Geospatial Analysis of Commission on Cancer-Accredited Centers Within Cancer Care Utilization–Based Catchment Areas

Abstract

PURPOSE

The Commission on Cancer (CoC) seeks to expand access to high-quality care through community engagement standards targeting centers’ catchment areas and efforts to accredit centers in more areas including rural hospitals. Little is known about the social, environmental, and geographic characteristics of their catchment areas. To support future investigation into the impact of CoC-accredited centers, this study compares characteristics of cancer care utilization–based catchment areas, termed Cancer Service Areas (CSAs), with and without CoC-accredited centers.

METHODS

Geocoded CoC-accredited centers and cancer care patient flows extracted from Medicare claims data were used to delineate CSAs using a spatially constrained community detection method. Characteristics including environmental justice index (EJI), social vulnerability index (SVI), rurality, travel time, and localization index (LI, a ratio of cancer care received by patients within a CSA) were aggregated by CSA. A logistic regression model was created to evaluate characteristics associated with the presence of a CoC-accredited center within a CSA.

RESULTS

Six hundred sixty-eight CSAs were defined, of which 511 CSAs had at least one CoC-accredited center. CSAs with CoC-accredited centers had lower health vulnerability (odds ratio [OR], 0.65 [95% CI, 0.427 to 0.993]) and lower racial and ethnic minority status vulnerability (OR, 0.61 [95% CI, 0.424 to 0.886]), but no differences for other components of the EJI or SVI. These CSAs also had higher LIs, meaning patients remained in their local CSA for care (OR, 9.00 [95% CI, 2.408 to 33.640] for high v low LIs).

CONCLUSION

Minority and comorbid populations may have more difficulty accessing cancer center care, further exacerbating observed variations in cancer outcomes. Cancer centers may address this by broadening their outreach into at-risk catchment areas.

INTRODUCTION

There is marked regional variation in cancer mortality in the United States.¹ Although multifactorial, one contributing factor is access to high-quality cancer care.^2,3 Care at an accredited cancer center is associated with greater receipt of evidence-based care and superior oncologic outcomes.^4–6 However, care at these centers may not be accessible to all patients. Moreover, lack of access to a cancer center may exacerbate existing differences in cancer care utilization and contribute to observed variation in regional cancer mortality. One strategy that some cancer centers are using is the creation of cancer care networks through relationships with affiliate sites.⁷ Effects of these relationships are mixed, and methods for optimal cancer care network design are unknown.⁸ To inform regional care delivery strategies and the role of cancer centers in expanding their regional networks, we sought to understand the characteristics of cancer center catchment areas, including their sociodemographic, environmental, and geographic factors.

Cancer care utilization–based catchment areas, or Cancer Service Areas (CSAs), were previously defined by a novel and scale-flexible method in Geographic Information Systems (GIS) using flows of patients with cancer (number of services received by patients with cancer from their care providers) generated from Medicare claims data.^9–11 Distinct from the Dartmouth hospital referral regions that were defined to represent health care markets for tertiary care (ie, cardiovascular and neurosurgery), CSAs are cancer-specific and effective in capturing regional cancer care markets.¹¹ Previous work examined the distribution of Association of American Cancer Institutes (AACI) centers, finding that CSAs with AACI cancer centers had a higher proportion of patients treated in these areas.⁹ However, AACI cancer centers are a select group of only 94 National Cancer Institute (NCI)–designated and academic cancer centers, predominantly located in populous and urbanized areas. These centers may be inaccessible to large swaths of the country, especially those living in rural or remote areas. By contrast, there are approximately 1,000 Commission on Cancer (CoC)–accredited centers, which are more evenly distributed than AACI cancer centers and include not only NCI-designated cancer centers but also accredited community cancer centers and local affiliate programs. Their catchment areas are expected to be distinct from those of AACI centers.¹² Examining the distribution of CoC-accredited centers within utilization-based CSAs may provide a better understanding of geographic access to cancer center care. Furthermore, these more numerous centers, if more accessible, may be more effective in addressing the observed regional variation in cancer outcomes.

To support future investigation into the impact of CoC-accredited centers, this study aimed to build upon our previous work⁹ by accounting for the presence of CoC-accredited centers to define utilization-based CSAs and to examine the socioeconomic, environmental, and geographic characteristics of CSAs with and without CoC-accredited centers. Cancer center presence within cancer care utilization–based catchment areas is a modifiable factor that can be targeted through health care system design, community engagement and outreach, and future policy decisions. For example, strategies such as building affiliate relationships with existing cancer centers, establishing new designated cancer centers within key regions, or promoting referral patterns that include accredited cancer centers could be used to improve how cancer care is delivered.

