MLTC-M Community of Practice in Statistical Methods for Multimorbidity Research

According to the NHS, two out of three of individuals aged over 65 have more than one long term health condition. Having two or more conditions, also known as multimorbidities, can cause complications as both conditions have to be taken into account when treating an individual. For example, we may want to give a patient a drug to treat one of their conditions but this treatment could potentially make their other conditions worse. Alternatively if someone is on two different drugs we don't always know how those drugs interact and whether having both drugs could cause adverse side effects.

Medical research generally focuses on one condition at a time and therefore results from medical studies may not be applicable in practice where patients may have many other conditions. To tackle this problem a number of studies have recently been set up to investigate multimorbidities. These include exploring which conditions tend to occur together, investigating multimorbidities that lead to hospitalisation or identifying biological processes that cause similar diseases to occur throughout the body.

There are a number of shared analytical methods that will be used in these studies and each study group contains specialists in cutting edge statistical techniques. We aim to set up a "Community of Practice" that will facilitate collaborations between these projects to ensure that the highest quality scientific methods are used for studying multimorbidities. This will allow expertise to be shared across the studies and to the wider scientific community.

This community will hold regular meetings to share developments in the field, as well as providing a platform for early career researchers to present their work in a supportive and constructive environment. Workshops will allow small teams of experts to develop protocols for best practice that can be directly implemented in each of the studies. These results will also be shared amongst the wider scientific community through online repositories and video tutorials. There will be dedicated support for early career researchers to develop skills in applying these methods and for general career development.

This community of practice will allow research into multimorbidities to be performed efficiently and to the highest standard to ensure the highest possible impact on patient outcomes.

The CoP will define best practices and coordination of effort for the following themes:

Defining multimorbidity: Each Research Collaborative will identify patients with multiple long term conditions. Defining these robustly from electronic medical records and cohort data is critical. Resources exist with published code lists, but consistency is lacking. Sharing best practice will add considerably to the robustness and efficiency in documenting and publishing these.

Statistical Genetics: Genome-wide association studies (GWAS) aim to determine association between genetic variants and a specific outcome. Extending GWAS to multimorbidity is an active area for the research collaboratives. "Genetic correlation" determines the overall shared genetics between traits, while "Mendelian randomisation" uses genetic instrumental variables to test for causality. The CoP will identify collaborative opportunities and share knowledge to apply these methods.

Translation to treatment: Each consortium aims to improve the treatment of individuals with multimorbidity. Genetic discoveries can identify targetable biological mechanisms and pharmacogenetics investigates the effect of genetic variants on treatment response, potentially aiding the development of personalised treatments. The Research Collaboratives will use diverse forms of clinical data, such as imaging, which are difficult to analyse using traditional methods. However, techniques such as machine learning (ML) can be used. ML aims to develop computer algorithms that improve with experience. Examples include processing and normalising raw data, integrating genomic information, exploring data structure, predictive modelling, generative modelling, and prioritising experiments.

Data resource curation: The CoP will actively work to share best practices, new data and data analysis code between the research consortiums. Output will be circulated to the wider scientific community through reports and educational resources.