Long-Term Modelling Tools for Adolescent Mental Health and Wellbeing Research
Lead Research Organisation:
University of York
Department Name: Centre for Health Economics
Abstract
Adolescents often experience mental health and wellbeing difficulties. When such difficulties persist or escalate over time, they cause misery to adolescents and their families, impose substantial costs on education, health and welfare services, and have lifelong effects on increasing the risk of numerous other bad outcomes including unemployment, crime, physical illness and premature death. They are also linked to adverse experiences in early childhood such as poverty and neglect that can perpetuate intergenerational transmission of inequalities in income, health and wellbeing.
Intervening to prevent and manage adolescent mental health problems is challenging, however. Research can help by using trials to evaluate the effects of different interventions on mental health. Unfortunately, however, trials typically only have one to three years of follow-up data. It is thus hard to know how far effects will persist in the long-term and how this will vary between different adolescents in different circumstances. Standard research methods for extrapolating long-term effects do not take into account emerging scientific knowledge about how the long-term persistence and consequences of mental health difficulties may vary for different adolescents in different circumstances. For some adolescents, short-term improvements in mental health fade out rapidly over time, ultimately yielding little or no benefit. But for others, improvements persist and accumulate into large lifelong benefits and public cost savings.
We propose to develop a better approach to long-term modelling of intervention effects on adolescent mental health and wellbeing. We will develop a versatile and reusable computer programme for predicting how adolescent health and wellbeing will develop from age 11 to 17 for different kinds of adolescents in different family, neighbourhood and school environments, known as a "microsimulation model". This model will be based on detailed data about the lives of more than 10,000 adolescents in the Millennium Cohort Study of children born in the year 2000, supplemented with further data where necessary. We will also link the new model to existing microsimulation models of childhood (age 0 to 11) and adulthood (age 17+) that capture the main early childhood causes and lifelong consequences of adolescent mental health and wellbeing difficulties. We will then test how well the integrated model can address a diverse range of research questions that cannot be answered using standard long-term modelling approaches, by conducting example studies of (i) the long-term effects of whole-school anti-bullying programmes, (ii) the long-term effects on adolescent mental health and wellbeing of tax-benefit reform options for reducing poverty in childhood, and (iii) the lifelong consequences of adolescent mental health problems for income, health, wellbeing and public cost in adulthood.
We will engage young people to ensure that our computer model meaningfully captures adolescent experiences and produces information relevant to young people. We will collaborate with youth social work apprentices and sixth formers in Bradford, a deprived area with relatively high rates of adolescent mental health problems, in a series of workshops throughout the project. We will also ensure scientific credibility and policy relevance by consulting experts from multiple disciplines across the adolescent mental health and wellbeing research field and with education, welfare and health policymakers across government. To help adolescent mental health and wellbeing researchers use and refine our approach in future, we will collaborate across research teams at five different UK universities and make our tool readily accessible via user-friendly web-based platforms.
Intervening to prevent and manage adolescent mental health problems is challenging, however. Research can help by using trials to evaluate the effects of different interventions on mental health. Unfortunately, however, trials typically only have one to three years of follow-up data. It is thus hard to know how far effects will persist in the long-term and how this will vary between different adolescents in different circumstances. Standard research methods for extrapolating long-term effects do not take into account emerging scientific knowledge about how the long-term persistence and consequences of mental health difficulties may vary for different adolescents in different circumstances. For some adolescents, short-term improvements in mental health fade out rapidly over time, ultimately yielding little or no benefit. But for others, improvements persist and accumulate into large lifelong benefits and public cost savings.
We propose to develop a better approach to long-term modelling of intervention effects on adolescent mental health and wellbeing. We will develop a versatile and reusable computer programme for predicting how adolescent health and wellbeing will develop from age 11 to 17 for different kinds of adolescents in different family, neighbourhood and school environments, known as a "microsimulation model". This model will be based on detailed data about the lives of more than 10,000 adolescents in the Millennium Cohort Study of children born in the year 2000, supplemented with further data where necessary. We will also link the new model to existing microsimulation models of childhood (age 0 to 11) and adulthood (age 17+) that capture the main early childhood causes and lifelong consequences of adolescent mental health and wellbeing difficulties. We will then test how well the integrated model can address a diverse range of research questions that cannot be answered using standard long-term modelling approaches, by conducting example studies of (i) the long-term effects of whole-school anti-bullying programmes, (ii) the long-term effects on adolescent mental health and wellbeing of tax-benefit reform options for reducing poverty in childhood, and (iii) the lifelong consequences of adolescent mental health problems for income, health, wellbeing and public cost in adulthood.
We will engage young people to ensure that our computer model meaningfully captures adolescent experiences and produces information relevant to young people. We will collaborate with youth social work apprentices and sixth formers in Bradford, a deprived area with relatively high rates of adolescent mental health problems, in a series of workshops throughout the project. We will also ensure scientific credibility and policy relevance by consulting experts from multiple disciplines across the adolescent mental health and wellbeing research field and with education, welfare and health policymakers across government. To help adolescent mental health and wellbeing researchers use and refine our approach in future, we will collaborate across research teams at five different UK universities and make our tool readily accessible via user-friendly web-based platforms.
Technical Summary
We will build a versatile and reusable discrete event microsimulation model ("LifeSim AMHW") of the co-evolution of a core set of AMHW and related outcomes for different adolescents in different circumstances from age 11 to 17, as well as the associated public costs across different policy sectors. Event transitions will be modelled based on mediating, moderating and confounding factors at individual-level, family-level, neighborhood-level and school-level.
We develop a conceptual framework of the main causal pathways involved using directed acyclic graphs (DAGS). We will estimate these pathways using longitudinal data from the Millennium Cohort Study. AMHW will be operationalised using latent class modelling of multiple manifest AMHW outcomes to address the problems that arise when relying on individual AMHW outcomes. Where necessary we will also use external data sources, for example large-scale administrative data on rare outcomes such as adolescent mortality, and estimates from a recent longitudinal study of school-level factors not available in MCS. To ensure accurate and up-to-date modelling of parental income during adolescence, which can mediate and moderate effects on AMHW, we will integrate our model with a UK tax-benefit model, UKMOD.
To test and validate the model, we will also conduct an example evaluation study of a school-based AMHW intervention, by mapping the short-term effects from an existing trial onto our simulated outcomes at age 14, and predicting the effects at age 17. To facilitate broader adoption of our tools we will also conduct example studies of the long-term AMHW consequences of tax-benefit reform options for reducing child poverty and heterogeneity in the long-term burden of AMHW difficulties for wellbeing and public cost in adulthood. We will also carry out external model validation against trial follow-up data and probabilistic sensitivity analysis of overall parameter uncertainty.
We develop a conceptual framework of the main causal pathways involved using directed acyclic graphs (DAGS). We will estimate these pathways using longitudinal data from the Millennium Cohort Study. AMHW will be operationalised using latent class modelling of multiple manifest AMHW outcomes to address the problems that arise when relying on individual AMHW outcomes. Where necessary we will also use external data sources, for example large-scale administrative data on rare outcomes such as adolescent mortality, and estimates from a recent longitudinal study of school-level factors not available in MCS. To ensure accurate and up-to-date modelling of parental income during adolescence, which can mediate and moderate effects on AMHW, we will integrate our model with a UK tax-benefit model, UKMOD.
To test and validate the model, we will also conduct an example evaluation study of a school-based AMHW intervention, by mapping the short-term effects from an existing trial onto our simulated outcomes at age 14, and predicting the effects at age 17. To facilitate broader adoption of our tools we will also conduct example studies of the long-term AMHW consequences of tax-benefit reform options for reducing child poverty and heterogeneity in the long-term burden of AMHW difficulties for wellbeing and public cost in adulthood. We will also carry out external model validation against trial follow-up data and probabilistic sensitivity analysis of overall parameter uncertainty.
