Sleep Health Data Commons
Building toward millions of nights of sleep, ready to learn from.
A national commons of de-identified clinical sleep data — polysomnography, EEG, home sleep tests, brain imaging and longitudinal health records — harmonized to common standards and open to credentialed researchers.
Two women, both 36. From a single night of sleep EEG, one brain shows the strong slow waves and spindles of a younger brain while the other shows the weaker, sparser patterns more common decades later — a difference researchers can read as a brain-health score. Measuring it across millions of nights is what the Commons is for. A research measure, not a diagnosis (Sun et al., 2019). De-identified data.
The Commons, as it grows
These are program goals. Confirmed figures will be published here before launch.
What’s inside
Clinical data as collected, then harmonized: signals and imaging in BIDS, health records in OMOP CDM v5.4.
Polysomnography
Attended overnight sleep studies: EEG, EOG, EMG, airflow, effort, oximetry, with staging.
EEG
Routine, ICU continuous, and epilepsy-monitoring EEG — a spectrum of acuity in one place.
Home sleep tests
Portable airflow, effort and oximetry recordings from the home setting, at population scale.
Brain imaging
Structural and other MRI linked to sleep and EEG for the same patients where available.
Longitudinal health records
Diagnoses, medications, labs and encounters over time, so sleep findings connect to outcomes.
Linked across modalities
Where a patient has studies in more than one modality, the records are linked, so a night of sleep can be studied alongside imaging and clinical course.
The code, not just the data
Explore the algorithms behind the Commons — 85 open-source tools for analyzing sleep and EEG data, from automated staging and spike detection to brain-age and outcome models. Each links to its code, and many to the dataset it was built on.
Browse the analysis code →How it works
Health system contributes
A site sends its archived sleep studies, EEG, imaging and linked health records.
Stanford de-identifies centrally
Records are de-identified in one secure place, with per-patient date shifting.
Harmonized to standards
Everything is converted to BIDS and OMOP CDM v5.4 so sites can be analyzed together.
Researchers analyze
Credentialed researchers query and analyze the Commons through bdsp.io.
Study sleep at national scale
- Pooled clinical sleep data across many health systems and levels of acuity.
- Signals and imaging in BIDS, health records in OMOP CDM v5.4 — analysis-ready.
- Access by credentialing and a data use agreement, through bdsp.io.
Contribute, and get your data back harmonized
- We handle de-identification and conversion centrally; your effort is kept light.
- You receive your own data harmonized to BIDS and OMOP, plus the conversion code.
- Your patients’ privacy is protected under a single IRB and a data use agreement.
Privacy and trust
Privacy is the project’s promise. It is built into how the Commons is run, not added on.
De-identification
Direct identifiers are removed, and each patient’s dates are shifted by a consistent per-patient offset, so timelines stay internally accurate without revealing real dates.
Single IRB review
The program operates under a single IRB of record (BRANY), so oversight is consistent across every contributing site.
Stanford’s high-risk data environment
Data are handled in Stanford’s secured environment for high-risk data, with controlled, auditable access.
Credentialed access, no re-identification
Researchers are credentialed and agree to a data use agreement that forbids any attempt to re-identify individuals.
A growing national network
Each point is a contributing health system; points fill as sites join. Institutions are named only with their permission.
Joined Goal
Placeholder network graphic (concept B). A geographic U.S. map will replace this once sites are confirmed.