Association of area sociodemographic characteristics and capacity for treatment with disparities in colorectal cancer care and mortality

Authors

  • Jennifer S. Haas MD, MSPH,

    Corresponding author
    1. Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts
    • Division of General Medicine and Primary Care, Brigham and Women's Hospital, 1620 Tremont Street, Boston, MA 02120-1613
    Search for more papers by this author
    • Fax: (617) 732-7072

  • Phyllis Brawarsky MPH,

    1. Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Aarthi Iyer MPH,

    1. Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Garrett M. Fitzmaurice ScD,

    1. Laboratory for Psychiatric Biostatistics, McLean Hospital and Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Bridget A. Neville MPH,

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, Massachusetts
    Search for more papers by this author
  • Craig Earle MD MSc FRCP(C)

    1. Cancer Care Ontario and the Ontario Institute for Cancer Research, Toronto, Ontario, Canada
    Search for more papers by this author

Abstract

BACKGROUND:

Disparities in treatment and mortality for colorectal cancer (CRC) may reflect differences in access to specialized care or other characteristics of the area where an individual lives.

METHODS:

Surveillance, Epidemiology and End Results Program–Medicare data for seniors diagnosed with CRC were linked to area measures of the sociodemographic characteristics and the capacity of surgeons, medical oncologists, and radiation oncologists. Outcomes included receipt of stage-appropriate CRC care and mortality.

RESULTS:

After adjustment, blacks and Hispanics were less likely than whites to undergo surgery (odds ratio [OR] 0.57, 95% confidence interval (CI) 0.52-0.63 and OR 0.82, 95% CI 0.70-0.95, respectively). Individuals who lived in areas with the highest tertile of surgeon capacity were more likely to undergo resection than those in the lowest, and use of surgery declined as the percentage of blacks in the area increased. Adjustment for the area measures resulted in a modest decline in disparities in care relative to whites (5.3% for black). Blacks also experienced greater all-cause and cancer-specific mortality than whites. Further adjustment for area sociodemographics and surgeon capacity reduced the disparity in mortality between blacks and whites. Although there was a similar black/white disparity in the use of adjuvant chemotherapy, the disparity remained after adjustment for area characteristics, although use of chemotherapy was greater in areas with the greatest capacity of medical oncologists.

CONCLUSIONS:

Sociodemographic characteristics and measures of the availability of specialized cancer providers in the area in which an individual resides modestly mediated disparities in the receipt of CRC care and mortality, suggesting that other factors may also be important. Cancer 2011;. © 2011 American Cancer Society.

Colorectal cancer (CRC) is the fourth most commonly diagnosed cancer and the second leading cause of cancer death in the United States, with approximately 150,000 diagnoses each year.1 But because cancer care is highly specialized, there are challenges to ensuring that each individual has access to evidence-based treatment and follow-up. This challenge is reflected by well-documented racial/ethnic and socioeconomic disparities in the treatment and mortality for CRC. For example, blacks receive less aggressive treatment and have greater mortality than whites.2-10 Although the causes of these disparities are complex, the availability of physicians with appropriate expertise within reasonable geographic proximity may be a factor. This is supported by studies from England, a country with national health insurance, that found that residents of more impoverished areas are less likely than residents of more affluent areas to undergo cancer surgery.11, 12 An ecologic study in the United States found that counties with a greater percentage of blacks have fewer radiation oncologists and gastroenterologists.13

Because the regional availability of specialists who treat CRC may be associated with the likelihood that an individual with CRC receives guideline-recommended, stage-specific CRC care, we hypothesized that individuals with CRC who live in an area with greater capacity for CRC treatment are more likely to receive recommended care and have lower mortality than individuals who live in an area with less capacity for CRC treatment. Furthermore, because local sociodemographic characteristics of the population may influence provider availability, we also examined whether these area characteristics mediate some of the racial disparities in CRC treatment and mortality.