METHODS

The study used various data sets and methods which were integrated into a workflow diagram in Appendix Figure A1. Details are provided in each subsection.

Creation of CSAs

Data used to define CSAs were CoC-accredited centers and ZIP-to-ZIP flows of patients with cancer in the United States (Fig 1). A list of CoC-accredited centers including their names, accreditation levels, and addresses was obtained from the 2021 programs listed on the American College of Surgeons’ Web site.¹³ Additionally, a list of NCI-designated cancer centers excluding basic laboratory cancer centers was obtained from the NCI’s Web site.¹⁴ These two lists were merged without duplications and their addresses were geocoded as points to represent the locations of CoC-accredited centers. For a secondary analysis, CSAs were classified into four groups: none (ie, no CoC-accredited center), low, middle, and highest, based on the highest level of CoC-accredited centers present in a CSA which incorporate case volume, teaching, and research. The low group of CSAs contained community cancer programs, hospital associate cancer programs, and/or integrated network cancer programs; the middle group of CSAs contained at least one comprehensive community cancer program or academic comprehensive cancer program; and the highest group of CSAs contained an NCI-designated comprehensive cancer center and/or NCI-designated network. ZIP-to-ZIP flows of patients with cancer were extracted from January 2014 to September 2015 Medicare enrollment and claims data provided by the Centers for Medicare and Medicaid Services with an approved Data Use Agreement and Institutional Review Board protocol. Data processing was completed as previously detailed.⁹ The travel time associated with each flow was extracted from a ZIP-to-ZIP drive time matrix.¹⁵

FIG 1.

CoC-accredited centers and major ZIP-to-ZIP flows of patients with cancer (flows with the number of services ≥ 20 and travel time ≤ 12 hours are shown) in the United States. Note: The two ends of the patients’ flows (or blue lines) represent the ZIP codes of patients with cancer and the ZIP codes of their care providers, and the width of the flows represents the number of services received by patients with cancer from their care providers. CoC, Commission on Cancer.

A spatially constrained community detection method developed in ArcGIS Pro was used to define CSAs with maximal cancer care utilizations within the unit and minimal utilizations between them, meaning that cancer care utilizations were concentrated within each CSA and rarely spanned two or more CSAs. To ensure each CoC-accredited center had a distinctive CSA serving a relatively large population, parameters were set including a minimum population of 120,000 within each CSA, and in cases where CoC-accredited centers within the same CSA were more than 60-minute apart from each other, the CSA was divided into multiple CSAs with each containing at least one CoC-accredited center to obtain the final CSAs.⁹

Explanatory Variables and Outcomes

Publicly available data were used to compute social, environmental, and geographic variables across CSAs (Appendix Table A1). The 2014 social vulnerability index (SVI) and 2022 environmental justice index (EJI) were obtained from the Centers for Disease Control and Prevention (CDC) at the census tract level.^16,17 The SVI comprises four themes: socioeconomic status, household characteristics, racial and ethnic minority status, and housing type and transportation. The EJI comprises three modules: social vulnerability (a module, as differentiated from the SVI), environmental burden, and health vulnerability. Values for each theme of the SVI or module of the EJI are percentile ranks and range from 0 to 1, with higher values indicating greater vulnera-bility¹⁸ or cumulative environmental burden.¹⁹ The 2020 health care infrastructure and access theme was obtained from the Office of Minority Health and CDC at the county level to measure access to general health care, including hospitals and primary care physicians.²⁰ Values are again percentile ranks and range from 0 to 1, with higher values indicating greater vulnerability.²⁰ Rurality was obtained from the Rural-Urban Commuting Area (RUCA) Codes provided by the USDA Economic Research Service at the census tract level.²¹ Primary RUCA ranges from 1 to 10, representing more metropolitan areas to more rural areas. Data for each variable were aggregated at the CSA level and weighted by the population within each CSA using ArcGIS Pro. Because EJI data were not available in Hawaii and Alaska, they were calculated for CSAs only within the 48 contiguous states. For the health care infrastructure and access theme, CSAs within the same county were each assigned the value of that county. Using ZIP-to-ZIP flows of patients with cancer and ZIP code areas, we calculated localization index (LI, the ratio of cancer care received by patients within a CSA), the weighted average travel time (minutes) of patients from each CSA, area size (meter²), population, and population density (persons/km²) for each CSA using ArcGIS Pro (Appendix Table A1). CSAs were used as the unit of geospatial analysis.

