How the study of networks informs knowledge translation ...advancing KT science Also summarized is the use of theory in this SNA research to demonstrate the fit of SNA with theoretical

SYSTEMATIC REVIEW Open Access

How the study of networks informsknowledge translation and implementation:a scoping reviewStephanie M. N. Glegg1,2,3* , Emily Jenkins4 and Anita Kothari5

Abstract

Background: To date, implementation science has focused largely on identifying the individual and organizationalbarriers, processes, and outcomes of knowledge translation (KT) (including implementation efforts). Social networkanalysis (SNA) has the potential to augment our understanding of KT success by applying a network lens that examinesthe influence of relationships and social structures on research use and intervention acceptability by health professionals.The purpose of this review was to comprehensively map the ways in which SNA methodologies have been applied tothe study of KT with respect to health professional networks.

Methods: Systematic scoping review methodology involved searching five academic databases for primary research onKT that employed quantitative SNA methods, and inclusion screening using predetermined criteria. Data extractionincluded information on study aim, population, variables, network properties, theory use, and data collection methods.Descriptive statistics and chronology charting preceded theoretical analysis of findings.

Results: Twenty-seven retained articles describing 19 cross-sectional and 2 longitudinal studies reported on 28 structuralproperties, with degree centrality, tie characteristics (e.g., homophily, reciprocity), and whole network density being mostfrequent. Eleven studies examined physician-only networks, 9 focused on interprofessional networks, and 1 reported ona nurse practitioner network. Diffusion of innovation, social contagion, and social influence theories were mostcommonly applied.

Conclusions: Emerging interest in SNA for KT- and implementation-related research is evident. The included articlesfocused on individual level evidence-based decision-making: we recommend also applying SNA to meso- or macro-levelKT activities. SNA research that expands the range of professions under study, examines network dynamics over time,extends the depth of analysis of the role of network structure on KT processes and outcomes, and employs mixedmethods to triangulate findings, is needed to advance the field. SNA is a valuable approach for evaluating key networkcharacteristics, structures and positions of relevance to KT, implementation, and evidence informed practice. Examininghow network structure influences connections and the implications of those holding prominent network positions canprovide insights to improve network-based KT processes.

Keywords: Knowledge translation, Evidence-based practice, Social network analysis, Information flow, Scoping review,Network, Implementation

© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

* Correspondence: [email protected] Sciences, The University of British Columbia, 212 - 2177Wesbrook Mall, Vancouver, BC V6T 1Z3, Canada2Therapy Department, Sunny Hill Health Centre for Children, 3644 SlocanStreet, Vancouver, BC V5M 3E8, CanadaFull list of author information is available at the end of the article

Glegg et al. Implementation Science (2019) 14:34 https://doi.org/10.1186/s13012-019-0879-1

http://crossmark.crossref.org/dialog/?doi=10.1186/s13012-019-0879-1&domain=pdf

http://orcid.org/0000-0002-8164-6983

http://creativecommons.org/licenses/by/4.0/

http://creativecommons.org/publicdomain/zero/1.0/

mailto:[email protected]

BackgroundThe “science” of knowledge translation (KT), namelythe study of the processes, determinants, and out-comes of KT and evidence informed practice (EIP)efforts [1], is of high interest to researchers because ofthe significant challenges that exist in getting researchinto practice. To date, KT scientists have targeted theirefforts primarily at identifying the key processes ofKT, understanding the influences on these processes,and evaluating the effectiveness of various strategiesto support them [2–5]. However, identifying the spe-cific mechanisms by which KT strategies, particularlycomplex multi-faceted ones, have been effectiverequires additional research. Furthermore, limited in-formation is available to clinicians and researchersabout the role of social relationships and network con-nections in facilitating KT [6].Social network analysis (SNA) is a research para-

digm concerned with the patterns of connections (i.e.,ties) between actors (i.e., people or entities) within aninterconnected group or network, and how this “socialstructure” impacts outcomes of interest [7]. Key SNAterms that provide the context for this review are de-fined in Table 1, along with their application to thestudy of KT. Actors may be individuals, organizations,countries, or other entities; ties reflect the connectionsor linkages between them [7]. The structural charac-teristics of both whole and individuals’ networks canbe studied. Included with the paradigm are theories,such as network, graph, diffusion, and social influencetheories, and a set of methodologies that can be ap-plied across a range of substantive problems.Visualization, or mapping, and the use of network de-scriptive statistics can provide a visual and empiricalbasis for comparison across networks, including

identifying important strengths, gaps, or differencesbetween networks that merit further explorationthrough qualitative means. Statistical and computa-tional modeling can also be used to explain and topredict network-related phenomena, and to simulatethe complexities inherent in network dynamics.SNA offers an alternate perspective to behavior

change theory-based approaches prevalent in KT sci-ence [5]. These latter approaches focus onindividual-level factors influencing behavior change,often from a social cognition perspective [5]. Con-versely, SNA proposes a network-level perspective thatexamines how connections among individuals orentities, and the nature of the associated interactions,influence an outcome (e.g., accessing or sharing evi-dence, changing practice behaviors based on evi-dence). The paradigm respects the socially drivennature of innovation uptake, and the value inherent inexamining not only the processes involved in KT butalso the social structures and characteristics of thenetworks of relationships within which KT occurs.Examples of network-related KT processes that can beexamined from a SNA lens include one-way versustwo-way exchange of information, the timing and pre-diction of evidence uptake by different types of indi-viduals, the influence of specific types of people onbehavior change, individuals’ capacity for change basedon their positions in the network, gaps in the flow ofor access to information or resources required for evi-dence use, and testing the effectiveness of strategies toaddress gaps or inefficiencies identified in thenetwork.Recent systematic reviews on SNA in health care

have focused on quality and patient safety initiatives[8, 9], on a single profession (i.e., nursing) [10, 11], oron only select network properties (e.g., the study ofbrokers) [12]. Some reviews focus on conditions (e.g.,obesity networks) [12] to explore possible networkinteractions for potential treatments. Given the com-plex and interprofessional nature of health care prac-tice, a study of the full breadth of health professionsand network properties is required. Furthermore,some of these reviews included non-health care litera-ture (e.g., from television production and corporatebusiness contexts) [13], or neglected to include socialsciences databases in which most SNA journals areindexed. The existing broad reviews of health profes-sional networks [6, 14, 15] do include some studies onKT-related phenomena (e.g., diffusion, knowledgetransfer); however, the majority of their content cen-tered on the study of social interactions that have im-plications for organizational functioning (e.g.,friendships, work task assignments, staff recruitment,social support trust), but were not linked directly to

Contributions to the literature

� This review synthesizes the KT literature employing a social

network analysis (SNA) approach to the study health

professionals, to demonstrate the utility of SNA for

advancing KT science

� Also summarized is the use of theory in this SNA research to

demonstrate the fit of SNA with theoretical approaches used

in KT research, including diffusion of innovation and

complexity theory

� This article acts as a reference tool for those considering

applying a SNA lens to their KT research, to support the

design of SNA-specific research questions, methodological

approaches, and measures

Glegg et al. Implementation Science (2019) 14:34 Page 2 of 27

Table 1 Social network analysis (SNA) terms and their implications for knowledge translation

SNA term (frequencycount)

Definition Implication for KT

Network An interconnected group of actors (e.g., people,organizations) [7]

Provides the social context within which KT occurs

Actor A point (node) in a network that represents an individual,organization or entity connected to other actors (throughties) [7]

Represents the people, teams, or organizations involvedin KT processes

Tie (2) The relations or connections among actors in thenetwork [79]

Represents the interactions, collaborations, orrelationships involved in KT Measures one- versus two-way communication, advice seeking, collaboration, etc.[7]

Dyad Pairwise relations between actors [7] Represents one of three levels of analysis for socialnetwork data (the others being individual node-leveland whole network level) [7]

Centralization

Whole networkcentralization (3)

Extent to which interconnections are unequal across thenetwork [21] (i.e., concentrated around one or more centralindividuals) [7]

Thought to enhance ease of knowledge sharing and topromote standard practices of existing protocols [80].Decentralization may support new innovations, butlead to mixed messaging and decreased clarity becauseof multiple information sources [72]

Centrality

Degree centrality (3) # of direct ties (connections) of an actor Seen as an indicator of visibility [81], prestige [39] orpower [79] resulting from lots of direct contact to manyothers

Indegree centrality (10) # of individuals who send (identify) ties to an actor Considered an index of importance [28] power orinfluence [40]

Outdegree centrality (5) # of direct ties an actor sends (identifies) to others [33] Used to quantify access to resources throughcolleagues, exposure to evidence and others’ practices;positively associated with EIP use [33]

Betweenness centrality(4)

Extent to which an individual is tied/connected to others whoare not connected themselves [40]

Used as a proxy for control of KT processes [39]; highvalues reflect a favorable position (e.g. brokeringpotential) [40] for information flow or power [79]

Flow betweennesscentrality (3)

How involved an actor is in all of the paths or routesbetween all other actors (not just those representing theshortest paths) [79]

Used to determine contributions of individuals towardteam decision-making; provides insights into structuralhierarchy [33] Used as a proxy for ease of bypassing thecore individuals in the network [39, 79]

Closeness centrality (2) Proportion of actors that can be reached in one or moresteps [79]

Proxy for degree of access to information [39] orefficiency in communicating with the network (relativereach) [7]

Bonacich centrality (1) Extent to which an actor is tied to others, weighted accordingto the centrality (e.g., popularity, importance) of those towhom the actor is tied/connected [79]

Proxy for power or hierarchy within a network; mayhelp to identify network fragmentation/brokeringopportunities [14]

Hubs and authoritiescentrality (1)

The structural prominence of individuals within a core-periphery structured network [32]

Proxy for importance [32]

Tie characteristics

Tie strength (7) Value associated with a tie/connection, e.g., frequencyof contact, emotional intensity, duration of connection, etc. [7]

Weak ties thought to increase access to newinformation/opportunities; strong ties seen as requiredfor innovation implementation [82]

Tie homophily (includesexternal-internal or EIindex) (13)

Similarity of connected actors/nodes on a given attribute [7] Similarities among people create conditions for socialcontagion (individuals may be more likely to modifytheir behaviors/attitudes to match those around them)[67, 83]

Tie hierarchy (1) Connections between actors dissimilar in their status (e.g.,according to profession, leadership or power position) [7]

Hierarchy may be a barrier to innovation adoption (e.g.,lack of interest from above/resistance from below [29]

Tie reciprocity (8) The extent to which directional ties to actors are reciprocated(i.e., are bi-directional) [79]

Reciprocity may reflect greater stability or equality(versus hierarchy) [79]

Euclidian distance (1) A measure of the dissimilarity between the tie patterns ofeach pair of actors in the network [79]

Can be used to identify key people by their dissimilarityto others (e.g., who has the most research productivity


the exchange or application of evidence to informpractice. Similarly, the emphasis on outcomes relatedto work satisfaction, leadership roles, professional be-haviors, protocol efficiency, patient flow, operatingroom layout, technology adoption, and workplace per-formance reduce the extent to which KT-specific out-comes can be explored. The review by Chambers et al.

