Article Text
PDF
Abstract
Background The value of a participatory approach to the generation of evidence for health and social services from a moral, methodological and policy level continues to gain recognition globally. Trust is a crucial mechanism in the participatory health research (PHR) process and is strongly influenced by context. However, gaps remain in conceptualising and operationalising trust over time in PHR partnerships.
Objective This case study seeks to address these gaps by exploring the evolution of trust multidimensionally across two timepoints.
Setting and participants Participants in a PHR project called the Public and Patient Involvement (PPI) Ignite Network in Ireland (n=57 (T1); n=56 (T2)) were invited to complete a network survey at two timepoints. The PPI Ignite Network had local and national partners.
Network measures Several core social network measures were calculated at both timepoints to characterise the differences between trust dimensions and between local and national partners.
Results Subtle changes were observed across most network measures over the two timepoints. While there was a slight decrease in the number of connections for each trust dimension throughout the PPI Ignite Network, connections that were consistently nominated in both timepoints increased slightly. Some trust dimensions, such as vulnerability and integrity, were more similar, while others, like integrity and shared values, visions and goals, differed greatly, where national partners consistently received more incoming connections compared with local partners.
Conclusion These findings (1) provide empirical support for using social network analysis to operationalise trust comprehensively and multidimensionally over time in a participatory partnership, (2) offer nuanced insights into the trust development process within the PPI Ignite Network and (3) enhance our understanding of trust in the community-based participatory research model.
- Community-Based Participatory Research
- Health
- SOCIAL MEDICINE
Data availability statement
The data that support the findings of this study are available from the corresponding author, JS, upon reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
This study provides empirical support for using tools and techniques from network science to clarify important conceptual and operational complexities of trust in participatory health research partnerships across two timepoints. In doing so, we help address critical ambiguities that hinder the application and evaluation of participatory health research in health promotion.
Our approach to measuring trust in participatory partnerships embraces its multidimensional nature, allowing us to see how trust unfolds, across all its dimensions, over two timepoints.
By exploring trust in this way, we embraced the partnership environment, which plays an important role in trust and partnership synergy and sustainability.
This case study used a small network with two timepoints over a year. Considering trust takes time to develop, it is possible that surveying trust at only two timepoints over a year is restrictive.
As trust is inherently contextual, its evolution will likely vary depending on the partnership of interest.
Background
The value of a participatory approach to the generation of evidence for health and social services from a moral, methodological and policy perspective continues to develop on a global scale.1–3 Participatory health research (PHR) can be defined as “systematic inquiry, with the collaboration of those affected by the issue being studied, for the purposes of education and taking action or effecting change”4 (pg. 43). In PHR, ‘those affected’ is intentionally broad encompassing individuals, community members or groups such as patients, public, health professionals and organisational representatives. These individuals/groups can be both directly or indirectly affected by a health issue.5
With roots grounded in principles of social action, justice and emancipatory philosophy, PHR has the potential to tackle complex health problems and achieve more meaningful and nuanced short- and long-term outcomes.5–7 PHR has been gaining recognition throughout research communities as an approach that serves to bridge the gap between research and practice.5 7 8 Specifically, PHR helps maximise the relevancy of research and usability of its products, while simultaneously building capacity and addressing issues of social justice and self-determination among end-user communities.5 7 The central tenet of PHR is its co-creation process, where those affected by the issue under investigation or who benefit from the knowledge being produced, are key to the knowledge production process, working as equitable partners with academics from idea conceptualisation to dissemination and beyond.5 9
In this article, we discuss PHR as an umbrella term for a variety of approaches (eg, participatory action research,10 participatory rural appraisal10 11 and community-based participatory research (CBPR)12 13). While terminologies may vary by country of origin, discipline and research goals,12 14 they all strive to bridge the gap between knowledge and practice by harnessing inclusivity and recognising the importance of actively and meaningfully engaging those who the research serves to benefit in the research process.5
One widely recognised approach to PHR5 15 is CBPR. A conceptual model for CBPR was developed16 and later adapted,7 providing a concrete framework for understanding how the CBPR process is influenced by contextual and process-related aspects that can affect the ability to achieve both short-term impacts (eg, stronger partnerships) and long-term outcomes (eg, improved health, community transformation and health equity). The intention of the model is to act as a dynamic tool that evolves with research and understanding of CBPR. This includes a deeper understanding of how context, partnership characteristics and processes contribute to research and intervention design, and ultimately lead to intermediate- and long-term outcomes.17 However, challenges in operationalising aspects of the model limit our understanding and evaluation of the PHR process. For instance, Oetzel et al 18 19 noted that additional longitudinal research is required to better understand how CBPR processes lead to outcomes and under what conditions, to further substantiate the mechanisms in the model.18 19
Trust is frequently identified as an important component of the CBPR model, described as “permeating and affecting all interactions and relationships in the partnership and as linking one [domain] to another”20 (pg. 14). Trust has been underscored as a crucial mechanism21 22 essential to the PHR process that can affect the ability to achieve both intermediate impacts and long-term outcomes.23 24 For example, seminal work by Jagosh et al 22 found that the building and maintenance of trust was a key mechanism for supporting partnership synergy, a universal feature of the collaborative process necessary for building and sustaining partnerships. Synergy has been described as “the power to combine the perspectives, resources and skills of a group of people and organizations” (pg. 183) and influences partnership effectiveness.25 However, defining, measuring and operationalising trust in PHR are challenging given the overwhelming variation in how it is defined.26 This reflects sentiments expressed by Misztal et al.27 (pg. 117), underscoring that of Wuthnow et al,28 describing trust as “one of the most complex, multidimensional and misunderstood concepts in the social sciences”27 28 (pg. 117).”
