Colorectal Cancer Screening Adherence in a General Population

In the United States, an estimated 147,500 incident cases of colorectal cancer (CRC) and 57,100 deaths from CRC will occur in 2003 (1). Only 37% of cases are detected early (2). Despite evidence that screening can reduce CRC incidence and mortality (3–11), these tests remain underutilized (12, 13).

To determine the extent of nonadherence to screening in the U.S. population, the timing of all use of all screening modalities must be identified. In this regard, the National Health Interview Survey (NHIS) and the Behavioral Risk Factor Surveillance System (BRFSS), the two most important sources for estimating screening nonadherence in the general U.S. population, have shortcomings (12, 13). Until recently, the NHIS reported only CRC testing done for screening. The BRFSS and NHIS questions do not differentiate between flexible sigmoidoscopy and colonoscopy. Nonadherence to current screening guidelines, which describe different screening intervals for these procedures, becomes impossible to accurately assess. Neither BRFSS nor NHIS addresses double-contrast barium enema, now an option for CRC screening. Lastly, reports from the BRFSS and NHIS use different and inconsistent screening periods to report outcomes (12–15). Thus, we cannot compare current reports on CRC screening nonadherence from different sources, make historical comparisons, or easily assess progress in improving CRC screening coverage.

The consistent use of current screening guidelines and capturing both diagnostic and screening use of tests can overcome these shortcomings. The U.S. Multisociety Task Force on Colorectal Cancer recommended screening for average-risk adults aged 50 and older by any of four modalities: annual fecal occult blood test (FOBT), flexible sigmoidoscopy every 5 years, colonoscopy every 10 years, or a barium enema every 5 years (16). These guidelines essentially agree with American Cancer Society recommendations and the U.S. Preventive Services Task Force guidelines (17, 18).

Prior to a trial of community-based interventions promoting CRC screening, we surveyed adults in five Minnesota counties to determine the proportion whose utilization of colorectal screening was not in accordance with current recommendations (16, 19). We assessed utilization rates for each of the four testing modalities within time frames specified by the 2003 U.S. Multisociety Task Force on Colorectal Cancer recommendations (16).

The University of Minnesota Institutional Review Board approved this study. Eligible participants were at least 50 years old and residents of Isanti, Kandiyohi, Rice, Steele, or Wright counties. Wright County—the intervention site for the trial—is semirural, with a large farming population and several small communities within commuting distance of a major metropolitan area. Its 1999 population was 87,864, with 21.6% (18,979) ≥50 years old (20). The other counties, selected as controls, approximated these characteristics.

We targeted a random sample of 2000 eligible participants (1500 from Wright and 500 from control counties) from the 1999 Minnesota State Driver's License and Identification Card database. Because some individuals in the database may have been ineligible (e.g., by death), we oversampled 2590 individuals. In 1997, 79% of Minnesota's residents had driver's licenses (21); 13% of Minnesota residents in the database have ID cards only, so the percentage represented by the database is probably larger than 90%.

A 55-item mailed questionnaire completed between February and October 2000 with telephone follow-up assessed participants' CRC screening practices and beliefs. The questions on CRC screening practices were developed to assess each of the four screening modalities. Because we focused on nonadherence, our survey could be simplified by not asking the purpose of testing. Each question was prefaced with a two-to-four-sentence description of the test (available upon request) to help respondents differentiate between the tests. The main distinguishing characteristics of the tests cited were as follows: (1) FOBT is performed at home, and the cards are then returned to the clinic; (2) flexible sigmoidoscopy (as opposed to colonoscopy) is performed with a shorter flexible tube, takes less time, is performed in the physician's office, and does not require medication; and (3) barium enema requires an X-ray after drinking a large amount of white fluid.

The first mailing was followed by up to three mailed reminders, including resending surveys. To maximize participation of persons for whom a mailed survey could be a barrier (e.g., the elderly and persons with reading problems or low literacy), we attempted to call all nonrespondents and ask the four screening test questions.

The primary outcome was overall nonadherence to the U.S. Multisociety Taskforce on Colorectal Cancer screening guidelines (16). Subjects were considered adherent if they reported having a FOBT within the previous year, a flexible sigmoidoscopy or barium enema within 5 years, or a colonoscopy within 10 years. All others were nonadherent.

To better represent the target sample estimates, screening probabilities were adjusted using a modified form of Horvitz-Thompson weighting (22) that simultaneously adjusts for nonresponse and unknown eligibility (23). Rather than assume that all nonrespondents are eligible, this modified method employs a flexible model based on both continuous and discrete covariates in the driver's license database (age, gender, license, and address status) to estimate the probability of eligibility for nonrespondents and differentially down-weight their contribution to the nonresponse adjustment. To incorporate variability both from sampling and from estimating the weights, confidence intervals (CIs) were constructed using bootstrap methods that include computing the estimated weights. CIs for adherence rates are based on 2000 bootstrap samples using the bias-corrected, accelerated method (24). Data analysis was performed on the data set current as of May 1, 2001.

