Structuring successful collaboration: a longitudinal social network analysis of a translational research network

In previous work [7, 10‐12], we examined the establishment and growth of a TRN in the field of T2 cancer research from its foundation in 2011 over its 4 years of operation. The TRN was established after achieving funding as a result of a competitive grant process and was largely built around the core foundational members’ current and prior research contacts. Its 68 members were drawn from a range of hospitals, universities and research centres, and members were from disciplines as diverse as palliative care, protein chemistry, health economics and genetics. A governing body made up of 14 members was established to guide and oversee the network. A network director, manager and several network operations staff were appointed. The TRN was focused on T2 cancer research: introducing clinically proven knowledge of cancer processes, diagnostic or treatment regimes into routine clinical practice. The TRN also provided access to funding for 1-year projects, conference and professional development grants, shared databases and facilities and support from project officers and translational research fellows. Membership was open to researchers, clinicians and students of the member organisations who applied and was free. A consumer advisory group made up of people with cancer or who have survived cancer was convened to have input into research priorities and design.

This previous work [7, 10‐12] demonstrated that before the start of the network, silos and tightly bound structures pervaded activities [11]. Relationships were based primarily on profession (clinicians or researchers) and geographic location. Using a social network survey (collaboration survey #1) in 2012, we were able to show that the presence of the TRN helped to mediate network relationships, appeared to close the profession gap between researchers and clinicians and influenced increased connectivity between partners. These analyses, supplemented by interviews [12], also highlighted how key players acted as go-betweens to link people and resources and that the network manager was the primary broker in this process [10]. Geographic location was associated with members’ choice of strategic partners, with network participants preferring to work with other local network participants [11]. At that early stage, it was not possible to identify and therefore to link this to any network outcomes. The collaboration survey #2, conducted in 2013 after 18 months of operation, was administered to a much larger network (268 members). The response rate for this second survey (43 %) was much lower than the first round (76 %) so the results of the social network component were viewed with caution. Over a quarter of the respondents in the sample (28 %) stated they had changed their practice (primarily around universal consenting for a tumour tissue bank and the inclusion of consumers in research design) and acknowledged personal benefits from TRN membership.

This paper reports on the results of collaboration survey #3 in May 2015 after 4 years of network operation. Over this time, the TRN had lost and gained some members but the core of original members present at the foundation in 2011 had remained largely constant. Notably, there were now nine consumer advisors. Many of the questions were the same as the previous two surveys allowing a comparison with previous years.

This study has two objectives. We firstly aimed to evaluate the TRN’s performance over the 4 years of its operation in terms of translational research planning, projects, outputs and dissemination. Few papers providing longitudinal evaluative data of a TRN have been published. Secondly, the study informs TRNs generally or any other organisation seeking to drive collaboration by examining the structure of collaborative relationships and looking at key player roles.

The study examines collaboration between members but defines it as both formal and informal links. Collaboration has been defined as the interaction of two or more independent partners, each of whom bring a unique view or expertise to a common problem [13]. Collaboration is a key success factor in knowledge translation [3] and implementation [1, 2]. Studies that consider formal collaboration such as co-authorship on papers or funded research partnerships [14, 15] may miss the important hidden work of a socio-professional network [16] in the context of the complex adaptive systems of health [17] and research: the corridor conversations, the advice over the phone, the go-between’s role in securing resources or introducing to a key person. We designed the study to capture these informal relationships and used them, rather than just the formal ties of funded project partners, to assess TRN collaborative outcomes.

The response rate for the on-line survey was 79 % (192/244). Of these, 21 (9 %) respondents formally refused consent (question 1), leaving 171 completed surveys (70 %, 171/244) although some questions were skipped by some respondents. The average time to complete the survey was 14 min. Respondents from all subgroups of the TRN were represented. As in the first two surveys, worksites were divided into three groups (designated central, satellite and peripheral) for analysis, based on their geographic and administrative proximity [23]. Central sites had high proximity, satellite sites medium, and peripheral low. Sixty-three percent of respondents worked at a central site, 28 % at a satellite site and 9 % at a peripheral site. A comparison of respondents and non-respondents showed that they were similar in gender distribution (χ ² (1, n = 244) 0.87, p = 0.35) but representation from central, satellite, peripheral sites were dissimilar (χ ² (2, n = 235) 6.97, p = 0.03). Satellite sites were least well represented and peripheral sites the best. All but one consumer participated (8/9). Table 3 summarises these results.