The primary outcome was the presence or absence of a CoC-accredited center within a CSA. In a secondary analysis, the association between LI and the highest accreditation level of center within a given CSA was evaluated to determine whether individuals were more likely to remain in their CSA for cancer care when it contained a higher level center, such as an NCI-designated center, compared with those with a lower level center, such as a community cancer program.

Statistical Analyses

For all explanatory variables across CSAs with and without CoC-accredited centers, Levene’s unpaired t tests and Mann-Whitney U tests were used to compare parametric and nonparametric continuous variables, respectively. A correlation matrix was used to assess potential multicollinearity among explanatory variables, and those with variance inflation factor (VIF) > 5 were either excluded from the model or accounted for using interaction terms. Multivariable logistic regression was used to evaluate associations between explanatory variables and the presence of a CoC-accredited center within CSAs. The outcome of the presence or absence of a CoC-accredited center was binary and did not differentiate CSAs with more than one CoC-accredited center. In a sensitivity analysis, we used multivariable linear regression to evaluate the density (or number) of CoC-accredited centers in each CSA. Additionally, an ordered logistic regression was used to evaluate the association of the highest accreditation level (none, low, middle, and highest) present within a CSA and LI.

In all regression models, SVI themes, EJI modules, health care infrastructure and access, and rurality were normalized, and CSA area size was log transformed. LI had a nonparametric association with the outcome, so the inflection points of its LOWESS smooth plot were used to divide LI into three categories: low (0 < LI < 0.3), moderate (0.3 ≤ LI < 0.8), and high (0.8 ≤ LI ≤ 1). All statistical analyses were performed in Stata, and significance was set to a .05 level with 95% confidence intervals (CIs) reported.²²

RESULTS

Overall Characteristics of CSAs

Overall, 1,002 CoC-accredited centers were mapped (Fig 1). They consisted of 67 NCI-designated clinical and comprehensive cancer centers, seven NCI-designated networks, 143 academic comprehensive cancer centers, 448 comprehensive community cancer programs, 230 community cancer programs, 99 integrated network cancer programs, and eight hospital associate cancer programs. A total of 668 CSAs were delineated with a median LI of 0.54 and a median population of 366,593 (Table 1). There were 157, 224, and 287 CSAs that contained zero, one, and two or more CoC-accredited centers, respectively. Table 1 provides additional descriptive statistics for characteristics of CSAs. The overall EJI is mapped by CSA in Appendix Figure A2, demonstrating greater environmental burden in the Southeastern region. Similarly, the overall SVI is mapped by CSA in Appendix Figure A3, demonstrating greater social vulnerability in the Southern and Western Coastal regions.

TABLE 1.

Unadjusted Analysis of CSA Characteristics by the Presence of CoC-Accredited Centers

CSA Characteristic	All CSAs (N = 668)	CSA Without CoC (n = 157, 24%)	CSA With CoC (n = 511, 76%)	P
EJI,a median (IQR)
Social vulnerability module	0.48 (0.39–0.60)	0.54 (0.44–0.66)	0.48 (0.37–0.58)	<.001 c
Environmental burden module	0.46 (0.34–0.62)	0.42 (0.31–0.56)	0.48 (0.35–0.63)	.002 d
Health vulnerability module	0.30 (0.17–0.43)	0.33 (0.19–0.50)	0.29 (0.15–0.42)	.001 d
SVI, median (IQR)
Socioeconomic status theme	0.49 (0.38–0.59)	0.52 (0.44–0.63)	0.49 (0.37–0.58)	<.001 c
Household characteristics theme	0.52 (0.39–0.62)	0.58 (0.44–0.66)	0.51 (0.38–0.61)	<.001 d
Racial and ethnic minority status theme	0.47 (0.32–0.64)	0.47 (0.34–0.67)	0.47 (0.32–0.63)	.313d
Housing type and transportation theme	0.52 (0.44–0.60)	0.56 (0.49–0.62)	0.51 (0.43–0.59)	<.001 c
Health care infrastructure and access theme, median (IQR)	0.35 (0.26–0.44)	0.36 (0.28–0.47)	0.35 (0.26–0.44)	.055d
Travel time (minutes), median (IQR)	85.93 (71.22–113.62)	95.29 (77.24–134.78)	84.40 (69.60–108.62)	<.001 d
Localization index, median (IQR)	0.54 (0.37–0.69)	0.44 (0.23–0.63)	0.56 (0.40–0.72)	<.001 d
Area size (m2), median (IQR)	4.77 × 109 (8.67 × 108–1.16 × 1010)	5.98 × 109 (1.13 × 109–1.48 × 1010)	4.46 × 109 (8.51 × 108–1.08 × 1010)	.241d
Rurality,b median (IQR)	1.59 (1.02–3.11)	2.28 (1.06–4.04)	1.45 (1.02–2.90)	<.003 c
Population, median (IQR)	366,593 (257,545–560,704)	257,416 (177,295–387,508)	403,457 (288,057–636,442)	<.001 c
Population density (persons/km2), median (IQR)	96.62 (29.80–453.05)	40.46 (18.68–286.82)	109.91 (37.68–471.62)	<.001 c

