The natural history of disease is how a pathological process progresses in a person over time, without treatment. This lecture links that classic concept to the work of Shmatko et al. (Nature, 2025).
In short: from previous diagnoses, treatments and lifestyle factors, it is possible to generate a statistical prognosis of the timing and risk of the next major health event.
Delphi-2M
The paper presents Delphi-2M, an open-source artificial intelligence model that estimates a person's risk of over 1,000 diseases, years in advance. Rather than analysing language, it treats diagnoses and medical events as items in a sequence, using the same architecture as language models.
The lecture reviews how this approach works and what it may mean for preventive medicine. It is a statistical tool: it does not replace assessment by a professional.