Geographic distance to Commission on Cancer (CoC)-accredited and non-accredited hospitals in the United States

Abstract

Purpose:

The Commission on Cancer (CoC) establishes standards to support multidisciplinary, comprehensive cancer care. CoC-accredited cancer programs diagnose and/or treat 73% of patients in the U.S. However, rural patients may experience diminished access to CoC-accredited cancer programs. Our study evaluated distance to hospitals by CoC accreditation status, rurality, and Census Division.

Methods:

All U.S. hospitals were identified from public-use Homeland Infrastructure Foundation-Level Data, then merged with CoC-accreditation data. Rural-Urban Continuum Codes (RUCC) were used to categorize counties as metro (RUCC 1–3), large rural (RUCC 4–6) or small rural (RUCC 7–9). Distance from each county centroid to the nearest CoC and non-CoC hospital was calculated using the Great Circle Distance method in ArcGIS.

Findings:

Of 1,382 CoC-accredited hospitals, 89% were in metro counties. Small rural counties contained a total of 30 CoC and 794 non-CoC hospitals. CoC hospitals were located 4.0, 10.1, and 11.5 times farther away than non-CoC hospitals for residents of metro, large rural, and small rural counties, respectively, while the average distance to non-CoC hospitals was similar across groups (9.4–13.6 miles). Distance to CoC-accredited facilities was greatest west of the Mississippi River, in particular the Mountain Division (99.2 miles).

Conclusions:

Despite similar proximity to non-CoC hospitals across groups, CoC hospitals are located farther from large and small rural counties than metro counties, suggesting rural patients have diminished access to multidisciplinary, comprehensive cancer care afforded by CoC-accredited hospitals. Addressing distance-based access barriers to high-quality, comprehensive cancer treatment in rural U.S. communities will require a multisectoral approach.

INTRODUCTION

Rural patients with cancer experience worse survival outcomes than urban patients,^1,2 due in part to fragmented cancer care delivery systems.^3–5 Care fragmentation can be addressed through coordinated cancer care delivered within a multidisciplinary framework and often involves multispecialty treatment plans (e.g. tumor boards), patient navigation, access to support services (e.g. genetic testing/counseling, nutrition, rehabilitation), and opportunities to participate in clinical trials.^6–8 It has been documented that rural patients have less access to cancer specialists and specific treatment modalities;^9–16 however, access to comprehensive, multidisciplinary care in rural areas is less understood. Previous studies have focused on access to NCI-designated cancer centers and academic medical centers,^9,17,18 which represent less than 1% of hospitals in the U.S. and are almost exclusively located in urban areas.

In contrast to the limited reach of the NCI-designated cancer centers and academic medical centers, there are approximately 1,500 hospitals, freestanding cancer centers, and cancer program networks accredited by the American College of Surgeons’ Commission on Cancer (CoC).¹⁹ CoC-accredited facilities diagnose and/or treat 73% of all cancers in the U.S.²⁰ As an accrediting body, the CoC is dedicated to improving survival and quality of life for patients with cancer through standard setting, prevention, research, education, and the monitoring of comprehensive quality care.²¹ Accreditation standards span the full spectrum of cancer care including diagnosis, treatment and support services.²¹ Recent evidence supports the positive impact of CoC standards on cancer care and outcomes. A national study of CoC-accredited hospitals found that compliance in one of their required quality measures (i.e., removal and examination of at least 12 lymph nodes during colon cancer resection) improved over time and was associated with improved survival.²² Two other studies conducted in states with large rural populations found CoC accreditation to be associated with better performance on four treatment-related quality measures.^23,24

