An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Lynn J. Andreae, PhD, MPH, Susan J. Andreae, PhD, MPH, Andrea L. Cherrington, MD, MPH, Joshua S. Richman, PhD, MD, and Monika M. Safford, MD.
Author Information and AffiliationsThe effects of medication nonadherence are especially profound in remote, economically depressed communities with a high burden of chronic diseases, such as the rural Alabama Black Belt region. We built on our previous work in this region to rigorously test a community member-delivered intervention designed in partnership with community peer coaches.
With our community partners, using qualitative research methods, we developed a medication adherence intervention delivered by trained community members (aim 1) and then tested this intervention in a cluster-randomized trial of individuals with diabetes seeking help with adherence to their medications (aim 2).
We first conducted qualitative research with experienced peer coaches and community members to learn their perspectives on diabetes and the medications used to treat it. Using this information, the Corbin and Strauss framework of the lived experience of illness, Bandura's social cognitive theory, and adult learning theory, we developed an 11-session telephone-delivered diabetes self-management intervention that stressed medication adherence, assessed barriers to adherence, and involved strategizing to overcome these barriers. The intervention, delivered by community peer coaches, also provided information on healthy eating, physical activity, and stress reduction, and encouraged communication with health care providers. We then conducted a cluster-randomized trial, using towns as clusters, to test the effectiveness of this intervention. The intervention was deployed by trained peer coaches who resided in the same communities as the participants. We collected baseline and 6-month follow-up data in the communities where participants were recruited. The primary outcomes were self-reported medication adherence and measures of hemoglobin A1c, blood pressure, low-density lipoprotein cholesterol, and body mass index. Secondary outcomes were quality of life (QOL), medication beliefs, and self-efficacy to use medications.
We recruited a total of 473 participants, 403 of whom completed follow-up (85.4% retention), which was within the design specifications for the study. The mean age of the trial population was 57.2 years, 78.2% were women, 90.6% were African American, 69.3% reported an annual income of <$20 000, 26.7% were employed, and 43.8% were taking insulin. In the control arm, 239 (89%) participants completed the study, and in the intervention arm, 164 (81%) completed the study. The intervention dose was high, with 81.8% of intervention participants completing all 11 sessions of the program. Self-reported medication adherence improved more in the intervention arm than in the control arm (P < .0001), but the other primary outcomes did not differ significantly between the trial arms. QOL improved similarly across both trial arms, but beliefs about the need for medications, concerns about medications, and medication use self-efficacy improved more in intervention than in control participants. Satisfaction with the program was high.
This intervention was highly engaging and improved self-reported medication adherence and self-efficacy, but it did not improve glycemic control or other physiologic parameters among mostly African American individuals desiring help with their diabetes medications and living in a remote, economically disadvantaged rural region.
Real-world challenges created delays in data collection, which may have influenced the study results. The self-reported availability of healthy foods was limited, possibly limiting dietary changes recommended in the intervention, such as increasing the consumption of fresh fruits and vegetables. The intervention was entirely community based and without a clinical component; thus, medication titration to achieve better glycemic control was not included as part of the intervention and may have limited the intervention's impact on physiologic parameters. The fact that the study was restricted to the Black Belt setting may limit the generalizability of the findings.
Medication nonadherence is both common and costly. Clinical trials have demonstrated that medications that lower blood glucose, lipids, and blood pressure (BP) can lower risks for blindness, kidney disease, amputation, stroke, and heart attack.1-3 However, adherence to these medications is suboptimal; as many as half of diabetic patients do not take recommended medications as directed,4 and a recent meta-analysis of studies examining adherence to oral diabetes agents estimated a pooled proportion of patients who were adherent (where adherence is defined by a medication possession ratio of ≥80%) to be 67.9%.5 In addition to negative health outcomes, medication nonadherence contributes to substantial economic burden. For example, Egede and colleagues6 examined medication adherence using medication possession ratios from Veterans Affairs pharmacy data for 740 195 veterans from 2002 to 2006; they estimated that improving adherence to 80% in nonadherent individuals would have resulted in a cost savings of $993 679 348 over the 5-year period, and improving adherence to 100% would have saved $1 158 009 119.6 The annual cost of nonoptimized medication use, including medication nonadherence, was estimated to be $290 billion in 2008,6,7 and a more recent analysis, published in 2018 by Watanabe and colleagues, estimated the annual cost of drug-related morbidity and mortality from nonoptimized medication use to be $528.4 billion.8
Medication nonadherence is especially problematic in hard-to-reach populations heavily burdened with chronic diseases like diabetes, such as residents of the rural Black Belt region, whose residents face the triple threat of being minorities, poor, and rural.9-11 The Black Belt is a region with dark soil amenable to agriculture where cotton plantations and slaves were in greatest concentration. Initially named for the color of the soil, since the Civil War, the name has taken on political meaning and is used with pride by its residents.12 The Black Belt runs in an arc across east central Texas to Maryland; in Alabama, it is a 200-mile-wide belt across the southern half of the state. There is a substantial mismatch of need and resources in this region (Table 1), and the resulting disparities in outcomes are evident; for example, the 2017 age-adjusted diabetes mortality rate was 35.6 per 100 000 for Black Americans compared with 15.5 per 100 000 for White Americans.13 Nonadherence is common in this population, as shown in Table 2, which contrasts participants of a past trial and Black participants with diabetes in a national epidemiology study.14,15 Other studies have reported lower medication adherence in minority patients with diabetes than in White patients.16-18
Characteristics of Target Communities in Rural Alabama Black Belt Counties.
Black Individuals With Diabetes From the REGARDS (National, 2003-2007) and ENCOURAGE (Alabama Black Belt, 2010) Studies.
Although the literature has identified some factors that may influence medication adherence, such as age, race, out-of-pocket costs, insulin use, depression, health literacy, and health beliefs,20,21 there is limited understanding of the most impactful barriers in this population and no data on effective strategies to improve the situation. Decades of research on how to improve nonadherence have yielded only modest changes in medication adherence. Since the 2008 publication of a Cochrane review of medication adherence interventions, which concluded that the strategies used up to that point had not been very effective and that innovative programs were needed to assist adherence to long-term therapy,22 recent reviews have shown that developing effective interventions for medication adherence remains challenging. A 2017 systematic review and meta-analysis by Conn and Ruppar revealed that interventions to date have yielded only modest effects on adherence.23 For diabetes specifically, a review of studies published between 2007 and 2014 concerning adherence to glucose-lowering agents showed that adherence rates had remained unchanged since 2007.20
To improve medication adherence, we undertook a fundamental re-examination of the biomedical model within which adherence is conceptualized. This model views medications as biochemical substances that modify diseased biological processes, improving disease outcomes. From this perspective, failing to take medications that have been proven effective is irrational. However, focusing the benefits of medications on disease outcomes may not resonate with patients. Some patients may not have fully accepted their diagnosis, in which case taking medication disrupts their sense of identity and life trajectory. Other patients may deal with numerous competing demands and view self-care, including medication adherence, as hindering rather than facilitating the many kinds of identity-relevant work that they undertake throughout the day, which also decreases the likelihood that individuals will adhere consistently to long-term therapies. Failing to respond to such patient perspectives may be a reason why adherence interventions have demonstrated such modest success to date. An alternative perspective is a biopsychosocial framework based on the work of Corbin and Strauss24 and of Charmaz.25 The interrelated triad of body-biography-conceptions (BBC) of self has been called the BBC chain to emphasize the interrelatedness of each element (Figure 1). A chronic illness diagnosis requires work of reconceptualizing ourselves in our various life roles, with attempts to preserve our identity and functioning. This work can be stressful, with some individuals never establishing equilibrium and resorting to denying the existence of the illness, sometimes even years after diagnosis. This stress may play a role in medication nonadherence, yet it is rarely acknowledged in medication adherence interventions.26-28 Corbin and Strauss'24 Chronic Illness Trajectory Framework stresses the importance of helping individuals come to terms with their illness and frames self-care behaviors, including medication adherence, in terms of maintaining or rebuilding their sense of identity and personal narrative; this framework also calls for assisting patients with integrating their illness-related tasks into everyday life tasks.29-31 We hypothesized that an intervention built on this patient-focused framework might facilitate changes in medication-taking behavior that would be maintained long term.
