Multimorbidity is defined as the presence of two or more conditions in one person at the same time and affects many people both in the UK and worldwide. Research in the past has focused around understanding single diseases and as a result, the findings may be less relevant to the challenges faced by those with multiple conditions. Much of the existing research in multimorbidity treats it simply as a count of the number of long-term conditions, which may miss important patterns in groups of conditions with shared characteristics.
This research aims to detect novel groups (or clusters) of people with similar patterns of long-term conditions, including information on how these conditions develop over time, by applying recently developed machine learning methods to GP health records.
Exploring patterns of multimorbidity and how different conditions group together, including how they develop over time, might lead to new opportunities for proactively preventing diseases occurring. Viewing diseases as part of clusters rather than as single conditions might also lead to more effective strategies for treatment, including reducing inappropriate medication use, and provide insights into how we could better design the healthcare system to support people with multimorbidity.
Multimorbidity is defined as the presence of multiple chronic conditions in one person and is an emerging priority globally and within the UK. Multimorbidity leads to greater healthcare utilisation and costs, and poses a challenge both to health systems, which tend to be organized around single disease pathways and guidelines, and to patients in navigating their care. Much of the existing literature on multimorbidity characterises it based on counts of the number of long-term conditions, and using a narrow set of conditions, which may miss underlying associations between conditions and between less common diseases. Furthermore, most approaches are cross-sectional, based on current disease burden, which may miss patterns between people with multimorbidity in the sequence of disease onset. Identification of ‘clusters’ of similar diseases or of people with similar conditions may provide greater insights into the aetiology of multimorbidity and is a research priority for the National Institute for Health Research.
This study aims to produce clusters of people with similar diseases and patterns of disease development. The study has two sub-studies: study 1 uses a cross-sectional approach based on current disease burden and study 2 aims to produce clusters of similar people incorporating information on sequence and trajectories of disease acquisition and examining the predictors of cluster membership. We will use a combination of machine learning methods developed from natural language processing which can handle large and high-dimensional datasets, benchmarked against existing clustering methods.
The intended public health benefit of identifying clusters is to aid our understanding of aetiology, disease progression and potential targeted approaches for prevention of multimorbidity. We also aim to explore the predictors of multimorbidity cluster and whether health inequalities exist, to inform public health policies to target the people most at risk from multimorbidity.
People will be assigned to clusters of people with similar disease patterns (type of disease and trajectory of acquisition); predictors of cluster assignment.
Thomas Beaney - Chief Investigator - Imperial College London
Thomas Beaney - Corresponding Applicant - Imperial College London
Azeem Majeed - Collaborator - Imperial College London
Jonathan Clarke - Collaborator - Imperial College London
Mark Cunningham - Collaborator - Imperial College London
Mauricio Barahona - Collaborator - Imperial College London
Paul Aylin - Collaborator - Imperial College London
Thomas Woodcock - Collaborator - Imperial College London
Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation