Multimorbidity (having multiple diseases at one time) is a growing problem in health and social care. Both research and actual recent events have shown that those suffering from multimorbidity have greater health and care needs and worse health outcomes. This is expensive for the health and social care system and often puts certain groups of people at a specific disadvantage. If we know more about what causes multimorbidity and how it develops, we will be better able to develop practical strategies to deal with it and to allocate funds fairly and usefully.
Often multimorbidity affects certain groups, especially the most vulnerable, but we do not yet have accurate ways of predicting the patterns. So far, most research has focused on how to define the problem. We aim to look at how multimorbidity actually develops and to see whether we can predict its development. Specifically, we do not understand how individual health journeys evolve: how do individual diseases, medicines, health behaviours, mental health, geography, income, etc. contribute to these patterns?
One of the reasons we do not know this is that the amount of data points is vast, and therefore difficult to compute. We now have the ability to program computers to learn as they go along, to more quickly and efficiently analyse large amounts of data.
This project aims to analyse data from a large set of patients using two different but complementary methods of computer learning. This study will look at a large set of general practice health care records (almost 40% of individuals with primary care records in the UK) using two different approaches to see which approach, or combination of approaches, most accurately predicts patterns of disease.
Approach 1: ‘biological’ age
The first approach is inspired by using an estimate of ‘biological’ age. Unlike a person’s actual age (that is to say: the actual number of years one has lived), ‘biological’ age is a combination of actual age and other information, including current and past diseases, other measures of health such as blood pressure, and life changes such as menopause (which happen at different ages for different people).
Approach 2: ‘longitudinal’ age
The second approach uses ‘longitudinal age, that is: data collected over several years from each individual to identify patterns in the groupings of these diseases. This will help identify those groupings that are most common and help improve the delivery of health care and ensure equitable allocation of resources.
By using these two methods together we plan to determine if they accurately pinpoint and predict which patients will develop certain groups of diseases, what diseases those might be, and in what patterns they might appear.
What we will do with the findings
This initial programme of work is what is called ‘Proof of Concept’, that is: it is a small project to see if the data that we have can be analysed in the way we hope and what that might tell us about the patterns of disease.
We need to confirm that these methods work with the sort of data currently available. We also need to see if their results are useful for clinicians and decision makers and what the advantages and disadvantages of each method might be. Finally, we need to see how they work together with each other, or with other additional types of analysis or data.
If the concept works, we plan to apply for funding for a follow-on larger project to develop tools to help policymakers decide what resources are needed for health and social care, and to help services provide better and more complete care for patients.
Who We Are
Lead applicant Professor Rafael Perera is Professor of Medical Statistics, Lead for the Thames Valley Applied Research Collaborative and Deputy Lead for the Oxford British Research Council multimorbidity themes and has studied long-term conditions. He has overall responsibility for managing the teams, ensuring cross-speciality communication and learning, and ensuring that the work progresses to time. He leads a team of data scientists, social scientists, statisticians, epidemiologists, psychiatrists who together have a broad spectrum of experience which they bring to addressing the problem of analysing the causes and development of multimorbidity.
Lead lay applicant Anica Alvarez Nishio is an editor and PPIE advocate focusing on health and education, who has first-hand experience in the practical delivery of health and mental-health care strategies, primarily making these fair and accessible. Additionally, she has hands-on/lived end-of-life and dementia care experience. She is currently undertaking a postgraduate course in Philosophy and has an interest in the effective usage of data and technology and particularly in the ethical issues surrounding the delivery of care. She will take responsibility for ensuring adequate support and training for lay members of the team as well as ensuring integration between the external steering and stakeholder group and the project.
Artificial Intelligence Machine learning; programming computers to learn as they go along
Biological Age: The medical or effective age of one’s body based on chronological age, plus physical health, mental health, social group, addiction issues, medications, etc.
Biomarker: An indicator which can be measured to signal the presence or severity of a disease
Chronological Age: The number of years one has lived
Multimorbidity: Having more than one disease at a time