Community-Based Peer Support for Diabetes Management: 24-Month Changes Relative to Comparison Communities Free

Diabetes has become a critical public health issue in China, with nearly one-quarter of global cases and the latest prevalence at 11.9% (1). As a fundamentally progressive disease, self-management is needed for the rest of the lives of individuals with diabetes (2). Diabetes self-management education is necessary but often insufficient to sustain improvements (3). Thus, diabetes self-management support is widely promoted by national and international guidelines for diabetes (4,5).

Peer support has established broad evidence for its effectiveness in various diseases, including diabetes (2,6–9). Past studies have demonstrated that trained peers can effectively encourage diabetes management in China (10,11). Nevertheless, compared with standardized implementation of programs in community settings (12,13), relatively little research in diabetes has examined how community organizations may take an active part in developing and shaping peer support programs. Additionally, most studies reported the 6- or 12-month effects of peer support in diabetes (7,8,10), with some studies evaluating benefits maintained over time. For example, one study reported the effectiveness of a peer leader program in sustaining benefits of a 6-month community health worker diabetes self-management education program for Latino adults at 18 months (14). Nevertheless, few studies have evaluated long-term outcomes of peer support in diabetes self-management.

Previously, we reported the development and evaluation of peer support for people with diabetes in China implemented through community health centers (CHCs) (8,9). Significant improvements in blood glucose (HbA_1c) and other clinical indicators and diabetes distress were found after a 12-month intervention in CHCs (8), after which most improvements were sustained to follow-up at 18 months (9). Further analyses showed that linkages with community resources were associated with greater reduction in HbA_1c (8), highlighting the value of including community partners in program expansion, especially when the demands on some CHCs may limit contributions.

Reflecting these findings of the value of linkage with community resources, development of the present program focused on integrating multiple community organizations in a community-based peer support program. With support from professional expertise, community organizations adapted and delivered standardized intervention components in accordance with individual communities’ needs and capabilities. We have previously reported on how this balancing of standardization and adaptation emerged through program planning and implementation (15). In this study, we aimed to assess and report the effects of this community-based peer support program over 24 months. Given their potential relationships with diabetes management, we also tested the interaction of age and sex with interventions to examine the robustness of outcomes and to examine possible age- or sex-specific benefits.

Shanghai has ∼240 million residents in 16 districts, including 240 subdistricts (8,9). Within each subdistrict, there usually exists one CHC, one subdistrict health promotion office, and several residential committees. The residential committee is the basic organization of shared community management, education, and services within a neighborhood. In addition, Shanghai has promoted >6,000 community self-management groups since 2008, through which residents share interests in various health issues (8). As noted before (15), multiple community partners, including district and subdistrict health promotion departments, CHCs, residential committees, and community self-management groups, and, in some cases, nonprofit health promotion agencies, worked together within and across communities. In addition, the health promotion department of the Shanghai Municipal Health Commission took a key collaborative role in the project, facilitating further broad adoption of peer support.

The current study was to commence in April 2019 and continue for 12 months through April 2020 in an intervention comparison design. However, the emergence of coronavirus disease 2019 (COVID-19) in early 2020 in China restricted individuals’ participation in program activities and refocused health staff from CHCs and subdistrict health promotion offices to COVID-19–related or emergency work. Thus, program group activities generally ceased until the fall of 2020, except for some continuous peer support through WeChat, a dominant social networking platform in China. Final assessments were then planned for 24 months after baseline.

The study compared outcomes in 12 intervention communities with 4 comparison communities. The 4 comparison communities were drawn from the same districts as those from which the 12 intervention communities were drawn to represent both urban and suburban settings, with 3 urban and 9 suburban settings among the 12 intervention communities and 1 urban and 3 suburban settings among the 4 comparison communities. Both the intervention and comparison communities received nationally mandated essential public health services for diabetes management and regular primary care in CHCs. In addition, the intervention communities included three program levels with peer support to enhance diabetes care in communities, as detailed below. This study was approved by the committees for the protection of human subjects at the Shanghai Sixth People’s Hospital (2019-028) and at the University of North Carolina at Chapel Hill (19-0804).

