Information seeking for making evidence-informed decisions: a social network analysis on the staff of a public health department in Canada

Study setting

The public health department serves primarily a large urban population. It has a total of 620 staff, and has assigned ‘professional consultants’ (n ~ 20) to practice-based teams to conduct rapid reviews of the literature to address practice-based issues. They conduct research, interpret and analyze data, and prepare specialized reports on the development and implementation of programs and policies.

Recruitment

All the staff members (n = 620) were invited to participate through an email sent from the Medical Officer of Health. The purpose of the project, details of methods, examples of a social network analysis, and the importance of their contribution were described.

Data collection

For the purpose of this survey, research evidence was defined as “primary studies, systematic reviews, and meta-analyses that evaluate the effectiveness of an intervention”. Discussions with colleagues, program evaluations, community needs assessments, client values and preferences, professional experience, and provincial guidelines were not considered research evidence in this study. In addition, we included a broad range of activities in our definition of ‘decision’; including decisions about how to implement programs/policies, how to address local issues that arise, and how to identify and respond to community needs.

The choice of five for the number of people to identify was made following the strategies of Coleman et al.[16] and West et al.[18]. Both pivotal studies on health professionals limited the number to three for each respondent, assuming that health practitioners tend to limit their networks to a small number of surrounding peers. We made a similar assumption. However, we chose a limit of five for the maximum number of names each person could provide, to reduce the possibility of ceiling effect. For each question, two additional response options were available: not connected to anyone, and prefer not to answer.

Two reminder emails, one week apart, were sent to those who had not completed the survey. Participation was voluntary and responses were kept confidential by use of participant code. The study was approved by the McMaster FHS/HHS Research Ethics Board, as well as ethics approval body at the study health department.

Data analysis and interpretation

The network analysis was performed using UCINET v6 [25] and NETDRAW [26] software.

Characteristics of study sample

The 196 respondents, on average had 10 years experience working in public health. Approximately 90% were female, and approximately two thirds had a baccalaureate degree. Respondents were most likely to work in the ‘chronic diseases’ (36%) and ‘communicable diseases’ (25%) divisions. Table 2 shows the baseline characteristics of all respondents to the surveys. Comparing the respondents and non-respondents to each network question, the only noticeable difference was the mean years of experience. Non-responders worked in public health for fewer years than responders (7.2 year vs. 10.7 year in information-seeking network).

The network sociograms are provided and explained in each section. Figures 1, 2, and 3 correspond to the information-seeking, recognition of expertise, and friendship networks respectively. At each network, the nodes represent the people, lines between nodes represent the connection, and the arrow heads signify the directionality of connection. The shapes of the nodes denote the organizational division of each person. The size of the nodes is proportional to specific centrality measures of that person, which will be explained separately in the subsequent sections. The participant codes of the most central people, according to various centrality definitions are also provided in dix 1, along with their basic characteristics.

Information seeking (Q1Q2)

Ninety-eight percent of respondents identified 4 or less peers as their sources of information. The merged Q1Q2 network consisted of 170 nodes and 332 ties, and showed a large connected body, which consisted of a seemingly highly connected patch of nodes in the middle, and two less dense wings at two sides (Figures 1a and 1b). It had a density of 1.2%, and a dyad-based reciprocity of 19.9%. We divided the network into a managerial and non-managerial subgroup. The managerial subgroup consisted of the supervisors, managers, director, and Associate Medical Officer of Health (AMOH) (n = 47). The reciprocity of information-seeking connections still remained low in the managerial subgroup, with a value of 23%, although this was higher than in the non-managerial subgroup (11.6%).

Recognition of expertise (Q3)

The Q3 network consisted of 163 nodes, which were connected by means of 310 ties. The density of the network was 1.2%. Ninety-four percent of respondents identified 4 or less peers as experts. The dyad based reciprocity of Q3 network was 2.3%, which was the lowest, in comparison to the information-seeking (Q1Q2) and friendship (Q4) networks. It is supported by the fact that the recognition of expertise is not generally a bi-directional relationship. The shape of the sociogram for Q3 (Figure2) was fairly similar to the information-seeking network.

Friendship (Q4)

The symmetrized friendship network had 159 nodes, 302 ties, and a density of 1.2% (Figure3). Of the respondents, 100% identified four or less other individuals as friends. The reciprocity of non-symmetric network was 32%, which was the highest among the four networks.

Brokerage

The characteristics of the actors with the highest frequency of each brokerage role are reported below:

Coordinator

Coordinators are staff who connect people in their own division to others in the same division. The actors with the highest coordinator roles, big circles in Figure4, were a ‘professional consultant’ in division #4, a supervisor in division #5, a public health nurse from division #5, a nutritionist in division #1, and two dental practitioners also from division #1 (Additional file 1).

Distribution of connections

The sociograms suggest that the nature of inter-personal connections in the health department is localized, and public health staff generally turn to a handful of people within their own division to obtain information to assist in making practice decisions. This tendency has also been reported among other health professionals. For example, Keating et al.[31] showed that, primary care physicians in a hospital based academic practice had influential discussions with an average of 4 others during six months prior to the study. The reasons why health professionals may limit their ties to a small group of informants might be due to the importance of the ease of access to the information source, and the tendency to form circles of trust comprised of peers that are similar with respect to values and concerns [32].