As explicated by Lucero et al.,29 “although numerous CBPR scholars have discussed the importance of trust and offer anecdotal suggestions, very few systematically research it” (pg. 160). Influential work by Lucero et al 24 29 30 has provided important advancements in the study of trust in participatory literature presenting, for the first time to our knowledge, an alternative to the binary view of trust in CBPR (ie, present or absent). As highlighted above, Lucero et al 24 29 30 operationalised trust as a typology of six categories from the lowest type being a trust deficit (suspicion) to the highest called critical reflexive trust (having the ability to discuss and move on after a misstep). However, more work is still needed, especially exploring trust types over time.24 With the recognition that trust is a dynamic, socially embedded process and extends beyond a simplified view as a variable, it requires a methodology that reflects this.21
One approach is to view PHR partnerships as a social network. A social network describes the relationships among people, organisations or other social actors.31 Social network analysis (SNA) is a methodology for describing and measuring contextual and relational dynamics among and between social entities like individuals or organisations.32 Trust is a type of relation that has been commonly explored in the network literature in diverse fields,33–49 such as in health50 and education.51 As mentioned by Zolin and Gibbons,49 “for a researcher, analysis of networks that are directly or indirectly related to trust may yield practical and theoretical insights that are not discoverable through other means …” (pg. 189). This is because, unlike other methods, SNA allows us to understand trust while embracing its social environment.49 This is a key strength of SNA as it extends beyond the behaviour of the individual, embracing the social aspects of behaviour. Using SNA, we can consider the interdependent nature of human data.31 52 Further, as trust is a type of relation, it is inherently embedded in a network of relationships. This creates opportunity to explore a variety of research questions about trust.49 For example, network questions can help us explore how trust is developed over time.49 Indeed, viewing PHR partnerships as a social network, applying SNA tools and techniques to explore trust over time, could help address the challenges that persist in operationalising trust in the CBPR model, and in turn, improve our understanding and evaluation of trust in the PHR process.
Recognising this potential, Gilfoyle et al 53 then proposed a novel and interdisciplinary conceptual triad in their scoping review, with trust in the centre, connecting PHR and SNA, to explore how trust can be conceptualised, operationalised and measured in PHR and social networks literature. Results from this review53 revealed two key findings. First, it found trust to be multidimensional, identifying several key trust dimensions. Second, it underscored a lack of conceptual and operational consistency of trust, particularly in the PHR literature. Gilfoyle et al 54 then empirically tested the merits of exploring trust in a PHR partnership, known as the national Public and Patient Involvement (PPI) Ignite Network (see section ‘Setting’), comprehensively and in a multidimensional way. This revealed important nuances between the different dimensions of trust between partners who had national and local roles, which become diluted when explored in combination.
However, authors54 were limited in that they explored trust cross-sectionally and without attention to specific attributes, such as partnership characteristics (eg, local or national partners). This is problematic because trust develops over time,21 24 30 and networks are dynamic as their membership and social contexts change.55 Further, network attributes, such as the roles of local or national partners, can influence collaborative behaviour,31 also warranting investigation over time. This is especially important in PHR where the importance of trust throughout all phases of the research process is underscored,20 30 especially for ensuring partnerships are effective, equitable and long-term.21 24 56 57 Thus, if some types of partners (eg, those who hold funding (cf. 17)) are developing trust and others (eg, either local partner or national partners) are not, important goals and ultimately outcomes of a partnership may be jeopardised. Addressing these limitations, this case study seeks to extend the findings from Gilfoyle et al 54 to better understand the evolution of trust in the context of a PHR partnership (see setting below). This is done by exploring specific features of a PHR network (the national PPI Ignite Network in Ireland) asking:
Do the trust characteristics of the PPI Ignite Network change from T1 to T2?
