Sleep does more than help you feel rested. It plays a key role in brain health, heart health, metabolism, and long-term disease risk. Doctors have known for years that poor sleep is linked to chronic illnesses, but predicting exactly which diseases someone might develop has been difficult.
A new study published in Nature Medicine suggests that just one night of detailed information about your sleep could help predict your risk for many serious health conditions years in advance. This research combines sleep science with artificial intelligence in a way that could change how doctors think about prevention.
Why Sleep and Chronic Disease Are So Closely Connected
Sleep affects nearly every system in the body. During sleep, the brain clears waste, the heart and blood vessels recover, hormones reset, and breathing patterns stabilize. When sleep is disrupted, these processes are affected.
Previous research has linked poor sleep to:
- Heart disease and stroke
- Dementia and other brain disorders
- Diabetes and kidney disease
- Mental health conditions
Many of these problems begin developing long before symptoms appear. This makes sleep an especially powerful early signal of health changes happening under the surface.
A New Way to Read the “Language” of Sleep
The researchers created a new computer model called SleepFM, designed to analyze full overnight sleep studies known as polysomnography (PSG). These studies record brain waves, heart rhythm, breathing, oxygen levels, and muscle activity during sleep.
Instead of focusing on one sleep problem—like sleep apnea—the model looks at all signals together, learning complex patterns that humans might miss.
Key features of the approach:
- It analyzes multiple sleep signals at once
- It does not rely on manual scoring alone
- It can handle different sleep study setups from different hospitals
The goal was to see whether sleep data could predict future disease risk, not just diagnose sleep disorders.
How the Study Was Done
The study analyzed an exceptionally large dataset:
- Over 585,000 hours of sleep recordings
- More than 65,000 participants
- Data from multiple sleep clinics and long-running health studies
The researchers trained the model using sleep data and then linked it with electronic health records. They looked at whether diseases appeared after the sleep study, making sure predictions were truly forward-looking.
The model was tested across:
- Heart disease
- Stroke
- Dementia
- Kidney disease
- Diabetes complications
- Cancer
- Mental and neurological conditions
Importantly, the model was also tested on a completely separate dataset it had never seen before, to make sure the results were reliable.
What the Results Revealed
The findings were striking. From a single night of sleep, the model could accurately predict the future risk of 130 different diseases.
Some of the strongest predictions included:
- All-cause mortality
- Dementia
- Heart failure
- Heart attack
- Stroke
- Chronic kidney disease
The model performed especially well for brain and heart conditions. For example, it showed high accuracy in predicting dementia years before diagnosis. This supports the idea that subtle changes in sleep may appear long before memory problems or other symptoms.
The model also performed better than predictions based only on age, sex, or body weight. This suggests that sleep holds unique health information not captured by routine checkups.
Why This Discovery Stands Out
This study is exciting because it shows sleep could become a non-invasive window into future health. Instead of waiting for symptoms, doctors may one day use sleep data to identify people at higher risk and intervene earlier.
Potential future benefits include:
- Earlier detection of chronic diseases
- Better targeting of prevention strategies
- Reduced need for invasive testing
- More personalized healthcare
The researchers emphasize that this tool is not ready for routine clinical use yet. The model needs further testing, especially in more diverse populations. Still, the results show what may be possible as sleep tracking and AI technology continue to improve.
How This Affects Everyday Health
For now, this research reinforces an important message: sleep is not optional for long-term health. Poor sleep may be an early warning sign, even when you feel otherwise healthy.
In the future, advanced sleep analysis could help doctors:
- Identify disease risk earlier
- Monitor health changes over time
- Personalize prevention plans
This study highlights sleep as more than rest—it may be one of the most powerful predictors of future health we have.
Source:
Thapa R, Kjaer MR, He B, et al. A multimodal sleep foundation model for disease prediction. Nature Medicine. 2025. doi:10.1038/s41591-025-04133-4. Accessed by: https://pubmed.ncbi.nlm.nih.gov/41495409/