Density

Innovations spread through social networks. People talk about new ideas and policies and share their concerns and experiences on a daily basis. Consequently, the level of dispersion of ideas and their internalization is higher in more connected organizations than detached isolated communities. High density and reciprocity of connections are indicators of strong ties among people, resulting in greater adoption of innovations [33]. However, the association between the network density and the success of adoption is not linear and straightforward. For example, when the innovation is intended to be diffused by community leaders, lower density may increase the adoption of new strategies due to strengthening the influence of leaders over their communities [34].

According to our study findings, the density of the information-seeking connections was very low (1.2%). However, it rose considerably when the divisions were studied separately. The densest division of the network was the office of the Medical Officer of Health, with a density of 13.6%. It was the smallest division in the network, consisting of 19 actors (of whom 11 responded to network questions), in which the people had relatively close connections with each other, as well as with the other divisions. Merrill et al. in a study on work-related communications between staff in a small public health department (with the size of 156 employees) reported a density of 15% [35]. Similar to our findings, they found that the overall density was much lower than the within-program densities (31-64%), and concluded that, between-program communications were much less than within-program ties.

However, it is not possible to make a conclusive statement as to whether the observed low density was a real feature of this network, or due to the under-reporting of connections and missing values. The large proportion of missing values, big size of the organization, reliance of respondents on memory instead of choosing from a roster list, and limiting the number of answers to 5 individuals, all signify that the findings of this study might have under-estimated the real interchange pattern of information in the health department.

Reciprocity

The reciprocity of information-seeking ties was around 20%. A priori, we did not expect a high reciprocity in the information-seeking network. Since, there were individuals in the department who professionally served as the consultants and information sources, the majority of the staff were turning to them for getting information; and reverse connections were not happening frequently. However, the reciprocity within the managerial subgroup was still low (23%). Creswick and Westbrook similarly reported reciprocity of 43% for work-related problem solving, and 26% for medication advice seeking in the emergency department staff of a hospital [36]. A high proportion of unreciprocated information-seeking is probably the result of compartmentalized professional activities in the department, meaning there are a few central information sources to whom many people turn to seek information. This may be due to the personal and professional capabilities of the central people in the network, lack of sufficient expertise for autonomous practice in the staff, the formal organizational bureaucracy, or a combination of all.

Central actors

A group of ‘professional consultants’, managers and epidemiologists were the most influential people in the department. ‘Professional consultants’ and epidemiologists were also recognized most often as being experts with respect to research evidence by staff. The centrality of ‘professional consultants’ in the information-seeking and recognition of expertise networks was expected from their formal responsibilities. The ‘professional consultants’, including project specialists and health promotion consultants, have the role of experts and professional advisors in the network. Their primary role is to conduct reviews of the literature (appraise, interpret and apply research evidence) so as to provide advice on the development and implementation of programs and policies.

The centrality of two ‘professional consultants’ in division #4 was extraordinarily higher than that of their counterparts in other divisions. The highly centralized roles of these two ‘professional consultants’ could be due to the geographical proximity of staff, the exceptional capabilities of these ‘professional consultants’, or specific tasks and responsibilities in this division. The high number of individuals seeking information from a few people may result in overload and reduced productivity over time. Reducing the work load of overly centralized people through redesigning alternative information flow routes could be considered as an improvement strategy [37]. Fujimoto et al.[34] in a network study on community leaders regarding adoption of evidence-based prevention strategies suggested that decreasing the centralization of advice networks and increasing the centralization of informal discussion networks improved adoption outcomes. In our study network, empowering the local information sources in division #4 in order to reduce the exclusivity of connections to two central actors, will probably provide the staff with more alternatives for obtaining the required information, and consequently may facilitate greater flow of information.

A number of managers in different divisions were among the people with highest centrality in information-seeking, recognition of expertise, and friendship networks. Managers have the responsibility of managing and controlling the programs and services. They have the financial and operational responsibility over the processes. They have the role of decision-makers and planners in the department. The mentioned formal roles explain the influential positions of managers who connected their divisions to the OMOH. West and Barron report a similar finding for hospital managers in acute-care hospitals in the United Kingdom [32]. They found that, both groups of doctors and nurses mainly turned to managers to discuss professional matters, and much less to each other. Managers had a considerable brokering role in connecting traditional clinical disciplines (physicians and nurses) in their hospital. They suggested managers could facilitate the communication between nurses and physicians as part of their professional role in the hospital.

Due to effects of organizational hierarchy, it was expected that seeking information from managers by staff with lower organizational status might be limited to formal connections. According to social exchange theory [38], people recognize the advisee’s status (importance) as an incentive to seek information; however, many people may be hesitant to consult prestigious figures because it highlights their own lower status and lack of knowledge and exposes them to the judgment of superiors. But surprisingly, the managers in each division were among the people with the largest network of friends in that division as well. This characteristic highlights the pivotal role of managers as central people in divisions, formally and informally; and could be considered as a valuable potential for organizational change. In a study on sales managers in a financial services firm, Mehra et al. found that the centrality of managers and group leaders in friendship networks with their sub-ordinates was positively and significantly associated with group performance and the leaders’ reputation [39].