BBC Chain.
Using key concepts from Corbin and Strauss' framework, Bandura's social cognitive theory, and adult learning theory, this program used peer storytelling and self-management support from peer coaches. Storytelling by fellow community members can facilitate behavior change through modeling; Bandura's social cognitive theory posits that behavior is influenced by watching others and the consequences of their behavior.32,33 In addition, hearing stories of “people like me with problems like mine” may assist individuals in coming to terms with their illness and change beliefs and norms regarding health-related behaviors.34 Viewers of peers telling stories often develop para-social relationships with the storytellers, “a sense of friendship, attraction, and involvement with the person or character,” achieving “homophily,” or a feeling of relatedness to the storyteller.35-38 Peer storytelling had a remarkable effect on lowering BP in an urban setting, but peer storytelling has not been examined for diabetes or in rural settings.39,40 Peers can also be trained as interventionists.41-52 They receive training, including motivational interviewing, to coach other community members with diabetes on how to improve self-management within the context of their own lives, helping them overcome challenges within their community (Figure 2). Interacting with a live peer coach is a potent experience because participants have discussions with someone like them, facilitating the process of internalizing the DVD content and stories they heard and working on an action plan to overcome their barriers to medication adherence. Peer support is being more widely evaluated for diabetes53,54 but, to our knowledge, is not within the Corbin and Strauss framework to improve medication adherence.53-56
Conceptual Model of CHWs and Their Roles.
Building on ongoing partnerships with people living in Alabama's Black Belt, we collaboratively developed a peer-delivered telephone intervention guided by the Corbin and Strauss framework, drawing on social cognitive theory and adult learning theory. We aimed to recruit 500 individuals with diabetes who reported medication nonadherence into a cluster-randomized trial of this intervention. The trial's primary outcomes were patient-reported medication adherence and physiologic measures, and secondary outcomes were quality of life (QOL), medication beliefs, and self-efficacy. The aims of the proposed study were as follows:
Aim 1: With our community partners, using qualitative research methods, build on already developed culturally tailored education material to develop the medication adherence intervention. The intervention will consist of educational DVDs with integrated storytelling about how community members accepted their disease and overcame barriers to medication adherence, plus one-on-one telephone peer coaching. Activities include conducting focus groups with patients, creating the DVDs and the coaching intervention protocol, training peer coaches, and pilot testing.
Aim 2: Conduct a cluster-randomized trial with 500 individuals with type 2 diabetes and medication nonadherence. The trial will compare the effect of usual care and the intervention on medication adherence and physiologic risk factors, including hemoglobin A1c (HbA1c), BP, and low-density lipoprotein cholesterol (LDL-C) as primary outcomes, and on QOL and self-efficacy as secondary outcomes.
Stakeholders who were engaged in the program included community members with diabetes and representatives from organizations that worked with our patient population, such as churches, nonprofit organizations, businesses, primary care offices, and health departments. We used 5 main strategies to engage stakeholders.
Two individuals from 2 communities located in the Black Belt were hired as community coordinators. As well as being longtime residents of their communities, the community coordinators had worked with the research team on 2 previous studies and therefore were knowledgeable about the research process. A third community member was hired as part of the study team as a data collector. As members of the study team, the community coordinators and the data collector provided day-to-day guidance and feedback on all aspects of the study protocol, from intervention development and participant recruitment and retention methods to the selection of data collection instruments. In addition to guidance at weekly meetings, they connected us with other key stakeholders and representatives from their communities. For example, early during the intervention development process, the community coordinators and data collector engaged a group of 14 individuals from their communities as study recruiters. The community recruiters and peer coaches from previous studies were engaged as needed to discuss specific questions about the study protocols and intervention development. Later during the intervention recruitment phase, recruiters worked closely with the community coordinators to refer potential study participants to the onsite staff in Birmingham, Alabama to conduct eligibility screening.
Early in the contract period, we conducted 3 discussion groups with members of a diabetes support group in Carrollton, Alabama, to solicit feedback on the diabetes medication educational video and diabetes medication messages to ensure cultural concordance.
We held 4 community coalition meetings that provided a chance for patients to meet the research team and develop trust. The 4 meetings had a similar structure, with a theme selected by patients, breakout sessions and activities led by community members, educational content, and feedback on intervention components. The first meeting was held in partnership with the 11th Annual Black Belt Institute Meeting in Camden, Alabama, and hosted by the West Alabama Community Health Improvement League and the University of Alabama at Birmingham (UAB) Center for Community Health. The 76 attendees included patients and representatives from various health clinics, churches, the Cooperative Extension, the Wilcox County Chamber of Commerce and Board of Education, the Linden Health Department, and numerous community organizations (Community Action, Alabama Tombigbee, Vredenburgh Outreach Center, Bama Kids, and Youth on the Move). The second meeting was held in Livingston, Alabama, and attended by 96 participants, including patients with diabetes, and included presentations by the West Alabama Area Health Education Center, the Livingston Health and Wellness Education Center, and the Cooperative Extension. The third meeting was held in Camden, Alabama, on September 24, 2015, in partnership with the 12th Annual Black Belt Institute Meeting with the West Alabama Community Health Improvement League and UAB Center for Community Health and was attended by 114 community members, including patients with diabetes. Our final meeting was held on April 23, 2016, in Livingston, Alabama, and was attended by 60 community members, including patients with diabetes.
This is described in the aim 1 subsection in the Methods section.
Finally, patients and organizations reached out to us with requests to provide health education at health fairs held at community churches. Our peer supporters led most of these activities, building a reciprocal trusting relationship with our communities.
During trial implementation, the relationships we developed with organizations during intervention development helped us identify and gain entry to community events for recruitment and data collection activities and develop recruitment and retention strategies and materials that positively impacted retention (Appendix H, Table 1). For example, the retention plan was modified to include mailing postcards every other month after receiving feedback from community members that telephone numbers often change for our targeted population, and mailing addresses are more stable. In addition, recruitment and retention scripts for data collection visits emphasized the individual who referred the patient to the study and the names of our community coordinators (full-time research staff members who lived in the community) so that patients were more comfortable with research staff coming into their homes for data collection. The data collection protocol was also impacted. After pretesting the survey assessment with community members, we shortened the instrument so that it could be administered in 30 minutes. Finally, program materials and study protocols were culturally concordant, as reflected in the high satisfaction ratings after program completion.
This project was designed in collaboration with our community member partners and built on a 5-year partnership of community-engaged research on diabetes peer coaching interventions. We tested the hypothesis that an intervention designed within the Corbin and Strauss framework and developed with community members improves self-reported medication adherence and health outcomes compared with usual care.
In aim 1, with our community partners, we used qualitative research methods to collaboratively develop the medication adherence intervention delivered by trained community members; in aim 2, we tested this intervention in a cluster-randomized trial of 500 individuals with diabetes seeking help with their medications.
Since 2008, before this study, we had conducted 2 trials of diabetes interventions in partnership with Black Belt communities. The ENCOURAGE trial (2008-2011)19 recruited 424 (target, N = 400) individuals with diabetes (75% African Americans) and engaged peer supporters. Responding to a call from our peer supporters, we developed the Living Healthy trial57 that tested community member-delivered interventions to overcome pain as a barrier to exercise in diabetes. The present study builds on a research partnership that includes a tested and durable collaborative approach to community-partnered research.
In June 2017, due to slow accrual of participants in rural areas, we received approval from the sponsor (PCORI) and expanded enrollment to clinics that serve low-income Black populations in the Birmingham area. The burden of chronic disease and difficulty accessing health care are similar to those faced by Black Belt residents, including transportation challenges and lack of availability of healthy foods. Recruitment efforts in this area focused on attending health fairs and posting flyers at community locations such as libraries and churches. The main focus of recruitment was in the county's safety net clinic, in which 63% of the clinic's 11 500 patients have an annual income <200% of the federal poverty level, and almost all are African American; 10% have Medicaid, 23% have Medicare, and the remainder are covered by the county's indigent care program, which includes pharmacy benefits as well as ambulatory services. A second site was the UAB Department of Family Medicine's clinic, which serves a similar patient base, with fewer uninsured patients but a similar demographic and income profile. Patients seen at both clinics are therefore vulnerable but still have access to pharmacy services required for the study, which focused on medication adherence.
