BETTER HEALTH: Durham -- protocol for a cluster randomized trial of BETTER in community and public health settings

The BETTER intervention is supported by a training program for prevention practitioners, based on the BETTER tools including baseline and followup patient questionnaires [8], and visual aids for explaining CDPS activities and assisting the patient to set goals for accomplishing CDPS activities as prioritized by the participant [8]. After their training, prevention practitioners discuss their experience with the BETTER intervention and Brief Action Planning among themselves, and with investigators or supervisors. Individuals may apply for certification in Brief Action Planning [10], but there is no method for ongoing quantitative assessment of fidelity to the principles of Brief Action Planning.

The goal of ‘BETTER HEALTH: Durham’ is to improve participation in CDPS actions in low income areas, by adapting the ‘BETTER’ intervention for delivery by the Durham Region Health Department (public health unit), and testing the effectiveness of the adaptation in a cluster randomized trial.

The specific objectives of BETTER HEALTH: Durham are:

(1) To detect a 30% improvement in adherence to evidence-based CDPS actions among individual participants in intervention clusters, 6 months after the BETTER intervention, compared to individual participants in wait-list control clusters, as observed in the first BETTER cluster randomized trial [5]. The 30% increase is measured by the composite BETTER index score and reflects eligible CDPS actions met within six months following enrollment, between the immediate intervention clusters compared to the ‘wait-list’ control clusters (as achieved in the original BETTER cluster randomized trial). All participants will complete an outcome survey six months following the baseline survey; those in ‘immediate intervention’ clusters will have been offered the intervention immediately after completion of the baseline survey whereas those in ‘wait-list’ control clusters will be offered the BETTER intervention after completion of the outcome survey.(2) To adapt and tailor the BETTER intervention to the populations of the eligible low income areas using a community engagement strategy and participatory research methods, including assessing the social determinants of health of participants and linking them to community resources.(3) To evaluate the implementation of the adapted BETTER HEALTH: Durham intervention among participants residing in the eligible low income areas.(4) To conduct integrated knowledge translation with participants residing in eligible low income areas, stakeholders and knowledge users throughout the project and at the end of the study [16].

Census dissemination areas in Durham Region, Ontario, Canada, classified in the lowest quintile of urban median household income, with low cancer screening participation and low access to primary care [1].

Males and females aged 40–64 living in eligible low income areas. Only one person per household will be eligible. Participants will be recruited using community-based strategies specific to the individual clusters, devised using participatory research methods [17]. Informed consent will be sought from individual participants.

Cluster randomized trial, randomly assigning clusters to the BETTER HEALTH: Durham intervention immediately, or to a six-month ‘wait-list control’ prior to the intervention (Table 1).

The sampling frame consists of 16 clusters, which are aggregations of census dissemination areas, identified using the methods described in our prior work [1]. All clusters are categorized in the lowest quintile of urban median household income. The average cluster size is 342 persons in the target age range 40 to 64 years. We will randomly sample 10 clusters to be randomly allocated to immediate intervention or six month wait list control. Random allocation of clusters will be conducted using a random number generator function in the statistical package R, conducted at the level of the cluster using simple random sampling. The complete statistical code will be written in R by the co-principal investigator for biostatistics and implemented simultaneously so that no allocation concealment is necessary.

We will use community-based participatory research (CBPR) principles to guide the adaptation of BETTER. CBPR is a “collaborative research approach that is designed to ensure and establish structures for participation by communities affected by the issue being studied, in all aspects of the research process to improve health and well-being through taking action including social change” [17]. A CBPR approach has improved the quality of interventions in a variety of community settings [17, 18]. We have convened a Community Advisory Committee consisting of a range of stakeholders and residents of low income areas to provide advice on the adaptation of BETTER, recruitment strategies and community engagement throughout the study.

The specific process of adaption of BETTER will follow the ADAPT-ITT model [19]. Stakeholders including prevention practitioners, nurse practitioners, public health professionals, primary care providers, and community stakeholders will be invited to a meeting to learn about BETTER HEALTH: Durham and overall study goals and approach. Subsequently, research team members will conduct small group meetings with members of the public who are potentially eligible for BETTER HEALTH: Durham, to discuss specific needs, and the fit of BETTER HEALTH: Durham to those needs. Extensive notes will be taken at the meetings to document all suggestions. Adaptations of BETTER HEALTH: Durham will be made considering the needs of the community, balanced with fidelity to core elements of BETTER, and discussed in follow up meetings.

