Detecting myocardial infarction
When high-risk patients are admitted into ICU, they have a 1 in 4 chance of having a heart attack during their admission. Yet 95% of these heart attacks are missed — partly because critically ill patients do not have symptoms typical of heart attacks, and partly because diagnostic tests are only done intermittently, typically daily. For the first time using continuous ICU monitoring, our aim is to identify heart attacks systematically.
Clinical phenotypes
Most adults who are admitted to ICU have at least one if not more chronic conditions, yet historically researchers have failed to take this into account when designing trials. Treatments that work in one group of patients may cause harmful side-effects in another. We will explore the relationship between chronic conditions and acute illness in critically unwell patients, along with further information that we acquire during their stay to understand how these impact on organ dysfunction and longer-term outcomes. We will use cutting edge causal inference techniques that will help us explore whether the outcome is caused by the exposure, or if an apparent association is caused by a different exposure.
Trust in AI
This information will help us to develop, deliver and evaluate tools to predict heart attacks before they happen, through using datasets of patients admitted to ICUs in Scotland, and test them in datasets of patients admitted to ICUs in England and Pittsburgh, USA. We ultimately aim to embed these algorithms within the NHS so that we can trial interventions in high-risk patients to prevent heart attacks from happening.