We conducted focus groups, semistructured interviews, and nominal group discussions with community members to develop the intervention. We started with an existing diabetes self-management program delivered by peer coaches58 based on social cognitive theory with educational content focusing on diabetes basics, healthy eating, physical activity, and stress management.19,32,59 In that program, intervention participants viewed educational videos, followed by structured telephone sessions with their peer coach that included setting goals.57 Participants monitored behavioral goals daily and watched the next session's video before their next telephone meeting with the peer coach. We retained those elements of this intervention that our peer coaches felt were particularly engaging and helpful.
The intervention was developed over 18 months, adapting our existing program of peer-delivered telephone diabetes self-management education and goal setting. Each telephone session was on a different self-management topic, preceded by an educational video that participants viewed at home before the telephone discussion with their peer coach. For this project, we developed new educational content and storytelling videos of community members sharing their own lived experience of diabetes following the Chronic Illness Trajectory Framework.24 We added the lived experience of diabetes to the educational videos by integrating short clips of community members sharing their experiences with diabetes self-management activities. Brief interviews were conducted with people at the park and at community events to include a wide variety of practical strategies used by fellow community members to remember to take medications, exercise, and eat healthy foods. Longer stories that addressed other themes were recorded in a one-on-one interview format. Videos were integrated throughout the sessions and were also used as a starting point for discussions by peer coaches and participants during each telephone session. In the session videos, stories were used to introduce education topics or reinforce health messages. For example, the video session focusing on medications began with a series of community members sharing their motivations for taking their medications (eg, being there for their children, wanting to live a long healthy life). Another video session that discussed BP medication titration included a story of a gentleman talking about the process he went through to find the right combination of medications to manage his BP. Stories were collected from both men and women. The majority of the storytellers were African American to reflect our intended study population. The final Living Well With Diabetes intervention consisted of 11 sessions with educational DVDs with integrated storytelling about how community members accepted their disease and overcame barriers to medication adherence, plus one-on-one telephone peer coaching. These 11 sessions made up the intensive-intervention phase. In addition to the 11 intensive-intervention sessions, we included 2 types of brief check-ins during the less-intensive maintenance phase of the intervention, during which coaches contacted participants monthly (Table 3).
Living Well With Diabetes Intervention Topics and Timing.
Community member participants were Black Belt adult residents taking medications for diabetes. Community coordinators approached members of their social networks and their communities to invite participation. Interested individuals were provided a telephone number to contact the study team or gave permission to be contacted. Purposive sampling ensured the inclusion of both men and women and of individuals experiencing and not experiencing difficulty with their diabetes medications.
Three focus groups were conducted in community settings in Wilcox County, Alabama, in 2014 (Figure 3). The purpose of the focus groups was to gain an understanding of the patient perspective on living with diabetes in our targeted communities so that this perspective could be integrated into the intervention. After obtaining informed consent, a moderator guided the discussion, with a co-moderator taking notes. The moderator guide was developed using concepts from the Chronic Illness Trajectory Framework24 and focused on participants' understanding of diabetes and diabetes complications, the impact of diabetes on the participants' lives, and strategies to live well with diabetes. Each group session was audio recorded and lasted between 30 and 60 minutes. Participants received lunch and a $20 gift card.
Wilcox County in Relation to Jefferson County and Birmingham, Alabama.
Focus groups were analyzed using open coding, an emergent process in which concepts, or codes, are derived from the text rather than applying codes that were determined a priori.60 Recordings were transcribed and analyzed using NVivo v.10 (QSR International). Transcripts were independently reviewed by 2 investigators to identify major themes and generate initial codes with definitions. Initial codes then formed the basis of a codebook that was used to code the transcripts through frequent meetings among investigators to resolve discrepancies.
To further explore themes emerging from the focus groups, we conducted 28 semistructured interviews with community members with diabetes. Because feedback from focus groups was used to identify various beliefs regarding diabetes and its impact on the daily lives of participants, we also interviewed some participants from the focus groups so that topics could be explored more in depth. For example, interviews explored the participants' transition from initial diagnosis to their current experience of living with diabetes. As with the focus groups, we analyzed the interviews using the open-coding methods described previously.
Focus groups and semistructured interviews were followed by nominal group sessions to identify and prioritize common questions about diabetes and self-management for integration into the intervention. Six nominal group sessions were conducted; 4 sessions were with community members with diabetes, and 2 sessions were with individuals who had worked with the research team as peer coaches in 2 previous diabetes self-management studies. Group discussions were conducted in our partnering communities. Participants provided informed consent and received a $20 gift card and a healthy lunch.
The nominal group technique is a semiquantitative method of structured group discussion that results in a prioritized list of responses to a specific question.61 Advantages over traditional focus groups include engaging each participant without discussion being dominated by a small number of participants. Participants first considered the question, “What are some questions that you have about your diabetes?” Peer coaches were asked to recall questions their clients had over the course of delivering interventions from our 2 previous trials. Each participant first silently considered their answers and wrote them down. The moderator then asked each participant to share an item from their list in round-robin fashion until no new items emerged. Items were displayed in full view of all participants, and discussion centered on understanding the meaning and distinctness of each item.
Because of the high functional illiteracy in our partnering communities, the 4 nominal groups of participants included the generation of a list of responses but did not include the ranking phase, because community members found the ranking to be too confusing. The 2 groups of peer coaches also generated a list of responses; in addition, they were asked to prioritize items. Each participating coach selected their top 3 choices, which were weighted (5 points for their top choice, 3 points for their second choice, and 1 point for their third choice). The choices were then totaled and the prioritized list displayed for the group's review. Peer coaches were also asked to consider a second question: “What are some topics your clients find difficult to talk to their doctors about?” They listed their responses and ranked them in a fashion analogous to the procedures described above.
Once collected, nominal group data were organized by research staff using card sorting. All items generated for the first question were listed, duplicates were identified, and unique items were sorted into groups by the researchers, who then met and came to consensus on the groupings, which were further categorized into larger domains. A similar approach was used to group the responses to the second question asked of peer coaches. The nominal groups yielded a list of questions that informed the development of the content of the intervention, emphasizing those questions prioritized by the coaches.
The Living Well With Diabetes program was developed iteratively, incorporating aspects of the Chronic Illness Trajectory Framework,24 Bandura's social cognitive theory, adult learning theory, and results from the qualitative research; additional discussions with community members; and community coalition meetings to develop an initial draft of the intervention consisting of an 11-session program with videos (Table 3), a peer coach manual, and a participant activity book.
The intervention development process combined peer coach training and intervention pretesting developed in a previous study and described in detail elsewhere.62 Training began with 2 in-person sessions that covered basic skills, like goal setting, motivational interviewing, and effective communication skills. This was followed by 3 months of training for pairs of peer coaches who role-played each intensive-intervention session and 1 maintenance session, once playing the role of a peer coach and once playing the role of a participant. Each coach was certified for each session by research staff, with the coach playing the role of coach and the research staff member playing the role of participant. Staff assessed (1) session fidelity, (2) understanding of session content, (3) relationship/rapport with the participant, and (4) other miscellaneous concerns identified by the certifier. Additional opportunities to practice with study staff or a community coordinator were offered to coaches who were not certified after the first evaluation session.
Study staff solicited suggestions for refinements at the certification session, which were incorporated into the intervention and data collection protocols. Our approach optimized peer coaches' confidence in their ability to deliver the intervention, empowerment, and engagement, with coaches feeling ownership over the program.62 This approach results in few peer coach resignations later on, which is a problem in many coaching intervention programs.63
Peer coaches were all women, were African American, had diabetes or cared for someone with diabetes, and were willing to work with participants by telephone. Peer coaches were hired as contractors by a community-based nonprofit organization, Health and Wellness Education Center of Sumter County. This organization had a subcontract with UAB to pay and supervise the peer coaches, providing ongoing support to coaches, offering encouragement, and identifying new coaches should a coach resign from the program.
Nineteen peer coaches completed training and certification. During the implementation period, 3 peer coaches dropped out (2 coaches due to health reasons and 1 coach due to an increase in work responsibilities not related to the study). Peer coaches were matched with an average of 14 participants over the intervention implementation period. The peer coaches dictated the number of participants with whom they wished to work at any one time, with most coaches opting for 3 to 6 participants at a time. Peer coaches had no previous medical experience and did not have a relationship with medical providers in the community.
