This study used the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development and Information, Centers for Medicare & Medicaid Services; Information Management Services (IMS), Inc.; and the SEER Program tumor registries in the creation of the SEER-Medicare database.
Professional societies recommend posttreatment surveillance for colorectal cancer (CRC) survivors. This study describes the use of surveillance over time, with a particular focus on racial/ethnic disparities, and also examines the role of area characteristics, such as capacity for CRC screening, on surveillance.
Surveillance, Epidemiology, and End Results (SEER)-Medicare data were used to identify individuals aged 66 to 85 years who were diagnosed with CRC from 1993 to 2005 and treated with surgery. The study examined factors associated with subsequent receipt of a colonoscopy, carcinoembryonic antigen (CEA) testing, primary care (PC) visits, and a composite measure of overall surveillance.
Of eligible subjects, 61.0% had a colonoscopy, 68.0% had CEA testing, 77.1% had PC visits, and 43.0% received overall surveillance. After adjustment, blacks were less likely than whites to undergo colonoscopy (odds ratio [OR] 0.76, 95% confidence interval [CI] = 0.69-0.83) and to receive CEA testing and overall surveillance, whereas white/Hispanic rates did not differ. Rates for all outcomes increased from 1993 to 2005, but black/white disparities remained. Individuals in areas with greatest capacity for CRC screening were more likely (OR = 1.09, 95% CI = 1.02-1.18) to receive colonoscopy, and those in areas with the greatest percentage of blacks were less likely (OR = 0.89, 95% CI = 0.83-0.95) to receive colonoscopy. Those living in areas with shortage of PC were less likely to receive PC visits (OR = 0.55, 95% CI = 0.48-0.64) and overall surveillance (OR = 0.83, 95% CI = 0.71-0.98).
Colorectal cancer (CRC) is the third most common type of cancer in the United States and the third leading cause of cancer death; approximately 140,000 individuals will be diagnosed and 49,000 will die of CRC this year.1 CRC survivors are at risk for local recurrences and second primary cancers even if the first cancer is completely removed; approximately 50% who undergo resection with a curative intent will have a recurrence within 3 to 5 years of an initial diagnosis.2, 3 Routine follow-up is recommended to detect new precancerous lesions in the colon or recurrent cancer when potentially curative surgical resection is possible.4
Professional society guidelines for postoperative CRC surveillance vary in the timing and components of surveillance. The National Comprehensive Cancer Network (NCCN) recommendations have included colonoscopy within 1 year of resection since 1996 for stage I, II, and III CRC.5 NCCN recommendations also include repeated colonoscopy in 3 years and then every 5 years, carcinoembryonic antigen (CEA) testing for stage II and III CRC every 3 to 6 months for 2 years and primary care (PC) visits every 3 to 6 months for 2 years for stage I, II, and III CRC.6, 7 Since 1999, the American Society of Clinical Oncology has recommended similar CEA testing and PC visits, but colonoscopy at 3 and then every 5 years after surgery.8, 9 In 2006, the American Cancer Society and the US Multisociety Task Force on Colorectal Cancer recommended colonoscopy within 1 year of resection and then at 3 and 5 years, but did not address CEA testing or PC visits.2
Compared with other race/ethnicities, CRC incidence and mortality rates are highest in black men and women; incidence rates are 20% higher and mortality rates are approximately 45% higher than those in whites.10 Black/white disparities in postoperative CRC surveillance have been documented.3, 4, 11-14 Earlier studies show blacks to be approximately 30% to 40% less likely to receive postoperative surveillance.4, 13 The purpose of this study is to further describe rates of postoperative surveillance among white, black, and Hispanic CRC patients who are covered by Medicare, by specifically examining whether racial/ethnic disparities have diminished over the period from 1993 through 2005 and by examining the role of area sociodemographic and health service characteristics, on disparities. We explore whether sociodemographic characteristics of the area where a subject resides, including income and ethnic/racial composition, and health service characteristics, including CRC screening and PC capacity, are associated with the receipt of postoperative surveillance. Our analysis is based on Andersen's conceptual model of health care use which considers the influence of the external environment and health care system on use of health services.15 In earlier work, also based on this model, we found that racial/ethnic disparities in CRC screening and care in the Medicare population were mitigated, to some extent, by characteristics of the area where an individual lives.16
MATERIALS AND METHODS
These analyses used several data sources. The Surveillance, Epidemiology and End Results (SEER) program collects information on all incident cancer cases for persons with cancer residing in SEER regions, including primary tumor site, cancer stage, patient demographics, and 2000 US census tract census data.17 The SEER data are linked to Medicare claims data by the National Cancer Institute.17 Medicare claims data include the National Claims History (carrier) file for physician claims, the Medicare Provider Analysis and Review (MEDPAR) file for hospitalizations, and skilled nursing facility admissions and outpatient files from outpatient facilities. The SEER-Medicare data were used to identify: 1) patients diagnosed with CRC from 1993 to 2005; 2) initial surgeries; and 3) postoperative colonoscopy, CEA testing, and PC physician visits (family practice, internal medicine, general practice, geriatric specialties) from 1993 to 2007 to create measures of postoperative surveillance.
