Our analysis adds to earlier work demonstrating regional variation in the use of CRC screening (13
) by examining the relationship between CRC screening use and the recommendations of primary care physicians, by hospital-referral region. Although higher proportions of physicians who correctly recommend CRC screening were associated with relatively small changes in the proportion of adults screened, increases in correct recommendations would result in many more people being screened. For example, if in each hospital-referral region the proportion of primary care physicians who recommend initiating screening at age 50 years increased by 30 percentage points, an estimated 1.5 million additional adults older than 50 years would be screened (based on an estimated U.S. population of 77 million older than 50 years, derived from the U.S. Census [www.census.gov/popest/national/asrh/NC-EST2005/NC-EST2005-01.xls
]). Similarly, a 30-percentage-point increase in the proportion of primary care physicians in each region who recommend screening at the correct interval could result in an additional 2.1 million people being current on screening.
Our work is consistent with earlier work suggesting that lack of provider counseling about CRC screening, rather than poor patient acceptance, is associated with lower rates of screening (28
). Patient recall of physician's recommendations is one of the strongest predictors of cancer screening (8
). Our findings suggest that population-based interventions directed at the CRC screening recommendations of primary care physicians may improve CRC screening use. Despite the endorsement of several influential national organizations and an awareness of the importance of CRC screening, however, many primary care physicians report screening practices that are inconsistent with the guidelines (12
). Practice guidelines alone may be limited in their effect on physician behavior for several reasons, including lack of awareness, lack of agreement with the recommendations, barriers to successfully implementing the guideline, and concerns about patient acceptance of the guideline (31
Several studies suggest that office-based systems may improve the prevalence of CRC screening in primary care practices (4
). One successful example, which was intended to increase cancer screening among disadvantaged patients, was based on the assignment of office responsibilities and the use of a cancer-screening checklist with chart stickers (35
). An intervention requiring quarterly feedback of a provider's CRC screening rates was also associated with increases in screening (36
). Although some studies suggest that local, practice-based physician-reminder systems may improve the delivery of CRC screening and other types of cancer prevention (4
), our results suggest a role for regional interventions to increase provider compliance with guidelines. Information on the feasibility of these types of interventions is limited, however, and one quality improvement program implemented by a managed care health plan to increase CRC screening was not successful (38
). Outreach and education by leaders in medical opinion (i.e., academic detailing), however, have been shown to improve adherence to guidelines for preventing myocardial infarction and other medical conditions (39
Our analysis has several limitations. The data do not allow us to examine the relationship between the recommendations of an individual's personal physician and that individual's screening behavior, and they are not intended to be a proxy for the recommendations of a specific physician. Rather, our findings reflect regional differences in physician recommendations. Although both the NHIS and the SCCSP are nationally representative, we included only respondents who lived in hospital-referral regions that were sampled in both surveys, and our results may not be generalizable to individuals in other areas. Finally, although we selected data from the NHIS that were collected several years after the SCCSP data that we used, some of the individuals in the NHIS may have been screened before the SCCSP was conducted. Unfortunately, NHIS does not allow identification of the precise year of a test.
Our findings indicate that regional differences in the recommendations of primary care physicians for CRC screening are associated with differences in screening use by individuals. For this reason, increasing the use of CRC screening in the United States may require interventions to improve the beliefs and recommendations of primary care physicians about CRC screening.