Constructs from the Consolidated Framework for Implementation Research associated with church enrollment and intervention adoption in a national implementation study of a faith-based organizational change intervention

Abstract

Background

Organizational adoption is a key but understudied step in translating evidence-based interventions into practice. The purpose of this study was to report recruitment strategies and factors associated with church enrollment and intervention adoption in a national implementation study of the Faith, Activity, and Nutrition (FAN) program.

Methods

We worked with partners using multiple strategies to disseminate intervention availability. Interested churches completed an online form. To enroll, the church coordinator (FAN coordinator) and pastor completed baseline surveys and then received intervention online training access. We compared enrolled vs. non-enrolled churches on how they heard about the study and church characteristics. We compared intervention-adopting vs. non-adopting churches on Consolidated Framework for Implementation Research (CFIR) constructs using Fisher’s exact tests, χ², or independent sample t-tests and reported differences where p < 0.10, d≥|0.35|, or the difference in percentage points was ≥ 10.

Results

We received 226 interest forms; 107 churches enrolled, and 85 churches adopted the intervention. Faith-based sources were the most, and paid media the least, effective in reaching churches, which were largely from the southeast with a Methodist or Baptist tradition (no differences by enrollment status). Enrolled churches were less likely to have 500 + worshipers and more likely to have attended a study information session than non-enrolled churches. Church (CFIR inner setting) and FAN coordinator characteristics, but not intervention characteristics, were related to intervention adoption.

Conclusion

Partnerships, relationships, and “face time” are important for enrolling churches in evidence-based interventions. Church and church coordinator characteristics are related to intervention adoption. Further work on adoption conceptualization and operationalization is needed.

Background

Dissemination and implementation (D&I) research conducted in community-based settings, as compared to clinical settings, is limited yet critical for improving population health [1]. Several factors make D&I research in community-based settings challenging, including that health program delivery is often not a primary mission of the setting, there are often fewer mandates and incentives for such work, revenue streams to pay for intervention delivery may be limited, and community-based organizations are usually organized in less formal or standardized ways [1]. These factors present challenges in reaching community-based organizations and facilitating an evidence-based intervention’s adoption and implementation. However, community settings are crucial partners in health program delivery as they strongly influence day-to-day behaviors in large populations [1]. Thus, researchers must partner with communities in identifying strategies to overcome these challenges.

The adoption of a given evidence-based intervention is essential for translating research into practice and scaling up the intervention [2]. Adoption is one of eight implementation outcomes in Proctor et al.’s taxonomy and is defined as “…the intention, initial decision, or action to try or employ an innovation or evidence-based practice” [3]. Adoption is also one of five dimensions of the RE-AIM model [4] where it is defined as “the absolute number, proportion, and representativeness of settings and intervention agents (people who deliver the program) who are willing to initiate a program, and why.” Further, CDC’s Knowledge to Action Framework includes the “decision to adopt” as a critical activity in translating evidence-based interventions into practice [2]. Despite the importance of adoption, reviews of interventions conducted in community-based settings such as schools [5] and faith-based organizations [6] indicate that adoption is not a frequently reported indicator. Further, behavior change interventions in areas such as physical activity often show poorer reporting of components related to external validity (e.g., reach, adoption) than internal validity (e.g., effectiveness/efficacy) [7,8,9], perhaps in part because individual-level (versus organizational level) interventions dominate the field.

Faith-based organizations represent a community setting with great potential for promoting population health as they are ubiquitous in communities, they reach populations with marked health disparities (e.g., racial/ethnic minorities, rural populations, older adults), they tend to be trusted sources of information, and many embrace physical health as a dimension of holistic health [1, 10,11,12]. However, large-scale D&I research in this setting is rare, likely due in part to the challenges noted earlier. The purpose of our paper was to report recruitment strategies and factors associated with organization enrollment and intervention adoption in a national implementation study. As such, this paper contributes to multiple gaps in the literature by addressing the following objectives: (1) describe strategies used to disseminate a faith-based, evidence-based intervention designed for Christian churches in a national study, (2) report how churches who completed an interest form heard about the intervention/study and which of these recruitment strategies were most common in churches that enrolled versus did not enroll, (3) examine church characteristics associated with study enrollment, and (4) examine factors that predicted intervention adoption, guided by the Consolidated Framework for Implementation Research (CFIR) [13, 14].

Methods

Intervention

The Faith, Activity, and Nutrition (FAN) intervention was developed using a community-based participatory research process where university and faith partners collaborated in all phases of the research [15]. FAN is an organizational change intervention, designed for Christian churches, based on Cohen et al.’s structural model of health behavior [16]. Church committees, comprised of a church coordinator (FAN coordinator), pastor (optional), and several church members, are trained to implement four core components in their church related to physical activity and healthy eating: (1) increase opportunities, (2) set policies, (3) share messages, and (4) enlist pastor support. They also receive materials to support the implementation of these components, including an Assessment and Planning Guide, a FAN Program Plan document (to plan 12 months of implementation activities), monthly newsletters, 12 months of materials for members (e.g., bulletin inserts and handouts), and technical assistance.

