A methodological approach to the analysis of egocentric social networks in public health research: a practical example

Qualitative research is not a new paradigm in public health. It has been recognized decades ago that public health problems result from complex social, economic, political, genetic and environmental causes, and that there different methods that need to address them (1). Many of these methods evolved in social sciences and offer effective tools for dealing with health challenges that public health research is focused on (1-3). Disease prevention and health promotion is only one of the topics where more individually based and contextualized knowledge is required (3).

In Slovenia, the use of qualitative research in medicine increased significantly in the last years. Studies were performed either with qualitative methods only (4-9), or by means of an integrated approach, together with quantitative measures (9, 11).

The social network analysis (SNA) is defined as a set of methods used for mapping, measuring and analysing social relationships between people, groups and organisations (12). It is a research approach uniquely suited to describing, exploring and understanding structural and relational aspects of health (13, 14). It incorporates the social context to explain individual or group outcomes (15). Historically, the SNA has been used mainly in social sciences, but its use has grown rapidly in recent years into many other areas, including public health (12, 14). Research on social networks in public health focuses less on the transmission of specific tangible elements and more on how social structures and relationships influence health and health-related behaviours (16). Much of what is studied in public health is relational by itself. The SNA has been used primarily to study the transmission of diseases and information, diffusion of innovations, the role of social support and social capital, the influence of personal and social networks on health behaviours and the inter-organisational structure of health systems (13). There have been discussions whether the SNA is a quantitative or qualitative method. Authors describe the evolution from the qualitative approach in the early beginnings of the SNA, to the development of advanced mathematical methods and computer programs that led to a predominance of quantitative approaches, and finally, the mixed approach which combines both (17-19). Quantitative approaches map and measure networks by simplifying social relations into numerical data, where ties are either absent or present. Qualitative approaches, on the other hand, enable to consider issues relating to characteristics of social ties (19).

There are two types of social networks, namely: complete and egocentric social networks. Complete networks consist of a group of units (people, organisations, etc.) and relationships between them. In egocentric social networks (ESN), the person of interest is referred to as the ego. The people he/she is appointing to his/her network – relatives, friends, advisors, etc., are referred to as alteri (15, 20, 21). While the sociocentric approach is used to study complete social networks, the egocentric approach is gaining popularity because of its focus on individuals, groups and communities (15). Personal relationships and networks, like family relations, friendships, or neighbourhoodrelationships, are essential for social integration. Although the ESN studies have a long history in sociology, their use in health studies is rare (22).

With the present study, we aimed to contribute to the consolidation of the SNA in public health research. Our goal was to demonstrate the analysis and use of computer programmes for the ESN analysis on a practical example.

The present study is part of a larger study on social networks and social support for exercise among participants of healthy lifestyle education workshops ‘I’m moving’, in the frame of the National Program of Primary Prevention of Cardiovascular Disease.

2.2

The Visualisation of the ESN

Three computer programs for the SNA were analysed. The document analysis approach was used for the qualitative analysis of programs’ characteristics. Document analysis is a qualitative approach where documents are interpreted by the researcher. Analysing documents, or in the present example, computer programs, incorporates the coding content into themes, similar to focus groups or interview analyses (24, 25).

The programs Egonet.QF (26) and E-net (27) were primary designed for the analysis of the ESN. The primary purpose of the tool Pajek (28) is the analysis of large social networks, but the analysis of the ESN is possible as well. The following themes, or in the present study, program characteristics, were used for the analysis: the availability of the program and manuals or instructions, preparation of input data, display of different strengths and elements of ties, calculation of numerical measures, and the export of network figures.