The trial schematic is shown in Figure 4. The trial was cluster randomized to reflect the clustering of patients within tightly knit small rural towns, with each town serving as the cluster. People talk frequently within these communities, and if 2 participants from the same town were in different trial arms and compared notes, the opportunity for contamination was great, which could change their behavior and threaten the validity of the study.
Schematic of the Living Well With Diabetes Trial.
Eligible adults were aged ≥18 years, had been told by a doctor or nurse that they had diabetes, were taking oral diabetes medications, reported nonadherence with their diabetes medications or wanted help with their diabetes medications, and had seen a primary care doctor in the past 12 months. Nonadherence was assessed using the modified Green 3-item scale (Table 4) with an additional question: “Would you like help with taking your diabetes or sugar medicines?” If the individual responded “yes” to any 1 of these 4 questions, he or she was eligible for the study. Adults who were not community dwelling, were <18 years of age, were pregnant or planned on getting pregnant in the next 6 months, had an end-stage medical condition with limited life expectancy, did not have a primary care doctor, were expected to move out of the area in the next 6 months, or were unable to communicate in English by telephone with their peer coaches were not eligible to participate.
Medication Adherence Scale Used in the Living Well With Diabetes Study.
We primarily used a chain referral sampling method for recruitment, an effective technique relying on social networks to engage hard-to-reach populations.65,66 Community coordinators also presented the study at local community events, placed newspaper advertisements and radio announcements, posted flyers in local venues, and made church announcements. Coordinators formed teams of community recruiters, provided them with study orientation, and then asked them to invite potentially eligible individuals in their social networks to participate.
In addition to community-based recruitment, we displayed flyers in local primary care practices along with interest cards. Research assistants or community coordinators were also stationed in the waiting rooms and screened interested patients referred by practice staff.
Control participants received a general health education DVD containing videos on dementia and Alzheimer disease, breast cancer screening, colorectal cancer screening, osteoporosis and fall prevention, eye health, oral health, foot care, and driving safety.
Intervention participants received the Living Well With Diabetes DVD covering 6 content areas and an activity workbook. The activity workbook included illustrations and activities to allow participants to follow each session's structure in coordination with the peer coach. Participants used the activity book with their peer coaches during each of the 11 intervention sessions (Table 3). Medications were discussed at all program sessions, with each session beginning with a review of medication barriers and the participant's behavioral goal related to their medications. Months 1 to 3 constituted the intensive-intervention phase; during weeks 1 to 8, contacts occurred weekly, and during weeks 9 to 12, contacts occurred biweekly. Months 3 to 6 constituted the maintenance phase, during which peer coaches called monthly to review goals, assess barriers, and problem solve to overcome barriers, and participants continued to work on goals set during the intensive phase.
The final Living Well With Diabetes program consisted of 11 sessions over 6 months. Before each session with the peer coach, participants watched a 15- to -30-minute video with that week's educational content with integrated stories told by community members. Telephone sessions with the peer coach reinforced the video's educational content through discussions and activities. During the session, peer coaches used a manual and client plan book, the latter being a new addition as a result of the feedback received during the training/pretesting. The plan book was used by the coach to track the participant's behavioral goals. Participants followed along during the session using their activity book and monitored their behavioral goals between sessions. Diabetes medication barriers, including adverse effects and cost issues, were assessed at every session, while healthy eating and physical activity goals were set during weeks 2 and 3 and monitored throughout the 6-month program.
Intervention fidelity was monitored weekly during one-on-one meetings with peer coaches and review of the program manuals. Fidelity was also monitored using random session recordings, reviewing completed program materials (as each participant completed the intervention, the peer coach submitted the manual used for that participant, which was then reviewed by the program staff with subsequent feedback to the coach), and meeting regularly with peer coaches throughout the implementation period. Approximately 10% of program sessions were recorded and reviewed by the program manager. Peer coaches individually met with a program staff member weekly. The primary purpose of this meeting was to provide support and encouragement to the peer coaches and to identify implementation issues quickly so they could be addressed in a timely manner. The agenda for this meeting included checking the progress of each of the peer coach's participants, identifying difficulties related to program content, and identifying logistical challenges (such as helping with scheduling program sessions or tracking down new contact information as needed). Challenges in program implementation that needed to be addressed as a group (eg, several peer coaches having similar problems or having input regarding potential strategies to overcome problems) were added to the weekly group meeting agenda. During the weekly group meetings, in addition to discussing implementation challenges identified during the individual meetings, time was spent on issues that were identified by study staff during review of the program materials or through listening to session recordings.
The study was powered to detect clinically meaningful differences in physiologic risk factors and had 4 primary outcomes. Medication adherence was self-reported using a modified version of a 3-item adherence scale developed by Green et al (Table 4).64 The reliability coefficient of the modified scale for our study sample was r = 0.41. The scale was scored by summing the number of “yes” responses, resulting in possible scores of 0, 1, 2, or 3. The scale is commonly dichotomized, with adherence defined as 0 “yes” responses.
The 3 physiologic measures were HbA1c, BP, and low-density lipoprotein cholesterol (LDL-C), the “A-B-Cs” of diabetes care that were a central focus of the intervention. Each measure was assessed by the research team as described below (see the “Data Collection, Randomization, and Data Sources” section). HbA1c and LDL-C were assessed via finger-stick blood samples, as described below.
Secondary outcomes were health-related QOL and self-efficacy. Generic QOL was assessed using the Short Form 12-item (SF-12) questionnaire, and diabetes-specific QOL was assessed using the Diabetes Distress Scale.67 Self-efficacy was assessed using the Self-Efficacy for Appropriate Medication Use Scale68 and the Perceived Diabetes Self-Management Scale,69 which is associated with HbA1c levels.
To understand the pathways through which the intervention exerted its effects, we collected additional measures guided by our conceptual framework (Table 5). Where available, we selected validated scales tested in populations with low literacy. We also transcribed the name and dose of all diabetes, BP, and lipid medications at baseline and follow-up.
Additional Measures Used in the Living Well With Diabetes Study to Understand Mechanistic Pathways.
All study measures were collected by trained research staff following standardized protocols with quality control. To the extent possible, data assessors were blinded to the participant's intervention assignment before data collection at baseline. Data were collected at baseline and again 6 months later.
Process measures were selected to understand which aspects of the intervention were particularly effective, assessing both program satisfaction and peer coach effectiveness. To assess intervention fidelity, we collected additional process measures from peer coach workbooks, including notes from each session and data entered for specific activities. Therefore, the peer coach workbooks were the source of the number of contacts with participants. We also monitored intervention implementation and fidelity through weekly teleconferences with peer coaches and weekly outreach to each peer coach throughout the intervention period. During these conference calls, study staff collaboratively troubleshot problems; ongoing advice often came from other peer coaches or community coordinators. Community coordinators were instrumental in finding participants whose telephone numbers had changed.
Power estimates accounted for clustering of patients within towns, using a variance inflation factor, conservatively estimating power for intraclass correlation coefficients (ICCs) of 0.01 to 0.05, which were taken from the ENCOURAGE cluster-randomized trial conducted in the same region, which also used towns as clusters. We assumed 20% attrition, or 200 participants analyzed per arm. We estimated detectable differences in adherence proportions and the difference in mean changes between arms detectable with 80% power using 2-sided χ2 and t tests with α = .05 (Table 6). We hypothesized that the intervention would result in improved medication adherence. Eligibility required participants to self-report problems with medication adherence at baseline, eliminating ceiling effects and allowing us to reasonably anticipate large changes in the intervention arm and only modest changes in the control arm, as was seen in a storytelling intervention directed at BP.40 Adherence was assessed using the modified 3-item Green scale (Table 4).64 Because data on the clinical meaning of group mean changes in adherence as assessed by the Green scale are limited, we designed the study to detect clinically important changes in physiologic measures. We estimated we would have 80% power to detect differences of 0.28 to 0.32 SD for continuous outcomes; this translated to detectable differences for the change in HbA1c from 0.41% to 0.48%, in systolic BP of 3 to 4 mm Hg, and for LDL-C of 6.7 to 8.1 mg/dL, as well as differences as small as 3.7% to 4.4% for change in the Diabetes Distress Scale.
