Use of social network analysis in health research: a scoping review protocol


Social connections are known to influence health.1 People with many supportive social connections tend to be healthier and live longer than people who have fewer supportive social connections, while social isolation, or the absence of supportive social connections, is associated with the deterioration of physical and psychological health, and even death.2–5 These associations hold even when accounting for socioeconomic status and health practices.6 Additionally, having a low quantity of supportive social connections is associated with the development or worsening of medical conditions, such as atherosclerosis, hypertension, cardiovascular disease and cancer, potentially through chronic inflammation and changes to autonomic regulation and immune responses.7–13 Unsupportive social connections can also have adverse effects on health due to emotional stress, which can then lead to poor health habits, psychological distress and negative physiological responses (eg, increased heart rate and blood pressure), all of which are detrimental to health over time.14 The health of individuals is therefore connected to the people around them.15

Social networks can influence health via five pathways.15 16 First, networks can provide social support, to meet the needs of the individual. Dyadic relationships can provide informational, instrumental (ie, aid and assistance with tangible needs), appraisal (ie, help with decision-making) and/or emotional support; this support can be enhanced or hindered by the overall network structure.17 In addition to the tangible aid and resources that are provided, social support—either perceived or actual—also has direct effects on mental health, well-being and feelings of self-efficacy.18–20 Social support may also act as a buffer to stress.16 19 The second pathway by which social networks influence health, and in particular health behaviours such as alcohol and cigarette use, physical activity, food intake patterns and healthcare utilisation, is through social influence.16 21 That is, the attitudes and behaviour of individuals are guided and altered in response to other network members.22 23 Social influence is difficult to disentangle from social selection from an empirical standpoint. That is, similarities in behaviour may be due to influences within a network, or alternatively, they may reflect the known phenomenon where individuals tend to form close connections with others who are like them.22 24 The third pathway is through the promotion of social engagement and participation. Individuals derive a sense of identity, value and meaning through the roles they play (eg, parental roles, community roles, professional roles, etc) in their networks, and the opportunities for participation in social contexts.16 The fourth pathway by which networks affect health is through transmission of communicable diseases through person-to-person contact. Finally, social networks overlap, resulting in differential access to resources and opportunities (eg, finances, information and jobs).15 16 An individual’s structural position can result in differential health outcomes, similar to the inequities that stem from differences in social status.16

There has been an explosion of literature in the area of social networks and health. In their bibliometric analysis, Chapman et al found that the number of studies that examine social networks and health has sextupled since 2000.25 Similarly, the value of grants and contracts in this topic area, as awarded by the National Science Foundation and the National Institutes of Health, has increased 10-fold.25 A turning point in the field was the HIV epidemic, where there was an urgent need to better understand its spread.25 The exponential rise in the number of studies since then that examine social networks and health appears to reflect a widespread understanding that an individual’s health cannot be isolated from his or her social networks and context. There is, however, significant heterogeneity in what aspect of, and how, social networks are being studied. For example, many health research studies use proxies for social connectedness such as marital status or living alone status (as these variables tend to be commonly included in health surveys), without considering the quality of those social connections, and without further exploring the broader social network and their characteristics.16 26 These proxy measures do little to describe the structure, quantity, quality or characteristics of social connections within which individuals are embedded. Another common approach in health research is to focus on social support measures and their effects on health. Individuals are asked about perceived, or received, social support (for example, through questions that ask about the availability of people who provide emotional support, informational support and/or assistance with daily tasks, with either binary or a Likert scale of responses).27 28 While important, social support measures do not assess the structure of social networks and represent only one of many different mechanisms by which social networks influence health.17 23

Social network analysis

Social network analysis (SNA) is a methodological tool, developed in the 1930s by social psychologists, used to study the structure and characteristics of the social networks within which individuals are embedded.16 29 It has evolved over the past 100 years and has been used by researchers in many social science disciplines to analyse how structures of relationships impact social life.29 30 SNA has the following key properties3 30 31: (1) it relies on empirical relational data (ie, data on actors (nodes) and the connections (ties) between them); (2) it uses mathematical models and graph theory to examine the structure of relationships within which individual actors are embedded; and (3) it models social action at both the group and the individual level arising from the opportunities and constraints determined by the system of relationships. The premise of SNA is that social ties are both drivers and consequences of human behaviour, and are therefore the object of study.15 16 23 32 Social networks are comprised of nodes, representing the members within a network, connected by ties, representing relations among those individuals.33 There are two types of SNA: egocentric network analysis and whole network analysis. Egocentric network analysis describes the characteristics of an individual’s (ie, the ‘egos’) personal network, while whole network analysis examines the structure of relationships among all the individuals in a bounded group, such as a school or classroom.3

