Stanford’s SleepFM AI model predicted 130 medical conditions from a single night of sleep-clinic recording, each with a C-index of at least 0.75. Trained on 585,000 hours of recordings, it suggests that sleep studies carry information about many diseases beyond sleep disorders.
What SleepFM Actually Found
Stanford’s model estimates future disease risk from one night of sleep recording. Mortality prediction reached a C-index of 0.84, dementia prediction 0.85, and myocardial infarction 0.81. The C-index measures how well the model ranks people’s risk across a group, not the chance that one person develops the disease. Atrial fibrillation prediction reached an AUROC of 0.81, close to the 0.82 an earlier study reported.
What sets SleepFM apart is its breadth. Earlier approaches typically focused on individual outcomes, while this model produced predictions for 130 conditions from the same recordings.
Limits and What Comes Next
SleepFM is a proof of concept, not a finished clinical product. It was built and tested on lab-grade sleep recordings, not consumer wearables. Its training data came mostly from people referred for sleep studies, so the authors note it does not represent the general population.
The authors write that as wearable sleep technology advances, models like SleepFM may offer noninvasive monitoring. This study did not test consumer devices.
The direction is worth watching, but for now it is a research result.