Do the dimensions of trust identified differ at the local versus the national level? How did this evolve from T1 to T2?
Methods
Patient and public involvement
This is one substudy that is part of a larger study in which a Research Advisory Group was involved. This group comprises four research partners representing academic, service or community organisations in the PPI Ignite Network (further described in this article). These partners were a subset of individuals interested in this work, who were already working with coauthors JS, AM and MG through a prior grant called PPI Ignite@UL. These partners provided input and approval for the research objectives of this study, ensured all content in the network surveys and interview guide were both accessible to participants and contextually relevant, reviewed and interpreted findings at a high level confirming from their perspective, if they agreed with the findings as a partner in the PPI Ignite Network, acted as a soundboard for brainstorming ways to address any research challenges, provided suggestions/feedback for ensuring dissemination materials and outputs (eg, conference posters and manuscripts) and were being communicated effectively for diverse audiences. One Research Advisory Group member has been further involved in the interpretation of the results as well as reviewing and revising manuscript content and language, and thus, authorship of this article (coauthor MMC). Coauthor MMC was also involved in the dissemination of this work at an international conference (cf. 58).
Setting
In 2017, five universities across Ireland were funded as individual PPI Ignite Teams by the Irish Health Research Board (HRB) and Irish Research Council (IRC) to build capacity for PPI in health research. Building on and consolidating this work, the HRB and IRC then funded the PPI Ignite Network (March 2021–2026), “aim[ing] to provide a shared voice for PPI across Ireland, aiming to change the research culture, and an important contributor to improving health outcomes for the public.59”
The PPI Ignite Network brings together academic, service and community organisations who co-designed the work programme and must collaborate in a synergistic and cohesive manner to plan, implement and evaluate the PPI initiatives set, where trust plays a central role. The PPI Ignite Network’s work focuses on five key areas: (1) building capacity for PPI in community and academic settings, (2) develop accredited education programmes for PPI, (3) enhance university policies and procedures to support PPI, (4) develop quality improvement and impact, and (5) create systems for national co-ordination and functioning (for further information on the PPI Ignite Network, see: https://ppinetwork.ie/about-us/).
The PPI Ignite Network (n=57 at T1 and n=56 at T2 at the time of sampling), a national PHR partnership, provides an ideal setting to better understand how trust evolves in a PHR partnership over time. At the time of sampling, the PPI Ignite Network included 7 universities (called lead sites, including the original five PPI Ignite Teams and two additional institutions), a national office, 10 national-level community partners contributing to national-level governance and activities, and 39 (at T1) and 38 (at T2) local-level partners contributing to governance and activities at one university in the PPI Ignite Network. This administrative structure of the PPI Ignite Network resembles a hub and spokes model. The national office acts as the hub, at the centre of the administrative structure, connecting with national partners and the seven universities, while the universities are further connected to their local partners. Within this structure, resource allocation and decision-making pertaining to goals and objectives are distributed across the network. All partners (ie, national and local) in the PPI Ignite Network interact through multiple avenues, including local partner meetings (ie, site leads and their local partners), PPI Ignite Network-wide meetings (all partners), five work packages each addressing a specific function central to the network’s goals (outlined as key work areas above) (open to all partners) and the National PPI Festival (please see: https://ppinetwork.ie/national-ppi-festival/) (open to all partners and external participants). The PPI Ignite Network functions as a participatory partnership where we explore the dimensions of trust in action for this study.
Data collection
A social network can be defined as the set of connections among people, organisations or other social actors.31 This study invited all 57 individuals at T1 (May 2021) and 56 individuals at T2 (May 2022), in the PPI Ignite Network, to complete the same network survey at these two timepoints. This timeframe was chosen based on discussions with the Research Advisory Group, recognising that the initial stages of partnership development are crucial for the trust development process,60 while also ensuring sufficient time for trust to build.