As described previously (8,15), the program intervention was organized at three program levels, with peer leader involvement in each. Level 1 was intended to reach a broad audience in neighborhoods. Standardized program curriculum elements were designed by the project team from Peers for Progress and the Shanghai Sixth People’s Hospital, with consultation from international, national, and local experts. These included “5 Key Messages” (8) that everyone should know about diabetes and “6 Diabetes Modules” codeveloped by the Shanghai Municipal Health Commission (15) as health education materials to improve health literacy and basic knowledge about diabetes among people with diabetes and those supporting them in communities. Level 2 included neighborhood groups and activities designed to provide people with diabetes a sense of community and to facilitate healthy lifestyles. It mainly included two types of community activities. One focused on diabetes-related activities to enhance education and self-management skills (e.g., healthy diet, exercise, sharing of diabetes information and experience). The second type focused on nondiabetes-related group activities to promote a sense of community and provide opportunities for support among group members (e.g., crafts, other interest groups). Level 3 involved peer leaders providing individual support to group members struggling with their diabetes and its management and to their family members.

For levels 2 and 3, the project team developed a series of one- to two-page tools focusing on specific topics in diabetes management or typical challenges in self-management to serve as the basis for group meetings on a topic or as informational handouts and to provide peer leaders with a key structure for counseling individuals. In addition to these curriculum elements, a program training and core manual provided instruction on peer support principles, keys to success, provision of effective support, engagement with family members, building of mutual trust, effective communication, support of insulin management, privacy and confidentiality, backup and support for peer leaders, materials and suggestions for organizing group activities, setting up a diabetes action plan, and program examples from the nine CHCs in the previous study (8).

Active support of program activities by the project team included trainings in October 2018 and February and June 2019 and meetings in August, October, and December 2019 and September 2020 with district and subdistrict health staff. These encouraged staff to take on greater roles in program leadership to direct program development and share local innovation and adaptation in each community (8,15). Training for final follow-up assessment was completed in January 2021.

To elicit active support from community organizations, local innovation and adaptation of the intervention were emphasized (15). This was based on standardized intervention components created by the project team. Refining these and decisions about which community partners would take the lead in implementing each component were then placed with each community to fit their strengths, needs, and preferences. With guidance from the project team, innovations in implementation were further disseminated through meetings and other communications among the 12 communities.

According to their needs and capacities, communities tailored their specific approaches to providing key intervention components, e.g., with some communities choosing peer leaders to provide diabetes education, some choosing clinicians, and some a combination of the two. In examples of this kind of tailoring from the previous report (15), communities averaged four to five approaches to disseminate the “5 Key Messages.” The frequency of activities of diabetes education classes of level 1 and diabetes-related and nondiabetes-related activities of level 2 varied from an average of once monthly to more than four times per month in communities. For level 3, all peer leaders in half of the communities and at least one peer leader in 11 of the 12 intervention communities provided individual or family support.

Individuals with type 2 diabetes >18 years old who were willing to participate in the program were recruited. Participants were excluded if, in the judgment of the individual CHCs, they had physical impairment or serious mental health problems that were likely to interfere with their participation in the program and its activities. Recruitment included announcements and posters in communities, oral invitations from primary care physicians in CHCs, recommendations by peers or community self-management groups, or individual recruitment through quarterly health assessments required as part of nationally mandated essential public health services for people with diabetes. Participants recruited to evaluation samples signed informed consent documents at the baseline assessment.

Key clinical indicators were extracted from the CHCs’ laboratory data, including HbA_1c, fasting plasma glucose (FPG), and LDL cholesterol (LDL-C). Other indicators included BMI, systolic blood pressure (SBP), and diastolic blood pressure (DBP). Surveys were drawn from consensus measures for evaluating peer support in diabetes, as developed by previous Peers for Progress investigators (16) and adapted to the Chinese and present context. These included demographic information, depressive symptoms as measured by the 8-item Patient Health Questionnaire (PHQ-8) (17), general quality of life as measured by the EuroQol 5-Dimension (EQ-5D) survey (18), and a four-item measure of diabetes-specific distress adapted from the Diabetes Distress Scale (19). For the EQ-5D, we used the EQ-5D-3L descriptive system that comprised five dimensions (i.e., mobility, self-care, usual activities, pain/discomfort, anxiety/depression) and three levels of scoring for each and used the utility scoring system developed for the Chinese population to calculate health utility scores (Supplementary Table 1) (20).

The primary clinical outcome included in this evaluation was HbA_1c. Secondary outcomes included BMI, FPG, SBP, DBP, and LDL-C as cardiovascular risks posed by diabetes and depression (PHQ-8 score), diabetes distress, and quality of life (EQ-5D score).

