A medical record is like a long diary: doctor visits, diagnoses, prescriptions, lab results. A human doctor reads it cover to cover to understand what's going on with the patient. Now, a machine reads this diary, and does it in seconds.

The developers created a model called Foresight-E. It was trained on de-identified records of about 61 million people: that's nearly the entire population of England. The model works like a text predictor: given past events, it predicts the next one. But instead of words, it uses medical codes: diseases, procedures, medications. In total, its vocabulary includes about 40,000 concepts.

What sets this model apart is that it predicts 30 days into the future. For example, the likelihood of hospitalization with COVID-19 or the risk of death. It was tested on data from 2023 that the model had never seen during training. It's like taking an exam on a textbook that was written after the class ended.

Currently, access to the data is suspended, so the exact performance numbers are unknown. But the approach itself, how to build such a model without compromising patient privacy, is described in detail. It's like the blueprint for a bridge that hasn't been built yet, but you can already learn from it.