Improving early recognition, management, outcome prediction, and antimicrobial prescribing for sepsis, using a Bayesian statistical learning and prediction model

Supervisors: Professor Andrew Dowsey, Bristol Veterinary School Dr Chris Bordeaux, NHS, Centre for Medical Education Dr Chris McWilliams, Department for Engineering Mathematics Dr Ranjeet Bhamber, Bristol Medical School  Email: e.barker@bristol.ac.uk Twitter: @gangly_ed

Research Project Summary:

My current research interests are in the use of integrating electronic health record information with wider data sources to aid the creation of clinical decision support systems, using machine learning to develop novel insights, digital phenotyping, earlier recognition, and risk scores. Specifically, my PhD will concern the development of a clinical decision support system to assess the risk of a bacterial infection pertinent to sepsis, to identify the pathogen causing this sepsis from electronic health record data integrated with primary & secondary care data, laboratory records, and pharmacy records, and to then prescribe an appropriate antibiotic, to both improve patient outcomes, and reduce antimicrobial resistance through antimicrobial stewardship. This will be achieved through the use of machine learning, in particular Bayesian networks. Machine learning methods will be utilised to address clinical interpretability, missing data, and longitudinal time series. I am also collaborating with the Personalised National Early Warning Score project and the Antimicrobial Resistance team, as well as the Laboratory Markers of Covid-19 severity study.

General Profile: 

I graduated in July 2019 with a degree in Artificial Intelligence (1st Class Hons) from the University Of Liverpool. This provided a broad base of knowledge in AI, including bio-computation, multi-agent systems & robotics, modal logics, expert systems, social networks, as well as a foundation in many traditional computer science concepts. Formal research experience began in Summer 2018 with an EPSRC research bursary, working with convolutional neural networks to help explain deep-learning based image classification. Here I developed a better practical understanding of machine learning, working with scikit-learn, TensorFlow and python data science libraries. My dissertation project in mid-2019 worked in the field of legal reasoning. This involved an expert system to present formal arguments for and against a case. Based on pre-existing work called the ANGELIC methodology, answers to questions corresponded to legal base factors, and these were structured in a weighted abstract dialectical framework to calculate formal arguments.

Interests: I like wild swimming, running, rock climbing, woodwork and pasta making!