Researchers have developed an AI model named Delphi-2M that predicts the risk of over 1,000 diseases up to 20 years before symptoms manifest. Published in Nature, the model achieves 76% accuracy for near-term predictions and maintains a 70% accuracy for long-term forecasts. Trained on data from 402,799 participants in the UK Biobank and validated with 1.9 million Danish health records, it distinguishes itself by assessing risks across multiple diseases simultaneously. Delphi-2M processes medical histories akin to language models, using recorded diagnoses and lifestyle factors to forecast health trajectories. It can simulate numerous health futures, providing data on population-level disease burdens. Though promising, limits exist; the model's accuracy decreases significantly for predictions extending beyond 10 years, especially for rare diseases and those influenced heavily by environmental factors. Privacy concerns and the necessity for validation across diverse populations remain substantial challenges before clinical deployment.

Source đź”—