Nearly half of the respondents (80 or 47 %) chose “researcher” as their self-title, 22 (12 %) respondents chose “clinician”, 16 (9 %) chose “manager”, 18 (11 %) chose “clinician-researcher” and 20 (4 %) chose “other”. These “others” were aggregated into one of the four categories based on their “If other please specify” response plus their response to their main activity. For example, research nurses were aggregated with “clinician-researcher” and doctoral students to “researcher”. Five percent (8/171) of respondents were consumers.

There were 162 responses to the question: “Briefly, what is the main objective of the TRN as you understand it?” Answers matched fairly closely to the stated overall goal of the TRN, i.e. “to develop a sustainable translational research engine and to apply it to identified areas of need”. Frequently used words were research (126 times), collaboration/network (38), translation/al (42), facilitate/support (31), consumers/patients (17), grants/funding (7) and bridge/bring together (6). Notable was the absence of the concept of sustainability; it was only mentioned in two answers.

Sixty-nine respondents (40 %) nominated 77 members who had invited or influenced them to join the TRN. The network director was nominated the most (16 times), followed by two researchers five times each, and a large peripherally located research group was named four times.

Personal impacts of the TRN were assessed using a Likert scale of responses (from “not at all or not yet” to “to a large extent”). About two thirds of the respondents felt that TRN had, at least to some extent, increased their knowledge and skills, increased their career and networking opportunities, provided access to research opportunities, support and expertise and had been involved in TRN activities. The most positively viewed statements were TRN has given me opportunities to meet and/or work with new people (68 % scoring “To some extent”—“To a large extent”); TRN has given me access to useful resources (68 % scoring “To some extent”—“To a large extent”) and The time I have invested in TRN has been useful (70 % scoring “To some extent”—“To a large extent”). On this last statement, 18 % answered “To a large extent” The least positively viewed statement was TRN has got me interested in taking part in research where 25 % answered “Not at all/not yet”.

Consumer engagement was seen as having a number of benefits. Examples of responses are consumers provide valued input into early research design and planning for trials, they provide helpful feedback on design of consumer aids and meeting consumers adds passion to a project when one can “put a face” on one’s objective.

Eighty-two percent of respondents (133/162) had contact with other TRN members or staff. For the 29 (18 %) who had not had any contact, this was the final question in the survey. Contact with staff (77 %, 103/133) and attendance at a formal TRN meeting (75 %, 100/133) were the two most nominated activities.

Most respondents (77 %, 103/133) had not been involved in one of the TRN-funded projects over the last 4 years. Twenty-nine (22 %) respondents nominated other projects not funded by TRN that had resulted nonetheless through TRN membership. Table 4 summarises the projects nominated, illustrating their broad nature.

Figure 1a–c show the collaboration network diagrams. Each point (node) represents a respondent, and each line (tie) represents a collaborative link. The structure of this network was analysed using UCInet [21]. The number of ties each respondent has (degree) and the number of times a respondent lies on the shortest path between two other respondents therefore being in a brokerage position (called betweenness centrality) were computed for each respondent.

Network size was 244, density was 3.9 % and there were 1658 ties. The average number of ties per node was 11, and the range was 0–168 ties. There were 987 ties between members who knew each other before the network started and 671 ties between people who did not know each other before the network started.

Respondents with the highest involvement with other respondents were the network manager (ID number, 206), a TRN staff member (213), a clinician-researcher (36) and the network director (131). For both outdegree and indegree (number of ties the member reports and the number of ties reported by others directed to them, respectively), 206 and 213 are the two most central actors. They were also the two people who were in the strongest brokerage positions, that is who were on the shortest path between two members who were not directly linked. Tables 5, 6 and 7 give numbers of ties and betweenness centrality measures.

