
Mental Health Data Science and Epidemiology

Data science stands at the forefront of mental health research, exploring the origins, progression, and personal impact of mental disorders. It offers a pathway to transform insights derived from data into tangible improvements for patient care.
The main aims of our group are:
- To maximise the scope and impact of our Data Science and Epidemiology research portfolio
- To collaborate with other groups within the University, NHS, commercial sector and a global network
- To work towards data linkage between mental health and other data source platforms
- To further develop mental health data science on theMasters and PhD programmes within the Institute of Mental Health
- To aim to be a global leader in Mental Health Data Sciences
Please contact the Mental Health Data Science and Epidemiology Research Theme Leads for further information:
Prof Steven Marwaha Lead
Dr Isabel Morales-Munoz Co-Lead
Our current research:
PREVENTA
PREVENTA
PREVENTA: PRevention of pErsistent high leVels of deprEssion across adolesceNce and young adulThood: the role of Active ingredients
Key people: Dr Isabel Morales-Munoz, Prof Steven Marwaha, Dr Pavan Mallikarjun, Dr Alexander Zhigalov, Prof Christopher Yau, Dr David Wong, Dr Anna Moore, Buse Durdurak, Beckye Williams
The aim of this project is to detect the combination(s) of active ingredients occurring by age 11 that associate with lower risk of persistent high levels of depression across adolescence and young adulthood (from 13 to 18 years), using the Avon Longitudinal Study of Parents and Children (ALSPAC) cohort. During the Discovery phase, we will apply Latent Class Growth Analyses, to detect the trajectories of depression across adolescence and young adulthood, including a group with persistent high levels of depression. Second, we will apply logistic regressions to investigate the associations between a range of active ingredients and different groups of depression. This will allow the characterization of the relevant active ingredients that are associated with lower risk of persistent depression, which forms the basis for the digital tool. During the Prototyping phase, we will build a digital tool comprising novel predictive machine learning models that will reliably predict the combination(s) of active ingredients that associate with lower risk of persistent high levels of depression. In the future, this digital tool will be validated against routinely collected data and will ultimately form the basis of a future digital stratification tool for clinicians, to guide early intervention of young people with depression. For this project, we are working with collaborators from the University of Birmingham, University of Oxford, University of Manchester and University of Cambridge. In addition, for the design and development of this project we are collaborating with the Youth Advisory Group, at the Institute of Mental Health, and we have a lived experience involvement lead withing our team.
Psychosis Immune Mechanism Stratified Medicine Study
Psychosis Immune Mechanism Stratified Medicine Study
Key people: Professor Rachel Upthegrove
In collaboration with the University of Cambridge, we have recently been awarded a major grant from the Medical Research Council to investigate the link between increased brain inflammation and psychosis. Evidence suggests that inflammation may be present before and during the early stages in some, but not all young people with psychosis. In this multicentre Psychosis Immune Mechanism Stratified Medicine Study, Professor Upthegrove and Dr Khandaker will lead a team of investigators, including University of Birmingham MDS Professors Nicholas Barnes, and George Gkoutos, to examine how immune dysfunction could cause psychosis and use advanced AI techniques to identify who might benefit most from novel immune targeted treatments.
Is this still a current project? (if so, add further details from additional page)
Developing and validating a prediction model for the onset of bipolar disorder
Developing and validating a prediction model for the onset of bipolar disorder
Using a machine learning approach to develop and validate a prediction model for the onset of bipolar disorder
Key people: Professor Steven Marwaha and Dr Pavan Mallikarjun
Bipolar disorder (BD) is a debilitating mental health condition, characterised by severe shifts in mood, that can range from disabling highs (i.e., mania/hypomania) to extreme lows (i.e., depression). Approximately 1-2% of the population are affected by bipolar, with most people experiencing the onset of mood symptoms prior to their 20s. Despite this, little is known about the predictors to bipolar disorder and hypomania symptoms, particularly among young people. Intervening early in the development of bipolar is a top clinical priority, and one that may have the potential to limit its functional and symptomatic impact on those affected. Thus, predicting the onset of bipolar/hypomania prior to its onset, may help clinicians/researchers to develop novel, tailored preventative strategies and interventions for young people.
We aim to develop and validate a clinically useable prediction model to predict the risk of onset of hypomania in young people aged 21/22. We will use machine learning approaches to develop prediction models to predict the onset of hypomania in young people aged 21/22.
