Machine learning unlocks the complexities of multimorbidity

New paper reviews how machine learning can support multimorbidity research to promote Longevity.

Longevity research has been able to leverage the power of machine learning to great effect in the last decade or so. Companies focused on Longevity therapeutics, such as Insilico Medicine and Juvenescence, are using machine learning to discover and assess drugs, predict outcomes and analyse patient populations. Now it’s time for multimorbidity research to benefit.

Longevity.Technology: Age is a huge risk factor for multimorbidity. Just as people are living longer, so there is an increase in long-term conditions and a rise in the number of people with two or more chronic conditions – multimorbidity. As well as having an impact on quality of life, it leads to poorer health outcomes and is responsible for disproportionate healthcare workload and costs. The key to treating multimorbidity and promoting Longevity lies in targeting the underlying age mechanisms and machine learning is key in analysing disease clusters, risk trajectories and treatment effects.

There has been much health-related research focused on the prevention and management of single medical conditions, looking at prevalence, cause and treatment in isolation.

Multimorbidity is on the rise; indeed it is now the norm in higher-income countries, with over 50 million people affected in the EU alone [1], and this means that research needs to understand the extent of the burden and to enable companies and institutions to develop strategies that can effectively treat disease clusters and interaction.

Researchers at the Medical Sciences Division of the Nuffield Department of Women’s and Reproductive Health at The University of Oxford examined how machine learning tools such as matrix factorisation, deep learning, and topological data analysis enabled access to a greater understanding of evolving patterns of multimorbidity.

Their paper on their work highlights the following findings [2]:

  • Machine learning is making significant contributions toward our understanding of the complex relationships between diseases;
  • Advanced models take a range of modalities from big datasets with little pre-processing or information loss at study design;
  • Developments in matrix factorisation and deep learning allow a better understanding of evolving patterns of multimorbidity.

There is still work to be done, however. The paper notes that data preparation and processing, incorporation of prior expert-knowledge and mapping the final results to clinical use and medical guidelines are areas that warrant improvement.

Additionally, the Oxford team posit, alternative factorisation techniques, representation learning, interpretable and causal ML with uncertainty estimation are new developments in machine learning that have not yet been used for the study of multimorbidity and could add to the success of methodology research in multimorbidity.

Machine learning is key to scientific and medical research; it can de-risk clinical trials, discover drugs and develop models of aging and disease. There are enormous biomedical datasets that are packed with information on the health, treatment histories and treatment outcomes of millions of individuals, as well as drug performance and clinical trial results.

Interrogating this data and analysing trends of disease where they are in co-existence and determining their dynamic relationships will not only allow better treatment of multimorbidity, but will have a significant impact on promoting both healthspan and lifespan.


Image credit: Arek SochaPixabay