Members, on average, did not collaborate with a member of the same site group (E–I index, −0.88; p = 0.28) or with someone with the same self-title (E–I index = −0.88; p = 0.40) more than they did with members from outside these groups. In other words, there is no obvious pattern of clustering around sites or professions when averaged across the whole network.

The social network questions were deliberately placed at the end of the survey, after questions about consideration of personal impacts, wider outcomes and specific involvement in TRN events and activities. These diverse aspects of TRN involvement “primed” people to answer the final question using a broad socio-professional understanding of collaboration: “In the next and final question we want to know with whom you are collaborating. By “collaboration” we mean either formally (e.g. on a funded project) or informally (e.g. have discussed aspects of research, supplied expertise, advice or equipment to others) … Please select those people with whom you are currently collaborating on a TRN activity, event or project …” This allowed us to capture informal collaborative ties as well as the formal.

Responses of the 16 participants who reported the most collaborative ties (over 20 each) were examined to drill into the specific activities they had in mind. The majority (75 %, 12/16) reported involvement in five or more of the seven activities listed (Fig. 2). This analysis also showed that half (50 %, 8/16) were actively involved in a TRN project (confirmed by their answer in the next questions about TRN-funded project involvement (Fig. 3), and 81 % (13/16) had taken part in an informal meeting, discussion or email exchange about the TRN. This confirms that respondents saw both formally defined activities (such as investigator on a funded project) as well as informal activities such as email discussions or giving advice as collaboration.

Detailed results from survey #1 have been reported elsewhere [10‐12]. Table 8 summarises the main findings from the three collaboration surveys. Over this 4-year period, the network size grew from 68 members, to 263, then dropped back to 244. Figure 4 compares respondents’ self-title across the three surveys. A Pearson’s chi-square test (with α = 0.05) showed that there was no significant difference in the proportions of the different self-titles taking part: χ ² (6, N = 271) = 10.72, p = 0.10. Density of the collaborative ties network in survey #1 (4 %) was the same as the density of collaborative ties in survey #3, but the size of these two networks was very different. For survey #1 only 26 members reported having a collaborative tie; there were 106 ties reported and only one tie was between two members who had not known each other before the TRN started. In contrast, collaborative ties in survey #3 involved 171 members and had 1658 ties, 671 of which were between people who did not know each other before the TRN started. The network manager appears in all three surveys as the central actor who interacts with the most other members. The manager is also the top broker in the network in both surveys #1 and #3.

Respondents of survey #1 were asked how they became a member. Half stated they were invited to join (50 %) and/or were involved in the original grant proposal for the network (47 %). Four percent joined on their own initiative while 10 % selected “other”. In survey #2, 38 (34 %) respondents stated they had been “involved from the start”, 44 respondents (44 %) stated they had been invited to join and 12 (11 %) said they joined on their own initiative. The network director and researcher 44 are key people in surveys #2 and #3 inviting others to join the TRN.

The survey provides evidence of a well-functioning network which has grown in size and complexity over the 4 years of its operation. Shared understanding of purpose is a key success factor in collaborative networks [24], and there is evidence that TRN members have this shared vision. The free text answers to the objectives of the TRN were closely aligned with the formally stated goal and emphasised research, translation, collaboration between researchers and clinicians, support, networking and consumer involvement.

There is also ample evidence of positive personal impacts of the TRN, wider network outcomes developing over time and a range of changes in practice. In addition to funded projects, TRN members are also collaborating on non-TRN-funded projects showing a positive influence on the wider research environment. Respondents reported attending formal events as well as being involved in less formal activities such as email exchanges about TRN issues.

Collaboration is widespread across the network both with existing and new contacts. Averaged across the whole network, clustering by site or within the self-title groups of clinician or researcher is not seen. A possible reason for this is the mixing effect of collaboration with consumers (who were coded as a separate site and self-title group) and the strong collaborative links with TRN Operational staff by many members of the network.