A network survey is a questionnaire designed to generate names and connections among individuals in a network.31 The network survey in this study was developed based on the dimensions of trust identified by Gilfoyle et al 53 and in collaboration with the Research Advisory Group to ensure its clarity and appropriateness (see online supplemental file 1 for the network survey). The survey was administered electronically via Qualtrics software (version May 2021 to December 2022). Survey questions included seven network questions corresponding to the dimensions of trust identified as important in previous work53 (shown in table 1).
Supplemental material
Conceptual and operational descriptions for each trust dimension (also presented in previous work53 54)
To generate each trust dimension network, all participants were asked to name up to seven organisations when responding to the network survey questions (the same seven organisations for each question), and to consider the individual representing each organisation in their responses. This distinction is critical as this case study is focused on trust within a collaborative partnership, not organisational trust.
The seven dimensions of trust and their corresponding network questions were informed by a scoping review by Gilfoyle et al,53 which comprehensively synthesised how trust had been conceptualised and operationalised in both the PHR and social network literature, and if/where convergence existed. This scoping review53 included a thematic analysis of the extracted literature to better identify the conceptual and operational linkages of trust across and within the PHR and social network literature through their thematic groupings. The relational constructs (ie, subthemes from this review), along with discussions from the Research Advisory Group, informed the dimensions of trust identified. An overview of how the dimensions of trust were conceptualised and operationalised is presented in table 1, informed by53 and also presented in previous work.54 A subsequent study provided empirical support for the scoping review findings.54
Analysis
Individual networks for each dimension of trust were constructed from participant responses to the seven trust statements included in the network survey at both timepoints. Responses were quantified by assigning edge (ie, connection between two individuals) weights from –1 to +1, based on a 5-point scale (strongly disagree to strongly agree), with intervals of 0.5. For example, a response of ‘agree’ was assigned a weight of 0.5, while ‘strongly agree’ received an edge weight of 1. Conversely, responses of ‘neither agree nor disagree’, ‘disagree’ or ‘strongly disagree’ did not result in an edge (ie, connection in the network). This is because, in alignment with the literature on trust, we did not want to infer neutral agreement or disagreement with each statement as an expression of distrust. Distrust is said to differ conceptually from trust61 and more specifically stated by Jones62 “the absence of trust is not to be equated with distrust”.62 Thus, by focusing explicitly on trust connections that were present, we avoid misinterpretation of neutral and negative responses.
Network measures
Several network measures were used to analyse each of the seven trust dimensions at both T1 and T2. Specifically:
In-degree measures the number of incident edges an individual in the PPI Ignite Network has, that is, the number of times a person in the network was nominated by another individual in the network. This metric helped identify trust relationships between individuals (ie, who trusts whom).
Weighted in-degree represents the total strength of agreement for each trust statement (described further in the analysis).
Average in-degree, as reported in this article, reflects the mean number of received nominations across the network, providing insight into the overall level of trust in the network. Additionally, average in-degree allowed us to assess changes in the number of incoming edges received in the network as a whole (ie, agreement or strong agreement for a specific dimension of trust) over time.
Clustering coefficient measures the extent to which individuals cluster together in the PPI Ignite Network, specifically examining the proportion of closed triads (ie, triangles) in the network.63 For example, if there are three individuals in a network, A, B and C and individuals A and B trust each other, and B and C trust each other, then, if this is a closed triad, A and C will also trust each other. The clustering coefficient tells us how frequently this occurs. This identifies how trust is shared within groups throughout the network.
Reciprocity at the network level measures the proportion of reciprocated edges in the network.63 A reciprocated edge occurs when trust is mutual (eg, both individuals agree or strongly agree on the same trust dimension). The concept of reciprocity is often recognised as a critical mechanism of trust.45 56
Freeman centralisation about the in-degree measures the positional importance (centrality) of individuals in the trust dimension network.64 A higher value for one (or a small number of) individual(s) suggests that they hold a position of influence for a specific trust dimension. Conversely, a decentralised indicated that influence is distributed across many individuals.
These measures allowed us to compare the structural properties of trust dimensions.
Additionally. a separate network was constructed to include only the connections present at both timepoints. This provided a basis to compare changes in average in-degree and centralisation about the in-degree for individuals who consistently selected the same people at both timepoints (ie, individuals whose rate of agreement remained stable over time). This enabled us to examine how trust evolved over time for those naming new individuals in T2 (eg, due to staff turnover, new partnerships or interactions driven by work package preference), compared with those who maintained their nominations from T1.