NOTE. Bold text indicates significance at the .05 level.

Abbreviations: CoC, Commission on Cancer; CSA, cancer service area; EJI, environmental justice index; RUCA, Rural-Urban Commuting Area; SVI, social vulnerability index.

^aEJI was calculated for 664 CSAs within the 48 contiguous states due to no available EJI data for Alaska and Hawaii.

^bRurality calculated using RUCA Codes.

^cLevene’s unpaired t test.

^dMann-Whitney U test.

Unadjusted Analysis of CSAs With and Without CoC-Accredited Centers

Among 668 CSAs, 511 (76%) CSAs contained at least one CoC-accredited center, and 157 (24%) CSAs did not contain any CoC-accredited center (Table 1). Compared with CSAs without CoC-accredited centers, CSAs with CoC-accredited centers had significantly greater environmental burden (median = 0.42 v 0.48, P = .002) and significantly lower social vulnerability (median = 0.54 v 0.48, P < .001) and health vulnerability (median = 0.33 v 0.29, P = .001). These CSAs also had significantly lower socioeconomic status vulnerability (median = 0.52 v 0.49, P < .001), household characteristics vulnerability (median = 0.58 v 0.51, P < .001), and housing type and transportation vulnerability (median = 0.56 v 0.51, P < .001). However, no significant difference was found in racial and ethnic minority status vulnerability or health care infrastructure and access between two groups of CSAs (P = .313 and P = .055, respectively).

In addition, CSAs with CoC-accredited centers had significantly higher LI (median = 0.56 v 0.44, P < .001; Table 1 and Fig 2) and shorter average travel time to cancer care (median = 84.40 v 95.29 min, P < .001; Table 1). These CSAs were more metropolitan (median RUCA = 1.45 v 2.28, P < .003) and populous (median = 403,457 v 257,416, P < .001) with relatively smaller area sizes (median = 4.46 × 10⁹ v 5.98 × 10⁹ m², P = .241) and thus had higher population density (median = 109.91 v 40.46 persons/km², P ≤ .001).

FIG 2.

LI of 668 CSAs and accreditation levels of 1,002 CoC centers in the United States, indicating CSAs with CoC-accredited centers have higher LI. Note: The CoC-accredited centers were grouped into four categories: no CoC-accredited centers (CSAs on the map without a CoC-accredited center, not shown in the legend), low (CSAs with only a community cancer program, hospital associate cancer program, or integrated network cancer program), middle (CSAs with at least one comprehensive community cancer program or academic comprehensive cancer program), and highest (CSAs with an NCI-designated comprehensive cancer center and/or NCI-designated network). LI was divided into three categories: low (0 < LI < 0.3), moderate (0.3 ≤ LI < 0.8), and high (0.8 ≤ LI ≤ 1). CoC, Commission on Cancer; CSA, cancer service area; LI, localization index; NCI, National Cancer Institute.

Multivariable Analysis of Factors Associated With the Presence of a CoC-Accredited Center

The social vulnerability module and household characteristics theme were found to have high multicollinearity with other variables. Due to the duplication of socioeconomic status in the SVI, the social vulnerability module was removed. However, the household characteristics theme was retained in the model given its clinical importance, and interaction terms were created between it and highly correlated variables. After adjusting for other covariates, lower health vulnerability (odds ratio [OR], 0.65 [95% CI, 0.427 to 0.993]), racial and ethnic minority status vulnerability (OR, 0.61 [95% CI, 0.424 to 0.886]), and area size (OR, 0.74 [95% CI, 0.556 to 0.978]) were associated with the presence of a CoC-accredited center (Table 2). There was no statistical significance for other components of the EJI, SVI, or interaction terms. Relative to CSAs with the lowest LI, CSAs with moderate (OR, 3.81 [95% CI, 2.081 to 6.981]) and high LIs (OR, 9.00 [95% CI, 2.408 to 33.640]) had a great association with the presence of CoC-accredited centers.