Clusters and bridges

The OMOH formed a central cluster in the department, bridging different divisions, with the highest density in the network (12.4%). The OMOH is responsible for the management of public health programs and services; and advises the Regional Council; which are consistent with the central bridging place of the OMOH in our study. The exclusively central role of the OMOH as a directing and coordinating division of the health department may affect the knowledge flow between divisions. Tsai showed that, in multi-unit organizations, the more control headquarters exercised on their sub-units, the less those units interacted with each other [40]. While the importance of the OMOH should be acknowledged and nurtured, efforts also should be made to directly connect divisions with each other to shorten the inter-divisional distance. Research on multi-unit organizations has shown that, by spanning sub-unit boundaries and promoting inter-unit connections, organizational units obtain new knowledge more easily and complete their projects faster [41].

Summarizing various centrality and brokerage roles, we can assume a four-level semi-formal hierarchy of the information flow in this health department (Figure5). These include 1) the staff who have a limited circle of information sources and friends; 2) practitioners in each division who play the role of local coordinators; 3) ‘professional consultants’ and some managers who are the central figures and also representatives of their division; 4) and the OMOH at the top as the central bridge connected to all divisions.

Hierarchy is an important structural determinant influencing the diffusion of innovations [42]. Increasing the levels of hierarchy makes communication between levels more difficult, and therefore hampers the flow of information. A few people at the higher levels of the hierarchy have more opportunity to control the flow of information, and there is less horizontal flow in the lower levels. As shown in Figure5, in each division there were different local coordinators, who were connecting their neighboring peers in the division, and might or might not be directly connected to managers, ‘professional consultants’, and the OMOH. They had a variety of job titles (including ‘professional consultants’, supervisor, public health nurse, nutritionist, dental educator). These people could only be identified as local coordinators by means of social network analysis, because their position in the network is not defined in their formal job descriptions. Identifying the local coordinators, who were mainly practitioners and clinical staff, and enhancing their connections to the professional information sources, including the ‘professional consultants’ and epidemiologists, could be an organizational intervention to reduce the number of hierarchical levels in information flow, and facilitate the adoption of EIDM as a sustained organization change.

Implications for practice

Future research

To our knowledge, this is the first study investigating the organizational structure and the formal and informal roles of staff of a public health department in Canada with respect to EIDM. The generalizability of the observed semi-formal hierarchy of information flow could be assessed in other public health organizations. There is also a need for more in-depth analysis of the roles and effectiveness of ‘professional consultants’ in public health organizations in building capacity for EIDM. In addition, the role of the OMOH as the bridging division, and its potentials for facilitating brokerage and communication between health disciplines should be investigated in interventional studies.

This study provides a foundation for longitudinal network analysis of the effect of an organization-wide, tailored KT intervention on the capacity of one health department for EIDM. The evolution of the network over time, as well as the role of a tailored KT intervention on the overall shape and specific characteristics of the network are being assessed in the larger study. The impact of various network characteristics on the effectiveness of the KT interventions being implemented in the larger study, and which have been influenced by these SNA results will be investigated by the primary author in future studies.

The notion of information seeking among the staff of the health department should also be investigated more thoroughly in a qualitative study, by seeing the social networks from the actors’ perspectives. While not commonly used in SNA context, a qualitative approach may provide important insights to aid in interpreting the findings from SNA studies.

Limitations

Some limitations in the current study were identified. People solely relied on their memory, and they might have easily forgotten to include someone on any or all of the four SNA questions. Recalling and enlisting the names of information sources and friends took much more time than was expected, which could explain why 32 people who participated in the larger study did not answer any of the social network questions.

While the majority of top level and influential staff answered the SNA questions, the large proportion of the staff who did not answer these questions is an important threat to external validity. It implies that, the findings regarding the managerial and professional staff in the network, who are in fact the main targets and promoters of the EIDM, are more dependable. The patterns seen in the peripheries should be used cautiously, since the people with the lower organizational ranks (practitioners and administrative staff) were under-represented in the sample. Nevertheless, other potential biasing factors which might have resulted in low response rate should also be taken into account; for instance, the non-responders might not be interested in EIDM, might not consider it relevant to their practice, might not see the importance of participation, or simply might be too busy to participate.

Additionally, there were a few managers and ‘professional consultants’ in each division who did not differ from other staff in terms of centrality. Due to the large proportion of non-respondents, it is not possible to conclude that managers and ‘professional consultants’ did not fulfill their consulting and coordinating roles very well. They might be central information sources for groups of people who did not answer the survey.

Even though we provided detailed explanations of the meanings of some terms like ‘evidence’ and ‘decisions’ in the questionnaire, we cannot exclude the possibility of miscomprehension and diversity in understanding of the meaning of these terms, as a threat to internal validity of observed patterns.