Detectable Differences With 250 Participants Per Trial Arm, 80% Power.
Data collection occurred at baseline and 6 months to ensure adequate time for changes in biologic outcomes and allowed us to assess whether behavioral changes were maintained, because all intervention content was covered during the first 8 sessions.
Referred individuals were screened by study staff for eligibility. Interested individuals were scheduled for a 45- to 60-minute telephone interview and were mailed the informed consent form. Telephone interviews were scheduled at the earliest convenient date after screening and were conducted by trained UAB study staff. For quality assurance, we randomly selected 10% of interviews for the program coordinator or data coordinator to listen to.
After completing the interview, participants were randomly assigned, and in-person data collection visits were scheduled. The clusters were towns (or, in Birmingham, neighborhoods) blocked on small (<1000 residents), medium (1000-1999 residents), and large (≥2000 residents) community sizes. For the Birmingham area, clusters were the 99 neighborhoods in Birmingham, all of which were considered to be large communities because all include ≥2000 residents. With 55 clusters in 2 regions participating in the 400-participant ENCOURAGE study,19 we conservatively estimated that this study would enroll participants from 80 to 100 clusters, with the addition of 2 new regions. The first member of a given cluster determined the study arm of that cluster. All subsequent residents were placed into the same trial arm as the first participant from that cluster.
Participants met research staff at a community location or in their homes for obtaining physiologic outcome data as well as additional study data. Data collectors were trained and certified by the study investigators. Participants were asked not to drink any caffeine (from coffee, tea, or soda), eat, do any heavy physical activity, smoke, or ingest alcohol for 30 minutes before the appointment. During the visit, the participant signed informed consent. The following measures were collected:
During the data collection visit, participants were asked to produce all medications currently taken; the name of the medication, dose, and frequency were recorded from the medication bottle.
After completing baseline data collection, participants were given a portable DVD player that was theirs to keep, as well as a health report card with their HbA1c, LDL-C, BP, and weight data. Participants in the control arm were given the general health education DVD. Participants in the intervention arm were given the name of their peer coach, an activity book, and the intervention DVD and were offered the use of a study cell phone for the duration of the study. Participants received a $20 Visa gift card at the 6-month in-person data collection visit. The procedures for data collection at the 6-month visit were identical to those used at baseline.
Data were entered into Research Electronic Data Capture (REDCap) database for management.83,84 Field-based data collectors used paper forms that were later entered into the data management system. Data collectors were trained and certified to maximize data completeness and accuracy, and data collection forms were reviewed by study coordinators, with retraining provided as needed.
During the 6-month intervention period, UAB research staff contacted all participants 2 to 3 times for brief calls to answer questions, update contact information, and provide help as needed in accessing study education materials. Figure 5 shows the study flow from referral to study completion.
Living Well With Diabetes Trial Flow.
The main study hypotheses tested were that intervention participants would have higher medication adherence, significantly greater improvement in HbA1c, BP, LDL-C, and measures of QOL, and greater self-efficacy than with control participants. Analysis began by calculating ICCs for each outcome to assess the magnitude of clustering; all regression models for outcomes with ICCs of ≥0.01 used generalized estimating equations (GEEs) to account for clustering. The intervention acted at the individual level; thus, analyses were at the individual level. Statistical significance for the comparison of the study arms was judged by P < .05 for the coefficient for study arm in regression models. For each continuous or ordinal outcome measure, we calculated the change from baseline to 6-month follow-up for each participant. These change scores were the units of analysis. We proceeded to examine summary statistics, histograms, and scatterplots of the measured outcomes for outliers and data trends. We then tested for unadjusted differences in changes between the intervention and control arms using t tests, Mann-Whitney-Wilcoxon tests, or regression models with GEEs to account for clustering as appropriate. We then adjusted for baseline values of outcome variables and any factors that were imbalanced at baseline. We used the complete-case approach to analyses because all individuals with baseline and follow-up data were included; those who did not complete follow-up were not included in the analyses. Among participants who participated in the 6-month follow-up, there were no missing data for HbA1c, QOL, and medication adherence; 3 participants were missing data for BP; 10 participants were missing data for BMI; and 76 participants were missing data for LDL-C. Appendix H, Table 2 shows the final analytic sample size for each outcome by treatment arm. All analyses were carried out in SAS v9.4.
Medication adherence was originally proposed to be assessed by self-report and by electronic pill bottles (MEMS TrackCaps). However, early during aim 1, our community partners expressed concerns that MEMS TrackCaps would disrupt the usual routine of medication taking (such as using pill boxes). Furthermore, most individuals enrolled in the study were expected to be taking multiple medications for diabetes as well as taking BP and cholesterol medications. The MEMS TrackCaps system would have required the participant to use a cap for each medication. We discussed these concerns with the funding agency, and the use of the MEMS cap was removed from the protocol. A second change was the expansion of the study's recruitment area to include Birmingham, which was not originally planned.
Three focus groups were conducted with 16 community members. Twelve participants were ≥51 years, 15 participants were women, 15 participants had a high school or higher education, and 10 participants were employed (Table 7). Twenty-one of the 28 interview participants were ≥51 years, 21 participants had high school or higher education, and 15 participants were men (Table 8). Findings from the focus groups and interviews identified topics to add to the education content and to guide storytelling videos and culturally appealing messaging related to medications (Table 9).
Participant Characteristics for the Focus Groups.
Characteristics of 28 Participants in Semistructured Interviews.
Themes for Storytelling Videos That Emerged From Focus Groups and Interviews.
Four nominal groups engaged 37 individuals with diabetes (Table 10), and 2 groups engaged 13 experienced peer coaches (Table 11). Nominal group findings were used to identify knowledge gaps about diabetes self-management and medications to guide education content development (Table 12).
Characteristics of Participants in Patient Nominal Groups in Camden (2 Groups), Monroe (1 Group), and Pine Apple (1 Group), Alabama.
Characteristics of 13 Peer Coaches in Nominal Groups Held in Camden (n = 9) and Livingston (n = 4), Alabama.
Content Included in the Diabetes Medication Session Based on Nominal Group Findings.
The drafts of the peer coach manual, participant activity book, and session videos were pretested and refined as part of peer coach training (Table 13). Tables 14 to 17 summarize information learned that resulted in changes made to the intervention based on stakeholder engagement. Many of these themes were integrated into the session content as points of emphasis (Table 13).
Content of the Living Well With Diabetes Intervention.
Focus group findings suggested that area residents viewed diet and exercise as more important than medications, which they viewed as required only if one is unsuccessful with diet and exercise. Many displayed a lack of knowledge that medications were used to lower risks of longer-term complications, and many believed that oral medications are only needed if morning glucose readings were abnormal. Three major themes emerging from the focus groups and interviews included the common perception of personal failure if diabetes medications continue to be needed, a profound lack of understanding of how medications work, and the belief that the need for medication was temporary and episodic (Table 14).
Themes From Focus Groups Regarding Misinformation and Misperceptions About Medications.
The focus groups also provided insights into the lived experience of diabetes for individuals living in our partnering communities. Participants reported feelings of fear, shock/disbelief, and devastation in reaction to a diabetes diagnosis. Many described periods of denial. They reported the need for extensive planning, particularly related to food and medicine. Fear of low blood sugar drove much of their behavior. Participants noted physical limitations, feeling that they could no longer do what they once were able to do. Participants also described coping strategies, including religion, exercise, support from family/friends, and active participation in self-management (Table 15).
Themes From Focus Groups Regarding Reactions to Diagnosis, Impact on Daily Activities, and Coping Mechanisms.
Findings from the nominal groups identified knowledge gaps that were organized into domains and subdomains of major topics (Table 16). The 2 subdomains that contained the most questions (n = 12) were “questions related to diet” and “effect of diabetes on the body.” The subdomains with the next largest number of questions contained only 6. Other subdomains with a significant number of responses were “medication effectiveness,” “logistics of taking medicine,” “blood sugar management,” and “heritability and causes of diabetes.” Peer coaches rated questions about “adverse effects,” “effect of diabetes on the body,” and “nervous about visit” as being the most commonly asked by participants. Almost all topics that were difficult to discuss with the doctor had a significant number of questions (Table 17). Peer coaches commented that patients often are nervous about the doctor visit in general and that the patient's “mind goes blank,” even for questions that were not noted to be particularly difficult to discuss.