Despite the large and important role of CoC hospitals in cancer care delivery, there are no published studies describing rural disparities in access to these facilities. Thus, the primary objective of this study was to estimate geographic distance to CoC hospitals, assessing potential rural-urban disparities in access. Although CoC hospitals diagnose and treat the vast majority of patients with cancer in the U.S., we hypothesized there are farther distances to these facilities from rural areas compared to urban areas, which could contribute to diminished access to high-quality, evidence-based multidisciplinary cancer care. Because differences in access to CoC hospitals may reflect either fewer rural hospitals in general or a different mix of hospitals in rural areas, we also examined travel distance to non-CoC hospitals. Finally, previous studies have reported variation in access to cancer care across regions of the U.S.,^18,25–27 but none have examined whether rural-urban disparities may also vary across geographic region. To account for potential heterogeneity in terms of available healthcare facilities across rural regions, we estimated distances to hospitals east and west of the Mississippi River, as well as by each census division.

METHODS

Data Sources

A publicly available county shapefile was extracted from the U.S Census Bureau website.²⁸ A feature class/shapefile with the names, county locations, and longitudinal and latitudinal coordinates of U.S. hospitals was extracted from publicly available Homeland Infrastructure Foundation-Level Data (HIFLD).²⁹ The CoC provided the study team a file in April 2021 for this project with the names, city, state and Zone Improvement Plan (ZIP) code of their accredited cancer programs. The HIDFL and CoC datasets were merged by name, city and ZIP code to categorize U.S. hospitals into two mutually exclusive categories: CoC-accredited hospitals (CoC hospitals) and non-CoC-accredited hospitals (non-CoC hospitals). Unmatched observations (N=42 entries from the CoC data) were merged manually with additional information extracted from public websites. Each county’s rural-urban designation was derived from the U.S. Department of Agriculture 2013 Rural-Urban Continuum Codes (RUCC). The RUCC classification scheme includes three metropolitan county categories (distinguished by the population size of their metro area) and six nonmetropolitan county categories (distinguished by their degree of urbanization and adjacency to metro areas).³⁰ Following other studies, we used the RUCCs to categorize counties into three groups: metro (RUCC 1–3), large rural (RUCC 4–6), and small rural (RUCC 7–9).^31–35 To ensure travel distances were not inflated by the geographic separation of Alaska and Hawaii, only counties within the continental U.S. (i.e., contiguous states) were included in the study.

Statistical Analyses

Longitudinal and latitudinal coordinates for county centroids were identified using the ArcMap application of ArcGIS Desktop (version 10.7.1; Esri; Redlands, CA). A full spatial join was performed using ArcMap to determine the nearest CoC and non-CoC hospital for each county. For each county, we separately calculated distances to the nearest CoC hospital and the nearest non-CoC hospital using Great Circle Distance in ArcGIS (i.e., “as the crow flies”) with the latitudinal and longitudinal coordinates of the hospitals and county centroids. These distance measures were summarized as means, standard deviation (SD), median, and interquartile range (IQR). A measure of relative distance was also calculated for each county centroid by dividing the distance to the nearest CoC hospital by the distance to the nearest non-CoC hospital. This unit-less measure helps standardize travel experience and burden across counties in different regions and rurality category. Relative distance was averaged across rurality categories using the PROC SURVEYMEANS procedure in SAS (version 9.4; SAS Institute Inc.; Cary, NC). Inferential statistics were calculated to compare distances across rurality. P-values from chi-squared tests (categorical variables) and ANOVA (means) are reported with statistical significance set at 5%.

Subgroup Analyses

To explore potential variation within rural counties across the U.S., our first subgroup analyses separately replicated our primary methods for counties east and west of the Mississippi River. Additionally, we compared mean distances to CoC and non-CoC facilities for metro, large rural, and small rural counties in each of the nine Census Divisions.