[6] described primarily the settings and outcomes ofthese studies, whereas the current study aims to de-scribe in detail the nature of the application of SNA tothe study of KT. Such an approach aims to advancethe science of KT by providing insight into worthwhilemethodological directions this literature can provide.Evidence for the effectiveness of specific KT

Table 1 Social network analysis (SNA) terms and their implications for knowledge translation (Continued)

SNA term (frequencycount)

Definition Implication for KT

relative to their connected peers) [28] (as a proxy ofinfluence)

Density

Whole network density(8)

An index of the proportion of existing ties relative to allpossible ties in a network [79]

Proxy for efficiency of information flow [79], solidarity[84], or cohesiveness within a network [21]

Ego network density (2)

Subgroups

Components/isolates (3) Portions of the network that contain actors connected to oneanother, but disconnected from actors of other subgroups[79]

Subgroups and isolates can be targeted to increaseconnectedness, share information, or mobilize action

Cliques (1) Maximum # of actors who share all possible connectionsamong themselves [79]

Can describe paths for fostering awareness andadoption of interventions [23]

Clusters (4) Dense sets of connections in a network [79] Identifying attributes that influence clustering helpsunderstand KT-related behaviors, such as informationseeking (e.g., experts; same department) [36, 41]

Network roles and positions

Brokers (1) Actors holding bridging positions in a network (i.e., play a rolein connecting subgroups) [79]

Can leverage brokers’ positions for efficient KT byleveraging their tie paths/connectedness [36, 37, 79]

Coreness/Core-peripheryindex (2)

The core of a network represents the maximally dense area ofconnections, whereas the periphery represents (to themaximum extent possible), the set of nodes withoutconnections within their group [79]

Power/influence at the core [39]. The most active EIPpractitioners may be found at periphery [32]

Structural equivalence(2)

When two actors/nodes have the same relationships to allother nodes in the network—they can be substituted withoutaltering the network [79]

These positions may generate social pressure within anetwork [24, 25]

Structural holes/constraint (ego network)(2)

Structural holes: absent ties in a network that limit exchangebetween actors; constraint: degree to which an actor is tiedto others who are themselves connected [79]

Inequality among actors can be identified and targetedthrough KT interventions; may have implications for EIPadoption [31] (e.g., many ties may restrict one’s actions/capacity) [79]

Transitivity/network closure (i.e., network structure related to triads)

Alternating k-stars (4) The tendency of actors to create ties [29] Used as an indicator of hubs within a network [37] orthe tendency to share/exchange knowledge [29]

Alternating k-triangles/transitive triads and/ornon-closurestructures (5)

The extent to which sets of 3 actors form patterns ofconnections that create larger “clumps” within the network[29, 79]

Assesses tendency to build relationships outside ofone’s local group—access to new knowledge [29]

Cyclic closure (1) The tendency for transitive triads (sets of three actors inwhich two ties exist) to lead to reciprocal ties within that triad[27]

Cyclic closure thought to reflect non- hierarchicalknowledge exchange, which is more effortful tomaintain and therefore less likely to be seen inknowledge sharing networks [27]

Alternating independenttwo-paths (2)

Assesses the conditions required for transitivity (i.e., ties thatform between each pair of actors in a set of three actors) [29]

Can determine the extent to which actors tend to buildsmall, closed, non-hierarchical connections that limitbroader access to new information [29]

SNA social network analysis, KT knowledge translation


interventions or for the identified relationships be-tween network properties and other variables relevantto KT can be sought elsewhere. Furthermore, none ofthese reviews examined the use of theory in their in-cluded body of literature specifically, despite this focusbeing an identified gap [6]. Given the rapid growth ofthe use of SNA in the health care context over thepast 8 years (see Fig. 2), an updated and more directedsearch is warranted. A targeted examination of theresearch specific to SNA in KT and EIP is required toinform the application of SNA methodologies in thisfield, with attention paid to the insights offered byboth the structural properties examined and the theor-etical perspectives applied.A SNA perspective can broaden our understanding of

the mechanisms by which KT efforts are effective byexamining the social structures and relationships that fa-cilitate or hinder KT and EIP. This understanding willaugment our knowledge base by expanding the range ofKT determinants worthy of consideration. As re-searchers gain interest in the social drivers of KT andEIP, this review will provide a foundation for developingkey research questions and SNA-driven methodologicalapproaches for KT research that are based on estab-lished and relevant theories. Furthermore, this reviewwill add to the current theorizing in the field related to asystems-focused understanding of KT and implementa-tion processes. Specifically, implementation happenswithin a complex system, and network approaches havebeen used to study complex systems; the link betweenthe two areas demands greater attention. Given thesegaps, the purpose of this article is to synthesize the waysin which SNA methodology can be used to advance thescience of KT.

MethodsThe purposes of a scoping review are to examine theextent, range, and nature of research in a given field; todetermine the utility of conducting a subsequent sys-tematic review; to summarize a body of research; and/or to identify research gaps, making it an ideal ap-proach to map the KT literature utilizing SNA [16].The specific objectives of this scoping review were (1)to describe the literature on SNA as it has been appliedto KT and EIP involving health care professionals, interms of its research design, methodology, and keyfindings; (2) to provide a critical analysis of the resultsin the context of existing theory; and (3) to identifystrengths and gaps to inform future research. The scop-ing methodology, as described by Levac, et al.’s [17]modification of Arksey and O’Malley’s [16] guidelines,was applied. Step six in the methodology, consultationwith key stakeholders, is optional and was not appliedin the current review.

Step 1: identify the research questionThe specific research questions developed for this reviewwere:

1. How has SNA been applied to health professionalnetworks in the field of KT/EIP with respect tostudy aims, data collection and analysis methods,and populations, context, variables, and structuralproperties under study?

2. What are the primary theoretical underpinningsthat explain the link between the networkproperties and KT/EIP?

3. What are the gaps in the literature that can informfuture research directions?

Step 2: identify relevant literatureThe search strategy involved a systematic search ofpeer-reviewed English SNA literature in November2015, repeated in July 2018, within five primary litera-ture databases: MEDLINE, Cumulative Index to Nurs-ing and Allied Health Literature (CINAHL), Embase,Web of Science (Science, Social Sciences Citation, andArts and Humanities Citation Indexes), and Socio-logical Abstracts. Wherever possible, keywords weremapped to subject headings, which were focused tonarrow the search (e.g., SNA terms) and exploded tobroaden the search’s scope (e.g., health careprofessional-related terms) to best capture relevant arti-cles. Keywords encompassed concepts related to socialnetworks, KT, implementation and EIP, as well ashealth care professionals. Because of the lack of consist-ent indexing terms addressing the concept of KT [18],keywords and subject headings were drawn fromempirically evaluated search strategies on this topic tofoster relevant results [18–20]. The Appendix providesthe detailed search strategy for MEDLINE; other data-base strategies are available on request.

Step 3: select the literatureRetrieved articles were screened for inclusion by two au-thors (SG and EJ). Inclusion criteria includedpeer-reviewed articles describing outcomes of researchstudies employing quantitative SNA methodology toexamine networks involving health care professionals inthe context of KT, implementation, or EIP (broadly de-fined as the exchange and/or application of informationto facilitate best practices in health care). The healthprofessional context was selected to narrow the scope ofthe review while maintaining high relevance to KT, ashealth professionals are common knowledge users orsubjects of implementation efforts. Outcomes of interestincluded, but were not limited to competencies (i.e., atti-tudes, knowledge, or skills) and behaviors by health pro-fessionals related to their sharing or use of evidence to


inform clinical decision-making. Dyadic (i.e., pair-level),ego-network (i.e., individuals’ networks), and whole net-work (e.g., departmental or organizational-level) proper-ties and variables were of interest. Exclusion criteriaincluded non-English articles for feasibility, and articlesthat did not quantify SNA data or network properties tofocus on articles that described, predicted, or explainednetwork-related phenomena in the context of KT inquantitative terms specific to SNA (e.g., empirical stud-ies whose analysis employed network data and analysismethods, as opposed to discussion papers). To target thescope toward evidence use by health professionals (i.e.,to maintain relevance to KT involving health profes-sionals within health care organizations), articles wereexcluded if they focused on online or social media-basednetworks (e.g., virtual communities of practice),policy-level KT, use of research by patients, focused oncommunication not explicitly involving research evi-dence or clinical decision-making about care based onevidence, or focused on the implementation of

non-clinical interventions (e.g., electronic medical re-cords, non-research-related quality improvementinitiatives).

Step 4: chart the dataSG extracted the data using a structured table developed apriori in accordance with the research questions. Informa-tion from each study was captured with respect to studyaim, population, sample size, variables, KT process andstructural properties examined, theoretical perspective anddata collection, and analysis methods employed. A secondreviewer (AK) screened the extracted data for accuracy.

Step 5: collate, summarize, and report resultsDescriptive statistics, including frequency counts andpercentages, were calculated to provide an overview ofthe literature’s breadth. Articles were charted by year ofpublication and country of first author to illustrate thechronological and geographical development of the field.Each network property identified in the reviewed articles

Fig. 1 Flow diagram of the article screening process


is presented with respect to the relational parameter itrepresents in the context of studying KT. Key findingsrelated to network properties were compiled and sum-marized narratively. An analysis of the use of SNA andthe theory that informed this body of research was thenperformed.

ResultsPublication characteristicsA total of 3531 articles were retrieved, of which 27met inclusion criteria. Figure 1 shows the PRISMAflow diagram of included and excluded studies.Figure 2 depicts the frequency of publications byyear. The USA and Italy (eight each) led in fre-quency, with Canada (4), Australia (2), theNetherlands (2), the UK (2), and Sweden (1) follow-ing. Study characteristics are presented in Table 2.Tallies and proportions of these studies presented inthe following paragraphs do not sum to 100% incases where the categories of characteristics are notmutually exclusive (e.g., a study may employvisualization of network properties while also pre-senting descriptive network property values).

Study design and data collectionThe 21 studies’ data sets were described in 27 articles,all but 2 of which employed cross-sectional designsyielding data at single time-points. Seventeen (81%) ofthe studies collected SNA data through surveys, 2 (10%)using a survey alongside interviews to support SNA datainterpretation [21, 22], and 1 (5%) through telephone

interviews [23]. Two (10%) studies employed documentreview (i.e., prescription, referral, or intervention re-cords); one for SNA data collection and one to supportoutcome measurement [24–26].

Networks and actorsPhysician-only networks were the most commonlystudied (11, 52%) [23–25, 27–35], followed by interpro-fessional networks of researchers and clinicians (6,29%) [36–41]. Only one study (5%) examined the net-work of nurses and physicians [22], and one a networkof nurse practitioners [42]. Two of the interprofessionalnetworks included public health officials [36–38], oneincluded leadership (i.e., directors, managers) and ad-ministrative support personnel [36, 37], and one in-cluded leadership and knowledge brokers [39]. In somecases, interprofessional network members’ professionsor formal roles within the network were not clearly in-dicated [21, 38, 39]. For two studies, the analysis of asubset of network members (i.e., managers/professionalconsultants; physicians with specific clinical workloads)was carried out [35, 40].Studies examined networks ranging in size from 13 to 784

participants, with a mean of 153. Just over half the studies(12, 57%) were conducted across organizational boundaries[23–28, 30–35, 43]. Eight (38%) were conducted within asingle health care organization [21, 22, 29, 36, 37, 39, 42, 44,45], one (5%) was conducted within a research-focused net-work [40, 41], and one (5%) within a health-specific field at anational level [38].

Fig. 2 Publications by year


Table 2 Characteristics of included studies

Citation Study purpose Type of network/setting Network size (#participants)

Datacollectionmethods

Theoreticalperspective

Zappa 2011 [29] To describe relationships forknowledge sharing about anew drug

Physician network within agroup of 338 hospitals

784 physicians Survey Diffusion ofinnovation

Yousefi-Nooraie2014 [37] (samestudy as 29)

To assess factors associatedwith information seeking inpublic health

Information seeking, expertiserecognition, and friendshipnetworks within an urbanpublic health department

15 managers and 13professional consultants (n= 28)

Survey Transactivememory theory;social exchangetheory

Yousefi-Nooraie2012 [36] (samestudy as 30)

To identify the structure ofintra-organizational knowledgeflow for evidence informedpractice

Information sharing network 170 directors, managers,supervisors, consultants,epidemiologists,practitioners, andadministrative support

Social influencetheory

Tasselli 2015 [22] To describe knowledgetransfer between professions,effectiveness of central actorsand brokers, and the influenceof organizational hierarchy onaccess to knowledge

Knowledge transfer network ina hospital department

n = 11853 physicians and 65nurses

Survey andinterviews

Sociology ofprofessions theory;SNA paradigm

Sibbald 2013 [21] To explore patterns ofinformation exchange amongcolleagues in inter-professionalteams

Information seeking andsharing networks within sixinterdisciplinary primary healthcare teams

n = 28 (nurses, physicians,residents, allied healthprofessionals, e.g., nurses,dietician, social worker);two sites: n1 = 19 and n2 =8.

