The AI Award team spoke to Dr James Teo, Clinical Director at AI and Data, KCH and GSTT and Prof Richard Dobson, Health Informatics, KCL and UCL about the health record reading technology, CogStack.
What is the technology?
CogStack is a technology that aims to provide an efficient way to read and code clinical records in the NHS. A team at King’s College London, King’s College Hospital and University College London developed CogStack and this NHS home-grown technology was highlighted by the Secretary of State for Health in his Healthtech Alliance speech in 2020.
What are the challenges CogStack is aiming to address?
Electronic health records are important for patient care. Researchers regularly study large numbers of health records to find ways to improve care. Clinicians often add information to health records in the form of narrative. Narrative text is unstructured making it very difficult to analyse in large volumes as each individual health record must be carefully read and standardised codes are assigned to the record based on particular words, conditions or treatments. This is called “Clinical Coding” and is normally done by ‘Clinical Coders’.
Clinical coding is a vital part of nearly all delivery of care, planning and service improvement for healthcare: from delivering specific health services to studying the health of the public.
As clinicians add more information to health records, the process of ‘clinical coding’ becomes more time-consuming, expensive and the risk of mistakes increases. Coding mistakes are important: they can underestimate the need for some clinical services and lead to incorrect conclusions in research.
How does CogStack address these challenges?
There is an alternative: we could teach computers to ‘read’ electronic health records, to summarise and suggest the standardised codes for clinical coders to verify.
By reducing labour-intensive tasks for NHS staff and providing automated clinical coding that is fast, efficient and accurate, CogStack could provide researchers with better quality health record information resulting in improved delivery of care.
The project will assess the magnitude of these potential efficiency savings and improved performance. People and resources freed up by CogStack will be able to improve many other activities across the NHS, such as health services delivery, patient safety, public health and clinical research.
What type of AI is used?
CogStack uses a type of Artificial Intelligence called Natural Language Processing (NLP). NLP has already been developed to read languages in other industries, but the technology needs tailoring to healthcare as medical records are filled with medical terms, jargon, acronyms, and local dialects.
What is the AI Award supporting you to achieve?
This project will build on the CogStack technology to establish a more efficient way to read and code records. This will involve working with staff, patients and members of the public at five NHS Trusts: Kings College Hospital, South London & Maudsley Hospital, Guys & St Thomas Hospital, University Hospital Birmingham and University College London Hospitals.
Patients and the public are central to this project and informed its design. A coordinated, cross-site patient and public involvement group will provide input to the CogStack learnings, building a patient involvement toolkit, ensuring public members are equipped to review NLP projects, and can embed patient priorities into the future use of this tool.
What’s next for the technology?
Through our partnerships the project will test and improve our NLP models across our partner Trusts, demonstrate benefit through our clinical coding use cases and develop a blueprint for adoption at other healthcare organizations.