MATERIALS AND METHODS

Data

This analysis used data from the Surveillance, Epidemiology and End Results (SEER) Program–Medicare file, the American Medical Association (AMA) Physician File, and the Area Resource File (ARF). The SEER Program collects information on all incident cancer cases for persons with cancer residing in SEER Program areas, including California, Connecticut, Hawaii, Iowa, Kentucky, Louisiana, New Jersey, New Mexico, Utah, rural Georgia, and the metropolitan areas of Detroit, Atlanta, and Seattle.14 SEER data, including cancer stage, primary tumor site, and patient demographics, are linked to Medicare claims data by the National Cancer Institute. Data were used for individuals diagnosed with cancer from 1992 to 2005 with Medicare follow-up through 2007. A restricted access version of these data was obtained so that we could use data on the characteristics of each individual's census tract of residence that is available only in the restricted version. The AMA Physician File was used to assign physician specialty to physicians with claims included in SEER-Medicare to create our variables for physician capacity defined below. The ARF includes information about population demographics, including the population by age in each county. This study was reviewed and approved by the institutional review board of Partners HealthCare.

Study Sample

Adults aged 66-79 years, whose race/ethnicity was reported as white, black, or Hispanic and who were diagnosed with American Joint Committee on Cancer stage 1, 2, or 3 CRC as a single primary between 1992 and 2005 were included in this analysis (N = 61,380). We excluded individuals over the age of 79 as there is less clinical certainty about the appropriate management of these individuals.15 Individuals who did not have continuous Medicare part A and B coverage or who were members of a health maintenance organization for any time within 1 year prediagnosis to 6 months after diagnosis were also excluded because these individuals may not have complete claims in SEER-Medicare (n = 19,284).16 We also excluded individuals who died within 6 months of diagnosis, because they might not have had the opportunity to initiate the recommended treatment (n = 3430). The final sample contained 38,666 individuals.

Variables

We examined 3 outcome measures: receipt of adequate CRC care, all-cause mortality, and CRC-specific mortality. Our measures for the receipt of adequate care for CRC were based on National Institutes of Health consensus guidelines for the treatment of CRC, and they reflect treatment recommendations during this time period.17 We examined 3 treatment outcomes, including: (1) surgical resection within 6 months of diagnosis for patients with stage I, II, or III colon or rectal cancer; (2) use of adjuvant chemotherapy within 3 months of surgery by patients with stage 3 colon cancer or from 3 months before surgery through 9 months afterward for stage 2 and 3 rectal cancer; and (3) use of radiation therapy by patients with stage 2 or 3 rectal cancer from 3 months before surgery to 9 months afterward.17 These outcomes were defined using Medicare claims.

Mortality was defined by the interval from the date of diagnosis to the Medicare date of death, and if that was missing, the SEER date of death. The interval from date of diagnosis to the end of the study period (December 31, 2007) was included in the survival analysis for individuals who were alive at the end of the study period. CRC mortality was defined by SEER cause of death.

Our principal independent variables were measures of the regional capacity of relevant specialists who treat CRC, defined as the number of unique general or oncology surgeons, medical oncologists, and radiation oncologists at the level of health service area (HSA). An HSA is defined as 1 or more counties, relatively self-contained with respect to the provision of hospital care.18 We used Medicare claims to determine all of the physicians seen by each patient. Each physician was then categorized by specialty training using data from the AMA Physician File linked to the claims with each physician's unique physician identification number (UPIN). Physicians were then assigned to an HSA based on the ZIP code of the practice. Physicians whose practices were located in more than 1 HSA were assigned to the HSA where they saw the greatest number of subjects. We included only HSAs that contained counties within the SEER program areas, with 160 HSAs represented. We then summed the number of unique UPINs for each of the above-named specialties for each HSA and created measures per the number of 100,000 residents aged ≥50 years from the ARF. These measures were categorized in tertiles: for surgeons, the categories were ≤229.1, 229.2-326.2, 329.3-1493.0; for medical oncologists, ≤18.4, 18.5-36.7, 36.8-278.0; and for radiation therapists, ≤11.3, 11.4-21.3, and 21.4-100).