Analysis of trust over two timepoints
To compare trust networks across T1 and T2, we only included responses from individuals who participated in both timepoints. First, we performed a two-sample Kolmogorov-Smirnov (KS) test on the degree distributions. The KS test, a non-parametric test for comparing two probability distributions, calculates the maximal difference between the cumulative forms of the two distributions. This provides a measure of the differences between the distributions, allowing us to assess the extent to which trust dimension networks differ over time. While degree distribution analysis reveals the nature of connections in the networks, it does not provide insights into specific patterns, such as who is connected to whom, or whether participants with a low in-degree tend to connect with others of a low in-degree or those with a high in-degree.
To further examine differences, we calculated the Hamming-Ipsen-Mikhailov (HIM) distance. This metric combines the Hamming distance,65 66 which measures the number of matching edges between two networks (ie, trust dimensions at T1 and T2), with the Ipsen-Mikhailov distance, a ‘spectral distance’, used to assess the differences in the overall network structure.67 Spectral distances are useful for assessing global structural differences but may overlook variations in smaller substructures. The HIM distance68 ranges from 0 to 1, where a score of 0 indicates identical networks and 1 indicates opposite networks. For example, a complete graph (a network where everyone is connected to everyone else) compared with a graph with no edges would yield an HIM distance of 1.
Local versus national partners
To determine whether trust dimension networks differed between local and national partners, we stratified each trust dimension network accordingly by type (ie, local or national). Due to the small sample size and given their role in governing the national network, we combined the national office (n=1) and lead sites (n=7) with the national partners (n=10) into a single overarching national category. We then calculated the network measures described above at both T1 and T2 to examine changes in the trust dimension networks over time for each local and national partner.
Results
In T1 (May 2021), 57 individuals from the PPI Ignite Network were invited to complete the network survey, with 43 participants (75%) responding. By T2 (May 2022), one organisation had left the network, reducing the total number of invitees to 56. Of these 56 individuals, 33 individuals (59%) participated in the T2 survey. A detailed breakdown of participation by partnership type is provided in table 2.
Response rate by partner type
Table 3 presents the network-level measures calculated at T1 and T2 for the seven trust dimensions, including new collaborations. Over time, the number of connections (those who agree or strongly agree with that trust dimension) and the average in-degree (the number of incoming edges) decreased. On average, participants received approximately one fewer incoming connection compared with T1. This indicates that individuals were agreeing and strongly agreeing slightly less often on trust statements at T2.
Network-level measures over time*
The mean clustering coefficient, which measures the number of trust triangles, also declined over time. Similarly, the number of reciprocal edges (where trust is mutual between two individuals) decreased at T2. For in-degree centralisation, there was a slight increase for most trust dimensions, except for trust dimension 5 (shared values, visions and goals) and trust dimension 7 (reciprocity). This suggests that over time incoming trust connections became slightly more concentrated among an individual/group of individuals. Despite this subtle increase, the networks remained relatively decentralised at both timepoints.
In general, the magnitude of change in each network measure varied depending on the trust dimension. For instance, trust dimension 6 (power-sharing and co-ownership) consistently reported one of the highest total number of connections (ie, the trust statement most likely to receive agreement or strong agreement) at both timepoints. However, trust dimension 6 also had the greatest decrease in connections over time, indicating the largest decline in agreement compared with other trust dimensions. In contrast, trust dimension 2 (integrity) had one of the lowest numbers of connections at both timepoints, reflecting fewer individuals agreeing or strongly agreeing with the integrity statement. Additionally, trust dimension 2 exhibited the smallest change in connections over time, suggesting relative stability across timepoints. For other network measures, such as mean clustering coefficient (ie, average number of trusted groups), there was no change from T1 to T2 for trust dimension 4 (ability). However, a relatively large reduction was observed for trust dimension 5 (shared values, visions and goals) over time.
Network measures also revealed that certain trust dimension networks exhibited similarities. For example, trust dimension 1 (vulnerability) and trust dimension 2 (integrity) showed comparable network measures, as did trust dimension 5 (shared values, visions and goals) and trust dimension 6 (power-sharing and co-ownership). However, while dimensions 1 (vulnerability) and 2 (integrity) were similar to each other, they were notably different from dimensions 5 (shared values, visions and goals) and 6 (power-sharing and co-ownership). This finding was consistent over time.
Kolmogorov–Smirnov (KS) test
After calculating the KS statistic, a non-parametric test for comparing two probability distributions, we did not find a statistically significant difference in the in-degree distribution across any of the trust dimensions from T1 to T2. This suggests that the two samples were drawn from the same distribution. Such consistency aligns with expectations, as participants ideally name others in a similar way across timepoints, rather than thoughtfully at T1 and randomly at T2.