For descriptive analyses, continuous variables were expressed as means with SDs. Categorical variables were expressed as numbers with percentages. All baseline characteristics were compared between the intervention and comparison communities using the independent t test for continuous variables and χ² test for categorical variables. For analyses of the follow-up values of outcome variables and changes from baseline in outcome variables at 24 months, independent t test for unadjusted analyses and ANCOVA for adjusted analyses were used to analyze the differences between the intervention and comparison communities. In adjusted analyses, we controlled for the corresponding baseline indicator for analysis of each outcome as well as CHC site and demographic variables for which the two groups of communities differed significantly at baseline, including age, sex, diabetes duration, employment, monthly family income, and oral glucose-lowering medication use. For interaction analyses of age or sex with intervention, multiplicative interactions between the intervention and comparison communities and age (<65 vs. ≥65 years) or sex (men vs. women) subgroups were investigated via ANCOVA by including the product term in adjusted models. Linear regression in adjusted analyses was used to investigate the differences of outcomes between intervention and comparison communities within age and sex subgroups. The forest plots included the unstandardized β-coefficients and 95% CIs for separate regression analyses by age and sex. Sensitivity analyses of outcomes were completed with the data set including the primary outcome of HbA_1c at both baseline and follow-up using multiple imputations with chained equations for missing values of other variables. A two-sided P < 0.05 was considered significant. Analyses were performed using SPSS Statistics for Windows, version 26.0 (IBM Corporation) and R, version 4.2.0 (R Foundation for Statistical Computing).

Figure 1 portrays the flow of participants in the intervention and comparison communities over the 24 months of the study. From all 16 communities, 1,441 adults with diabetes completed baseline assessment between 30 March 2019 and 9 December 2019, among whom 1,066 (74%) completed the follow-up assessment between 12 February 2021 and 3 November 2021. Among these, 967 (67% of the 1,441 at baseline, 91% of the 1,066 completing follow-up) had no missing values for any variables and so were included in the final analyses. Baseline and follow-up values for HbA_1c were available for 1,027 participants (71% of the 1,441 at baseline, 96% of the 1,066 at follow-up) who were included in sensitivity analyses, with multiple imputations for missing values.

Compared with participants in the comparison communities (n = 242), the baseline data in Table 1 show that those in the intervention communities (n = 725) were older (mean age 66.43 vs. 63.45 years), were more likely to be female (57.1% vs. 45.5%), had a longer duration of diabetes (mean 7.95 vs. 6.40 years), and were more likely to be retired (88.4% vs. 80.2%) and to report lower monthly family income, all statistically significant (P ≤ 0.017). The communities did not differ significantly on education levels or marital status. Also at baseline, those in the intervention communities were significantly more likely to be using oral glucose-lowering medications (86.6% vs. 81.0%, P = 0.033) but, nonsignificantly, were less likely to be using insulin (17.9% vs. 21.1%).

Table 1 includes analyses of clinical and quality-of-life indicators at baseline, showing that participants in the intervention communities had significantly lower levels of mean HbA_1c (7.46% [58 mmol/mol] vs. 7.74% [61 mmol/mol]) and BMI (25.35 vs. 25.88 kg/m²) than those in the comparison communities but higher levels of both SBP (138.16 vs. 133.79 mmHg) and DBP (80.10 vs. 77.54 mmHg, P ≤ 0.041). They also had significantly higher levels of diabetes distress (1.18 vs. 1.09) and depressive symptom (1.88 vs. 0.92) scores and lower EQ-5D scores (0.94 vs. 0.96), with P ≤ 0.001.

As can be seen in Table 2, adjusted analyses show that participants in the intervention communities had lower mean HbA_1c (7.42% [58 mmol/mol] vs. 7.95% [63 mmol/mol], P < 0.001) than those in the comparison communities at 24-month follow-up, controlling, as noted above, for baseline levels, CHC site, and variables on which the two groups of communities differed. They also had lower mean BMI (25.31 vs. 25.94 kg/m², P = 0.028), FPG (7.91 vs. 8.59 mmol/L, P < 0.001), and reported depressive symptom scores (1.39 vs. 1.41, P = 0.022). Analyses of the estimated means of follow-up values of outcome variables and changes from baseline outcome variables at 24 months are shown in Supplementary Table 2. Sensitivity analyses (Supplementary Table 3) with missing data imputed showed similar findings except for minor differences regarding diabetes distress for which greater benefits were observed in the intervention communities (P = 0.047).

Figure 2 presents analyses of interactions between the intervention and comparison communities and age (<65 vs. ≥65 years) (Fig. 2A) and sex (men vs. women) (Fig. 2B). The forest plots include the unstandardized β-coefficients and 95% CIs for separate regression analyses by age and sex of 24-month values regressed on intervention versus comparison communities, again controlling for corresponding baseline values of outcome variables, CHC site, and the demographic variables controlled in previous analyses.