Other independent variables included individual characteristics, including sex, age at diagnosis (continuous), race/ethnicity (white, black, Hispanic), marital status (married, not married), Charlson comorbidity index (categorized as 0, 1, 2, ≥3),16 whether an individual was of “low income” (based on eligibility for state assistance with Medicare premiums and copayments), cancer type (colon, rectal), stage, and year of diagnosis. Stage was categorized in accordance with the American Joint Committee on Cancer staging system as used by the SEER Program, based on the extent of the tumor, the number of lymph nodes involved, and the presence of metastasis.19 Area characteristics included whether the county of residence was rural or urban based on the rural/urban continuum codes included in the SEER-Medicare data. Additional measures of area characteristics, characterized in tertiles, included the percentage of the census tract population that was black (≤0.9%, 0.91-4.2%, 4.3-100%), the percentage Hispanic (≤2.2%, 2.3-7.4%, 7.5-99.8%), and median household income (<$38,044, $38,045-$53,819, $53,820-$200,008).

Statistical Analysis

To examine the effect of regional capacity for CRC treatment on treatment for CRC, we used multilevel logistic regression models, clustered by HSA, using SAS version 9.2 (SAS Institute, Cary, NC). We examined the 3 treatment outcomes defined above. For each of these outcomes, initial models adjusted for individual sociodemographic and clinical characteristics including, sex, age at diagnosis, race/ethnicity, marital status, comorbidity, eligibility for state buy-in coverage (as an indicator of personal disadvantage), cancer stage, and year of diagnosis. Final models were adjusted for the area physician capacity measures, urban/rural designation, and sociodemographic variables; if inclusion of these measures decreased the racial/ethnic effect, physician capacity and area sociodemographics were considered to be mediators of disparities in CRC treatment. We used Cox proportional hazard models, clustered by HSA, with SUDAAN version 9 (RTI International, Research Triangle Park, NC) to model all-cause and CRC-specific mortality with the same variables described above.

RESULTS

Sample Characteristics

Compared with whites with CRC, both blacks and Hispanics were less likely to be married, more likely to be disadvantaged, and more likely to have a higher burden of comorbid conditions (Table 1). Blacks were more likely, and Hispanics less likely, than whites to be diagnosed with colon (vs rectal) cancer, and both were diagnosed with later-stage disease. Blacks were most likely to live in an urban area (95.1%) and whites least likely (87.8%). Hispanics were more likely than both blacks and whites to live in areas with fewer surgeons and radiation oncologists. Blacks were more likely to live in an area of higher capacity of surgeons and radiation oncologists. Both blacks and Hispanics lived in census tracts with a lower household income than whites. Blacks were less likely than whites and Hispanics to undergo surgical resection, adjuvant chemotherapy, and radiation therapy. Median survival was shorter for blacks compared with both whites and Hispanics.

Table 1. Characteristics of the Sample by Race/Ethnicity
 WhiteBlackHispanicPa
  • HSA indicates health service area; CRC, colorectal cancer.

  • a

    Chi-square test, accounting for clustering within HSAs.

  • b

    Linear regression model, clustered by HSA.

  • c

    Missing: married, 1343; percentage blacks in census tract, 363, percentage Hispanics in census tract, 363; median income in census tract, 364.

  • d

    Includes individuals with stages 1, 2, or 3 colon or rectal cancer.

  • e

    Includes individuals with stage 3 colon cancer or stage 2 or 3 rectal cancer.

  • f

    Includes individuals with stage 2 or 3 rectal cancer.

  • g

    Proportional hazards model, clustered by HSA.

  • h

    Number of patients who died: white, 13,894, black, 1421; Hispanic, 792.

