A Review of Volunteer-Based Peer Support Interventions in Diabetes

The literature search was designed and implemented by a librarian (G.R.) in November 2010. To identify peer-support interventions in diabetes, searches were conducted in three databases: MEDLINE, Web of Science (i.e., Science Citation Index), and PsychINFO. Permutations of keywords and controlled vocabulary search terms were used in combination and represented three concept elements: “diabetes” AND “peer support” AND “outcomes.” All three database searches used similar search terms to maintain consistency of search strategies and ensure relevant search retrieval. Wherever possible, the search strategies included search limiters to restrict retrieval to English-language articles, those published from 1990 to the present, and those conducted in adults. A total of 523 original article citations were retrieved. Figure 1 presents the literature search process.

The initial review involved two individuals (T.S.T. and A.C.) independently evaluating abstracts and titles of articles to determine potential inclusion. For studies deemed potentially eligible, the full-text articles were obtained to perform a more comprehensive examination. Final selection was rendered after discussion and consensus. Additional studies were identified via backward-searching strategies and contact with authors of included studies (regarding recently published articles).

This review focused on patient-related health outcomes of diabetes interventions involving volunteer peer supporters. We considered peer supporters to be volunteers if they were not full- or part-time employees of the clinics, organizations, or universities in which they served. We did include programs in which peer supporters received a stipend or honorarium used to offset the costs of participation (e.g., transportation and childcare). For studies that did not explicitly report information regarding employment status, we contacted the authors to obtain this information.

We also applied other criteria for study inclusion. Specifically, studies also had to 1) describe an intervention focused on self-management rather than prevention; 2) include an explicit training component for peer supporters; 3) report clinical, behavioral, knowledge, and/or psychosocial outcomes; and 4) use an RCT or quasi-experimental (e.g., pre-/post- or case-comparison) study design. Studies involving family members as peer supporters were excluded.

Data abstraction was conducted in a two-stage process. First, one of the authors (T.S.T.) reviewed the selected studies and abstracted the relevant data using a standardized outline. Second, two other authors (G.X.A. and A.C.) performed a subsequent review to confirm the accuracy of the abstraction. Any disagreement was discussed by at least two of the reviewers and resolved.

Abstracted data were organized into two general categories: study characteristics (Table 1) and study outcomes (Table 2). Study characteristics included the study design, participant selection and characteristics, peer supporter selection and characteristics, peer supporter training, and description of the intervention. Participant demographic characteristics were presented for the total study sample at baseline. In cases where studies reported demographic information separately for the intervention and control groups, we took an average of the two values. Study outcomes included the length of follow-up and retention rate, clinical outcomes, behavioral and knowledge outcomes, psychosocial outcomes, and program satisfaction.

Twelve intervention studies met the inclusion critieria in 16 publications (four were companion papers).^{3–14,33–36}  The 12 interventions included seven RCTs, four studies using a pre-/post-design,^6,8,10,13  and one study using nonrandomized allocation of intervention and comparison groups.¹⁴ 

Face-to-face, group-based programs. Seven (58%) studies were face-to-face, group-based interventions.^{3,5,6,10–12,14}  They ranged in length from six weekly 90-minute group discussions³  to 46 weekly 90-minute group discussions.¹⁰  In all but a few cases, the intervention was delivered by a single peer supporter.^3,5,10,14  None of the interventions appeared to involve other health care providers in the delivery of the peer-support component. However, all of the group-based interventions incorporated skills-development components such as goal setting, action-planning, and problem solving.

Telephone-based platforms. Five studies tested telephone-based interventions, with two interventions engaging peers to help patients follow through on physician treatment recommendations.^4,7  Two studies used an interactive voice response (IVR) platform to facilitate communication and exchange reciprocol support between two peer partners.^8,9  And, one study involving CHWs used telephone support as one modality of intervention delivery.¹³ 

Six studies were conducted in a community setting.^6,10–14  Two studies were conducted in a clinic setting.^3,5  The four studies using a telephone/IVR platform were conducted in a home setting.^4.7–9  Eight studies were conducted in the United States;^4,6,8–13  three were conducted in the United Kingdom;^3,5,7  and one was conducted in Canada.¹⁴ 