Questions People With Diabetes Had About Diabetes.
Topics Patients Find Difficult to Discuss With Their Doctor.
Recruitment resulted in 1735 community members being referred for screening, with 553 individuals (32%) from community organization partners, 412 individuals (24%) from community coordinators and UAB staff, 340 individuals (20%) from previous studies who wanted to be informed about future studies, 160 individuals (9%) from flyers distributed in the community, 125 individuals (7%) referred by other community members, 85 individuals (5%) referred by study peer coaches, and 60 individuals (3%) for whom referral information was unknown or missing (Figure 6). Of the 1735 individuals referred for screening, 473 (27%) were enrolled, 402 (23%) declined to participate, 507 (29%) were not eligible, and 353 (20%) could not be contacted (Figure 7). Eighty-two individuals (17.4%) in the sample were enrolled from Birmingham. Individuals who declined to participate in the study before completing the eligibility screening gave as reasons lack of interest (n = 315 of 401 [78.6%]), no time to participate (n = 53 [13.2%]), illness or a family member who was ill (n = 26 [6.5%]), or offered no specific reason for declining (n = 8 [2.0%]). Table 18 shows reasons for ineligibility among those screened. The most common reason for ineligibility was not being prescribed a pill for diabetes.
Sources of Referrals of Enrolled Participants.
CONSORT Diagram for the Cluster-Randomized Trial of the Living Well With Diabetes Intervention.
Reasons for Ineligibility.
Of 473 participants randomly assigned, 203 were allocated to the intervention arm and 270 to the control arm. Of these 473 participants, 85.4% completed the 6-month follow-up (81% of intervention arm participants and 89% of control arm participants). The reasons for discontinuing the study are listed in Figure 7. The final number of clusters in the trial was 114. Of the 58 intervention clusters, the largest cluster had 32 participants, and the smallest cluster had 1 participant. Of the 56 control clusters, the largest cluster had 41 participants, and the smallest cluster had 1 participant. The imbalance in participants by study arm reflected variability in recruitment rates by cluster.
There were no significant differences in baseline demographic characteristics between the 69 participants who did not complete 6-month follow-up and those who did (Appendix H, Table 3). Examination of the baseline demographic characteristics for the Black Belt and Birmingham study populations showed that the Birmingham population had slightly more non-Black persons, fewer were married or living with a partner, and more were taking insulin. There were no significant differences in age, sex, education, income, or employment (Appendix H, Table 4).
Table 19 presents baseline characteristics of the 403 study participants who completed follow-up, overall and by study arm. Their mean age was 57 years, 78% were women, 91% were Black, 56% had a high school education or less, and 69% had an annual income of <$20 000. Only race differed significantly by trial arm. At baseline, 42% (n = 171) of participants reported that they were adherent to their medications, using the 3-item Green scale, but still wanted help with their diabetes medications. The mean HbA1c was 8.4%, mean systolic BP was 129 mm Hg, the mean LDL-C was 83.0 mg/dL, and the mean BMI was 36.7.
Baseline Characteristics of 403 Participants With Follow-up Data in the Living Well With Diabetes Trial.
The unadjusted differences in the primary outcomes of the trial are shown in Table 20. All unadjusted outcome changes favored the intervention arm, with no significant differences.
Mean Unadjusted Change From Baseline to 6-Month Follow-up in Primary Outcome Measures in the Living Well With Diabetes Trial.
The main results from the trial for the primary outcomes are shown in Table 21, which shows parameter estimates from analysis of covariance (ANCOVA) models adjusted for baseline values and covariate imbalance across treatment arms (race). The ICC for medication adherence was 0.055, with a 95% CI of −0.018 to 0.145, thus not statistically significant. However, the lower bound of the 95% CI was very close to 0; thus, to be conservative, we also present adjusted results that include baseline values and race, and we account for clustering using GEE models. As can be seen, differences in medication adherence were statistically significantly different between treatment arms in both the analyses with and without adjustment for clustering. However, the observed changes in adherence were relatively small in magnitude and of uncertain clinical importance, because the study arms did not differ in HbA1c, BP, and LDL-C outcomes. Point estimates for other outcomes were all better in the intervention than in the control arm, but these differences were not statistically significant in either model.
Parameter Estimates (95% CI) and P Values for Tests of Differences in Primary Outcomes From Adjusted Analyses Comparing Control With Intervention Arms in the Living Well With Diabetes Trial.
The unadjusted secondary outcome results are shown in Table 22. For the 2 domains of QOL, changes in SF-12 scores were very modest and did not differ significantly by trial arm. For the 4 domains of medication beliefs, changes in the intervention arm were in the anticipated direction, indicating a favorable treatment effect for the domains, and were statistically different from the control arm for necessity, concerns, and harm but not overuse. Medication use self-efficacy increased significantly more in the intervention arm than in the control arm.
Unadjusted Arm Mean Changes in Secondary Outcome Measures From Baseline to 6-Month Follow-up in the Living Well With Diabetes Trial.
The adjusted results for the secondary outcomes were similar to the unadjusted results (Table 23). The exception was the harm domain of the medication beliefs scale; the ANCOVA led to a nonsignificant difference between trial arms, and the GEE analysis led to a significantly greater difference in the intervention arm, indicating fewer beliefs about the harms of medications.
Parameter Estimates (95% CI) and P Values for Tests of Differences in Secondary Outcomes From Adjusted Analyses Comparing Control With Intervention Arms in the Living Well With Diabetes Trial.
Program satisfaction was high for both control and intervention participants (Appendix H, Table 5). As noted previously, 165 (81.3%) of the 203 participants allocated to the intervention arm completed follow-up data collection. Uptake of the intervention was robust, with 166 (81.8%) of the 203 intervention participants completing all program sessions, 8 participants (3.9%) completing 6 to 9 sessions, 15 participants (7.4%) completing 1 to 5 sessions, and 14 participants (6.9%) completing no sessions. Satisfaction with program components was high (Appendix H, Table 6). Eleven participants who completed all program sessions did not complete the follow-up. Of the 165 participants who completed the follow-up, 154 participants completed all program sessions. Those who did not complete all 11 sessions had higher baseline HbA1c (9.3%) than did program completers (8.4%) (Table 24). Furthermore, for those who completed the follow-up, HbA1c did not change much (−0.1% HbA1c) between baseline and follow-up for those completing <11 sessions, whereas HbA1c improved (0.4% HbA1c) for intervention completers with both baseline and follow-up HbA1c values; these differences were not statistically significant. This difference between completers and noncompleters should be interpreted in light of the known limitations of a per-protocol analysis.
Program Completion, Timing of Data Collection, and Unadjusted Change in HbA1c Between Baseline and 6-Month Follow-up in Intervention Participants of the Living Well With Diabetes Trial.
There were differences in the time elapsed between baseline data collection and session 1 completion. Our goal of completing session 1 within 30 days of enrollment was often not realized for various reasons, such as participant availability and logistical issues. As a result, 106 (52%) of the 203 intervention participants completed session 1 within 30 days, and 166 (81.8%) intervention participants completed session 1 within 60 days of enrollment. Differences in the change in HbA1c were not significantly different between these those who completed session 1 within 30 days and those who did not (P for difference = .99), or between those who completed session 1 within 60 days and those who did not (P for difference = .49) (Table 24).
After the final program session with the peer coach, our goal was to complete follow-up within 30 days. HbA1c reflects a participant's average blood glucose over the previous 90 days, with greater impact for the previous 30 days; if the intervention exerted a beneficial effect that decayed over time, delays in data collection could dilute intervention effects. In addition to the effectiveness of the intervention attenuating quickly, it is also possible that those participants with whom it was difficult to schedule the follow-up study visits may also be less adherent to their medications in general, perhaps reflecting an overall difficulty with adherence or dealing with competing demands. Finally, because the delays in follow-up data collection could be an indicator of engagement of the participants in the intervention itself (ie, those who were challenging to contact and schedule for the data collections were less engaged in the intervention), we conducted exploratory analyses examining the impact of the timing of data collection.