In our three previous studies, church committees were trained to implement FAN via a full-day, in-person training [17,18,19,20]. In the original effectiveness trial conducted in partnership with African Methodist Episcopal churches in South Carolina, the intervention resulted in moderate to large changes in church organizational practices (implementation outcomes) and small but significant improvements in members’ health behaviors [17, 18], leading to its inclusion in the National Cancer Institute’s Evidence-Based Cancer Control Programs [21]. In two more recent D&I studies of FAN, one conducted in a single county among churches with diverse religious denominations and the second with churches from a single denomination throughout South Carolina, we found similar positive outcomes [19, 20].

From 2018 to 2020, we converted the FAN training to an online format and conducted a pilot study of the training [22]. We changed the training delivery mode, not the intervention itself nor the primary implementation strategy (i.e., training church committees). This decision to offer online training resulted from increased interest in FAN from churches within and outside our state and our desire to scale up the intervention efficiently and cost-effectively. In a previous paper, we described the iterative and lengthy process to convert the training delivery mode to online [22]. In brief, we engaged potential end-users and other audiences, including our Community Advisory Board, throughout the process, incorporated best practices for online training and e-learning, hired a design team to ensure a professional product, and incorporated engaging and interactive content throughout. This process led to an eight-lesson training, delivered once per week to committee members who completed the prior week’s lesson, had a satisfactory score (≥ 80%) on a knowledge check, and completed the brief post-lesson feedback survey. The design team used Articulate Storyline software to create the lessons. The lessons and supporting materials were housed in Moodle, an online learning management system. Committee members received a username and password where they logged onto the site to complete lessons, download resources, and post to the discussion board. Our pilot study [22] evaluated the online training with nine individuals from nine different churches across seven states. All participants in the pilot study completed the lessons within the target time and reported highly favorable scores for domains within our evaluation categories of effectiveness as well as design and functionality [22].

Recruitment strategies

Prior to the start of the study, we formed two groups to guide church recruitment and other study activities (see Table 1 for more details). First, our Community Advisory Board, comprised of faith, health, academic, and economic development representatives, met throughout the five-year study via Zoom. They served as advisors for all study activities, including recruitment. Second, we formed a group of translation partners, comprised of faith and health representatives. Their role was more limited to providing ideas for reaching churches and sharing recruitment materials and thus expanded our dissemination reach. Across these two groups, representatives included both long-time and new partners.

Table 1 Types of organizations represented by community advisory board (CAB) and translational partners

Our study priority population was churches with predominantly African American congregations, but we did not limit participation to this racial/ethnic group. Due to the breadth and depth of our recruitment efforts, their duration, and their interconnectedness (e.g., a partner often connected us with a person/organization who might further connect us with another), we cannot describe every person or organization we contacted. Furthermore, for our recruitment efforts that focused on asking partners to share information via their networks, we had no way of tracking distribution or how recipients used or shared materials. Thus, in the following sections, we summarize our recruitment activities according to broad categories and include examples for each. Unless otherwise noted, recruitment activities were conducted from July 2020 to October 2022.

FAN flyer, brochure, and website

In June 2020, in collaboration with our Community Advisory Board and translation partners, we created a recruitment flyer, language to accompany emails with our flyers, and a study website. These partners helped to select images, content, and messages that would resonate with our priority population. For example, they suggested we emphasize that FAN is an established and “tested” program that includes resources, is free and convenient, addresses health concerns, and focuses on health promotion rather than illness or disease. The flyer underwent revisions to keep it “fresh” and relevant (e.g., New Year’s message, the link between health behaviors and COVID-19). The study website included an overview of the intervention and study requirements, a study flyer and brochure, an inspirational video from a pastor, church spotlights from previous studies, and links to complete an interest form (which generated follow-up from our staff) and to sign up for an information session (described later).

Emails to partners

Our initial recruitment efforts (beginning July 2020) focused on disseminating information about FAN and the study via our Community Advisory Board, translation partners, organizations suggested by our board and partners, and other existing networks and partners (faith- and public health-related). We shared the flyer, sample email language, and a link to our website with these partners and asked them to disseminate the materials through their networks. We also shared these materials with churches that participated in our pilot study, with community health advisors from a previous study, and with the Prevention Research Center listserv (i.e., a network of 26 academic centers across the US funded by CDC; our national translation study is funded through this mechanism). Throughout the recruitment period, we repeated these emails. Our emails over the two years often resulted in new connections, expanding our network.

“Cold contact” emails

We searched extensively online to identify faith-based networks or health-related organizations that likely had ties to churches or faith-based networks. In the case of large networks with public email addresses (e.g., Cooperative Extension offices, councils/unions of churches), we sent mass emails with the study flyer. For other contacts, we sent personalized emails about the intervention and study and invited further conversation (e.g., faith denomination wellness, benefits, and pension groups; Black Greek Letter Organizations in the Southeast; predominantly African American seminaries; national public health and chronic disease prevention networks).