Six studies recruited samples that were largely white.^5,7–10  Five studies targeted ethnic minority populations,^4,6,11–13  and one study did not report the racial or ethnic breakdown of the sample.¹⁴  With exception of studies by Heisler et al.^8,9  involving male Veterans Administration (VA) patients, paticipants across all studies were largely female (range 42–78%) and middle-aged to older (range 52.9–72.7 years). Participants were recruited from the clinic registry,³  VA databases,^8,9  primary care practices or community health centers,^4,5,7,13,14  and community-based organizations and sites.^6,10,11 

Of the six studies that reported language proficiency as a demographic descriptor of the sample,^{3,4,7,8,11,14}  41% of participants in one study had limited English proficiency,⁴  whereas a majority of participants from another study were proficient in both Spanish and English.¹¹  Of the three studies that reported employment status,^5,6,13  employment rates ranged from 12 to 39%. Of the five studies that reported income,^6,8–10,13  the percentage of participants who earned < $30,000 per year ranged from 48 to 73%.

Peer supporters were recruited from multiple sources, including advertisements in diabetes-focused magazines,³  diabetes care centers,^3,7  primary care practices,⁵  physician referrals,¹³  membership at a senior citizen center,⁴  previous diabetes education or health promotion programs,^6,11,14  community presentations,¹²  community postings,⁶  and word of mouth.⁶  In two studies,^8,9  the peer supporters were patients trained to support another patient with diabetes.

Several studies explicitly outlined a set of criteria used to identify and select potential peer supporters. Desirable characteristics and attributes included being actively engaged in community-based activies,⁴  being willing and available to be trained and conduct the subsequent intervention,^5,6  demonstrating proficiency in the target community's primary language,⁶  having good interpersonal skills,^6,13  possessing similarities to target participants,¹⁰  being respected in the community,¹⁰  having previous experience facilitating groups,¹⁰  having the ability to motivate others,¹⁰  having good listening skills,¹⁰  having basic problem-solving skills,¹⁰  living in the community being served,^11,12  and being willing to help the community.¹³  Although some studies only involved peer supporters who were diagnosed with diabetes,^5,8–10  other studies included peer supporters who had cared for or who were living with someone with diabetes.^6,13 

Peer supporter training varied substantially with regard to duration and intensity. We classified training intensity level into three categories (low, moderate, and high) based on the number of hours or days alotted for training. Six studies had training programs that were categorized as low intensity.^4,7–10,14  The low-intensity trainings ranged from a brief 3-hour session teaching empowerment-based communication skills and behavior-change strategies^8,9  to a 2.5-day workshop addressing the intervention's goals and objectives, program content and curriculum, adult learning principles, and facilitation techniques.¹⁴ 

Three studies had trainings that were moderate in intensity.^6,11,12  As an example, in the program by Lorig et al.,^11,12  peer supporters completed a 4-day workshop that trained them how to deliver the tightly scripted six-session diabetes self-management program (2.5 hours per session), conduct intervention sessions using role-play and simulation techniques, present diabetes education information, set goals, develop action plans, solve problems, and address problematic group dynamics.^11,12 

Finally, three studies had peer supporter training programs that were high intensity and required peer supporters to start with basic training followed by more specialized training modules.^3,5,13  For example, the training for the program by Baksi et al.³  consisted of a basic 18-session (27 hours in total) program that addressed the roles and functions of peer advisors in diabetes (PADs), core diabetes education content areas (e.g., diabetes disease process, nutrition, and acute and chronic complications), and communication skills. This basic training was followed by an apprentice component during which potential PADS conducted 33 simulated education sessions supervised by a health professional.

There was also a range of training topics taught, including core diabetes education content areas,^3,5,6,10  the stages-of-change model,⁴  behavior change,¹⁰  patient empowerment,^7–9  facilitation skills,^10,13,14  group dynamics and process,^6,11,12  motivational interviewing,^3,7  communication skills,^3,8,9,13  active listening skills,^3,7  goal setting,⁶  action planning and problem solving,^9,10  communication with health care providers,⁶  adult learning principles,¹⁴  instructional methods,^11–13  use of an IVR,^8,9  and delivery program curriculum.^{3,5,6,11–14} 

Training programs employed various instructional approaches, with several studies describing the opportunity for peer supporters to receive feedback from a professional,^3,10–12  the opportunity to observe a professional,^6,11,12  and supervision during the initial phases of intervention delivery.⁵ 

Only four studies described an evaluation component to the peer support training.^3,10–12  For example, to be certified as a peer advisor in diabetes, participants must have completed a 40-minute oral examination conducted by eight health care providers.³  Examination domains included factual knowledge, written knowledge, understanding, application of knowledge, and observed practice sessions. Other studies such as that of Lorig et al.^11,12  required peer leaders to conduct as least two practice sessions in which a trainer observed their performance and provided feedback. Similarly, Klug et al.¹⁰  required peer supporters to lead at least one supervised session.