TABLE 2.

Logistic Regression Model Demonstrating Factors Associated With the Presence of a CoC-Accredited Center Within a CSA (reference is absence of a CoC-accredited center within a CSA; n = 664), Adjusting for EJI Modules, SVI Themes, Health Care Infrastructure and Access, Average Travel Time, LI, Area Size, Rurality, Population, Population Density, and Household Characteristics Theme With Interaction Terms

CSA Characteristic	OR (95% CI)	P
EJIa,b
Environmental burden module	1.21 (0.897 to 1.621)	.215
Health vulnerability module	0.65 (0.427 to 0.993)	.046
SVI
Socioeconomic status theme	1.09 (0.665 to 1.798)	.724
Household characteristics theme	2.16 (0.051 to 92.467)	.687
Racial and ethnic minority status theme	0.61 (0.424 to 0.886)	.009
Housing type and transportation theme	0.94 (0.667 to 1.319)	.712
Health care infrastructure and access theme	0.87 (0.696 to 1.079)	.200
Travel time (minutes)	1.00 (0.995 to 1.005)	.926
LI
Low	Reference	–
Moderate	3.81 (2.081 to 6.981)	<.001
High	9.00 (2.408 to 33.640)	.001
Area size (m2)	0.74 (0.556 to 0.978)	.034
Rurality	0.99 (0.628 to 1.575)	.982
Population	1.00 (1.000 to 1.000)	<.001
Population density (persons/km2)	1.00 (1.000 to 1.000)	.164
Household characteristics theme and area size interaction	0.97 (0.814 to 1.154)	.727
Household characteristics theme and rurality interaction	1.07 (0.740 to 1.552)	.714

NOTE. Bold text indicates significance at the .05 level.

Abbreviations: CoC, Commission on Cancer; CSA, cancer service area; EJI, environmental justice index; LI, localization index; OR, odds ratio; SVI, social vulnerability index.

^aEJI was calculated for 664 CSAs within the 48 contiguous states due to no available EJI data for Alaska and Hawaii.

^bThe social vulnerability module of the EJI was excluded from the model due to multicollinearity.

The sensitivity analysis evaluating the density of CoC-accredited centers within a CSA yielded similar results that racial and ethnic minority status (P = .001) and LI (P = .009 and P = .020 for moderate and high LI) remained significant. However, health vulnerability module and area size were no longer significantly associated with the density of a CoC-accredited center (P = .860 and P = .173, respectively; Table 3).

TABLE 3.

Multivariable Linear Regression Demonstrating Factors Associated With the Density of CoC-Accredited Centers in a CSA (reference is a density of 0, ie, no CoC-accredited centers; n = 664), Adjusting for EJI Modules, SVI Themes, Health Care Infrastructure and Access, Travel Time, LI, Area Size, Rurality, Population, Population Density, and Household Characteristics Theme With Interaction Terms

CSA Characteristic	Coefficient (95% CI)	Beta Coefficient	P
EJIa,b
Environmental burden module	0.07 (−0.055 to 0.202)	0.053	.264
Health vulnerability module	0.02 (−0.175 to 0.210)	0.012	.860
SVI
Socioeconomic status theme	−0.17 (−0.388 to 0.057)	−0.120	.144
Household characteristics theme	1.10 (−0.579 to 2.778)	0.795	.199
Racial and ethnic minority status theme	−0.27 (−0.439 to −0.109)	−0.198	.001
Housing type and transportation theme	0.04 (−0.110 to 0.190)	0.290	.599
Health care infrastructure and access theme	−0.08 (−0.173 to 0.017)	−0.056	.108
Average travel time (minutes)	0.00 (−0.003 to 0.001)	−0.031	.446
LI
Low	Reference	–	–
Moderate	0.37 (0.092 to 0.646)	0.118	.009
High	0.54 (0.084 to 0.989)	0.122	.020
Area size (m2)	−0.09 (−0.208 to 0.038)	−0.113	.173
Rurality	−0.04 (−0.254 to 0.172)	−0.029	.707
Population	0.00 (0.000 to 0.000)	0.533	<.001
Population density (persons/km2)	0.00 (0.000 to 0.000)	−0.018	.726
Household characteristics theme and area size interaction	−0.05 (−0.127 to 0.028)	−0.762	.210
Household characteristics theme and rurality interaction	0.04 (−0.130 to 0.209)	0.027	.650

NOTE. Population and population density values are greater than zero, but displayed as 0.000 based on rounding. Bold text indicates significance at the .05 level.

