Evaluation of the Families SHARE workbook: an educational tool outlining disease risk and healthy guidelines to reduce risk of heart disease, diabetes, breast cancer and colorectal cancer

The first step in creating the Families SHARE workbook was to develop an accurate, yet simple, algorithm for assessing increased common disease risk due to FHH. The four diseases included in the workbook are among the 15 leading causes of death in the United States [28] and were selected for having both familial and modifiable behavioral risk factors. While there are a multitude of complicated algorithms used to assess disease risk [29‐31], these algorithms may not be easily implemented within a community-based public health intervention and interpreted by participants [32]. Thus, this first step focused only on FHH based risk, constructed from a review of the literature and practitioner guidelines for stratifying screening recommendations in clinical practice.

A preliminary literature search yielded numerous FHH-inclusive risk assessment models for heart disease, type 2 diabetes, breast cancer, and colorectal cancer. The language and statistics used to communicate disease risk varied across these models, as did the criteria for risk assessments bearing the same label. For example, Snyder et al. ascribe a “high risk” of breast cancer to individuals with at least one first-degree relative (FDR) diagnosed under the age of 50 [33], while Warner et al. consider the same FHH to warrant only “moderate risk” [34]. Other models eschew verbal descriptions and instead convey risk in odds ratios or risk points, generally referring to an increased risk [35].

Thus, it was decided that the risk algorithm developed for the Families SHARE workbook should incorporate the “least common denominator” of risk criteria while being stringent enough to preclude overestimation of risk. To determine these criteria, a systematic literature search was conducted throughout the Academic Search Premier and PubMed databases. The terms “‘family history’ and risk and (assessment or algorithm or calculation) and (breast cancer or colo* cancer or diabetes or heart disease or myocardial infarction or coronary artery disease)” yielded 1,329 results published between January, 1982 and April, 2011 (see Fig. 2).

A manual search was then conducted through articles’ bibliographies to account for sources not captured by the original search terms, yielding another 20 articles for review. In total, 65 articles were included for review that 1) offered an original set of criteria for assessing disease risk based on FHH only or controlling for demographic characteristics; or 2) revised an existing set of criteria for assessing disease risk based on FHH only or controlling for demographic characteristics. Articles in which family history contributed points towards a total risk score were excluded from the review [35‐41]. Similarly, risk assessments that considered FHH in a multivariate model including clinical or behavioral risk factors (e.g. BMI, breast density, biomarkers, or lifestyle factors) were not synthesized within the review. Twenty three of the 65 articles featured categorical risk language, while the remainder focused on continuous risk variables (e.g., odds ratios, relative risks, hazard ratios) [42‐84]. Risk criteria and risk assessment language were cataloged from these articles, and similar terms were then rephrased to facilitate grouping, e.g. “one FDR relative with CRC” [85] and “family history of colorectal cancer: any FDR” became “affected FDR” [77], a factor associated with various categorical expressions including “high risk,” “increased risk,” and “moderate risk,” depending on the source paper (Additional file 1).

The risk criteria and risk assessment language indicated an increased risk based on the same criteria for each of the four diseases: subjects are ascribed “increased risk” for any of the four diseases if they have at least one affected FDR or two affected second-degree relatives (SDRs) with a specific disease. This “increased risk” category generally corresponds to ratios or relative risks at or above 2.0 [42, 44, 45, 47‐49, 52, 53, 55, 57, 59, 60, 63‐65, 68‐70, 72‐74, 76, 77, 79, 83]. Thus, Families SHARE adopted this risk algorithm based on this aggregation of risk criteria and risk language, as well as the recommendations of two genetic counseling experts. This increased risk label is consistent with increased frequency of screening in clinical practice [86‐88].

In addition to the risk algorithms, information that would facilitate conversation about risk and promote encouragement of risk-reducing strategies was also included [89]. Specifically, as presented in Table 1, workbook content maps to fundamental concepts within the Health Belief Model (HBM) [90]. To improve knowledge regarding family risk of common, complex disease, a personalized three-generation health pedigree was provided with the Families SHARE workbook which includes information on how to interpret key components of the pedigree, along with an example. The risk algorithm provided users with a way to update their perceived susceptibility of disease, with a focus on increased risk versus population-based risk. Each disease is defined and a list of behavioral risk factors is provided. Finally, to encourage positive health behaviors, the workbook presented information intended to increase the perceived benefits of and cueing behavioral actions aimed at disease prevention or early detection. The workbook highlighted demonstrated, modifiable risk-reducing strategies that have been identified as leading causes of death in the United States – namely, smoking behavior, poor diet, physical inactivity, and alcohol use [91]. Such factors have been shown to be associated with the focal diseases presented in the Families SHARE workbook [2, 92‐96].

Messaging was developed based on a review of the literature, government resources (e.g CDC, NIH, AHRQ), and expert websites (e.g., Mayo Clinic). All content was presented at a Grade 8 reading level (Flesch-Kincaid). The first version of the workbook was sent to experts in health literacy and genetics education for review and comment. Following initial revision, materials were distributed to a sample of key stakeholders (mothers of young children) from the community for evaluation.