Number33,59730512018 
Individual characteristics    
 Men49.1%39.8%50.9%<.0001
 Age, y, mean (range, 66-79)72.972.672.2<.0001b
 Marriedb62.9%40.4%56.0%<.0001
 Comorbid conditions    
  None69.2%59.7%63.7%<.0001
  1 condition21.3%25.6%24.0% 
  2 conditions5.7%7.5%6.8% 
  3 or more conditions3.8%7.2%5.5% 
 Ever eligible for state buy-in coverage12.8%38.8%49.2%<.0001
 Type of cancer   <.0001
  Colon72.7%78.4%69.2% 
  Rectal27.3%21.6%30.8% 
 Stage of cancer   <.0001
  Stage 134.0%32.9%29.5% 
  Stage 236.1%33.3%38.6% 
  Stage 329.9%33.8%31.9% 
Area characteristics    
 Physician capacity in HSA per 100,000 population    
  Surgeons   .0004
   Low29.8%31.1%33.5% 
   Middle31.8%20.1%51.8% 
   High38.3%48.7%14.6% 
  Medical oncologists   .1524
   Low25.2%15.8%30.8% 
   Middle52.3%67.4%58.6% 
   High22.4%16.8%10.7% 
  Radiation oncologists   .0008
   Low24.2%27.3%28.0% 
   Middle40.8%26.5%58.6% 
   High34.9%46.2%13.4% 
 Urban county87.8%95.1%92.9%<.0020
 Percentage of blacks in census tractc   <.0001
  Low36.11.529.8 
  Middle35.94.538.3 
  High27.994.031.8 
 Percentage of Hispanics in census tractc   <.0001
  Low33.8%47.7%3.3% 
  Middle36.1%18.6%10.1% 
  High30.1%33.7%86.7% 
 Median income in census tractc   <.0001
  Low29.4%67.4%46.7% 
  Middle34.7%21.6%30.4% 
  High35.9%10.9%22.9% 
Cancer care and survival    
 Received adequate CRC care    
  Surgical resectiond89.5%85.4%88.0%<.0001
  Chemotherapy and surgical resectione66.6%54.2%63.7%<.0001
  Radiation therapy and surgical resectionf37.6%27.8%38.1%.0026
 Median survival time (days)349327293226<.0001g
 Death from CRC (among patients who died)h42.1%47.3%46.1%.0001

Outcome 1: Surgical Resection and Mortality for Stage 1-3 Colorectal Cancer

After adjustment for personal and clinical characteristics, blacks were less likely than whites to receive surgery (odds ratio [OR], 0.57; 95% confidence interval [CI], 0.52-0.63), as were Hispanics (OR, 0.82; 95% CI, 0.70-0.95) (Table 2). Disadvantaged individuals were also less likely to receive surgery, as were individuals with rectal as opposed to colon cancer. Use of resection increased with stage. Individuals who lived in an area with the highest tertile of surgeon capacity were more likely to undergo resection than individuals in the lowest tertile, and use of surgery declined as the percentage of blacks in the area increased. Adjustment for the area measures resulted in a modest decline in disparities in care relative to whites (5.3% for black and 7.3% for Hispanics).

Table 2. Surgical Resection and Mortality: All Colon and Rectal Cancer Patients, Stages 1, 2, and 3 (N = 38,666)
 Surgical ResectionaAll-Cause MortalitybColorectal Cancer Mortalityb
Model with Individual Covariates OnlydModel with Individual and Area CovariatesModel with Individual Covariates OnlycModel with Individual and Area CovariatesModel with Individual Covariates Only1Model with Individual and Area Covariates
  • a

    Data are presented as odds ratio (95% confidence interval).

  • b

    Data are presented as hazard ratio (95% confidence interval).

  • c

    All models also adjusted for sex, age (continuous), marital status, comorbid conditions, and year of diagnosis and clustered by health service area.

  • d

    All models also adjusted for sex, age (continuous), marital status, comorbid conditions, and year of diagnosis and clustered by HSA.

