AIMS
The project aims to investigate whether a tool can be built using a specific computer approach called “reinforcement learning” to help healthcare professionals make better decisions when caring for people with Community Acquired Pneumonia.
BACKGROUND
Community-acquired pneumonia (CAP) is a common and severe lung infection, especially among the older population. In the UK, CAP is becoming more common and sadly up to 15% of patients admitted to hospital with CAP die within 30 days.
The severity of CAP varies: some patients can be safely discharged with oral antibiotics, whilst others will be critically unwell, requiring intensive care admission.
The treatment of patients with CAP is complex. Key decisions relate to the antibiotics used, the way antibiotics are given (in a tablet or by a drip) and the place of care (home, hospital and in hospital, a normal ward or intensive care).
RESEARCH
Artificial intelligence uses computer systems to perform tasks that require intelligence, such as understanding images, recognising speech and decision-making. Machine learning is a branch of artificial intelligence where systems learn to perform tasks without being told (programmed) exactly how to do the tasks.
A particular type of machine learning is Reinforcement Learning (RL). This looks at scenarios where multiple decisions need to be made and where decisions may affect other decisions.
The aim of this project is to train systems using RL to make the best possible decisions when people come to hospital with CAP, using the routine health data captured from the admissions (such as blood pressure readings, oxygen levels and blood results).
The project will provide important insights into how RL might be used more broadly within the healthcare setting and in the specific case of treatment of CAP.
PATIENT INVOLVEMENT
The researchers have been working with the patient-led Clinical Ambassador Research Group based at Heartlands Hospital.
They seek valuable patient/public input into their research design and are hoping to co-develop ways of explaining RL in healthcare to patients and public.
APPROVAL
This project was supported unanimously by the PIONEER Data Trust Committee.
This work is led by Professor Giovanni Montana, Professor of Data Science at University of Warwick.
