Mechanisms Activated in the Interaction between Peer Supporters and Peers: How do the Peer Supporters Perceive and Perform Their Role in an Intervention Targeted Socially Vulnerable People with Type 2 Diabetes: A Realist-Informed Evaluation

,


Introduction
People with a chronic disease such as type 2 diabetes (T2D) beneft from peer support as it enables them to share knowledge, experience, and emotions with like-minded people [1]. "Peer support" is defned as "support from a person who has experiential knowledge of a specifc behaviour or stressor and similar sociodemographic characteristics as the target population" [2]. Tus, a peer supporter ofers support on a level that a healthcare professional cannot give. Tis makes peer support a popular approach, complementing and enhancing the established healthcare system to support people with T2D in managing their disease [3,4]. Studies have shown improvements in health outcomes from glycaemic control and blood pressure regulation [3,5] to mental health challenges such as depression and anxiety [3,6]. Lastly, peer support has been highlighted for its potentially benefcial efects on socially vulnerable people and their management of T2D [7][8][9]. Socially vulnerable people are characterised by low socioeconomic status, e.g., a low income, education, unemployment, and poor social network [7,[9][10][11]. Moreover, this group experiences more complications related to T2D, receives less care, and faces multiple barriers to accessing health care services [8,12,13].
However, how and why peer support programmes work, especially for socially vulnerable groups, remain unclear, and deeper insights into the underlying mechanisms generating outcomes are needed [14,15]. As programmes are implemented in complex settings with multiple layers of contextual factors that either facilitate or inhibit how the mechanisms operate, the outcomes often have a mixed pattern [16]. Terefore, it is important not only to rely on a single outcome measure, such as intermediate outcomes, which often are used to measure health-behavioural improvements of peer support programmes but also to measure immediate outcomes (changes in awareness, understanding, and skills), which usually come before behavioural outcomes [17]. Otherwise, valuable outcome patterns may be left unnoticed. Tis may be particularly important when investigating the efect of peer support programmes among vulnerable groups where healthbehavioural improvements may not be achievable [14]. Drawing on the methodological guidance for evaluating complex interventions [18] and the principle of realist evaluation (RE), it is possible to gain a deeper understanding of the underlying mechanisms that make peer support programmes work or fail in diferent contexts, and how they generate immediate and intermediate outcomes among peers in peer support programmes. Tis study presents the fndings of a realist-informed evaluation of the Danish peer support intervention "Together on Diabetes" (ToD). Te study is based on a previous study that found that the ToD intervention only improved diabetes selfmanagement(DSM) and use of health care services (the intended intermediate outcomes) if individual contextual factors among peers, such as their occupation and fnancial situation, health condition, energy, and other life events, facilitated their engagement in the intervention [14].
However, the study also found that regardless of context, all peers increased their self-care awareness (immediate outcome). Tus, our interest in this study was to explore the mechanisms generating the immediate outcome of increased self-care awareness among peers and not only the intermediate outcomes of health behaviours. We investigated how the peer supporters, as the key provider of the intervention, perceived and performed in the intervention (the resource) and how this impacted the mechanisms (reasonings) activated in the interaction between peer and peer supporters.

Study Aim.
Tis study aimed to investigate how nonprofessional volunteers perceived and performed as peer supporters in the Danish "Together on Diabetes" peer support intervention, and how their performances activated mechanisms that generated immediate outcomes of self-care awareness among socially vulnerable peers with type 2 diabetes.

Te "Together on Diabetes" Intervention.
Te empirical setting for this study was the Danish peer support intervention "Together on Diabetes" developed and implemented in 2017 by the Municipality of Copenhagen's Centre for Diabetes and the Danish Diabetes Association and evaluated by the University of Copenhagen. Te ToD intervention aimed to improve DSM and increase the use of health care services among socially vulnerable people with T2D (peers) through peer support meetings with nonprofessional volunteers with T2D (peer supporters). Te intervention is part of a larger evaluation study of three diabetes interventions developed within the Cities Changing Diabetes Copenhagen partnership programme [19].
Te intervention contained fve components: (1) recruitment (peers and peer supporters), (2) a two-day mandatory training course for peer supporters, (3) matchmaking peers and peer supporters, (4) six months of individual biweekly peer-to-peer meetings between peers and peer supporters, and (5) bimonthly supervision, and network meetings for peer supporters. Te meetings were intended to give the peers social and emotional support, assistance in daily diabetes management, and support in accessing health care services. Te training course aimed to introduce and equip the peer supporters to the intervention. Likewise, the supervision and network meetings aimed to assist them with professional support and the opportunity to share and discuss experienced dilemmas and challenges. Te matchmaking was carried out by the project manager who matched them based on initial conversations with each of them and on their responses in a matching survey (attached as supplementary material). If both peers and peer supporters agreed to the suggested match, the project manager facilitated a match meeting where they could introduce themselves and discuss expectations and wishes for the peerto-peer meetings. Te inclusion criteria for peer supporters were having well-regulated glycaemic control, basic knowledge of T2D and the Danish healthcare system, and an interest in supporting a socially vulnerable person with T2D.
Inclusion criteria for peers were poorly regulated glycaemic control, multimorbidity, no employment, low/no education, living alone with no/minimal social network and difculties assessing the healthcare system [14].