This pattern of widespread collaboration is evidence of a well-functioning network. For translational researchers, there is a clear need to collaborate. The very title “translational” points to moving from one domain to another or more graphically bridging the “valley of death” between bench and bedside [25]. Expertise from the research and clinical domains are both needed yet the increasing specialisation and “niche” nature of many research and clinical fields widens the valley [26, 27] and means that silos emerge more readily. Collaboration has long been hindered by these silos as they become isolated and introspective [28‐30]. Language and concepts can be so specialised that it can be difficult to search effectively for a partner. Funding bodies expect these collaborations, not just across disciplines but to include consumers who can make a valuable contribution to research design.

Finding an appropriate collaborative partner is recognised as a barrier for practice-based TR involving more than just picking a name from a Google search or even a dedicated TRN website. Schleyer and colleagues [31] investigated the process of how biomedical researchers find potential interdisciplinary partners. They found that researchers highly prized compatibility in a potential collaborator in both personality and work style. This had a strongly social element in that they would more likely trust a personal recommendation from a colleague than a compatibility “hit” from a search engine on a website. The social structure of the TRN allows this level of facilitative social interaction. Another barrier is that disciplines can be so complex and specialised that researchers can often not know what they are looking for outside their own field. Again, the assistance of a boundary spanner in the TRN with some experience of both fields is invaluable [12]. The TRN members in key brokerage roles appear to be mediating these introductions in the TRN. Researchers may also suffer from power differentials in seeking a collaborator. Junior researchers or “less powerful” professions such as nursing or allied health with a limited social professional network can find it hard to find an appropriate partner, and it may be intimidating or awkward if a marked power differential exists [31]. Again, having go-betweens that can broker these collaborative introductions is a clear advantage in the TRN.

The number of new links between people who did not know each other before the start of the TRN has grown markedly since the first collaboration survey in 2012. At that time, only one respondent reported a collaborative tie with a person they had not known before involvement with the TRN. In this survey, 671 new ties are reported, giving clear evidence of the networking and collaborative opportunities TRN gives members. The TRN manager, director and staff have strong brokerage potential as do many of the governing body members. We know from previous work with the governing body [12] that their network position as a broker is enacted though giving advice, acting as a go-between or facilitator and sometimes as a “gate-jumper” (the opposite of a “gate-keeper”, i.e. achieving access to difficult to get resources or people). A strong case for the influence of the TRN on building this interdisciplinary collaboration can therefore be made.

Considering these 671 new collaborative ties, further, we examined whether these new ties were limited to TRN staff who may reasonably be expected to make new contacts as part of their job. By removing the network manager (206), the director (131) and operational staff (213) and (222), 283/671 ties (42 %) remained between respondents who did not know each other before the TRN started. This shows both that the TRN staff have been instrumental in building connectivity in the network and that other members have been able to find collaborative partners within the network through other means. The TRN website, for example, is frequently updated and features the work of members, providing a ready way to contact potential partners. Survey #3 showed members were involved in a range of TRN events and activities, which would also increase their chances of meeting a potential partner.

This survey has been able to show that the TRN’s impact goes well beyond outcomes from formal TRN-funded projects. Only a small proportion of respondents were formally involved over the last 4 years with one of the TRN-funded projects (Table 9). This contrasts with our broad definition of collaboration that the majority of respondents reported (also reflected in differences between the members saying they were involved in TRN-funded projects and those officially listed (Table 9)). It also contrasts with the significant number of respondents reporting collaboration on other non-TRN-funded projects (Tables 5, 6 and 7). The diversity of these projects is large.

There is evidence that most people attribute positive personal impacts and wider outcomes to their involvement with the TRN. The most positively received statement was “The time I have invested in the TRN has been useful” closely followed by the “The TRN has given me opportunities to meet and/or work with new people”. This latter statement is borne out by the number of new ties reported in the network. Translational research is focused on patient outcomes through changes in clinical practice so the long list of instances of changed practice that members attribute to the TRN is encouraging.