RESULTS

Overall, a quarter of the 5,509 hospitals included in this study (N=1,382) were CoC-accredited. However, this varied across rurality, with 35%, 11% and 4% of the hospitals in metro, large rural and small rural counties, respectively, having achieved CoC-accreditation as of April 2021 (Table 1). In the continental U.S., a third of the counties were designated as small rural. These 1,052 small rural counties contained 30 CoC hospitals (2% of all CoC hospitals) and 794 non-CoC hospitals (19% of all non-CoC hospitals). Metro counties also composed a third of U.S. counties and these 1,160 metro counties contained 1,226 CoC hospitals (89% of all CoC hospitals) and 2,327 non-CoC hospitals (56% of all non-CoC hospitals). Non-CoC hospitals were more common in small rural (96%) and large rural (89%) counties, than in metro counties (65%). CoC hospitals were in 47%, 15%, and 3% of metro, large rural, and small rural counties, respectively. A third of small rural counties had no hospitals, compared with 18% of metro counties and 9% of large rural counties.

Table 1.

Descriptive statistics and geographic distance to U.S. hospitals, by rurality^*

	Metro	Large Rural	Small Rural	p-value
Sample
U.S. counties [row %]	1,160 [37%]	896 [29%]	1,052 [34%]
No hospital (col %) [row %]	206 (18%) [33%]	77 (9%) [12%]	342 (33%) [55%]	<0.001
Any CoC hospital (col %) [row %]	544 (47%) [78%]	123 (14%) [18%]	29 (3%) [4%]
Only non-CoC hospital(s) (col %) [row %]	410 (35%) [23%]	696 (78%) [39%]	681 (65%) [38%]
U.S. hospitals [row %]	3,553 [64%]	1,132 [21%]	824 [15%]
Hospital accreditation status
CoC-accredited (col %) [row %]	1,226 (35%) [89%]	126 (11%) [9%]	30 (4%) [2%]	<0.001
Not CoC-accredited (col %) [row %]	2,327 (65%) [56%]	1,006 (89%) [24%]	794 (96%) [19%]
Average geographic distance to nearest U.S. hospital
CoC hospital, miles
Mean (SD)	18.8 (21.0)	37.5 (28.2)	61.1 (40.2)	<0.001
Median (IQR)	14.8 (5.6–24.0)	31.8 (22.0–46.4)	52.1 (31.9–79.4)
Non-CoC hospital, miles
Mean (SD)	10.8 (8.9)	9.4 (9.4)	13.6 (11.7)	<0.001
Median (IQR)	8.6 (3.8–15.6)	6.1 (3.3–12.8)	10.4 (4.3–19.6)
Relative distance**
Mean (SD)	4.0 (10.8)	10.1 (17.0)	11.5 (44.1)	<0.001

Abbreviations: U.S., United States; col, column; %, percent; CoC, Commission on Cancer; SD, standard deviation.

^*Rurality defined at the county level using Rural-Urban Continuum Codes (RUCC) codes: Metro (1–3), Large Rural (4–6), Small Rural (7–9).

^**Relative distance was defined as the ratio of travel distances between the nearest CoC and non-CoC hospitals.

^***P-values from chi-squared (categorical variables) and ANOVA (means) tests.

Distances to CoC and non-CoC Hospitals

Distance to the nearest CoC hospital ranged from an average of 18.8 miles for metro counties to 37.5 and 61.1 miles for large rural and small rural counties, respectively (Table 1). Average distance to the closest non-CoC hospital was 10.8, 9.4, and 13.6 miles for metro, large rural, and small rural counties, respectively. The relative distance calculations revealed that metro, large rural and small rural counties were 4.0, 10.1, and 11.5 times farther from a CoC hospital than a non-CoC hospital.

Most metro counties (92%) were within 25 miles of a CoC hospital (Figure 1). In contrast, only 41% and 23% of large rural and small rural counties were within 25 miles of a CoC hospital. About 10% of small rural counties were 100+ miles from a CoC hospital. Far less variation was observed in average distance to closest non-CoC hospitals by rurality. Almost all metro (97%), large rural (94%), and small rural (91%) counties were within 25 miles of a non-CoC hospital.

Figure 1.

Distribution of geographic access to CoC and non-CoC hospitals by rurality

Figure 1 shows the proportion of metro, large rural, and small rural counties <25-miles, 25–49 miles, 50–99 miles, 100+ miles to the nearest CoC and non-CoC hospital.