Research Setting: Peer Support in a Healthcare System
Context. Te Danish healthcare system is universal and built on the principle of free and equal access for all citizens, as general taxes fnanced health care services [20]. Te healthcare system includes three political and administrative levels: the state, regions, and municipalities. Te municipalities are in charge of diabetes rehabilitation outside hospitals, disease prevention, and health promotion [21].

Teoretical Framework.
Following the RE principle [22] and the Intervention-Context-Actor-Mechanism-Outcome (ICAMO) confguration by Mukumbang et al. [23], we developed an empirically tested initial programme theory to uncover the causal relationship between the contexts in which the ToD intervention was implemented, the mechanisms activated within the peers and peer supporters, and the immediate and intermediate outcomes. Te ICAMO confguration is an expanded version of the original CMO confguration by Pawson and Tilley [22] as it includes two additional components, "Actor" and "Intervention." According to Mukumbang et al., it is important to have an explicit focus on the actors involved and how they engage in the intervention, as "programmes can only work when the relevant actors adopt either all or parts of the intervention" [23]. Te initial programme theory was based on an overall programme theory for the ToD intervention [14] and fndings from our previous study described previously [14]. A graphic illustration of our initial programme theory is shown in Figure 1.
Key terms and frameworks within RE methodology and how they were applied are described in Table 1.

Method
We used a qualitative multimethod case study design [26,27] comprising in-depth interviews with peers, peer supporters, and the project team; participant observations of training courses, supervisions, and network meetings; and project documents (match survey for peers and peer supporters, and logbook for peer supporters). In total, 12 pairs of peers and peer supporters completed the ToD intervention during the study period (February 2018 to July 2019). Ten peer supporters gave consent to participate in the evaluation whereas one of them had completed peer relationships with three peers, representing three cases. Another case consisted of two peer supporters as one of them had to withdraw from the intervention due to a poor health condition. Overall, it gives a total of 11 cases.
Te peers and peer supporters were recruited to the ToD intervention using diferent strategies. Te project manager raised awareness of the intervention through fyers, information meetings, newsletters, and cofee meetings with relevant stakeholders, such as health professionals and civil society organisations who work with socially vulnerable groups with T2D. Based on the latter contacts, snowball sampling was applied, relying on word of mouth among potential participants.
Peer supporters were recruited from the Centre for Diabetes, who had participated in rehabilitation services (N � 5), members from the Danish Diabetes Association (N � 2), the general practitioner (N � 1), or via the ToD interventions web page (N � 2).
Peers were mainly recruited through the Centre for Diabetes (N � 6). Tey were recruited by health professionals who considered them too vulnerable to participate in the centre's regular diabetes rehabilitation services. Te remaining peers were recruited either through home care (N � 2) or by their general practitioner (N � 1).

Data Collection.
Te study consisted of a triangulation of individual semistructured interviews, participant observations, and project documents, all conducted between February 2018 and April 2020.

Observation of Peer Supporter Training, Supervision, and Network Meetings.
Observations of training courses (n = 3), supervisions (n = 2), and network meetings (n = 2) for peer supporters were conducted with the observer as the participant [28]. An observational guide was used to observe the supervisions with a professional supervisor and the network meetings' plenary discussions on challenges and dilemmas experienced in the peer-to-peer meetings. All observations were recorded in feld notes.