We compared HbA1c changes between baseline and 6-month follow-up for those who completed follow-up within 30 and 60 days of their last program session. Again, due to scheduling challenges, only 38 (25.3%) of 150 participants with final follow-up and completion of the program completed follow-up within 30 days, and 104 (69.3%) participants completed follow-up within 60 days. The point estimate for change in HbA1c was greater for program completers who completed follow-up within 30 days than for those who did not (Table 24). Similar differences were evident for those who completed follow-up within 60 days vs those who did not. Neither of these differences reached statistical significance, but both differences were near the prespecified detectable differences in HbA1c (0.4%) for the trial, suggesting that the timing of final data collection may have influenced the outcomes of the trial.
To further explore the impact of the timing of data collection, we conducted an analysis of group differences in change in HbA1c between baseline and 6-month follow-up for only the 121 intervention participants who completed the intervention and final data collection within 60 days of completing the intervention and for the 239 control arm participants; we adjusted for baseline values and race and separately accounted for clustering (Table 25). These results demonstrated differences in the point estimates for change in HbA1c between baseline and 6-month follow-up (−0.26 using ANCOVA and −0.28 using GEE) between control and intervention arm participants who both completed the program and completed follow-up within 60 days of completing the intervention; P values approached but did not reach statistical significance. The magnitude of these differences was less than the 0.4% HbA1c detectable difference used to design the trial, but the 95% CIs around the change did include 0.4.
HbA1c at Baseline and Follow-up, and Adjusted Change in HbA1c Among All Control Participants Compared With Intervention Participants Who Completed Follow-up Data Collection Within 60 Days of Their Last Program Session.
Satisfaction with the study was high, with 92% of the control and 95% of the intervention participants expressing that they were extremely satisfied or satisfied with the program (Table 26). More than 99% of all participants found the study staff to be helpful and friendly, and >90% of all participants expressed interest in participating in future similar studies. Notably, although more intervention participants than control participants reported that they discussed the results of their first report card with their doctor, only 36.4% of intervention participants did so, despite this being a topic in the intervention.
Program Satisfaction and Program Evaluation Questions.
Intervention participants expressed high program satisfaction (Table 27). Most participants found the program materials to be helpful, with 92.6% reporting that they used the program activity book and found it helpful and 98.2% of participants watching the videos and finding them helpful. Furthermore, most participants reported that working with a peer coach was a positive experience: 88.8% reported that working with a peer coach was helpful, 89.7% reported that it was easy to reach their peer coach, 91.3% reported that it was easy to talk to their peer coach, 91.9% reported that the support they received from their peer coach was good or great, and 91.9% reported that their peer coach understood them. Overall, 96.3% would recommend their peer coach to a friend or relative with a similar health condition.
Living Well With Diabetes Intervention and Peer Coach Evaluation Questions.
In aim 1, we collaboratively adapted an existing intervention based on social cognitive theory to integrate the Corbin and Strauss Chronic Illness Trajectory Framework and peer storytelling. The final Living Well With Diabetes intervention consisted of educational DVDs with integrated storytelling about how community members accepted their disease and overcame barriers to medication adherence, plus one-on-one telephone peer coaching with a trained peer coach to encourage medication adherence in the context of a diabetes self-management intervention. The findings provided insights in our target communities regarding commonly held beliefs regarding medications, the impact of diabetes on the daily lives of individuals living with diabetes, and common questions about diabetes and diabetes medications.
Although community members acknowledged the importance of controlling their diabetes, our results showed that they may misconstrue the role of medications in diabetes self-management. Moreover, focus group results provided a deeper understanding of the lived experience of individuals living with diabetes in rural Alabama by their thoughts and attitudes about diabetes, what kinds of coping mechanisms they used, and the major impacts that diabetes had on their daily lives. Three overarching themes that emerged included reaction to diagnosis, the impact of diabetes on daily activities, and coping strategies.
Participants reported a range of reactions when diagnosed with diabetes, including shock, denial, fear, and total devastation. Being diagnosed with diabetes had a huge psychological impact on participants, consistent with published reports indicating that being diagnosed with diabetes significantly impacts one's emotional health and well-being.85 Delivering the diagnosis of diabetes should be approached with empathy by a provider who knows the patient when possible, as many participants described enduring something similar to the grieving process after being diagnosed with type 2 diabetes. Such an approach may not be routinely implemented in most current primary care contexts.
Similarly, we found that diabetes had a significant impact on daily activities. Diabetes was seen as a burden that restricted spontaneity. Participants emphasized the need for daily planning regarding both diet and medications. Medication management and the complexity of planning ahead, knowing how to modify medications based on blood glucose readings, and remembering to plan and take medications were all common themes. Although participants recognized the benefits of a healthy diet, they expressed how difficult it is to follow such a diet, with cited challenges including traveling, which is a frequent need in rural areas. While interventions are unlikely to completely overcome these barriers related to planning and diet, increased empathy and understanding by health care providers could better engage patients and more effectively encourage them to persevere in their self-care efforts.
Participants also reported burdensome physical limitations attributed to diabetes, such as pain in their feet, increased fatigue, and impaired vision. The literature demonstrates that a greater proportion of people with diabetes have physical limitations than do people without diabetes. One study found that people with diabetes had a higher proportion of physical limitations than did people without diabetes overall (66% vs 29%, respectively; P <.001).86 In fact, disability is a key indicator of the degree of morbidity associated with chronic diseases such as type 2 diabetes.86 While the current literature suggests that physical activity can improve insulin sensitivity, glycemic control, and cardiovascular risk factors in people with diabetes,87 physical limitations were especially frustrating to many participants who stated that they knew physical activity was key to controlling their disease but that the disease itself was preventing them from being physically active. Although physical activity may alleviate some of these concerns, undoubtedly, many people with diabetes and physical limitations must learn to cope with their condition, and supportive, empathic health care providers could ensure that they remain engaged in their self-management. Notably, few studies have tested interventions specifically designed to increase provider empathy, especially in those who care for rural Black individuals.
Among the more prevalent coping strategies are prayer and religiosity. The use of solitary prayer has been cited numerous times in articles and was reported as the most common source of complementary and alternative medicine used by patients, regardless of their chronic condition.88 Previous studies also suggest that many patients and physicians believe that personal spiritual practices can play an important and beneficial role in coping with health and illness.88 The current literature suggests an important role for peer or social support in managing chronic illnesses like type 2 diabetes, as cited by our study's participants. A study by Tang et al suggested a role for social support in diabetes-specific QOL and self-management.78 That study in conjunction with our findings supports the importance of involving family and friends as well as religion in supporting this population.
One final coping strategy expressed by participants in our study was active participation in health care. The current literature supports this finding: Participatory decision-making during primary care encounters by patients with type 2 diabetes resulted in improvements in HbA1c and LDL-C levels by improving patient activation, which in turn improved medication adherence.89
Finally, nominal group findings provided insights into common questions about diabetes and other topics that patients find difficult to discuss with their doctor. Providers' understanding of these questions and how to best answer them is key to engaging and empowering patients and thereby improving their outcomes. The nominal group results suggest that patients have many questions about topics they find difficult to discuss with their doctor. This finding indicates that many patients may not feel comfortable obtaining the information they need about diabetes during their doctor visits. Studies need to explore how making such information available outside the office encounter impacts patient engagement and activation. With limited access to the internet, Black Belt-area residents often rely on information from family or friends, and we learned that misinformation is common. Peer coaches are 1 solution to this dilemma, because they are already integrated into their community and can share information through direct contact. They are therefore a potential resource when asking the doctor is too difficult. However, community health workers and peer coaches have limited medical training, and these types of health extenders are usually trained to avoid answering medical questions or giving medical advice. When we trained peer coaches, we emphasized the importance of referring patients back to their doctor or nurse for such questions, but coaches reported that their clients frequently asked questions anyway, especially about medications. Resources with reliable, easily understood information that can support community peer coaches would be helpful. Physicians can also reduce the information gap through understanding a patient's reluctance to ask questions, being proactive when talking with patients, and ensuring that patients feel comfortable discussing their concerns.
The Living Well with Diabetes intervention resulted in significant improvements in self-reported medication adherence and in consistent but not statistically significant improvements in HbA1c, systolic BP, LDL-C, and BMI. The intervention did not impact physical or mental functioning, but it resulted in significant improvements in several other secondary outcomes, including beliefs about medications and self-efficacy in adhering to medications.