Six studies did not provide a stipend or honorarium to peer supporters.^{4,5,8–10,14}  Five studies provided peer supporters with a stipend or honorarium for participating in training^6,7  or conducting the intervention.^3,6,7,11,12  One study provided peer supporters with gift cards.¹³ 

Nine studies examined glycemic control as an outcome measure. Of the six RCTs,^3–5,7,9,12  two reported improvements in glycemic control compared to the control or comparison condition,^9,12  and four found no between-group differences.^3–5,7  Although four RC Ts found no between-group differences, one of these studies compared a peer-led intervention to a professional-led intervention and found no differences by delivery source.³  The only study that used a pre-/post-intervention design found a signficant reduction in A1C after the peer support intervention.¹³ 

Three studies measured partial or full lipid profiles. However, only two reported results, and neither found intervention effects.^5,13  Of the four studies that assessed blood pressure, three yielded no between- or within-group differences in systolic or diastolic blood pressure.^5,9,13  One study that compared a peer-led intervention to a professional-led intervention found significant reductions in systolic and diastolic blood pressure after intervention for both groups but no group-by-time interaction.³ 

BMI was assessed in five studies,^5,7,10,13,14  but only four of these reported results. Three studies found no between- or within-group changes in BMI,^5,10,13  and one found improvements in BMI for both conditions.¹⁴ 

The only study that assessed knowledge observed improvements although no group-by-time interaction.³  Nutrition behaviors were measured in four studies.^5,6,10,12  Of the two RCTs that did so, one did not find any improvements in nutrition behavior,⁵  and the other found the peer support condition to be associated with following a healthy diet and reading food labels.¹²  The two studies using a pre-/post-intervention design that assessed nutrition behaviors found peer support to have a positive impact on fruit and vegetable consumption⁶  and frequency of following a healthy eating plan.¹⁰ 

Six studies assessed self-reported physical activity.^{4,6,10–12,14}  Of the three RCTs that did so,^4,11,12  two found no differences in physical activity between the peer support and control or comparison conditions,^4,11  and one found that participants in the peer support condition slightly increased their total minutes of aerobic activity compared to the usual care condition.¹²  A nonrandomized two-group study found increases in the number of steps taken per day for both conditions combined.¹⁴  Two of the six studies used a pre-/post-intervention design and found participation in peer support interventions led to increases in the number of days exercising at least 30 minutes a day and engaging in a specific exercise activity.^6,10 

Although nutrition behaviors and physical activity were the most frequently measured self-care behaviors, other behaviors were evaluated. Neither RCT demonstrated significant improvements in medication use⁹  or in practicing good self-care behaviors as a whole.³  There were mixed results for blood glucose testing, with one study reporting no between-group difference¹¹  and another reporting a slight improvement in blood glucose testing for the peer support condition compared to the usual care condition.¹²  Of the studies using a pre-/post-intervention design, relationships were found between peer support and performing foot exams⁶  and practicing good self-care behaviors as a whole.

Five studies measured quality of life.^5–7,9,12  Of the four RCTs that did so, one found a significant reduction in health-related distress for the peer support group.¹²  The only study using a pre-/post-intervention design did not report any improvement in overall quality of life.⁶  However, when examining the five domains of quality of life separately, Commellas et al.⁶  found a positive association between peer support and feeling “more active and vigorous.”

Self-efficacy was examined as an endpoint in five studies.^7,8,10–12  Of the three RCTs that did so, two studies found significant improvements for the peer support condition compared to the control condition,^11,12  and one study found increased self-efficacy in the peer support condition but no group-by-time interaction.⁷  Increased self-efficacy after the peer support intervention was noted in two studies using a pre-/post-intervention design.^8,10 

The two studies that asssessed depression^8,12  yielded inconsistent results. The RCT found a significant reduction in depression for the peer support condition compared to usual care.¹²  The study using a pre-/post-intervention design did not report any changes in depression scores after the intervention.⁸ 

As peer support becomes increasingly recognized as a viable and promising model for long-term diabetes self-management, greater attention has been focused on evaluating the health-related impact of peer support interventions and understanding the underlying mechanisms. Previous reviews of peer support studies in diabetes^{1,15,16,21,32}  examined peer support programs in general, irrespective of the compensation status of the peer supporter. This approach limits the identification of strategies that may be unique to volunteer-based programs. The objective of this review was to investigate implementation strategies and diabetes-related health outcomes associated with peer support interventions led by volunteers.