Individual Interviews.
We conducted 27 individual semistructured interviews across the 11 cases. Te informants contained peer supporters (n = 10), peers (n = 9), the project manager, and the diabetes nurse. Each type of informant was interviewed per case to gain various perspectives on mechanisms and contextual factors in the peerto-peer meetings. Te project manager and the diabetes nurse, who visited the peers before and after the intervention, were interviewed three times each.
Te interviews were conducted immediately after the 6month intervention. Te interview guide was informed by the initial programme theory and feld notes from the observations. Te guide was semistructured and focused on the peers' and peer supporters' reasons for and expectations towards participating in the intervention; how the peer supporters perceived and performed their role; how they experienced the training course, network meetings, and supervisions; refections on the peers' improvements; and how contextual factors in the peers' and peer supporters' everyday life afected their engagement. Only the informant and the interviewer were present during the interviews. Interviews with peers were conducted in their own homes. Interviews with the remaining informants took place at the Centre for Diabetes. One peer supporter represents three cases. However, in two of the cases, interviews with peers are missing as they were not interested in participating. Due to restrictions related to the COVID-19 pandemic, seven interviews with peer supporters (n � 3), peers (n � 3), and the project manager (n � 1) were collected by telephone.

Project Documents.
Project documents such as match surveys and logbooks were included in the study to provide information on the peer relationship and whether any contextual factors in the peer and peer supporters' life circumstances infuenced the relationship. Te match survey consisted of 12 questions and aimed to ensure that the best matches were made between the peers and peer supporters. Te surveys gave insights into peers' and peer supporters' sociodemographic characteristics, language skills, residence, motivation, peers' perceived needs, and how they have been recruited. Also, whether they had any requirements for the match (e.g., age, sex, ethnicity, allergies, a nonsmoker, geographic distance, and type of vulnerability). Peer supporters completed the matching survey at the training course, while peers completed the matching survey with assistance from the project manager at an initial meeting. Furthermore, the logbooks contained ten questions and provided insights into peer supporters' refections on the peer-to-peer meetings, including activities, their peerrelationship, improvements among their peers, and whether they needed support to tackle any challenges. Te logbooks were flled out by the peer supporters after each peer-to-peer meeting.

Data Analysis.
Te semistructured interviews were recorded and transcribed verbatim. Field notes of the participant observations were recorded in writing. All transcripts and project documents were managed in NVivo 12 [29]. Systematic text condensation [30] was used to analyse data, consisting of the four following analytical steps: (1) reading transcripts to obtain a general impression of data and identify preliminary themes of relevance to the study aim; (2) identifying and sorting meaning units connected to the preliminary themes; (3) condensation of units and themes; and (4) synthesising data into themes with similar code groups. Informed by the themes and code groups, the initial programme theory was thoroughly discussed by all authors and refned. First author SG conducted the frst two analytical steps. Te other steps were conducted in collaboration with SG, CG, UC, and SV.

Ethical Considerations.
Tis study followed the reporting standards for realist evaluations by RAMESES II (Realist and Meta-narrative Evidence Syntheses: Evolving Standards) [31] and the codes of ethics in the Helsinki II Declaration. Ethical approval was obtained from the Danish Data Protection Agency (Rec. No: 2015-55-0630). All participants received verbal and written information about the study and gave their written consent to participate. Tey were able to withdraw from the study at any time. Confdentiality and anonymity of all participants were maintained throughout the collection of all interviews, analyses, and reporting. Table 2, the ten peer supporters included six males and four females. Tey were middle-aged, and the majority were of Danish origin.   Intervention "A combination of program elements or strategies designed to produce behaviour changes or improve health status among individuals or a group" [17] In this study, the intervention was the biweekly peer-to-peer meetings Context "An irreducible set of factors infuencing when and how an intervention is delivered and how mechanisms are triggered" [22] We investigated how individual contextual factors infuenced peer supporters' perception of and performance in the intervention Individual contextual factors Pawson categorises context into four layers (individual, interpersonal, institutional and infrastructure). Te individual layer includes the actors of the programme's sociodemographic characteristics, capacities, and life circumstances [24] Actors "Individuals, groups, and institutions who play a role in the implementation and outcomes of an intervention" [17] In this study, the actors were peer supporters and peers Mechanism A mechanism is defned as "a combination of resources ofered by the social programme under study and stakeholders" We investigated how peer supporters perceived and performed in the intervention (resource) and how this resulted in changes (reasonings) in the encounter with peers "Reasoning in response" [22]. Tus, resources and reasoning are mutually constitutive of a mechanism, but to help operationalise the diference between a mechanism and a context Dalkin et al., recommend disaggregating them as separate concepts [25] Resources Te resource that an intervention provides Reasoning Te actors' reasoning and response to the resources [25] Outcomes Outcomes of a programme can take many forms, be intended and unintended, as well as they can be multiple and vary across the target group(s) depending on the mechanisms and context [16] Inspired by Mukumbang et al. [23], we used the terms "immediate outcome" and "intermediate outcome" to disaggregate between changes in awareness and behavioural changes among peers Immediate outcomes: Refers to changes in awareness, skills, or understanding. Tese types of changes usually come before behavioural changes [17] Intermediate outcomes: Refers to behavioural changes that follow the immediate outcomes [17] Health & Social Care in the Community groups and/or within health care services. Te nine peers included two females and seven males. Tey were primarily middle-aged men of Danish origin, outside the labour market, with short and intermediate education backgrounds. All peers and peer supporters were diagnosed with T2D and had one or more diabetes complications (cardiovascular diseases, hypertension, neuropathy, and nephropathy). Furthermore, all peers and most of the peer supporters had other diagnoses, such as mental health disorders (depression, stress, and anxiety), arthritis, and KOL. In the following sections, the study fndings are presented. First, the immediate and intermediate outcomes of the intervention are described, as well as how peers' contexts infuenced these. Ten, key contextual factors and mechanisms (resources and reasonings) are presented. Finally, the refned programme theory and examples of the causal relationship between the diferent components are unfolded.