Telephone contacts

Partner emails and our “cold contact” emails sometimes resulted in further discussions about FAN and the study. Most “cold contact” organizations that expressed an interest in FAN wished to determine if it was a good fit with their mission and culture before sharing the flyer with their network of churches. We offered these contacts the opportunity to view the online training lessons. Some contacts were highly fruitful. For example, the dean of a seminary requested a phone conversation in response to a “cold contact” email and phone call. During the call, the fit between FAN and the mission of a new international role he was about to play in his denomination became apparent, and we discovered mutual collaborators, which further enhanced trust. As a result, he invited us to author an article for a global missionary magazine and present our study during a Zoom call with a large group of pastors. Other contacts were less fruitful. For example, we had several calls with a representative from a national but smaller denomination. Although there was overlap in our missions and goals, they prioritized pastor (vs. congregational) health promotion and did not want FAN to conflict with other initiatives they had planned.

Presentations

Existing partners and new contacts invited us to present to their network, typically to a group of churches and occasionally to a group of health-related organizations. These presentations allowed us to share information about the study requirements and the enrollment process (for churches) or how they could get the word out via their networks (for organizations). We also hosted 36 online FAN information sessions for churches and religious and health-related organizations. We promoted these sessions through the earlier described networks and on our website, with a link to register. During the information sessions, we delivered a PowerPoint with a video demonstrating the training, and we encouraged questions about the intervention and the study. We followed up via email with churches and organizations who attended the sessions. In addition to these targeted presentations, the study PI included a slide about the study (with the website) in most of her academic and community presentations and asked for assistance in getting the word out. With the exception of several conference presentations by the PI, all other presentations were delivered virtually.

Social media and e-media

We adapted our study flyers (same images and titles but with abbreviated text and a link to our website for more information) for use in social media and e-media recruitment advertisements that we shared with partners willing to distribute via their organization’s social media platforms, post on their websites, or include in their e-newsletters. We also contacted enrolled churches with an active social media presence (e.g., Facebook) to ask them to post our social media recruitment ads on their site. Our Prevention Research Center and School of Public Health shared advertisements on their social media platforms (Facebook and Twitter) and via an listserv and e-newsletter. We conducted a paid Facebook Ad Campaign from June to August 2021, but did not continue this approach due to low yield.

Other paid advertising

We also paid for radio ads on Praise/Gospel stations that reached African American listeners in Jackson, MS, Atlanta, GA, Chicago, IL, and Detroit, MI in May/June of 2022. We selected these cities based on guidance from advertising specialists in our School of Journalism and Mass Communications. In Atlanta and Detroit, the media groups also shared e-blasts and promoted our study on their social media platforms.

Word of mouth

We encouraged enrolled churches to share the flyer with family and friends from other churches and made the recruitment flyer available on our FAN online training site.

Procedures

The University of South Carolina Institutional Review Board reviewed and approved the study protocol and deemed it exempt. We recruited churches in 10 groups so that they would have a cohort of churches with whom they could share ideas and questions. Interested churches completed a brief interest form (online survey accessed via our website). If more than one person from a church completed an interest form, we used the first complete form. In response, the study project coordinator emailed additional information about FAN and study participation. Additional communication occurred via email and telephone to ensure churches understood what participation would entail, to identify their FAN coordinator, and to address questions.

For churches that remained interested in enrolling in the study, and five weeks before we released the first online training lesson to the cohort, we emailed the FAN coordinator a link to the first online baseline survey. The survey asked if the coordinator wished to participate in the study, agreed to complete baseline and 12-month surveys, and agreed to form a church committee for training. This survey also requested permission to contact their pastor. Upon positive endorsement of these questions, we emailed the pastor a link to their first online survey. This brief survey asked if the pastor approved of and supported their church’s participation in FAN and if they agreed to complete baseline and 12-month surveys. These FAN coordinator and pastor surveys included introductory language typical in an informed consent form (e.g., what participation entailed, the voluntary nature of the study, the right not to answer the question(s), the right to discontinue the survey and study at any time, who to contact with questions). After this information was presented, they were asked “Do you wish to take part in the study?” An answer of “yes” indicated online consent to participate. An answer of “no” indicated they did not wish to participate, and the survey was terminated. Written informed consent was not obtained. After completing the first baseline surveys, we sent the FAN coordinator and pastor a link for their second and final baseline survey. Administering the baseline survey in two parts reduced burden for FAN coordinators and pastors in instances where the pastor did not grant approval. When pastor and FAN coordinator surveys were complete, we shared a link to a brief online survey for the remaining church committee members that collected their contact information and basic sociodemographics. We used their name and email address to create an account for the online training. No financial incentives were provided for participation in these church leadership surveys or for completing training. This paper uses data from the interest form and the FAN coordinator baseline surveys.

Before committee members received access to the first online training lesson, they had to post an introduction on the discussion board and watch two brief orientation videos (e.g., explaining how to navigate the site and lessons). We released one lesson each week pending completion of the prior week’s lesson, receipt of a satisfactory score on a 10-item knowledge check, and completion of the lesson feedback survey. Moodle captures all participant logins to the online site. Within each lesson, the learner must click on an arrow to advance to the next frame, and the option is available only after the voice narration for the frame is complete. The final slide includes a trigger to capture the completion of the lesson and navigated the learner to the lesson knowledge check and feedback survey.