The workbook materials were further refined based on the responses from key stakeholders in two Phases. This evaluation was approved by the Institutional Review Board of the National Human Genome Research Institute (NHGRI: 12-HG-0023) and registered in ClinicalTrials.gov (NCT01498276). The procedures and results for these two Phases of evaluation are detailed below.

Demographic characteristics, including age, marital status, race/ethnicity, household income and education were assessed at baseline.

Self-reported height and weight were assessed at baseline and converted into Body Mass Index (BMI). Additionally, self-reported lifestyle behaviors were assessed at baseline, including vigorous and moderate physical activity, cups of fruit and vegetables eaten each day, and smoking behavior [97, 98]. These items were converted to assess the percentage of the sample meeting Centers for Disease Control and Prevention (CDC) recommendations [99].

In a free response format, respondents indicated elements of the workbook that were confusing and offered suggestions for improvement.

At follow-up, participants completed a series of questions assessing workbook usability and understanding. Usability questions asked whether participants were able to use the algorithms for themselves and other family members. Understanding questions asked participants to indicate on a 7-point scale their understanding of the workbook overall and their understanding of individual components of the workbook (e.g. sample pedigree, personal pedigree, disease risk worksheets, etc.), with 1 representing “I did not understand at all”, three representing “I understand a little”, five representing “I mostly understood”, and seven representing “I completely understood.” A 7-point scale was used to maximize variability and reliability in responses [100]; in addition, intermediary responses were provided to ease translation of response options.

At baseline and follow-up, participants were asked to indicate: “From the following list, are there any lifestyle behaviors that you would like to improve?” The list included: limit alcohol consumption, stop smoking, increase fruit and vegetable consumption, increase fiber consumption, increase physical activity, none of the above, other – free response. In this report, we focus on dietary behaviors and physical activity; a dichotomous variable was created indicating whether the participant intended to change the selected behavior. For each indicated behavior, participants were asked to rate their confidence in making changes within the next year on a 7-point scale, where one is “not at all confident” and seven is “very confident.” These items were mapped to the lifestyle messages presented in the workbook; a 7-point scale was used to maximize variability and reliability based on guidelines put forth by Krosnick and Presser [100]. The intentions and confidence items were combined into one measure for each behavior, whereby those indicating no intention to change a particular behavior were assigned a confidence score of 0, and others assigned the value selected for their confidence rating.

Survey data were analyzed using SPSS and R. Descriptive measures, including means, standard deviations, and ranges, were computed for sample characteristics, shifts in confidence to modify behavior, understanding workbook components, and utilization of the workbook for self and other family members. Where appropriate, paired t-tests were computed to evaluate differences between baseline and follow-up assessments for each Phase; hypothesis testing was conducted using a type 1 error rate of .05. A content analysis was conducted to aggregate free response data.

Focus group audio recordings were transcribed and investigated using thematic analysis, based in an interpretative phenomenological framework that is extensively used in health psychology research [101]. This framework was utilized to understand the lived experience of the participants and adopts a shared understanding from the perspective of the researcher and the interviewer [102]. As emergent themes were identified, the transcripts were continually reviewed and similar concepts were grouped into superordinate themes [101]. Overarching super-ordinate themes emerged from the identified component themes through an iterative process, and data saturation was reached. Prior theory served as a resource for interpretation of themes, and component and super-ordinate themes were discussed amongst authors and confirmed against the transcript data. Two researchers performed thematic analysis separately using NVivo (version 9.2) and strong inter-rater reliability was found, with substantial Kappa values (0.77) and 97.92 % agreement averaged across the themes.

Table 3 describes the samples from both Phases. There were no significant differences between the two samples on demographic or health behavior characteristics. The majority was between the ages of 31-40, white, married, had completed post-graduate education, and had annual household income that exceeded $100,000. In terms of health status and associated behaviors, participants were largely healthy. The majority met recommendations for daily moderate physical activity and did not smoke. Approximately half met daily fruit and vegetable intake recommendations.

Through open-ended questions, as well as feedback from focus groups, we identified several key areas for change based on participant confusion, requests for additional information, or requests for clarity on presented information. These recommended changes, as well as the consequent workbook modifications to address recommendations, are summarized in detail in Table 4.

The main themes that emerged from focus groups included 1) confusion about the purpose of the workbook, 2) the need for interactive worksheets, 3) definition of risk and becoming aware of risk, 4) reducing generic information, 5) focusing on children’s disease risk and behaviors, 6) aesthetics, and 7) application of workbook.

The detailed feedback collected from participants through interviews and focus groups engendered substantial revisions to the workbook between Phase 1 and Phase 2. Here and in Table 4, we highlight some of the key revisions motivated by the insights we synthesized from this feedback. First, the introduction page was thoroughly reworked to provide participants with greater context about the research study and the purpose of the workbook (Theme 1). Participants were also encouraged to use this workbook as a launching point to other critical actions, such as sharing this information with family members and a physician, and visiting the study website to access risk evaluation worksheets for other family members, such as children (Theme 5).