Individual characteristics      
 Race/ethnicity      
  WhiteReferenceReferenceReferenceReferenceReferenceReference
  Black0.57 (0.52-0.63)0.60 (0.55-0.66)1.16 (1.09-1.23)1.08 (1.02-1.16)1.37 (1.26-1.49)1.25 (1.14-1.38)
  Hispanic0.82 (0.70-0.95)0.88 (0.77-1.01)0.92 (0.85-0.99)0.88 (0.81-0.95)1.04 (0.93-1.16)0.98 (0.88-1.10)
 Ever eligible for state buy-in coverage (vs never eligible)0.92 (0.82-1.03)0.95 (0.85-1.06)1.26 (1.21-1.31)1.23 (1.18-1.29)1.0 (0.93-1.07)0.96 (0.9-1.03)
 Type of cancer      
  Rectal (vs colon)0.34 (0.32-0.37)0.34 (0.32-0.37)1.25 (1.21-1.30)1.25 (1.21-1.30)1.53 (1.45-1.61)1.53 (1.45-1.61)
 Stage      
  1ReferenceReferenceReferenceReferenceReferenceReference
  25.13 (4.67-5.63)5.16 (4.70-5.66)1.45 (1.40-1.51)1.45 (1.40-1.51)3.09 (2.84-3.35)3.09 (2.85-3.36)
  37.38 (6.29-8.67)7.42 (6.33-8.70)2.40 (2.31-2.50)2.41 (2.32-2.51)7.74 (7.16-8.37)7.78 (7.19-8.42)
Area characteristics      
 Urban county (vs rural) 0.98 (0.85-1.13) 1.05 (0.99-1.12) 1.03 (0.95-1.13)
 % Black      
  Low Reference Reference Reference
  Middle 0.92 (0.84-1.01) 0.99 (0.95-1.03) 0.98 (0.92-1.04)
  High 0.87 0.77-0.98) 1.03 (0.98-1.07) 1.05 (0.97-1.11)
 % Hispanic      
  Low Reference Reference Reference
  Middle 0.95 (0.87-1.04) 1.00 (0.96-1.04) 0.98 (0.92-1.04)
  High 0.91 (0.83-1.01) 1.06 (1.01-1.11) 1.05 (0.98-1.13)
 Median household income      
  Low Reference Reference Reference
  Middle 1.05 (0.97-1.14) 0.93 (0.89-0.97) 0.90 (0.85-0.96)
  High 1.05 (0.97-1.14) 0.87 (0.84-0.92) 0.83 (0.78-0.89)
Surgeon capacity      
 Low Reference Reference Reference
 Middle 1.06 (0.94-1.20) 0.97 (0.92-1.01) 0.98 (0.92-1.05)
 High 1.19 (1.05-1.35) 1.00 (0.96-1.04) 0.99 (0.92-1.05)

For eligible patients, blacks also experienced greater all-cause mortality (hazard ratio [HR], 1.16; 95% CI, 1.09-1.23), but Hispanics had lower morality (HR, 0.92; 95% CI, 0.85-0.99) than whites (Table 2). Disadvantaged individuals and those with rectal rather than colon cancer also had higher mortality. Further adjustment for area sociodemographics and surgeon capacity reduced the disparity in mortality between blacks and whites by 6.9% (HR, 1.08; 95% CI, 1.02-1.16) and increased the survival advantage of Hispanics compared with whites (HR, 0.88; 95% CI, 0.81-0.95). Mortality was higher in areas with the highest percentage of Hispanics, lower for individuals who lived in higher income areas, and was not associated with the availability of surgeons in the HSA. Blacks also experienced greater CRC-specific mortality compared with whites (HR, 1.37; 95% CI, 1.26-1.49); further adjustment for area sociodemographics and surgeon capacity reduced this disparity by 8.7%. There was no difference in CRC-specific mortality between Hispanics and whites.

Outcome 2: Adjuvant Chemotherapy

Among eligible patients, blacks were less likely than whites to receive adjuvant chemotherapy (OR, 0.61; 95% CI, 0.50-0.73) (Table 3). Usage did not differ between Hispanics and whites. In addition, patients diagnosed at stage 3 rather than stage 2 were more likely, and disadvantaged patients less likely, to undergo chemotherapy. There was little change in the black/white disparity with further adjustment for area variables. Chemotherapy use was greater in urban areas, decreased as the percentage of Hispanics increased, and increased as area income and the supply of medical oncologists increased (OR, 1.21; 95% CI, 1.03-1.42 [third tertile compared with first]). Among eligible patients, all-cause and CRC-specific mortality was greater for blacks compared with whites (HR 1.18, 95% CI 1.09-1.28 and HR 1.30, 95% CI 1.18-1.44, respectively). Additional adjustment for area characteristics reduced the disparity in all-cause mortality by 7.6% (HR, 1.09; 95% CI, 1.00-1.20) and by 10.8% in CRC-specific mortality (HR, 1.16; 95% CI, 1.04-1.3).

