Introduction

A regular series of open invitation talks from leading voices in data science, AI, healthcare and those with lived experience dealing with multiple long-term conditions (MLTC). This seminar series is part of the AI for multiple long-term conditions: Research Support Facility project. 


Event title: New findings from age-dependent topic modelling of Multiple Long-term Conditions in UK Biobank

Speakers: 

Chris Holmes, The Alan Turing Institute

Xilin Jiang, University of Cambridge

About the event

Longitudinal data from electronic health records (EHR) has potential to improve clinical diagnoses and personalised medicine (Abul-Husn et al. 2019 Cell). A key challenge is to extract biologically useful information from age-dependent patient disease histories spanning many distinct diseases.

In this session, we will introduce an age-dependent topic modelling (ATM) method that provides a low-rank representation of longitudinal records of hundreds of distinct diseases in large EHR data sets. The model learns, and then assigns to each individual, topic weights for several disease topics, each of which reflects a set of diseases that tend to co-occur as a function of age (quantified by age-dependent topic loadings for each disease). We use variational Bayesian methods to estimate the parameters of the model. Simulations show that ATM outperforms other approaches in distinguishing distinct age-dependent comorbidity profiles.

Results of the ATM applied to 282,957 UK Biobank samples, analysing 1,726,144 disease diagnoses spanning 349 diseases with ≥1,000 incidences identified 10 disease topics optimizing model fit. Topic loadings were highly age- dependent, implying differences in disease etiology for early-onset vs. late-onset disease. Further analyses reveal distinct polygenic risk score (PRS) subtypes across the topics.

Register now

Everyone is welcome at this session, however, the target audience is clinicians, policymakers and health data science researchers. We hope that attendees will learn more about this ATM method and how this learning can be applied to multiple long-term conditions research.

Register now

By registering for the event, you are agreeing to the events code of conduct:
Events code of conduct | The Alan Turing Institute

Speakers

Xilin Jiang

British Heart Foundation Career Development Fellow & University of Cambridge

Organisers

Dave Chapman

Programme Manager, AI for Multiple long-term conditions Research Support Facility