In egocentric network analysis, a list of ‘alters’ (ie, nodes) to whom the ego is connected, is obtained through a name generator. Name generator questions ask for a list of alters based on role relations (eg, friends or family), affect (eg, people to whom the ego feels close), interaction (eg, people with whom the ego has been in contact) or exchange (eg, people who provide social and/or financial support).34 These are followed by name interpreters, where the ego is asked questions about the characteristics of each named alter.35 Analyses of these data involve constructing measures that describe these egocentric networks. Such measures include network size, network density (ie, how tightly knit the network is), the strength of relationships (ie, the intensity and duration of relationships between ego and alter), network function (ie, the resources and/or support provided through the network) and the diversity of relations within the network (‘heterogeneity’).23 36 In whole network analysis, the network boundary is determined a priori and network members are known, for example, through membership lists or rosters.37 Each network member is surveyed, to identify the other network members with whom they are connected and/or affiliated; attributes of each member are obtained through surveying the network members themselves. Variables are constructed at the individual and network levels. Individual-level measures include the number of ties to other network members (‘degree’), types of relationships, and the strength and diversity of relationships. Network-level measures include but are not limited to: density (representing how tightly knit or ‘glued’ together the network is), reciprocity (ie, the proportion of network ties that are reciprocated), isolates (ie, nodes with no ties to other network members), centralisation (or the extent to which the network ties are focused on one node or a set of nodes), cliques and equivalence (ie, sets of nodes that have the same pattern of ties and therefore occupy the same position in the network).33 38 The constructed measures can then be included in statistical models to explore associations between individual and/or network-level measures, and outcomes.33 39

Study rationale

In medicine and health research, there has traditionally been a dichotomy between the individual and the context in which the individual is situated—such as in their relationships with others.40 As such, epidemiology of diseases has historically focused on individual-level traditional risk and protective factors—such as biological markers, genetics, lifestyle and health behaviours, and psychological conditions.41 While criticisms of this individualistic focus abound, attempts to develop and use different approaches in medicine and research have lagged behind.42 The use and adoption of methods, like SNA, that frame issues of health and wellness differently, has the potential to offer new insights and solutions to clinical and healthcare delivery problems,42 by more holistically considering ‘different levels of change’ beyond the individual.41 We seek to examine the extent to which SNA has transcended the boundaries of its disciplines of origin in the social sciences, into health research. For example, while Chapman et al have clearly shown an explosion of publications at this intersection,25 it remains unclear whether these studies use SNA tools (which were developed specifically to interrogate the nature and characteristics of social networks), or whether they suffer from the known problem of conflation of constructs like social support, social capital and social integration.15 43 Many studies that report the impact of ‘social networks’ on health outcomes do not use SNA methods but rather use self-reported network size (without probing the network and its structure),44 45 social support,46 marital status47 48 and/or household members47 as proxies.

We will therefore undertake a scoping review to map the use of SNA as a data collection and analytical method in health research. More specifically, the scoping review will examine how SNA has been used to study associations across social networks and individual health and well-being (including both physical and psychological health), health knowledge, health engagement, health service use and health behaviours. Scoping reviews are a knowledge synthesis approach that aims to uncover the volume, range, reach and coverage of a body of literature on a specific topic.49 They differ from systematic reviews, another type of knowledge synthesis, in their objectives. Systematic reviews seek to answer clinical or epidemiological questions and are conducted to fill gaps in knowledge.50 Systematic reviews are used to establish the effectiveness of an intervention or associations between specific exposures and outcomes. On the other hand, scoping reviews do not seek to provide an answer to a question, but rather, aim to create a map of the existing literature.49 They are used to provide clarity to the concepts and definitions used in literature, examine the way in which research is conducted in a specific field or on a specific topic, and uncover knowledge gaps.49 A scoping review, therefore, is well suited as a research method to address our research question, of mapping the ways in which SNA has been used in health research. This scoping review can identify areas (eg, specific populations and specific health outcomes) where there has been a plethora of SNA research warranting future systematic reviews. It can also identify areas within health research where the use of SNA is scarce, highlighting topics, populations or outcomes for future study.

