Skip to main content

Table 3 Variables, network properties and key findings

From: How the study of networks informs knowledge translation and implementation: a scoping review

Citation Primary data analysis method Variables of interest Findings
Attributes Structural or relational parameters Network property used as proxy for structural parameter
Descriptive/exploratory studies
 Yousefi-Nooraie 2012 [36] (same study as [37]) Deriving network properties to describe the network _ Connectedness Whole network density Low density (1.2%) observed
Information exchange Tie reciprocity Head management division identified as central cluster bridging organizational divisions, with hierarchical information flow.
Expertise recognition and information seeking clustered within divisions; friendships spanned departments;
Friendship and expertise recognition predicted information seeking ties;
Network-identified brokers should receive same interventions and supports as formal brokers
Prestige (key actors) Indegree centrality
Mediating power of actors Betweenness centrality
Subgroups of connected actors Clusters
Brokers (actors connecting distinct teams/clusters of alters) Brokerage patterns (measured by which groups the information source amd its recipients belonged)
 Sibbald 2013 [21] Deriving network properties to describe the network Cohesiveness related to giving and seeking research-related information Whole network density Low density for information seeking and giving (7–12%) observed; suggested these behaviors not a central focus of the interprofessional relationships
Profession Key players (prestige) in giving and seeking research information Indegree centrality Medical residents prominent in knowledge exchange; physician seen as primary implementer of evidence; nurses as intermediaries between physicians and support staff; allied health more likely to draw information from external networks
 Quinlan 2013 [42] Deriving network properties to describe the network Profession tenure of the team
Occupational distance among members
Number of team members
Communicative power (i.e., the facilitation of mutual understanding among other team members) Flow betweenness centrality True interprofessional decision-making attributed to low structural hierarchy. Nurse practitioners (in newly formed teams) and registered nurses (in established teams) tended to have greatest communicative power. Mutual understanding and professions’ involvement varied across clinical decision-making episodes
Change in flow betweenness centrality between clinical decisions
 Long 2014 [41] (same study as [40]) Descriptive SNA; correlations among specific network properties and/or attributes Geographic proximity profession (e.g., clinicians/researchers) Grouping based on similarity in attributes # components Geographical proximity, professional homophily associated with clustering (past collaborations); geographical proximity and past collaborative ties influenced current and future collaborations. Intended future collaborations were more interprofessional.
Network density varied across networks (current 4%, future 27%, past 31%).
Weak ties and reputation associated with intention for future collaboration; strong ties associated with current collaborations.
External-internal (E-I) indices based on tie homophily
Clustering coefficients (comparing ego-and whole network density)
Past strong collaborations; Current or future collaborations Past collaboration network tie strength; Current or future collaboration ties
Previous experience in the field Current or future collaborations Current or future collaboration ties
Actor’s reputation Indirect contacts Future collaboration network tie strength
Future intended collaborations Future collaboration ties
 Long 2013 [40] (same study as [41]) Chi square analyses to test for association between attributes and network position (i.e., key actor status) Current workplace
Gender
Membership in other networks
Qualifications
Approach to work within the network
Key actors (with respect to power, influence or connectedness) Indegree centrality
Betweenness centrality
A manager, and specific researchers and clinicians identified as key players.
Apart from expert status and valuing adequate network resources, network-identified central players and formal brokers had little in common.
Planned interventions and support for formal broker roles may be misdirected if not also offered to network-perceived central players or brokers. Network members may be better able to correctly identify central actors than formal brokers
 Guldbrandsson 2012 [38] Traditional descriptive statistics, e.g., frequency counts, percentages Profession
Work organization
Geographical region
Information seeking about child health promotion Tie homophily (%)
Indegree centrality (frequency of being named)
Organization and professional field were shared in nearly half of all information seeking ties
 Doumit 2014 [34] Percentage of people nominating an actor; descriptive statistics (frequency counts, percentages) Influence by central actors Degree centralization Six individuals with high credibility influenced 85% of the network, suggesting opinion leaders have potential for supporting evidence use
Reasons for change in medical approach (proportions)
Barriers to clinical decision-making (proportions)
 D’Andreta 2013 [39] Deriving network properties to describe the network (descriptive SNA) KT model adopted Prestige within the network Degree centralization KT teams with different models of KT (i.e., focus on research dissemination vs. knowledge co-production and brokering vs. integrated research-clinical collaboration) varied in their structural properties (e.g. the prominence and control of leaders in KT processes)
Control over knowledge Betweenness centralization
Access to knowledge Closeness centralization
Alternate paths for knowledge flow that circumvent central actors Flow betweenness centralization
Organizational role (e.g., director, support staff) Core actors—dominant individuals with frequent knowledge exchange Coreness scores (core-periphery algorithm)
Predictive/explanatory studies
 Zappa 2011 [29] Descriptive SNA External communication (# visits from drug representatives); research orientation (# publications); clinical experience; hierarchical position (administrative role)
Medical specialty*
Hospital affiliation*
Colleagues with whom knowledge is discussed and transferred Network density Low network density (0.3%)
Components Multiple small components suggested lack of strong opinion leaders to drive treatment adoption. Findings suggest physicians tend to build small, closed, non-hierarchical internal, and external connections within their professional group, potentially limiting broader access to new information
Isolates tended to be clinically experienced and active in research.