Survey andsemi-structuredinterviews

SNA paradigm

Racko 2018 [47] To examine the influence ofsocial position on knowledgeexchange over time

Knowledge exchangenetworks within threeacademic-clinical KT programs

three surveys: n1 = 66; n2 =70; n3 = 42 clinicians andacademics

Surveys Social capitaltheory

Quinlan 2013 [42] To explore mechanisms ofinformation sharing acrossprofessional boundaries

Knowledge contribution todecision-making by memberswithin multidisciplinary primaryhealthcare teams (two clinicaldecisions, so two networks foreach of four clinical teams)

Nurse practitioners (n = 13or fewer)

Onlinesurvey

Habermas’ theoryof communicativepower

Paul 2015 [35](portion of datafrom study [45])

To test a model examining therole of triadic dependence onreciprocity and homophily

Influence network 33 physicians Surveys SNA paradigm

Patient care network 135 physicians

Menchik 2017 [44] To explore the type ofknowledge valued byphysicians and the influence ofhospital prestige on evidence-seeking behavior and per-ceived esteem by peers

Information seeking andclinical case discussionnetworks, within six hospitals

126 physicians Survey Social influencetheory

Mascia 2018 [27] To explore theoreticalmechanisms explainingnetwork formation acrossclinical sectors

Advice-seeking networkswithin two regional healthauthorities

97 pediatricians Survey Balance theory;structural holesperspective;homophilyprinciple

Mascia 2014 [33] To explore the associationbetween connectedness withcolleagues and frequency ofevidence use within aphysician network

Collaboration networks within5 health authorities

104 pediatricians Survey Diffusion ofinnovation; socialinfluence theory;social contagion;strength of weakties [82]

Mascia 2015 [48](same data asstudy [30, 32])

To explore the influence ofhomophily on tie formation

Mascia 2011a [30](same data asstudy [32, 48]

To determine the associationbetween attitudes toward EIPand network structure and toidentify predictors of

EIP advice-sharing networkswithin 6 hospitals

297 physicians Survey Homophilyprinciple


Table 2 Characteristics of included studies (Continued)




collaborative ties

Mascia 2013 [32](same data asstudy [30, 48])

To explore the relationshipbetween attitudes toward EIPand network position

Core/peripherymodel; structuralholes

Mascia 2011b [31](same study as[30, 32, 48])

To explore the associationbetween network structureand propensity to adopt EIP

207 physicians Social contagion;structural holesperspective

Long 2014 [41](same data asstudy [40])

To examine the influence ofclustering on past, present,and future collaborationswithin a translational researchnetwork

Past, present, and intendedcollaboration networks withina research network

68 researchers andclinicians

Onlinesurvey

SNA paradigm

Long 2013 [40](same data asstudy [41])

To identify key players within aresearch network, theircommon attributes, and theirperceived influence, power,and connectedness

Research collaboration anddissemination networks withina research network

Keating 2007 [45] To describe the network ofinfluential discussions amongphysicians and to predictnetwork position

Frequency of influentialconversations relevant topractice within primary care

38 physicians Survey SNA paradigm

Heijmans2017 [26]

To explore relationshipsbetween network propertiesand quality of care

Information exchangenetworks within 31 generalpractices

180 health professionals(physicians, residents,nurses, pharmacyassistants, social workers)

Surveydocumentreview (i.e.,interventionand referralcharting)

SNA paradigm

Guldbrandsson2012 [38]

To identify potential nationalopinion leaders in child healthpromotion

Discussion network withinnational child healthpromotion context

153 researchers, publichealth officials,pediatricians and otherindividuals

Emailedsurvey item

Diffusion ofinnovation

Friedkin 2010 [25] To examine the associationbetween discussion networks,marketing, and physicianprescribing practices

Advice and discussionnetworks of physicians (re-analysis of Coleman, Katz andMenzel, 1966 historical data onmedication adoption)

125 physicians Documentreview (i.e.,prescriptionrecords ofpharmacies)

Diffusion ofinnovation; socialcontagion;cohesion;structuralequivalenceBurt 1987 [24] To test social contagion theory

by examining cohesion versusstructural equivalence asdrivers of tie formation

Doumit 2014 [34] To identify opinion leaders andtheir impact on EIP

Advice networks of craniofacialsurgeons within 14 countries

59 craniofacial surgeons Onlinesurvey

Diffusion ofinnovation

Di Vincenzo2017 [28]

To explain the impact ofresearch productivity on tieredundancy (i.e., connectionsthat lead to the same people/information)

Advice seeking networkswithin and external to a healthauthority containing 6hospitals

228 physicians Survey Structural holesperspective;homophilyperspective

Bunger 2016 [43] To evaluate change in adviceego-network composition andits impact on whole networkstructure following implemen-tation of a “learning collabora-tive” model in improve carequality

Advice networks of clinicians(psychologists, social workers,others) and leadership in 32behavioral health agencies

132 clinicians, supervisorsand senior leaders

Surveys SNA paradigm

Ankem 2003 [23] To understand communicationflow and its influence onawareness/adoption of atreatment, and to identifyopinion leaders with influence

Frequent discussion networkswithin a sample drawn froman online physician directory

32 interventionalradiologists

Phoneinterviews

Diffusion ofinnovation; SNAparadigm


Study purposes and data analysis methodsA SNA perspective has been used to explore the patternsand efficiencies of information sharing within and acrossprofessions, to identify positions of influence and determinetheir effectiveness, to predict or to explain patterns of tiesbased on attributes or network structure, to compare thestructural characteristics of different models of KT, and toexamine the relationship between network properties andEIP attitudes and behaviors. Using a longitudinal approach,SNA has been used to evaluate the influence of social struc-ture on knowledge exchange over time, as well as to evaluatenetwork change following a quality improvementintervention.The majority of articles focused on information flow,

while individuals’ adoption of a clinical practice, involve-ment in collaborations that support KT or EIP, and evi-dence-informed group decision-making garnered lessinterest. No research was available on other KT activ-ities, such as the processes involved in developing guide-lines or other KT tools, adapting knowledge to the localcontext, assessing barriers to change, facilitating depart-mental- or organizational-level practice or service deliv-ery changes (including de-implementation), monitoringevidence use, evaluating KT effectiveness, or sustainingchange over time [46].

Describing networksEight (29%) of the 27 articles described networks by derivingnetwork properties from relational data [21, 22, 30, 36, 39,40, 42]. Two articles (10%) used conventional descriptive sta-tistics (e.g., frequency counts, proportions) to describe socialnetwork data [34, 38]. Network visualizations illustrated thedata in 13 (48%) articles [22, 30, 32, 34–37, 40, 42–45]. Ofthese, ten articles presented whole network graphs (i.e., illus-trations of the network structure), two presented networkgraphs of subgroups within the network, and one mappedwhole networks within graphs that accounted for covariates.Network properties represented in the graphs included cen-trality (i.e., connectedness, brokers (i.e., bridgers)), core ver-sus periphery structure (i.e., central areas of highconnectivity versus peripheral areas of lower connectivity),and tie strength (e.g., frequency of contact); and attributes,

such as gender, professional role, size of clinical practice, anddivision, department or team, and organization or site. Visu-alizations were used to depict network property configura-tions (i.e., illustrated the definitions of network properties) intwo articles. Conventional charts (e.g., boxplots, scatterplotsand bar, line, and area charts) were also used in six articles tovisualize relationships between variables (e.g., network prop-erties with one another, diffusion or adoption over time, cen-trality versus adoption timing, or receipt of usefulinformation, percentage of ties by strength at different timepoints).

Testing hypotheses (network modeling)Fourteen articles described the use of regression, includingordinary least squares [24, 28, 32, 44, 47], ordinal logistic re-gression [33], multi-level modeling [25, 26, 37], P2 logistic re-gression [35, 45], linear regression [22], and MR-QAPanalysis [30, 48]. Regression was used to test theory aboutthe cause of social structure [24, 25, 28, 35, 48], to explainthe impact of social structure on knowledge exchange behav-iors [47], ease of knowledge transfer [22], receipt of usefulknowledge [22], or quality of care [26], and to describe theinfluence of hospital prestige on evidence-seeking behaviorand perceived esteem by peers [44]. Three articles reportedpaired t tests or Wilcoxon ranks to evaluate differences be-tween groups [22, 26] or time points [43], the latter of whichalso employed analysis of variance. Two articles describedthe use of the Chi-square test to examine associations be-tween attributes and network position, or between two attri-butes [23, 40]. Exponential random graph models were usedin three instances to predict or to explain the formation ofties based on attributes and network structure [27, 29, 37],and a single study employed factor analysis to constructgroupings of individuals based on frequency of informationexchange [23]. No studies employed stochastic actor-basednetwork modeling to examine network change over time.Sample hypotheses relating to tie formation included

predictors, such as homophily, existing ties (leading toreciprocity), and having a formal mechanism within theorganization for interacting. Further hypothesis examplesincluded that higher professional status would be associ-ated with more knowledge exchange, tie homophily (i.e.,

Table 2 Characteristics of included studies (Continued)




D’Andreta2013 [39]

To compare the networkstructures of three research/KTprogram initiatives

Informal advice giving andseeking networks within eachof three academic-clinical KTprograms

n = ~ 260 (directors,managers, programleaders, knowledgebrokers, researchers, andothers unspecified)

Onlinesurvey

SNA paradigm;Epistemicdifferencesperspective

SNA social network analysis, EIP evidence informed practice, KT knowledge translation. Where indicated by the articles’ authors, the dependent variable isdesignated using bold text


Table 3 Variables, network properties and key findings

Citation Primary dataanalysismethod

Variables of interest Findings

Attributes Structural orrelational parameters

Network propertyused as proxy forstructuralparameter

Descriptive/exploratory studies

Yousefi-Nooraie 2012[36] (samestudy as [37])

Derivingnetworkproperties todescribe thenetwork

_ Connectedness Whole networkdensity

Low density (1.2%) observed

Informationexchange

Tie reciprocity Head management division identified as centralcluster bridging organizational divisions, withhierarchical information flow.Expertise recognition and information seekingclustered within divisions; friendships spanneddepartments;Friendship and expertise recognition predictedinformation seeking ties;Network-identified brokers should receive sameinterventions and supports as formal brokers

Prestige (key actors) Indegree centrality

Mediating power ofactors

Betweennesscentrality

Subgroups ofconnected actors

Clusters

Brokers (actorsconnecting distinctteams/clusters ofalters)

Brokerage patterns(measured bywhich groups theinformation sourceamd its recipientsbelonged)

Sibbald 2013[21]


– Cohesiveness relatedto giving andseeking research-related information

Whole networkdensity

Low density for information seeking and giving(7–12%) observed; suggested these behaviors nota central focus of the interprofessional relationships

Profession Key players (prestige)in giving and seekingresearch information

Indegree centrality Medical residents prominent in knowledgeexchange; physician seen as primary implementerof evidence; nurses as intermediaries betweenphysicians and support staff; allied health morelikely to draw information from external networks

Quinlan 2013[42]


Profession tenureof the teamOccupationaldistance amongmembersNumber of teammembers

Communicativepower (i.e., thefacilitation of mutualunderstandingamong other teammembers)

Flow betweennesscentrality

True interprofessional decision-making attributedto low structural hierarchy. Nurse practitioners(in newly formed teams) and registered nurses(in established teams) tended to have greatestcommunicative power. Mutual understandingand professions’ involvement varied across clinicaldecision-making episodes

Change in flowbetweennesscentrality betweenclinical decisions

Long 2014[41] (samestudy as [40])

DescriptiveSNA;correlationsamongspecificnetworkpropertiesand/orattributes

Geographicproximityprofession (e.g.,clinicians/researchers)

Grouping based onsimilarity in attributes

# components Geographical proximity, professional homophilyassociated with clustering (past collaborations);geographical proximity and past collaborative tiesinfluenced current and future collaborations.Intended future collaborations were moreinterprofessional.Network density varied across networks (current4%, future 27%, past 31%).Weak ties and reputation associated with intentionfor future collaboration; strong ties associated withcurrent collaborations.

External-internal(E-I) indices basedon tie homophily

Clusteringcoefficients(comparing ego-and whole net-work density)

– Past strongcollaborations;Current or futurecollaborations

Past collaborationnetwork tiestrength; Currentor futurecollaboration ties

Previousexperience in thefield

Current or futurecollaborations

Current or futurecollaboration ties

Actor’sreputation

Indirect contacts Futurecollaborationnetwork tiestrength

Future intendedcollaborations

Futurecollaboration ties


Table 3 Variables, network properties and key findings (Continued)





Long 2013[40] (samestudy as [41])

Chi squareanalyses totest forassociationbetweenattributesand networkposition (i.e.,key actorstatus)

CurrentworkplaceGenderMembership inother networksQualificationsApproach towork within thenetwork

Key actors (withrespect to power,influence orconnectedness)

Indegree centralityBetweennesscentrality

A manager, and specific researchers andclinicians identified as key players.Apart from expert status and valuingadequate network resources, network-identifiedcentral players and formal brokers had little incommon.Planned interventions and support for formalbroker roles may be misdirected if not alsooffered to network-perceived central players orbrokers.Network members may be better able tocorrectly identify central actors than formalbrokers

Guldbrandsson2012 [38]

Traditionaldescriptivestatistics, e.g.,frequencycounts,percentages

ProfessionWorkorganizationGeographicalregion

Information seekingabout child healthpromotion

Tie homophily (%)Indegree centrality(frequency ofbeing named)

Organization and professional field wereshared in nearly half of all informationseeking ties

Doumit 2014[34]

Percentageof peoplenominatingan actor;descriptivestatistics(frequencycounts,percentages)

– Influence by centralactors

Degreecentralization

Six individuals with high credibility influenced85% of the network, suggesting opinion leadershave potential for supporting evidence use

Reasons forchange inmedicalapproach(proportions)

– –

Barriers to clinicaldecision-making(proportions)

– –

D’Andreta2013 [39]

Derivingnetworkproperties todescribe thenetwork(descriptiveSNA)

KT modeladopted

Prestige within thenetwork

Degreecentralization

KT teams with different models of KT (i.e., focuson research dissemination vs. knowledgeco-production and brokering vs. integratedresearch-clinical collaboration)varied in their structural properties (e.g. theprominence and control of leaders in KTprocesses)

Control overknowledge

Betweennesscentralization

Access to knowledge Closenesscentralization

Alternate paths forknowledge flow thatcircumvent centralactors

Flow betweennesscentralization

Organizationalrole (e.g.,director, supportstaff)

Coreactors—dominantindividuals withfrequent knowledgeexchange

Coreness scores(core-peripheryalgorithm)

Predictive/explanatory studies

Zappa 2011[29]

DescriptiveSNA

Externalcommunication(# visits fromdrugrepresentatives);researchorientation (#publications);clinicalexperience;hierarchicalposition(administrative

Colleagues withwhom knowledge isdiscussed andtransferred

Network density Low network density (0.3%)

Components Multiple small components suggested lack ofstrong opinion leaders to drive treatmentadoption. Findings suggest physicians tend tobuild small, closed, non-hierarchical internal, andexternal connections within their professionalgroup, potentially limiting broader access to newinformationIsolates tended to be clinically experienced andactive in research.Advice sharing more likely if physicians shared amedical speciality, geographic proximity but

ExponentialRandomGraphmodels (p*models)

Tendency to exchangeinformation with anumber of sources

Alternating k-stars

Tendency to shareknowledge within asmall peer group(network closure)

Alternating k-triangles; alternat-ing independenttwo-paths







role)Medicalspecialty*Hospitalaffiliation*

differed in research productivity or years inpractice.