Some participants in BETTER HEALTH: Durham will not have a regular PCP, and others who have a regular PCP may have difficulty accessing their PCP. A primary care engagement strategy will identify PCPs willing to accept participants lacking a PCP who wish to have one.

The entry survey will be a modification of the BETTER Health Survey (First Visit) [8], and will be administered to consenting participants as an interview by a research assistant, rather than self-administered, as in the first BETTER trial, because of anticipated low levels of functional literacy. The categories of self-reported income to be collected by the survey will be reset to reflect the focus of low income areas. The survey will include components of the RAND Medical Outcomes Study Social Support Scale items [20], and items pertaining to income security [21] and food security [22]. Responses to these items may inform the approach of the prevention practitioner to the individual participant, and may be associated with the participants’ choices of CDPS goals and their accomplishment.

Eligibility for CDPS actions will be determined exclusively by self-report of previous and current participation in CDPS actions in the survey, unlike in the first BETTER trial, in which eligibility, previous, and current participation were determined from EMRs and the BETTER Health Survey (First Visit) [8]. The maximum number of eligible CDPS actions will be reduced from twenty-eight in the first BETTER trial to seventeen in ‘BETTER HEALTH: Durham’ because of the difference in data sources for determination of eligible CDPS, as demonstrated in Table 2. Those CDPS actions excluded from the first BETTER trial required information from the EMR, for which there is no equivalent self-report.

Survey data will be entered electronically directly to a database application particularly created for BETTER HEALTH: Durham using the REDCAP platform [23] and housed on a secure server the Applied Health Research Centre, St Michael’s Hospital, Toronto. The database application will automatically determine the eligible CDPS actions for each participant based on an adaptation of the BETTER Algorithm [8] and generate an adapted and populated version of the BETTER Prevention Prescription (a list of eligible CDPS actions and recommended periodicity) [8].

At the completion of the baseline survey, all participants in all clusters will be offered educational materials available to the public from the Durham Region Health Department promoting cancer screening, safe alcohol use, tobacco cessation, weight control, healthy eating and physical activity. Participants residing in the immediate intervention clusters will be offered the adapted BETTER intervention immediately after the baseline survey. Participants who choose screening tests as short-term goals will go to their PCPs to have these done, if they have a PCP; for those who do not have a PCP and cannot be linked promptly to one, the prevention practitioner will arrange to have the chosen screening tests facilitated by a nurse practitioner engaged by the study for this purpose.

Six months after the baseline survey interview, the research assistant will re-administer an outcome survey based on a modification of the BETTER Health Survey (Followup Visit) [8] to participants in both intervention and ‘wait list’ control clusters, and enter the responses online. The database application will compute the number of CDPS actions completed for each participant automatically after entry of the six-month data. Participants residing in ‘wait-list’ control clusters will be offered the adapted BETTER intervention immediately after completion of the outcome survey.

Differences in the collection of information and flow of participants between the first BETTER trial and BETTER HEALTH: Durham are summarized in Table 3.

We have adapted the approach taken for the evaluation of BETTER 2 in Newfoundland to the evaluation of the adaptation of BETTER [12]. The key questions to guide the qualitative evaluation are: 1. How was BETTER adapted? 2. What has been the impact of BETTER HEALTH: Durham as perceived by stakeholders? 3. What barriers and enablers of BETTER HEALTH: Durham have been encountered? 4. What are the benefits and disadvantages of BETTER HEALTH: Durham?, and 5) How can the implementation of BETTER HEALTH: Durham be sustained?

We will use a qualitative approach based on grounded theory [24] and the Consolidated Framework for Implementation Research (CFIR) [25] to evaluate the adaptation and implementation of BETTER (Fig. 1). Grounded theory is a well-known qualitative method suited to examining processes, such as the adaptation of BETTER and the implementation of BETTER HEALTH: Durham in the context of a public health unit in low income areas. CFIR was developed from existing implementation frameworks and illustrates interrelationships among five different domains (Fig. 1). Given the complexity of the adaptation and implementation processes for BETTER, all potentially relevant domains will be considered. Constructs for each domain of CFIR have been described which will facilitate interpretation during data coding and analytic processes [25] .