Regional Differences

Figure 2 displays the distribution of CoC and non-CoC hospitals across the U.S. More than half of the metro and large rural hospitals were located east of the Mississippi River (55% and 52%, respectively; Table 2). In contrast, 67% of small rural hospitals were located west of the Mississippi River. The distribution of CoC hospitals also varied by geography and rurality. East of the Mississippi River, CoC hospitals comprised 41%, 15%, and 8% of all hospitals in metro, large rural and small rural counties, respectively. West of the Mississippi River, CoC hospitals comprised 26%, 7% and 1% of all hospitals in metro, large rural and small rural counties, respectively. The 676 small rural counties west of the Mississippi River housed a total of 8 CoC hospitals.

Figure 2.

Location and rurality of CoC hospitals (Panel A) and non-CoC hospitals (Panel B)

Figure 2 shows the distribution of CoC-hospitals (A) and non-CoC hospitals (B) by rural status.

Table 2.

Descriptive statistics and average geographic distance to U.S. hospitals east and west of the Mississippi River, by rurality^*

	Metro	Large Rural	Small Rural	p-value
Sample
East of the Mississippi
U.S. counties (col %)^	741 (64%)	489 (55%)	376 (36%)
U.S. hospitals (col %)^	1,967 (55%)	583 (52%)	269 (33%)
Hospital accreditation status
CoC accredited (col %)^^	816 (41%)	89 (15%)	22 (8%)	<0.001
Non-CoC accredited (col %)^^	1,151 (59%)	494 (85%)	247 (92%)
West of the Mississippi
U.S. counties (col %)^	419 (36%)	407 (45%)	676 (64%)
U.S. hospitals (col %)^	1,586 (45%)	549 (48%)	555 (67%)
Hospital accreditation status
CoC accredited (col %)^^	410 (26%)	37 (7%)	8 (1%)	<0.001
Non-CoC accredited (col %)^^	1,176 (74%)	512 (93%)	547 (99%)
Average geographic distance to U.S. hospitals
East of the Mississippi
Distance (SD) to nearest CoC hospital, miles	13.1 (10.4)	26.2 (14.7)	34.5 (18.8)	<0.001
Distance (SD) to nearest non-CoC hospital, miles	9.9 (7.8)	8.4 (7.6)	11.1 (8.6)	<0.001
Relative (SD) distance**	3.4 (8.2)	8.8 (13.1)	7.0 (14.6)	<0.001
West of the Mississippi
Distance (SD) to nearest CoC hospital, miles	28.9 (29.5)	51.0 (33.9)	75.9 (41.3)	<0.001
Distance (SD) to nearest non-CoC hospital, miles	12.4 (10.4)	10.6 (11.1)	14.9 (13.0)	<0.001
Relative (SD) distance**	5.2 (14.2)	11.7 (20.6)	13.2 (53.7)	<0.001

Abbreviations: U.S., United States; col, column; %, percent; CoC, Commission on Cancer; SD, standard deviation.

^*Rurality defined at the county level using Rural-Urban Continuum Codes (RUCC) codes: Metro (1–3), Large Rural (4–6), Small Rural (7–9).

^{^}Percentage of variable (counties, U.S. hospitals) East or West of Mississippi River.

^{^^}Percentage of each hospital type (CoC or non-CoC) in that region (East or West of Mississippi River).

^**Relative distance was defined as the average ratio of travel distances between the nearest CoC and non-CoC hospitals.

^***P-values from chi-squared (categorical variables) and ANOVA (means) tests

Across all rural designations, distance to CoC-accredited facilities was shorter in counties east of the Mississippi (Table 2). Metro and small rural counties were 13.1 and 34.5 miles, respectively, from the nearest CoC hospital east of the Mississippi River, compared with 28.9 and 75.9 miles west of the Mississippi River. In contrast, distance to the nearest non-CoC accredited facility was similar across areas. Relative distances (from the closest CoC hospital compared to the closest non-CoC hospital) were larger in the western half of the continent as well, ranging from 3.4 (metro counties) to 7.0 (small rural counties) in the east, versus 5.2 (metro counties) to 13.2 (small rural counties) in the west.