In tests of the interaction of age-stratified subgroups with the intervention versus comparison communities (Fig. 2A), none of the fully adjusted models were significant for any outcome. Inspection of the unstandardized β-coefficients and 95% CIs in Fig. 2A indicated trends for those <65 years old to benefit more from the intervention on measures of HbA_1c, FPG, and depression as indicated by PHQ-8 score and for those ≥65 years old to benefit more from the intervention for BMI and DBP. Similar findings were found in sensitivity analyses with missing data imputed (Supplementary Table 4).

For interactions of intervention versus comparison communities with sex, the fully adjusted models showed significant interactions for DBP (P = 0.023) and LDL-C (P = 0.026) (Fig. 2B). Inspection of the unstandardized β-coefficients and 95% CIs indicated significantly lower DBP for men and significantly higher LDL-C for women in the intervention communities compared with those in the comparison communities. No other tests of interactions were significant. Sensitivity analyses with missing data imputed exhibited similar findings except for the test of LDL-C, which was of borderline significance (P = 0.05) (Supplementary Table 4).

This study found significant intervention effects for HbA_1c as the primary outcome, as well as improvement in BMI, FPG, and depressive symptoms among participants with diabetes receiving community-based peer support intervention at 24 months. With diabetes often viewed as more difficult in daily self-management among older adults, the lack of significant differences in response to the intervention between younger (<65 years old) and older (≥65 years old) adults points to the robustness of findings. Also suggesting robustness was the emergence of only two interactions with sex. Relative to the comparison communities, women showed a minor increase in LDL-C compared with men, while men benefited more from DBP reduction than women. Additionally, the sensitivity analyses with imputation of missing data supported the robustness of the findings.

These and findings from our previous studies (8,9,15) may be considered around key elements of the RE-AIM model of Reach, Effectiveness, Adoption, Implementation, and Maintenance model (21). As noted previously, the 71.2% retention in the community-based intervention at 24 months, from before to after the COVID-19 pandemic, supports reach. Meanwhile, adoption and implementation details of varied program features at all three program levels in the current 12 communities were documented earlier (15). Along with our report of the sustainable benefits of peer support on diabetes management to 18 months in nine CHCs (9), this study further shows benefits on clinical and quality-of-life measures to 24 months, indicating effectiveness as well as maintenance of changes in individual outcomes. Finally, through the collaboration in the project of the health promotion department of the Shanghai Municipal Health Commission, maintenance and spread of these approaches are reflected in the Shanghai Health Authority’s “Establishing a Peer Support Network for Health Management Covering All Communities” as routine policy of the Healthy Shanghai Initiative (2019–2030) “Promoting Residents’ Health Self-Management.”

These findings build on prior research in several key ways. Previous research of our group showed effectiveness of the culturally tailored, peer-supported, and/or peer-led diabetes self-management education and support program, ranging from 6 to 12 months (7,8,10,22) with sustained benefits at 18 months (9,14,22). The current study extends these findings by demonstrating that the improvement of clinical and psychosocial outcomes can be sustained to 24 months through a peer support program based in communities.

This collaborative model provides an operational base for further scaling up by integrating standardization along with adaptability to the resources and needs of individual communities. The balance of standardization with flexibility and adaptability is important. To ensure accuracy, reliability, and consistency of core program content, this project included standardized intervention structure (levels 1–3) with components such as the “5 Key Messages” and “6 Diabetes Modules” for health promotion and education. At the same time, community partners demonstrated high levels of adaptation and innovation in the delivery of these standardized components along with designing and implementing group activities to best suit the needs and capabilities of their communities. Additionally, due to the advantages of expanding program leadership beyond CHCs, the innovation, dedication, and resourcefulness of many other community partners led to continuous diabetes education and peer support via the online medium WeChat during the cessation of activities and isolation many experienced during the COVID-19 pandemic. The connections and positive working relationships among multiple community partners and dedicated and responsible community and subdistrict- and district-level staff and alignment between the project and the priorities of these organizations were key factors supporting the feasibility of the program implementation.

This study also reinforces the importance of examining outcomes beyond the findings from short-term diabetes self-management education programs. Building on findings from 6-, 12-, and 18-month evaluations (7–9,14,22), this study documented additional value of peer support in diabetes self-management with improved effects in terms of clinical, psychological, and quality-of-life outcomes.

