A Randomized Controlled Trial in a 14-Month Longitudinal Design to Analyze the Effects of a Peer Support Instant Messaging Service Intervention to Improve Diabetes Self-Management and Support

Abstract

Background/Objectives: The outcomes of diabetes therapy depend largely on how well patients can implement medical advice in their lives. The main aim of the DiabPeerS study was to evaluate a peer support instant messaging service (IMS) approach to diabetes self-management education and support (DSMES) for people with type 2 diabetes mellitus (T2DM). Methods: Participants with T2DM took part in a randomized controlled trial. Both the intervention group (IG) and the control group (CG) received standard therapy, but the IG additionally participated in the peer support IMS intervention. The duration of the intervention was 7 months, succeeded by a follow-up 7 months later. Eleven biochemical, six behavioral, and six psychosocial parameters were measured at four times. Results: The targeted sample size could not be reached, and 68 participants took part. The following results have been found for the main hypotheses: No influence on HbA1c was detected (IG: −0.27, CG: +0.06, p > 0.05). Diabetes self-management behaviors were unaffected (IG_diet: +0.02, CG_diet: +0.46, p > 0.05; IG_exercise: −0.72, CG_exercise: +0.44, p > 0.05; IG_bloodsugar: −0.21, CG_bloodsugar: +0.65, p > 0.05; IG_footcare: +0.37, CG_footcare: +1.13, p > 0.05). Quality of life increased during the intervention in both the IG (KSK: +8.92, PSK: +7.41, p < 0.001) and the CG (KSK: +8.73, PSK: +7.48, p < 0.001). Medication adherence increased in the IG (+3.31, p < 0.01), although these participants were still classified as non-adherent. Conclusions: A peer support IMS intervention is a promising approach, but we recommend combining the online setting with an initial face-to-face situation.

1. Introduction

In type 2 diabetes mellitus (T2DM), which is the most common type of diabetes, hyperglycemia results from an inadequate production of insulin and insulin resistance. In the long run, diabetes can damage the heart, blood vessels, eyes, kidneys, and nerves and increase the risk of cardiovascular diseases. Global health organizations such as the World Health Organization [1] and the International Diabetes Federation [2] as well as leading scientific works like the Global Burden of Disease Study [3] or the Non-Communicable Diseases Risk Factor Collaboration [4] highlight that T2DM is a rising threat to health at a global level. Furthermore, scientific data points to a further increase in its prevalence over the next decades. In figures, this means that 537 million adults worldwide suffered from diabetes mellitus in 2021, and a further increase to 783 million people with diabetes is estimated by 2045 [2]. While diabetes was in fourteenth place of leading causes of death worldwide in 1990, it subsequently moved up the list and was in tenth place in 2021. Furthermore, diabetes was in fifth place among leading causes of disease burden in 2022 and is predicted to rise to third place by 2050 [3]. Besides these individual consequences, T2DM costs USD 10,801 per capita in high-income countries, and USD 242 per capita in low-income countries. These expenses relate to direct costs (e.g., medical management), indirect costs (e.g., loss of productivity), and intangible expenses (e.g., reduced quality of life [5]).

Due to the chronic character of diabetes, it requires continuous therapy, regular medical appointments, and good adherence in those affected. Consequently, diabetes self-management (DSM)—which refers to patients’ ability to manage symptoms, therapy, physical and psychosocial consequences, and lifestyle changes inherent in living with a disease [6]—plays a major role in this context. As a consequence, diabetes self-management education (DSME) and diabetes self-management education and support (DSMES) are core elements of person-centered diabetes care [7]. The DSMES consent report defines DSME as “The ongoing process of facilitating the knowledge, skills, and ability necessary for diabetes self-management as well as activities that assist a person in implementing …” these behaviors, and DSMES as the support to “… implement informed decision making, self-management behaviors, problem solving, and active collaboration with the health care team …”. Furthermore, the report describes four critical times for DSMES, which are (1) at diagnosis, (2) annually and/or when not meeting treatment targets, (3) when complicating factors develop, and (4) when transitions in life and care occur [8]. Undoubtably, DSMES has several benefits without negative side effects. These are, on the one hand, reductions related to HbA1c (=glycated hemoglobin; it reflects the average plasma glucose level over the previous 8–12 weeks and is the preferred test for diabetes control [9]), mortality, hypoglycemia, diabetes-related distress, and hospital stays. On the other hand, DSMES provides essential diabetes-related education and support leading to a healthier lifestyle, weight maintenance or loss, increased self-efficacy, empowerment, healthy coping, and overall improved quality of life [8,10,11,12].

Support can be provided by several stakeholders like healthcare professionals, family members, friends, or others living with the same disease (i.e., peers) [13,14]. Therefore, DSMES provided by peers could be a cost-effective way to reduce the burden on healthcare systems as the prevalence of diabetes increases, and to provide evidence-based treatment for people with diabetes. Peer support can be provided in different ways including face to face, telephone-based, web-based [15], or via other new technologies like instant messaging services (IMS).

IMS has two major advantages compared to other mHealth solutions: Firstly, the majority (84.4%) of the age group of 45–64 years, which is most affected by the onset of T2DM, use IMS, while only 51.0% of them use social media tools [16]. Secondly, compared to standard therapy, long-term support via IMS can be provided easily [17] and promptly [18], is inexpensive [17], and requires less effort to attend [19,20]. These factors are particularly relevant in the context of rising prevalence, rising healthcare costs, and a shortage of healthcare professionals. In addition, time- and location-independent support services for people with diabetes are a promising approach for less urbanized regions, such as Lower Austria. In contrast, at the time the trial was designed and registered, we were not aware of any peer support IMS trial.

Therefore, the aim of the present DiabPeerS study was to analyze the effects of a seven-month peer support IMS intervention in addition to antidiabetic therapy according to current Austrian guidelines [21] (hereafter referred to as ‘standard therapy’) on HbA1c in patients with T2DM.