The 2000 Medicare carrier and outpatient files were used to identify persons who had a screening sigmoidoscopy or colonoscopy in 2000 to create aggregate measures of county CRC screening capacity.16 The Area Resource File (ARF) file was used to create measures of county PC capacity and ascertain county population for 2000. This study was reviewed and approved by the Institutional Review Board of Partners HealthCare (Boston, Mass).
Persons aged 66 to 85 years, whose race/ethnicity was reported as white, black, or Hispanic and who were diagnosed with CRC as a single, first cancer between 1993 and 2005 were included. Other racial/ethnic groups were not included due to low numbers. We chose the 12-year study period, starting prior to institution of NCCN Guidelines, to allow for detection of any changes in surveillance over time. Stage was classified according to the extent of tumor, extent of spread to the lymph nodes, and the existence of distant metastasis (TNM). The cohort was restricted to those diagnosed with local (TNM stage 1 or 2) or regional (TNM stage 3) CRC,18 and whose primary course of treatment included surgery (N = 56,880). We excluded subjects younger than age 66, because they may not have complete Medicare comorbidity data, and persons older than age 85 at diagnosis, because they may not be candidates for colonoscopy in accord with the American Cancer Society and US Multisociety Task Force Guidelines on Colorectal Cancer, which recommend consideration of discontinuance of surveillance colonoscopy for persons with advanced age or comorbidities, who have < 10 years of life expectancy.2
We excluded subjects who did not have both Medicare part A and B coverage or who were members of a health maintenance organization (HMO) for 13 months prior to diagnosis because comorbidity data may be incomplete (N = 6986).19 For each time period under study, we excluded persons who did not have both Medicare part A and B coverage or who were HMO members for the entire time period, because their Medicare claims may not be complete.19 We also excluded persons who were not eligible to receive the outcome during the entire time period due to: death; study period end (December 31, 2007); or attainment of age 86.
Our outcome measures were generally in accord with NCCN guidelines,6, 7 the earliest and more comprehensive guidelines. These include 1) at least 1 colonoscopy within 15 months of surgery for all subjects, 2) at least 2 CEA tests within 2 years of surgery for subjects diagnosed at stage II or III, and 3) at least 2 PC visits within 2 years of surgery for all subjects. We also examined a composite measure of all 3 measures, which we defined as “overall surveillance”: colonoscopy and PC visits for stage I or all 3 for stages II and III. We defined the index surgery date as the last surgery within 2 months prior through 6 months after diagnosis. Procedures and office visits within 3 months of surgery were excluded to avoid including routine postoperative visits and tests to evaluate postoperative complications.5, 11 Fifteen months was chosen as the time interval for colonoscopy, because subjects may not have been able to schedule tests at exactly 12 months. Two CEA tests in 2 years and 2 PC visits in 2 years were considered compliant to give subjects time to complete adjuvant chemotherapy, which may last up to 9 months after resection. Outcome measures were identified by Medicare claims. The sample eligible for each outcome was: 1) colonoscopy, 38,889 persons (10,187 excluded due to lack of Medicare or HMO membership at any time for 18 months after surgery, 818 excluded due to death, age, or end of study period); 2) CEA testing, 23,093 persons, and 3) outpatient visits and overall surveillance, 33,648 persons.
