King’s Health Partners informatics tool CogStack featured in NHSX report on AI
Dr James Teo, Clinical Director of Data Science and Consultant Neurologist at King’s College Hospital NHS Foundation Trust and King’s Health Partners [pictured below, right], gives some background on the project.
Pioneering teams across the NHS are leading the way in using open source data technologies to transform patient care. NHS trusts are starting to demonstrate that it is now possible to get meaningful insight from data, that can result in significant savings and improved patient care. A recently published report by NHSX, ‘Artificial Intelligence: How to get it right’, showcases one such King’s Health Partners innovation – CogStack.
Making sense of the data
The huge volume of data created by clinicians every day highlights the need for a consolidated, rationalised and easily accessible repository. As with many other trusts, the data at King’s College Hospital NHS Foundation Trust is made up of vast numbers of electronic patient records, structured data, such as pathology reports, blood pressure readings, etc, and unstructured data, for example physicians’ handwritten notes.
Finding ways to accurately index and connect this type of data has always been a challenge. The sheer volume of data available prompted the Precision Medicine team at King’s College Hospital, led by Dr Anwar Al-Haq and Professor Richard Dobson’s Health Informatics group from the Department of Biostatistics and Health Informatics, the Centre for Translational Informatics (CTI), and the National Institute for Health Research Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London (NIHR Maudsley BRC), to innovate and extract value from clinical data.
In 2016, through clinic-academic collaborations across King's Health Partners and later as part of Health Data Research UK London, the teams developed and deployed a new, wide-ranging open source healthcare and search analytics tool – CogStack – which has enabled King’s College Hospital to pool 20 years’ of clinical data, including physicians’ handwritten notes.
CogStack is an information retrieval and extraction platform. It implements open-source enterprise search, natural language processing, analytics, and visualisation technologies to unlock the health record and assist in clinical decision making and research.
Electronic health record systems are often closed, proprietary and contain incomplete and unstructured data. This results in the wealth of information potentially available within health records often being inaccessible and underused.
In addition to tools built by the NIHR Maudsley BRC and CTI, CogStack also uses components from major open-source platforms, like the Elastic Stack, and open Elastic’s X-Pack extensions for security, alerting and monitoring. It works by consolidating and transforming all current and historical patient data - electronic records alongside written notes - into valuable insight and delivered in real-time to improve patient care. A detailed description of the technology and deployment can be found in CogStack - Experiences Of Deploying Integrated Information Retrieval And Extraction Services In A Large National Health Service Foundation Trust Hospital.
By implementing new data mining techniques within NHS trusts, it has the ability to search any clinical data source (unstructured and structured), and natural language processing (NLP) applications, developed to automate information extraction of medical concepts.
These tools allow clinical text to be searched for specific terms using simple or complex syntax, rapidly retrieving the data needed to answer complex queries such as “has this patient received any high cost treatments that have not been captured in their discharge summary?” or “provide me a list of females of childbearing age who may have been exposed to a potentially toxic teratogenic drug”.
So far, more than twelve million free text documents and 250 million diagnostic results and reports have been processed within CogStack. The speed at which queries can be developed and results returned and refined is very powerful, this allows for easier identification of patients with certain diseases or symptoms, and lets clinicians proactively intervene in care.
One example of how it is being used was in a fracture outpatient clinic, where the CogStack team has piloted the software to identify under-coding. The pilot enabled the team to triple their depth of coding within a month (from ~10% of cases to 30% of cases having procedures recorded accurately). This translated to £1,260,000 of financial activity per annum. This work was presented at the NHS Expo in September 2019.
NHS homegrown AI technologies
This type of work and the value it can provide is one of the reasons CogStack was highlighted in NHSX’s recent ‘Artificial Intelligence: How to get it right’ report as it lays the foundation for further infrastructure for AI development. CogStack is one of the core technologies for Health Data Research UK London, the London Medical Imaging AI Centre for Value-Based Healthcare and the NIHR Applied Research Collaboration (ARC) for South London.