Table 3. Chemotherapya and Mortality: Patients With Colon Cancer, Stage 3 and Rectal Cancer, Stages 2 and 3 (N = 4741)
 ChemotherapybAll-Cause MortalitycColorectal Cancer Mortalityc
Model with Individual Covariates OnlydModel with Individual and Area CovariatesModel with Individual Covariates OnlyModel with Individual and Area CovariatesModel with Individual Covariates OnlyModel with Individual and Area Covariates
  • a

    Chemotherapy among patients who had resection, within 4 months of diagnosis for colon cancer stage III and within 9 months of diagnosis for rectal cancer stage II and III.

  • b

    Data are presented as odds ratio (95% confidence interval).

  • c

    Data are presented as hazard ratio (95% confidence interval).

  • d

    All models also adjusted for sex, age (continuous), marital status, comorbid conditions, and year of diagnosis and clustered by HSA.

Individual characteristics      
 Race/ethnicity      
  WhiteReferenceReferenceReferenceReferenceReferenceReference
  Black0.61 (0.50-0.73)0.62 (0.51-0.76)1.18 (1.09-1.28)1.09 (1.00-1.20)1.30 (1.18-1.44)1.16 (1.04-1.3)
  Hispanic1.05 (0.84-1.32)1.13 (0.95-1.35)0.99 (0.90-1.09)0.95 (0.85-1.05)1.03 (0.91-1.17)1.0 (0.87-1.14)
 Ever eligible for state buy-in coverage (vs never eligible)0.62 (0.55-0.70)0.65 (0.58-0.73)1.10 (1.04-1.17)1.08 (1.01-1.15)0.92 (0.84-1.0).088 (0.81-1.26)
 Rectal cancer (vs colon)0.97 (0.87-1.08)0.98 (0.87-1.09)1.13 (1.07-1.20)1.13 (1.07-1.20)1.17 (1.09-1.26)1.18 (1.10-1.26)
 Stage 3 cancer (vs stage 2)3.33 (2.97-3.74)3.35 (2.98-3.71)1.39 (1.30-1.49)1.40 (1.31-1.50)1.75 (1.59-1.91)1.76 (1.60-1.93)
Area characteristics      
 Urban county (vs rural) 1.18 (1.02-1.36) 1.03 (0.95-1.12) 1.02 (0.91-1.13)
 % Black      
  Low Reference Reference Reference
  Middle 0.93 (0.84-1.04) 0.99 (0.93-1.04) 1.01 (0.93-1.09)
  High 0.93 (0.82-1.06) 1.04 (0.98-1.11) 1.09 (1.01-1.19)
 % Hispanic      
  Low Reference Reference Reference
  Middle 0.86 (0.79-0.94) 0.96 (0.91-1.02) 0.95 (0.88-1.02)
  High 0.85 (0.73-0.98) 1.02 (0.97-1.08) 1.01 (0.93-1.09)
 Median household income      
  Low Reference Reference Reference
  Middle 1.08 (0.97-1.19) 0.92 (0.87-0.98) 0.89 (0.82-0.96)
  High 1.13 (1.0-1.28) 0.88 (0.83-0.94) 0.85 (0.78-0.92)
Medical oncologist capacity      
 Low Reference    
 Middle 1.03 (0.88-1.19) 1.05 (0.99-1.12) 1.08 (1.0-1.17)
 High 1.21 (1.03-1.42) 1.00 (0.94-1.08) 1.03 (0.94-1.13)

Outcome 3: Radiation Therapy

Among those eligible for radiation therapy, blacks were less likely to undergo radiation therapy compared with whites (OR, 0.68; 95% CI, 0.53-0.88) (Table 4). There was no statistically significant difference in receipt of radiation therapy between Hispanics and whites. Individuals with stage 3 disease were more likely than those with stage 2 to undergo radiation therapy. These relationships remained consistent with the addition of area variables to the model. Counterintuitively, we found that use of radiation therapy was lower in areas of higher income.