Patient characteristics included age at diagnosis (66-70, 71-75, 76-80, 81-85 years), sex, race/ethnicity (white, black, Hispanic), marital status (married, not married), stage at diagnosis (I, II, III), cancer location (colon, rectal), and whether an individual was of “low-income status,” based on eligibility for state assistance with Medicare premiums and copayments (state buy-in). We also included the Charlson comorbidity index, a summary measure of comorbid conditions, each assigned a weight according to its influence on mortality, with higher scores indicating greater comorbidity (categorized as 0, 1, 2, ≥ 3).19
Area characteristics from the SEER-Medicare file included the SEER registry and measures of 2000 census tract characteristics, categorized in tertiles as follows: the percentage of the population who were black (≤ 0.88%, 0.89%-3.98%, 3.99%-100%), the percentage of the population who were Hispanic (≤ 2.15%, 2.16%-7.20%, 7.21%-99.2%), and median household income (≤ $35,431, $38,432-$54,349, $54,350-$200,008). We also included a measure of county-level CRC screening capacity in 2000, defined as the number of screening colonoscopies/sigmoidoscopies per 100,000 residents aged 50 years and older,16 categorized in tertiles (6.06-1001.90, 1001.91-1787.75, 1787.76-6509.08), and an ARF measure of PC capacity in 2000, categorized as none, part, or all of the county designated as a PC shortage area. Tertiles were based on the colonoscopy sample.
We used bivariate analyses to examine the associations of an individual's race/ethnicity with demographic, area, and outcome variables; chi-square tests were used to determine statistical significance. We determined the statistical significance in unadjusted temporal trends in outcome by the Mantel-Haenszel chi-square. In order to examine rates of each outcome, we developed logistic regression models for the time intervals specified above, using SAS, version 9.2 (SAS Institute, Cary, NC). For each outcome, models were adjusted for individual characteristics (age at diagnosis, sex, race/ethnicity, marital status, comorbidity, eligibility for state buy-in coverage, stage, year of diagnosis); census tract characteristics (percent black, percent Hispanic, median income); SEER registry; county characteristics (CRC screening and PC capacity); and a time variable to account for changes in the slope of the relationship between diagnosis year and the outcome. For instance, colonoscopy rates generally increased for the first 7 years, from 1993 to 1999, and then remained fairly flat for the next 6 years, from 2000 through 2005. For colonoscopy models, a piecewise linear time trend with a breakpoint at 1999 was assumed, allowing for a different rate of change in the rates before and after 1999. In order to test whether racial/ethnic effects changed over time, we included 2-way interaction terms of race/ethnicity and 1) year of diagnosis, and 2) the time trend variable. Interactions were removed from the models if they were not statistically significant. The same models were used to compute predicted rates of outcomes for each racial/ethnic group, for each year. In these models, covariates were fixed at the mean values observed across the sample.
Compared to whites with CRC, both blacks and Hispanics were more likely to be eligible for state buy-in coverage, and have more comorbid conditions (Table 1). Blacks were more likely to be diagnosed with stage III disease (34.2%), compared to whites (29.2%). Blacks were most likely to live in areas with high numbers of other blacks (94.6%). Similarly, Hispanics were most likely to live in areas with high numbers of other Hispanics (87%). Both blacks and Hispanics were more likely to live in areas with a lower household income than whites. Whites were more likely to live in counties that were not classified as PC shortage areas (34.0%) compared with blacks (18.5%) and Hispanics (17.4%). Hispanics were more likely to live in counties with the lowest screening capacity (59.0%) compared with whites (32.9%) and blacks (31.8%), and whites were most likely to live in counties with the greatest screening capacity.