Our pre-registered main hypotheses were:

• H1: Peer support IMS intervention reduces HbA1c in patients with T2DM compared to standard therapy.
• H2: Peer support IMS intervention helps to maintain diabetes self-management behaviors in patients with T2DM compared to standard therapy.
• H3: Peer support IMS intervention improves the quality of life of patients with T2DM compared to standard therapy.
• H4: Peer support IMS intervention improves medication adherence in patients with T2DM compared to standard therapy.
• H5: Extraversion correlates positively with the benefits of peer support IMS intervention as measured by the frequency of IMS usage, quality of life, and HbA1c levels.

2. Materials and Methods

2.1. Study Design and Setting

The DiabPeerS study (i.e., “improving glycaemic control in patients with T2DM through peer support instant messaging: a randomized controlled trial”) was a randomized controlled trial addressing DSMES in persons recently diagnosed with T2DM. It was conducted in Lower Austria by the St. Pölten University of Applied Sciences, the Karl Landsteiner University of Health Sciences, and the Austrian Health Insurance Fund in Lower Austria from 2020 to 2024. This study was approved by the Ethics Committee of Lower Austria (Ethics number: GS4-EK-4/569-2018; approval date: 26 March 2019) and is registered at ClinicalTrials.gov (ClinicalTrials.gov Identifier: NCT04797429). The DiabPeerS intervention design, study recruitment, and measures have been reported elsewhere [22]. In brief, the IG and CG both received standard therapy, but the IG additionally participated in the peer support IMS intervention. Within the intervention, the peer support aspect was realized and facilitated via moderators who guided IMS groups, meaning initiating and moderating the IMS exchange and discussion. These moderators were people experienced in the management of their own T2DM, received specific training by the study staff for their task, and were continuously guided by a dietitian. The duration of the intervention was 7 months, succeeded by a follow-up 7 months later. Biochemical, behavioral, and psychosocial parameters were assessed four times during this period: at baseline (T0), three months after the start of the intervention (T1; T0 + 3 months), at the end of the intervention (T2; T0 + 7 months), and at the follow-up 7 months after the intervention (T3; T0 + 14 months).

Written, informed consent to participate was obtained from all participants.

Participants were eligible to participate if they (1) were over 40 years of age; (2) were diagnosed with T2DM [21] and participated in standard therapy [21]; (3) had an HbA1c level of ≥6.5% (48 mmol/mol); and (4) had been receiving oral hyperglycemic agents for a maximum of three years prior to the start of the trial. Inclusion criteria for moderators were (1) being over 60 years of age (based on the assumption that this age group is retired and has more time to devote to this than working people); (2) having been diagnosed with T2DM [21]; (3) having an HbA1c level of ≥6.5% (48 mmol/mol); (4) participating in standard therapy [21]; having received oral hyperglycemic agents for a minimum of three years prior to the start of the trial; and (5) regularly participating in the Austrian disease management program ‘Therapie aktiv—Diabetes im Griff’ [‘Active therapy—Diabetes under control’]. Exclusion criteria for both participants and moderators were (1) insulin dependency; (2) hospitalization or vacation of more than 3 weeks during the intervention; (3) eye disorders that severely limit vision and, hence, inability to read the display; (4) severe illnesses such as kidney, liver, heart disease, malignant cancer, or neurological or mental illness; (5) illicit drug use or non-medical use of prescription drugs; (6) limited German language skills; and (7) pregnancy.

An a priori calculation was performed using G-Power [23] (version 3.1.) for the primary outcome variable HbA1c, using an ANCOVA at 5% level of significance and 80% power. Based on previous research, we assumed a small to medium effect size for group differences in HbA1c means (partial eta squared = 0.05, which corresponds to an effect size f = 0.229) [24,25,26]. Considering a 30% dropout ratio, the goal was to include 98 randomly assigned participants in the IG as well as the CG (196 in total). Furthermore, nine moderators were to be additionally recruited (205 persons in total, 196 participants plus nine moderators).

2.2. Target Outcomes

Based on the defined hypotheses, the primary outcome was the HbA1c [%] which was measured using the TOSOH G8 (Sysmex Austria GmbH, Vienna, Austria) apparatus.

Secondary outcomes were:

Psychosocial parameters:

• Social support was measured using the ‘Fragebogen zur Sozialen Unterstützung’ (F-SozU) [‘Questionnaire on social support’], [27]: the F-SozU operationalizes social support as perceived or anticipated support from the social environment. The short form consists of the following subscales: ‘emotional support’, ‘practical support’, ‘social integration’, ‘stress from the social network’. The F-SozU consists of14 items using a five-point Likert scale with the endpoints ‘1’ (does not apply) and ‘5’ (accurate).
• Self-efficacy was measured using the ‘General Self-Efficacy Scale’ (GSE) [28]: The GES consists of ten items designed on a four-point Likert scale with the endpoints ‘1’ (not at all true) and ‘4’ (completely true) and assesses optimistic self-beliefs to cope with several challenges in life.
• Depression was measured using the ‘Patient Health Questionnaire-9’ (PHQ-9) [29]: The PHQ-9 asks for all nine criteria of depression as defined in the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) using a four-point Likert scale with the endpoints ‘0’ (not at all) and ‘3’ (nearly every day).
• Diabetes distress was measured using the ‘Diabetes Distress Scale’ (DDS) [30]: The DDS includes four dimensions of distress (‘emotional burden’, ‘regimen distress’, ‘interpersonal distress’, ‘physician distress’). The DDS consists of 17 items using a six-point Likert scale with the endpoints ‘1’ (not a problem) and ‘5’ (a very serious problem).
• Quality of life was measured using the ‘Short-Form Health Survey’ (SF-12) [31]: The SF-12 includes eight dimensions (‘physical functioning’, ‘role limitations due to physical problems’, ‘bodily pain’, ‘vitality’, ‘general health perceptions’, ‘social functioning’, ‘role limitations due to emotional problems’, ‘mental health’). The summary scores ‘Physical Summary Scale’ and ‘Psychological Summary Scale’ (0–100 scales) were calculated from the specified scales.
• Diabetes knowledge was measured using the ‘Diabetes Knowledge Test’ (DKT) [32] with forward and retranslation: The DKT consists of 20 statements about diabetes which have to be rated as ‘true’, ‘false’ or ‘I don’t know’. Based on the answers, a difficulty index (percentage of patients who scored correctly) was calculated.