2.2.1

Egonet.QF

In Egonet.QF, no preparation of data is needed, as the data about the ego and alteri are entered into the program directly in order to create a network card. A network card is created by adding alteri to the ego. First, the number of circles is selected. The circles represent the closeness of alteri to the ego. In our example, we used the circles as a measure for the strength of ties. In the analysis of the overall support, the closest circle means the highest value of strength – 4. In the analysis of separate ties, where the value is only 0 and 1, we used only one circle. As it is not possible to show different elements of ties in one network card, we created five different cards: four for each different element and one for the overall support. Different sectors of the network card can be selected. The sectors can be specified by different means. In our example, we divided the sectors according to the ties of alteri to the ego (family, relatives, friends, workshop members or others). The nodes – alteri–are placed manually into the specified sector and circle. Additionally, information about alteri can be added: the ID, label, role and sex. As a result, the ego is presented as the centre of the network card with its alteri in different circles and sectors. All attributes of the ego and alteri can be seen in a separate window. Figures of the networks can be exported as digital image files (JPEG).

2.2.2

E-net

In the program E-net, data can be imported in two formats, row-wise as a single text in a raster image file (VNA) format, or column-wise as an Excell sheet. According to the way we organised our data, the row-wise format has been more convenient. The rows represent the collected egos and alteri, while the columns represent their attributes. The VNA file can be created by copying-and-pasting Excell matrices into a text editor and saving the document in the VNA format. In the VNA file, two kinds of data must be identified by an asterisk and matrix title (*ego data and *alter data) (Figure 1). After importing the VNA file in E-net, the data about ego and alteri are shown in a table. The visualisation of the network shows the ego in the centre, connected with the alteri. The lines between the ego and alteri do not convey any information about the type or strength of the connection. Information about each alter (age, sex, relationship, strength of all different elements) is shown by clicking on the respective alter in the network. Figures of the networks cannot be exported.

Figure 1

2.2.3

Pajek

The program Pajek works with specific plain text files, called Pajek files. They can be created from text files by using the program txt2Pajek (29). For demonstration, we used only the connection ‘overall support’. Figure 2 shows how data are entered into a text file. The label ‘from’ presents the ego, the label ‘to’ different alteri, and the label ‘weight’ the strength of the tie, in our case the variable ‘overall support’. In the next step, the text file is imported in txt2Pajek. The Pajek file is created and saved automatically in the same folder as the text file. By choosing the option ‘drawing network’, the network is shown with the ego and its connections to alteri. The strength of the connections is demonstrated by different sizes of the lines. Figures of the networks can be exported as JPEG files.

Figure 2

2.5

The Preparation of the Data for the Analysis

Each of the elements of the tie has the value 0 or 1, considering the answer to the question. In questions 1, 2 and 4, the answer ‘yes’ has the value 1 and the answer ‘no’ has the value 0. In question 3, the answer ‘no’ has the value 1 and the answer ‘yes’ has the value 0. Each of the four elements can be analysed and visualized separately. For an overall view on the strength of social support, the variable ‘overall support’ was created, being a summary of separate elements’ values, and had values from 0 to 4. Figure 3 shows how to properly organize data on the ego and alteri for entering the SNA.

Figure 3

Results of the numerical analysis of the ESN give an insight into the network’s main characteristics regarding the structure and content. We are able to obtain quantitative and qualitative data about the ESN’s size, composition and structure (23). Degree is a measure very easy to compute, but informative in many applications (20). It gives the first overview of the individual’s social network. Regardless of the element or strength of the ties, it shows how many people are connected to the ego, or in our example, how many people support the individual. The average strength of ties has an added value because it summarizes information on the strength of separate ties. The ‘gap’ between the degree and average strength is valuable information, especially in exploring social support. Together with the strength of separate ties, it shows the eventual potential for increasing the support in a very detailed way. The network’s composition shows the characteristics of people within the ESN. It tells what are the preferences or possibilities of the individual at choosing persons into her/his network.

