Using deep learning approaches to examine serious mental illness and physical multimorbidity across the life-course: from mechanisms towards novel interventions

Research question: What are the clinically important clusters of serious mental illness-physical multimorbidity that influence patient outcomes, and the social, environmental, behavioural, clinical and biological correlates of these? Background The prevalence of people living with multimorbidity is increasing globally.

Patients with multimorbidity often have complex health needs, live in lower socio-economic circumstances, experience reduced life expectancy and receive fragmented care, due to the dominant single-disease-oriented approach in healthcare.

This often leads to incomplete, inefficient, and costly healthcare delivery, and contributes to poor health outcomes in these individuals.

Serious mental illness (SMI, defined as severe depression, bipolar disorder, and schizophrenia) combined with physical illness is one of the most common multimorbidity clusters observed across patients, and is associated with substantially lower life expectancy.

Improving our understanding of the factors associated with multimorbidity would help inform resource allocation, address healthcare inequities, and inform holistic patient care.

Aims and objectives The broad aims of this project are to identify clinically important SMI-physical multimorbidity clusters within nationally representative datasets, and understand the social, environmental, behavioural, clinical and biological correlates of these.

Developing a biological and mechanistic understanding of multimorbidity could also help identify new targets for treatment and primary and secondary prevention, and potentially identify the impact of polypharmacy on illness, thus optimising drug treatments for patients.

The specific objectives of this development grant are to build a multi-disciplinary collaboration, develop frameworks for data access and governance, and undertake feasibility work to inform our research focus, facilitate future work, and enable the delivery of the aims of the full Research Collaboration.

Methods In order to meet our objectives, we will consolidate our multidisciplinary collaboration through workshops to inform detailed research plans for the Research Collaboration.

We will also undertake feasibility work, including PPI, risk assessments, and set up of data governance frameworks for large-scale clinical, and genomic data.

Our pilot analysis of the UK Biobank data, using state-of-the-art deep learning methods to identify clinically important multi-morbidity clusters, and their correlates will inform the focus of our future work for our full Research Collaboration proposal.

Anticipated impact and dissemination Understanding the national social, behavioural, environmental, epidemiological, and spatial correlates of multimorbidity will inform treatment pathways, including the allocation of resources needs across the UK, and public health policy.

Our research will help fill key gaps in knowledge about multimorbidity clusters within ethnic minorities in the UK; this may allow targeted social and healthcare interventions within these communities, with patient engagement and education around prevention.

Furthermore, prediction algorithms arising from the proposed work will be validated against existing risk scores, and provide clinical decision support through easy-to-use and interpretable web portals.

These are expected to directly assist clinical-decision making for people with multi-morbidity, through identification of those at higher risk of poorer outcomes, informing patient management.

Our work on understanding the impact of chronic treatments, and polypharmacy on multimorbidity will also help us optimise patient management. Ultimately, this work will inform the provision of more holistic patient care in the context of multi-morbidity.