NAIROBI, Kenya – Artificial intelligence could soon forecast people’s health years ahead, predicting the likelihood of more than 1,000 diseases based on medical records, researchers have revealed.
In an article by BBC, the new model — Delphi-2M — has been likened to a “weather forecast” for human health.
Instead of predicting exact dates of illness, it calculates the probability of conditions such as heart attacks, type 2 diabetes, and sepsis over time.
“We can now say there’s a 70% chance of disease in the same way we say there’s a 70% chance of rain,” said Prof Ewan Birney, interim executive director of the European Molecular Biology Laboratory, which co-led the project. “And we can do that not just for one disease, but for all diseases at the same time.”
How it works
Delphi-2M uses similar technology to popular AI chatbots, which are trained to predict sequences of words.
Instead, the model analyses patterns in anonymous health data — hospital admissions, GP records, and lifestyle habits — to anticipate disease progression.
The system was developed using data from more than 400,000 participants in the UK Biobank project and tested against the records of 1.9 million people in Denmark.
According to Prof Birney, the results held up well: “If our model says it’s a one-in-10 risk for the next year, it really does seem to turn out to be one in 10.”
Potential uses
Although not yet ready for clinical use, scientists hope the AI tool could one day help:
- Identify high-risk patients early so doctors can intervene with medication or lifestyle advice.
- Inform disease screening programmes by showing who is most at risk.
- Plan healthcare resources by forecasting how many cases of a disease may occur in specific areas years ahead.
“This is the beginning of a new way to understand human health and disease progression,” said Prof Moritz Gerstung, head of AI in oncology at the German Cancer Research Centre. “Generative models such as ours could one day help personalise care and anticipate healthcare needs at scale.”
Challenges ahead
The researchers, writing in Nature, acknowledge the technology is still experimental and needs refining.
Because the UK Biobank data is drawn mostly from people aged 40–70, the model may not yet represent the wider population. Work is underway to expand it with genetics, imaging, and blood analysis.
Prof Gustavo Sudre, a neuroimaging and AI researcher at King’s College London, said the findings were a “significant step towards scalable, interpretable, and — most importantly — ethically responsible predictive modelling in medicine.”
Prof Birney cautioned that strict regulation and testing will be needed before the system can be used in clinics.
But he believes it could follow the same trajectory as genomics, which took a decade to become a routine part of healthcare.