However, the KS statistic revealed some variation among the trust dimension networks over time. For instance, trust dimensions 3 (reliability) and 6 (power-sharing and co-ownership) showed the largest KS statistic (0.20), indicating relatively greater changes over time, while trust dimension 2 (integrity) had the smallest KS statistic (0.09), suggesting minimal change. Although these changes in KS statistic were subtle overall, the variation highlighted differences across the trust dimensions and how they evolved from T1 to T2. For plots, see online supplemental file 2.
Supplemental material
Hamming-Ipsen-Mikhailov (HIM) distance
As our networks had a low density of connections, we recognised that HIM distance, exploring whether connections between individuals change over time, would never be close to one. Therefore, we focused less on the overall magnitude of the HIM distance value and more on the relative differences across trust dimensions. We observed a small range in HIM distance across the dimensions of trust, ranging from 0.08 to 0.12. Specifically, networks for trust dimension 2 (integrity) (HIM=0.08) were more similar from T1 to T2 compared with trust dimension 5 (shared values visions and goals) (HIM=0.12). See online supplemental file 3 for further details.
Supplemental material
Indeed, both the KS statistic and HIM distance revealed greater differences across trust dimensions than within each dimension from T1 to T2.
Local versus national comparison
Findings for the weighted mean in-degree and clustering coefficient by type of node (local vs national) T1 and T2, are presented in table 4. We observed a decrease in the weighted mean in-degree (ie, average number of incoming connections) for both local and national partners across trust dimensions over time. However, the weighted mean in-degree was higher for national partners than the local partners at both timepoints, across all trust dimensions. This suggests that individuals who were national partners or site leads in the PPI Ignite Network received more trust nominations (ie, more people agreeing or strongly agreeing with trust statements about them) compared with local partners. We noted the largest difference between local and national partners in trust dimension 5 (shared values, visions and goals) at both T1 and T2, while trust dimension 2 (integrity) showed the smallest difference between these groups over the same periods. In contrast, the clustering coefficient did not show consistent trends across partnership type. For example, at T1, some local partners exhibited a higher clustering coefficient (ie, more trust triangles) compared with national partners. However, by T2, these trends reversed, with local partners having a lower clustering coefficient for certain trust dimensions. This was evident in trust dimensions 5 (shared values, visions and goals), 6 (power-sharing and co-ownership) and 7 (reciprocity).
Network measures for trust dimensions at T1 and T2 stratified by local versus national node type
Figures 1 and 2 illustrate the networks for these two trust dimensions over time. The networks appeared less dense for trust dimension 2 (integrity) over time (figure 1), while they became denser for trust dimension 5 (shared values, visions and goals) (figure 2). This indicates that the rate of agreement differed across these trust dimensions, highlighting an important nuance detected when examined as distinct networks. Additionally, we observed that more partners were disconnected from the networks in T2 compared with T1 for both dimensions of trust. This is particularly pronounced for trust dimension 2 (integrity). The disconnection of partners suggests that trust connections for these partners no longer existed at T2 for the respective trust dimension.
Trust dimension 2—integrity—at T1 and T2. Note: Blue arrows indicate T1 connections. Green arrows indicate T2 connections. The size of the node pertains to the number of incoming nominations for that individual. A larger node has more people ‘agreeing’ or ‘strongly agreeing’ with that statement of trust about them. NP = National Partner; LP = Local Partner; SL = Site Lead; DP = nominated but did not participate in network survey; NN = nominated but not in the PPI Ignite Network.
Trust dimension 5—shared values, visions and goals at T1 and T2. Note: Blue arrows indicate T1 connections. Green arrows indicate T2 connections. The size of the node pertains to the number of incoming nominations for that individual. A larger node has more people ‘agreeing’ or ‘strongly agreeing’ with that statement of trust about them. NP = National Partner; LP = Local Partner; SL = Site Lead; DP = nominated but did not participate in network survey; NN = nominated but not in the PPI Ignite Network.
Persistent connections in T1 and T2
Findings comparing network measures of persistent connections (ie, the same person nominated in both T1 and T2) are outlined in table 5 (non-weighted properties such as clustering coefficient and reciprocity are not included as they would not change over time as we are only including persistent connections). In this analysis, individuals who formed new collaborations and nominated new individuals in T2 were excluded. This approach allowed us to focus specifically on trust connections persisted over time. Interestingly, the average number of weighted incoming connections across all trust dimensions increased from T1 to T2. This suggests that, for individuals who were nominated consistently across T1 and T2, the level of agreement regarding trust statements increased.
