The data source for this study was the Cancer Control Module of the 2000 National Health Interview Survey (NHIS). The NHIS is an in-person health interview conducted by the U.S. Census Bureau for the National Center for Health Statistics.24
The NHIS uses a multistage probability sample design to represent current health information for the civilian, noninstitutionalized, household population of the United States. For the year 2000 survey, the overall sample adult response rate was 72.1%.25
In 2000, a supplement to the core NHIS was the Cancer Control Module.24
This module consisted of seven sections covering health behavior, family cancer history, and cancer screening. For the 2000 Cancer Control Module, 32,374 persons, 18 years or older, were interviewed. A total of 5,564 respondents were eligible for our analysis (Fig. ).
Study population CRC = colorectal cancer.
All adults over the age of 40 were asked about CRC screening procedures and the date of the most recent examination. A respondent was considered to have undergone CRC screening if they reported ever having had a colonoscopy, sigmoidoscopy, or home-based fecal occult blood testing (FOBT). Because office-based fecal occult testing is not considered an appropriate CRC screening strategy, we did not include this within our analysis.9
The 2000 Cancer Control Module did not inquire about barium enema testing, and as such, this variable could not be included within our analysis.
We considered respondents to be current with FOBT if they reported having at least one test within the past year. Sigmoidoscopy and colonoscopy were collapsed into a single endoscopy variable. Respondents were considered to be current with endoscopic screening if they reported having at least one of these procedures within the previous 10 years. Because sigmoidoscopy screening is recommended to occur at 5-year intervals, we performed a secondary analysis in which we defined being current with endoscopic screening as reporting at least one procedure during the previous 5 years. Models were then constructed using both definitions (10 and 5 years) for endoscopic screening. The primary analysis included the 10-year screening definition. When data are presented relating to the 5-year interval definition, this is indicated specifically within the text.
To determine if an individual might be a candidate for early initiation of cancer screening, we used guidelines available in 2000.26–28
At this time, the American Cancer Society was recommended initiating CRC screening at age 40 in individuals with either a first-degree relative affected with CRC before age 60 years while the American Gastroenterology Association (AGA) suggested early initiation of CRC surveillance for any individual with an affected first-degree relative.27
For the purpose of this analysis, we used the conservative AGA recommendations and considered any individuals with an affected first-degree relative, regardless of age, as a candidate for early initiation of CRC screening.
Respondents who reported undergoing endoscopy or FOBT were asked to give the reason for undergoing the procedure. Responses were categorized as: (1) part of a routine physical exam/screening test, (2) because of a specific problem, (3) follow-up test of an earlier test or screening exam, (4) family history, or (5) other. Respondents also rated their own perceived likelihood of developing cancer by responding to the question, “Would you say your risk of getting cancer in the future is low, medium, or high.” For the purpose of our analysis, we dichotomized this response into “low likelihood” and “medium or high likelihood.”
We used self-reported demographic information to construct our independent covariates. Individuals who reported their typical source of health care as a clinic, health center, doctor’s office, health maintenance organization, or a hospital outpatient department were categorized as having a usual source of care. Respondents who indicated that they had no usual source of care or obtained their typical health care through an emergency room were categorized as having no usual source of care. We grouped self-reported educational levels into four categories: less than high school, graduated high school, some college, and graduated college with or without professional or graduate school. We categorized respondents into those with insurance coverage, defined as individuals with private insurance, military insurance, other governmental insurance, Medicare or Medicaid, and those reporting no insurance coverage whatsoever.
Screening rates were stratified by family history and race. We only included respondents whose self-reported race was either “white” or “black.” We did not include respondents reporting other races within the analysis, as these groups were small. We used χ2 tests to compare screening rates between groups.
To determine the impact of race on early initiation of CRC screening in increased risk individuals, we constructed a separate logistic multivariate model for each CRC screening outcome (endoscopy within the preceding 10 years and FOBT within the past year). These models were adjusted for potential confounders to CRC screening including age, race, family history, gender, education, insurance status, and having a usual source of care.24
To further determine the impact of a positive family history for white and African-American respondents, we created two additional logistic multivariate models which were stratified by race. All data analyses were weighted to reflect national population estimates. To account for the NHIS multistage probability sample design, standard errors were adjusted using SAS-callable SUDAAN software, version 9 (Research Triangle Institute, Research Triangle Park, NC).