It is clear that consumer engagement facilitated by the TRN has had a significant impact. All nine consumers are people currently living with, or who have survived cancer, or people who have cared for a family member with cancer. They have all undergone research facilitation training. All formal grant proposals are presented to the consumer group for feedback. Comments on consumer engagement given in the survey included:

Consumer input to our project and having a cancer survivor speaking to the students was invaluable. When you put a face to the problem, it becomes personal and this made everyone in the lab more motivated to find a solution (Researcher, University 1)

My engagement with consumers has helped formulating my research proposal (Researcher, Laboratory Setting)

I believe researchers have a new understanding of the value of consumer representative involvement in their research - in all the contacts I have had with researchers in [the university] and other institutions (Consumer)

We have become more aware of the need to involve consumer advocates/representatives in our grant proposals. Hearing about how they think about issues of relevance has been an eye-opener. The TRN's consumer panel has been a useful avenue through which we were introduced to this (Researcher Research Group 4)

I have helped one researcher create forms for patients to become involved in trials, and will be helping another in the coming months (Consumer)

The tumour tissue bank was nominated frequently by clinicians as being a catalyst for change in their practice. Both the bank and consumer involvement were mentioned in survey #2. New to this current survey is work being done on diagnosis and management of hereditary cancers.

This wealth of longitudinal data shows that silo-spanning, translational collaboration can be facilitated by strategic social interactions to result in positive personal impacts and wider outcomes. TRNs can become large and unfocussed if many new people join in an unstructured way. New members that join through the personal invitation of key members allow a manageable and strategic growth in numbers.

TRN staff such as the director, manager, operational staff and other key players can facilitate introductions of members from different areas to one another thereby solving a major barrier to strategic collaboration. Their knowledge of the wider research and clinical community allows them to target people to link up rather than a broader scattergun approach. These key players also have a significant role in facilitating access to resources. This has two implications for those managing TRNs: (1) people recruited to these roles should be good communicators and be active in making personal connections with members. (2) orientation to the organisations and familiarisation with the context in which they are working is important for them to have that overarching strategic role, and time should be invested in this. Once the introductions have been made, collaborations appear to flourish as seen by the number of non-TRN-funded partnerships members have initiated with other members.

A warning is also apposite here: the key players in the network need to share the role so that all the brokerage activity is not being done by a single actor. If this solo actor leaves the TRN, their absence is likely to disrupt the function of the network, fragment groups, or at the least slow the formation of new ties.

Consumer involvement is a powerful tool for designing high-quality, relevant studies as well as driving positive personal impacts and wider outcomes. Recommended is a formal process for recruiting and training appropriate consumer representatives to optimise their input into research design and practice. Again, these collaborative links between consumers and network members can be brokered by TRN staff.

The study achieved a strong response rate (79 %) after careful preparation and concerted follow-up. The comparison of non-respondents with respondents showed that most sites were well represented. The lowest proportion of responses (6/18 members, 33 %) came from the hospitals 2 and 3. Groups with high response rates were research group 3 (8/9 members 89 %—ninth member formally opted out), the consumer advisory group (8/9 members, 89 %) and university 2 (23/33 members, 70 %).

Social network surveys collect self-reported ties. The reliability of this was maximised by using a roster style format, where names of all members were provided. This meant that respondents were not relying on recall alone. The list of 244 members had the potential to impose a huge respondent burden, but this did not seem to be a problem. The average time of completion was 14 min, and verbal feedback from the pilot verified that the roster of members, structured around research or clinical groups, was easy to work through.

While the formal measure of reciprocity of ties was low at 32 %, we can argue that this is not a significant issue in this context. The type of professional collaborative interaction we wanted to capture was deliberately broad, i.e. not just collaboration on formally funded projects but such things as casual conversations leading to loan of equipment, introductions to other key members and giving advice. The report of a tie by one party in this context could be argued to be indicative of an interaction of some sort which in this socio-professional context could justifiably be defined as collaboration. One striking example of this non-reciprocity is in the indegree and outdegree data of the most central actors. The network manager and TRN staff member 213 have nominated three to four times the number of ties to other members than ties others have directed to them (see Tables 5 and 6). This means that many people are forgetting or not defining as collaboration, contact with the network manager and staff. As TRN staff, they may be in a better position to accurately remember their interactions with other members.

This study examined a single translational network, but longitudinal data over 4 years shows a rich picture of how to foster cross-silo collaboration more generally. Additional file 2 discusses the methodological issues of adequate whole network response rates and respondent concerns over confidentiality.