Although not an outcome, a remarkable finding of this study was the extraordinarily high adherence to and satisfaction with this peer-delivered supportive intervention founded on the shared experience of living with a chronic illness. One potential reason for this finding is that a Corbin and Strauss-inspired intervention strategy holds promise, integrating the lived experience of illness through peer storytelling about how difficult it can be to accept the diagnosis of diabetes, the impact of diabetes on individuals' lives, and the importance of setting achievable goals and tracking progress. In our ENCOURAGE trial,19 the intervention was also supported and delivered by trained community members, but it did not incorporate the Corbin and Strauss approach; also, the mean number of completed sessions was 13.8 (with a wide SD of 8.1 sessions) in a 17-session yearlong program, and only 32% of participants completed all 17 sessions.19 This contrasts with the 82% completion rate of all 11 sessions of the Living Well With Diabetes intervention, suggesting that incorporating the shared experience of chronic illness into an intervention may be more engaging.
The results of this study were potentially impacted by the realities of collecting data in real-world settings. We specifically selected geographically remote, “hardly reached” settings due to their burden of chronic diseases like diabetes and the paucity of medical resources; the choice of this setting came with challenges that may have contributed to the lack of significant findings for glycemic control. Specifically, the timing of data collection relative to the time of intervention completion was particularly challenging, a problem that was previously reported.90 The variation in data collection timing may have impacted the results, but our sensitivity analyses suggested only a modest influence on the results.
The multiple influences on glycemic control are important to consider when interpreting our findings. Although medications are a cornerstone of glycemic control once HbA1c levels rise above normal, diet and physical activity are also critical. A highly compliant patient may still not achieve glycemic control if they do not eat a healthy diet or engage in regular physical activity. Our intervention included information on healthy eating and encouraged goal setting to improve diet, but it was not a lifestyle intervention per se. Participants had many questions about what to eat and about how to afford and access healthy foods, because many buy the majority of their groceries at convenience stores, such as dollar stores or gas stations. Questions related to diet were also among those rated as the most difficult for patients to discuss with their doctor. Additional practical information about how to follow a healthy diet in rural food deserts, such as the Alabama Black Belt, is needed.
Additionally, the intervention emphasized the importance of physical activity, but many participants expressed concerns about their inability to engage in enough physical activity due to pain. Our previous intervention successfully overcame pain as a barrier to exercise in patients with diabetes with chronic pain, but it did not change glycemic control. Future interventions integrating the successful elements of the Living Well With Diabetes intervention, our past intervention on physical activity, and more emphasis and practical information on healthy eating may prove more effective at achieving glycemic control than would an intervention focused on only 1 of these 3 vital components of successful diabetes management.
Clearly, this study demonstrated the feasibility of training community members to deliver a supportive and well-liked telephone intervention in these communities. It is noteworthy that 40% of our study participants were insulin users, among whom glycemic control can be more difficult to achieve. Future work building on the findings of the Living Well With Diabetes study are clearly warranted.
Although this study was not designed to examine the response to the intervention vs the control in specific subgroups (heterogeneity of treatment effect), several sensitivity analyses were conducted on subgroups based on the timing of data collection and intervention completion, as discussed previously. Although behavioral intervention studies often conduct analyses of individuals who received the full intervention dose, this was not pursued here because so many participants received all 11 sessions of the intervention and so few received less than the full dose of sessions.
Limitations include the delays in data collection that could impact the HbA1c findings; our sensitivity analyses suggest that these effects, if present, were not large. The intervention was delivered by community members by telephone, and although we did monitor some of these interactions, we were not able to monitor all interactions, creating the possibility of lapses in intervention fidelity. Greater emphasis and more practical advice on healthy eating may have resulted in a greater impact on HbA1c. We were unable to assess medication adherence by objective means and therefore had to rely on self-reported adherence. Past studies show that both pharmacy-derived medication adherence measures and self-reported measures are correlated with levels of physiologic measures, but the need to rely only on self-reported adherence is a clear limitation of this study.20 There was a real possibility that the most nonadherent patients were least likely to enroll in a randomized trial given the deep-rooted mistrust that many area residents have toward the health care system and especially medical research.91,92 It is possible that this type of intervention could have the largest effect on such individuals, so that underenrollment of such individuals could dilute the intervention effect observed in this study. This intervention did not engage physicians, because failure to titrate medications would influence physiologic measures. Randomized trials ideally should use blinding where feasible, but this is often not practical in behavioral interventions such as the one described here. Self-reported measures were used for secondary outcomes and process and program evaluation assessments, which are subject to bias. We conducted this trial in rural Black residents of the Black Belt, possibly limiting generalizability, a limitation counterbalanced by the dire health care and economic needs of the residents of the Black Belt region, especially in light of their historical significance, which we feel warrants research to benefit this population specifically. Finally, participants in this sample were mostly women, possibly limiting generalizability to men.
This was an intensive intervention with excellent adherence and good fidelity. Even though it did not result in statistically significantly improved clinical outcomes, the intervention did provide evidence that lay members of the community with no previous medical training could be trained as peer coaches and can provide the support and education to change medication self-efficacy and negative beliefs regarding medications, which is a critical first step in improving medication adherence. Because of the multiple and complex influences on clinical outcomes such as HbA1c levels (medication adherence, medication regimen, weight, diet, physical activity), future interventions should more closely link community peer coaches to health care providers for medication titration and place greater emphasis on weight loss, dietary modification, and physical activity. To provide ongoing, timely support for self-management alongside evidence-based care, there have been increasing efforts to integrate lay health workers into health care delivery teams, with varying degrees of integration. Programs range from nurse case manager-community health worker teams93,94 to state agency initiatives that connect patients to community health workers, such as the Clinical-Community Health Worker Initiative, a program of the Mississippi State Department of Health aimed at decreasing heart disease and stroke in the Mississippi Delta region.95 Supervised by a program manager and registered nurses, the community health workers work with patients who are referred from federally qualified health centers and rural clinics, and they offer counseling on healthy behaviors, reduce barriers to health care access, and contact clinical systems in the event of elevated BP readings. Further research is needed to determine optimal strategies for lay health worker supervision and sustainability.
The Living Well With Diabetes intervention resulted in improved self-reported medication adherence, more accurate beliefs about medications, and increased confidence in medication use self-efficacy. However, it did not result in significant improvements in glycemic control, BP, or LDL-C. Our qualitative research revealed that participants had substantial knowledge deficits and many questions about diabetes, especially about what to eat and how to afford healthy diets, suggesting that more educational resources are needed. We also learned that many individuals experience emotional distress when first diagnosed with diabetes, signaling the importance of empathy when delivering this diagnosis. Last, this study demonstrated the promise of training community members to deliver a telephone-based diabetes medication adherence intervention in this remote area.
Thank you to all of the participants who volunteered for this research study. This project would not have been possible without the work of our community coordinators, Ms Ethel Johnson and Ms Debra Clark, and our community data collector, Ms Sheree Moultry.
Research reported in this report was funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (#AD-1306-03565-IC). Further information available at: https://www.pcori.org/research-results/2013/testing-coaching-program-help-adults-diabetes-living-rural-alabama-take-their
General Health Education Video Handout (PDF, 210K)
Peer Advisor Candidate Screening Form (PDF, 247K)
Recruitment Flyer, Participant Interest Card, Patient Eligibility Screening Script (PDF, 572K)
Telephone Interview (PDF, 590K)
Biometric Protocol, Consent Form, In-Person Data Collection Form, Health Report Card (PDF, 2.3M)
Participant Retention Postcards (PDF, 1.1M)
Appendix Tables (PDF, 223K)
Living Well with Diabetes study participant retention protocol (PDF, 145K)
Analytic sample size for each outcome by treatment arm (PDF, 161K)
Baseline characteristics of those who completed and who didn't complete the study (PDF, 206K)
Characteristics of Birmingham area participants versus participants from other areas (PDF, 218K)
Program satisfaction and program evaluation questions (PDF, 189K)
Living Well with Diabetes intervention and peer coach evaluation questions (PDF, 201K)
Andreae LJ, Andreae SJ, Cherrington AL, Richman JS, Safford MM. (2021). Testing a Coaching Program to Help Adults with Diabetes Living in Rural Alabama Take Their Medicine as Directed. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/11.2020.AD.130603565IC
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