Network-level measures over time*
In summary, findings are highlighted below:
An SNA approach revealed subtle changes over time when exploring trust multidimensionally in the PPI Ignite Network. On average, there was a slight decrease in trust connections across each trust dimension from T1 to T2 on a global level. This indicates that, at the second timepoint, fewer individuals agreed or strongly agreed with a given trust statement about the individual they nominated in the network compared with the first timepoint. However, trust connections that remained consistent over time showed an increase across all dimensions of trust.
More distinct differences emerged when stratifying trust by partnership type (ie, local or national partners). National partners and site leads in the PPI Ignite Network received more trust nominations, meaning more people agreed or strongly agreed with trust statements about them, compared with local partners.
Discussion
This case study extends the work by Gilfoyle et al 53 54 by comparing the dimensions of trust across two timepoints, stratified by local or national partnership types, and more broadly, by contributing to the conceptual and operational gaps related to trust in participatory research partnerships.53 54 By analysing the different dimensions of trust as separate networks, we identified changes across these timepoints and provided empirical support for a comprehensive, multidimensional exploration of trust as it evolved within the PPI Ignite Network.
Our analysis revealed a general decrease in the number of trust connections across most trust dimensions over the two timepoints at the network level. However, trust connections that were consistent from T1 to T2 showed increases across all trust dimensions, suggesting that when partnerships were maintained from T1 to T2, trust increased. Comparatively, the slight overall decrease in trust across the PPI Ignite Network may reflect the formation of new collaborations (eg, new employees or partners given staff and partnership turnover and/or interacting with new people depending on their work package), where trust had not yet been established/sustained. This aligns with existing literature, which emphasises that trust must be built and sustained over time, while new collaborations or changes in personnel can impact its development and maintenance.21 69
We also observed that some trust dimensions were more similar both visually (eg, network maps) and across network measures (based on the KS test and HIM distances), such as vulnerability and integrity. Others were markedly different, like integrity and shared values, visions and goals with a higher number of incoming connections for national partners compared with local partners. These findings contribute meaningfully to the literature by providing empirical support for using SNA to operationalise trust in a comprehensive, context-sensitive and multidimensional way over time. This approach avoids treating trust as a composite measure, which can overlook the unique influence of individual’s trust dimensions in a PHR partnership. This distinction is critical as PHR emphasises the need for contextually derived and driven knowledge production to address the needs of the communities,7 70 as highlighted in the CBPR conceptual model.17 Operational techniques must, therefore, consider the partnership context so that partners can both understand and evaluate if their goals are being met and if they are on a trajectory toward success. A lack of contextual consideration is a limitation of traditional quantitative methods,71 yet a strength of SNA. By incorporating both individual and system-level perspectives, SNA captures complex social-relational processes, like trust, while accounting for the social context and its influence on individuals within it.72
Using SNA provided valuable “insight[s] into the relationships, positions, structure and strength of [the] network”73 (pg. 4) across two timepoints. Through network maps, we observed where trust connections existed or were absent in the PPI Ignite Network over time, while also gaining an understanding of the implications of individual positions and the overall network structure. For example, central actors—individuals occupying highly connected positions within the network—are often viewed as opinion leader with prestige and influence.31 These actors play a critical role in the diffusion of ideas and behaviour,31 which has important implications for the trust-building process. By equipping the PPI Ignite Network members with a better understanding of their network structure, SNA can guide strategic interventions (ie, strategic actions that or remove links between social entities74) within the trust dimension networks to ensure trust is built and maintained throughout the next 5 years of working together and beyond.
For instance, partners can identify areas of weakness in the trust dimension networks, such as areas of fewer connections or individuals positioned on the periphery of the network, and take deliberate action to strengthen these areas. This could include fostering strategic collaboration opportunities between central individuals (thought to have higher influence) and peripheral individuals (who have fewer connections in a given network).
Equipping partners with an enhanced understanding of the trust development process within their specific context could in turn dictate the strategic allocation of (often limited) time and resources to enhance trust and, ultimately, partnership functioning. For instance, given the finding that local partners were less central compared with national partners (ie, had fewer collaborations and incoming connections across trust dimensions), immediate interventions could include creating opportunities for local partners to have more influence within the PPI Ignite Network. This might involve offering local partners leadership roles in key initiatives or work packages. If partnership capacity is an issue, interventions could focus on the (re)distribution of resources and providing more targeted supports for local partners. Indeed, conceptualising and operationalising trust in this manner also helps to address a significant gap in the PHR literature. As noted, “the majority of trust and community-based participatory research literature conceptualised trust as an outcome and acknowledges that research on trust development is lacking”29 (pg. 62).