Stratifying analyses by Census Division revealed additional patterns. On average, all counties were within 25 miles of a non-CoC hospital, with small differences across county rurality and Census Division (Figure 3). Greater variation was observed in distance to CoC hospitals. Average distance to the closest CoC hospital ranged from 8.5 miles for metro counties of the New England Division to 41.8 miles for metro counties in the Mountain Division. For small rural counties, the average distances ranged from 30.8 miles in East South Central to 99.2 miles in the Mountain Division.

Figure 3.

Distance to nearest CoC and non-CoC hospitals by Census Division by rurality

Figure 3 shows the average distance to the nearest CoC and non-CoC hospital for metro, large rural, and small rural counties within each of the Census Divisions. Whiskers denote 95% confidence intervals and p-values reported from ANOVA.

DISCUSSION

This study was the first to examine rural-urban disparities in geographic access to multidisciplinary, comprehensive cancer care at CoC-accredited hospitals. Overall, a quarter of hospitals were CoC-accredited, but considerable variation was observed across counties by rurality: 35% of hospitals in metro counties were CoC-accredited, compared with 4% in small rural counties. The vast majority of CoC hospitals (89%) were in metro counties and only 2% of all CoC hospitals were located in small rural counties. Consistent with previous studies that examined access to oncology specialists and NCI-designated cancer centers,^9,17,26 we found travel distance to the closest CoC hospital to be greater in rural areas. Subgroup analyses revealed additional variation by geographical region. Disparities in access to CoC hospitals were more pronounced in counties west of the Mississippi River, especially in the small rural counties of the West North Central, West South Central and Mountain Divisions. Our results also indicate the observed disparity in access to CoC hospitals is not simply the result of fewer hospitals in rural areas; nearly all counties were located within 25 miles of a non-CoC hospital and distance to the closest non-CoC hospitals did not differ meaningfully across rurality or geographic region.

CoC-accreditation uses standards to provide a framework for multidisciplinary cancer care and can be useful for hospitals to optimize the cancer care they deliver. The most recent accreditation standards include requirements related to structure, process, and outcomes and address cancer care from prevention to survivorship and end-of-life.¹⁹ The disparities in resources, capacity, and sustainability of oncology workforce, services, technologies, and infrastructure across rurality are all well documented,^16,36–38 and the lack of CoC-accredited hospitals in rural counties may reflect the substantial resources and institutional support necessary to obtain and maintain accreditation in the context of these larger structural and financial challenges.³⁹ Important supportive services, such as patient navigation, nutritional assessment and support, rehabilitation services, and genetic counseling are included alongside treatment guidelines in the current CoC standards. These services impact both quality of life and survival but can be difficult to offer in less-resourced rural settings, given the current staffing models and reimbursement challenges. Smaller surrounding populations in rural areas result in lower patient volume which impacts hospital viability and further challenges the possibility of achieving accreditation. Indeed, administrators and providers of rural hospitals in one study reported difficulties pursuing and maintaining CoC accreditation.³² Even so, the respondents in this study acknowledged that the accreditation process provided opportunities and incentives to expand coordinated and high-quality care services beyond treatment, and stressed the importance of CoC accreditation for its accountability to standards and high-quality patient care.³²

An integral aspect of CoC accreditation is the collection and reporting of standardized cancer data elements.¹⁹ This information is used for monitoring of delivery of care, treatment patterns and outcomes as well as in cancer prevention and control activities.^21,40 High-quality data are essential for accurate assessment, effective improvement efforts, and efficient use of resources. However the resources required to create, interpret and act on these data impede accreditation for rural hospitals.³² Although cancer is a reportable condition, rural, unaccredited facilities may be unlikely to employ their own cancer registrar or analyze their own data for quality improvement purposes, given the lack of incentive and high financial and human resource costs. There is some evidence to support this. Performance on a well-established surgical quality measure did not change over time for rural, non-CoC hospitals in one study, remaining almost 10 percentage points lower than the 85% quality threshold set by CoC.²⁴

