Screening has been shown to decrease the risk of mortality for colorectal cancer (CRC) [1–4]. Although the use of CRC screening has increased in the US, particularly over the past decade [5, 6], for many groups, screening rates are below the Healthy People goal of 70% [7]. Primary care physicians (PCPs) play an important role in the delivery of CRC screening services [6, 8–12] by advising, recommending, performing and/or referring patients for screening [13]. It is therefore not surprising that studies have consistently reported a strong association between healthcare provider recommendations for and receipt of CRC screening for persons with a usual source of care [8, 9, 14]. This supports the hypothesis that variations in the delivery of CRC screening services among PCPs have a negative impact on efforts to increase screening rates in the US. However, there are very few published studies on PCP-related variations in CRC screening that simultaneously assess screening rates and physicians' delivery of screening, as reported by patients.

Despite the limited literature on PCP-related variation in screening, the existing evidence about differing quality of preventive health care among PCPs [15–18] suggests that similar differences in quality of CRC screening may contribute to underuse of screening in some populations. One study, which included analyses on CRC testing by PCP specialty, found that patients receiving their usual routine care from family/general practitioners were less likely to have had CRC testing than those receiving care from a general internists [18]. However, the findings of that study were limited by the use of insurance claims data alone and the exclusion of fecal occult blood test (FOBT) use in the definition of CRC screening. Thus, there is paucity of published studies that simultaneously addressed both a patient's awareness and use of CRC screening services and a physician's recommendation and delivery of CRC screening through interviews with patients. This would lead to a better understanding if it is clarified whether or not low rates of screening are the result of patient factors such as education or income or a physician's failure to offer screening. Additionally, understanding the variations in screening practices according to PCP specialties will provide insights into some of the reasons behind suboptimal rates of CRC screening in the US. A key question is whether a particular PCP specialty group has achieved the US national CRC screening objectives.

In this study, we compared the CRC screening practices of patients receiving care from FPs to those receiving care from internists. We examined individual's awareness of CRC screening tests and the utilization of such screening, according to whether the patient received their usual medical care from an FP or internist, and related those differences to patient factors.

The data for this study were obtained from the Medicare Current Beneficiary Survey (MCBS) and matched Medicare claims. The design and data collection methods for the MCBS are described in detail previously [19, 20]. In brief, the MCBS is an ongoing annual survey with four-year rotating cohorts of nationally representative samples of Medicare beneficiaries who are interviewed in-person three times a year [19]. We used data on non-institutionalized beneficiaries who were 50-75 years of age at the beginning of 2006, were continuously enrolled in Medicare, participated in the interviews in both 2006 and 2007, and had no personal history of renal disease or CRC in 2006 or 2007. Our study population therefore also included persons 50-64 years of age who were in Medicare as a result of disability. Our upper age cutoff was based on current U.S. Preventive Services Task Force CRC screening guidelines [21, 22]. This study was reviewed by the institutional review board (IRB) of the University of Massachusetts Medical School (Worcester, MA) and was considered exempt from a full IRB review.

Measures of colorectal cancer awareness and screening

In the fall of 2007, MCBS respondents 50 years of age or older, who did not report a history of CRC, were asked whether they had ever had CRC screening test (sigmoidoscopy, colonoscopy, and/or a home FOBT), and if so, the date of the most recent test (as shown in Figure 1.) We defined a nominal screening variable by first considering the receipt of a sigmoidoscopy/colonoscopy within five years of the interview date and then FOBT within one year, in a mutually exclusive manner. We also created a combined outcome of CRC screening defined as receiving colonoscopy or sigmoidoscopy within 5 years and/or FOBT within one year. This approach was based on how questions were asked on the MCBS as described previously [20]. Respondents were also asked if they were ever given an FOBT kit (categorized as yes vs. no), and those who had received an FOBT kit were then asked if they had returned the last kit (categorized as yes vs. no). Patients who had not previously received a kit were asked if they had "ever heard of this home testing kit" (categorized as yes vs. no). In addition to questions on FOBT use, those who had not previously undergone a colonoscopy or sigmoidoscopy were asked if they had "ever heard of" sigmoidoscopy/colonoscopy and if so, whether his/her physician recommended that he/she should have the exam (categorized as yes vs. no). Participants who had previously heard of or received CRC screening were asked: "Before today, did you know that Medicare now helps pay for the cost of screening tests".