We will collect qualitative data three times during the project: 1) at the beginning of the study to assist with adaptation and start-up issues (Year 1); 2) after the intervention is established to understand perceived enablers and barriers, benefits and disadvantages (Years 2–3), and 3) near the end of the study (Year 4) to explore sustainability and impact. Data collection strategies will include focus groups and one-to-one, semi-structured interviews with a range of participants: 1) members of the public eligible for BETTER who live in low income clusters that will not be participating as well as clusters that will be participating; 2) prevention practitioners who are providing the BETTER intervention; 3) Medical Officer of Health (MOH) and public health unit; 4) primary care providers; and 5) other community stakeholders. Interview guides will be based on the study objectives and revised periodically to seek contrasting and supporting data. Interviews will be recorded and transcribed verbatim and field notes will be created to document non-verbal and contextual information. The number of interviews or focus groups depends on data saturation. Saturation occurs when data categories are dense and no new or relevant data are being observed [26].

Consistent with grounded theory, we will use the constant comparative method for data analysis. Initially, two team members will code approximately 2–3 transcripts in each data collection phases using an editing style of coding [27]. From the codes identified during this process, a preliminary coding guide will be developed and reviewed with all team members. Subsequently, a research assistant will code the remaining transcripts using the coding guide. We will hold periodic analysis meetings with several team members to review the codes, sort codes into categories and identify main themes [25, 28]. Furthermore, team members will create memos that will document emerging relationships among the codes and categories. We will use data management and analysis software (NVivo 10, QSR International). An audit trail including interview summaries and memos will improve trustworthiness through transparent documentation of all major decisions made during data collection and analysis [29]. Involvement of several members of the research team during the analytic process will ensure that identified themes are consistent with coded data. We will triangulate the data from several sources to provide a full description of the themes.

The main outcome is a composite index score, consisting of the ratio (multiplied by 100) of the number of eligible CDPS actions at baseline (denominator) that have been accomplished at follow-up (numerator) according to self-report, calculated at the level of the individual participant [5]. This index is modelled on the Summary Quality Index (SQUID) method, developed by Nietert in order to evaluate the quality of interventions administered in primary care settings [11]. At study enrollment, the eligibility of each participant for CDPS actions will be determined according to health history and health status. Six months later, the number of ‘eligible’ CDPS actions which are met by each participant will be determined [5]. Analysis of the primary outcome (absolute percent increase in eligible CDPS actions completed or engaged in) will account for the correlation among outcomes which may arise from individuals within the same cluster, by implementing a two-level hierarchical regression model [30]. Specifically, a generalized linear random effects regression model will be constructed in which a cluster-specific random effect, arising from a normal distribution, will be included to account for the dependency among outcomes of individuals within the same cluster [31‐34]. The main binary exposure in the hierarchical model will be immediate BETTER compared to ‘wait-list’ control clusters; furthermore any characteristics that were not balanced (between the immediate BETTER and ‘wait-list controls’) from the randomization process will be adjusted for in the hierarchical regression model. In addition to obtaining estimates of the regression parameters, the hierarchical regression model will allow us to investigate the percent of residual or unexplained variation attributable to each level of the hierarchy.

With an equal number of participants in five intervention clusters and five wait-list control clusters (the average number of eligible residents per cluster being 342), a total sample of 120 participants (on average, 12 participants per cluster) will be required to detect an increase of 30% or greater in the composite index score (reflecting additional actions met among the intervention clusters (compared to the control clusters), as observed in the first BETTER cluster randomized trial [5]. The calculation is based on a two-sample comparison of means with 80% power and alpha = 0.05. The calculation accounts for the design effect (correction factor determined as (1 + (m-1)*rho)) arising from the clustered design, with intracluster correlation coefficient = 0.237, as in the original BETTER cluster randomized trial [5]. The ICC is a measure of the relatedness of the clustered data [31].

Integrated and end-of-grant KT activities are central to the adaption and implementation of the BETTER intervention in a community context [16]. We will use the Knowledge to Action process [35] as a guiding framework (Fig. 2). Key stakeholders including the BETTER HEALTH: Durham Community Advisory Committee will be involved in the development of comprehensive integrated and end-of-grant KT plan.