Behavioral parameters:

• Medication adherence was measured using the ‘A14-scale’ [33]: The A14 consists of 14 items of non-adherent behaviors phrased in a non-threatening and non-judgemental way using a five-item Likert scale with the endpoints ‘4’ (never) to ‘0’ (very often).
• Dietary behavior and alcohol consumption during the past four weeks was measured based on a food frequency questionnaire using a nine-item Likert scale with the endpoints ‘0’ (never) to ‘9’ (more than four times a day).
• Smoking was measured using the questionnaire ‘Aktivrauchen—Kurzversion (Erwachsene)’ [‘Active smoking—short version (adults)’] [34]: The questionnaire assesses the current smoking behavior per week and that of the last six months. Participants indicated the number of cigarettes, cigarillos, cigars, or amount of tobacco (grams) used for pipes or hand-rolled cigarettes.
• Physical activity was measured using the ‘International Physical Activity Questionnaire Short Form’ (IPAQ-SF) [35,36]: The IPAQ-SF asks seven questions to assess ‘vigorous-intensity’ and ‘moderate-intensity’ physical activity as well as ‘walking’ and ‘sitting’. Participants indicate the time in minutes or hours for each activity level. Based on this information, three levels of physical activity (low, moderate, high) were calculated and expressed in the metabolic equivalent of task (MET) minutes per week.
• Diabetes self-management behaviors were measured using the ‘Summary of Diabetes Self-Care Activities German’ (SDSCA-G) [37]: The SDSCA-G focuses on the past seven days related to the diabetes self-care activities ‘nutrition’, ‘physical activities’, ‘blood glucose testing’, ‘footcare’, and ‘smoking’. The SDSCA-G consists of 11 items, and participants marked the number of days that the respective behavior was performed on an eight-point Likert scale with the endpoints ‘0’ (0 days) to ‘7’ (7 days). While the first ten items were calculated to a score and four sub scores (diet, exercise, blood glucose testing, foot care), the eleventh item focuses on smoking habits.
• Clinic and communication visits (health professional visits in the past six months, hospital stays in the past six months).

Biochemical parameters:

• Fasting blood glucose [mg/dL], total cholesterol [mg/dL], high-density lipoprotein (HDL), low-density lipoprotein (LDL) [mg/dL], and triglycerides [mg/dL] [mg/dL] were measured using Cobas 8000 (Roche Austria GmbH, Vienna, Austria).
• Blood pressure [mmHg] was measured using Boso Medicus uno OA.
• Body height [cm], body weight [kg], and body fat [%] were measured using Seca mBCA 555 (seca gmbh & co. kg, Hamburg, Germany). The device measures body weight with a scale, body height with ultrasound length measurement, and body composition by the voltage drop of the alternating current in one step. Body height and body weight were converted into Body Mass Index (kg/m²) and categorized according to the WHO classification [38]. Waist circumference [cm] to calculate the waist-to-height ratio and visceral body fat (Seca mBCA 555) was measured using an ergonomic, stepless, and extendable measuring tape.

Additionally, demographic data and personality traits were surveyed once during the baseline assessment (T0). Demographic data include age, gender, education, living arrangement (marital status), income, employment status, immigration background [39], and clinical information (age at diagnosis, on insulin, other medical therapies, list of prescribed medications). Personal traits were assessed using the Big Five Inventory—German form (BFI-2) [40], which consists of 60 items using a five-item Likert scale with the endpoints ‘1’ (strongly disagree) to ‘5’ (strongly agree).

Additionally, participation in the disease management program ‘Therapie aktiv—Diabetes im Griff’, drug abuse as well as pregnancy for non-menopausal women were evaluated at each measurement point.

Communication data (real-time):

Communication protocols containing the communication data of the different IMS groups were retrieved on a daily basis. Chat logs were exported as Excel files. Numbers of responses (replies, reactions, reply actions) were calculated to quantify user engagement.

2.3. Statistical Analyses

Statistical analyses were performed using SPSS version 29 (SPSS Inc., Chicago, IL, USA). Except for the IPAQ-SF, all internal consistencies of the questionnaires used were satisfactory (see Table S1). Therefore, the IPAQ-SF was excluded from further analyses, and physical activity was evaluated only via the SDSCA-G and the subscale “Exercise”.

Continuous variables were reported as mean with standard deviation, and ordinal variables as numbers (percentage). Regarding the analysis of differences in demographic parameters between the IG and the CG, the mean differences between the two groups were calculated using the t-test for independent samples. Where the criteria were not met, these analyses were performed with the Mann-Whitney U-test. Contingency tables with χ²-tests and standardized residuals were created for statements on distribution differences in nominal scaled variables. Where the requirements for χ²-tests were not met (e.g., expected frequencies less than 5), Fisher’s exact test was calculated.

Regarding the main hypotheses, tests for homogeneity of variances were all non-significant (Mauchly-W test: all ps > 0.069; except Medication adherence although Greenhouse-Geisser not significant). Therefore, we used parametric procedures (e.g., repeated measures ANOVA) for analyzing the five hypotheses. Hypotheses were analyzed according to the intention-to-treat principle.

Missing Data

To test whether missing values were at random, MCAR tests (missing completely at random) were calculated. Out of 37 tests, only three were significant (DKT T1, SF-12 T3), which would have been expected based on an α error rate of 5% (remaining MCAR tests: χ² = 2.6–498.5; all ps > 0.063; for detailed results, see Table S2).

3. Results

Unfortunately, and despite extensive efforts, it was not possible to recruit the calculated number of participants based on the power analysis. The study started with only 41 participants in the IG and 27 participants in the CG. None of the participants had to be excluded due to any of the outlined exclusion criteria (see Section 2.1). 47 participants stayed until the end of the study (68% men, 32% women). 10 participants dropped out after the baseline assessment (T0), 7 participants after the start of the intervention phase (T1), and 4 participants after the end of the intervention phase (T2) (see Figure 1).

