The Big Data for Complex Disease consortium is delighted to host this symposium to bring together researchers from across HDR UK’s Driver Programmes and wider Institute who are interested in using health data modelling to understand risk and disease trajectories, better-informing prediction, prevention, and treatment across the population. The symposium be an exciting opportunity to hear from researchers working with novel methodologies and tackling challenges in working with health data across various diseases, co-morbidities, and types of EHRs.

The symposium will feature two keynote talks, a panel discussion on the challenges of creating prediction models that deliver benefits for patients, a virtual poster session and short talks selected from submitted abstracts. We encourage participants to present on a range of topics to discuss the challenges faced (and potential solutions) when applying modelling approaches to longitudinal EHRs, including:

Data challenges

  • Dealing with diverse data, e.g., from multiple centres, primary and secondary care settings, and different data modalities
  • Addressing biases in data – methods to account for data missingness, loss to follow-up
  • Defining disease phenotypes and potential causality using longitudinal EHRs, including symptom-level data

Modelling approaches

  • Modelling of disease progression trajectories to understand co-morbidities and risk predictors
  • Dynamic risk prediction models and causal risk prediction models
  • Machine learning models for temporally sequenced data and repeated measurements

Implementation

  • Developing model evaluation metrics
  • Patient and public involvement and engagement (PPIE) in longitudinal EHR studies
  • Deploying EHR longitudinal models in clinical practice

Speakers:

We are very pleased to have agreed to two excellent keynote speakers who will discuss statistical and machine learning approaches to modelling longitudinal data.

Professor Tingting Zhu is an Associate Professor in AI for Digital Health in the Department of Engineering Science at Oxford. Her research interests lie in machine learning for healthcare applications, and she has developed probabilistic techniques for reasoning about time-series medical data. Her work involves the development of machine learning for understanding complex patient data, with an emphasis on Bayesian inference, deep learning, and applications involving the developing world.

Professor George Ploubidis is Professor of Population Health and Statistics at the UCL Social Research Institute. He is a multidisciplinary quantitative social scientist and a longitudinal survey methodologist. His research interests relate to socioeconomic and demographic determinants of health over the life course and the mechanisms that underlie generational differences in health, well-being and mortality. His methodological work in longitudinal surveys focuses on applications for handling missing data, causal inference and measurement error.

Short Talks

The symposium organizers have selected six Short Talks from the submitted abstracts, and these presenters have also been offered a virtual poster position. The selected Short Talks are listed below:

  1. Longitudinal Analysis of Disease Trajectories of Multimorbidity through Process Mining Techniques – Areti Manataki, University of St Andrews
  2. Longitudinal Study of Neighbourhood Characteristics and Mental Health in Childhood and Adolescence in England – Niloofar Shoari, Great Ormond Street Institute of Child Health, University College London
  3. Characterising post-myocardial infarction disease trajectories using non-negative matrix factorisation: a nationwide analysis – Jonathan A. Batty, Leeds Institute for Data Analytics, University of Leeds
  4. Use of national data collections to understand Lynch syndrome diagnostic testing pathways in endometrial cancers – Catherine Huntley, Institute of Cancer Research
  5. Inequalities in asthma outcomes in England: a national cohort study – Zakariah Gassasse, Imperial College London
  6. Mortality in adolescents with and without neurodisability in England: a national cohort study using linked health and education data from ECHILD – Louise Macaulay, University College London  

Patient and Public Involvement and Engagement

Patients and the public who kindly provide their health data are fundamental to our research and we are delighted that they are part of the symposium programme. We will have public contributors read and review the submitted abstracts as part of the panel for short talk selection. We are pleased to invite a public contributor to serve as a member of our panel discussion on the potential benefit disease prediction models can have on patients and members of the public.

We are happy to welcome any lay representatives who would like to attend the symposium online. Please contact Amy Hodgkinson (amy.hodgkinson@hdruk.ac.uk) to ensure any accessibility requirements and support are in place for attendance.

If you have any queries or questions regarding this event please contact Alexis Webb at alexis.webb@hdruk.ac.uk.