Measures

Sources of recruitment

The interest form asked how they heard about FAN. Thirteen options were provided, plus “other” (see Table 2 for a full list; “Newspaper” and “Television” are not listed because we did not use these strategies and they were not endorsed by any respondent). Respondents could choose as many options as applied. If “other” was chosen, they described how they heard about FAN. We coded open-ended responses from the “other” option into categories. If an open-ended response matched one of the provided options, we counted the response for that option rather than coding it to a new category.

Table 2 How enrolled and non-enrolled churches heard about the study

Church characteristics

The interest form collected general church-related characteristics, including church name, denomination, address, race/ethnicity, worship attendance numbers, and presence of an active health ministry. We categorized denominations into faith traditions where possible, based on consultation with a Community Advisory Board member who is a pastor with a doctorate in divinity. For example, AME, AMEZ, UMC, and Methodist were all categorized as “Methodist.” We categorized church address (state) into geographical regions of the US (Southeast, Northeast, Midwest, Southwest, West). We categorized the predominant race/ethnicity of the congregation using a technique from Polson and Dougherty [23]. The respondent estimated the percentage of worshipers who were Hispanic/Latino and the percentage who belonged to various race groups. If a single race or ethnicity comprised ≥ 80% of worshipers, we coded the predominant race/ethnicity of the congregation [23]. The pastor also reported congregation race and ethnicity. In four instances, the pastors’ report led to a different categorization than that reported on the interest form. We used the pastors’ reports for these four discrepancies. Respondents reported how many people attended worship service(s) each week, with response options of 1–19, 20–99, 100–499, or 500 or more. Finally, the respondent reported whether their church had an active health ministry, defined as “a recognized team of people who sponsor regular educational events and experiences that promote well-being.” Response options were yes, no, and maybe.

Consolidated Framework for Implementation Research (CFIR)

The CFIR [13, 14] is a widely-used implementation science framework. Although the CFIR conceptualizes contextual factors (barriers and enablers) that influence implementation of an intervention, we used it to guide our examination of factors that influenced intervention adoption. We included items that mapped onto constructs from the domains of intervention characteristics, inner setting, and characteristics of individual (FAN coordinator). A paper from our previous D&I study described how we identified and adapted CFIR items from the literature for a faith-based setting [20]. In the current study, we prioritized constructs shown to predict adoption [24] and implementation [20]. All CFIR items were rated on a 4-point Likert scale: strongly disagree (1), disagree (2), agree (3), and strongly agree (4). As needed, we recoded items so that a higher score always indicated a more favorable response. An option of “refuse to answer” was included for most questions.

The specific CFIR items and their source are included in Table 3. Within the intervention characteristics domain, we assessed the constructs of complexity (2 items: ease of providing healthy food choices and physical activity opportunities) and cost (4 items: financial costs and time costs to provide healthy food choices and physical activity opportunities). Within the inner setting domain, we assessed the constructs of church culture (composite of 2 items: FAN coordinator’s rating of their pastor’s sense of personal responsibility for improving congregant health and openness to changes in church practices), networks and communication (composite of 3 items: FAN coordinator’s rating of their pastor’s working relationships with lay leaders in the church, pastor sharing information and knowledge with the church, and pastor involving members when decisions are made), tension for change (1 item: acceptance of new ideas in their church), compatibility (1 item: providing healthy food choices and opportunities for physical activity matches their church priorities), relative priority (1 item: health ministry is as important as the spiritual ministry in their church), and congregant needs (2 items: serving healthy food choices and providing physical activity opportunities would be well-received by their members). Structural characteristics within the inner setting included the presence of a health ministry (yes/no), recent experience with health promotion (led or co-led health promotion efforts in past year, yes/no), church size (number of worshipers each week: 1–19, 20–99, 100–499, 500+), predominant race and ethnicity of the congregation (reported earlier), and duration of pastor tenure at church (< 1 year, 1–5 years, 6–10 years, 11–20 years, > 20 years). Finally, within the characteristics of individual domain, we assessed self-efficacy (4 items: confidence as well as skills needed to help their church provide healthy food choices and physical activity), beliefs about the intervention (2 items: serving healthy food choices and providing physical activity would be valuable to their church), perceived benefits (their church would benefit from healthier food choices and opportunities for physical activity), and identification with the organization (2 items: desire to perform to the best of their ability for their church and sense of commitment to their church). Additional characteristics (personal attributes of FAN coordinator) included duration of church membership (< 1 year, 1–5 years, 6–10 years, 11–20 years, > 20 years), age (years), gender, self-rated health (1 = poor, 5 = excellent), and body mass index (BMI, kg/m², computed from self-reported height and weight).

Table 3 FAN coordinator’s baseline ratings for adopting (n = 85) and non-adopting (n = 22) churches, guided by the CFIR

Table 4 Characteristics of enrolled and non-enrolled churches

Church enrollment

If both the pastor and FAN coordinator completed their two baseline surveys within the five-week enrollment period, their church was considered to have enrolled in the study.

