Datasets and access¶
YAIB works on preprocessed cohorts in a common parquet format. To run it on the full ICU datasets, you need three steps:
- Get access to the datasets (this page).
- Generate the cohorts with YAIB-cohorts.
- Run YAIB on the generated cohorts.
To try YAIB first, use the demo cohorts in the repository; see the Quickstart.
Supported datasets¶
| Dataset | MIMIC-III / IV | eICU-CRD | HiRID | AUMCdb |
|---|---|---|---|---|
| Admissions | 40k / 73k | 200k | 33k | 23k |
| Version | v1.4 / v2.2 | v2.0 | v1.1.1 | v1.0.2 |
| Frequency (time series) | 1 hour | 5 minutes | 2 / 5 minutes | up to 1 minute |
| Originally published | 2015 / 2020 | 2017 | 2020 | 2019 |
| Origin | USA | USA | Switzerland | Netherlands |
Getting access¶
All four datasets require a credentialing procedure.
- MIMIC-III, MIMIC-IV, eICU-CRD and HiRID are available on PhysioNet. The eICU getting-started guide describes the process.
- AUMCdb has its own access procedure.
We are not involved in these access procedures and cannot answer requests for data access.
Other datasets¶
Any dataset can be used once it is in YAIB's format; see Adding datasets. Since the datasets were collected independently, they do not share a data structure or identifiers. YAIB-cohorts uses the ricu R package to harmonize them.