Social assortativity and contagious processes in modern Britain.

Lead Research Organisation: University of Liverpool
Department Name: Institute of Infection and Global Health


Assortativity is the property describing when individuals tend to mix with others who have similar (social or demographic) properties to themselves. As such it is a key structural property of social networks. The extent to which interactions between individuals are assortative has important implications for contagious processes which operate on such interaction networks, including the spread of infectious diseases and the propagation of ideas or beliefs. The spread of infection provides a good indicator of assortativity and network structure as data on disease is regularly recorded across a range of scales. In addition, accurately reflecting assortativity in models of infection is necessary if such models are to robustly inform public-health policy.

Recent work has shown that there is strong assortative mixing by age in the UK population: most of the social encounters made by children are with similarly aged children, while adults tend to interact with a broad age range of other adults. This insight into social mixing helps explain some of the patterns of infection observed in past epidemics, including the early stages of the 2009 influenza pandemic. However, age assortative mixing alone does not explain the inequalities and clustering of infection often seen in epidemics - a better quantitative understanding of how interactions within modern society are structured by age, gender, socioeconomic status, education and ethnicity would improve models of the spread of infection and beliefs.

There is a wealth of information on the assortative nature of modern British society collected by a range of social and economic surveys but, to date, this information has not been incorporated into mathematical models of social networks. Current mathematical models of infection, routinely used to inform health planning and predictions, do not yet include realistic social structures. There is, therefore, an opportunity to exploit these datasets and considerably improve the realism of models of the social structure of the UK. It is likely that there is assortative structure within household, workplace and community settings by a variety of individual metrics. Strong assortativity would have important impact on the dynamics of infection and other contagious processes.

We shall compile and incorporate information on assortativity from a range of sociological cross-sectional and cohort studies to analyse the character and strength of assortativity of interactions made by people in modern Britain. We will assess the assortativity and dissassortativity of British society within home, workplace, social and community settings, as well whether significant correlations exist between different assortative structures. By linking assortativity and interaction data with census and travel information, and with information from two recent GB-based social encounter studies we will develop models of social interactions and contagious processes: the spread of both infections and belief.

Our analysis will provide ourselves and other researchers, particularly those in strategic health planning and computational sociology, with an improved understanding of which social structures are important for the spread of diseases or beliefs. By incorporating this extra information into models of contagion we expect to considerably improve our understanding of observed patterns of infection within communities and how they may be best controlled in terms of efficacy and resources. In addition, we hope to lay a quantitative foundation with which researchers can model and explore contagion of information and belief within the UK. All final models, parameters and collated data generated through the project will be released to the academic community and other researchers.

Planned Impact

There is considerable interest in the development of mechanistic models of social interactions or to parameterize social networks, primarily from the need to improve predictive mathematical models of infection spread. This work has direct impact on public-health policy; for example the age-structured POLYMOD study (Mossong 2008) has been extensively used to help inform vaccination policy in the UK and elsewhere. We envisage that our study will have similar benefits, allowing us to substantially improve the models used to predict the impact of policy. While the application and data of the research are UK-focussed, the methodology and modelling approaches are applicable to any population: academic researchers and allied health professionals from a range of countries where similar data is available would find this work directly relevant to their research and policy-advisory tools.

We will ensure health professionals are aware of our work through publication in high impact academic journals, presentation of the work at an international conference, and through the team's direct links with researchers in the Health Protection Agency (to be Public Health England), and government advisory bodies such as the Joint Committee on Vaccination and Immunisation and the Scientific Pandemic Influenza Advisory Committee. Given that vaccination and other public-health policy is under continual review and change is common, we believe our work will be have an ongoing important impact on health policy. The general public will therefore benefit indirectly through improved models of social interactions contributing towards better infectious disease control.

In addition, beyond the modelling of infection, our work will help to generate important insights into the spread and maintenance of opinions, beliefs and behaviours within the UK. This would be helpful in understanding and possibly changing a range of health-relevant behaviours such as cessation of smoking, diet and lifestyle choices, and refusal to accept vaccination or treatment.
Description We have conducted an extensive search of surveys and other other publicly-available data sources to identify information relating to the social interaction of individuals. In total, we identified and reviewed 147 publicly-available information sources, of which we identified 44 which have directly relevant information. These identified data sources will be of immediate use for researchers wishing to model the structure of society and social interactions, in particular to model the spread and control of infectious diseases.

Our analysis focused on 3 studies with direct information about social interactions, and 8 information sources which contained detailed information regarding co-location of individuals (which types of people tend to interact or share common space). Of these 8, we conducted extensive analysis of 4 studies which are large studies of co-habitation and describe which types of people live with each other.

We devised a framework of individual attributes and characteristics with which to amalgamate information from different sources and studies in a consistent manner. Data were collated under this framework to produce a very large data-set of sampled households, the individuals who live in each household, and various attributes of those individuals and their households. These individual attributes include the age, sex ethnicity and occupation of individuals. For households, attributes include the number of people living in the household as well as information such as number of bedrooms -- information which has previously not been incorporated into mathematical models of epidemics, yet may have a great impact on the transmission dynamics of certain infections, such as influenza.

Using the collated data-set, we developed a realistic mathematical simulation model of social interactions and transmission of infection within and between households. With this model, we have shown that striking inequalities in disease risk arise naturally as a consequence of the choices people make in who they live with.
Exploitation Route Our work will be of interest to many researchers attempting to model infectious disease transmission and control (including the cost-benefit analysis of new vaccine policies) within Great Britain.

To this end, we will make available demographic-attribute based 'mixing' matrices for a variety of attributes. These tables describe the probability of individuals of different demography cohabiting or co-locating, and are based on secondary analysis of identified data sources. They will be presented in a form suitable for modellers to 'plug-in' to existing models. We anticipate that they will be considered for use by the modelling teams at Public Health England.

