Overview
Current activity

The majority of my current work is in the field of applied AI, as part of the Explainable AI in Healthcare Group at the Digital Foundry (White City Campus), with Professor Alessandra Russo (Computing) I am involved in the following areas, particularly where they involve commercial partners and translation into healthcare.

1. Learning from multimodal data using transformer-based machine learning and knowledge graphs to drive AI for diagnosis in primary care. Creating hybrid explainable Neurosymbolic AI.  AI-DIP project in partnership with QMUL.

2. Medical device reggulatory approval study of  pre-GP access virtual AI-driven history taking apps for capturing symptoms and signs from patients. AI-DIP project in partnership with SOAP Health.

3. Natural Language Processing of EHR text data to construct knowledge graphs, including the use of Large Language Models.

4. Answer Set Programming to provide explainable guidelines. ICHT BRC.

Past work:

My work is primarily focused on research around the concept of ‘The Learning Healthcare System’ (LHS). The LHS recognises that in order to create a safe, effective and efficient healthcare system, it is necessary to move beyond the compartmentalisation of ‘Patient Safety’, ‘Evidence-based practice’ and ‘Translational Medicine’. The LHS envisions the harnessing of routine healthcare processes and systems (especially electronic health record systems (EHRs), to create a continuous cycle of improvement, routine data (including randomisation at the point of care) being processed into knowledge and applying that knowledge to achieve improved and safer practice. The LHS encompasses a number of areas that have in the past tended to remain separate; these include decision support and the study of medical error, extraction and manipulation of routine health data, integration of clinical research functions into clinical systems and methods for providing practice recommendations at the point of care. Research into these activities require staff with a wide range of expertise, clinical, psychology, human computer interface and design, computer science/programming and informatics. Although the group has grown from the application of the LHS in Primary Care, the relevance of this work stretches beyond just Primary Care and into all evidence-based medical disciplines. The group has developed through a combination of core expertise, project-funded staff and a very wide international collaboration. We seek funding for both large programmes and smaller projects around the LHS theme, where synergies can be built across several distinct sub-themes as below:
Clinical Research Informatics

Research informatics, and epidemiological research using electronic health records, represent a major focus of research. Our current research is active in linking and analysing large datasets from electronic records, as well as developing their application to clinical and community trials. We led a large European Union FP7 Integrated project (TRANSFoRm, www.transformproject.eu, Scientific Director, Delaney), that includes Imperial College, University of Dundee and CPRD (Clinical Practice Research Datalink) amongst its 21 partners. TRANSFoRm developed an informatics infrastructure for the Learning Healthcare System in General Practice, including phenotype-genotype studies, randomised controlled trials and diagnostic decision support, all built into primary care electronic health record systems. We have guided significant in-house developments in specialist electronic health records, via the NIHR Health Informatics Collaborative, with the establishment of data integration, archiving, search and retrieval tools for both structured/coded data and text.

Collaborators

The group’s international collaborations are wide and varied but largely take the form of collaborative grant acquisition.

UK: We formed part of a highly successful clinical research partnership in infectious disease in primary care, led from Bristol and Cardiff. Very large observational studies aimed at developing prediction rules for risk stratification in childhood urinary tract infection (DUTY-HTA) and cough (TARGET- NIHR Programme Grant) have been conducted in several hundered practices, using informatics tools developed and hosted at KCL by the group, and in both projects London practices were the lead recruiters.

Europe: We led a major EU programme in translational medicine and patient safety (FP7 TRANSFoRm, 22 partners, Delaney) 2009-15. This EU project contained a number of critical UK collaborations, most particularly with Univ of Dundee (and hence health Data Research UK), Univ of Warwick, and Imperial College. Within the EU, major partners were the Karolinska Institute, University of Geneva, Univ of Antwerp, and Univ of Dusseldorf.

International: We worked with several US institutions (Prof Charles Friedman, Univ Michigan and Prof Jyoti Pathak, Weill Cornell) around the concept of the ‘Learning Healthcare System).

Commercial partners: Working with commercial vendors of EHR systems and other parties such as ‘information brokers’ and suppliers of decision support systems is essential in translating academic work into practice in a sustainable way. Knowledge is a commodity in the modern healthcare system and resource is needed to create, deploy, translate and curate it.