Most of the information in hospital health records is not recorded in a way that can be easily analysed. This makes it difficult for hospitals to plan their services or for doctors to get a complete overview of a patient. CogStack uses artificial intelligence to analyse text records and make them accessible for hospitals to improve care.
Every time a patient is seen in hospital, codes are added to their health record indicating diagnoses or procedures. These codes are recognised across healthcare providers and are used to plan funding to support services where there is a need.
However, it is estimated that only 20% of the information in health systems is entered in this ‘structured’ way. The vast majority of information is stored in an ‘unstructured’ way, for example, in doctors’ notes, letters, reports. This includes vital information but it cannot be easily analysed. In addition, health data can be locked away on different systems across the hospital, making it difficult to access.
Specialists in health records called clinical coders read the medical notes of each person and select the corresponding codes out of 70,000 options. As with any manual labour-intensive process, clinical coding can be inefficient leading to undercounting of conditions. This, in turn, can affect healthcare planning and funding such as too few diabetic nurses being available to care for patients.
Researchers at University College London and King’s College London and NHS partners including South London and Maudsley NHS Foundation Trust and Guy’s and St Thomas’ NHS Foundation Trust have developed a computer program called CogStack that uses artificial intelligence to make health records more useful for hospitals.
CogStack is compatible with various databases across healthcare, so it can access the necessary information and standardise it to create an index that is easier for clinicians to search.
The program can also analyse doctors’ notes and letters to recognise their meaning and accurately assign the correct code. It was designed to be run by NHS hospitals to screen health data very quickly and can help can help people code large quantities of records more efficiently. coders to be more efficient and cover more records.
CogStack is completely secure as it does not leave the hospital electronic system. Discussions with the public and other stakeholders found that there was strong support for this project and allowed the team to address any privacy concerns.
Professor Richard Dobson at University College London, co-lead on the project, said:
“Imagine if we had computers but no internet and how limiting that would be. The hospital has all these computers with data on them; once you connect them and make it possible to search across all the systems, it then becomes really useful.
“The public was quite surprised that clinicians can’t already access all patient medical data. So, they were keen to ensure clinicians were able to do this to improve the care they provide.”
Having this data has provided many benefits to hospitals, like helping to predict surges of COVID-19 cases, identifying bottlenecks in patient journeys through the system and facilitating recruitment to clinical trials. The system can also be used to flag patients who are at risk should they have a particular combination of symptoms or show signs they are suffering side effects from their medication.
CogStack helped one NHS Trust in 2022 to maintain services following a data hack that left electronic health records inaccessible. The system offered clinicians access to data, helping to reduce the damaging consequences for patient care.
CogStack was recognised by several prestigious national and international awards and was included in reports by the Chief Medical Officer, an NHS AI report, the NHS Tech Plan and keynote speeches by the Health Secretary.
UK health data landscape review commissioned
31 May 2023
The UK’s ability to use data to improve health care is to be put under the microscope in an independent review commissioned by the Chief Medical Officer for England and UK Government’s Chief...
Multi-generational households, living alone and ethnicity – older people and the risk of severe COVID-19 outcomes
21 February 2023
Researchers carried out an analysis of older people and the risk of severe COVID-19 outcomes using the OpenSAFELY research platform
Digital device link to health records could be key to answering major health challenges
8 February 2023
Health insights gained from smart tech and wearable devices could be bolstered dramatically by safe and secure linkage to electronic health records, according to researchers at the BHF Data...