Abbreviations: CoC, Commission on Cancer; CSA, cancer service area; EJI, environmental justice index; LI, localization index; SVI, social vulnerability index.

^aEJI was calculated for 664 CSAs within the 48 contiguous states due to no available EJI data for Alaska and Hawaii.

^bThe social vulnerability module of the EJI was excluded from the model due to multicollinearity with socioeconomic status theme, household characteristics theme, and housing type and transportation theme.

Multivariable Analysis of Association of Accreditation Level of CoC Centers and LI

The ordered logistic regression evaluating the association between LI groups and CoC accreditation level demonstrated that relative to CSAs without CoC-accredited centers, those with higher accreditation levels of CoC were associated with higher LIs (middle accreditation group OR, 3.60 [95% CI, 1.990 to 6.505] and highest accreditation group OR, 7.62 [95% CI, 3.049 to 19.059] for low v moderate or high LI, or low or moderate v high LI) when adjusting for all other explanatory variables and interaction terms (Table 4).

TABLE 4.

Ordered Logistic Regression Demonstrating Association Between CoC-Accreditation Level and LI Groups (low, moderate, high) Within a CSA (N = 668), Adjusted for EJI Modules, SVI Themes, Health Care Infrastructure and Access, Travel Time, Area Size, Rurality, Population, Population Density, and Household Characteristics Theme With Interaction Terms

CSA Characteristic	OR (95% CI)	P
Center type with highest level of accreditation in CSAa
None (no CoC-accredited center)	Reference
Low (community cancer program, hospital associate cancer program, integrated network cancer program)	1.68 (0.765 to 3.705)	.196
Moderate (comprehensive community cancer program, academic comprehensive cancer program)	3.60 (1.990 to 6.505)	<.001
High (NCI-designated comprehensive cancer center or NCI-designated network)	7.62 (3.049 to 19.059)	<.001
EJIb,c
Environmental burden module	1.08 (0.792 to 1.485)	.614
Health vulnerability module	1.63 (0.972 to 2.730)	.064
SVI
Socioeconomic status theme	0.77 (0.438 to 0.137)	.378
Household characteristics theme	0.00 (0.000 to 0.044)	.001
Racial and ethnic minority status theme	0.46 (0.300 to 0.721)	.001
Housing type and transportation theme	1.83 (1.245 to 2.678)	.002
Health care infrastructure and access theme	0.84 (0.666 to 1.068)	.157
Average travel time (minutes)	0.99 (0.982 to 0.993)	<.001
Area size (m2)	2.89 (2.085 to 3.974)	<.001
Rurality	1.00 (0.559 to 1.776)	.991
Population	1.00 (1.00 to 1.00)	<.001
Population density (persons/km2)	1.00 (1.00 to 1.00)	.095
Household characteristics theme and shape area interaction	1.38 (1.133 to 1.688)	.001
Household characteristics theme and rurality interaction	0.46 (0.284 to 0.749)	.002
Household characteristics theme and health vulnerability module interaction	0.69 (0.477 to 1.005)	.053
Household characteristics theme and socioeconomic status theme interaction	1.47 (1.092 to 1.971)	.011

NOTE. OR indicates incremental likelihood of each change in LI group. Bold text indicates significance at the .05 level. Abbreviations: CoC, Commission on Cancer; CSA, cancer service area; EJI, environmental justice index; LI, localization index; NCI, National Cancer Institute; OR, odds ratio; SVI, social vulnerability index.

^aCoC-accredited centers were grouped into four accreditation levels (none ie, no CoC-accredited center, low, middle, and highest) based on CoC definitions that incorporate analytic case volume, teaching, and research. The low group consisted of community cancer programs, hospital associate cancer programs, and/or integrated network cancer programs; the middle group consisted of at least one comprehensive community cancer program or academic comprehensive cancer program; and the highest group consisted of NCI-designated comprehensive cancer centers and/or NCI-designated networks.

^bEJI was calculated for 664 CSAs within the 48 contiguous states due to no available EJI data for Alaska and Hawaii.

^cThe social vulnerability module of the EJI was excluded from the model due to multicollinearity with socioeconomic status theme, household characteristics theme, and housing type and transportation theme.