Church adoption

Commonly cited definitions of adoption lack clarity and often conflate the intention or commitment to implement the evidence-based intervention with some level of implementation. We opted, a priori, to conceptualize adoption as distinct from and occurring prior to implementation. Consistent with RE-AIM and other writings [25,26,27], we conceptualized church adoption as involving two levels of decisions within the organization: (1) the organizational leader’s decision to try the intervention and (2) the decision to attempt intervention implementation by key organizational staff who will ultimately be responsible for intervention implementation. Accordingly, we operationalized church organizational adoption as being present if two conditions were met: (1) the pastor provided their approval and support for their church to take part in FAN, and (2) the FAN coordinator completed the online training (i.e., introductory videos and seven or eight of the eight lessons as the first seven lessons covered all of the content that would be necessary to implement FAN). Because all pastors of enrolled churches provided their approval and support for their church to participate in FAN, this condition of adoption was met for all enrolled churches. Thus, FAN coordinators’ training completion was the only condition that varied across enrolled churches.

Statistical analyses

We used statistical tests appropriate with the variable structure to examine differences in recruitment source and church characteristic across enrollment status. We used Fisher’s exact tests to examine whether each reported recruitment source differed by enrollment status (i.e., 2 × 2 tables). We also compared whether churches that enrolled were more likely to report more than one recruitment source than those that did not. To examine church characteristics associated with study enrollment, we used Fisher’s exact tests or χ² tests of significance to compare churches that enrolled in the study with churches that did not enroll by region of the country, denomination, race/ethnicity, worship attendance numbers, presence of an active health ministry (yes/no/maybe), and participation in a FAN information session.

Several approaches were used to examine characteristics that predicted intervention adoption among enrolled churches. Because many of the CFIR constructs were measured with single-item Likert scales, we used the nonparametric Wilcoxon rank-sum test to determine if the distributions of responses differed for adopting and non-adopting churches. For most items, FAN coordinators “agreed” or “strongly agreed” with the CFIR items (i.e., positive ratings). Thus, we present the percentage of FAN coordinators who “strongly agreed” with the CFIR items in adopting and non-adopting churches to simplify the presentation of the data. For dichotomous variables (e.g., yes/no options, gender), we report p-values from a Fisher’s exact test. For continuous variables (e.g., age, BMI), we report p-values from independent sample t-tests. Due to the small sample size of non-adopting churches and the exploratory nature of the analyses, we report patterns of differences that appear to be meaningful. Thus, we noted variables where the difference in percentages between groups is ≥ 10, where p < 0.10, or where Cohen’s d (continuous variables) >|0.35| (i.e., between a small and medium effect size) [28].

Results

Church enrollment and adoption

Figure 1 shows the study flow chart. We received 226 completed and unduplicated interest forms from churches. Of these, 107 churches enrolled (47.3%) and 119 (52.6%) did not. The most common reason for non-enrollment was that the church was no longer interested in the program (64/119; 54%). Consistent with the advice to form a committee of two to five members who would undergo the online training, 442 people across the 107 churches registered for the online training (4.1 committee members/church). Of the 107 enrolled churches, 85 were classified as adopting churches (79.4%) and 22 (20.6%) as non-adopting churches. Among the adopting churches, 83 (97.6%) FAN coordinators completed all eight lessons, and 2 (2.4%) completed seven lessons.

Fig. 1

Study flow chart. A church was considered to have enrolled if both the pastor and FAN coordinator completed their two baseline surveys within the five-week enrollment period. A church was considered to have adopted the intervention if the pastor provided their approval and support for their church to take part in FAN (all enrolled church pastors did so), and the FAN coordinator completed the online training (i.e., introductory videos and seven or eight of the eight lessons)

Recruitment sources for churches that enrolled in the study versus churches that expressed interest but did not enroll

As shown in Table 2, how churches heard about FAN did not differ markedly by enrollment status. Among enrolled churches, an email or newsletter from a faith-based group (not their church or denomination) (19.6%) followed by an email or newsletter from their church or denomination (18.7%) were the most commonly reported recruitment sources from the list of options. For churches that did not enroll, the most common source was also an email or newsletter from their church or denomination (22.7%), but three other options tied for second: email or newsletter from a faith-based group (not their church or denomination) (10.9%), faith-based event or conference (not their church) (10.9%), and social media from their church (10.9%). Notably, only one church (non-enrolled) reported hearing about FAN via a radio advertisement. “Other” was the most common way churches heard about FAN, representing 34% of all churches. The most frequently cited “other” sources were from a faith-based health community (a group focused on health but with a religious affiliation or identity) (8.4% enrolled, 11.8% non-enrolled), a faith community (a church or group primarily focused on faith/religion) (10.3% enrolled, 5.0% non-enrolled), or a health coalition (6.5% enrolled, 4.2% non-enrolled).