The program Egonet.QF is suitable for creating network diagrams, which make use of concentric circles (26). The use of circles is an excellent tool to present the strength of ego-to-alter ties. We used different sectors of the network to present different relationships between the ego and alteri. The division based on other attributes is possible too, so users can choose the option that is best suited to the purpose of their study. The visualization with different circles and sectors is attractive and gives a good overview on the ties between the egos and different alters. E-net is a program designed specifically for the analysis of personal networks (27). The VNA file needed for the import of data is easy to create and allows importing data of more than one ego and her/his alters in one file. This is useful when analysing more ESNs. The visualization is very basic and shows only the ego, alteri and connection between them. Attributes of the actors are shown in the window beside the picture, by clicking on the respective node. The attributes, in our case, included strengths of separate elements of support by every alter, therefore this option gives a good overview of the source of support for the individual. The attributes can be displayed only separately for each actor. Pajek is a program for analysis and visualization of large social networks (28). Despite that, a visualisation of an ESN is possible too. Two steps are required in the preparation of data, including the use of an additional program for the transformation of data (29). The visualization is very basic and shows only the ego, alteri and connections between them. The only information about the ties is their strength, shown in different sizes of the lines. The analysis of ESNs could be performed in Pajek as well, but only if the ESN is analysed within an analysis of a complete social network.

Despite offering a selection of numerical analysis tools, none of the programs calculates the degree or average strength of the network, and only E-net calculates homophily and homogeneity. This is not considered as an important disadvantage, as the numerical analysis of the ESN is easily performed in other widely accessible programmes, such as Excell.

The analysis of the ESN provides a feasible way of obtaining network information, but it has several limitations. Demonstration of one single case limits representativeness. Relying on a single case is considered acceptable for a study as ours, as the purpose was to demonstrate the use of a particular methodology (17). Another limitation is related to the bias of recall. Respondents are often asked to recall behaviours that took place over a broad period of time. Recent contacts are more likely reported into the network, and forgetting to name people is another potentially significant problem (14, 15). In the analysis of the ESN, the example relies on the accuracy of reports from the focal actor, the ego. As we do not observe the status of relationships from the perspective of alteri, the perceptions of alteri could be different (17, 22). In our study, we focused only on ties between the ego and alteri, and did not explore the ties between alteri. Those ties can potentially affect the ego. As the purpose of our study is to demonstrate the use of the ESN analysis with social support, we limited our example only on ego-to-alter relationships.

With the analysis of different methodological approaches to the analysis of ESNs, we contributed significantly to public health research. We pointed out or introduced measures most suitable for the analysis of the ESN related to public health issues. The approaches are applicable to a wide range of ESN types and questions related to ESN research, including personal networks and social support, professional networks atthe personal level or networks between organizations.

Using the ESN analysis helps us learn about how individuals correspond with their social networks (31). Data obtained from the numerical network analysis, at the population level, can give an overview of social networks and social support in populations of interest. Data can serve for the monitoring and evaluation of programs, where social networks and social support present important components for program success. Especially health promotion or health education programs, where the focus lies on behaviour change, depend highly on the social support of participants. It is important to keep in mind that individuals are involved in a social environment and in a series of social relationships. Improvement and maintenance of health is influenced not only by the individual’s behaviour, but also by the behaviours of others in the network (16). The methodological approaches demonstrated in the present study can be used for assessing for any kind of behaviour change. Public health experts working in those programs can benefit from information on their participants’ social networks. A network card obtained with one of the visualization tools can not only serve as a presentation of results, but can also help the individual as cognitive support in network exploration, to help keeping the overview of relationships (32). The existing and potential sources of support or barriers can be identified, and can help maintaining or improving social support. Besides social support, many other relations can be explored at the individual level, such as seeking advice or information from a professional. Studies on the ESN can be performed autonomously, when the focus is on a small sample or specific environment. When gathering personal network information on large scale, ESN studies can be embedded within population surveys to enable population-representative data (22). When performing the ESN analysis at the population level, the numerical analysis is preferred, as it is more suitable for handling large amounts of units, and as the data can be further analysed by using statistical methods (eg. a comparison between egos or ego-alter relations, time dynamics, etc.).

We demonstrated the analysis of ESNs, including measures and visualization, and compared three different computer programs for the analysis of ESNs. Results show that both numerical analysis and visualization have their benefits. The decision as to which approach to use is dependent on the purpose of the ESN analysis. While the numerical analysis can be used in large scale population-based studies, visualization of personal networks can help health professionals at creating, performing and evaluating preventive programs, especially if focused on behaviour change. The analysis of social networks, including ESNs, is used widely in the area of public health, and there is an enormous potential in the use of it.