Despite these recommendations, CRC screening remains low in the U.S., with only 61% of average-risk individuals being current with screening recommendations. In Appalachian Pennsylvania (PA), the CRC screening rate was less, at only 44.4% [5]. Spanning all or a portion of 13 states in the eastern U.S. including Pennsylvania, Appalachia is characterized by high rates of poverty, low health insurance coverage, and limited access to health care. In addition, CRC incidence and mortality rates have increased [6-8].

Receiving a physician recommendation for CRC screening is a primary predictor for patient adherence with screening guidelines [9,10]. Recognizing the pivotal role of the primary care provider, the Task Force on Community Preventive Services (Task Force) reviewed reports of provider-oriented interventions to increase screening for breast, cervical and colorectal cancer [11]. The Task Force found sufficient evidence to recommend two interventions: provider assessment and feedback, and provider reminders (inform provider that client screening is due) and recall (signifies client screening is overdue). Provider assessment and feedback interventions evaluate provider performance in delivering or offering screening to clients (assessment) and present providers with information about their performance in providing screening services (feedback) [10-14]. However, the studies upon which the recommendations related to CRC screening were based only upon FOBT as the screening modality and were relatively dated, having been published in 1986, 1989, and 1990 [15-17].

One interventional strategy to disseminate these evidence-based provider-oriented interventions is academic detailing (AD), a method that transfers knowledge or methods to physicians, nurses or office staff through the visit(s) of a trained educator. Academic detailers can provide methods to improve usage of the targeted service through assessment and feedback, and recalls and reminders. Indeed, physicians reported that AD is more likely to persuade them to implement CRC management guidelines than other strategies [18]. Acceptance of AD by primary care physicians has been generally high; 88% of Belgian physicians who had experience with AD indicated they would welcome similar visits in the future [19].

As a dissemination strategy, AD can be multi-component, which is also recommended by the Task Force, possibly including client-oriented interventions (e.g., client reminders, small media) and interventions to reduce structural barriers (e.g., inconvenient hours/locations, need for multiple practice visits) [14]. However, there are only a few reports of the efficacy of AD to increase cancer screening. In a 2x2 factorial randomized controlled trial in ten urban primary care practices, AD with patient reminders improved CRC screening with endoscopy by 14% [20]. In a second randomized controlled trial of underserved urban participants within eight community health centers, AD significantly improved CRC screening (any screening method) by 16% [21]. Another randomized controlled trial of urban populations served by 264 medical offices found a 7% increase in colonoscopy rates for the AD intervention arm [22]. A group randomized implementation trial in 32 primary care practices found a 4.9% (adjusted) increase in CRC screening (colonoscopy, FS, and FOBT) and 7.9% increase in CRC screening recommendation due to an AD with audit and feedback intervention. For breast cancer screening, a randomized block study of 168 primary care physicians found that AD intervention increased recommendation for mammography and practical breast examinations in an urban community [23]. Finally, a randomized controlled trial of 16 practices found that AD with chart audit and feedback was associated with a significant increase in recommendation for and receipt of breast cancer screening in Oklahoma [24].

We recruited primary care practices (n = 4) in four counties of north-central PA through existing partnerships with ACTION Health Cancer Task Force, a community-based coalition of Northern Appalachia Cancer Network (NACN) and Penn State Ambulatory Care Research Network (PSARN). The NACN is a regional member of the National Cancer Institute (NCI)-funded Appalachia Community Cancer Network, whose goal is to develop, test and disseminate evidence-based interventions that measurably reduce cancer-related health disparities [7,8]. PSARN aims to link primary care research with patient care, medical education, and community service in central PA [25]. Three practices are members of a large multi-practice primary care health network which shares a full electronic medical record (EMR), but are geographically separated, while the fourth practice is affiliated with a major academic health system and uses a paper-based record.

We planned the study using the PRECEDE/PROCEED conceptual model for population health improvement, which guides users through a systematic process to identify factors that influence health and health behaviors [26-29]. The development of the study procedures included key informant interviews of medical office assistants, nursing staff, office administrators, and providers to assess current policies for preventive health care and screening practices at each participating practice. In addition to interviews, we conducted an environmental assessment of the practice, including screening tools, medical record flow sheets, computerized databases, patient education handouts, and patient brochures on CRC screening.

The health care providers, physicians and nurse practitioners, at each practice answered a health care provider survey. The survey instrument collected information on screening practices, follow-up and referral procedures, and characteristics of the practice. A lead nurse at each practice distributed the survey after the practice agreed to participate in the study and again after data collection was completed at each site. The health care provider survey sought to provide the research team with important information about the early detection of CRC, including screening recommendation behavior, opinions about screening, and some general questions about the individual medical practice. Once the survey was completed, the answers were transcribed and saved to a database. Unique identifiers for the respondents were established by the lead nurse of the clinic so surveys before and after the intervention could be linked. The list of identifiers was not shared with the research team providing anonymity for respondents.

Upon completion of the pre-intervention practice assessment, each of the four practices received three separate AD visits that reviewed four different learning modules. The intervention was delivered by a physician-led team including, a senior nurse and PhD-level behavioral health specialist who were trained by the lead physician. The academic detailers covered various topics (Table ​(Table1),1), reinforced standard medical practice, supplied providers and staff with county-specific cancer incidence and mortality data and ACS educational tools (e.g., wall poster, client reminders, scripts).