Church characteristics associated with enrolling in the study

Our recruitment efforts reached churches from 33 states across the US. Enrolled churches came from 23 states, and non-enrolled churches came from 27 states. South Carolina and Georgia were the most frequently represented in both groups. As shown in Table 4, the Southeast region was most represented in both groups (75.7% enrolled, 69.0% non-enrolled), followed by the Northeast (10.3% enrolled, 11.2% non-enrolled). Most enrolled and non-enrolled churches had predominantly African American/Black congregations, with the proportion significantly higher in enrolled churches (74.8% vs. 59.1%, p = 0.03). Enrolled churches represented 21 unique denominations (17 for non-enrolled churches). Denominations/faith traditions were similar in enrolled and not-enrolled churches, with Methodist (48.6% enrolled, 44.2% non-enrolled) and Baptist (23.4% enrolled, 27.0% non-enrolled) traditions most common. A Pentecostal/Holiness tradition appeared more highly represented in enrolled (8.4%) than non-enrolled (1.7%) churches, and while 5% of non-enrolled churches were Catholic, no Catholic churches enrolled in the study. Most churches, whether enrolled or not, were smaller (< 100 worshipers/week) in size; large (500+) churches were more common in non-enrolled than enrolled churches (7.8% vs. 0.9%, p = 0.08). About half of the enrolled and non-enrolled churches reported having an active health ministry. Finally, while a small proportion of churches who completed an interest form attended a FAN information session hosted by the research team, enrolled churches were significantly more likely than non-enrolled churches to have attended a session (21.5% vs. 9.2%, p = 0.01).

Predictors of church adoption

Table 3 shows comparisons between adopting and non-adopting churches for FAN coordinator ratings. Within the intervention characteristics domain, FAN coordinators from adopting churches rated the complexity, time costs, and financial costs similar to FAN coordinators from non-adopting churches. Compared to other CFIR items (described next), however, responses to these items were more variable (but not different by adoption status), with more FAN coordinators choosing “strongly disagree” or “disagree” (costs were reverse scored): 12.5% for healthy eating complexity, 20.0% for physical activity complexity, 54.4% for healthy eating financial costs, 31.1% for physical activity financial costs, 24.0% for healthy eating time costs, and 35.6% for physical activity time costs (data not shown in table).

Within the inner setting domain, FAN coordinators of adopting churches rated their church culture more favorably than non-adopting churches (d = 0.50). They also differed on structural characteristics: adopting churches were more likely to have a health ministry (54% vs. 43%), were more likely to have a FAN coordinator who led or co-led health promotion efforts in the past year (60% vs. 41%), had more worshipers (100 + people, 35% vs. 23%), were less likely to have predominantly African American congregations (72% vs. 86%), and had pastors who had been serving their church for a longer duration (6 + years, 53% vs. 41%). However, with respect to congregant needs, FAN coordinators from adopting churches were less likely to “strongly agree” that serving healthy food choices (27% vs. 41%) and providing physical activity opportunities (18% vs. 36%, p = 0.02) would be well-received by their members (note that ≤ 10% of adopting and non-adopting FAN coordinators “disagreed” or “strongly disagreed” with these statements). Differences were not evident (statistically, difference in percentages, or d) for the constructs of networks and communication, tension for change, compatibility, or relative priority.

Within the domain of characteristics of individual, adopting FAN coordinators were more likely to “strongly agree” than non-adopting FAN coordinators that providing healthy food choices would be valuable for their church (beliefs; 63% vs. 50%) and would benefit their church (perceived benefits; 70% vs. 50%), that providing physical activity opportunities would benefit their church (perceived benefits; 65% vs. 45%), and that they wanted to perform to the best of their ability for their church (identification with organization; 78% vs. 68%). In terms of personal attributes, FAN coordinators from adopting churches were church members for a longer duration (> 20 years, 61% vs. 48%), were older (p < 0.001, d = 0.83), and had a lower BMI (d=-0.40). Differences were not evident for the constructs of self-efficacy, beliefs (for physical activity), identification with organization (sense of commitment to church), or the personal attributes of gender and self-rated health (Table 3).

Discussion

Adoption is the first necessary decision for an organization to put an evidence-based program into practice [2], yet few community-based studies report organizational adoption and factors associated with it. For example, a review of 76 studies that examined organizational characteristics associated with the adoption and/or implementation of innovations found that only 34% assessed adoption, 49% were conducted in healthcare settings, 86% were cross-sectional, and only 55% used a conceptual framework to guide the study [29]. Reviews have shown low rates of reporting organizational adoption in school-based interventions for physical activity [5], physical activity interventions for children and youth [30, 31], interventions to reduce sedentary behavior in worksites [32], and faith-based physical activity and healthy eating interventions [6]. This paper addresses gaps in the literature by reporting recruitment strategies and constructs from the CFIR associated with church enrollment and intervention adoption in a national implementation study. As part of the larger implementation study, each FAN coordinator also completed an implementation survey 12 months after the completion of the online training. These findings, which will be critical in evaluating the effectiveness of our online training, will be reported in a future paper.

Strategies used to disseminate FAN and differences in recruitment for enrolled and non-enrolled churches