Findings from this review indicate that volunteer-based peer support interventions were predominantly group-based or by telephone. This may reflect a desire to contain costs, particularly costs associated with delivering the intervention by the volunteer peer supporters. This is consistent with our experience with both volunteer and paid models of peer support.³¹ 

Although peer support interventions involving volunteers may be the most viable strategy for low-resource agencies, there are a number of practical difficulties with implementing a peer-led intervention. Retention of volunteers is crucial to program success and yet extremely difficult if the volunteers are employed or have other significant obligations. In addition, retention is positively associated with a fulfilling experience, requiring program administrators to invest resources in providing support to the peer supporters.³⁷  Intervention programs wishing to assess reach or dose³⁸  may experience difficulty obtaining these process evaluation data given that the volunteers are not paid to conduct administrative work.³¹ 

As with the review of CHW intereventions by Norris et al.,²¹  our review yielded inconsistent findings with regard to diabetes-related health outcomes. With the exception of the study by Lorig et al.,¹¹  data from the few RCTs using usual-care control subjects provided little evidence that peer support interventions improve metabolic and cardiovascular outcomes. However, some studies found that peer-led interventions are as effective as,⁷  if not more effective than,⁹  professional-led interventions in lowering A1C.

Although particiption in peer support interventions is positively associated with improvements in self-care behaviors such as diet and physical activity,^6,10  generally, these changes do not differ from those observed in control conditions.^4,5,11  Indeed, the most favorable findings are for psychosocial outcomes, with peer support interventions leading to greater self-efficacy,^9–12  higher diabetes-specific quality of life, and fewer depressive symptoms¹²  even when compared to usual-care control groups.^11,12 

Additionally, participant satisfaction ratings for interventions included in this review were generally positive. These findings are not unexpected given that peer supporters are often selected based on their motivation to help. In addition, peer supporters are trained to engage in a number of supportive behaviors consistent with social support theories.^39,40  For example, several programs reported that the peer supporters provided appraisal support in the form of feedback to participants on their progress in goal-setting. Informational and emotional support were common strategies used in the telephone-based components. Considering these findings, peer support interventions may have the greatest potential for improving psychological and emotional functioning.

On further examination, there were no specific characteristics common to successful interventions compared to unsuccessful interventions. However, given the heterogenity in intervention characteristics (e.g., intervention duration and delivery; peer support selection, recruitment, and support structure; and roles and responsibilites of peer supporters) across studies included in this review, it is unlikely that we would be able to identify predictors of success.

In contrast to previous reviews, a large percentage of studies (33%) in this review compared peers versus professionals rather than peers versus controls.^3,7,9,14  Overall, peer-led interventions performed as well as professional-led interventions. Moreover, in one study,⁹  participants in the peer support intervention had greater improvements in clinical and psychosocial outcomes than those in the intervention led by a nurse case manager. These findings suggest that, although peers cannot substitute for health care professionals, they may serve a vital role in low-resource settings in which access to health care professionals and health care is poor.

Although there has been greater interest in comparing peers versus professionals, the prevailing view of organizations such as the Institute of Medcine, the American Public Health Association, and the American Association of Diabetes Educators is that CHWs function in a complementary role to health professionals or serve as important members of the health care team. In fact, some studies have found that health care professionals have a positive view of peer supporters⁴¹  and recognize the value of peer-to-peer communication in self-management efforts.⁴² 

Consistent with an overview of peer training curriculums in diabetes by Nettles and Belton,⁴³  the training process for peer supporters in this review also varied substantially. Generally, the duration and intensity of training was commensurate to that of the intervention itself. For example, interventions that required peer supporters to take a primary role in leading weekly, face-to-face, group-based DSME sessions (90–120 minutes per session) conducted training during a period of several days.^3,5,11,12  In contrast, interventions that engaged peer supporters to provide brief telephone-based support conducted training during several hours.^4,8,9 