Advice sharing more likely if physicians shared a medical speciality, geographic proximity but differed in research productivity or years in practice.
Exponential Random Graph models (p* models) Tendency to exchange information with a number of sources Alternating k-stars
Tendency to share knowledge within a small peer group (network closure) Alternating k-triangles; alternating independent two-paths
Tendency to interact with similar others
Prominence as a knowledge source in the network
Tie homophily/hierarchy; indegree centrality
 Yousefi-Nooraie, 2014 [37] (same study as [36]) Descriptive SNA   Relative connectedness of actors of a given role Indegree centrality
Outdegree centrality
Managers identified as key brokers in KT interventions and EIP implementation processes.
Public health professionals preferred to limit advice seeking and expert recognition to a small number of peers; advice seeking limited to own division.
Strong friendship ties a significant predictor of information seeking ties.
EIP scores not predictive of information seeking or expert recognition ties.
*Role (e.g., manager)
*Organizational division
Score on EIP implementation scale
Key individuals Degree centrality
Organizational division Tendency to connect to peers from other units E-I index
Tendency to reciprocate expert recognition and information seeking ties Tie reciprocity
Exponential random graph modeling (ERGM) *Role (e.g., manager)
*Organizational division
Score on EIP implementation scale
Tendency to connect with those with similar attributes Tie homophily
Reciprocity Tie reciprocity
Formation of information seeking and expertise-recognition ties Ties and direction of ties (in vs. out)
Tendency for network to have highly connected nodes (hubs) Alternating in-k-stars
Alternating out-k-stars
Friendship connections Ties
Multilevel logistic regression *Role (e.g., manager)
*Organizational division
Score on EIP implementation scale
Tendency to connect with those with similar attributes Tie homophily
Formation of information seeking and expertise-recognition ties Ties
Friendship connections Ties
 Tasselli, 2015 [22] Paired t test
Linear regression
*Gender
*Tenure
*Profession
*Organizational unit
*Rank (i.e., leadership role)
Ease of knowledge transfer
Perceived receipt of useful knowledge
Connectedness
Hierarchy
Network fragmentation
Individual reach
Brokerage potential
*Network size
Tie strength
Mean degree centrality
Bonacich centrality
Mean betweenness centrality
Closeness centrality
Betweenness centrality
*Degree centrality
Knowledge tends to transfer within rather than across professions; nurses’ networks were denser and more hierarchical; closeness centrality positively associated with ease of knowledge transfer; brokering positions increased access to useful knowledge
 Menchik 2010 [44] OLS regression # medical literature database searches per month
# journals read regularly
*Age
*Gender
*Tenure at hospital
*Medical school
*% clinical time
*Sub-specialization Prestige of hospital (published rankings)
Relational esteem by colleagues Indegree centrality Physicians in higher prestige hospitals were less likely to be named as advice givers. Prestige in these settings associated with medical school attended.
In lower-prestige hospitals, regularly reading a range of journals, and less time spent on clinical work increased likelihood of high esteem by colleagues
 Mascia 2014 [33] Ordinal logistic regression Self-reported frequency of EIP use
*Gender
*# patients in caseload
*area of clinical practice (e.g., asthma, urology, etc.)
*# article subscriptions
*perceptions of barriers to availability of evidence
*perceptions of difficulty applying evidence to practice
*Organizational affiliation
*Affiliation to formal groups
*Collaborative nature of actor’s medical practice
Degree of collaboration with colleagues Outdegree centrality Degree centrality directly associated with physicians’ EIP use
 Mascia 2018 [27] Exponential random graph models *Past task force involvement
*Tenure
*Gender
*Geographic distance
*Association members
*Health district
Tendency to reciprocate advice
Tendency to seek advice from an indirect tie
Tendency for local, generalized exchange of advice
*Tendency to form advice ties
*Popularity as an advice source
*Advice-seeking activity
*Brokering
Tie reciprocity
Transitivity (path closure)
Cyclic closure
*Density
*Indegree centrality
*Outdegree centrality
*Formation of non-closure structures
*Tie homophily (of attributes)
Advice ties unlikely unless reciprocated; advice ties tended to organize around clusters—driven by transitivity, not popularity; Tendency against exchange of advice in cyclic structures; positive relationship between ties and association homophily in one health authority, and between ties and district/task forces in the other; tendency toward homophily related to tenure and distance, but not gender
 Mascia 2015 [48] (same data as [31], [30] and [32]) Multiple regression-quadratic assignment procedure (MR-QAP) Age
Gender
Tenure/seniority
Frequency of collaboration
Similarity of professional role, institution and geographical location
Tie strength
Tie homophily
Ties more likely if specialization, institution were the same between individuals; less likely if similar roles, greater difference in time since graduation and further geographic distance; professional homophily better predictor than institutional homophily
 Mascia 2013 [32] (same data as [48], [30] and [31]) Descriptive SNA Age*
Gender*
Hospital tenure*
Tenure in health authority*
Managerial role*
Geographical distance from colleagues*
Affiliation with other organizations*
Self-reported EIP adoption (i.e. frequency of database searching)
Connectedness Whole network density Low density (5.7%) observed
OLS regression Network authority (i.e., importancerelevant and popular) Hubs and authorities centrality The most active EIP practitioners likely to be found at network periphery (i.e. least central)
Degree of coreness in the network Network coreness score (degree centrality and core-periphery position)
 Mascia 2011b [31] (same data as [48], [32] and [30]) Ordinal logistic regression Tendency to adopt EIP (self-reported frequency of peer-reviewed research use)
Age*
Gender*
Tenure in health authority/organization*
Managerial role*
# publications*
Perceived access to evidence*
Hospital affiliation*
Extent to which a given tie is redundant because of concurrent ties with another alter Ego-network constraint Physicians with greater network constraint (i.e., many redundant ties) reported decreased EIP adoption. May be related to information bias—tendency of physicians to interpret the information in a way that is congruent with their previous knowledge or opinion.
High degree centrality associated with EIP use.
Individual’s network size* Total # of ties in ego-network (indegree + outdegree centrality)*
 Mascia 2011a [30] (same data as [32, 48]) Descriptive SNA Average number of advice exchange colleagues Mean ego-network density Advice sharing most likely when physicians shared a medical specialty, geographic proximity, similar attitudes toward EIP, or had co-authored publications.
Collaboration less likely when actors held similar managerial roles, or were at different hospitals/clinical/geographical areas.
Tendency for colleagues to both give and receive advice with one another Tie reciprocity
Multiple regression quadratic assignment procedures (MR-QAP analysis) Advice exchange among pairs of physicians Ego-network ties
Similarity between pairs of tied actors in:
Geographical distances
Gender*
Age*
Medical specialization*
Clinical experience*
Tenure in health authority/organization*
Managerial role*
# publications*
Co-authorship*
EIP adoption (self-reported frequency of peer-reviewed research use)*
Tie homophily
 Paul 2015 [35] Extended p2 model with Bayesian modeling and estimation *Age
Patient age
Patient sex
Patient race
Patient health status
Patient intensity of care
Relative # shared patients
*Same gender
*Same specialty
*Same location
Reciprocity
Social dependence (clustering)
Density
Tie homophily
Tie reciprocity
Alternating independent two-paths
Transitive triads
Alternating k-stars (two-stars)
Low network density (0.10) observed.
Triadic clustering higher than chance.
Ties not associated with gender/specialty homophily.
Location positively associated with ties.
Complementary expertise positively associated with patient sharing.
Transitivity may account for reciprocity
Gender
Clinic
% female patients
Self-identify as expert
# clinics per week
# years practicing in the city
Tenure at hospital
Years clinical experience
Location of training
Involvement in influential discussions Whole network density Low density (0.154) observed; reciprocity more likely than not—may be an artifact of transitivity; high triadic clustering observed; same clinic and gender, expert, higher clinical caseload increased tendency for tie formation
 Keating 2007 [45] P2 logistic regression analysis Self-identified experts seen as more influential; no relationship between # years in practice or location of work or training.
Clustering with respect to EIP knowledge exchange observed between those with greater # of patients and higher frequency of clinical sessions.
High reciprocity observed in the absence of opinion leaders with high centrality.