Tendency to interactwith similar othersProminence as aknowledge source inthe network

Tie homophily/hierarchy;indegree centrality

Yousefi-Nooraie, 2014[37] (samestudy as [36])

DescriptiveSNA

Relativeconnectedness ofactors of a given role

Indegree centralityOutdegreecentrality

Managers identified as key brokers in KTinterventions and EIP implementation processes.Public health professionals preferred to limit adviceseeking and expert recognition to a small numberof peers; advice seeking limited to own division.Strong friendship ties a significant predictor ofinformation seeking ties.EIP scores not predictive of information seeking orexpert recognition ties.

*Role (e.g.,manager)*OrganizationaldivisionScore on EIPimplementationscale

Key individuals Degree centrality

Organizationaldivision

Tendency to connectto peers from otherunits

E-I index

– Tendency toreciprocate expertrecognition andinformation seekingties

Tie reciprocity

Exponentialrandomgraphmodeling(ERGM)


Tendency to connectwith those withsimilar attributes

Tie homophily

Reciprocity Tie reciprocity

Formation ofinformation seekingand expertise-recognition ties

Ties and directionof ties (in vs. out)

Tendency fornetwork to havehighly connectednodes (hubs)

Alternating in-k-starsAlternating out-k-stars

Friendshipconnections

Ties

Multilevellogisticregression


Tendency to connectwith those withsimilar attributes

Tie homophily

Formation ofinformation seekingand expertise-recognition ties

Ties

Friendshipconnections

Ties

Tasselli, 2015[22]

Paired t testLinearregression

*Gender*Tenure*Profession*Organizationalunit*Rank (i.e.,leadership role)

Ease of knowledgetransferPerceived receipt ofuseful knowledgeConnectednessHierarchyNetworkfragmentationIndividual reachBrokerage potential*Network size

Tie strengthMean degreecentralityBonacich centralityMeanbetweennesscentralityClosenesscentralityBetweennesscentrality

Knowledge tends to transfer within rather thanacross professions; nurses’ networks were denserand more hierarchical; closeness centralitypositively associated with ease of knowledgetransfer; brokering positions increased access touseful knowledge







*Degree centrality

Menchik 2010[44]

OLSregression

# medicalliteraturedatabasesearches permonth# journals readregularly*Age*Gender*Tenure athospital*Medical school*% clinical time*Sub-specializationPrestige ofhospital(publishedrankings)

Relational esteem bycolleagues

Indegree centrality Physicians in higher prestige hospitals were lesslikely to be named as advice givers. Prestige inthese settings associated with medical schoolattended.In lower-prestige hospitals, regularly reading arange of journals, and less time spent on clinicalwork increased likelihood of high esteem bycolleagues

Mascia 2014[33]

Ordinallogisticregression

Self-reportedfrequency of EIPuse*Gender*# patients incaseload*area of clinicalpractice (e.g.,asthma, urology,etc.)*# articlesubscriptions*perceptions ofbarriers toavailability ofevidence*perceptions ofdifficultyapplyingevidence topractice*Organizationalaffiliation*Affiliation toformal groups*Collaborativenature of actor’smedical practice

Degree ofcollaboration withcolleagues

Outdegreecentrality

Degree centrality directly associated withphysicians’ EIP use

Mascia 2018[27]

Exponentialrandomgraphmodels

*Past task forceinvolvement*Tenure*Gender*Geographicdistance*Associationmembers*Health district

Tendency toreciprocate adviceTendency to seekadvice from anindirect tieTendency for local,generalizedexchange of advice*Tendency to formadvice ties*Popularity as anadvice source*Advice-seekingactivity

Tie reciprocityTransitivity (pathclosure)Cyclic closure*Density*Indegreecentrality*Outdegreecentrality*Formation ofnon-closurestructures*Tie homophily (ofattributes)

Advice ties unlikely unless reciprocated; advice tiestended to organize around clusters—driven bytransitivity, not popularity; Tendency againstexchange of advice in cyclic structures; positiverelationship between ties and associationhomophily in one health authority, and betweenties and district/task forces in the other; tendencytoward homophily related to tenure and distance,but not gender







*Brokering

Mascia 2015[48] (samedata as [31],[30] and [32])

Multipleregression-quadratic as-signmentprocedure(MR-QAP)

AgeGenderTenure/seniority

Frequency ofcollaborationSimilarity ofprofessional role,institution andgeographical location

Tie strengthTie homophily

Ties more likely if specialization, institution werethe same between individuals; less likely if similarroles, greater difference in time since graduationand further geographic distance; professionalhomophily better predictor than institutionalhomophily

Mascia 2013[32] (samedata as [48],[30] and [31])

DescriptiveSNA

Age*Gender*Hospital tenure*Tenure in healthauthority*Managerial role*Geographicaldistance fromcolleagues*Affiliation withotherorganizations*Self-reported EIPadoption (i.e.frequency ofdatabasesearching)

Connectedness Whole networkdensity

Low density (5.7%) observed

OLSregression

Network authority (i.e.,importance—relevantand popular)

Hubs andauthoritiescentrality

The most active EIP practitioners likely to be foundat network periphery (i.e. least central)

Degree of coreness inthe network

Network corenessscore (degreecentrality andcore-peripheryposition)

Mascia 2011b[31] (samedata as [48],[32] and [30])

Ordinallogisticregression

Tendency toadopt EIP (self-reportedfrequency of peer-reviewed researchuse)Age*Gender*Tenure in healthauthority/organization*Managerial role*# publications*Perceived accessto evidence*Hospitalaffiliation*

Extent to which agiven tie isredundant becauseof concurrent tieswith another alter

Ego-networkconstraint

Physicians with greater network constraint (i.e.,many redundant ties) reported decreased EIPadoption. May be related to informationbias—tendency of physicians to interpret theinformation in a way that is congruent with theirprevious knowledge or opinion.High degree centrality associated with EIP use.Individual’s network

size*Total # of ties inego-network(indegree + outde-gree centrality)*

Mascia 2011a[30] (samedata as [32,48])

DescriptiveSNA

– Average number ofadvice exchangecolleagues

Mean ego-networkdensity

Advice sharing most likely when physicians shareda medical specialty, geographic proximity, similarattitudes toward EIP, or had co-authoredpublications.Collaboration less likely when actors held similarmanagerial roles, or were at different hospitals/clinical/geographical areas.

Tendency forcolleagues to bothgive and receiveadvice with oneanother

Tie reciprocity

Multipleregressionquadraticassignmentprocedures(MR-QAPanalysis)

– Advice exchangeamong pairs ofphysicians

Ego-network ties

Similarity betweenpairs of tied actors in:GeographicaldistancesGender*Age*Medicalspecialization*

Tie homophily







Clinical experience*Tenure in healthauthority/organization*Managerial role*# publications*Co-authorship*EIP adoption (self-reported frequencyof peer-reviewed re-search use)*

Paul 2015 [35] Extended p2model withBayesianmodelingandestimation

*AgePatient agePatient sexPatient racePatient healthstatusPatient intensityof care

Relative # sharedpatients*Same gender*Same specialty*Same locationReciprocitySocial dependence(clustering)

DensityTie homophilyTie reciprocityAlternatingindependent two-pathsTransitive triadsAlternating k-stars(two-stars)

Low network density (0.10) observed.Triadic clustering higher than chance.Ties not associated with gender/specialtyhomophily.Location positively associated with ties.Complementary expertise positively associatedwith patient sharing.Transitivity may account for reciprocity

GenderClinic% femalepatientsSelf-identify asexpert# clinics perweek# years practicingin the cityTenure athospitalYears clinicalexperienceLocation oftraining

Involvement ininfluential discussions

Whole networkdensity

Low density (0.154) observed; reciprocity morelikely than not—may be an artifact of transitivity;high triadic clustering observed; same clinic andgender, expert, higher clinical caseload increasedtendency for tie formation

Keating 2007[45]

P2 logisticregressionanalysis

Self-identified experts seen as more influential; norelationship between # years in practice or locationof work or training.Clustering with respect to EIP knowledgeexchange observed between those with greater #of patients and higher frequency of clinicalsessions.High reciprocity observed in the absence ofopinion leaders with high centrality.

Being perceived asinfluential

Indegree centrality

Perceiving others asinfluential(information seeking)

Outdegreecentrality

Reciprocity Tie reciprocity

Heijmans 2017[26]

Pairedsample ttests/WilcoxontestsLogisticmulti-levelanalyses

*Patient age*Patient sex*Patient group*Patient illnessstatus*Treatment/control group forparallelrandomizedcontrolled trial

ConnectednessFrequency of contactInfluence ofcoordinatorSimilarity in attitudesrelated to treatmentgoalsPresence of opinionleader*Network size

DensityTie strengthDegree centralityHomophily (E-Iindex)% of possible in-degree ties

Low density (0.37 and 0.38) observed.Most ties between those who did not valueachieving treatment goals.General practitioner most likely named as opinionleader.Nurse performance associated with consistentlyidentified opinion leader.Lack of tie homophily for positive attitudesassociated with poor clinical outcomes

Friedkin 2010[25]

Randominterceptmulti-levelevent historymodel

Professional ageChief or honoraryposition (yes/no)Number ofjournals readValue keeping upwith scientificdevelopmentsPhysicians’adoption of anew antibiotic(i.e., prescribingbehavior)Marketingpatterns of drugcompanies

Influence of advisors/discussion partners

Contact networkrole (CNET)—asummativemeasure of 4measures ofstructural cohesionand structuralequivalence;position in themedical advicenetwork

Cohesion and structural equivalence werecorrelated, and may be useful in combination toimprove reliability in the evaluation of networkstructures across settings







Proportion ofpreviousadopters at agiven time(“internalcontagion”)

Di Vincenzo2017 [28]

OrdinaryLeastSquaresregression

# publications*Tenure*Managerial role*Geographicdistance*Hospitalaffiliation*# publicationsfrom same-specialtycolleagues

Dependence onothers/access to newinformationRelative productivityamong ego-networkcolleaguesSame roleSame specialty

Ego-networkconstraintEuclidian distance*Ego-network size

Young employees appeared to have moreredundant networks (greater need for advice).Hospital affiliation (i.e., context) influencedconstraint.Constraint negatively associated with ego-networksize and relative productivity (mediated by profes-sional group membership), positively with product-ivity, Euclidean distance, role/specialty homophily(augments impact of productivity on prestige)

Burt 1987 [24] Ordinaryleast squaresregressionwithlikelihood-ratio chi-squared test

Timing ofadoptionRelative timing ofadoption withinthe networkProfessional ageContact withdrug companyNumber ofjournals readNumber ofhouse calls vs.office visitsValue keeping upwith scientificdevelopments

Position in themedical advice/discussion network

Structuralequivalence

Adoption by others in equivalent positions withinthe network was a stronger predictor of adoptionthan adoption by those in an individual’s advice ordiscussion networks.Early adopters tended to participate in a range ofEIP behaviors.Adoption by prominent physicians seen to berelated to their desire to avoid being late adopters.

Influence of advisors/discussion partners

Structuralcohesion

Ankem 2003[23]

Chi-squarestatistics

PreferredinformationsourceTiming ofawareness of theinterventionTiming ofinterventionadoption

– – Clinical networks were most prominent in fosteringawareness and adoption of a clinical intervention,but research and social networks also likely toinfluence these processes.Early adopters tended to rely on journals andconferences for information informing practicechange; late adopters to a greater extent bynetwork contacts.