3.1. Dropout Analysis

A detailed dropout analysis did not reveal any substantial differences regarding biochemical parameters (Table S3), behavioral parameters (Table S4), psychosocial parameters (Table S5), or the variables relevant for the hypotheses (Table S6). The only exceptions were total cholesterol and LDL which were higher in the dropout group (for details, see Table S3) as well as differences regarding physical activity (Exercise, Quality of life—physical summary scale; see Table S6). What was also interesting was that participants in the dropout group gradually deteriorated in their diabetes knowledge over time prior to dropout (Table S5), which suggests a declining commitment to the study. Also, no dropout differences were found for full responders between the IG and the CG (IG: from 41 to 27; CG: from 27 to 20; χ² = 1.09, p = 0.779).

Concerning the drop-out causes, the majority stopped their participation due to health reasons (e.g., discovery of cancer), IMS participation-generated stress, technical problems, and disappointment because they were not in the IG.

3.2. DiabPeerS Participants

As for general information, 69.10% of all participants were male, and 30.90% were female with an average age of 61.65 years (SD = 8.61 years). The majority were married/in a partnership (76.50% of participants), had a completed apprenticeship as their highest educational level (33.80% of participants), and were retired (50.00% of participants). The IG and the CG did not differ significantly in terms of demographic factors (see

Table 1. Summary of participant demographics (n = 68).

	IG	CG
	M (SD)	M (SD)
Age (years)	62.00 (8.54)	61.11 (8.85)
	p = 0.816
HbA1c (%)	7.33 (0.98)	7.16 (0.94)
	p = 0.543
	n (%)	n (%)
Gender
Male	27 (65.9)	20 (74.1)
Female	14 (34.1)	7 (25.9)
Diverse	0 (0.0)	0 (0.0)
	p = 0.473
BMI according to [38]
Underweight	0 (0.0)	0 (0.0)
Normal weight	4 (14.8)	2 (10.0)
Overweight	7 (25.9)	6 (30.0)
Obesity	14 (51.9)	12 (60.0)
Missing *	2 (7.4)	0 (0.0)
	p = 0.543
Family status
Single	3 (7.3)	2 (7.4)
Married/partnership	33 (80.5)	19 (70.4)
Separated	1 (2.4)	0 (0.0)
Divorced	1 (2.4)	3 (11.1)
Widowed	1 (2.4)	2 (7.4)
Other/missing	1 (2.4)	1 (3.7)
	p = 0.422
Education
No formal education	0 (0.0)	1 (3.7)
Secondary school	4 (9.8)	2 (7.4)
Apprenticeship	14 (34.1)	9 (33.3)
Master craftsman qualification	4 (9.8)	2 (7.4)
A-level	11 (26.8)	5 (18.5)
University degree	3 (7.3)	6 (20.2)
Other/missing	5 (12.2)	2 (7.4)
	p = 0.563
Occupation
Employed	15 (36.6)	10 (37.0)
Self-employed	3 (7.3)	2 (7.4)
Unemployed	1 (2.4)	1 (3.7)
Retired	20 (48.8)	14 (51.9)
Other/missing	2 (4.8)	0 (0.0)
	p = 0.943
Net household income
>1000 €	2 (4.9)	2 (7.4)
1000–2000 €	11 (26.8)	7 (25.9)
2000–3000 €	11 (26.8)	8 (29.6)
3000–4000 €	6 (14.6)	5 (18.5)
<4000 €	8 (19.5)	5 (18.5)
Missing	3 (7.3)	0 (0.0)
	p = 0.990
Immigration background †
Yes	0 (0.0)	1 (3.7)
No	33 (80.5)	22 (81.5)
Missing	8 (19.5)	4 (14.8)
	p = 0.411

Note. HbA1c = glycated hemoglobin; BMI = Body Mass Index; € = Euro. * BIA measurement was not possible due to contraindication (cardiac pacemaker); † immigration background = both parents born outside Austria [39]. p-Values refer to the results of a χ² test for group differences.

and Table S7) and the study’s variables (see Tables S8 and S9).

In general, the participants were all rather well-adjusted regarding their HbA1c values when compared to the Austrian guidelines [41] (see Table 1 and

Table 2. Mean and standard deviation of all hypotheses-relevant scales over time.

			T0		T1		T2		T3
		N	M (SD)	N	M (SD)	N	M (SD)	N	M (SD)
HbA1c (H1)	IG	41	7.33 (0.98)	31	7.05 (0.59)	27	7.04 (0.67)	27	7.06 (0.77)
	CG	25	7.16 (0.94)	24	7.21 (1.08)	22	6.79 (0.81)	19	7.22 (1.86)
SDSCA-G: Diet (H2)	IG	40	3.72 (1.47)	33	4.25 (1.35)	28	3.96 (1.50)	27	3.74 (1.40)
	CG	27	4.08 (1.58)	23	4.21 (1.21)	22	4.17 (1.32)	20	4.54 (1.57)
SDSCA-G: Exercise (H2)	IG	41	3.22 (1.82)	33	3.08 (1.95)	29	3.31 (1.94)	27	3.09 (1.92)
	CG	27	3.78 (1.87)	23	3.85 (1.77)	22	4.18 (1.83)	20	3.30 (1.68)
SDSCA-G: BST (H2)	IG	39	3.49 (2.91)	32	4.31 (2.75)	29	3.53 (2.87)	26	3.90 (2.74)
	CG	26	2.90 (2.76)	23	3.04 (3.00)	21	3.45 (2.78)	20	3.45 (2.91)
SDSCA-G: Footcare (H2)	IG	41	1.66 (1.99)	33	2.00 (1.86)	29	1.98 (1.81)	27	2.35 (2.05)
	CG	27	2.37 (2.34)	23	2.37 (2.34)	22	2.59 (2.38)	20	2.83 (2.56)
SDSCA-G: Overall (H2)	IG	38	3.09 (1.27)	32	3.57 (1.13)	28	3.34 (1.36)	26	3.30 (1.08)
	CG	26	3.41 (1.37)	23	3.57 (1.32)	21	3.75 (1.45)	20	3.73 (1.32)
SF-12: QoL—KSK (H3)	IG	41	41.41 (5.36)	41	43.71 (7.76)	39	49.78 (7.72)	41	50.33 (6.57)
	CG	27	41.70 (5.08)	27	44.20 (6.88)	26	50.21 (6.36)	27	50.43 (6.30)
SF-12: QoL—PSK (H3)	IG	41	49.88 (5.66)	41	51.97 (6.57)	39	55.03 (7.70)	41	57.36 (6.75)
	CG	27	47.33 (6.97)	27	48.26 (8.21)	26	53.83 (9.71)	27	54.74 (8.99)
A14: Medication ad. (H4)	IG	36	45.81 (9.59)	31	45.26 (12.03)	28	46.71 (10.11)	26	49.12 (6.68)
	CG	26	49.27 (7.99)	22	51.41 (5.04)	21	51.95 (5.04)	19	50.84 (8.04)