These matrices will be useful for research beyond the realm of infectious disease modelling, in that they will be available and relevant to any researcher attempting to characterise contagious processes within Britain; we anticipate they will be important for modelling the influence of peers on lifestyles and health (such as smoking and obesity) as well as the spread and maintenance of political and other beliefs.
Sectors Education,Healthcare,Leisure Activities, including Sports, Recreation and Tourism,Government, Democracy and Justice,Retail,Other

Description Due to the short time of the project, no findings have been used by policy makers. We are in communication with Public Health England regarding our findings as well as their data and analysis needs for modelling of infectious diseases. We have also been in communication with the equivalent organisation in the USA, the Centers for Disease Control and Prevention, to make them aware of our work and the potential for a similar study using US data. We are currently preparing two manuscripts arising from this work: the first will describe the systematic identification of data sources and to make the modelling community aware of the information available to parameterise models of society; the second will present the findings from the detailed household model of disease transmission. Methodologies developed during this project will be used in two forthcoming PhD projects: (1) a NIHR funded Health Protection Research Unit studentship developing mathematical models of susceptibility and infection in England and Wales, for use in outbreak scenario planning; (2) an ESRC Advanced Quantitative Methods studentship will develop statistical methods and simulation approaches to model Malawian population and society from census and other information.
First Year Of Impact 2014
Sector Healthcare
Impact Types Policy & public services

Description Mathematics in Healthcare
Amount £2,004,298 (GBP)
Funding ID EP/N014499/1 
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Academic/University
Country United Kingdom
Start 12/2015 
End 11/2019
Title Social mixing mactrices 
Description Demographic-attribute specific co-location matrices 
Type Of Material Database/Collection of data 
Provided To Others? No  
Impact We will be making these matrices available to other modelling groups, and in particular the modelling units at Public Health England. 
Description Collaboration with Public Health England 
Organisation Public Health England
Country United Kingdom 
Sector Public 
PI Contribution Attending meetings; identifying research priorities and knowledge gaps; providing information on available data and information needed to address research aims; analysis leading to publication and other outputs.
Collaborator Contribution Attending meetings; collaborative discussion; identifying research priorities and analysis needs.
Impact Publications: Todd S, Diggle PJ, White PJ, Fearne A, Read JM. The spatio-temporal association of non-prescription retail sales with cases during the 2009 influenza pandemic in Great Britain. BMJ Open 2014;4:e004869 Mutli-disciplinary in nature: epidemiology; mathematical modelling, clinical interpretation; economic analysis; social geography Pinsent A, Read JM, Griffin JT, Smith V, Gething PW, Ghani AC, Pasvol G, Hollingsworth TD. Risk factors for UK Plasmodium falciparum cases. Malaria Journal 13:298 Mutli-disciplinary in nature: epidemiology; mathematical modelling, clinical interpretation; social geography House TA, Baguelin M, van Hoek AJ, White PJ, Sadique Z, Eames K, Read JM, Hens N, Melegaro A, Edmunds WJ, Keeling MJ. Modelling the impact of local and reactive school closures on critical care provision during an influenza pandemic. Proceedings of the Royal Society, B. 2011; 278(1719):2753-2760 Mutli-disciplinary in nature: epidemiology; mathematical modelling, clinical interpretation; social geopgraphy
Start Year 2009
Description US State department 
Organisation United States Department of Health & Human Services
Country United States 
Sector Public 
PI Contribution Outreach activity following request by US HHS to model the risk of Ebola introduction to USA during early stages of the outbreak (Sept 2014), The model developed by Read used a secondary data analysis of airline transportation network and published WHO case data to estimate the risk of importation to the US for late 2014 and early 2015, and was presented to US State department and HHS partners at the US state department, Washington DC.
Collaborator Contribution Provided useful feedback on the modelling work, insight into findings, and provide travel funds.
Impact A paper is in preparation. Presentation provided to the US Ebola modelling consortium, Oct 2014. Poster on work presented at Epidemics conference 2015.
Start Year 2014
Description Epidemics conference 2015 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Posters and talks presented, and arising discussion with academic beneficiaries and PG students generated further research ideas and plans. Estimated audience ~200 people, representing international community.
Year(s) Of Engagement Activity 2015
Description Public consultation regarding Census data 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? Yes
Geographic Reach Regional
Primary Audience Public/other audiences
Results and Impact There was substantial interest in the research and significant public discussion of the use of data occurred.

Year(s) Of Engagement Activity 2013
Description The King's Fund -- Making better decisions for public health: Insights from secondary data 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Policymakers/parliamentarians
Results and Impact This was a one day event presenting research findings from six of a group of 15 projects exploring Health and Wellbeing, supported by the Economic and Social Research Council (ESRC)'s Secondary Data Analysis Initiative.

The aims and findings from the study were presented to an audience composed of health professionals, social and geographical researchers, and policy makers.

Year(s) Of Engagement Activity 2014
Description US HHS Ebola Modelling Consortium 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Policymakers/politicians
Results and Impact Invited presentation to consortium of international researchers and US-based government agencies, concerning the risk of non-infected countries importing Ebola from affected West African countries, during the height of the epidemic in 2014. Estimate audience ~150. Follow-up discussion occurred with various research groups (including commercial organisations), and a specific request from the Department of Health and Human Services to model the import risk during the Christmas period.
Year(s) Of Engagement Activity 2014
Description US state department presenation 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Policymakers/politicians
Results and Impact Invited visit to the US State Department to present modelling work on risk of importation of Ebola to USA and other countries. Estimated audience ~30 from US government, military and CDC. Very engaged discussion followed, and follow-up work continues.
Year(s) Of Engagement Activity 2015