shorthand: eICU-CRD, eICU

Pollard, T., Johnson, A., Raffa, J., Celi, L. A., Badawi, O., & Mark, R. (2019). eICU Collaborative Research Database (version 2.0). PhysioNet. https://doi.org/10.13026/C2WM1R.

Original publication: https://www.nature.com/articles/sdata2018178 The eICU Collaborative Research Database, a freely available multi-center database for critical care research. Pollard TJ, Johnson AEW, Raffa JD, Celi LA, Mark RG and Badawi O. Scientific Data (2018). DOI: http://dx.doi.org/10.1038/sdata.2018.178.

https://www.google.com/maps/embed?pb=!1m18!1m12!1m3!1d62926689.41650223!2d-161.4961388481041!3d36.100716450581736!2m3!1f0!2f0!3f0!3m2!1i1024!2i768!4f13.1!3m3!1m2!1s0x54eab584e432360b%3A0x1c3bb99243deb742!2sUSA!5e1!3m2!1sde!2sde!4v1736764536627!5m2!1sde!2sde

NOTES

Determining the first ICU stay

https://github.com/CUB-CORR/reprodICU/blob/main/helpers/A_extract/A_extract_eicu.py#L234

Whilst eICU includes information about sequential ICU stays during the same hospital stay (patient → unitVisitNumber), there is “no systematic method for chronologically ordering patientHealthSystemStayID for the same patient within the same year”.

To still calculate a global sequential stay number to ensure a common, fixed procedure for selecting probable first ICU stays, reprodICU orients itself along established procedures.

Specifically, ICU stays are ordered the following way:

  1. ICU mortality must be "increasing" (i.e. alive [= false / 0] first)
  2. Admission year must be increasing
  3. Admission age must be increasing
  4. within the same hospital stay
    1. unitVisitNumber must be increasing

MAPPINGS

there are no pre-existing mappings of eICU values to established ontologies.

custom mappings