Developing and applying innovative methodologies of longitudinal network data and agent-based modelling to public health

Everyone is embedded within a social network of friends, work colleagues and family. The people who we are surrounded by can influence our thoughts and choices we make in life, and indeed our health and well-being. This study investigates the effect that the social networks within a school environment have on adolescent risk-taking behaviours by looking at the self-reported levels of alcohol, tobacco and drug use as well as the evolving social networks the young people are embedded in. Friendships can influence adolescent health behaviours, but are these due to peer selection (like associates with like) or peer influence (adoption of peer behaviours)? The aim of this study is to explore further these ideas and investigate how the school environment, gender segregation of friendships, and key personal or parental attributes, might influence the peer network and affect the magnitude or ?mechanisms? of the network effects on substance use behaviour. This study will use a number of different research methods including longitudinal network analysis and agent-based modelling to provide a clearer understanding about the effects of social networks on substance misuse and to enhance public health interventions.

Background: It is widely accepted that the social network within which one is embedded can affect one?s health and behavioural choices. Within a school environment peer influence and selection are two computing forces on behavioural choices and substance misuse.
Research aims and objectives: To determine how social network and behavioural dynamics affect and alter interventions to prevent substance abuse in adolescents and to investigate what factors shape the evolution of young people?s school social network and consequent influence on health behaviours.
Design: A longitudinal panel study over 5 waves of 8th through to 12th graders has been conducted 2001-2005 (when 11/12 to 15/16 years old) within the Belfast Youth Development Study (BYDS). BYDS documents complete network data of 5457 school children from 43 schools of single gender and coeducation, from the integrated, controlled and maintained systems and from 3 demographic areas, who have completed annual questionnaires (2013 completing all 5 waves) on lifestyle and substance abuse behaviours (alcohol, tobacco, cannabis, cocaine, ecstasy, heroin, LSD, speed) and includes the naming of up to 10 friends from within their year group.
Methodology: A three-fold analysis will take place: (i) traditional Social Network Analysis (SNA) techniques to observe the peer networks in place at the individual, personal and group level at any one time; (ii) longitudinal study of the dynamics of SNA using the recently developed stochastic actor-driven models for the evolution of social networks developed by Snijders et al. (SIENA) and (iii) to capture the adaptive dynamic interdependence of actors and their behaviours, an agent-based model (ABM) will be developed to take account of diversity and feedback within the pupil population. Research questions will be addressed through the investigation and analysis of single and multiple-wave network data. ABMs will be developed in accordance to the findings of the network analysis to improve our design of future interventions to mitigate their potential harm.
Scientific and medical opportunities: The purpose of the actor-driven and agent-based models is to provide greater flexibility in investigating how various characteristics of individuals and their social networks may affect the spread of substance abuse behaviours within schools, and how different policies might affect average levels of substance abuse. The analysing of such cohort data is important so as to feedback into the public health system through intervention design and policy