Given the fundamentally progressive nature of diabetes and the age of participants (intervention communities mean of 66.43 years vs. comparison communities mean of 63.45 years), it is encouraging that those in the intervention communities showed a difference of 0.53 points for HbA_1c compared with the comparison communities (7.42% [58 mmol/mol] vs. 7.95% [63 mmol/mol], respectively) at 24-month follow-up (adjusted P < 0.001). This exceeds the 0.5-point difference in HbA_1c generally taken as clinically meaningful in research. Framing these findings are those of the UK Prospective Diabetes Study (23), which found that sustained improvements in glycemic control are associated in a linear manner with benefits in mortality as well as other end points. Thus, changes, such as those observed here, that extend not only across small research samples but also across communities of individuals with diabetes may have substantial public health importance.

Additional differences favoring the intervention included BMI (25.31 kg/m² vs. 25.94 kg/m²), FPG (7.91 mmol/L vs. 8.59 mmol/L), and depression scores (1.39 vs. 1.41) at 24 months in adjusted analyses. Since these factors, including metabolic status and depression, contribute to higher risks of cardiovascular diseases (24,25), the findings in this study add substantially to broader health benefits. In addition, the possible benefits for diabetes distress from the sensitivity analyses indicate benefits in this area as a topic for future research.

Due to diversities in biology, culture, lifestyle, environment, and socioeconomic status related to differences between sexes (26), examination of sex differences in diabetes management is needed. Similar to other studies (14,22), relatively lower participation of men (42.9%) was found in the intervention communities. Nevertheless, similar benefits of glucose control, including HbA_1c and FPG, were found for men and women, with no statistically significant interactions. Greater benefit with DBP was found among men. Amid a higher lifetime risk of cardiovascular disease in men in China (27) and with diastolic hypertension independently influencing the risk of adverse cardiovascular events (28), it is encouraging that the findings in this study of community-based peer support point to potential benefits for men. The only other significant interaction with sex was a minor difference in LDL-C, with only women in the intervention showing a significantly greater value. It should be noted that other than general promotion of a healthy diet, the intervention for this study did not emphasize dietary steps for lipid reduction.

As noted, no interactions were found by age (<65 vs. ≥65 years). Nevertheless, because most of the participants in the intervention were retired (88.4%) and the average age was in the mid-60s, more studies are needed to explore the adaptability of peer support for diabetes self-management for younger and employed individuals, especially given emerging trends of higher disease burdens and more difficulties in diabetes self-management of early-onset type 2 diabetes (age <40 years) (29).

Among limitations, the generalizability of this study is compromised by study participants recruited from CHCs and communities, with most being older adults and/or retired. Additional limitations included the fact that this was not a randomized controlled study and that some of the baseline data were not comparable between the two groups of communities. Other limitations included the reliance on routine clinical records for clinical data. As with other interventions, individuals willing to participate may differ from nonparticipants, with nonparticipants often having worse clinical measures. Finally, although few sex and no age differences in benefits were found in this study, approaches that may better suit men or women or younger adults need to be explored.

In conclusion, this study provides support for culturally appropriate community-based peer support to improve diabetes self-management at 24 months in individuals with diabetes in China. It offers encouraging evidence for a community organization approach to peer support for diabetes management by combining standardization of key components and approaches along with adaptation and innovation across differing communities (6,30,31).

Funding. This project was supported by National Science and Technology Major Project grant 2024ZD0523300, Shanghai Key Discipline of Public Health grant GWVI-11.1-20, Sanming Project of Medicine in Shenzhen grant SZSM202311019, Shanghai Municipal grant 2022ZZ01002, National Key Clinical Specialty Discipline Construction Program of China grant Z155080000004, and National Science and Technology Major Project grant 2024ZD0523300 (to W.J., principal investigator), China Medical Board grant 24-559, a China Scholarship Council grant and Shanghai Municipal Health Commission grant JKKPYC-2022-12 (to Y.L., principal investigator), Merck Foundation and Sanofi China grants (to E.B.F., principal investigator), and UNC-Michigan Peer Support Core of the Michigan Center for Diabetes Translational Research grant P30 DK092926.

The Merck Foundation and Sanofi China were not involved in the design of the study, selection of outcome measures or other methods, analyses of study data, or preparation of the manuscript.

Duality of Interest. E.B.F. has received consultation fees from Eli Lilly. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.L., J.T., L.S., and E.B.F. researched data, contributed to the discussion, and wrote the first draft of the manuscript. C.C. and W.J. reviewed and edited the manuscript. P.Y.T., M.M.C., H.C., M.S.E., Y.Q., W.Y., Xiaoy.W., and Xiaob.W. contributed to the discussion and reviewed and edited the manuscript. All authors approved the final version of the manuscript. E.B.F. and W.J. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were John B. Buse and Stephanie L. Fitzpatrick.