Note. HbA1c < 7% is well adjusted, meaning adequate micro- and macrovascular protection [41] IG = Intervention Group; CG = Control Group. SF12 = Short Form Health Survey, KSK = Physical Summary Scale; PSK = Psychological Summary Scale (standardized mean = 50, SD = 10 (range 0–100)). A14 = Medical Adherence (non-adherent score < 50, adherent score ≥ 50). SDSCA-G = Summary of Diabetes Self-Care Activities (All scale scores range from 0 to 7 with higher scores suggesting better self-management). BST = Blood sugar test.

).

3.3. Verification of DiabPeerS Hypotheses

The main hypotheses are verified in the following paragraphs.

Hypothesis 1.

Peer support IMS intervention reduces HbA1c of patients with T2DM compared to standard therapy.

As can be seen in

Table 3. Results of the mixed analysis of variance for Hypotheses 1 to 4.

	Group/Between	Pre-Post/Within	Interaction
Hypothesis 1
HbA1c	F(1.38) = 0.06, ηp2 = 0.002	F(3.114) = 2.08, ηp2 = 0.052	F(3.114) =1.40, ηp2 = 0.036
Hypothesis 2
SDSCA-G: Diet	F(1.40) = 0.78, ηp2 = 0.019	F(3.120) = 0.86, ηp2 = 0.021	F(3.120) = 3.87 **, ηp2 = 0.088
SDSCA-G: Exercise	F(1.42) = 1.11, ηp2 = 0.026	F(3.126) = 3.25 *, ηp2 = 0.072	F(3.126) = 1.52, ηp2 = 0.035
SDSCA-G: Blood sugar	F(1.39) < 0.01, ηp2 < 0.001	F(3.117) = 0.44, ηp2 = 0.011	F(3.117) = 1.98, ηp2 = 0.048
SDSCA-G: Footcare	F(1.42) = 2.95, ηp2 = 0.066	F(3.126) = 0.64, ηp2 = 0.015	F(3.126) = 0.50, ηp2 = 0.012
SDSCA-G: Overall	F(1.37) = 2.87, ηp2 = 0.072	F(3.111) = 0.93, ηp2 = 0.024	F(3.111) = 1.54, ηp2 = 0.040
Hypothesis 3
SF-12: QoL—KSK	F(1.63) = 0.16, ηp2 = 0.002	F(3.189) = 41.75 ***, ηp2 = 0.399	F(3.189) = 0.06, ηp2 = 0.001
SF-12: QoL—PSK	F(1.63) = 2.31, ηp2 = 0.035	F(3.189) = 21.48 ***, ηp2 = 0.254	F(3.189) = 0.32, ηp2 = 0.005
Hypothesis 4
A-14: Medication ad.	F(1.34) = 7.46 **, ηp2 = 0.180	F(3.102) = 0.53, ηp2 = 0.015	F(3.102) = 1.36, ηp2 = 0.038

Note. QoL = Quality of Life. SDSCA-G = Summary of Diabetes Self-Care Activities. SF12 = Short Form Health Survey, KSK = Physical Summary Scale; PSK = Psychological Summary Scale. Significant effects are shown in bold. Effect sizes: small: η_p² = 0.010; medium: η_p² = 0.060; large: η_p² = 0.140. * p < 0.05, ** p < 0.01, *** p < 0.001 (one-sided).

and Figure 2, neither main effects nor interaction were significant (for descriptives, see Table 2; for differences between IG and CG, see

Table 4. Differences between IG and CG regarding variables which are relevant for the hypotheses.