Churches have diverse organizational structures. Some churches are organized into broader networks (typically denominations or church councils), whereas others have no or only loose affiliations with other churches (e.g., independent and non-denominational). Thus, the process of scaling up an evidence-based program in this setting is inherently complex because it is difficult to share its availability efficiently. We used multiple strategies to disseminate FAN availability to churches in this national study. Initially, we relied on help from long-standing and newly formed relationships and partnerships with health- and faith-based organizations. These partners had the advantage of being known and trusted in their networks, which facilitated not only the recruitment of churches but also introductions to other organizational leaders who subsequently shared the availability of FAN through their networks. Much time was spent talking with prospective partners and giving presentations to organizational leaders, groups, and churches. Although we also contacted health- and faith-based networks for which we did not have established relationships, these attempts yielded mixed outcomes. Paid advertising (radio and Facebook) proved ineffective in recruiting churches, at least within our relatively short campaigns. It is possible that longer campaign durations are needed. Our experiences underscore the importance of trust and relationships in recruiting communities and organizations to participate in programs and research [33, 34] and the need for “face time,” which occurred primarily via Zoom calls due to the national reach of our study. Although it would have been interesting to map the spread of our reach through health- and faith-based networks, we were not able to track how each organization shared information. Although we asked how church leaders heard about the program, it was not possible to trace responses back to specific organizations (unless this was noted in response to the “other” prompt) due to multiple levels of sharing information. We did not find differences in how enrolled and non-enrolled churches heard about the intervention/study. Churches most commonly heard about the intervention/study from faith-based networks via emails/newsletters, faith-based events or conferences, social media, and faith-based contacts (including faith-based contacts with a health focus). Health-related groups more generally were also cited, but less often. To enroll large numbers of churches, it is essential to allocate sufficient time and staff resources to enact a variety of strategies through multiple channels while capitalizing on existing social and organizational networks. Accordingly, we found that partnerships, relationships, and ‘face time’ were important for enrolling churches in this evidence-based intervention.

Church characteristics associated with study enrollment

Of all churches that completed an interest form, 47% (n = 107) ultimately enrolled in the study. We surpassed our study goal of enrolling 100 churches. By design, most churches had predominately African American congregations, and this was truer for churches that enrolled than those that did not. This finding is consistent with two of our previous studies, where participation was greater among churches with predominantly African American congregations relative to white congregations [19, 20], which may reflect a more holistic approach of the Black church that includes addressing issues of health and social welfare inequities [35]. We also found that more of the enrolled churches had attended one of our information sessions. It could be that churches who attended information sessions had greater initial interest than those who did not, or it could be that the information session helped address questions about the program and fit with their church. There was also a trend for smaller churches to enroll, particularly relative to churches with more than 500 weekly worshipers. Larger churches may already have health-promotion resources (e.g., gyms) available and greater congregational capacity. Both enrolled and non-enrolled churches were primarily located in the Southeast, although churches from 23 states representing all regions of the country enrolled. Some of our long-standing partners who helped promote the study were located in the Southeast, and the proportion of African Americans, our priority population, is also greater in the Southeast. Lastly, while 21 different denominations were represented in the study, churches from Methodist and Baptist traditions accounted for the majority of enrolled and non-enrolled churches. Our Community Advisory Board included African Methodist Episcopal and United Methodist Church leaders, facilitating great reach into these denominations. Further, the National Baptist Convention is the largest African American religious organization in the US.

Predictors of church adoption

Lastly, this study examined factors that predicted organizational adoption of the intervention among churches enrolled in the study. Definitions of adoption often conflate the decision or commitment to try the evidence-based intervention with the actual implementation of the intervention. We conceptualized adoption a priori to reflect the organizational decision to try the intervention. Few faith-based interventions focus on organizational change and report organizational adoption [6]. In studies that have reported adoption, church adoption ranges from 12 to 100% [19, 24, 34, 36,37,38,39,40], although definitions of adoption and denominators used vary greatly. For example, some studies use the number of churches invited to participate as the denominator (e.g., [40]), whereas others use all churches in a designated target area as the denominator (e.g., [19, 24]). Our study was different still in that we did not have a sampling frame of churches, nor did we know the number of potential churches we might have reached with our flyers – we relied on partners to share information about the intervention/study with their networks. Therefore, we operationalized adoption as pastor agreement to participate in the study/intervention and FAN coordinator completion of at least seven of the eight lessons. Nearly 80% of churches that enrolled in the study adopted FAN based on this definition. The number of training lessons completed was fairly strict – for example, we could have simply based it on the FAN coordinator completing any training.

A strength of our study is that we used a conceptual framework (the CFIR [12, 13]) to guide our selection of constructs that might predict organizational adoption. We studied the domains of intervention characteristics, inner setting, and characteristics of the individual. It is important to note that the CFIR is an implementation (not adoption) framework, and its utility for predicting adoption has received little scrutiny. Perhaps not surprisingly, FAN coordinators rated most CFIR items very favorably, regardless of adoption status. All FAN coordinators had expressed interest in FAN, completed baseline surveys, and had pastors who supported participation and completed baseline surveys. Thus, they were a highly interested group. The CRUZA study [40], which also examined whether CFIR items varied by participation status among churches in a cancer control program, found only one CFIR item that differentiated these churches (change commitment). Like our study, they found that participating and non-participating parishes rated most CFIR items favorably. Because we were underpowered to detect differences between adopting and non-adopting churches (i.e., we had only 22 non-adopting churches), we used a more liberal p-value (p < 0.10) and we also reported differences that appeared meaningful (small to medium effect size or difference of 10% points between groups). Although there were very few differences between adopting and non-adopting churches that reached statistical significance, we found seemingly meaningful differences between adopting and non-adopting churches in a number of constructs within the CFIR domains of inner setting and characteristics of the individual (FAN coordinator).

