Introduction

Whilst a heart attack can be fatal, a large number of people survive their heart attack. Because patients are living longer after their heart attack there is a higher chance of them developing further conditions later in life. The full extent of which conditions are likely to develop following a heart attack is unknown. Therefore, this project aims to study millions of anonymised hospital records for patients across the whole of England using artificial intelligence techniques. The research may lead to the discovery of previously unknown consequences of a heart attack, for which new treatments or interventions could be developed to ultimately improve more patient lives.

Explaining the science

Major advances in our understanding and treatment of disease mean that people increasingly live with diseases that until recently were rapidly fatal. For example, it is now common to survive over a decade after the diagnosis of myocardial infarction (heart attack). However, this shift change in survival means there is a higher chance of developing further conditions later in life following a heart attack. The full extent of which conditions are likely to develop following a heart attack is unknown. 

Current statistical techniques to look at sequences of disease rely on existing knowledge of the diseases we wish to study and in which order we expect them to occur. This therefore limits scientific discovery of previously unknown consequences of a heart attack. Another limitation is that such methods often do not work well when a high number of different diseases are of interest. 

Therefore, this project aims to use machine learning techniques to investigate a large number of possible diseases which occur after a heart attack. The methodology we will investigate is called process mining - which allows us to assess millions of hospital records to determine all possible pathways of disease according to the date on which they were recorded. We will investigate how useful the results from process mining may be to the clinical community and how best to present them. We will also consider if additional or new methods are required to fully understand post heart attack disease trajectories.

Project aims

Using machine learning techniques to search through millions of records containing timestamped patient data has the potential to provide rich information on the sequenced and temporal pathways between many diseases, whilst also providing potential for discovery of novel disease trajectories which may otherwise never be detected. Thereby, there is a greater chance of identifying potential areas for prevention or interruption of the most fatal disease pathways. 

However, the most appropriate methods needed to study large volumes of data of potentially thousands of unique disease codes in a sequence at unstructured time intervals are not yet known. Therefore, this pilot study will investigate how well process mining techniques - originally designed as a business analytics tool - perform on this volume and complexity of real-world patient data. The outcomes of this research will inform the wider research community of the utility of these methods as well as identify areas where efforts need to be made for the development of new or additional methods to determine complex disease trajectories and pathways. 

Applications

There is currently a lack of detailed clinical evidence of the long term health consequences of a heart attack. Determining the appropriate tools and methods to tackle this challenge has the potential to impact upon many aspects of healthcare. Detailed insights into disease trajectories can provide information for commissioners for use in future health care services planning. Moreover, it will inform  clinicians and their patients of the most likely disease trajectories, thereby providing the opportunity to determine the best care pathways and interventions to prevent or interrupt the most fatal disease trajectories to improve patient lives. 

Organisers

Contact info

Dr Marlous Hall
[email protected]

Funders

Collaborators