	IG	CG
	M (SD)	M (SD)	t	Cohen d
HbA1c [%] (T0)	7.33 (0.98)	7.16 (0.94)	0.674	0.966
HbA1c [%] (T1)	7.05 (0.60)	7.21 (1.08)	−0.704	0.841
HbA1c [%] (T2)	7.04 (0.67)	6.80 (0.81)	1.162	0.737
HbA1c [%] (T3)	7.06 (0.77)	7.22 (1.86)	−0.408	1.325
SDSCA-G: Diet (T0)	3.72 (1.47)	4.08 (1.58)	−0.104	0.024
SDSCA-G: Diet (T1)	4.25 (1.21)	4.21 (1.21)	−1.425	0.034
SDSCA-G: Diet (T2)	3.96 (1.50)	4.17 (1.32)	−1.425	−0.151
SDSCA-G: Diet (T3)	3.74 (1.46)	4.54 (1.57)	−3.285 *	−0.542
SDSCA-G: Exercise (T0)	3.80 (1.80)	2.87 (1.54)	1.413	−0.303
SDSCA-G: Exercise (T1)	3.14 (2.00)	3.76 (1.71)	−1.216	−0.411
SDSCA-G: Exercise (T2)	3.31 (1.93)	4.18 (1.84)	−1.629	−0.461
SDSCA-G: Exercise (T3)	3.08 (1.91)	3.31 (1.70)	−0.436	−0.114
SDSCA-G: Blood sugar (T0)	3.96 (2.95)	3.00 (3.07)	0.815	0.205
SDSCA-G: Blood sugar (T1)	4.02 (2.94)	3.46 (2.88)	0.702	0.444
SDSCA-G: Blood sugar (T2)	3.53 (3.01)	3.45 (2.55)	0.101	0.029
SDSCA-G: Blood sugar (T3)	3.75 (2.87)	3.65 (2.76)	0.119	0.161
SDSCA-G: Footcare (T0)	1.93 (2.17)	1.75 (2.10)	0.221	−0.334
SDSCA-G: Footcare (T1)	1.92 (1.92)	2.65 (2.20)	0.221	−0.262
SDSCA-G: Footcare (T2)	1.97 (1.81)	2.61 (2.37)	−1.318	−0.294
SDSCA-G: Footcare (T3)	2.30 (2.08)	2.88 (2.48)	−0.890	−0.208
SF-12: Quality of life—KSK (T0)	41.41 (5.36)	41.70 (5.08)	−0.222	−0.055
SF-12: Quality of life—KSK (T1)	43.71 (7.76)	44.20 (6.88)	−0.266	−0.066
SF-12: Quality of life—KSK (T2)	49.78 (7.72)	50.21 (6.36)	−0.237	−0.060
SF-12: Quality of life—KSK (T3)	50.33 (6.57)	50.43 (6.30)	−0.059	−0.015
SF-12: Quality of life—PSK (T0)	49.88 (5.66)	47.33 (6.97)	1.657	0.411
SF-12: Quality of life—PSK (T1)	51.97 (6.57)	48.26 (8.21)	2.062	0.511
SF-12: Quality of life—PSK (T2)	55.03 (7.70)	53.83 (9.72)	0.554	0.140
SF-12: Quality of life—PSK (T3)	57.36 (6.75)	54.74 (9.00)	1.372	0.340
A-14: Medication adherence (T0)	45.81 (9.60)	49.27 (8.00)	−1.503	−0.387
A-14: Medication adherence (T1)	45.26 (12.02)	51.41 (5.04)	−2.258	−0.629
A-14: Medication adherence (T2)	46.71 (10.11)	51.95 (5.04)	−2.177	−0.628
A-14: Medication adherence (T3)	49.12 (6.68)	50.84 (8.04)	−0.785	−0.237
BFI: Conscientiousness	3.872 (0.56)	3.73 (0.62)	0.616	0.585
BFI: Neuroticism	2.24 (0.47)	2.31 (0.87)	−0.270	0.670
BFI: Agreeableness	3.82 (0.51)	3.87 (0.38)	−0.269	0.460
BFI: Open-mindedness	3.50 (0.66)	3.77 (0.58)	−1.033	0.632
BFI: Extraversion	3.60 (0.66)	3.30 (0.500)	1.220	0.600

Note. T0 = baseline assessment; T1 = 3 months after the start of the intervention; T2 = post intervention; T3 = 7 months after post-intervention/follow-up. SDSCA-G = Summary of Diabetes Self-Care Activities. SF-12 = Short Form Health Survey. KSK = Physical Summary Scale; PSK = Psychological Summary Scale. BFI = Big Five Inventory. * p < 0.05 (two-sided).

). In general, HbA1c values were above the Austrian recommendations of less than 7% for sufficient micro- and macrovascular protection [41]. If Hypothesis 1 were to be true, we would have expected a significant interaction—which we did not find. Therefore, Hypothesis 1 cannot be confirmed. The peer support IMS intervention did not significantly reduce the HbA1c value in the IG compared to the CG without any peer IMS support.

Hypothesis 2.

Peer support IMS intervention helps to maintain DSM behaviors in patients with T2DM compared to standard therapy.

As can be seen in Table 3 and Figure 3, the subscores for blood sugar, foot care, and the overall SDSCA-G sum were not significant. For the subscores of diet and time, a statistically significant interaction was observed, and for the subscore of exercise a significant decrease for the IG and for the CG (for descriptives, see Table 2; for differences between IG and CG, see Table 4). Nevertheless, although the IG had a better diet behavior than the CG 3 months after the start of the intervention, the positive effect was already gone at the end of the intervention phase (after 7 months = T2), and it was also descriptively lower at T3 (= post-intervention assessment). Therefore, Hypothesis 2 is not supported, either.

Hypothesis 3.

Peer support IMS intervention improves the quality of life of patients with T2DM compared to standard therapy.

As can be seen in Table 3, once again no significant effects were found except for a strong main effect for the retests (for descriptives, see Table 2; for differences between IG and CG, see Table 4). For participants of both groups, the quality of life improved continuously over time. Again, this effect was only expected for the IG, and not the CG. Therefore, Hypothesis 3 cannot be confirmed, neither for the physical health subscale nor the psychological subscale (see Table 3 and Figure 4).

Hypothesis 4.

Peer support IMS intervention improves medication adherence in patients with T2DM compared to standard therapy.

As can be seen in Table 3, only the main effect of IG vs. CG was significant with a strong effect size, while the main effects for retests and interaction were not significant (see Figure 5; for descriptives, see Table 2; for differences between IG and CG, see Table 4). The CG already had a substantially higher medical adherence than the IG at T0. Furthermore, the CG is classified as adherent for T1–T3, while the mean adherence scores of the IG remains within the non-adherent range for T0–T3 [33].

Hypothesis 5.

Extraversion correlates positively with the benefits of peer support IMS intervention as measured by the frequency of IMS usage, quality of life, and HbA1c levels.

For this hypothesis, only the IG was relevant. Unfortunately, only 14 out of a group of 41 participants filled in the personality questionnaire at very end of the study. Given the overall low final sample size for this hypothesis, we decided not to conduct further analyses.

In addition to the main hypotheses, we also analyzed the associated relevant parameters (see Section 2.2 and Table S10). The Mixed Analysis of Variance showed significant results concerning interaction DDS emotional burden (p < 0.05), DDS physician-related distress (p < 0.05), and time. Furthermore, DDS emotional burden (p < 0.05), regimen-related distress (p < 0.001), and diabetes distress (p < 0.01) were significant pre-post test within the IG and CG. Besides, DKT diabetes knowledge was significant pre-post test within the IG and CG (p < 0.001). We observed no differences concerning biochemical parameters. Regarding anthropometric and body composition data, significant pre-post test within the IG and CG were calculated for BMI (p < 0.01), waist circumference (p < 0.001), fat mass (p < 0.05), skeletal muscle mass (p < 0.05), and visceral fat (p < 0.001) (see Table S10).