Limitations
Although embracing context is important, readers should consider this when interpreting and/or applying findings to their own research. This case study examines a small network with two timepoints over a year. Considering that trust takes time to develop, surveying trust at only two timepoints may be restrictive. Additionally, not all partners in the PPI Ignite Network participated, and some who did participate did not complete both network surveys. To facilitate comparisons across timepoints, those who did not complete both network surveys were excluded, resulting in a smaller sample size. As such, the views reflected in case study might not be representative of the entire PPI Ignite Network and should be interpreted accordingly. However, consistent with findings from previous work,54 network properties differed only at the second decimal place, suggesting that the smaller sample size likely had minimal impact on the results. Furthermore, as trust is inherently contextual, its evolution will likely vary depending on the partnership of interest. This variability should be considered when applying these findings to other settings. Finally, while this case study employed a novel approach to operationalising trust across different contexts, it does not reveal why the networks evolved as they did. This limitation is addressed in a follow-up study published elsewhere (see75).
Future research
Areas of future work could investigate if the conceptualisation and operationalisation of trust within the PPI Ignite Network led to improved partnership outcomes. For instance, ‘readiness’ for public and patient involvement at a national level and within individual institutions was a priority outcome of the network. Future studies could examine whether changes in trust networks are associated with achieving the PPI Ignite Network’s objective of building capacity for PPI readiness. Additionally, future research could explore whether certain trust dimensions (among the seven identified) are particularly relevant to certain aspects of the CBPR model.7 16 For example, the CBPR model emphasises power dynamics as a critical factor influencing both context and partnership processes.76 With our enhanced understanding of trust—particularly the trust dimension ‘power-sharing and co-ownership’—it may be possible to identify where power dynamics exist by pinpointing asymmetrical trust relationships within this trust dimension network. Finally, as this is a case study exploring trust in one context, future work could expand to explore the trust development process in other PHR partnerships to compare findings across multiple study contexts.
Conclusion
This case study employs a novel and interdisciplinary lens, integrating insights from both the social network and PHR literature, to further clarify important conceptual and operational complexities of trust. By extending the findings of Gilfoyle et al,53 54 we consistently and comprehensively analysed trust over time in a real-world partnership, the PPI Ignite Network. The findings provide empirical support for using SNA to examine the evolution of trust as a multidimensional concept in PHR partnerships over time. Future research could consider exploring trust over more extended periods to gain deeper insights into its development and sustainability in different contexts.
Data availability statement
The data that support the findings of this study are available from the corresponding author, JS, upon reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by the University of Limerick Education and Health Sciences Research Ethics Committee (#2021_03_16_EHS). Participants gave informed consent to participate in the study before taking part.
Acknowledgments
We acknowledge the collaborating partners from the PPI Ignite@UL team for their contributions, including members of the Research Advisory Group, as well as members of the national PPI Ignite Network for their support for the study.
References
Footnotes
X @jsalsb
Contributors All authors have made substantive intellectual contributions to the development of this study. MG conceptualised and led the study, drafted and edited the final manuscript. MG is responsible for the overall content as guarantor. PMC analysed the data and contributed to the study design, data analysis and interpretation, writing and editing of the manuscript. JS secured funding, contributed to the study conceptualisation, data analysis and interpretation, and contributed to the writing and editing of the final manuscript. AM contributed to the study conceptualisation, writing and editing of the final manuscript. MMC contributed to study conceptualisation, interpretation of results, reviewed and approved the manuscript. ChatGPT4o was used only in the revision stages of this manuscript to help with grammar and sentence structure, like removing redundant language, suggesting word synonyms to reduce repetitiveness and suggestions to reduce long sentences into shorter, more digestible sentences. All authors verified and reviewed final content. This has also been added to Contributorship Statement in line with the BMJ AI use policy.
Funding This work was supported by the GEMS-10 scholarship to MG from the University of Limerick (Ireland); and a scholarship from the Integrated Knowledge Translation Research Network Canada: CIHR Foundation Grant; [FDN #143237 to MG]; and by Science Foundation Ireland [grant number 16/IA/4470 to PMC].
Competing interests None declared.
Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.