Eligible respondents numbered 1689, 89.6% of confirmed eligible individuals. However, correcting for unknown eligibility lowered the percentage to 86.3%. Of the remaining 901 in the sample, 383 (42.5%) were dead before their survey mailing date and therefore ineligible, 150 (16.6%) were ineligible due to incorrect addresses in the driver's license database or relocation before their survey mailing date, 196 (21.8%) were eligible but did not respond to the survey adequately for determination of overall screening compliance status, and 172 (19.1%) from whom surveys were not obtained and there was not enough other contact information to verify study eligibility. Respondents' mean age was 63.7 years, 53% were female, 79% were married, 17% had a college degree or higher (Minnesota = 24.7%, United States = 22.8% for those 45 years and older), and 15% had annual household incomes less than $15,000 (Minnesota = 14.5%, United States = 23.2% for those 45 and older; Ref. 25).

Response-adjusted, self-reported CRC test utilization rates are in Table 1. Overall, 44.7% of respondents reported having no CRC examinations within screening guidelines and thus were nonadherent. Twenty-six percent were adherent to only one test and 29% were adherent to multiple tests: 19% reported two, 8% reported three, and 2% reported all four methods.

Participants 65 years or older were significantly less likely than those 50–64 years old to be nonadherent overall and for each testing method (Table 1). Women were significantly less likely than men to be nonadherent overall and for each testing method except FOBT.

CRC screening is underutilized in this population; 44.7% of those surveyed self-reported not meeting recommendations for screening. Because all tests were reported regardless of why they were done, the reported level of nonadherence is independent of any intention to screen.

Our findings approximate those from a population-based Massachusetts study where, overall, 49.1% had not been tested by any method (26). There were notable differences in our results and the Massachusetts data by testing type. Our sample reported lower rates of FOBT (24.5% Minnesota versus 32.8% Massachusetts) and higher rates of flexible sigmoidoscopy (33.8% Minnesota versus 24.2% Massachusetts), colonoscopy (29.3% Minnesota versus 9.0% Massachusetts), and barium enema (13.7% for a 5-year interval in Minnesota versus 4.8% for a 10-year interval in Massachusetts). This disparity may be due to different sociodemographics, question wording, or true differences in screening practices.

Collectively, these data highlight a need for efforts aimed at overcoming barriers to CRC screening and reaching the large proportion of nonadherent individuals. Our finding of gender- and age-specific differences in CRC examination rates demonstrate the importance of reaching adults aged 50–64 years and males with future interventions.

Our study is limited in its reliance on self-reported screening data, which could have resulted in overestimates of screening rates (27–29). In a recent study, improved accuracy in responses was demonstrated when questions were designed specifically to differentiate between CRC screening modalities—sensitivities ranged from 89% to 96% for individual tests and specificities ranged from 86% to 97% (30). Our survey questions, although developed before the publication of this study, were worded consistently with these validated questions.

For simplicity, we did not distinguish between tests done for screening and for diagnosis, which is appropriate for the focus on nonadherence. However, the rate of testing done strictly for screening cannot be determined, a shortcoming of this approach. Although not relevant to the questions addressed here, motivation for test use may be important in determining the specific mechanisms or effects of screening promotion efforts.

The generalizability of these results may be limited. These semirural Minnesota counties might differ from urban areas or other regions of the United States in factors such as access to care or health beliefs. Our population was somewhat more affluent than the United States overall and had lower levels of educational attainment. Nonetheless, our FOBT screening rate was only slightly higher than national rates (24.5% Minnesota versus 20.6% national 1999 BRFSS data), and our flexible sigmoidoscopy rate was essentially the same as national rates (33.8% Minnesota versus 33.6% nationally; Ref. 12).

Our study's strengths include its relatively large population-based sample and high participation rate. Previous studies have reported response rates of 65% or less (12, 13, 26) compared with our 86.3%. Horvitz-Thompson methods that adjust for nonresponse were modified to account for eligibility rates among nonrespondents. Moreover, the survey assessed all four screening options according to current guideline intervals, with questions that allowed differentiation between flexible sigmoidoscopy and colonoscopy. We recommend that future studies examining completeness of CRC screening in populations assess all use of all four screening modalities according to current guideline intervals and report summary nonadherence rates to promote greater consistency among research findings and to provide the most accurate assessment of the number of people who need screening.

We thank Dr. Ann Kinney for help in the preparation of the health questionnaire and Dr. Anne Marie Weber-Main for editorial assistance.