Outcome measures were collected through medical chart abstraction performed prior to the AD intervention and six-months after the intervention. Abstraction was completed by abstractors hired from the participating practices. Abstracted data were entered directly into NCI's BioInformatics Grid^® (caBIG) through Oracle C3D. Developed to support translational and clinical research in oncology, the caBIG is a nationwide computer-based network "grid" of data-sharing tools and medical data intended to improve cancer care [30]. The abstraction protocol guided the abstractor through each step, including selection of medical records, determination of eligibility, recording screening results, and follow-up care.

Eligibility for medical record abstraction included patients seen in the practice during the previous two months who were aged 50 years or older. Ineligibility was conferred to patients seen for acute reasons, those with symptoms of CRC (e.g., rectal bleeding), and those with a prior diagnosis of CRC. The primary outcome measures were being current with CRC screening recommendation using FOBT, FS, or colonoscopy and having received a CRC screening of FOBT, FS, or colonoscopy during the previous year. We also documented the presence of a recommendation of CRC screening during the previous year as a secondary outcome measure. Information abstracted from the chart included age group in increments of 5 years, family history of CRC, interval since the previous physician visit, and any completed CRC screening tests along with their results. Patient names or other identifiers were not recorded in caBIG.

The prevalence of the outcome measures were calculated by dividing the number of patients who had a CRC screening (or recommendation) by the total number of eligible medical records reviewed. The statistical significance of differences in the estimates of the pre- and post-intervention screening rates were compared using chi square, or Fishers exact tests if the expected cell size was small. We used Wilcoxon tests to examine differences in median time to colonoscopy follow-up.

To assess acceptability and participant experiences in the intervention, a qualitative evaluation with key informant interviews was used to examine intervention processes, including participant perceptions of the program and consistency with programmatic goals. This was accomplished at the completion of chart data collection.

Semi-structured in-depth, face-to-face interviews with physicians, other health care providers and medical office directors were conducted by a trained qualitative researcher who was not a member of the research team, to insure open dialogue with interviewees. The semi-structured interview format sought to elicit specific information about the content and conduct of the intervention, including the acceptability of the intervention to providers and office staff, perceptions of what did and did not work as desired, and ways in which the provision of academic detailing could be improved.

Our pilot work did not show a significant change in screening rates overall. Three reasons may be possible. First, study sites had a relatively high pre-intervention rate of CRC screening. Second, ceiling effects for cancer screening have been described. Mammography screening rates have been reported to have ceiling effects between 90-95%, and CRC screening rates near 60% [31]. Finally, our follow-up chart audits occurred just six months after the intervention, yet had a look-back period of 12 months. In our pilot work, only four practices were recruited, and small sample size may have contributed.

Despite increased number of screenings in the charts audited after the intervention, the percentage of charts with a recommendation for CRC screening decreased. We are uncertain of the reason for decreased recommendation, but it may be that providers omitted documentation for a recommendation because they proceeded to schedule screening tests for their patients [32]. As with any study, random error due to chart selection may also have caused this observation.

While this study was not designed to examine the impact of practice characteristics upon CRC screening, these characteristics are worthy of discussion. First, the impact of AD on CRC screening differed between EMR practices and the paper-based medical record practice. A significant increase (16% to 64%) in records that had documentation of CRC screening currency was observed in the EMR practices but was not observed in the paper record practice. Similarly, when combining screening or recommendation for screening, the practices using an EMR had significantly higher rates combined than the paper record practice. This differential increase may be attributed to AD motivated providers to act on the EMR health maintenance reminders more frequently than before the intervention [33]. Second, practice processes may affect CRC screening rates. For example, the practices using the EMR required a healthcare team member (usually a nurse) to annually review and document age-appropriate disease prevention in the EMR to keep reminders current at the point of care for each patient [34]. Third, patient characteristics may also be a factor in the screening rates. For example, patients at the paper record site were significantly younger cohort in the post-intervention chart review [32,35,36].

Importantly, nearly one-third of completed screening tests in the EMR practices did not have documented results in either pre- or post- intervention review, compared to less than 4% of completed tests being undocumented in the paper record practice. At the transition from paper records to an EMR, paper data is often scanned into a database that does not have searchable data elements which makes searching for information cumbersome. Data abstractors may have overlooked documentation due to the cumbersome nature of these databases. A specified time range may also be employed to convert paper data into digital format. If a colonoscopy was performed before the start date of the specified window, results would not be available within the EMR. Future studies will need to include augmented abstractor training to insure the most accurate chart reviews of archived data.

As with all studies, our study has limitations. First, our study did not include comparison sites to control for secular trends in CRC screening. Second, only four sites participated in our study. Third, our follow-up period was limited to six months, thus inhibiting conclusions regarding the continued impact of the intervention. Finally, we were not able to examine the effect of the individual components of our AD intervention.

The study also has important strengths. First, the study was the first to examine AD to increase CRC screening in a rural population. Second, we developed methods that can be used in future studies. Importantly, we used the caBIG Oracle C3D web data collection tool which facilitated data entry within the practice sites and provided the research team with a data set for analysis.

The common data elements in caBIG will allow the NCI to aggregate information across a variety of research settings [37,38]. Finally, the study identified important areas for future research, including practice characteristics, such as the effects of medical recording systems, and patient characteristics.