Factoranalysis

SpecializationHospitalCityTiming ofadoptionFrequency ofcommunicationwith colleagues

Types of relationswithin the network(e.g., clinical, research,leisure)

Ties

Groupings ofinformation exchangerelations

Cliques

Longitudinal evaluative studies

Racko 2018[47]

Ordinaryleast squaresregression

Professionalstatus (ranking)*Professional role*Gender*Education*Organizationalstatus

ResearchcollaborationJoint decision-makingConnectedness tohigh-statusindividualsConnectedness toknowledge brokers

Ego-network sizevia tie heterophilyTie strengthMean status scoreof ego-networkrelative to wholenetwork% of possible ties*Tie homophily

Higher social status associated with more researchcollaboration at all time points, and joint decision-making in early phases.Higher-status ties with peers, ties to formalknowledge brokers and ties to unfamiliar peersinconsistently predicted knowledge exchange,research collaboration and joint decision-makingover time.Formal knowledge broker presence may facilitate


sharing the same profession with a connection) would beassociated with greater knowledge transfer ease, the pres-ence of brokers (bridgers) would be associated with anincrease in the receipt of useful information, particularlyto managers, and that greater connectivity, frequency ofcontact, homophily, the presence of a highly connectedclinical coordinator, and being an opinion leader would beassociated with an increase the use of best practices. Formore information about the full array of correlational,dependent and independent variables and covariates (bothrelational variables and attributes) identified in the in-cluded studies, refer to Table 3.Information seeking patterns varied across professions

and networks, with health professionals from some disci-plines having a tendency to form small, closed sub-groups, while others demonstrated greater connectivityand reach within the network, increased hierarchy (e.g.,reliance on “gatekeepers” of information spreading it ina top-down approach), or relied more on sources of in-formation or support external to the network (e.g., otherorganizations). Available information suggests that iso-lated individuals and those less connected at the periph-ery of the network may have more clinical experienceand be more evidence-based in their practices than thoseat the network core [29, 32]. In the absence of acore-periphery structure (i.e., a more highly connectedcenter with a less connected network periphery), degreecentrality (i.e., the number of connections an individualhas) may be a key factor associated with EIP use, at leastfor physicians [33]. A network-identified broker or opin-ion leader may increase access to useful knowledge [14],improve practice performance [26], and facilitatenetworking across professions [47].

Network propertiesTable 3 summarizes the key variables under study, thenetwork properties that were derived from relationaldata, and the relational parameters (i.e., the constructsfor which the network properties were acting as prox-ies). For example, network density was used as a proxyfor connectedness in one study, and for representingthe number of shared patients in another. Attributes ofinterest (which included individual characteristics, suchas profession and gender; environmental characteristics,such as organization; social attributes, such as per-ceived reputation; and KT-related measures, such asEIP attitude scores) are also presented to offer a sum-mary of the nature of non-SNA variables that have beenanalyzed alongside network properties. Study findingsare presented in the final column for interest.Eleven articles explored only a single or pair of net-

work properties. Although 28 network properties wereidentified during data extraction, the majority ofauthors examined centrality, tie characteristics (e.g., thedirections of the interactions; similarity in characteris-tics among pairs of connected individuals), and density(i.e., the proportion of ties relative to all possible ties)as their network properties of interest. Tie homophily(i.e., similarity of connected individuals on a given attri-bute, such as gender), indegree centrality (i.e., the num-ber of people naming an individual as being connectedto them), whole network density, the presence of ties,and tie reciprocity (i.e., bi-directionality in reportedinteractions or connections) were the most prevalentstructural properties studied. The study authors’ discus-sions about the influences of these network propertieswere clearly linked to prominent theoretical






Connectedness tounfamiliar peers*Intra-professionalwhole network size

interprofessional networking

Bunger 2016[43]

Paired t tests;one-wayanalysis ofvariance; de-scriptive SNA

Role (facultyexpert, internalcolleague,external peer,privatepractitioner,other)

ConnectednessLack of adviceseeking/sharingReciprocitySimilarity inconnectednessTendency for sub-groups to formSame institutionFrequency ofcommunication

DensityIsolatesTie reciprocityIndegreecentralizationClusteringTie homophilyTie strength

Ego-network size decreased, more markedly forsenior leaders.Exposure to private practitioners and “others”decreased; exposure to experts increased.Substantial turnover in dyads was reported, withgreater tie density around central core of experts.Reciprocity and tie heterophily increased over time

KT knowledge translation, EIP evidence informed practice. Where indicated by the article’s author, italic text = dependent variable; * = covariate


perspectives. Less emphasis was placed on the analysisof centralization (i.e., the evenness of the distributionof connections), subgroups (i.e., groups of connectedindividuals not connected to other groups within thenetwork), and transitivity (i.e., patterns related to setsof three individuals and their tendencies to share con-nections with one another).Attributes, such as research versus clinical product-

ivity, professional field or specialty, leadership roleand organizational prestige [29, 30, 35, 38, 44, 48], aswell as the presence of other types of ties (e.g.,friendship, expertise recognition, previous collabora-tions) [30, 36], appear to be predictive or explanatoryfactors for the formation of information seeking orresearch collaboration ties. Conflicting findings re-garding the influence of EIP attitudes, experience,gender, and geographical proximity on tie formationwere identified [27, 30, 35, 37, 45].

Use of theoryDiffusion of innovation was the theory most fre-quently applied (seven articles) [23–25, 29, 33, 34,38]; social contagion theory [24, 25, 33] and social in-fluence theory [33, 36, 44] were used in four andthree studies, respectively. A model combining trans-active memory theory and social exchange theory[37], as well as Habermas’ theory of communicativepower [42], social capital theory [47], sociology ofprofessions theory [22], balance theory [27], and anepistemic differences perspective [39], were applied inone instance each. Most commonly, theory was usedto select network properties to examine and to de-velop hypotheses to test. In addition, theory was usedto provide background information about SNA or thetopic under study, to assist in the interpretation offindings, and to develop and test new analyticalmethods to advance the field of SNA.Several other articles employed SNA-specific theor-

etical perspectives to exploratory analysis, includingapplying Granovetter’s strength of weak ties perspec-tive [33], examining the association between structuralholes (i.e., areas lacking connections) and the estab-lishment of brokers that bridge network gaps [27, 28]or between a lack of ties and EIP attitudes [32],examining the role of social pressure on tie formation[24, 25], exploring the influence of being within ahighly connected core of the network versus a lessconnected peripheral area on attitudes toward EIP[32], and evaluating network dynamics relative to thehomophily principle (i.e., the tendency of people toform connections with similar others) [27, 28]. Sevenarticles employed a SNA paradigm without referenceto a specific theory [21, 26, 35, 40, 41, 43, 45].

DiscussionPublication characteristicsWhile SNA has a long history in fields, such as soci-ology, mathematics, and psychology, its popularity inhealth care has shown momentum since the early2000s [6]. Annual publication trends suggest an emer-ging interest in SNA as an approach to study KTprocesses and determinants since 2010. This concen-tration follows work by Thomas Valente and colleaguesin the mid-late 2000s. Their work applied SNA inhealth care to identify and to evaluate brokers andopinion leaders as a means of promoting behaviorchange, linked communication networks and diffusionprinciples to health promotion research, and appliednetwork principles to the study of community-basedcancer research [49–51]. Increasing interest in the util-ity of SNA across a range of health care contexts, andthe relative maturity of this field, may contribute to itscontinued use in the newer field of KT to embed newpractices across settings.

Study design and data collectionMore longitudinal research would allow us to determinethe direction of the causal relationships betweennetwork structures and attribute variables, such as EIPattitudes and behaviors, for better prediction of imple-mentation outcomes. Such research would also enablethe evaluation of changes in network connection pat-terns over time. This approach can be used to assess theimpact of network interventions. Network interventionscan use socially based strategies to identify and targetnetwork gaps [52] (e.g., interactive forums to help estab-lish connections for isolated individuals or groups) toenhance KT processes, organizational capacity, or adher-ence to desired behavior (e.g., through social influence)[53]. Network interventions may also harness strengthsin a network [52, 53] (e.g., engaging highly connected in-dividuals to exert influence or to share resources) to bet-ter mobilize EIP attitudes, behaviors, or informationflow for sustained implementation.The simulation investigation for empirical network

analysis (SIENA) framework offers a means of evaluatingwhole network dynamics, particularly with networks de-fined by connections that persist over time, which couldbe particularly helpful in identifying barriers to the suc-cessful introduction of interventions in specific settings[54]. The SIENA framework can evaluate change in re-peated measures of a network with respect to actors andties, as well as the interplay of these network changesalongside changes in the behaviour of network members[54]. For example, a network intervention aimed at redu-cing isolation within the network, or at increasing theconnectedness of individuals who are positioned well toinfluence many others, can be evaluated for its


effectiveness using the SIENA framework. The evidenceof these changes would serve as supporting mechanismsfor new evidence-based interventions. Computationalmodels, such as agent-based modeling, can also be usedto represent individuals and their interactions. Multiplesimulation experiments that are programmed based onattribute data and structural characteristics allow re-searchers to specify and to control the parameters of thecomputational algorithms in order to determine the ef-fects of specific variables [55].Survey use was the dominant method of SNA data

collection employed, which is consistent with thebroader SNA field [7]. Although interviews were notprominent in the reviewed articles, they can gather thesame relational data as surveys, while allowing the par-ticipant and interviewer to clarify question and re-sponse meaning (e.g., defining operational terms inmore depth; describing the reasoning behind namingspecific people in one’s network) [56]. A qualitative datacollection approach may have implications for the sizeof the sample, but the opportunity for discussion canincrease response validity, while the relational networkdata can be quantified for analysis [56]. Qualitative in-terpretive strategies can also be used to understandcontext and meaning within the network and thephenomenon of interest [56], such as the extent towhich network structure and/or specific attributes orcontextual factors are perceived to influence attitudes,knowledge, or behavior related to the introduction of anew intervention. Mixed methods can be used to tri-angulate findings to validate results, and to strengthenthe explanatory power of the research by exploring thecomplexities involved to a greater depth [56].Limited use of document review was also observed, none

of which involved electronic data. The use of secondarydata (e.g., email records, social media interactions, medicalrecords) to quantify networks may be more or less time in-tensive than primary data collection; limitations may alsoexist in the types of variables available to study [7]. How-ever, particularly with electronic data, documentation ofnetwork activity may be readily available through regularquality monitoring, and span a time period to enable longi-tudinal analysis [7]. For example, data related to evidencesharing communication patterns and subsequent use ofbest practices by health professionals (e.g., intervention ap-proaches, treatment intensity, or dosage) within and acrossclinical teams can be leveraged to identify networkstrengths and gaps, and to monitor KT strategyeffectiveness.Observation is a fourth means of SNA data collec-

tion, which was absent from the reviewed studies. Al-though more resource-dependent and not without riskof observer-influenced behavior change, observationmay enable the identification of ties not captured

through self-report [7]. For example, interpersonal dy-namics during a meeting may be recorded by a thirdparty more objectively than meeting participants mayrecall, while concurrently focusing on the content ofthe meeting. Self-report data also presents potentialbias related to recall, particularly when respondentsare asked to think back to interactions in the past, orto report their frequencies [7]. While network rosterscan be used to help mitigate this problem in clearlybounded networks (e.g., an organization), in larger net-works this strategy can create excessive burden on re-spondents [7]. Careful attention to the way questionsare worded, and consideration of the number of altersrequested of respondents must be made to gathermeaningful and accurate data [7].

Networks and actorsWith more than half of the included studies examiningphysician-only networks, and only a handful studying inter-professional health care teams, great opportunity exists toexpand the range of professions under study. Because of thegrowing shift in health services delivery from profession-based to collaborative practice models involving interprofes-sional teams [57], further research is needed to evaluate thegeneralizability of findings beyond physician networks, aswell as in other health care contexts. The diversity of net-work sizes and settings, however, demonstrates the utility ofSNA for a broad range of applications, from the study of in-terventions in small hospital health care teams to largemulti-organizational or national networks. The examinationof inter-organizational networks (i.e., organizations as nodes)in the context of KT was beyond the scope of this review.Further study may be warranted to scope out this literatureand to determine its implications for informing KT from anorganizational network perspective.

