Study setting
In Indonesia, most people with diabetes receive their regular diabetes treatment at the primary-care level at public health posts called
puskesmas. These are government-mandated primary-care providers and the first point of contact for people seeking care in the public health system in Indonesia. Each
puskesmas normally serves one sub-district, which has a population of 30,000 to 50,000. The main role of
puskesmas is the delivery of primary outpatient care, but they are also used to promote and realize public health measures such as immunization, nutrition education and health information campaigns. About one-third of
puskesmas in Indonesia provide basic inpatient care for emergency obstetric and neonatal care. Outpatient care efforts have mainly focused on communicable diseases, so that many
puskesmas, especially in rural areas, provide a low quality of diabetes services [
2]. This is characterized by a limited capacity to detect diabetes via diagnostic tests, and to treat and manage diabetes and its complications. Diabetes care is mainly limited to the use of medication to control the disease, and mostly disregards patient education on topics such as diet or exercise [
2,
3]. The diabetes burden in Aceh is mostly unknown, but judging from national data, substantial [
2]. This study is taking place at
puskesmas in the districts Banda Aceh and Aceh Besar in the north of Sumatra, with Banda Aceh consisting of mostly urban areas and Aceh Besar of rural areas.
Study design
The study is a cluster randomized controlled trial with a parallel group design, with the clusters consisting of
puskesmas from Banda Aceh and Aceh Besar Regency.
1 Peer education groups were randomly established in 50% of
puskesmas after they joined the study and baseline data collection had been terminated. This study design allows us to establish two groups of participants (treatment and control group) so that we can causally identify the effect of the peer education on the primary and secondary outcomes. Blinding was possible at baseline, which preceded the allocation of clusters into treatment and control groups. However, after the randomization and allocation, the blinding of the treatment allocation of participants, project managers, and investigators is no longer possible. For the duration of the study, the control group will not receive any intervention beyond being informed about their blood test results. The Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) checklist is provided as an Additional file
1.
Intervention: peer education
Peer support has been defined as “support from a person who possesses experiential knowledge of a specific behavior or stressor and similar characteristics as the target population” [
9]. It has been shown to help reduce or prevent problematic health behaviors and alleviate vascular disease, HIV, and Parkinson’s disease among others [
5]. The appeal of peer support is its ability to create nonhierarchical reciprocal relationships through the sharing of similar life experiences between the peer educator and the peers. Additionally, because peer education relies on non-professionals to improve the health outcomes of patients, it may be significantly less resource intensive than trying to achieve the same effects with professional health workers.
The intervention was designed in cooperation with the local expert team as well as experts experienced with the implementation of peer education in a low-income context in Mali [
10]. Furthermore, qualitative interviews and focus group discussions with nurses working with diabetes patients at
puskesmas informed the intervention design, in particular regarding practicable ways to train peer educators and to provide them with the means to transfer their knowledge successfully to their peer groups.
Training of peer educators
Selected peer educators will receive training before and after the start of the implementation. A 2-day intensive training session by local physicians on diabetes and nutrition was carried out at the beginning of April 2019 before the start of the peer education sessions. It provided general information about diabetes as a disease, its risks, and the ways to treat it. This initial information session will be followed up by monthly half-day training sessions for the peer educators until the end of the study. These training sessions for peer educators will be led by two specially trained nurses, who, prior to each training session, will be educated by a member of the research team on the specific topic to be discussed. The topics and structure of the peer educator training sessions will be guided by the peer leader manual published by the International Diabetes Federation [
11]. The goals of these additional education sessions [
1] are to distribute the burden of training sessions for peer educators over a longer period of time [
2], to maintain the motivation and commitment of peer educators over time [
3], and to use feedback from the peer educators after their peer education sessions to adapt the training to the needs of the peer educators and their peers.
Use of peer education
Depending on the number of patients recruited per puskesmas and the number of potential peer educators, one or two peer educators will be selected per puskesmas, to limit the group size to 13 participants. Peer education sessions are planned to be held once a month for 18 months. They will be conducted by the peer educator only, without the presence of a trained nurse or a member of the research staff, to avoid any potential changes in the behavior of the peer educator or the patients. To preserve some flexibility in how peer educators conduct the peer education sessions and react to the needs of the group, we will refrain from closely monitoring each session. Rather, we will use feedback from the peer educators about the sessions to determine if they are successful in discussing the planned topics.
