AI model can predict risk of more than 1,000 diseases up to two decades ahead

A new artificial intelligence tool developed by researchers at the European Molecular Biology Laboratory, the German Cancer Research Centre and the University of Copenhagen can forecast a person’s risk of developing more than 1,000 diseases, including cancer, up to 20 years into the future. Named Delphi-2M, the model aims to identify high-risk patients early and guide preventive care, the researchers said in the journal Nature.

Delphi-2M analyzes anonymized medical records for patterns of “medical events” in a patient’s history alongside lifestyle factors such as smoking, alcohol use and weight to forecast health trajectories over the next decade and beyond. The approach relies on recognizing sequences of health events that precede serious disease and translating those patterns into probabilistic forecasts for future years. Researchers described the model as designed to predict multiple diseases at once and across long time horizons, rather than focusing on a single condition.

The model was trained and tested on patient data from about 400,000 participants in the UK Biobank study and 1.9 million people in the Danish national patient registry, enabling researchers to gauge its ability to forecast long-range outcomes and potential shifts in demand for screening and treatment. Researchers say the technology could eventually be used in primary care to help clinicians identify high-risk patients and offer targeted interventions before a disease takes hold. In a keynote framing of the approach, Professor Ewan Birney, executive director of the EMBL, described a scenario in which a clinician could say: “Here are four major risks that are in your future and here are two things you could do to really change that.” He added that weight loss and smoking cessation could appear as standard recommendations in a patient’s data, informing personalized counseling rather than broad, one-size-fits-all advice.

While the model shows promise for diseases with clear progression—such as type 2 diabetes and heart disease—it is less precise for incidents with more stochastic origins, such as certain infections. Still, researchers contend that the Delphi-2M framework could help anticipate healthcare demand at scale, aiding planning for services, screenings and interventions across populations.

Professor Moritz Gerstung, an expert in computational cancer biology at the German Cancer Research Centre, said the work marks “the beginning of a new way to understand human health and disease progression.” He added that the model could also inform disease-screening programs by projecting, for instance, how many people in a city may experience a heart attack in the coming year, helping health systems allocate resources more efficiently.

But several experts urged caution about drawing causal conclusions from the model’s forecasts. Professor Justin Stebbing, a consulting oncologist at Anglia Ruskin University, warned that the tool may reproduce biases present in its training data, including healthy-volunteer and selection biases. Professor Peter Bannister, a healthcare expert and fellow at the Institution of Engineering and Technology, said that even as a promising research concept, the immediate challenge is ensuring robust digital infrastructure and equitable access so that the benefits reach all populations, not just those with better data coverage or resources. “There is a long way to go before this translates into improved healthcare for everyone,” he said.

Proponents of AI in medicine also stress the need to expand the data inputs beyond anonymized clinical records. Professor Gustavo Sudre, an expert in neuroimaging and AI at King’s College London, argued that the current system relies on clinical records but was designed with future growth in mind. He noted that incorporating biomarkers, imaging data and genomics could make the Delphi platform a multimodal, precision-medicine tool capable of adapting to richer data sources as they become available.

The study arrives as advocacy groups warn that cancer incidence could reach record highs by 2040. One Cancer Voice has projected that cancer cases will rise sharply, with cancer detected in a patient roughly every two minutes at peak. The forecast highlights the most common cancers—breast, prostate and lung—along with a projected rise in cases among children and young people, with more than 63,000 pediatric and young-adult cases anticipated. While these projections underscore the scale of future demand for screening and treatment, researchers say tools like Delphi-2M could help preemptively tailor screening programs and resource planning if deployed with careful oversight and strong data governance.

Experts emphasize that the Delphi-2M system is not a stand-alone diagnostic tool but a decision-support technology intended to augment clinical judgment. Its developers stress that real-world use will require rigorous validation across diverse populations, transparent reporting of uncertainties and ongoing monitoring for biases. They also note that the technology’s ultimate value will depend on the healthcare ecosystem’s ability to offer timely, equitable interventions to those identified as high risk, including access to preventive care, lifestyle support and appropriate screening.

As researchers continue to refine the model and expand the data types it can ingest, Delphi-2M represents a notable step in a broader shift toward AI-assisted prediction in medicine. If integrated thoughtfully, the system could help clinicians identify early warning signs, optimize preventive strategies and better anticipate demand for healthcare services—an aim at the heart of modern technology-enabled precision health.