Immediate and Intermediate Outcomes.
In the interviews, peers described how sharing experiences, emotions, and concerns helped them to recognise the importance of taking better care of their T2D, such as taking their medication as prescribed, making healthier food choices, and being more physically active. Te latter is illustrated in the following quote: "I have become more aware that I need to go for walks and exercise a bit." (Peer, Case 3).
Tis increased self-care awareness was identifed among all nine peers independent of their context. However, in more than half of the cases, contextual factors in peers' everyday lives, such as an unstable fnancial situation, a lack of energy, a poor health condition, and/or other harmful events, hindered the increased self-care awareness from leading to behavioural improvements in their DSM. In the following quote, a peer describes how he is aware of his diabetes situation and knows what to eat and how to behave to achieve better treatment goals. However, contextual factors, such as his fnancial situation on cash benefts combined with poor mental health, prevented him from making the behavioural changes required to improve his DSM [14]: Although improved DSM was only achieved by some peers, the empirical fndings showed that many peers achieved other types of intermediate outcomes, such as improved cleaning and tidying up in their homes, better personal hygiene, and getting dressed in the morning. Tus, the outcome patterns in the nine cases show how the intervention initiated processes among peers that could be the frst steps towards making behavioural improvements in their DSM but also has value in itself.

Contextual Factors.
Te empirical fndings confrmed that the contextual factors in our initial programme theory ( Figure 1) afected how nonprofessional volunteers perceived and performed as peer supporters and, thereby, how the mechanisms generating increased self-care awareness were activated in the peer relationships. For example, peers' and peer supporters' sociodemographic characteristics (sex, age, ethnicity, and education level) infuenced the extent to which the peer supporters used themselves in the relationships. Te number of sociodemographic similarities increased the level of personal engagement. Furthermore, peers' health conditions (multiple diagnoses and diabetes-related complications) and life circumstances (e.g., occupation and fnancial situation, level of energy, and other life events) were found to either facilitate or inhibit the meeting frequency and the activities. In addition to the initial programme theory, peer supporters' lived experience (personal experience living with T2D and working experience) was found to be a contextual factor, as they used their competencies, knowledge, and skills in the encounter with peers. Finally, the health of peer supporters was also discovered to be a contextual factor. In a few cases, sharing the same diagnosis became a barrier as some peer supporters had other diagnoses and/or diabetes-related complications. Consequently, this caused interruptions in the peer relationships as the peer supporters had to cancel some of the biweekly meetings due to poor physical and/or mental health conditions. In one case, peer supporters' poor physical health conditions hindered them from participating in the intervention.