Study purposes and data analysis methodsThe examination of information exchange processes, keyplayers, reasons for tie patterns, associations betweennetwork properties and various attributes, and the evalu-ation of KT intervention outcomes are crucial for under-standing how to successfully embed a new practice.However, because of its examination of information flow,predominantly, this body of research presents a narrowview of KT that focuses primarily at the individual levelof evidence-based decision-making. This limitation re-lates in part to the scope of the review, as well as theconsideration that other actors (e.g., health leaders, re-searchers) rather than health professionals may typicallymanage many of the KT activities that were not repre-sented. Extending the application of SNA to broaderorganizational or group processes and through a widerrange of KT-related activities and actors will advanceour understanding of the network dynamics involved in


all aspects required to move evidence into action. Suchphenomena of interest may include the collaborativeproduction of KT tools, barriers assessments, implemen-tation processes, and evaluation efforts. Examining theseprocesses from a network perspective has the potentialto identify strengths and gaps in the network that needto be addressed, to explore the structural characteristicsand associated attribute-level variables that might con-tribute to their success, to describe the role of healthprofessionals in these processes, and to evaluatenetwork-level KT interventions to facilitate them.The limited number of network properties (i.e., three

or fewer) examined in more than half of this body ofliterature suggests that the potential for greater SNA-related insights from these studies to inform futureresearch and practice in KT remains largely untapped.Simply describing networks or examining a single net-work property (e.g., tie homophily, centrality) and itsassociation with attribute variables fails to leverageSNA’s full potential. As KT scientists, we are interestedin not only what is happening, but why it occurs, andthe processes involved. With this information, we arepositioned more effectively to design network-basedKT interventions.For example, basic SNA can be used to identify key

players with influence within the network; subsequentanalyses can be used to explain how these individualscame to hold these positions. Knowledge of an individ-ual’s structural position may also help to determine fromwhom they may seek evidence or KT support. This in-formation can be used to develop KT interventions thattarget specific health professionals or groups of individ-uals based on their network structure or key attributesto strengthen KT processes. For instance, influential in-dividuals can be leveraged as champions or knowledgebrokers to improve the efficiency of information ex-change or behavioral influence within a discipline group.Individuals with attributes in common with key playerscan be selected to lead KT interventions within an inter-professional health care team. Alternate paths for effi-cient information exchange or behavioral influence canbe accessed if resistance by specific individuals isencountered.An understanding of relational influences can also

advance the science of KT by improving the specificityof KT interventions, and by supporting their evalu-ation. For instance, KT intervention fidelity (e.g.,intended versus actual information flow) can be moni-tored using SNA, and the KT intervention can be ad-justed accordingly over time to address gaps orbarriers. Network-specific outcomes of a KT interven-tion (e.g., increased connectedness, access to informa-tion) can also be evaluated empirically based onrelational data. Collaboration between KT and SNA

researchers may enable a more in-depth examinationof the data available from KT research, to bring newinsights from a network perspective.Visualizations are an asset in SNA research because of

their ability to represent the data in a way that makes itmore accessible to those less familiar with SNA method-ologies [58]. Surprisingly, fewer than half of the articlespresented network maps, which suggests that re-searchers could do more to elucidate descriptive rela-tional findings for readers. While more complex thandescriptive analyses, graphing the results of p2 modelscan illustrate the relationship between binary networkdata and covariates, while factoring in network structure[59]. Stacked correspondence analysis of matrices repre-senting different time periods can be used to visualizenetwork data at different time points [7]. Supplementalgraphing using conventional visualization methods (e.g.,bar, scatterplot, line charts) is also available as an ap-proach to visualize the relationships among networkproperties and attributes that has yet to be fully lever-aged. Appropriate visualization methods and techniquesmust be selected to answer the research questions ofinterest, while preserving clarity [7, 60].While a range of analytical techniques was identified,

many studies employed traditional analytical techniquesdesigned for data meeting assumptions of independence.By their nature within the SNA paradigm, dyadic datado not meet these assumptions. Techniques designed toaccount for interdependencies in the data, includingquadratic assignment procedure (QAP) analysis and ex-ponential random graph models (ERGM), are consideredmore robust for those analyses of specific hypotheses in-volving dyadic ties or network characteristics. These ap-proaches enable the modeling of relationships betweendyadic (i.e., relational) variables (e.g., information ex-change) and attribute variables (e.g., gender), betweendyadic variables (e.g., similarity in EIP attitudes, and en-gagement in research collaboration), or at the whole net-work level (e.g., density of communication ties relativeto time to evidence adoption) [7]. An example from theincluded literature is the use of ERGM to help determinewhether particular individuals—say those with similarpersonal characteristics—are connecting for informationsharing more than expected due to chance [29].With the inclusion of longitudinal designs, analysis ap-

proaches, such as stochastic actor-based network models(SABM) that examine network change over time, canbegin to be represented in this body of literature. SABMcan represent both ties and individual attributes toexamine network change. As an example,Yousefi-Nooraie et al. [61] used SABM to determine theeffect of their intervention (evidence-baseddecision-making skills) on participants’ status as know-ledge brokers.


Network propertiesFurther KT-related research that includes analyses ofcentralization, subgroups, and transitivity may afford amore in-depth understanding of the network-relatedinfluences on KT among health professionals.Centralization (i.e., the unevenness of connectivityacross the network) can be calculated for the wholenetwork, or for departments or sectors within anorganization for the purposes of comparison. Subgroups(e.g., smaller connected groups within a network) can beidentified and addressed individually during a KT inter-vention. For example, isolated individuals can be en-gaged to form connections with colleagues to benefitfrom their knowledge or influence. Different subgroupsmay receive different KT interventions based on theircharacteristics and what evidence or theory suggeststheir influence might be. Efforts to link or to expandsubgroups may precede implementation efforts to estab-lish an environment more conducive to change.Transitivity has been used to examine the tendency of

individuals to exchange information with a small versusa large number of sources, and for the network to formhighly connected hubs. This analysis can inform the de-sign of KT interventions to improve the efficiency of in-formation sharing or influence (e.g., identifying targetsfor the intervention and relying on transitive processesto spread the information rather than targeting all net-work members). Such a strategy can then be comparedto alternatives, to test hypotheses about the influence ofdifferent network properties on the effectiveness of KTinterventions. This evaluative work is critical to improveour understanding of network influences on KT pro-cesses and outcomes.The range of structural properties examined suggests

that researchers consider multiple structural phenom-ena to be relevant to KT processes and outcomes. Con-siderable overlap also existed in the real-worldphenomena being evaluated by proxy through theseproperties. This diversity provides a foundation onwhich to build a stronger knowledge base about net-works’ multiple influences on research use. This ap-proach aligns closely with current discussion in the KTliterature about complex health systems, and the needto use “complexity-informed approaches” to embedevidence-informed changes in the health care system[62–64]. Such systems models assume that health careorganizations are dynamic, interdependent, containsub-systems with feedback loops, and exhibit emergentproperties. A combination of a complex adaptive sys-tem lens and SNA modeling to measure and explainfeatures of networks and individuals, and most criticallythe relationship between networks, sub-networks (likecliques), and individuals as they change over time, is anunderutilized approach to KT and implementation. The

approach moves from a mechanical understanding ofKT barriers and facilitators to a much more complexpicture of what is required to introduce and to sustainchange in health care organizations.While complex statistical models are more difficult to

apply, they present the benefit of multivariate analysesto better examine the interaction between various fac-tors, as well as the opportunity to control for covariatesthat may be inflating or masking key effects. As statis-tical models for SNA continue to emerge, these toolswill become increasingly important in clarifying the rela-tive influence of various network properties and attri-bute effects thought to influence KT.An understanding of the influences of individual at-

tributes and different types of ties (e.g., friendship)can support the structuring of the health care envir-onment to strengthen network density within homo-philous groups (e.g., discipline groups), or to fostergreater diversity of collaborations within the network(e.g., interprofessional clinical, project, or professionaldevelopment work). Knowledge of network structurecan be used to target EIP behaviors using a social in-fluence approach that introduces innovation into thenetwork’s core or brokers to reduce pervasive infor-mation bias, or that facilitates greater density overall.Unsurprisingly, different approaches to KT (e.g.,leader-driven vs. collaborative; researcher-led dissem-ination vs. researcher-clinician collaboration) maypresent different structural properties (e.g., hierarch-ical vs. clustering). This finding has implications forinformation flow, as well as for the development ofnetwork interventions to address gaps or to leverageprominent actors in the network to champion theinnovation. Those with formal health care leadershiproles may not be the only individuals with influence;informal brokers may be more recognizable by net-work members (particularly peers) as central to KTprocesses than formal leaders [37, 40]. Alternatively,some networks may not present with prominent cen-tral actors or opinion leaders for EIP [29]. A SNAlens can assist in identifying these individuals whenthey exist, analyzing the extent of their reach, deter-mining the reasons for their prominence, and devel-oping a network-informed plan to leverage theirposition to advance KT.Normative group processes and structural position

can also explain the timing of adoption, which maybe useful in identifying early adopters, opinionleaders, and shared attitudes within a network [23,24]. Differences reported in the preferred informationsources and network influences for early versus lateadopters from a diffusion of innovation perspectivecan be used to guide differential approaches to behav-iour change.


Theoretical insightsVarious theories, frameworks, and models guide KT re-search efforts [5], although only in a small proportion(3–6%) of primary research articles in the broader KTfield [65, 66]. The included articles demonstrated abroad range of theoretical approaches drawn from thefields of sociology and psychology (e.g., diffusion ofinnovation, social influence, social contagion, social ex-change), as well as from the field of SNA itself (e.g.,perspectives that explain the role of weak ties, struc-tural holes, cohesion, or tie homophily on network dy-namics). The variety of approaches used suggests thatdiverse theories may merit exploration for their utilityin KT-related SNA research, and that multiple theoriesmay be applied in a single study (e.g., [24, 25, 33]).Diffusion of innovation theory is applied commonly

in the KT literature [5], so its frequent application herewas not unexpected. The theory’s principles lend them-selves well to a SNA paradigm, in that the theory wasdeveloped to predict or to explain how information orinnovation spreads within social systems [67]. The col-lection of attribute data about network members per-mits the analysis of characteristics that influence anindividual’s adoption of innovation, relative to their net-work position and other contextual factors [68]. Whilethe majority of KT strategists have adopted an educa-tional approach (i.e., by implementing KT interventionsbased on an “information dissemination paradigm”) [5,69] to improve awareness, understanding, and attitudesas a means of influencing uptake [70, 71], the role ofsocial networks in their impact has yet to be compre-hensively studied. Studies that pair traditional educa-tional and behavioral outcomes research with SNA mayprovide greater insight into the social mechanisms thatinfluence these outcomes.Social contagion/influence theories are common in

the SNA literature, and purport that actors share atti-tudes, knowledge, or behaviors because of their tiesto others who influence them [72]. This perspectivehighlights the role of peers and others in fosteringbehavior change, attitudes, and identities [69]. Thestudy of opinion leaders, audit and feedback [5, 66,71], and mentoring [73, 74] may follow a social influ-ence perspective [69], as does a growing body of re-search on knowledge brokering as a human mediatorof research uptake in health care [75]. Social capitaltheory, another prominent SNA theory, attributes tieformation to an individual’s need for social capitalfrom others (e.g., resources, information, power). Thiscontrasting approach was used in the context ofknowledge brokering; its application in tandem withsocial influence theory may provide insight into thedirection of causality (i.e., ties form because of attri-butes vs. attributes are the result of ties) [7].

Transactive memory theory, social exchange theory,Habermas’ theory of communicative power, sociology ofprofessions theory, balance theory, and an epistemic dif-ferences perspective have logical applications to SNA.Each of these perspectives addresses social factors ap-plicable to networks. The transactive memory-social ex-change model describes how information seekingbehavior is influenced by the awareness and valuing ofanother individual’s knowledge or skills, their accessibil-ity, and the cost or effort involved in seeking the infor-mation [37]. Communicative power is meant to describethe influence of a third party on the mutual understand-ing achieved between a pair of individuals [42]. Soci-ology of professions theory describes how members of ashared profession tend to develop a collective statusbased on their unique health care jurisdictions, commontraining, and mutual knowledge, which can limit theirtendency to interact outside this peer group [22]. Bal-ance theory contends that individuals tend to developbalanced relationships (e.g., tie reciprocity) in order tocircumvent unease [27]. An epistemic differences per-spective purports that individual attributes in combin-ation with structures, processes, and other features ofthe environmental context (including power), influenceindividual performance and experiences [39]. These dif-ferences create diversity within the network that gener-ate opportunities for novel sharing and innovation [39].By understanding the structural network factors thatcontribute to information seeking, epistemic differences,power dynamics, and communication processes, we maygain a more holistic understanding of KT. Further ana-lysis is required to determine the utility other theoriesand frameworks from the KT and SNA literature toSNA-related KT research.The study of networks is particularly relevant when

considering the role of social evaluation in the diffu-sion of innovation. Gartrell [76] argues that socialevaluation is a key driver of decision-making, notingthe role of networks in providing norms and compari-son opportunities that influence behavior. Social influ-ence theory, and the principles of tie homophily,social contagion, and structural equivalence parallelthis line of thought. While the role of context is ad-dressed relatively commonly in KT and implementa-tion models and frameworks [75], the concept ofsocial evaluation is largely absent from the non-SNAKT literature. SNA can be applied to identify the at-tributes, structural positions, or nature of the rela-tionships that are most influential for KT processesfrom a social evaluation perspective, including therole of brokers, homophily, hierarchy, centrality, andother network properties. These network characteris-tics offer insights into the patterns and restrictions inthe flow of information as well as into power


structures, key actors, and their reach and efficiency.Normative group processes and structural positioncan also explain the timing of adoption, which maybe useful in identifying early adopters, opinionleaders, and the development of shared attitudeswithin a network [24, 76].Social constructivist and cognitive learning theories

have also emerged in the KT literature to explore themeans by which health professionals interact withintheir social context to construct and to understandknowledge [66, 77, 78]. The network perspective has notbeen combined with these approaches to enrich our un-derstanding of the role of this social context on evidenceuptake. Also missing from the included articles is refer-ence to behavior change theories that have been preva-lent in the KT literature [5]. These theories emphasizethe various barriers and facilitators to change, includingpersonal, organizational, and system-level factors [65].Integrating a SNA lens to behavior change models mayaugment their utility by including social structure as adeterminant of behaviour change.