Table 4. Radiation Therapya and Mortality: Patients With Rectal Cancer, Stages 2 and 3 (N = 6083)
 Radiation TherapybAll-Cause MortalitycColorectal Cancer Mortalityc
Model with Individual Covariates OnlydModel with Individual and Area CovariatesModel with Individual Covariates OnlyModel with Individual and Area CovariatesModel with Individual Covariates OnlyModel with Individual and Area Covariates
  • a

    Radiation therapy among patients who underwent resection within 6 months of diagnosis for rectal cancer stage 2 and 3.

  • b

    Data are presented as odds ratio (95% confidence interval).

  • c

    Data are presented as hazard ratio (95% confidence interval).

  • d

    All models also adjusted for sex, age (continuous), marital status, comorbid conditions, and year of diagnosis and clustered by HSA.

Individual characteristics      
 Race/ethnicity      
  WhiteReferenceReferenceReferenceReferenceReferenceReference
  Black0.68 (0.53-0.88)0.68 (0.51-0.92)1.27 (1.10-1.46)1.15 (0.98-1.35)1.41 (1.18-1.69)1.23 (1.01-1.5)
  Hispanic1.12 (0.89-1.40)1.17 (0.95-1.45)0.97 (0.84-1.12)0.90 (0.78-1.05)0.97 (0.79-1.18)0.93 (0.75-1.14)
 Ever eligible for state buy-in coverage (vs never eligible)0.86 (0.74-1.01)0.85 (0.73-0.99)1.16 (1.05-1.28)1.12 (1.02-1.24)1.05 (0.92-1.20)1.0 (0.87-1.14)
 Stage 3 (vs stage 2)1.81 (1.63-2.01)1.81 (1.63-2.02)1.41 (1.32-1.51)1.42 (1.32-1.52)1.78 (1.62-1.96)1.79 (1.63-1.97)
Area characteristics      
 Urban county (vs rural) 1.07 (0.84-1.36) 1.01 (0.88-1.14) 0.99 (0.84-1.17)
 % Black      
  Low Reference Reference Reference
  Middle 0.92 (0.82-1.03) 0.94 (0.86-1.02) 0.97 (0.86-1.11)
  High 0.87 (0.77-1.00) 1.01 (0.91-1.11) 1.08 (0.95-1.23)
 % Hispanic      
  Low Reference Reference Reference
  Middle 0.98 (0.84-1.14) 0.99 (0.90-1.08) 0.98 (0.87-1.11)
  High 0.89 (0.73-1.08) 1.06 (0.95-1.17) 1.0 (0.87-1.14)
 Median household income      
  Low Reference Reference Reference
  Middle 0.96 (0.84-1.10) 0.88 (0.80-0.96) 0.85 (0.75-0.96)
  High 0.83 (0.69-0.99) 0.83 (0.76-0.92) 0.81 (0.71-0.93)
Radiation oncologist capacity      
 Low Reference    
 Middle 1.01 (0.82-1.25) 0.93 (0.84-1.03) 0.95 (0.83-1.09)
 High 1.06 (0.86-1.30) 0.92 (0.82-1.02) 0.92 (0.80-1.06)

For these eligible individuals, all-cause and CRC-specific mortality was greater for blacks than whites (HR 1.27, 95% CI 1.10-1.46 and HR 1.41, 95% CI 1.18-1.69, respectively) but not different for Hispanics, compared with whites. Mortality was also greater for patients diagnosed at a later stage and for disadvantaged patients. Addition of area sociodemographics and supply of radiation oncologists to the model reduced the black/white disparity in all-cause mortality by 9.4% in CRC-specific mortality by 12.8%. Mortality decreased as area income increased but was not changed by increasing supply of radiation oncologists.