Use of Lived Experience.
Peer supporters used their lived experiences, particularly their personal experiences with T2D, to provide social and emotional support to their peers. However, data showed a pattern in their personal engagement depending on whether they had previous work-related experience as a support person or not. Peer supporters who did not have any previous experiences were more personally engaged. Tey identifed themselves as a befriender with the primary aim to assist their peer with social and emotional support. In contrast, those with previous experiences identifed themselves as a diabetes support person in accordance with the intervention. According to this group, their role was beyond providing social and emotional support to assist their peer with their daily diabetes management and access to health care services. Furthermore, from their experiences as a support person, they knew that the relationship was temporary. Terefore, they tried to keep a professional distance to their peer: "I draw on my experiences from work (...), and you make closer contact to some of them than others, but you do always have to keep that professional distance." (Peer supporter, case 4, 6 and 11).
However, even the most experienced peer supporters found it difcult to distance themselves completely. Contextual factors, such as peers' sociodemographic characteristics, such as sex, age, ethnicity, and educational level, infuenced their personal engagement. Tis was especially shown in the cases represented by the same peer supporter. Even though the peer supporter had working experience as a support person, she performed her role diferently depending on how much she could mirror herself with her peers. In the case where she mirrored herself the most (e.g., same educational level, sex, and ethnicity), she was noticeably more personally engaged: "I have gotten more of myself into play (...) We have shared knowledge on a completely diferent level than I have done with the other two. In other words, the equality between us (that equality around the fact) that one person knows something, and the other person knows something else" (Peer supporter, Case 4, 6 and 11).

Adapting the Role.
Most peer supporters who identifed themselves as a diabetes support person experienced a discrepancy between their expectations of the role presented at the training course and what the role entailed. In almost half of the cases, contextual factors in peers' life circumstances, such as unstable fnancial or occupational situations combined with poor health conditions, hindered them from interacting as intended, thereby achieving the intended outcomes. Consequently, this created frustration among many peer supporters as they felt they did not fulfl their role as expected: "I don't think I've moved him that much (...) Well, I expected something else. (...) I didn't succeed in changing his diet or anything like that" (Peer supporter, Case 5).
By sharing their frustration at supervisions and network meetings, they understood that this was a common challenge and realised that they still had an important role to play. Tus, they adapted their performance, as well as their expectations of outcomes to their peers' capabilities, resources, and perceived needs by taking their peers' circumstances and health condition into account. For example, in the following quote, a peer supporter described how she changed her role perception from being a diabetes support person to a befriender.
"I was about to say stop because I didn't feel he needed me. I then readjusted mentally; he doesn't need help with his diabetes. He needs help in his loneliness." (Peer supporter, Case 5).

Reasonings.
Te empirical fndings showed how the combination of the diferent resources created trust, respect, empathy, care, and honesty in the encounter between peers and peer supporters as well as an understanding of each other's situation.

Experience of Being Equal with a Like-Minded Person
Activated Mutual Trust, Respect, Empathy, Care, and Honesty. Te fact that the peer supporters were unpaid and nonprofessionals who used their personal experience with T2D to provide social and emotional support activated an experience of being equal: "You feel more equal than if you are with professionals. Tere you might be dependent on something from the professionals." (Peer supporter, Case 1).
Furthermore, it created an authentic relationship based on trust, respect, empathy, care, and honesty, making the peers feels safe and engaged in the intervention: Across all cases, these mechanisms permeated the relationships, which enabled them to talk about personal issues Health & Social Care in the Community and challenges they would not share with family members or friends. Moreover, many peers described how their peer relationship difered from consultations with healthcare professionals as they, to a larger extent, felt met with care and empathy, and without any judgement, assessment, or held accountability for anything they said: "After all, we relate to each other as people, and not clients, and that gives a human face, which I like." (Peer, Case 11).
Te ability of the peer supporters to adapt their role and expectation of the programme to ft the complex lives of the peers as described under resources had critical importance for the activation of mechanisms. Te adaptation created trust, care, and empathy in the sense that the peers felt heard, understood, recognised, and not judged. Tis transformation of the peer supporters' role was conducted without the programme losing meaning to the peer supporters. Finally, the experience of being equal created mutual respect. Tey showed each other respect by attending the biweekly meetings, complying with their agreements, being honest, and did not waste each other's time: "We had that mutual respect, right? (...) At frst, he thought I was getting paid for it (being a peer support, red.) When he found out I didn't get paid for visiting him, he realised that we shouldn't waste each other's time." (Peer supporter 2, Case 7).