Table 1. Description of the Sample Eligible for Colonoscopy by Race/Ethnicity
Screening capacity (number of colonoscopies or sigmoidoscopies per 100,000 residents age 50 and older in county)b
Overall, 61.0% of individuals had a colonoscopy within 1 year of surgical resection, 68.0% of stage II and III subjects had at least 2 CEA tests within 2 years, 77.1% of all patients had at least 2 primary care visits within 2 years, and 43.0% received overall surveillance (Table 2). Blacks were less likely than whites and Hispanics to receive a colonoscopy and overall surveillance. A total of 17,991 patients were excluded from the colonoscopy outcome due to HMO membership before (6986 patients) or after surgery (10,187 patients), and 818 more were excluded due to death, age, or end of the study period. Blacks (38.0%) and Hispanics (41.3%) were more likely to be excluded compared with whites (30.5%).
Table 2. Surveillance After Resection for Colorectal Cancer, by Race/Ethnicity
Abbreviation: CEA, carcinoembryonic antigen.
P value for chi-square: colonoscopy, office visits, overall surveillance <.0001; CEA testing .0001.
After adjustment, blacks (OR = 0.76, 95% CI = 0.69-0.83) were less likely than whites to have a colonoscopy (Table 3), whereas there was no difference between Hispanics and whites. Men were less likely than women to have a colonoscopy (OR = 0.83, 95% CI = 0.80-0.87) as were individuals eligible for state buy-in insurance. The odds of having a colonoscopy decreased with increasing age. Individuals with colon compared to rectal cancer were more likely to have colonoscopies (OR = 1.25, 95% CI = 1.19-1.32). Those who lived in an area with highest percentages of blacks were less likely to have colonoscopies, and those in areas of highest income and greater screening capacity were more likely to have colonoscopies.
All models also adjusted for comorbidity, year of diagnosis, a time variable, marital status, and SEER registry.
OR and CI in boldface type are significant at P = .05.
Eligible for state buy-in insurance
Primary care shortage area
Not a shortage area
Partial shortage area
After adjustment, blacks were less likely than whites (OR = 0.77, 95% CI = 0.69-0.87) to have at least 2 CEA tests within 2 years of surgery (Table 3). Men, individuals eligible for state buy-in insurance, and older individuals were also less likely to have the requisite CEA tests. Colon cancer patients were less likely to have CEA testing (OR = 0.82, 95% CI = 0.76-0.88) compared with those with rectal cancer
Hispanics were more likely than whites to have 2 PC visits within 2 years following diagnosis (OR = 1.14, 95% CI = 1.0-1.30). Rates for blacks and whites were similar (Table 3). Similar to results for colonoscopy and CEA testing, men and those eligible for state buy-in insurance were less likely to have primary care visits. Individuals older than 66 to 70 years and those with colon cancer versus rectal cancer were more likely to have PC visits. In addition, those living in PC shortage areas were less likely to have PC visits (OR = 0.55, 95% CI = 0.48-0.64) compared with individuals living in areas with no PC shortage.
Compared with whites, blacks were less likely (OR = 0.75, 95% CI = 0.67-0.83) to receive overall surveillance in the 2 years following CRC surgery, whereas there was no difference between whites and Hispanics (Table 3). Overall surveillance decreased with increasing age. Similar to results for the individual components, men and those eligible for state buy-in insurance were less likely to receive overall surveillance. Compared with subjects with rectal cancer, subjects with colon cancer were more likely to receive this care. Individuals who lived in PC shortage areas were also less likely to receive overall surveillance (OR = 0.83, 95% CI = 0.71-0.98) compared to individuals living in areas with no PC shortage.
Racial Disparities Over Time
Rates of all outcomes increased over the study period for each racial/ethnic group (data not shown). From 1993 to 2005, colonoscopies increased from 52.2% to 61.5%, for whites, 44.4% to 53.2% for blacks and 50.0% to 60.2% for Hispanics. Over the same period, rates of overall surveillance increased from 41.4% to 45.0% for whites, 29.1% to 38.6% for blacks, and 34.5% to 41.3% for Hispanics. Adjusted predicted probabilities over time for overall surveillance show that racial/ethnic differences in the adjusted rates were consistent from 1993 to 2005 (Fig. 1), suggesting that disparities did not diminish. Racial/ethnic differences in the adjusted rates for all other outcomes were also consistent over the time period (data not shown).