In the body of the first version of the workbook, participants were provided a sample pedigree populated with hypothetical FHH information, to facilitate understanding of how to navigate and interpret FHH-based risk. In response to participant feedback, the revised sample pedigree exercise is considerably shorter and more straightforward, and delineates first- and second-degree relatives with both text and color to define these essential concepts more clearly (Theme 6). The disease risk evaluations also evolved substantially between Phase 1 and Phase 2 (see Fig. 4). Rather than clustering all four disease risk evaluations on a single page, the revised version of the workbook dedicates two full pages to each disease, allowing participants to go directly to the disease most salient to them. This new organization scheme also afforded space to define each disease clearly and provide disease specific risk factor and health screening information (Theme 3). The process of calculating risk became simpler and more interactive, simply prompting participants to count and record the number of first- and second-degree relatives who have certain diagnoses and directing those participants at increased risk to contact a physician (Theme 2).

To facilitate participants building mental models of how health behaviors influence disease risk, we included a relevant behavior and screening message below each risk evaluation. For the final section of the workbook, dedicated to providing participants with health behavior and screening information, we replaced generic information with more concrete, creative information to make it more actionable and empowering (Theme 4). For example, the revised version of the message to be physically active included information on the impacts of physical activity on the body, such as strengthening bones and muscles and lowering blood pressure.

Sixty-seven percent of the participants in the Phase 2 evaluation, in contrast to 32 % in Phase 1, had no recommendations for workbook modifications to improve understanding, demonstrating significant improvement based on Phase 1 feedback (p = .004). Most recommendations for improvement did not converge on specific changes, but rather represented a single participant’s recommendation. Examples include simplifying the disease fact sheets (n = 1), providing more disease facts (n = 1), providing more background regarding why the four conditions are the focal point (n = 1) and why FHH is important (n = 1). Three participants recommended a more quantified risk assessment, with one wanting integration of lifestyle habits as a potential adjustment to the risk assessment. Participants requested more tips on how to implement both lifestyle and screening recommendations, particularly when at increased risk of disease (n = 4). Additionally, participants requested more information to facilitate family communication (n = 1) and a child version of the packet (n = 1).

The main themes emerging from the Phase 2 focus group included: 1) disease interactive worksheets, 2) focus on children’s disease risk and behaviors, and 3) sharing with family.

Table 5 provides summary statistics related to understanding of the workbook components and engagement in workbook concepts across the two phases. Overall, participants indicated high levels of understanding of workbook components. In Phase 1, lowest levels of understanding were indicated for the sample pedigree (M = 5.89, SD = 1.49); following revision, Phase 2 participants indicated increased understanding of the sample pedigree (M = 6.61, SD = 0.90). The ranges and standard deviations of responses were much narrower and the means were higher in the Phase 2 evaluation, suggesting improvement in understanding across all workbook components.

The majority indicated that they were able use the risk algorithm to calculate their own risk and the risk of at least one family member during both evaluation Phases. While about a third of participants indicated gathering FHH information from their family or their child’s father during both evaluations, there was a slight reduction in collecting FHH information from family and a slight increase in collecting FHH information from their child’s father for Phase 2 participants. Almost 20 % discussed FHH based disease risk with their family at Phase 2 compared to 6 % during Phase 1, a 3-fold increase.

Table 6 demonstrates a general trend in increased intention and confidence to change behavior between baseline and follow-up assessment for both evaluation Phases. Workbook content appeared to consistently increase intention and confidence to improve dietary behavior. However, intention and confidence to improve physical activity was not affected by the Families SHARE workbook. Given that baseline confidence was highest for physical activity, this null effect for increased intention and confidence to improve physical activity may reflect a ceiling effect.

The Families SHARE workbook is a key step forward in understanding how to better make disease risk information accessible and actionable for families. However, the homogeneity of the stakeholder samples engaged in the development of the workbook potentially limits the usability of the workbook and generalizability of the findings to more diverse populations. The majority of participants had completed a bachelors or post-graduate degree and represented a higher income population. As previously discussed, reducing and simplifying the text throughout the workbook was a key revision to the workbook between Phase 1 and Phase 2, as well as defining technical terms, such as first- and second-degree relatives. However, the materials will need further evaluation if anticipating use in lower literacy populations, or with participants from under-resourced communities who may face greater barriers in adopting the recommended behaviors. Additionally, the workbook was evaluated by a primarily white sample; cultural context may be important, and thus the materials may need to be culturally tailored for other populations. Moreover, although research suggests that mothers may be optimally positioned for genomics health educator interventions [19, 20, 113], men and family members in older generations (e.g. grandparents) may also hold optimal positions in some families [14]. Thus, understanding and usability in these constituencies may warrant future investigation. Finally, the literature search informing the risk criteria and assessment language used in the workbook was relatively narrow, and as such may have missed relevant papers.