The digital epidemiology of social media behaviour and mental wellbeing in the Avon Longitudinal Study of parents and children

点击翻译按钮获取中文摘要

There have been recent and startling rises in reported mental health difficulties, particularly in young adults. This means it is more crucial than ever to understand the origins of mental health and wellbeing to inform public health interventions and governmental policies that promote good mental health and prevent the far-reaching consequences of mental illness, an urgency recognised by both government and major research funders. Social media and other aspects of digital footprint have huge potential to facilitate the study of the causes of mental ill health through access to ecologically valid real time data on human behaviour. For example, the literature includes examples of automated methods for the detection of mental health conditions such as depression, Post-Traumatic Stress Disorder, suicidal ideation and Seasonal Affective Disorder. These measurements from social media have the potential to complement traditional questionnaire and diagnostic interview measures of these dimensions and disorders because they are based on large numbers of real social interactions and do not rely on retrospective recall or introspection.Validating these digital phenotyping approaches against ground truth is key to realising this potential. However, participants in social media research are usually unknown, so it has been historically difficult to do this effectively, a limitation that has so far restricted the usefulness of research using social media and other aspects of the digital footprint. To overcome this, recent efforts have linked social media data in UK birth cohorts such as the Avon Longitudinal Study of Parents and Children (ALSPAC). Studying digital phenotypes in well characterised populations with established ground truth measures allows us to put social media analysis on a sound epidemiological footing.This project aims to address limitations of earlier research by performing analysis using linked Twitter data for approximately 900 participants from ALSPAC. These data will be used to explore associations between Twitter behaviours (e.g. time of day when tweeting, language used in tweets, interactions with other users) and mental health and well-being. Where possible, we will use state-of-the-art approaches to make directional causal inferences about these relationships. Methodologically, this will involve machine learning techniques including Natural Language Processing and causal inference methods such as Granger Causality and Mendelian Randomization.This project could enhance research into the relationship between social media and mental health outcomes. This could enable exploration of the various proposed causal mechanisms between social media and mental health, and the identification of potentially harmful social media usage. This could inform the development of interventions aiming to modify social media behaviour. Furthermore, knowledge of these relationships could help improve the robusticity of machine learning models aiming to predict mental health status from behaviour on social mediaThis project has three main research questions:1. What is the relationship between timing of Twitter activity and mental health outcomes?2. What is the relationship between Twitter network characteristics and mental health outcomes?3. What is the relationship between incoming Twitter content and mental health outcomes