4. Discussion

In this RCT, we have implemented a peer support IMS intervention aiming to improve DSM in people with T2DM. Unfortunately, and despite holistic efforts, it was not possible to reach the targeted sample size. This might be due to the fact that this RCT focused on DSMES and lifestyle changes. The latter can be difficult for patients although they are aware of the importance of behavior changes for therapy success [42,43]. The majority of people continue showing unfavorable lifestyle habits even after being diagnosed with T2DM [42]. Even if they are willing to engage in healthy lifestyle habits, committing to behavioral changes seems to be more challenging if they are required implement more than one habit at a time, as this can be perceived as overwhelming and could be one reason our RCT intervention design might have been less attractive for people with T2DM. Furthermore, the recruitment phase of the study started within the COVID-19 pandemic where people were likely to show more resentments towards clinical studies and face-to-face meetings in public places and health institutions. On the other side, COVID-19 restrictions led to an overwhelming number of social activities being held online, which might have had a saturating effect on people’s willingness to participate in yet another online activity. In support of this explanation several participants of the IG stated their reluctance to practice online communication in general and in particular via their mobile phone [44].

Our main hypothesis was that the peer support IMS intervention would reduce the HbA1c in the IG. However, no significant changes were observed during the duration of the intervention between the IG and CG or pre-post within both groups. This is in contrast to recent reviews and meta-analyses which describe an HbA1c reduction for digital interventions such as mobile health, social media, or web-based care compared to usual care [45,46]. According to this research, digital health-led DSMES can lead to an HbA1c reduction in patients with T2DM of −0.48% for six months after baseline measurement and −0.457% for 12 months after baseline [45].

Furthermore, this result is also opposite to the scientific consensus regarding the effect of DSMES, which is interlinked with an HbA1c reduction [8,10,11,12]. One reason for this discrepancy might be that the intensity and frequency of IMS interaction was partly subject to strong fluctuations based on the moderators’ resources—mainly time, group dynamics and group interests. However, recent reviews and meta-analyses show that the intensity of support impacts HbA1c levels [47,48]. Besides, Kerr et al. [47] highlight that a substantial reduction of HbA1c with digital health intervention seems to be challenging because only half of the included RCTs with digital health interventions compared to standard care achieved a significant reduction in HbA1c. The potential for impacting HbA1c values is proportional to the initial levels of HbA1c. Specifically, elevated HbA1c levels are associated with a greater likelihood of measurable reduction [48]. He et al. [46] summarized that patients with baseline HbA1c ≥ 9% benefited more than those with a lower HbA1c. Therefore, further substantial reductions were more difficult to achieve for the majority of DiabPeerS participants, who were significantly below an HbA1c of 9%.

Concerning hypotheses two to four, we found inconsistent results:

Participants of the IG and CG had slightly lower average SDSCA-G scores for the subscore diet (except CG T3), blood sugar test, foot care, and higher scores for the subscore exercise (see Table 2) when compared to a similar cohort regarding age and HbA1c levels but without further comorbidities of a German RCT, which aimed to analyse the effectiveness of an IT-supported care management (diet_mean: 4.34 ± 1.67; exercise_mean: 3.07 ± 1.85; blood sugar test_mean: 5.47 ± 2.23; foot care_mean: 2.77 ± 2.22). Especially the lower subscores for blood sugar test and foot care as well as the higher subscore exercise are probably based on the comorbidities or the insulin dependency of one third of all participants in the German RCT [49]. Disease severity influences DSM significantly [50] besides motivation to DSM, favorable attitude to DSM, disease knowledge, medication and behaviors associated with DSM, competences, self-efficacy/perceived behavioral control, and social support [51]. Similar to the DiabPeerS study, the German RCT did not find significant effects of the IT-supported care management on DSM [49]. Another study, the Taiwan RCT TMU-LOVE implemented an educational program for patients with T2DM via a social media platform during the COVID-19 pandemic. After three months of intervention self-care activities measured via SDSCA (overall score) increased in the IG. The authors mention that improvements in self-care activities were more difficult to achieve than changes regarding diabetes knowledge [52]. An Iranian quasi-experimental with a face-to-face peer-to-peer training for 8 weeks led to an increase in all subscores of the SDSCA [53]. Within the RCT, Leong et al. [52] conducted also patient interviews. The analyses showed that highly individualized approaches are needed to help patients with T2DM to improve their self-care behaviors [52]. This conclusion is supported by numerous studies with individualized programs for patients with T2DM [54,55,56]. It can be assumed that the needed individual 1:1 support was not available to a sufficient extent during the DiabPeerS study.

Regarding the SF-12, our study participants had a lower physical summary scale and a higher psychological summary scale than the comparative values for patients with T2DM (KSK_mean = 45.85 ± 10.41; PSK_mean = 46.56 ± 11.46) [31]. In our study, both groups experienced a significant increase in quality of life (KSK and PSK). Quality of life in diabetic population groups is influenced by a variety of factors including physical activity, blood glucose testing, comorbidities, duration of the disease, diet quality, depression, or treatment regime as recent reviews and meta-analyses show [57,58]. Furthermore, the amount and quality of social support are essential for the quality of life of diabetic people [59]. As the peer support IMS intervention was designed to increase social support and DSMES within the IG participants, the increase of KSK and PSK might be partly related to the intervention. Regarding the CG, we assume that the regular measurement appointments and dietetic consultation as well as the exchange between study participants could have led to an increased quality of life. Furthermore, COVID-19 restrictions were alleviated throughout the course of the participation. This situation could also be part of the general observation regarding quality of life in our study.

