'AI trained on NHS data 'may predict ill health in advance''

 An artificial intelligence model that is being trained on de-identified data from 57 million people in England has the potential to predict future health risks and help improve healthcare, according to its developers.

The AI – known as Foresight – is being put through its paces in a pilot study carried out by researchers at University College London (UCL) and King's College London (KCL) and uses a generative AI (GenAI) approach similar to ChatGPT.

The hope is that it will allow a national predictive healthcare approach that can identify people who may be at high risk and open a window of opportunity to intervene earlier and save lives, which would be a help to the NHS's focus on disease prevention as one of the pillars of reforms alongside moving care from hospitals to the community and embracing digital technologies.

It also ties into the UK's ambition to harness health data for medical research, backed by a £600 million ($764 million) investment by the government announced last month.

"AI models are only as good as the data on which they're trained," commented Dr Chris Tomlinson, one of the lead researchers from UCL.

"So if we want a model that can benefit all patients, with all conditions, then the AI needs to have seen that during training," he added. "Using national-scale data allows us to represent the kaleidoscopic diversity of England's population, particularly for minority groups and rare diseases, which are often excluded from research."

Ultimately, the hope is that Foresight will be able to predict hospitalisations, heart attacks or a new disease diagnosis, according to the researchers, who believe this is the first time that an AI has been trained on health data on a national scale.

The data used to train the AI – including vaccination records, GP visits, hospital admissions, and A&E attendances – is being provided under the NHS England Secure Data Environment (SDE) to ensure patient privacy and can only be run on NHS computer systems.

The privacy measures controls were welcomed by Dr Luc Rocher, senior research fellow at the Oxford Internet Institute (OII) at the University of Oxford, who cautioned, however, that "the very richness of data that makes it valuable for AI also makes it incredibly hard to anonymise."

"These models should remain under strict NHS control where they can be safely used," he said.

Meanwhile, Dr Wahbi El-Bouri, senior lecturer in digital twins and in silico trials at the University of Liverpool, said that while this sort of data is vital to help improve health outcomes in the UK, "developing these AI models requires good quality data."

He added that "researchers who have worked with NHS data will know that data quality can often be poor, with large amounts of missing data, or incorrect reporting," and warned that "NHS data is the wrong type of data to tackle prevention as when someone has visited the NHS it is because something is already wrong."

Another researcher working on the project, KCL's Prof Richard Dobson, acknowledged that currently the data in the pilot is "broad but shallow, and ultimately we'd like to harness the expertise and AI platforms behind Foresight by including richer sources of information like clinicians' notes, or results of investigations such as blood tests and scans if they become available."

https://pharmaphorum.com/news/ai-trained-nhs-data-may-predict-ill-health-advance