Outcome measures
The primary outcome is the change in HbA1c levels from baseline to the final assessment. HbA1c will be collected at baseline, midline and the final assessment using point-of-care testing devices allowing for the immediate measurement of HbA1c levels. Secondary outcomes are:
-
changes in lipids (total cholesterol, high-density lipoprotein, and triglycerides) collected using point-of-care testing devices at baseline, during the trial, and at the final assessment
-
blood pressure
-
waist circumference
-
diabetes knowledge
-
medication adherence (five-item Medication Adherence Scale, MARS-5 [
12])
-
diabetes distress (Diabetes Distress Scale 2 [
13])
-
healthy behaviors, such as smoking status and number of cigarettes per day, and physical activity levels (WHO global physical activity questionnaire [
14])
-
dietary diversity (dietary diversity questionnaire published by the Food and Agriculture Organization of the U.N. [
15])
Finally, to assess the cost-effectiveness of the intervention, changes in health-care costs and changes in quality-adjusted life years (based on EQ-5D-3 L questionnaire [
16]) will be used to calculate the incremental cost-effectiveness ratio of the intervention [
16].
Regarding potential mediators, we specifically focus on the behavioral characteristics of the participants. In particular, we measure risk and time preferences as well as trust in other people using unincentivised questions. We also use the Collective Self-Esteem Scale, which measures the ability of participants to function in and identify with social groups [
17], and the 13-item Self-Control Scale [
18].
Number of participants and power calculation
The study aimed to include patients from all
puskesmas in Banda Aceh and Aceh Besar.
2 Power calculations indicated that with 680 participants in 34 clusters, an assumed intra-cluster correlation of HbA1c of 0.37, and a pooled standard deviation of HbA1c of 1.25, we would have 80% power to detect a difference of 0.5 in mean HbA1c using a significance level of
p = 0.05 and 90% power to detect a difference of 0.686 using a significance level of
p = 0.01. We chose a difference of 0.5 in mean HbA1c, as this has been determined as a clinically meaningful threshold [
19]. At the end of the recruitment phase, we had been able to recruit 534 participants from 31
puskesmas. However, two
puskesmas could not be used as separate cluster units since their low recruitment numbers prevented the formation of a peer education group. We decided to assign the participants from these
puskesmas to the geographically closest
puskesmas. This reduced the number of clusters from 31 to 29, which gives 80% power to detect a difference of 0.57 in mean HbA1c using a significance level of
p = 0.05 and 90% power to detect a difference of 0.78 using a significance level of
p = 0.01.
Inclusion criteria
While previous studies have shown that there is a real benefit of peer education for people with very elevated HbA1c levels, we aim to include any person with type 2 diabetes. There are two main reasons for this decision. First, many puskesmas in Indonesia are currently not equipped to test HbA1c levels and would, hence, be unable to select patients based on their HbA1c levels. We, therefore, think that our intervention should reflect the reality of the current health system and investigate the suitability of measuring HbA1c levels for later implementation. Second, we suspect that there is also a benefit of joining peer education for those with better HbA1c levels, by helping them to prevent a worsening of the disease over time.
Therefore, we used the following inclusion criteria for patients in the peer education groups or the control groups:
-
Patients treated in puskesmas for type 2 diabetes in the intervention area
-
Patients who agreed to undergo the whole process of peer education
-
Patients who agreed to carry out all biological and survey assessments as required by the protocol
-
Patients aged 20–79 years
-
Patients not enrolled in another research program
The peer educators will need to fulfill the following requirements:
-
Can commit to attending 20 h of training
-
Are willing to organize activities with other patients every month
-
Have basic diabetes self-management knowledge and supportive non-judgmental communication skills
-
Willing to lead
-
Literate
Study procedures
The project has several phases. First, to inform the intervention design, qualitative interviews with nurses and diabetes patients were carried out. Then, for each puskesmas, patients with diabetes were recruited into the peer education program. An interview was carried out in each facility to receive the facility head’s consent to participate, to gather general information and information specific to diabetes treatment and, if possible, to obtain a list of diabetes patients. Additionally, health workers in the puskesmas and in the villages as well as patients who had already been interviewed were asked to suggest further people with diabetes who were associated with the relevant puskesmas.
