The HPRN is a practice-based research network that covers the 16 counties of eastern Colorado. Of these counties, 10 are designated "frontier" (or less than 6 people per square mile), and the remaining 6 are "rural". The entire HPRN consists of 55 primary care practices, 16 hospitals, approximately 150 providers, and the communities that reside in this region. The 3 communities involved with this particular evaluation included 2 smaller communities (populations provided below), each with 1 primary care practice each. The third town is the largest community in the area and is home to 6 practices, including private solo practices and a hospital-based clinic. Ranching and farming are dominant sources of income and permeate the culture of this region. Residents living in the 3 counties included in this evaluation are older than the state as a whole, with a median age of 39 versus 34 for the state and a percent of the population age 65 and older of 16% versus 10% [10].

The HPRN launched a community-based, multi-component colon cancer prevention intervention in the spring of 2006. "Testing to Prevent Colon Cancer in Rural Colorado" aims to increase CRC screening behaviors, knowledge, and attitudes in rural northeastern Colorado. The intervention consisted of an awareness and educational campaign that encouraged local residents to talk to their providers about colon cancer testing options, targeting residents age 50 and older.

Community-based participatory research (CBPR) methods were used to develop, implement, and test the intervention messages and materials. While the definition of community may vary by project, CBPR research involves a collaborative partnership with a group to increase the relevance of the research to the community. The HPRN uses CBPR approach regularly, having formed a Community Advisory Council (C.A.C.) in 2003 to help to guide the research it conducts. The C.A.C. consists of 9 residents of rural northeast Colorado, including farmers, teachers, ranchers, a home health visitor, and a retired administrator. Specifically for this intervention, the C.A.C. was expanded to include 2 local physicians, 2 public health workers, and a hospital administrator from northeast Colorado. The resulting group, named the Joint Planning Committee (JPC), designed the intervention's main messages, materials, and implementation strategies.

The JPC developed a 4-point message to address key catalysts of behavior change, including relevance, education, facilitation/encouragement, and action. The main messages are below.

1. Colon cancer is the 2^nd leading cause of cancer death in the U.S.

2. Colon cancer is preventable.

3. Testing is worth it.

4. Talk to your doctor today.

The program's 4 main messages were incorporated into the intervention's materials, a set of 8 mostly bilingual components. Table 1 provides a description of each component and implementation. The JPC selected program components that could be implemented in each intervention community and that linked community members back to their local primary care providers. Methods were selected that tapped into the communication culture of rural communities, including the use of local community members in program components, local newspapers and adaptations of familiar small print materials (communication methods engrained into every day life in rural towns), and local organizations that are very common and valued in rural communities. The JPC helped identify local residents to have their photos taken for the palm card series and series of ads in the local newspapers, to share their personal stories in local newspapers, and to co-present community talks with a local health care provider. All local newspapers agreed to run the 3 newspaper components of the intervention. Local residents and clinic staff were also recruited to distribute the small print materials, such as the palm cards and farm auction flyers, and recorded the number of copies left at each location. All primary care practices in the intervention communities agreed to distribute the "got polyps?" travel mugs to residents who redeemed a palm card or newspaper ad for a mug at the clinic.

A random digit dial (RDD) telephone survey was conducted to estimate exposure to the community-based CRC prevention intervention in the target population and to determine if exposure increased intentions to talk to a doctor about CRC screening and to get tested.

The survey was conducted with a stratified random sample of residents living in the first 3 communities and surrounding areas to receive the intervention 3 months after the intervention had been implemented. Participants were sampled by age group (40 – 49 years of age and ≥ 50 years of age) and stratified by intervention community using zip codes. Two towns are substantially smaller than the third. The smaller towns' populations were over-sampled to produce more reliable estimates.

A Colorado market research firm conducted interviews for 1 week in July 2006 using 3 sample lists from Survey Sampling International: 1) household phone numbers from 3 communities having at least 1 adult age 50 or older, 2) household phone numbers from 3 communities having at least 1 adult age 40–49, and 3) household phone numbers for respondents living in surrounding areas within each county with at least 1 adult age 40 or older. Respondents aged 40–49 were included in this study since they are approaching the screening-eligible age of 50 or could be eligible for screening based on family history. A small sample from the surrounding areas was included to provide preliminary estimates of the reach of the intervention outside of the 3 primary communities. Age-eligible respondents with the most recent birthday were selected within each household.

Information about exposure to the intervention during the previous 3 months was collected. Exposure questions included both recall and recognition. Respondents were first asked to freely recall sources of colon cancer prevention information that they had seen, heard, or read in the past 3 months (recall). Respondents were then read a description of each intervention component not freely recalled and asked if they had seen, heard, or read the component (recognition). To estimate the depth of exposure to individual components, the number of exposures to each component was also recorded.