Sharing the Same Diagnosis Evoked an Understanding of Each Other's Situation.
Sharing the same diagnosis facilitated the relationship as they mirrored themselves in each other because they had gone through similar life-changing events. Tis created an understanding of each other's life situations, to which many peers were a relief. Tey did not need to explain themselves nor talk about the illness to make their peer supporter understand their situation: "It means so much that the person who steps in has an insight into your faws. So, it's such a huge relief that you don't have to talk about it (T2D, ed.) at all, but just act in the areas where it's necessary, right?" (Peer, Case 1).
Likewise, many peer supporters addressed how sharing the same diagnoses, including what it entails of behavioural changes (e.g., taking diabetes medication daily), created a mutual understanding of everyday life with T2D: "It came as a shock to him that he suddenly had to take medicine. It did that for me too. And then we talked a little about it (...) It meant that we also here were in the same situation." (Peer supporter 2, Case 7).
As illustrated in the refned programme theory, contextual factors infuenced how peer supporters identifed themselves and how they adapted their performance in the intervention. However, although they identifed themselves and performed diferently, the fact that the peer supporters were nonprofessionals, unpaid, and shared the same diagnoses created an experience of "being equal," which activated mechanisms, such as trust, respect, empathy, care, and honesty between peers and peer supporters. Tis is also shown in the ICAMO confguration as follows (Figure 3), which also demonstrates the infuence of peer supporters' work-related experiences and the number of sociodemographic similarities between the peer and the peer supporter.
Te ICAMO matrix in Table 3 shows how each component is linked.

Main Findings.
In this multimethod case study, we explored how being with a like-minded person with T2D increased self-care awareness (immediate outcomes) and subsequently, depending on the context, generated behavioural changes (intermediate outcomes) among socially vulnerable people with T2D. Furthermore, we explored how the peer supporters drew upon their lived experience (personally and work-related) to identify themselves in the role and set up personal boundaries for their engagement as well as to provide social and emotional support. Similarly, the fndings revealed how peer supporters were able to adapt their performance as well as their expectations of outcomes by taking into account their peers' circumstances and health conditions. Four individual contextual factors were found to impact the peer supporter's perception of and performance in the intervention: peer supporters' lived experience, peers' life circumstances, and peers' and peer supporters' sociodemographic characteristics and health conditions. By focusing on the peer supporter, as the key provider of the intervention, we achieved a deeper understanding of how and why the ToD intervention activated mechanisms, such as trust, respect, empathy, care, and honesty, which then led to increased self-care awareness and among some cases improved DSM and use of health care services. Tus, when evaluating complex health interventions, we emphasise the importance of measuring both immediate and intermediate outcomes [32]. Tis is because behavioural changes are usually preceded by changes in awareness, skills, or understanding. Te growing body of literature on peer support tends to focus on measuring the intervention efect (efect evaluation) [6] or fdelity issues in process evaluations [33]. Terefore, we contribute with novel knowledge to this feld as behavioural changes may not be achievable, especially not for socially vulnerable groups [9]. However, peer support interventions targeting socially vulnerable people do have promising potential. We believe that the analysis of mechanisms is at a level of abstraction that goes beyond the ToD intervention, as some study fndings have been reported elsewhere. For example, the experience of being equal is reported in other 8 Health & Social Care in the Community studies on peer support for people with T2D [34,35] and on peer support within mental health services [15,36]. However, comparing our fndings to the existing literature on T2D peer support often is difcult as many peer support interventions include other approaches (e.g., group-or telephone-based) targeted at other groups. Furthermore, they are carried out by other stakeholders, such as churches and community organisations [37,38] Figure 2: Refned programme theory for the ToD intervention (using the ICAMO confguration by Mukumbang et al. [23]).
Because peer supporters are non-professional, unpaid and share the same diagnosis (T2D), trust, empathy, respect, care and honesty (reasonings) were established between peer and peer supporter (actor), resulting in an increased self-care awareness among peers (immediate outcome) and, depending on peers' life circumstances (context), behavioural improvements in diabetes self-management and use of healthcare services (intermediate outcome) Prior work-related experience ICAMO 1B: No prior work-related experience Peer supporters (actor) with work-related experience as a support person (context) identify themselves as a diabetes support person, with the primary aim to assist peer in improving their diabetes selfmanagement and help them navigate within the healthcare system (resource), which made them keep a professional distance to peer in the bi-weekly peer support meetings (intervention) Peer supporters (actor) with no prior workrelated experience as a support person (context) identify themselves as a visiting friend, with the primary aim to provide social and emotional support (resource), which made them engage personally in the biweekly peer support meetings (intervention) If peer and peer supporter (actors) share similar sociodemographic characteristics (sex, age, ethnicity, and education level) (context), then the peer supporter is more likely to become personally engaged (resource) in the biweekly peer support meetings (intervention) If the peer and peer supporter (actors) do not share similar sociodemographic characteristics (sex, age, ethnicity, and level of education) (context), then peer supporter is more likely to maintain a professional distance from the peer (resource) in the bi-weekly peer support meetings (intervention)