At baseline, the patients recruited were interviewed by trained enumerators using a questionnaire. They were invited to the puskesmas on one predetermined day, during which the trained enumerators used point-of-care devices to test their HbA1c and lipid levels. The participants were informed of their test results. Blood was then drawn from the arm of participants by health-care professionals, which was used for laboratory as well as point-of-care tests of HbA1c and lipid profiles to determine the accuracy of the point-of-care test devices in the study setting under field conditions. In the middle and final assessments, only point-of-care tests using venous blood will be carried out.
In choosing potential peer educators, all participants were asked during the interviews if they would be interested in serving as a peer educator. Further, health facility staff were asked during the facility interview to suggest patients for this role. Out of the participants recruited, we determined potential peer educators based on three criteria: (1) their willingness to take on this role, (2) a recommendation from the health facility staff, and (3) how well they were already controlling their diabetes based on the HbA1c level from the baseline data.
Following data collection, the
puskesmas were randomized. Because recruitment took place before randomization, the researchers, participants, and facility managers did not know which facilities would be allocated to the treatment and control groups. We used covariate-constrained randomization to ensure the treatment and control arms were balanced in terms of baseline covariates while maintaining the randomness of the allocation [
20,
21]. The covariates took into account the composition of the group with regards to size of the group, and the age, sex, education, participation in other health programs,
3 and diabetes control (mean HbA1c values) of the members, as well as the location of the
puskesmas (rural or urban). Randomization was carried out using the statistics program Stata.
Peer education groups were formed at each puskesmas in the treatment group. Peer educators will receive continuous training in diabetes management skills and will be supported in the administrative processes (e.g., room booking) required to run their peer education group in their local sub-district.
To ensure that the effect of using peer education can be observed in comparison to normal patient education efforts, the control group will not receive any additional education apart from standard therapy.
As detailed in Fig.
1, 9 months after the first peer education session, a first follow-up will assess the implementation and the effects of peer education. The final assessment will take place 18 months after the first peer education session to assess the long-term effects of the intervention and the experience of participants.
We will then compare the outcomes between treatment and control groups, carry out a cost-effectiveness analysis, present the results to local health authorities, and discuss how to implement the program more widely if it proves to be effective.
Economic evaluation
Costs related to the implementation of peer education will be collected from the the documented project expenditures. Data on resource use for diabetes treatment and care, including equipment, drugs, and doctor and hospital visits, will be collected retrospectively using cost information from the health insurance provider if possible, or via the expert opinion of local specialists. These data will be supplemented with information on participants’ health-care seeking behavior and utilization, drug use, and out-of-pocket payments collected through the surveys.
To estimate cost-effectiveness, the incremental cost-effectiveness ratio will be calculated from the health-care system perspective and the societal perspective taking into account out-of-pocket expenditures and costs related to participating in peer education. The incremental cost-effectiveness ratio is defined by the difference in costs between receiving no intervention plus standard diabetes care and the standard care plus intervention costs, divided by the difference in their effect. Here it represents the average incremental cost for peer education associated with one additional quality-adjusted life year.
Statistical methods
Random sampling and random assignment of intervention units to the treatment and control groups will allow us to interpret differences in mean outcomes as causal effects of the intervention. By comparing outcomes between participants in the peer education groups compared to study participants in the control group, we will be able to draw conclusions of the intent-to-treat effect and the size of the effect.
Because covariate-constrained randomization was used to balance the baseline characteristics, it is necessary to account for this design at the analysis stage [
21,
22]. We will use general (generalized) linear mixed models to incorporate data structures that are both hierarchical and longitudinal and adjust for the covariates used during the covariate-constrained randomization [
21,
22]. Standard errors will be clustered at the intervention unit level (
puskesmas).