DISCUSSION

In this nationwide analysis of cancer care utilization–based catchment areas, we evaluated the association of social, environmental, and geographic factors with the presence of CoC-accredited centers within catchment areas. This study is novel in that it used actual cancer care utilization data and an innovative GIS method to delineate CSAs^9–11 as the utilization-based catchment areas for CoC-accredited centers, and the CSA schema was high-quality as evidenced by high median LI values. Additionally, it evaluated regional variation in several pertinent exposures including environmental and social vulnerability. This work enables future investigation into the mechanisms associated with regional variation in cancer incidence, prevalence, and outcomes, which are known to be affected by environmental and social risk factors.

Lower levels of health and racial and ethnic vulnerability were associated with the presence of a CoC-accredited center within a CSA. This suggests that CoC-accredited centers are more often present in racially and ethnically homogeneous areas where people tend to be healthier with fewer comorbidities. These findings have important implications for reducing variations in cancer care as disparate access to cancer care is observed for people with greater health vulnerability and racial and ethnic vulnerability. Expanding the impact of CoC-accredited centers may address this issue. Strategies may include establishing affiliate relationships with other centers or hospitals to improve cancer care delivery in underserved areas, increasing existing centers’ accreditation level through expanded capacity and resources, or improving access to areas where these CoC-accredited centers already exist.²³ Additionally, CoC-accredited centers can establish benchmarks for quality care across demographics, minimize barriers through social needs screenings and insurance coverage policies, and conduct greater community engagement and outreach such as recruiting patient groups with limited participation in clinical trials within their catchment areas.²⁴

A key finding of this study is that CoC-accredited centers’ presence is associated with greater LI, suggesting that the presence of a CoC-accredited center may influence patients’ decisions to seek care in their catchment area. This catchment area may provide better access to quality care for their patients and make them less likely to seek care outside. This could also be due to physician referral patterns, where physicians refer their patients to local providers through their networks to facilitate easier access to care. Furthermore, the highest accreditation level of a CoC-accredited center within a CSA was associated with greater LI, suggesting that not just a CoC’s presence but its accreditation level also influences where patients receive care.

The evaluation of the density of CoC-accredited centers and CSAs’ characteristics demonstrated that health vulnerability and area size were no longer significant. Although health vulnerabilities influence whether a CoC-accredited center is established, they may be less relevant in determining their numbers in a given area. Given that the accreditation of a cancer center should meet rigorous standards including community engagement requirements, one accredited center might be adequate to provide high-quality care in a relatively smaller catchment area.

The study has some limitations. Data used to define CSAs were restricted to Medicare enrollees and may not represent the CSAs of other populations. The ZIP codes of patient residence may not reflect where they lived when they received cancer care. For example, patients may rent or stay with family or friends near their care center. CoC-accredited centers were collected in 2021 which may change over time, although the degree of flux in accreditation status is expected to be minor and mostly affect community and local affiliate centers that may have more difficulty maintaining their accreditation. Additionally, patients who received care in a CSA may not attend a CoC-accredited center; rather, some may seek care at a nearby nonaccredited center in the same area. Due to data availability, some data including EJI and SVI were obtained in different years, which may affect the results. Finally, the study is ecologic in nature, performed at the CSA level but using data from patients with cancer treated within the CSA. Although this limits our ability to make direct causal inferences, it enables us to draw broader conclusions about the characteristics of catchment areas that are associated with the presence of CoC-accredited centers.

In conclusion, understanding the social, environmental, and geographic characteristics of cancer center catchment areas, including the role of their accreditation levels and distributions, is critical to help inform cancer centers’ strategic planning about where to offer quality cancer care as they expand their regional networks. These findings may be used by cancer centers to better address the needs of their catchment areas through outreach activities or strategic affiliations with regional centers. Furthermore, this work will enable future evaluation of the mechanisms associated with regional cancer incidence, prevalence, and outcomes. This may provide a foundation for future work to understand the mitigable care delivery factors that contribute to variation in regional cancer incidence, prevalence, and mortality such as expanded accreditation options for cancer center affiliates and strategies to disseminate evidence-based practices to regional sites.

CONTEXT.

Key Objective

To examine social, environmental, and geographic characteristics of cancer care utilization–based catchment areas, termed Cancer Service Areas (CSAs), with and without Commission on Cancer (CoC)–accredited centers.