Strengths and Limitations of the Study.
Tis study has several strengths. First, including socially vulnerable people as the target population is a strength. Tis group is often hard to reach and engage and is thus understudied [39].
Using the ICAMO confguration [23], we discovered that peer supporters identifed and engaged themselves diferently in diferent contexts. Tese fndings would not be possible to discover by only evaluating changes in outcomes. Moreover, the use of case studies, recommended when analysing complex interventions [31,40], allowed us to test the initial programme theory and verify whether propositions could be reproduced in diferent cases. Also, the triangulation of methods and data sources enhancing the credibility of the fndings gave us a broader and more nuanced understanding. Lastly, our relatively large sample size of eleven cases with four diferent informants (n = 27) ensured sufcient information power [41] to conduct the study. Our study does also have some limitations. Due to restrictions related to the COVID-19 pandemic, seven interviews (three with peers) were performed via telephone, limiting the interviewer's access to verbal nuances and nonverbal communication. Nevertheless, comparing these with the other interviews, we do not fnd that this had consequences for our analysis. Also, our analysis focused on individual contextual factors. Another context focus (e.g., the institutional level) [42] could have provided us with other analytical fndings.

Implications for Practice and Future Research.
Our study helps to fll the gap in the research literature on mechanisms underpinning peer support targeting vulnerable groups with T2D [15]. Te fndings showed how most of the mechanisms were reciprocally activated in the encounter between peers and peer supporters. Tus, we fnd it relevant that future research focuses on investigating the mechanisms activated within all key actors and not only the target group. Moreover, instead of having a traditional focus on behavioural improvements, we also recommend measuring changes in awareness, understanding, and skills as such changes may be left unnoticed. Likewise, we suggest policymakers be aware that behavioural improvements may not be achievable for all target populations, especially socially vulnerable groups when designing interventions. Finally, we recommend peer support as a supplement to the established healthcare system to provide social and emotional support to socially vulnerable people with T2D.

Conclusion
In this study, we explored the underlying mechanisms activated between peers and peer supporters in a Danish peer support intervention targeting socially vulnerable people with T2D. By focusing on how peer supporters, as the intervention's key providers identifed themselves with the role and performed in the intervention, we gained a better understanding of why and how the intervention activated mechanisms (e.g., trust, respect, care, honesty, and empathy) between peers and peer supporters. Tree contextual factors infuenced how the peer supporters perceived and performed their role, and thereby, how the mechanisms were at stake. Using principles from realist evaluation and by disaggregating mechanisms into "resources" and "reasonings," our study provides insight into which peer support programmes work, for whom and under what circumstances. Furthermore, the study contributes with novel, in-depth fndings about how to reach socially vulnerable groups in complex health interventions; groups that healthcare systems, even in a universal welfare system as the Danish, cannot reach.

Data Availability
Te data and transcripts used during the study are available from the corresponding author upon reasonable request.

Additional Points
What is known about this topic and what this paper adds? What is known? (1) Peer support targeting people with chronic diseases, such as type 2 diabetes, has shown positive efects on health.
(2) Peer support is increasingly implemented to supplement the established healthcare system to support people with type 2 diabetes in managing their disease. (3) Limited research has investigated the underlying mechanisms causing a positive efect. What the paper adds.
(1) Te usefulness of a realist-informed evaluation to explore a peer support programme for socially vulnerable people with type 2 diabetes. (2) Insights into key mechanisms and contextual factors enlightening how and why the programme has potential. (3) Insights on the importance of measuring both immediate and intermediate outcomes in evaluations, as intermediate outcomes, such as healthbehavioural improvements may not be achievable for socially vulnerable groups.

Ethical Approval
Tis study followed the codes of ethics of the Declaration of Helsinki and was approved by the Danish Data Protection Agency. Rec. No: 2015-55-0630.

Consent
Te participants received verbal and written information about the study and gave written consent to participate. Tey were informed that they could withdraw from the study and guaranteed anonymity.
Health & Social Care in the Community