Much NHS patient data is recorded in an unstructured form that is descriptive. It is a vast asset but difficult to collate and interpret. Overcoming this challenge offers the opportunity to quickly and accurately answer an enormous variety of healthcare-related questions.
CogStack has been developed (with support from HDR UK) by the PHIdatalab team at NIHR Maudsley BRC and NIHR UCLH BRC led by Richard Dobson – an HDR UK associated researcher and Professor of Health Informatics at KCL and UCL with clinical lead James Teo, Clinical Director of Data and Ai at KCH. It uses best-of-breed enterprise search, natural language processing (NLP), analytics and visualisation technologies. Advanced NLP technology allows it to read and understand unstructured records, such as clinicians’ case notes.
CogStack has created one of the largest NHS trained language models and is system agnostic, so can be widely used. It operates in near real-time and is user-friendly, allowing data to be accessed by asking straightforward questions.
It can be used for everything from large-scale research and business intelligence to planning patient personalised care.
Spread and use
Accelerated by the need for real-time data, the open-source CogStack platform has been deployed at King’s College Hospital NHS Foundation Trust and South London and Maudsley NHS Foundation Trust (SLaM), Guy’s and St Thomas’ NHS Foundation Trust and University College London Hospitals NHS Foundation Trust.
Fast follower adopters include international health and care providers and academic groups such as Monash Partners Academic Health Science Centre, Australia who were awarded AUS$1.92m for a local deployment.
Impact and outcomes
CogStack is allowing the NHS to implement new data mining techniques. Some 17 million free text documents and over 250 million diagnostic results and reports have already been processed at King’s College Hospital alone.
- Improving the safety of prescribing: identifying rheumatology patients taking methotrexate with abnormal pathology results suggesting potential renal failure
- COVID-19 pandemic insights:
- Demonstrating ACE inhibitors/ARBs do not increase risk of severe outcomes (Bean et al 2020)
- Evaluating and improving ‘early warning’ scores for hospitalised patients (Carr et al 2021)
- Investigating links between ethnicity, pre-existing health conditions and COVID-19 (Teo et al 2020)
- Hospital symptom trackers to detect future surges (Teo et al 2021).
- The findings informed the UK Government’s SAGE committee, the HDR-UK team and Impact of the Year finalists.
- Services/coding and audit: streamlining coding for orthopaedic procedures led to 20% increase in procedures identified (£1,26 million per annum hospital revenue).
Recognition of CogStack
The importance of the platform was recognised in Professor Dame Sally Davies’s eighth report to government as Chief Medical Officer (CMO) and the use-case of an academic-developed NHS-deployed open-source technology has been recognised as a flagship case study in the NHSX AI report, NHS Tech Plan and a number of several keynote speeches by Secretary of State for Health and Social Care. Major funding has been awarded by the Office for Life Sciences, InnovateUK and NHSX to support adoption at other healthcare organisations and development of a model for commercialisation.
Matt Hancock MP, who was then Secretary of State for Health and Social Care, highlighted CogStack during his keynote address to the Health Tech Alliance, describing it as a system that “can perform manual coding and data collection tasks in a tenth of the time that it takes a human analyst”.
Biobank conversion allows greater UK contribution to international research
21 February 2023
A team of researchers processed and converted over 1.3 billion rows of UK Biobank data to the Observational Medical Outcomes Partnership (OMOP) common data model (CDM), improving its usability for...
Paving the way for quicker, simpler and less expensive use of scans to diagnose heart damage
20 February 2023
Researchers have applied machine learning to improve the detection of heart damage through cardiac magnetic resonance (CMR) scans.