This scoping review will be limited to studies that use SNA in exploring network components and their associations with non-communicable diseases and health and well-being outcomes, for three reasons. The first is feasibility, given the large volume of studies anticipated, based on Chapman et al’s bibliometric study on this topic.25 Second, the use of SNA in understanding disease transmission of communicable diseases (such as sexually transmitted infections) is well established; its application to HIV was in fact one of the catalysts, as previously mentioned, to its broader uptake in health research.25 Third, SNA in health research has shifted from focusing on communicable diseases to focusing on non-communicable diseases and their risk factors; SNA is now being applied much more frequently to the latter conditions than the former ones.51

Methods and analysis

The scoping review will be informed by the framework developed by Arksey and O’Malley52 for conducting scoping reviews, as well as the additional recommendations made by Levac et al.53 Arksey and O’Malley’s framework recommends that the review process be organised into the following five steps: identifying the research question; identifying relevant studies; study selection; charting the data; and collating, summarising and reporting the results.52 The reporting of this review will adhere to the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews.54

Patient and public involvement

No patients will be involved.

Step 1: identifying the research question

A preliminary search of the literature identified a gap related to SNA and how it has been used to study the relationship between social networks and individual well-being and health outcomes. This led to the development of the research question that will guide this scoping review: how have social network analytical tools been used to study the associations between social networks and individual patient health? In this case, SNA is defined as a data analysis technique that uses either an egocentric or whole network analysis approach. For egocentric network analysis, we will include studies that involve peer nomination (ie, use of a name generator) and the collection of one or more characteristics of alters (ie, use of name interpreter(s)).

Step 2: identifying relevant studies

A search strategy will be constructed through consultation with an academic librarian (JW). The main concepts from the research question will be used for a preliminary search in Google Scholar. Additionally, the lead authors will provide the librarian with key studies that will be text-mined for relevant terms. These key studies will include a variety of populations (across different countries and age groups) and health outcomes.55–58 Key studies will be searched in Ovid MEDLINE for appropriate subject headings. In consultation with team members, the librarian (JW) will construct a pilot search strategy. A title/abstract/keyword search will be conducted in Ovid MEDLINE against the known seed/key studies. Table 1 lists example keywords and terms relating to social networks that will be used, with the full search strategy detailed in online supplemental appendix A.

Supplemental material

Table 1

Search terms relating to social network analysis

Due to a significant number of irrelevant articles surrounding communicable diseases using this search strategy, we will exclude records with these terms in either the title or keyword fields. Table 2 lists the terms related to communicable diseases.

Table 2

Search terms relating to communicable diseases

Of note, the search strategy will not include terms that relate to health-related outcomes of interest (outside of excluding communicable diseases). Prior literature has shown that the inclusion of outcome concepts in a search strategy reduces the recall and sensitivity of a search strategy.59 60 This problem is further exacerbated when only generic health terms (for example, ‘morbidity’ or ‘health status’) or specific health terms (eg, specific diseases or conditions such as ‘diabetes mellitus’) are used.61 Because the objective of this scoping review is to examine and map the use of SNA in health research, the outcomes of interest are very broad, including: physical health and well-being, psychological health and well-being, healthcare engagement, health knowledge, health behaviours, healthcare access and use, disease prevalence and outcomes (spanning every organ system), and mortality. It will be impossible for a search strategy to be sufficiently comprehensive, to capture all possible generic and specific terms relating to this broad range of outcomes. In keeping with recommendations to minimise the number of elements in a search strategy62—and in particular outcome elements63—our search strategy will entail searching for SNA terms in health databases without specifying health outcomes.

The search strategy will first be created in Medline (Ovid), then translated and adapted for the databases: (1) EMBASE (Ovid), (2) APA PsycInfo (Ovid) and (3) CINAHL (EBSCO). A search will be completed in April 2024. No date filters will be applied to the search. However, animal-only studies will be excluded. The current version of the search strategy including limits and filters, for all databases, is included in online supplemental appendix A.

Step 3: study selection

The criteria that will be used to determine which studies to include are as follows:

  • Studies that employ SNA as a data collection and/or analysis technique, as defined above. Of note, studies that elicit only the number of friends or other social contacts, without collection of any information about these social contacts, are not considered to be SNA and are therefore not included in the scoping review.

  • Studies that explore the social networks of individuals in whom the health outcome is measured.

  • Studies must include the exploration of non-communicable health outcomes. Examples include self-rated health or other global measures of health (including measures of physical health, mental health and well-being), health practices (eg, physical activity, dietary patterns, smoking, alcohol use, substance use), sexual and reproductive health, healthcare-seeking behaviours (eg, medication adherence, acute care use, attachment to a primary care provider), health knowledge, health beliefs, healthcare engagement, non-communicable disease prevalence and mortality.