Knowledge Generated

Six hundred sixty-eight CSAs were defined, of which 511 CSAs had at least one CoC-accredited center. CSAs with CoC-accredited centers had lower health vulnerability, lower racial and ethnic minority status vulnerability, and higher localization indices, meaning patients were more likely to remain in their region for care.

Relevance (M. Behera)

This study enhances our knowledge of how CoC-accredited cancer centers serve their surrounding populations, identifies geographic and social disparities in access to high-quality cancer care. Insights from this study can help inform targeted interventions and resource allocation strategies to improve equity and outcomes in cancer treatment and care delivery.*

*Relevance section written by JCO CCI Associate Editor Madhusmita Behera, PhD.

APPENDIX

FIG A1.

Workflow diagram of integrating various data sets, GIS methods, and statistical analyses. CDC, Centers for Disease Control and Prevention; CMS, Centers for Medicare and Medicaid Services; CoC, Commission on Cancer; CSA, cancer service area; EJI, environmental justice index; GIS, geographic information systems; LI, localization index; NCI, National Cancer Institute; OMH, Office of Minority Health; RUCA, Rural-Urban Commuting Area; SVI, social vulnerability index; USDA, US Department of Agriculture.

FIG A2.

EJI of CSAs overlaid with 1,002 CoC-accredited centers in the United States. EJI was classified by quartiles, with higher quartiles indicating greater environmental burden. CoC, Commission on Cancer; CSA, cancer service area; EJI, environmental justice index.

FIG A3.

SVI of CSAs overlaid with 1,002 CoC-accredited centers in the United States. SVI was classified by quartiles, with higher quartiles indicating greater social vulnerability. CoC, Commission on Cancer; CSA, cancer service area; SVI, social vulnerability index.

TABLE A1.

Data Sources, Geographies, Publication Year, and Definitions of Variables Used in Analyses

Variable	Data Source	Geographic level	Publication Year	Definition
CoC-accredited centers	American College of Surgeons	Point	2021	Names and street addresses of CoC-accredited centers which were geocoded to points representing each center’s location
EJI	CDC	Census tract	2022
Social vulnerability module				Comprises population-level indicators representing racial/ethnic minority status, socioeconomic status (eg, poverty, unemployment), household characteristics (eg, senior, disability, English proficiency), and housing type (eg, group quarter, mobile home)
Environmental burden module				Comprises population-level indicators representing air and water pollution, potentially hazardous and toxic sites, built environment, and transportation infrastructure
Health vulnerability module				Comprises population-level indicators representing preexisting chronic disease burden (eg, asthma, cancer, heart disease, diabetes, and poor mental health)
SVI	CDC	Census tract	2014
Socioeconomic status theme				Comprises population-level indicators representing poverty, unemployment, housing cost burden, education, and health insurance
Household characteristics theme				Comprises population-level indicators representing age, disability, single-parent households, and English language proficiency
Racial and ethnic minority status theme				Comprises population-level indicators representing difference races and ethnicities
Housing type and transportation theme				Comprises population-level indicators representing housing type (multiunit and mobile homes), density (crowding, group quarters), and vehicle access
Health care infrastructure and access theme	OMH and CDC	County	2020	Comprises population-level indicators representing infrastructure and access for hospitals, urgent care clinics, pharmacies, and primary care physicians
Travel time (minutes)	Calculated from the ZIP-to-ZIP travel time matrix15	CSA	2018	Weighted average time for patients to travel from the ZIP code of their residence to the ZIP code of their cancer care providers within a CSA weighted by associated patients’ flows
LI	Calculated from ZIP-to-ZIP flows of patients with cancer from Medicare claims data	CSA	2014	Proportion of cancer care received by patients within a CSA
Area size (m2)	Calculated from ZIP code areas	CSA	2014	CSA shape area, in m2, of a CSA
Population	Calculated from the population at the census blocks	CSA	2010	Population of a CSA
Population density (persons/km2)	Calculated by dividing population by area size	CSA	2010	Ratio of total population to total area of a CSA
Rurality	RUCA codes from USDA, ERS	Census tract	2010	Measure of population density and urbanization; delineates metropolitan, micropolitan, small town, and rural commuting areas

Abbreviations: CDC, Centers for Disease Control and Prevention; CoC, Commission on Cancer; CSA, cancer service area; EJI, environmental justice index; ERS, economic research service; LI, localization index; OMH, Office of Minority and Health; RUCA, Rural-Urban Commuting Area; SVI, social vulnerability index; USDA, US Department of Agriculture.