DISCUSSION

This study extends our understanding of disparities in CRC care and mortality for blacks compared with whites2-9 by suggesting that sociodemographic characteristics and measures of the availability of specialized cancer providers in the area where an individual resides modestly mediate these disparities. Among individuals eligible for surgical resection, we found that seniors who lived in areas with a greater percentage of blacks were less likely to undergo resection, that all-cause mortality was higher for individuals who lived in an area with a greater percentage of Hispanics, and that both all-cause and CRC-specific mortality declined with ecological income measures. Among individuals eligible for adjuvant chemotherapy, we found that individuals who lived in areas with a greater percentage of Hispanics were less likely to receive adjuvant chemotherapy and that individuals with a higher median household income were more likely to receive adjuvant chemotherapy and had lower all-cause and CRC-specific mortality. Among individuals eligible for radiation therapy, we found that—counterintuitively—usage was lower in areas with a higher household income but that all-cause and CRC-specific mortality was also lower in these areas. We also found that seniors were more likely to undergo resection and appropriate adjuvant chemotherapy if they lived in an area with more surgeons and medical oncologists, respectively, but we did not see an association between area capacity of radiation oncologists and use of radiation therapy.

Although many factors contribute to persistent disparities in health care,20, 21 the literature increasingly suggests that the characteristics of the place where an individual resides may be a mediator, particularly for health care services that are highly specialized (eg, cancer care). Prior studies have demonstrated that residents of more impoverished areas are less likely than those from affluent areas to receive cancer surgery.11, 12 In Georgia, urban or suburban residence was associated with greater likelihood of receiving recommended chemotherapy for CRC, but black/white disparities in care were greatest in urban areas, largely because the use of chemotherapy by rural whites was lower than any other group.22 Black/white disparities in resection for patients with nonmetastatic rectal cancer were most apparent in lower income areas.10 Our findings of black/white disparities in cancer resection and use of adjuvant chemotherapy are consistent with work by Hayanga et al13 that found a negative association in a county-level analysis between the percentage of blacks and the percentage of radiation oncologists and gastroenterologists. Our work extends these studies to show that area sociodemographic characteristics and availability of specialized physicians are associated with the outcomes of individual patients with CRC.

The current study also adds to our previous work, which found that residential segregation mediates some of the black/white disparities in breast cancer care, but not mortality,23 and that disparities in the use of end-of-life care for seniors with cancer are associated with the prevalence of blacks and Hispanics in a community.24 Our findings are also consistent with reports suggesting that differences in access to medical oncologists are associated with black/white disparities in the use of chemotherapy among individuals with lung cancer.25

Our study has some potential limitations. The variables we used to measure physician capacity may underestimate regional capacity, particularly in areas bordering SEER areas, because we counted only those physicians who treated individuals covered by fee-for-service Medicare in the SEER-Medicare data. Although Medicare procedure volume is an underestimate of total procedures, Medicare claims are widely used to define the volume of procedures at both the hospital and geographic area,26-31 and studies have demonstrated that Medicare volume and total volume are highly correlated and produce similar volume rankings.28 Although we had information on the socioeconomic status of the area where an individual lived, we were limited to information about eligibility for state buy-in coverage for Medicare as a marker of individual low-income status. Physician capacity does not guarantee access to care; minority and disadvantaged patients who live in areas of high capacity may face additional barriers to care. In addition, our findings may not be generalizable to younger or uninsured individuals, because our analysis focused on seniors with Medicare coverage. There may be greater disparities in care and survival among younger patients with more varied health insurance.

Reducing disparities in CRC care and mortality for blacks may require attention to differences in the resources available in the areas where individuals reside. Although access to specialized health care providers was modestly associated with some of these differences in CRC care, area sociodemographic characteristics were also associated with receipt of CRC care and mortality. Further work should examine how these and other measures of area characteristics may mediate these differences.

FUNDING SOURCES

Supported by grants from the American Cancer Society (RSGT CPHPS-114979).

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary