Evaluating Policy Implementations TO Predict MEntal health [EPITOME]: a Bayesian hierarchical framework for quasi-experimental designs in longitudinal settings

Mental health problems are a leading cause of ill-health, with 1 billion experiencing mental health or substance use disorders worldwide. In the UK, anxiety and depression contributed 8% of the years lived with disability in 2017 and are associated with the second largest costs to society after dementia. This morbidity is accompanied by substantial suicide mortality, which rose by 10.8% in 2018, with even faster rises amongst young people. Socio-economic disadvantage is strongly associated with mental ill-health. Austerity and immigration policies (i.e. systemic "shocks") implemented in the UK since 2010 may have increased mental health problems, particularly in disadvantaged populations, including people from ethnic minority backgrounds and may have been further exacerbated by the COVID-19 pandemic. However, causal evidence is missing. Data typically available to evaluate such issues are observational (without random allocation) and population-wide (without controls), biasing straightforward estimation of causal effects. To advance causal inference of systemic shocks (policies, COVID-19) on population health outcomes, we will develop a generalisable statistical framework to overcome biases inherent to observational data. We will then apply this to evaluate the cumulative long-term causal impact of these shocks on mental health outcomes using longitudinal and spatial data, with particular focus on ethnic inequalities.