Evidence suggests that CoC accreditation for rural hospitals is feasible with additional support from local higher-resourced institutions. In our study, the largest concentration of small rural CoC hospitals was found in Kentucky, the catchment area of the University of Kentucky Markey Cancer Center Affiliate Network (MCCAN). This network leverages the robust resources of its NCI-designated cancer center to support cancer care and accreditation and requires affiliate sites to obtain CoC accreditation within three years of membership if not already accredited.³² MCCAN has effectively supported affiliate sites to improve compliance on treatment-based performance measures and achieve CoC accreditation.²³ Some viewed network affiliation as a tangible way for rural hospitals to demonstrate their quality and performance, and to help fill otherwise-unmet needs for their rural patients with cancer.³² Such an intervention has the potential to improve the quality of cancer care in other rural contexts but requires further study to assess its effectiveness when translated to other settings.

Addressing distance-based access barriers to high-quality, comprehensive cancer treatment at scale in rural U.S. communities will require a multisectoral approach which adapts interventions to local contexts and leverages existing health system resources. MCCAN core functions were recently identified and published⁴¹ and the cancer care intervention is indeed a complex, multi-level intervention. Current adaptation methods only include single-level evidence-based interventions,⁴² and additional work is needed to understand how best to implement systemic interventions such as MCCAN in a different context.

Limitations

There are several limitations to our study. First, multidisciplinary care may occur in the absence of CoC accreditation. We chose this proxy because of the rigor and comprehensiveness of CoC standards regarding patient-centered cancer care delivery and quality. Additionally, cancer care services do not always occur in a hospital. Chemotherapy, genetic counseling, nutrition, and rehabilitation services along with other supportive care can be provided in a clinic or community setting. The data limited our ability to identify the specific services provided at each hospital as well as any services available outside the hospital setting. This study analyzed distance to, but not utilization of services from a CoC or non-CoC hospital. While distance is important, other factors that we did not attempt to measure could influence a patient’s decision to utilize healthcare at a specific hospital. Neither should we assume that patients only receive care at the nearest hospital using Great Circle Distance (i.e., “as the crow flies”). The results reported in this study are the lower-bound estimate of actual driving distances and travel time, which are likely much longer, especially for residents traversing mountainous regions or commuting through densely populated cities. Our study sample included county-level population data, not actual cancer patient data. Distances were calculated from county centroids, not residential address, so the actual distances traveled could differ depending on how the population is distributed within a county and if this differs systematically across rurality or geography. The same distance can also reflect very different travel experiences across regions and rurality (i.e., the burden of driving 5 miles in New York City is not the same as 5 miles on a rural highway in Montana). Additionally, our analyses excluded Alaska and Hawaii. Distance to CoC and non-CoC hospitals in these states may differ from the continental U.S., so our results should not be generalized to these two states. Finally, as we reaffirmed in our subgroup analyses, health access measures can vary within rural populations. While we focused on identifying variation by region, there could also be variation by rural definition. Our categories of metro, large rural, and small rural were consistent with prior geographic analyses of RUCCs but we acknowledge that alternative definitions or categories could yield different results given the heterogeneity of rural populations in the U.S.⁴³ Moreover, the changing population dynamics and corresponding rural definition updates could also influence future analyses measuring access to care.

CONCLUSION

Despite having relatively similar distance to non-CoC hospitals, CoC hospitals are farther away from large and small rural counties than metro. These novel findings identified disparities in geographic access to multidisciplinary cancer care. Multisectoral approaches to enhance cancer care in existing rural hospitals should be explored.