A new study suggests artificial intelligence (AI) could hold the key to predicting future patient health conditions. The newfangled technology, researchers claim, can play a crucial role in helping doctors monitor patients or make decisions around diagnosis.
Foresight, The AI tool, belongs to the same family of AI models as ChatGPT, which is trained on publicly available information. In contrast, Foresight leverages the vast NHS electronic health records dataset to make its predictions.
A team of researchers from King's College London (KCL), University College London (UCL), King's College Hospital NHS Foundation Trust and Guy's and St Thomas' NHS Foundation Trust developed Foresight.
A new study led by @KingsCollegeLon highlights the potential of a generative #AI tool, Foresight, to predict the health trajectory of patients by forecasting future disorders, symptoms, medications & procedures.
— Institute of Psychiatry, Psychology & Neuroscience (@KingsIoPPN) March 21, 2024
Read the full story➡️https://t.co/03zGorFp2Z#IoPPNNews @zeljkokr pic.twitter.com/o8EZ3mOtZZ
Researchers trained three variations of the AI using data from over 811,000 patients to develop Foresight. This data came from two London NHS trusts and a publicly available US dataset. Based on their medical records, Foresight AI was then challenged to predict ten potential health conditions a patient might develop.
The AI tool's accuracy varied depending on the data source. For instance, it correctly identified potential health conditions between 68 per cent and 76 per cent of the time while analysing NHS records. However, the tool achieved an impressive 88 per cent accuracy in predicting future illnesses using a US dataset.
"The proposed purpose of Foresight is not to enable patients to self-diagnose or predict their future, but it could potentially be used as an aid by clinicians to make sure a diagnosis is not missed or for continual patient monitoring for real-time risk prediction," Zeljko Kraljevic, first author and Research Fellow in Health Informatics at King's IoPPN said.
Kraljevic, lead author of the study published in The Lancet Digital Health, highlights that Foresight's precision in predicting patient health journeys suggests its potential value in aiding medical decision-making and clinical research.
The potential and pitfalls of AI health prediction
One of the main advantages of the AI tool, according to Kraljevic, is that it can quickly scale to more patients, hospitals or disorders without any significant modifications. "The more data it receives, the better it gets," he said.
Kraljevic also highlighted Foresight's ability to predict patient health trajectories with high precision, as confirmed by assessments by medical professionals. Doctors created 34 mock patient cases with potential health events to check real-world applicability. The tool's predictions aligned with the doctors' forecasts in 93 per cent of these cases.
@Cogstack's #GenerativeAI AI Foundation Model of Patients published in @LancetDigitalH: https://t.co/YfxT557lr2
— Prof James Teo (@jthteo) March 21, 2024
Unlike #LLM's that simply predict the next word, Foresight forecasts possible futures for patients, representing possible Multiverses to understand diseases /1 pic.twitter.com/WIfqYKR22b
Professor Richard Dobson, senior author, Professor of Medical Informatics at King's IoPPN and Theme Lead for Informatics at the NIHR Maudsley BRC, said: "It is an exciting time for AI in healthcare, and to develop effective tools, we must ensure that we use appropriate data to train our models and work towards a shared purpose of supporting healthcare systems to support patients."
Foresight isn't the only AI tool making a difference in healthcare and other industries. For instance, Sam's Club is reportedly gearing up to introduce a new AI-powered checkout system. However, the rise of AI also raises concerns about job displacement. Earlier this month, Klarna laid off 700 customer service agents, citing its AI chatbot's ability to manage the workload.