LimitationsThe search and screening process used in this reviewlimited inclusion to studies involving quantitative ana-lysis of SNA data; a greater depth of understandingabout SNA’s utility for KT may also be gained from the-oretical discussion papers. Also omitted was the contri-bution of qualitative data from network research, whichcan provide insight into how social structures, networkproperties, and the nature of relationships influence KTfrom the perspectives of network members. A broaderset of inclusion criteria that encompassed studies on theadoption of non-clinical best practices, including tech-nologies, training strategies, quality improvement initia-tives and other innovations, as well as KT in the policycontext, may also inform aspects of KT research and im-plementation initiatives, but was beyond the scope ofthis review.This discussion only addressed these studies’

SNA-specific findings. Important non-SNA-related find-ings (e.g., the most important sources of evidence toaugment awareness and adoption [23]) have not beensummarized but can also inform KT research and prac-tice. In addition, the scope of this paper prevented anin-depth discussion of the full breadth of theoretical per-spectives represented in the SNA KT literature, whichmay warrant a separate review. Finally, because of thevariability in study designs, the lack of inclusion of qual-ity appraisals in scoping review methodology, and the in-consistency of network properties examined across thestudies, results should be interpreted with caution untilfurther research can evaluate their quality andgeneralizability.

ConclusionsGiven the diversity and complexity of health professionalKT networks, optimal strategies for KT may vary depend-ing on the structure of a professional or organizationalnetwork, as well as on professional identities and personalattributes. Within a given setting, interprofessional dy-namics, hierarchy, social influence, centralization, brokers,and other important structural properties are worthy ofconsideration. The SNA paradigm offers a broader lens bywhich to examine and to influence KT processes and out-comes across contexts, while drawing on established the-ories known to the KT science field. SNA extends thescope of KT influence to include social relationships andstructural characteristics of individual and whole net-works. However, its full potential has yet to be realized.Suitable for relatively small (e.g., a dozen) to larger net-

works of several hundred members or more, SNA can beused to describe or to evaluate groups within or acrossdepartments, organizations, countries, or beyond. Longi-tudinal research, a more representative range of popula-tions, the use of interviews, document review andobservation for data collection, greater depth of analysis,and the leveraging of network visualizations can augmentthe contributions of SNA to the KT science knowledgebase.Understanding how network properties can be used as

proxies to measure social processes (e.g., information ex-change, best practice adoption, decision-making, influ-ence) can help KT scientists to apply SNA effectively toexpand the range of measures that can be used to evaluateKT efforts. The approach can be used to describe a net-work as a precursor to a KT intervention, as a means ofsupporting planning (e.g., identifying target groups or in-dividuals), as well as for testing hypotheses. Evaluating in-formation sharing, positions of influence, relationshipsbetween network connection patterns and individual attri-butes (e.g. attitudes) or behaviors, and the effectiveness ofKT interventions relying on or targeting networks are allfeasible. Predicting or explaining patterns of connections,comparing groups, time points or contexts are alsopossible.Finally, while this article did not present a comprehensive

overview of the use of theory across the entire body ofSNA-related KT literature, it does offer a starting point forconceptualizing theory-based SNA applications in KT re-search. In keeping with a systems or complexity theory ap-proach, SNA can offer a wider spectrum of determinants toexamine in evaluating KT processes by addressing socialfactors.The targeted SNA research outlined here may help

to highlight the role of interactions, relationships, andother social dynamics throughout the full scope of ac-tivities and processes required to move evidence intoaction within health care settings and beyond.


AbbreviationsCINAHL: Cumulative Index to Nursing and Allied Health Literature;EIP: Evidence informed practice; KT: Knowledge translation; SIENA: Simulationinvestigation for empirical network analysis; SNA: Social network analysis

AcknowledgmentsThe authors would like to acknowledge the contributions of Dr. DavidTindall in helping to shape the refinement of a previous version of thismanuscript, and Andrea Ryce for her contributions to the development ofthe search strategy.

FundingSG was supported by a Vanier Canada Graduate Scholarship, a CanadianChild Health Clinician Scientist Program Career Enhancement Award, a Four-Year Fellowship, Faculty of Medicine Graduate Award, Rehabilitation SciencesDoctoral Scholarship and Public Scholars Award from the University of BritishColumbia, and funding from the Sunny Hill Foundation.

Availability of data and materialsAll data generated or analyzed during this study are included in thispublished article.

Authors’ contributionsSG designed the study, conducted the search, screened for inclusion,extracted data, synthesized the literature, and drafted the manuscript. EJscreened for inclusion, contributed to study design refinement and revisionsto draft manuscript. AK verified data extraction and contributed to studydesign refinement and revisions to draft manuscript. All authors read andapproved the final manuscript.

Authors’ informationSG is a doctoral candidate in Rehabilitation Sciences at the University ofBritish Columbia, and the Knowledge Broker Facilitator with the ChildDevelopment & Rehabilitation Evidence Centre at Sunny Hill Health Centrefor Children. EJ is an Assistant Professor in the School of Nursing at theUniversity of British Columbia. AK is an Associate Professor in the School ofHealth Studies at Western University.

Ethics approval and consent to participateNot applicable.

Consent for publicationNot applicable.

Competing interestsThe authors declare that they have no competing interests. This studyrepresents work conducted in partial fulfillment of SG’s doctoral thesis.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Author details1Rehabilitation Sciences, The University of British Columbia, 212 - 2177Wesbrook Mall, Vancouver, BC V6T 1Z3, Canada. 2Therapy Department, SunnyHill Health Centre for Children, 3644 Slocan Street, Vancouver, BC V5M 3E8,Canada. 3BC Children’s Hospital Research Institute, 938 West 28th Ave,Vancouver, BC V5Z 4H4, Canada. 4School of Nursing, The University of BritishColumbia, T201-2211 Wesbrook Mall, Vancouver, BC V6T 2B5, Canada.5School of Health Studies, Western University, Arthur and Sonia Labatt HealthSciences Building, Room 222, London, ON V6A 5B9, Canada.

Received: 17 August 2018 Accepted: 11 March 2019

References1. Graham ID, Tetroe J. Some theoretical underpinnings of knowledge

translation. Acad Emerg Med. 2007;14:936–41.2. Menon A, Korner-Bitensky N, Kastner M, McKibbon KA, Straus S. Strategies

for rehabilitation professionals to move evidence-based knowledge intopractice: a systematic review. J Rehabil Med. 2009;41(13):1024–32. [cited2015 Aug 18]. Available from: https://medicaljournals.se/jrm/content/abstract/10.2340/16501977-0451.

3. Nilsen P. Making sense of implementation theories, models and frameworks.Implement Sci. 2015;10(1):53.

4. LaRocca R, Yost J, Dobbins M, Ciliska D, Butt M. The effectiveness ofknowledge translation strategies used in public health: a systematic review.BMC public health. 2012;12(1):751. Available from: BMC Public Health

5. Colquhoun HL, Letts LJ, Law MC, MacDermid JC, Missiuna CA. A scopingreview of the use of theory in studies of knowledge translation. Can JOccup Ther. 2010;77(5):270–9.

6. Chambers D, Wilson P, Thompson C, Harden M. Social network analysis inhealthcare settings: a systematic scoping review. PLoS One. 2012;7(8):e41911.

7. Borgatti SP, Everett MG, Johnson JC. Analyzing social networks. Los Angeles:SAGE Publications; 2013. p. 296.

8. Cunningham FC, Ranmuthugala G, Plumb J, Georgiou A, Westbrook JI,Braithwaite J. Health professional networks as a vector for improvinghealthcare quality and safety: a systematic review. BMJ Qual Saf. 2012;21(3):239–49.

9. Bae S, Nikolaev A, Seo JY, Castner J. Health care provider social networkanalysis: a systematic review. Nurs Outlook. 2015;63(5):566–84.

10. Benton DC, Perez-Raya F, Fernandez-Fernandez MP, Gonzalez-Jurado MA. Asystematic review of nurse-related social network analysis studies. Int NursRev. 2015;62(3):321–39.

11. Hurlock-Chorostecki C, Forchuk C, Orchard C, van Soeren M, Reeves S.Hospital-based nurse practitioner roles and interprofessional practice: ascoping review. Nurs Health Sci. 2014;16:403–10.

12. Zhang S, de la Haye K, Ji M, An R. Applications of social network analysis toobesity: a systematic review. Obes Etiol. 2018;19(7):976–88.

13. Long JC, Cunningham FC, Braithwaite J, Braithwaite J, Braithwaite J,Westbrook M, et al. Bridges, brokers and boundary spanners in collaborativenetworks: a systematic review. BMC Health Serv Res. 2013;13(1):158.

AppendixTable 4 OvidSP MEDLINE search strategy

Concept Keywords MeSH Headings

Health careprofessionals

Clinician* OR “health care professional*” OR“health professional*” OR therapist* OR physician*OR doctor* OR nurse* OR “allied health”

Allied Health Personnel OR MedicalStaff Hospital

Knowledgetranslation

Knowledge translation OR evidence based practiceOR evidence informed practice OR disseminationOR organi*ational innovation OR implementationadj3 research OR research utili*ation OR researchuse OR knowledge utili*ation OR knowledge transferOR knowledge mobili*ation OR knowledge exchange

Knowledge Management OR TranslationalMedical Research OR Diffusion of InnovationOR Evidence-Based Practice OR ProfessionalPractice OR Guideline Adherence ORSocial Change

Socialnetwork analysis

Social Network, social networks, network theory Interprofessional Relations


https://medicaljournals.se/jrm/content/abstract/10.2340/16501977-0451

https://medicaljournals.se/jrm/content/abstract/10.2340/16501977-0451

14. Tasselli S. Social networks of professionals in health care organizations: areview. Med Care Res Rev. 2014;71(6):619–60.

15. Sabot K, Wickremasinghe D, Blanchet K, Avan B, Schellenberg J. Use ofsocial network analysis methods to study professional advice andperformance among healthcare providers: a systematic review. Syst Rev.2017;6:208.

16. Arksey H, O’Malley L. Scoping studies: towards a methodological framework.Int J Soc Res Methodol. 2005;8(1):19–32.

17. Levac D, Colquhoun H, O’Brien KK. Scoping studies: advancing themethodology. Implement Sci. 2010;5(1):69.

18. KA MK, Lokker C, Wilczynski NL, Haynes RB, Ciliska D, Dobbins M, et al.Search filters can find some but not all knowledge translation articles inMEDLINE: an analytic survey. J Clin Epidemiol. 2012;65:651–9.

19. McKibbon KA, Lokker C, Wilczynski NL, Ciliska D, Dobbins M, Davis DA, et al.Cross-sectional study of the number and frequency of terms used to referto knowledge translation in a body of health literature in 2006: a tower ofbabel? Implement Sci. 2010;5(1):16.

20. Lokker C, McKibbon KA, Wilczynski NL, Haynes RB, Ciliska D, Dobbins M, etal. Finding knowledge translation articles in CINAHL. Stud Health TechnolInform. 2010;160(Pt 2):1179–83.

21. Sibbald SL, Wathen CN, Kothari A, Day AMB. Knowledge flow and exchangein interdisciplinary primary health care teams (PHCTs): An exploratory study.J Med Libr Assoc. 2013;101(2):128–37. Available from: https://www-ncbi-nlm-nih-gov.ezproxy.library.ubc.ca/pmc/articles/PMC3634376/.

22. Tasselli S. Social networks and inter-professional knowledge transfer: thecase of healthcare professionals. Organ Stud. 2015;36(7):841–72.

23. Ankem K. Influence of information sources on the adoption of uterinefibroid embolization by interventional radiologists. J Med Libr Assoc. 2003;91(4):450–9.

24. Burt RS. Social contagion and innovation: cohesion versus structuralequivalence. Am J Sociol. 1987;92(6):1287–335.

25. Friedkin NE. A multilevel event history model of social diffusion: medicalinnovation revisited. J Math Sociol. 2010;34(2):146–55. Available from:https://www.tandfonline.com/doi/abs/10.1080/00222500903221589,

26. Heijmans N, van Lieshout J, Wensing M. Information exchange networks ofhealth care providers and evidence-based cardiovascular risk management:an observational study. Implement Sci [Internet]. 2017;12(1):1–12. Availablefrom: https://doi.org/10.1186/s13012-016-0532-1

27. Mascia D, Pallotti F, Dandi R. Determinants of knowledge-sharing networksin primary care. Health Care Manag Rev. 2018;43(2):104–14.

28. Di Vincenzo F, Mascia D. Knowledge development and advice networks inprofessional organizations. Knowl Manag Res Pract. 2017;15(2):201–13.

29. Zappa P. The network structure of knowledge sharing among physicians.Qual Quant. 2011;45(5):1109–26. Available from: https://healthcouncilcanada.ca/files/2.48-2006_AnnualReport_1.pdf.

30. Mascia D, Cicchetti A, Fantini MP, Damiani G, Ricciardi W. Physicians’ propensityto collaborate and their attitude towards EBM: a cross-sectional study. BMCHealth Serv Res. 2011;11(1):172.

31. Mascia D, Cicchetti A. Physician social capital and the reported adoption ofevidence-based medicine: exploring the role of structural holes. Soc SciMed. 2011;72(5):798–805. Available from: https://doi.org/10.1016/j.socscimed.2010.12.011

32. Mascia D, Cicchetti A, Damiani G. “Us and Them”: a social network analysisof physicians’ professional networks and their attitudes towards EBM. BMCHealth Serv Res. 2013;13(1):429.