Concerning medication adherence (A-14), the IG as well as the CG were classified as non-adherent at T0. This result is contrary to other studies where good medication adherence was observed in varying degrees from 33.5% in a cross-sectional study in Ghana [60], 48.6% in a cross-sectional retrospective study in Saudi Arabia [61], 55.2% in a cross-sectional survey in Malaysia [62], and 97.8% in a pilot study in Bulgaria [63]. Medication adherence is influenced by a variety of factors such as good diabetes knowledge [64], years lived with the disease [63], diabetes distress [60], or ethnicity [62]. Although the medication adherence of the IG increased, these participants were still classified as non-adherent. Nevertheless, this result is in line with findings of recent reviews and meta-analyses [65,66,67,68,69] which describe positive effects of interventions via mobile phones on medication adherence.

Besides our main hypotheses, we collected and analyzed several diabetes-relevant parameters. While we could not calculate significant results for the biochemical parameters, we found inconsistent results for the psychosocial parameters. Furthermore, we calculated significant improvements for all anthropometric/body composition data pre-post within both the IG and the CG, although we had expected these positive results only for the IG. This might be related to the fact that all participants had the possibility to discuss their anthropometric/body composition data with a dietitian directly after their measurements. This assumption is supported by recent reviews and meta-analyses which point out that dietetic counseling positively impacts anthropometric and body composition of patients with pre-diabetes or diabetes. These studies also highlight that dietetic counseling improves HbA1c and might have a positive influence on other biochemical parameters like cholesterol [70,71]. However, these parameters seem to be unaffected by the dietetic counseling in our study or by the peer support intervention. For the dietetic counseling the missing effects can be explained by the thematic restriction on the results of the BIA measurements. Regardless of this, the anthropometric/body composition results are surprising, as no positive changes in diet or physical activity were observed via the SDSCA-G or with the FFQ. Furthermore, the DKT results indicated a surprising decrease in diabetes knowledge in both IG and CG.

In addition, diabetes-related distress seemed to be reduced in both groups. Stress is a factor known to influence body weight and body composition—especially in people with T2DM [72,73,74,75]. This is because of the diabetes-related distress that patients with T2DM face, e.g., the multitude of things they have to consider on a daily basis, which can lead to anxiety or depression [76,77]. Stress can negatively affect a variety of psychological processes and it can lead to changes in the human body’s homeostastis [78]. Highly relevant for T2DM is that high level of stress can lead to increased cortisol and glucose levels. In addition, stress can decrease insulin release or influence sensitivity and resistance of insulin. On behavioral level, it can affect adherence to lifestyle changes negatively such as exercise or recommended dietary behavior. Generally, high cortisol levels strongly correlate with increasing body fat—especially visceral body fat—and obesity [77]. Therefore, the reduction of diabetes-related distress in the DiabPeerS study could be the cause for the improvements regarding anthropometric/body composition data.

The main limitation of this RCT is the insufficient sample size. Despite extensive promotion of the study on a variety of channels extending beyond the borders of Lower Austria, the requisite number of participants was not attained. The restrictive inclusion criteria undoubtedly resulted in a considerable number of interested individuals being excluded from participating in the study. Concurrently, it is imperative for the design of an RCT to formulate inclusion criteria that facilitate the measurement of an effect use of specific parameters.

It is evident that enhancing collaboration with general practitioners and providing educational resources on studies and interventions would prove to be a highly effective strategy for ensuring successful recruitment in forthcoming projects. On the other hand, the burden on medical staff in the public healthcare system is the most important factor that adversely affects this scenario. It is likely that multi-site recruitment, higher financial compensation or the use of a research consultancy would also lead to higher numbers of study participants. Furthermore, targeted recruitment of groups particularly affected by T2DM, such as people with a migrant background, and designing the intervention specifically for them could both increase the number of study participants and enhance the applicability and robustness of the conclusions. Additionally, co-design approaches with the target group could help make the intervention even more appropriate and ensure that more people feel addressed.

Therefore, the hypotheses can be neither confirmed nor rejected correctly, and all results must be discussed with caution. Besides, and in line with other behavioral interventions, participant recruitment includes potential selection bias, e.g., generally more motivated people want to participate.

With regards to the results of our study, it would also be beneficial to consider the extent to which limitations were imposed by the intervention itself. Qualitative research approaches were applied in order to investigate the perception of the intervention by participants and moderators [44]. For instance, some participants in the IG appeared reluctant to communicate online and participate in the group interaction due to the anonymity of the group and the unfamiliar nature of the communication. This, in turn, has the potential to reduce the effectiveness of peer support. It is therefore recommended that participants in DSMES online interventions be better prepared for the intervention and the type of communication, and that joint face-to-face familiarization be facilitated so that greater commitment can be achieved.

5. Conclusions

The present study describes the results of the DiabPeerS study which aimed to improve HbA1c and diabetes-specific parameters via a peer support IMS approach addressing DSMES. New technologies like IMS can be a useful tool for DSMES, but interested participants have to be prepared for this type of intervention and a combination of face-to-face settings and online structures might help participants to overcome online communication barriers. According to our observations, it seems that people with type 2 diabetes are interested in connecting with others facing similar challenges and sharing insights on managing the condition.

Clear recommendations for future implementation were derived and published from the qualitative data analysis regarding the limitations and weaknesses that emerged during the implementation of the intervention [44]. Central results are the identification of different types of users of IMS technology and online communication in general among the participants, some of whom were reluctant to participate in online communication and many of whom had inhibitions about online communication. The desire for supplementary face-to-face meetings was expressed by many and is also recommended for future interventions of this kind to counteract group anonymity and declining participation and to increase engagement. Furthermore, support for the moderators by health professionals should remain part of the intervention and at the same time attempts should be made to keep the moderators’ workload low and to promote spontaneous action leading to greater participant engagement.

As all project materials are open access published for future interventions, an accompanying program guide was developed, which contains all of these recommendations (https://phaidra.fhstp.ac.at/o:5700 (accessed on 14 March 2025)).

In the future, it may be beneficial to explore ways to facilitate networking and peer support and identify most influencing factors, particularly online, to support the management of the disease. We hope that studies such as the DiabPeerS study provide valuable insights to inform the design of suitable interventions for the target group.