpful assistance provided by the SEER-Medicare reviewers, Information Management System coordinator Elaine Yanisko, and all the patients whose valuable data contributed to this study. Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under award number R37CA233777 (to T.F. Akinyemiju).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.