Respondents reporting exposure to any type of colon cancer prevention information were asked questions about intended behavior change as a result of this information. A 5-point Likert scale assessed how likely respondents were to talk to their doctors about CRC screening and how likely they were to get screened. Demographic questions included age, gender, ethnicity, and race and if they had been tested for colon cancer in the past 5 years. This question did not specify type of screening procedure. A copy of the survey is available at: .

Descriptive analyses were performed for the demographic variables age, gender, race and ethnicity, geographic location, and having been tested for colon cancer in the past 5 years. Chi-squared tests were performed to determine general associations between the demographic variables and exposure ("exposed" and "not exposed"). Respondents recalling or recognizing at least one of the 8 intervention components were considered exposed. Respondents who did not report any sources of colon cancer information or who exclusively reported CRC screening materials other than our 8 intervention components were considered "not exposed". Frequencies were run on the number of personal stories, newspaper ads, and medical articles read. Total and individual intervention component exposure rates were calculated.

A general linear model was performed to determine multivariate associations between exposure to the intervention and intended CRC screening behaviors. CRC screening behaviors included the likelihood to talk with a doctor about being tested and the likelihood to get tested for colon cancer. CRC screening behaviors were run as a continuous outcome variable, ranging from 1 = "strongly disagree" to 5 = "strongly agree". Exposure was defined by an ordinal aggregate variable ranging from 0 (exclusively reported CRC screening materials other than the intervention components) to 6 (reported at least 6 intervention components) and is referred to as "dose". Since the components were designed to be implemented as a whole intervention, this analysis measured the effectiveness of the intervention collectively. Models adjusted for age, gender, past 5-year testing status, race and ethnicity, and intervention dose. To adjust for over-sampling, responses were weighted by the inverse of the sampling fraction of each town. Weighted estimates are reported. Interaction between intervention dose and past 5-year testing status was also adjusted in each model. Prior to the final analysis, the ordinal exposure variable was tested for linear, quadratic, and cubic trends for each outcome. A p-value of < .05 was considered statistically significant.

All analyses were performed using SAS version 9.1. The survey instrument and methodology was approved by the Combined Institutional Review Board at the University of Colorado Denver and Health Sciences Center.

Of the 460 respondents, 68% reported exposure to at least one intervention component. Females were more likely to report exposure to the intervention materials than were males (74% vs. 63%, p = .02). Respondents who reported having been tested in the past 5 years were more likely than those not reporting testing in the past 5 years to have been exposed to the materials (78% vs. 58%, p < .001). The percent of the total respondents reporting exposure to each intervention component is displayed in Figure 1. The 3 newspaper components exhibited the most reach.

Based on these results, additional analysis was conducted to determine how many respondents saw only the newspaper component(s), a combination of components, or only non-newspaper components (Table 3). Of the total sample, 30% read at least one newspaper component only, 30% saw some combination of newspaper and other intervention components, and 8% saw other intervention components only. Overall, 60% of the sample was exposed to at least one of the newspaper components, and 38% saw at least one of the small-scale print materials or community talk.

Among respondents exposed to the intervention, the average number of unique components recalled or recognized was 2.8. Of the total sample, only 32% reported no exposures. Comparable proportions of respondents were exposed to 1, 2, 3, and 4 components (17%, 16%, 13%, and 12% respectively). Approximately 8% of the respondents reported exposure to 5–8 intervention components.

Multivariate analyses determined an increasing linear trend between the number of intervention items seen and the intent to "talk to a doctor about colon cancer testing" and to "get tested". Figures 2 and 3 display the adjusted rate of increase for each outcome by intervention dose. Results for intentions did not differ by age, gender, or race, including those respondents age 40–49. Respondents who had not been tested in the past 5 years reported lower intentions to talk to a doctor about colon cancer and to get tested for colon cancer than their tested counterparts at low intervention doses. However, as intervention dose increased, intention to talk to a doctor about colon cancer increased at a significantly faster rate in respondents who had not been tested than in those who had been tested in the past 5 years (p = .025; Figure 2). As shown in Figure 3, respondents who had not been tested in the past 5 years reported overall lower intentions to get tested compared with those who had been tested (p < .001). However, intentions to get tested increased at a significantly similar rate for both groups (p < .001).

Limitations of this study include the use of self-reported data. Although the standard method for this type of community-based survey, some error in product recall or recognition may have occurred. We cannot assess exposure in people who do not have access to a phone or who use cell phones as their household phone. However, we estimate that more than 90% of the target population in this region has a landline telephone. Due to the nature of random digit dial telephone surveys, we do not have information about non-responders and are unable to determine how results may be biased as a result. This intervention was designed to be administered as a whole unit, with pieces complimenting each other or reaching certain groups. Thus, we did not analyze the effect of individual components on CRC screening intentions. Finally, while the intervention materials were largely bilingual, this phone survey was conducted with English-speakers only. However, the proportion of Spanish-speaking only community members in our targeted age range is small. The vast majority of Spanish-speaking only people in this region are under age 40.