Previous qualitative studies conducted in the faith-based setting have shown that barriers to adoption or participation in health programs include lack of time, lack of financial resources, competing church events, lack of motivation/interest from members, lack of knowledge and self-efficacy related to the health promotion topic, and lack of support from leaders, while facilitators have included positive views of the program, leadership support, having a champion, mission compatibility, interest in health, having a health ministry, and learning about the program from a trusted source [34, 36, 41,42,43,44,45,46]. In contrast to these studies, we did not find that the intervention characteristics of time costs, financial costs, or complexity differentiated adopting from non-adopting churches. FAN was developed using a community-based participatory research approach that engaged church partners to design an intervention that had optimal characteristics for adoption, implementation, and sustainability [15]. Churches in our study were required to rate the complexity and costs (monetary and time) of providing healthy foods and physical activity opportunities “if they decided to do so.” Although these constructs had more variability in responses than others (i.e., less weighted toward “strongly agree” responses), they did not differ by adoption status. It will be interesting to see if these perceptions change after training and implementation attempts, and whether they predict actual intervention implementation. As described in the following paragraphs, some of our findings were consistent with these previous qualitative studies, with some exceptions.

Within the inner setting domain of the CFIR, we found that church culture and structural characteristics differentiated adopting from non-adopting churches. FAN coordinators of adopting churches reported a church culture where the pastor valued congregational health and practice changes. This finding is consistent with our results from an earlier study (different sample of churches): a positive culture facilitated adoption in an analysis that used qualitative comparative analysis to identify combinations of CFIR-derived church and pastor characteristics associated with FAN adoption [24]. Most structural characteristics showed patterns of differences between adopting and non-adopting churches in the expected direction. Adopting churches were more likely than non-adopting churches to have an active health ministry, a FAN coordinator with health promotion experience, a pastor with longer duration, and worship attendance greater than 100 people (most churches in both groups had fewer than 100 worshipers). However, somewhat unexpected and opposite to our enrollment findings and findings from our previous studies [19, 20], non-adopting churches were more likely to have predominantly African American congregations (the majority of churches in both groups had predominantly African American congregations). In contrast to an earlier evaluation of FAN, where the belief that serving healthy food choices and providing physical activity opportunities would be well-received by their congregations was positively associated with 12- and 24-month implementation [20, 47], we found that non-adopting churches held this belief more strongly in the current study. Tension for change, compatibility with priorities, and relative priority were similar in adopting and nonadopting churches.

Within the characteristics of individual domain (FAN coordinator), several constructs had patterns of associations consistent with the previously described qualitative studies. Adopting churches had stronger favorable beliefs than non-adopting churches about the potential value and benefit of healthy food and physical activity opportunities for their church, and they reported stronger identification with their church in terms of performing to the best of their ability. Other personal attributes differed: FAN coordinators from adopting churches, compared to non-adopting churches, were older (and thus perhaps less likely to be working), longer-time members of their churches, and had a lower mean BMI. Based on the previously cited studies, we expected that the self-efficacy of FAN coordinators would differentiate adopting from non-adopting churches. However, we did not observe this finding, although we recognize that this construct may be important for predicting intervention implementation.

Study limitations and contributions

There are several limitations to this study. Most significant is the small number of non-adopting churches, which limited power to detect differences between adopting and non-adopting churches. Thus, we focused more on identifying patterns of meaningful differences between adopting and non-adopting churches by using effect sizes and differences in percentages. Results should be viewed cautiously, however, particularly given the number of comparisons we made. Second, we could not measure how many churches were exposed to our recruitment materials because we used a network approach to reach churches. While this is a limitation, one could argue that this approach is needed to scale up interventions in this setting. It would be inefficient and costly to reach out to individual churches. Third, although we attempted to disseminate the availability of our study/intervention across the U.S., the majority of churches we reached were in the southeastern U.S. and were from a Methodist or Baptist tradition. Religiosity is higher in the Southeast, and health disparities are marked, but nonetheless, strategies to reach other regions and denominations more effectively are needed. Lastly, FAN was designed for Christian churches and thus this study did not encompass religious organizations and communities more broadly. Although the organizational change components of FAN likely apply across organizational settings and faith traditions, we would need to revise and tailor the content for broader use.

Conclusions

This study contributes to the literature as it is the only implementation study of which we are aware attempting to scale up a faith-based organizational intervention at a national level. Churches have broad reach in many communities, and many incorporate health and wellness into their larger ministry [1, 10,11,12], making them important partners in improving population health and contributing to eliminating health disparities [10, 35, 48,49,50,51]. Adoption is a critical step in scaling up evidence-based interventions, but working through networks is inherently complex, and guidance in the literature regarding conceptualizing adoption and understanding its influences is lacking. Further work on adoption conceptualization and operationalization is needed. This study provides valuable information about strategies to successfully recruit churches, underscoring the importance of partnerships, trust, relationships, and “face time” in outreach efforts. We also reported enrollment and adoption percentages in the study and associated factors. Adoption is under-reported in studies and even fewer report factors that predict study enrollment and intervention adoption at the organizational level. This study highlighted constructs from the inner setting and characteristics of the individual (FAN coordinator) domains of the CFIR that appeared to differentiate adopting and non-adopting churches despite generally positive ratings from all churches.