33. Mascia D, Dandi R, Di Vincenzo F. Professional networks and EBM use: astudy of inter-physician interaction across levels of care. Health Policy (NewYork). 2014;118(1):24–36.

34. Doumit GD, Papay FA, Moores N, Meisler E, Zins JE. Opinion leaders andevidence-based medicine in craniofacial surgery. J Craniofac Surg. 2014;25(1):106–10.

35. Paul S, Keating NL, Landon BE, O’Malley AJ. Reprint of: results from using anew dyadic-dependence model to analyze sociocentric physician networks.Soc Sci Med. 2015;125:51–9. Available from: https://doi.org/10.1016/j.socscimed.2014.08.027

36. Yousefi-Nooraie R, Dobbins M, Brouwers M, Wakefield P. Informationseeking for making evidence-informed decisions: a social network analysison the staff of a public health department in Canada. BMC Health Serv Res.2012;12:118.

37. Yousefi-Nooraie R, Dobbins M, Marin A. Social and organizational factorsaffecting implementation of evidence-informed practice in a public health

department in Ontario: a network modelling approach. Implement Sci.2014;9:29. Available from: https://implementationscience.biomedcentral.com/articles/10.1186/1748-5908-9-29.

38. Guldbrandsson K, Nordvik MK, Bremberg S. Identification of potentialopinion leaders in child health promotion in Sweden using networkanalysis. BMC Res Notes. 2012;5:42. https://doi.org/10.1186/1756-0500-5-424,https://www.ncbi.nlm.nih.gov/pubmed/22873749.

39. D'Andreta D, Scarbrough H, Evans S. The enactment of knowledgetranslation: a study of the Collaborations for Leadership in Applied HealthResearch and Care initiative within the English National Health Service. JHealth Serv Res Policy. 2013;18(Suppl 3):40–52.

40. Long JC, Cunningham FC, Carswell P, Braithwaite J. Who are the key playersin a new translational research network? BMC Health Serv Res. 2013;13:1.

41. Long JC, Cunningham FC, Carswell P, Braithwaite J. Patterns of collaborationin complex networks: the example of a translational research network. BMCHealth Serv Res. 2014;14(1):225. Available from: https://bmchealthservres.biomedcentral.com/articles/10.1186/1472-6963-14-225.

42. Quinlan E, Robertson S. The communicative power of nurse practitioners inmultidisciplinary primary healthcare teams. J Am Assoc Nurse Pract. 2013;25(2):91–102.

43. Bunger AC, Hanson RF, Doogan NJ, Powell BJ, Cao Y, Dunn J. Can learningcollaboratives support implementation by rewiring professional networks?Adm Policy Ment Health. 2016;43(1):79–92. Available from: https://doi.org/10.1007/s10488-014-0621-x

44. Menchik DA, Meltzer DO, Menchik DA, Meitzer DO. The cultivation ofesteem and retrieval of scientific knowledge in physician networks. J HealthSoc Behav. 2017;51(2):137–52.

45. Keating NL, Ayanian JZ, Cleary PD, Marsden PV. Factors affecting influentialdiscussions among physicians: a social network analysis of a primary carepractice. J Gen Intern Med. 2007;22(6):794–8.

46. Graham ID, Logan J, Harrison MB, Straus SE, Tetroe J, Caswell W, et al. Lostin knowledge translation: time for a map? J Contin Educ Heal Prof. 2006;26(1):13–24.

47. Racko G. Knowledge exchange in the UK CLAHRCs: the enabling role ofacademics and clinicians’ social position. J Health Organ Manag. 2018;32(2):246–62.

48. Mascia D, Di Vincenzo F, Iacopino V, Fantini MP, Cicchetti A. Unfoldingsimilarity in interphysician networks: the impact of institutional andprofessional homophily. BMC Health Serv Res. 2015;15(1):1–9.

49. Valente TW, Pumpuang P. Identifying opinion leaders to promote behaviorchange. Heal Educ Behav. 2007;34(6):881–96. Available from: http://journals.sagepub.com/doi/10.1177/1090198106297855

50. Valente TW, Fujimoto K. Bridging: locating critical connectors in a network.Soc Networks. 2010;32(3):212–20. Available from: https://doi.org/10.1016/j.socnet.2010.03.003

51. Valente TW, Fujimoto K, Palmer P, Tanjasiri SP. A network assessment ofcommunity-based participatory research: linking communities anduniversities to reduce cancer disparities. Am J Public Health. 2010;100(7):1319–25.

52. Valente TW, Palinkas LA, Czaja S, Chu K, Brown CH. Social network analysisfor program implementation 2015;1–18. Available from: https://doi.org/10.1371/journal.pone.0131712

53. Valente TW. Network Interventions. Science. 2012;337(6090):49–54.54. Snijders TAB. The SIENNA webpage [Internet]. [cited 2017 Nov 28]. Available

from: https://www.stats.ox.ac.uk/~snijders/siena/. Accessed 15 July 2018.55. El-Sayed AM, Scarborough P, Seemann L, Galea S. Social network analysis

and agent-based modeling in social epidemiology. Epidemiol PerspectInnov. 2012;9(1):1.

56. Hollstein B. Mixed methods social networks research: an introduction.In: Dominguez S, Hollstein B, editors. Mixed methods social networksresearch: design and applications. New York: Cambridge UniversityPress; 2014. p. 3–33.

57. Health Council Canada. Health care renewal in Canada: Clearing the road toquality. Annual report to Canadians [Internet]. Eglinton, ON; 2005. Availablefrom: http://healthcouncilcanada.ca/accord_theme.php?mnu=2&mnu1=36&mnu2=29. Accessed 15 July 2018.

58. Glegg SMN. Facilitating interviews in qualitative research with visual tools: atypology. Qual Health Res. 2018;29(2):301–10. https://journals.sagepub.com/doi/abs/10.1177/1049732318786485.

59. van Duijn MAJ, Snijders TAB, Zijlstra BJH. P2: A random effects model withcovariates for directed graphs. Stat Neerl. 2004;58(2):234–54.


https://www-ncbi-nlm-nih-gov.ezproxy.library.ubc.ca/pmc/articles/PMC3634376/

https://www-ncbi-nlm-nih-gov.ezproxy.library.ubc.ca/pmc/articles/PMC3634376/

https://www.tandfonline.com/doi/abs/10.1080/00222500903221589

https://doi.org/10.1186/s13012-016-0532-1

https://healthcouncilcanada.ca/files/2.48-2006_AnnualReport_1.pdf

https://healthcouncilcanada.ca/files/2.48-2006_AnnualReport_1.pdf

https://doi.org/10.1016/j.socscimed.2010.12.011




https://implementationscience.biomedcentral.com/articles/10.1186/1748-5908-9-29

https://implementationscience.biomedcentral.com/articles/10.1186/1748-5908-9-29

https://doi.org/10.1186/1756-0500-5-424

https://www.ncbi.nlm.nih.gov/pubmed/22873749

https://bmchealthservres.biomedcentral.com/articles/10.1186/1472-6963-14-225

https://bmchealthservres.biomedcentral.com/articles/10.1186/1472-6963-14-225

https://doi.org/10.1007/s10488-014-0621-x

https://doi.org/10.1007/s10488-014-0621-x

http://journals.sagepub.com/doi/10.1177/1090198106297855

http://journals.sagepub.com/doi/10.1177/1090198106297855

https://doi.org/10.1016/j.socnet.2010.03.003

https://doi.org/10.1016/j.socnet.2010.03.003

https://doi.org/10.1371/journal.pone.0131712

https://doi.org/10.1371/journal.pone.0131712

https://www.stats.ox.ac.uk/~snijders/siena/

http://healthcouncilcanada.ca/accord_theme.php?mnu=2&mnu1=36&mnu2=29

http://healthcouncilcanada.ca/accord_theme.php?mnu=2&mnu1=36&mnu2=29

https://journals.sagepub.com/doi/abs/10.1177/1049732318786485

https://journals.sagepub.com/doi/abs/10.1177/1049732318786485

60. Knoke D, Yang S. Social network analysis. 2nd ed. Thousand Oaks: SAGEPublications; 2008. 133 p.

61. Yousefi-nooraie R, Dobbins M, Marin A, Hanneman R, Lohfeld L. Theevolution of social networks through the implementation of evidence-informed decision-making interventions: a longitudinal analysis of threepublic health units in Canada. Implement Sci. 2015;10:166. Available from:https://doi.org/10.1186/s13012-015-0355-5.

62. Best A, Holmes B. Systems thinking, knowledge and action: towards bettermodels and methods. Evid Policy A J Res Debate Pract. 2010;6(1):145–59.

63. Holmes BJ, Best A, Davies H, Hunter D, Kelly MP, Marshall M, et al.Mobilising knowledge in complex health systems: a call to action. EvidPolicy A J Res Debate Pract. 2017;13(3):539–60.

64. Braithwaite J, Churruca K, Long JC, Ellis LA, Herkes J. When complexityscience meets implementation science: a theoretical and empirical analysisof systems change. BMC Med. 2018;16:63.

65. Michie S, van Stralen MM, West R, Grimshaw J, Shirran L, Thomas R, et al.The behaviour change wheel: a new method for characterising anddesigning behaviour change interventions. Implement Sci. 2011;6(1):42.Available from: http://implementationscience.biomedcentral.com/articles/10.1186/1748-5908-6-42

66. Colquhoun HL, Brehaut JC, Sales A, Ivers N, Grimshaw J, Michie S, et al. Asystematic review of the use of theory in randomized controlled trials ofaudit and feedback. Implement Sci. 2013;8(1):66.

67. Rogers E. Diffusion of innovations. New York: Free Press; 1995.68. Greenhalgh T, Robert G, Macfarlane F, Bate P. Diffusion of innovations in

service organizations: systematic review and recommendations. Milbank Q.2004;82(4):581–629.

69. Mano-Negrin R, Mittman B. Theorising the social within physician decisionmaking. J Manag Med. 2011;15(4–5):663–7.

70. Yamada J, Shorkey A, Barwick M, Widger K, Stevens BJ. The effectiveness oftoolkists as knowledge translation strategies for integrating evidence intoclinical care: a systematic review. BMJ Open. 2015;5(4):e006808.

71. Jones CA, Roop SC, Pohar SL, Albrecht L, Scott SD. Translating knowledge inrehabilitation: systematic review. Phys Ther. 2015;95(4):663–7.

72. Friedkin NE. A structural theory of social influence. Structural analysis in thesocial sciences. New York: Cambridge University Press; 1998.

73. Gagliardi AR, Fiona W, Perrier L, Bell M, Straus S. Exploring mentorship as astrategy to build capacity for knowledge translation research and practice: ascoping systematic review. Implement Sci. 2014;9(1):122.

74. Abdullah G, Rossy D, Ploeg J, Davies B, Higuchi K, Sikora L, et al. Measuringthe effectiveness of mentoring as a knowledge translation intervention forimplementing empirical evidence: a systematic review. WorldviewsEvidence-Based Nurs. 2014;11(5):284–300.

75. Glegg SMN, Hoens AM. The role domains of knowledge brokering: a modelfor the healthcare setting. J Neurol Phys Ther. 2016;40(2):115–23.

76. Gartrell CD. Network approaches to social evaluation. Annu Rev Sociol.1987;13:49–66.

77. Thomas A, Menon A, Boruff J, Rodriguez AM, Ahmed S. Applications ofsocial constructivist learning theories in knowledge translation forhealthcare professionals: a scoping review. Implement Sci. 2014;9(1):54.

78. Brehaut JC, Eva KW. Building theories of knowledge translationinterventions: use the entire menu of constructs. Implement Sci. 2012;

79. Hanneman RA, Riddle M. Introduction to social network methods [internet].Riverside: University of California, Riverside; 2005. 332 p. Available from:http://faculty.ucr.edu/~hanneman/nettext/.

80. Baldassarri D, Diani M. The integrative power of civic networks. Am J Sociol.2007;113:735–80.

81. Burt RS. Structural holes. In: structural holes. Cambridge: CambridgeUniversity Press; 1992.

82. Granovetter MS. The strength of weak ties. Am J Sociol. 1973;78(6):1360–80.83. Coleman JS. Social capital in the creation of human capital. Am J Sociol.

1988;94:S95–120.84. West E, Barron DN, Dowsett J, Newton JN. Hierarchies and cliques in the

social networks of health care professionals: implications for the design ofdissemination strategies. Soc Sci Med. 1999;48(5):633–46.


https://doi.org/10.1186/s13012-015-0355-5

http://implementationscience.biomedcentral.com/articles/10.1186/1748-5908-6-42

http://implementationscience.biomedcentral.com/articles/10.1186/1748-5908-6-42

http://faculty.ucr.edu/~hanneman/nettext/

Documents

How the study of networks informs knowledge translation ...advancing KT science Also summarized is the use of theory in this SNA research to demonstrate the fit of SNA with theoretical