This study examined the delivery and receipt of CRC screening among a nationally representative sample of Medicare beneficiaries according to the specialty of their usual PCP. We found that the use of CRC screening tests, defined in our study as having completed FOBT testing within one year or sigmoidoscopy or colonoscopy within 5 years, was lower than the national goal of 70% for both internists and family physicians despite high levels of CRC screening awareness among beneficiaries. We found that patients receiving care from FPs had a lower rate of testing than those receiving care from internists, despite equal or higher awareness of screening tests among FPs' patients. PCP specialty differences were observed for all testing outcomes examined and remained statistically significant even after accounting for socioeconomic status of enrollees and factors related to access to health care. Surprisingly, the magnitude of the differences in use of screening tests was similar across many of the subgroups examined. The stability of the results across multiple outcomes, subgroups and multiple sources of information strengthens our findings.

We found a wide spectrum in terms of testing rates in this population. In 2007, 68% Medicare beneficiaries in higher income groups who received care from internists had undergone testing, which is nearly at the national screening target proposed for 2020. We observed the lowest rate, 34%, among disabled enrollees (50-54 years old) receiving care from FPs, which was 50% lower than the rate for this group of enrollees who had received care from internists. Only about half of eligible patients had received an FOBT kit, with proportionally fewer of FPs' patients receiving one. Less than a fifth of beneficiaries who received care from FPs received a screening colonoscopy recommendation, compared to about a third of those who received care from internists.

To our knowledge, no previous studies have simultaneously examined specialty of usual healthcare provider and patients' receipt of CRC screening in a nationally representative sample. That said, our findings with respect to specialty differences in screening rates are consistent with previous studies showing that among adult PCPs, internists have higher rates of providing preventive care services including cancer screening than FPs [16, 17]. A study using Medicare claims data found that patients receiving care from FPs/GPs were less likely to undergo CRC screening compared to patients of internists [18]. Our study used data from in-person interviews with patients, supplemented with insurance claims, and included analyses on multiple CRC screening outcomes at both the patient and provider levels. This provides a more detailed analysis of patient characteristics, patient knowledge of screening services and use of screening tests based on the primary care specialist seen.

The most plausible reason for PCP specialty differences in screening is that family practitioners are less aggressive about offering CRC testing to their patients. Patients receiving care from FPs were equally knowledgeable about CRC screening and Medicare's benefit for screening, as those receiving care from internists. In fact, compared to internists, a higher percentage of patients receiving care from FPs had heard of colonoscopy, and yet screening-eligible patients of FPs were less likely to have been offered a home FOBT kit or a recommendation for colonoscopy. Our findings suggest that once offered screening, the rates of test completion were similar for both groups of patients. These findings suggest that the lower rates of CRC screening are due, in part, to potentially remediable healthcare provider variations in screening practices and not solely from patient-related factors or systematic refusal of testing by patients. Our results show that eliminating PCP variations in CRC practices has the potential to substantially increase utilization of CRC screening among a diverse group of screening-eligible adults.

There are other possible explanations for the PCP specialty variations found in this study. Some previous studies suggest that FPs may be more likely to provide safety-net care in non-academic and rural settings [17, 18]. This could have a significant impact on recommendation for colonoscopy as FPs may have less access to subspecialist referral networks necessary for colonoscopy. Further, FPs provide a wider scope of services [17], which may lead to more competing and disparate demands in their practices than for internists [13]. However, in this study, the differences persisted even after controlling for patient factors, suggesting that differences in use of screening tests cannot be attributed to differences in the complexity of patients' medical care needs alone.

Organization and style of physician practices, such as tracking systems for monitoring delivery of CRC screening services, may contribute to the differences observed in this study [17, 18]. Such practice-related barriers to providing CRC screening may present greater hurdles for FPs than internists, particularly in rural and underserved areas. Tailored practice-based interventions [25] including reminder systems [26], clinical outreach, programs to monitor disparate care in practices, and the use of preventive medicine specialists, supported by information technology solutions and incentive programs, may increase screening within primary care offices. Mitigating such hurdles has the potential to increase recommendations of CRC screening in primary care practices.

The use of CRC screening tests was lower than the national goal despite high levels of CRC screening awareness irrespective of whether patients received care from internists or family physicians. Our results suggest that potentially remediable variations in practice style or practice environment contributed to the differences in screening rates according to PCP specialty. Our study shows the potential to greatly increase screening rates nationally through greater efforts to increase physicians' offering of CRC screening. However, additional studies, including mixed methods research that capture the experiences of patients in primary care who have not undergone testing, may provide data that can be used to develop interventions to improve delivery of CRC screening in primary care setting.