What comes next for your health, and when?
AI is transforming medicine. With more and more accuracy, it’s helping doctors diagnose patients. Our medical history offers valuable insight into potential health issues. But what if AI could reliably estimate our next diagnosis, complication or even the timing of death? A research team from the European Molecular Biology Laboratory (EMBL), the German Cancer Research Center (DKFZ) and the University of Copenhagen created an AI model called Delphi-2M that’s able to predict medical diagnoses over a decade ahead. The innovation is presented in the journal ‘Nature’(opens in new window).
What your clinical future looks like
Delphi-2M is capable of assessing risk for more than 1 200 conditions, including cancer, diabetes, heart disease and respiratory illnesses. It’s less reliable when it comes to more random conditions, such as mental health disorders and pregnancy. It doesn’t calculate exact dates, but rather estimates the likelihood of diseases. Unlike ChatGPT and similar AI chatbots, Delphi-2M predicts outcomes, not words. Medical events usually follow patterns that are predictable. It learns those patterns in order to forecast future health outcomes. Delphi-2M in a way provides a health forecast, similar to a weather app. “So, just like weather, where we could have a 70% chance of rain, we can do that for healthcare,” EMBL interim executive director Ewan Birney told the ‘BBC’(opens in new window). “And we can do that not just for one disease, but all diseases at the same time – we’ve never been able to do that before. I’m excited.” The researchers trained Delphi-2M on data from the United Kingdom’s Biobank – a huge biomedical database with information on about half a million participants. They demonstrated Delphi-2M’s performance by testing it against data from nearly 2 million people in Denmark’s public health database. “Our AI model is a proof of concept, showing that it’s possible to learn many of our long-term health patterns and use this information to generate meaningful predictions,” Birney commented in a DKFZ press release(opens in new window). “By modelling how illnesses develop over time, we can start to explore when certain risks emerge and how best to plan early interventions. It’s a big step towards more personalised and preventive approaches to healthcare.”
A health story unfolding over time
Delphi-2M accepts the patient’s previous health history as a starting point. It then predicts the probability of the next health event in their life and the time to reach that event. “Just as large language models can learn the structure of sentences, this AI model learns the ‘grammar’ of health data to model medical histories as sequences of events unfolding over time,” explained Moritz Gerstung, head of the Division of AI in Oncology at DKFZ. “This is the beginning of a new way to understand human health and disease progression,” concluded Gerstung. “Generative models such as ours could one day help personalise care and anticipate healthcare needs at scale. By learning from large populations, these models offer a powerful lens into how diseases unfold, and could eventually support earlier, more tailored interventions.” Delphi-2M isn’t ready for clinical use just yet. But do we even want to know when our end will come? And do we want machines to be the ones to tell us?