Anxiety and mood disorders in young people
Anxiety and mood disorders in young people
A multivariate approach using the ALSPAC cohort
Key People: Professor Steven Marwaha
In this project, we will investigate the early life psychological, environmental and biological factors that may increase the risk of developing mood disorders (broadly defined) in late adolescence and early adulthood and the associated poor outcomes. Multi-factorial analyses will examine prospective impacts of diet, sleep, parental mental health, cognitive function, inflammatory markers, metabolomics and genetic susceptibility on the earliest symptoms of mood disorder.The aims of this project are:1. To investigate the longitudinal relationship between early explanatory factors and the development of affective symptoms, mood and anxiety disorders2. Develop and test predictive models of mood and anxiety disordersDescriptive statistics will be used to summarise data. A combination of univariate and multivariate modelling and advanced AI techniques will be used; for example, risk relationships will be investigated using general linear modelling (GLM), path analysis and supervised machine learning to predict depression, anxiety and recovery outcomes accounting for important confounding variables.
Adult Psychiatric Morbidity Survey (APMS) 2014
Adult Psychiatric Morbidity Survey (APMS) 2014
Mood Disorders and Personality Disorders
Key People: Professor Steven Marwaha
The purpose of the study is to complete research on mood disorders and personality disorders, in order to ultimatelyimprove the health care that this group receive. The aims of the research are to analyse the APMS data to betterunderstand the causes of mental disorders, health services access, inequalities and links to other disorders.
Prediction modelling of Outcomes in Psychosis
Prediction modelling of Outcomes in Psychosis
Key People: Dr Pavan Mallikarjun
Psychotic disorders, including schizophrenia, are among the 20 leading causes of disability worldwide. People with psychosis have heterogeneous outcomes with more than 40% not achieving good outcomes. Early identification of individuals with a higher risk of poor outcomes at initial clinical contact may facilitate personalized interventions, reduce time to their initiation and improve utilization of resources. Pavan is working with collaborators including from Centre for computational biology (UoB), University of Glasgow, University of Melbourne, University of Leiden and World Psychiatric Association in developing and validating outcome prediction models in psychosis using Artificial Intelligence and Machine Learning techniques.
Outcome of mood disorders using electronic health records data
Outcome of mood disorders using electronic health records data
Key people: Professor Steven Marwaha, Dr Danielle Hett
The mood disorders lab is currently investigating the outcomes of those with bipolar disorder and depression using the de-identified Clinical Record Interactive Search (CRIS) system (https://crisnetwork.co/uk-cris-programme). We have two CRIS projects currently underway, the first investigating long term outcomes of people with bipolar disorder under the care of secondary mental health services in Birmingham and Solihull; the second project examines the relationship between ethnicity and key clinical outcomes (e.g., social functioning) among people diagnosed with depression.
Sleep and Mental Disorders
Sleep and Mental Disorders
Key people: Dr Isabel Morales-Munoz
Sleep problems in childhood are clearly associated with mental health problems, but it remains unclear whether early sleep problems precede the development of mental disorders, and the nature and specific pathways of these associations are still unknown. Early sleep interventions may prevent or ameliorate the development of future psychopathology. Isabel is collaborating with the Institute for Mental (UoB), the Child Sleep Cohort (Finnish Institute for Health and Welfare, Helsinki), the FinnBrain Cohort (University of Turku, Finland), and University of Umeå (Sweden) to understand how sleep develops in early childhood and the impact of sleep in mental health.
PRONIA: Personalise Prognostic Tools for Early Psychosis Management
PRONIA: Personalise Prognostic Tools for Early Psychosis Management
Key People: Professor Rachel Upthegrove, Professor Nicholas Barnes, Professor Stephen Wood, Dr Sian Lowri Griffiths, Mr Paris Lalousis, Dr Katherine Chisholm, Dr Renate Reniers, Dr Pavan Mallikarjun
PRONIA Collaborative Lead: Professor Nikolaos Koutsouleris
Affective and non-affective psychoses have a major negative impact on human society. They account for 6.3% of the global burden of disease and cost €207 billion per year in Europe alone, making them the most expensive brain-related disorders and even more expensive than cardiovascular diseases. Reliable and accessible prognostic tools have the potential to alleviate this burden by enabling individualised risk prediction, thus facilitating preventive treatment tailored to the needs of the individual patient. PRONIA aims to use routine brain imaging, clinical, proteomic, metabolic and neurocognitive data to optimise candidate biomarkers for the prediction and staging of psychoses and and depression that generalises well across different mental health services. The University of Birmingham PRONIA partner is particularly interested in investigating heterogeneity, depression comorbidity and transdiagnostic immune markers in emerging mental health disorders within the PRONIA.
Data Analysis Experience
This currently includes:
- Longitudinal data analysis
- Prediction modelling / Machine Learning
- Network analysis
- Path analysis
- Latent Class Growth Analysis
- Multi-modal analysis