Nevertheless, conclusions must be drawn with caution given that the description of the peer training process also differed across studies. Some studies provided a detailed description of peer training either in the article itself or in a companion article.^3,6  On the other hand, other articles offered only a brief and succinct overview of training, making it difficult for researchers to replicate the training process. Similarly, although there is a great deal of published research on peer support interventions, curriculums used to train peer supporters is not as easily accessible or available to the public.⁴³ 

Only recently have articles been published describing the development and implementation of programs training peers to facilitate DSME and DSMS programs.^44,45  For example, in one article,⁴⁴  the authors provided an in-depth “how-to” guide for developing an empowerment-based peer training program. The article not only presented the theoretical underpinnings of the training process, but also outlined the core components of the training program (e.g., knowledge acquisition, skills development, and experiential learning), specific competencies peer supporters are expected to learn and master (e.g., communication, facilitation, and behavior-change skills), instructional methods used for training (e.g., role-play, simulation, and brainstorming), assessment instruments for formative and summative evaluation, and preestablished competency criteria for successful graduation.⁴⁴  Notably, few studies in this review included a formal evaluation component of the peer training process.^3,10–12 

Without knowing what competencies peer supporters are expected to achieve and how those competencies are assessed, we have no assurance that peer supporters have acquired the knowledge and skills to perform their roles and duties successfully. Consequently, even the most well-designed interventions may not produce positive results if peer supporters are not adequately trained.

Although not formally employed or paid, volunteer peer supporters from 6 of the 12 studies^{3,6,7,11–13}  received a stipend or honorarium for the training or intervention phase of the study. Interestingly, many of the studies that provided stipends required more intensive involvement of peer supporters in terms of roles, responsibiites, or time commitment. Alternatively, of the six studies that did not provide a stipend or honorarium,^{4,5,8–10,14}  four were telephone-based interventions,^4,8,9,14  which are typically less labor-intensive and present fewer barriers to peer supporter participation. Based on these findings, it would appear that there is an implicit recognition regarding work burden and volunteerism. Specifically, with greater monetary incentive (whether in the form of salary, wage, stipend, or honorarium) comes the expectation of greater involvement, responsibility, and commitment on the part of peer supporters.

What is most compelling about the peer support model is its potential to pick up where health professionals leave off. In other words, in a health care system where professional-led DSME interventions are time-limited because of insurance coverage contraints, the peer support model may offer greater longevity.

Intervention sustainability is dependent on multiple factors, one of which is peer supporter retention. Although volunteer-based peer support does not rely on a continuous funding source for viability, this model is not without its drawbacks. For example, peer support interventions involving volunteer peer supporters are associated with a higher dropout rate than those involving paid peer supporters.⁴⁶  Therefore, in the absence of stipends or any form of monetary incentive, are volunteer-based programs sustainable over the long term? Because the length of follow-up for interventions in this review ranged from 5 weeks to 12 months, this question cannot be answered.

Nevertheless, prior studies do shed some light on potential strategies for engaging and retaining peer supporters. For example, recognition within the community and concrete acts of appreciation such as luncheons and newspaper articles are sometimes as important as or more important than financial remuneration.³¹  Also, attention to peer supporters' needs such as their desires for continuing education and interaction with the medical team could provide support and stem burnout and turnover.⁴⁷  Future research should investigate the extent to which factors such as the aformentioned are associated with the sustainability of volunteer-based peer support programs, as well as what motivates individuals to volunteer as peer supporters.⁴⁸ 

Preliminary evidence for volunteer-based peer support interventions in diabetes is promising, but limited. Of note, there is little available information to guide the integration of peer support efforts with those of clinical professional teams, and stratiegies vary widely. Emerging evidence suggests that peer supporters value interaction with the medical team.⁴⁹  However, provider preferences regarding interaction with peer supporters and how those preferences may differ by discipline (i.e., nurses vs. case managers vs. physicians) have not been adequately explored. Thus, practices wishing to implement peer support interventions will likely need to assess the preferences within their own organization to allow appropriate communication between peers and providers for their specific setting.

Although the peer support model offers greater flexibility and customization compared to the professional-led model, these characteristics also make the empirical examination of peer support challenging. In addition to quantitative investigations, more qualitiative research is needed to understand how incentives of any form affect peer supporters' initial and continued motivation to participate in interventions. In addition, greater transparency is required regarding the peer training process and the assessment methods used to evaluate peer supporters' skills and competency. Without more extensive examination, we cannot replicate, nor can we understand, the underlying mechnisms of the peer support model.

This study was supported by grants from Peers for Progress and the American Association of Family Physicians Foundation.