The population in the UK is ageing and as a result we are seeing more patients with more than two long-term health conditions, termed multimorbidity. Multimorbidity is associated with taking lots of medicines (polypharmacy). Some medicines are used to treat symptoms (such as pain or depression) whilst others try to prevent diseases happening further down the line, called preventative medicines. Most studies for new medicines dont include women, very old or very young people or those that have many different conditions. Therefore it is hard to know whether these medicines are working as well in these groups meaning some patients are at risk of experiencing medication-harm (side effects) and may not be getting the benefit that we believe them to be. We therefore need to develop better ways to manage patients with both multimorbidity and polypharmacy. Although we know that multimorbidity is on the rise, it is not clear at the moment which diseases commonly occur together and importantly, how long it takes from developing two conditions until the third, fourth, fifth occurs. We want to develop a computer approach (called machine learning) that can look at patient records and cluster the commonly occurring diseases together to tell us which clusters are the most common, whether this differs for different age groups, and what factors lead to the development of a disease cluster. Understanding the make-up of multimorbidity will allow us to re-design clinic services to better meet the needs of the population and to try to reduce medication-related harm.
Background: The management of patients with multimorbidity is increasingly complex. Reliance on guidelines for the management of single diseases is a major contributor to increasing healthcare burden and polypharmacy. The use of multiple specialists to manage individual disease is costly and burdensome for both the health care service and patients alike, with oversight and support falling to the over-stretched GP. Understanding the scale of multimorbidity across the UK, and in particular the key determinants of different disease clusters along with the factors associated with worse outcomes, is critical to the design of sustainable future services for those with complex care needs that will reduce pill burden, inform future drug development, and provide better models of care.
Aim: To better understand the prevalence and distribution of different patterns of multimorbidity in the UK and the effect on healthcare utilisation and prognosis. Understand the patterns of medicines prescription in relation to multimorbidity clusters, and the key determinants and age distribution of different disease clusters.
Design: Observational descriptive cohort study involving electronic health care records. A novel machine learning cluster algorithm based on Frequent Pattern Mining, similar to supermarket basket analysis, will be used to cluster commonly co-occurring conditions. Comparison to exploratory hierarchical cluster analysis and Cox regression will be undertaken validate the accuracy of the machine approach.
Outcomes: The primary outcome will be time to accumulation of subsequent disease state(s) following acquiring an initial diagnosis of multimorbidity. Secondary outcomes include a) a taxonomy of commonly occurring disease clusters at the numerical level (2, 3, 4 diseases etc) and age group level (18-24 years, 25-30, etc) and b) Identify the determinants of multimorbidity (such as gender, socioeconomic status, smoking, alcohol, etc) in order to better understand the biological basis for disease clustering.
The primary outcome will be morbidity prevalence and time to accumulation of subsequent disease state(s)
following acquiring an initial diagnosis of multimorbidity (occurrence of second long-term disease).
Secondary outcomes include (a) a taxonomy of commonly occurring disease clusters and (b) Identify the determinants of multimorbidity (such as gender, socioeconomic status, smoking, alcohol, etc)
For analyses relating to multimorbidity incidence, the case cohort will be dynamic and cases will enter at the point at which they acquire a second long-term disease.
Lauren Walker - Chief Investigator - University of Liverpool
Lauren Walker - Corresponding Applicant - University of Liverpool
Christian Mallen - Collaborator - Keele University
Frans Coenen - Collaborator - University of Liverpool
Girvan Burnside - Collaborator - University of Liverpool
Kate Fleming - Collaborator - University of Liverpool
Munir Pirmohamed - Collaborator - University of Liverpool
HES Accident and Emergency;HES Admitted Patient Care;HES Outpatient;Patient Level Index of Multiple Deprivation