Being perceived as influential Indegree centrality
Perceiving others as influential (information seeking) Outdegree centrality
Reciprocity Tie reciprocity
 Heijmans 2017 [26] Paired sample t tests/Wilcoxon tests
Logistic multi-level analyses
*Patient age
*Patient sex
*Patient group
*Patient illness status
*Treatment/control group for parallel randomized controlled trial
Connectedness
Frequency of contact
Influence of coordinator
Similarity in attitudes related to treatment goals
Presence of opinion leader
*Network size
Density
Tie strength
Degree centrality
Homophily (E-I index)
% of possible in-degree ties
Low density (0.37 and 0.38) observed.
Most ties between those who did not value achieving treatment goals.
General practitioner most likely named as opinion leader.
Nurse performance associated with consistently identified opinion leader.
Lack of tie homophily for positive attitudes associated with poor clinical outcomes
 Friedkin 2010 [25] Random intercept multi-level event history model Professional age
Chief or honorary position (yes/no)
Number of journals read
Value keeping up with scientific developments
Physicians’ adoption of a new antibiotic (i.e., prescribing behavior)
Marketing patterns of drug companies
Proportion of previous adopters at a given time (“internal contagion”)
Influence of advisors/discussion partners Contact network role (CNET)—a summative measure of 4 measures of structural cohesion and structural equivalence; position in the medical advice network Cohesion and structural equivalence were correlated, and may be useful in combination to improve reliability in the evaluation of network structures across settings
 Di Vincenzo 2017 [28] Ordinary Least Squares regression # publications
*Tenure
*Managerial role
*Geographic distance
*Hospital affiliation
*# publications from same-specialty colleagues
Dependence on others/access to new information
Relative productivity among ego-network colleagues
Same role
Same specialty
Ego-network constraint
Euclidian distance
*Ego-network size
Young employees appeared to have more redundant networks (greater need for advice).
Hospital affiliation (i.e., context) influenced constraint.
Constraint negatively associated with ego-network size and relative productivity (mediated by professional group membership), positively with productivity, Euclidean distance, role/specialty homophily (augments impact of productivity on prestige)
 Burt 1987 [24] Ordinary least squares regression with likelihood-ratio chi-squared test Timing of adoption
Relative timing of adoption within the network
Professional age
Contact with drug company
Number of journals read
Number of house calls vs. office visits
Value keeping up with scientific developments
Position in the medical advice/discussion network Structural equivalence Adoption by others in equivalent positions within the network was a stronger predictor of adoption than adoption by those in an individual’s advice or discussion networks.
Early adopters tended to participate in a range of EIP behaviors.
Adoption by prominent physicians seen to be related to their desire to avoid being late adopters.
Influence of advisors/discussion partners Structural cohesion
 Ankem 2003 [23] Chi-square statistics Preferred information source
Timing of awareness of the intervention
Timing of intervention adoption
Clinical networks were most prominent in fostering awareness and adoption of a clinical intervention, but research and social networks also likely to influence these processes.
Early adopters tended to rely on journals and conferences for information informing practice change; late adopters to a greater extent by network contacts.
Factor analysis Specialization
Hospital
City
Timing of adoption
Frequency of communication with colleagues
Types of relations within the network (e.g., clinical, research, leisure) Ties
Groupings of information exchange relations Cliques
Longitudinal evaluative studies
 Racko 2018 [47] Ordinary least squares regression Professional status (ranking)
*Professional role
*Gender
*Education
*Organizational status
Research collaboration
Joint decision-making
Connectedness to high-status individuals
Connectedness to knowledge brokers
Connectedness to unfamiliar peers
*Intra-professional whole network size
Ego-network size via tie heterophily
Tie strength
Mean status score of ego-network relative to whole network
% of possible ties
*Tie homophily
Higher social status associated with more research collaboration at all time points, and joint decision-making in early phases.
Higher-status ties with peers, ties to formal knowledge brokers and ties to unfamiliar peers inconsistently predicted knowledge exchange, research collaboration and joint decision-making over time.
Formal knowledge broker presence may facilitate interprofessional networking
 Bunger 2016 [43] Paired t tests; one-way analysis of variance; descriptive SNA Role (faculty expert, internal colleague, external peer, private practitioner, other) Connectedness
Lack of advice seeking/sharing
Reciprocity
Similarity in connectedness
Tendency for sub-groups to form
Same institution
Frequency of communication
Density
Isolates
Tie reciprocity
Indegree centralization
Clustering
Tie homophily
Tie strength
Ego-network size decreased, more markedly for senior leaders.
Exposure to private practitioners and “others” decreased; exposure to experts increased.
Substantial turnover in dyads was reported, with greater tie density around central core of experts.
Reciprocity and tie heterophily increased over time
  1. KT knowledge translation, EIP evidence informed practice. Where indicated by the article’s author, italic text = dependent variable; * = covariate