This work demonstrates that many CRC survivors do not get recommended surveillance and that black/white disparities in surveillance persist despite increasing use of surveillance over time. We also found that there were few differences in CRC surveillance between Hispanics and whites. This manuscript extends previous work on surveillance for CRC3-5, 12-14, 20, 21 by examining more recent trends in use, all tests currently recommended for surveillance, and exploring whether area characteristics are associated with surveillance. Our overall rate for surveillance colonoscopy (61%) is comparable to previously reported rates for CRC survivors, ranging from 52% to 61%,11, 14, 21, 22 and our finding of increasing colonoscopy use over time, reflecting changes in the recommendations of professional societies,2, 5-8 is also consistent with earlier studies of temporal trends in colonoscopy.5, 20
In addition, our findings of black/white disparity in surveillance colonoscopy3, 14, 20, 21 and CEA tests4, 13, 21 are comparable to prior analyses, even though time frames and frequencies varied. Our data show strong racial/ethnic segregation, lower colonoscopy rates in areas with the highest percentage of black residents, and higher rates in areas with greater supply of providers. The black/white disparities in colonoscopy use may reflect differences in access to providers, neighborhood social support, or beliefs about and preferences for treatment, consistent with suggestions that socioeconomic, health care access, and cultural factors may contribute to black/white disparities in CRC screening among the Medicare population.23 Lack of access to testing has been identified as a barrier to postoperative CRC follow-up, and regional variation in capacity has been associated with regional variation in the use of CRC screening in the general population.16, 24
Our data show that PC office visits are associated with PC supply and that there is no black/white disparity in PC visits in the 2 years following surgery. Office visits provide the opportunity for coordination of care, including surveillance tests, and for discussion of relevant health information, including benefits associated with continued surveillance. Black/white disparities may be associated with differences in or compliance with recommendations or differences in patient/physician communication. Previous studies of racial differences in physician/patient communication among cancer patients has found whites to be more satisfied than blacks with the health information received and the coordination of care.25, 26 Perception that risk or CRC recurrence is low and confusion about which provider is responsible for tests have been cited as a barriers to surveillance after surgery.24, 27, 28 Further research into these issues related to communication is needed.
Our study has several limitations. Although we had information on area socioeconomics, we were limited to information about eligibility for state buy-in coverage for Medicare as a marker of individual low-income status. We also did not have information on individual access to health care. Our analysis focused on seniors with Medicare fee-for-service coverage. Thus, our findings may not be generalizable to younger or uninsured persons or HMO members. Because a higher percentage of HMO members are black or Hispanic, our results would be biased if HMO members were more likely than fee-for-service patients to receive surveillance. We did not include sigmoidoscopies and barium enemas in our study; others have found rates to be low, declining over time.20 Furthermore, we were not able to distinguish surveillance from diagnostic testing. However, previous analyses found low sensitivity for using diagnostic codes to identify diagnostic examinations and office visits and guideline-recommended testing to be generally performed for routine surveillance among cancer survivors. We also note the conflicting professional society guidelines for surveillance among CRC survivors. Those who have a colonoscopy at 3 years after resection would be consistent with American Society of Clinical Oncology guidelines but would not meet our definition for overall surveillance.
Analyses of randomized trials of intensive follow-up after surgery for CRC, including colonoscopy and CEA testing, show improved overall survival.29 Our study demonstrates low rates of surveillance among CRC survivors and ongoing black/white disparities in surveillance. Continued timely surveillance is vital for CRC survivors. Further studies are needed to identify the underlying reasons for the low rates and continued disparities in surveillance, and to further define the role of area characteristics on these disparities.
This study was supported by a grant from the American Cancer Society (RSGT CPHPS-114979).