The criteria that will be used to exclude studies are as follows:

  • Studies that explore the social networks of organisations or healthcare providers, rather than the social networks of the individual about whom the health outcome is measured or reported.

  • Studies that describe or use data analysis techniques other than SNA (eg, using proxies for social networks/social support that do not include peer nomination (such as marital status or living alone status), or studies where study participants report the number of social contacts but where no other information about each social contact is collected).

  • Studies that focus exclusively on online social networks (eg, social media, online forums, online support groups).

  • Studies related to prevention, transmission or outcomes of communicable diseases.

  • Non-English studies, for feasibility purposes.

We will not limit studies based on the study population or country in which the SNA is conducted. Studies in paediatric and adult populations will be included. The reasons for excluding SNA studies that focus solely on social media and online networks are twofold. First, we anticipate a very large number of articles, given the broad populations and outcomes of interest, and for feasibility purposes, we have needed to narrow the research objective to in-person and/or offline social networks only. Second, there are likely inherent differences in online and offline social networks. Individuals use health-related social networking sites and online networks primarily for information seeking, connection with others who share a similar lived experience while being able to maintain some emotional distance and interacting with health professionals64; this differs from in-person networks, which individuals go to more for emotional and tangible or instrumental support. Friends met on online networks vary from friends met in person in other important ways. They tend to have less similarity in terms of age, gender and place of residence,65 and the network ties more commonly arise spontaneously—that is, without common acquaintances or affiliations.66 The social patterns and interactions among individuals and their online network contacts are also different—with entire relationships built on text-based interactions.66 Therefore, while online social networks are an important area of study, they appear to be inherently different from the study of offline social networks, and are therefore excluded from this scoping review.

For the first step of the screening process, after removing duplicate articles, two reviewers will independently assess the titles and abstracts of the studies to determine whether they meet the inclusion criteria. Any studies that do not meet the inclusion criteria will be excluded from the review. Studies that either one of the two reviewers feels are potentially relevant will be included in the full-text review, to ensure that no article is prematurely excluded at this stage. During the second step of the screening process, two reviewers will independently review the full texts of the studies to ensure they meet the inclusion criteria. Conflicts will be resolved by third and fourth reviewers with expertise in SNA (JG) and health outcomes (KLT). The number of studies included in each step of the screening process will be reported using the Preferred Reporting Items for Systematic reviews and Meta-Analyses diagram.67

Step 4: charting the data

A data charting document (online supplemental appendix B) will be created to extract data from the studies in the review. This document will include information about the authors, year of publication, study location, study population characteristics, outcomes of interest to this scoping review, and the scales and measures used for each outcome. Data about the social network analytical method will also be extracted, including whether studies used egocentric versus whole networks, the name generator used (in egocentric network studies) or the relationship being explored, the maximum number of peer nominations allowed, the lookback period used, whether (and which) alter attributes were collected, and whether alter-to-alter tie data were collected. Data extraction will be performed by at least one reviewer, with a second reviewer separately checking and confirming the inputted data. Disagreements in data extraction will be resolved through a consensus, and through the input of reviewers with content and methods expertise (KLT, JG).

Supplemental material

Step 5: collating, summarising and reporting results

The results of the review will be presented in the form of figures and tables and will include descriptive numerical summaries. The numerical summary will include information about the number of studies included in the review, where the studies were conducted, when they were published and characteristics of the populations, such as the sample sizes and mean age. It will also include characteristics of the SNA conducted in these studies, including the number that are whole network studies versus egocentric network studies, the data sources used and the attributes of the social connections that are collected and analysed. Results will be synthesised in text, as well as through tables and figures.

Ethics and dissemination

This review does not require ethics approval. Data will be extracted from published material. Once the scoping review is complete, an article will be written to convey the findings of this review, and it will be submitted for publication in a peer-reviewed journal. We anticipate the results of this review will map out the ways in which SNA has been used in health research. Specifically, this scoping review will identify areas of potential saturation where SNA has been heavily used, opportunities for future systematic reviews (where there is a large body of primary research studies requiring synthesis) and health research gaps (eg, the health outcomes where SNA has been minimally used). The scoping review will also shed light on characteristics of SNA that have been used (eg, whether egocentric networks vs whole networks are used and in what settings, and whether a broad range of social network characteristics are captured and analysed), which will serve to inform the conduct of future SNA studies in health research.

This post was originally published on