Chengda Zhang, MD

The Analysis of eICU Collaborative Research Database (eICU-CRD)

1. Setting up local database

   8/9/2020-8/10/2020

   The instruction for installing local eICU database is posted in the eICU-CRD website. However while doing so, I encountered large work volume in setting up local tables with appropreate variables and the mission became impossible when the size of a single csv file exceeded 1Gb. Therefore, I have created and shared a way to set up local postgreSQL database with python. Using this method, I have successfully inserted all my csv into database, including the vital sign table which had over 140 million rows. My source code is available on Github.

2. Side notes of eICU database

   While working on my eICU database, I realized some features of this database:

   • Definitely take a look at apacheApsVar, apachePatientResult and apachePredVar before you start digging into the specific tables. Somehow the creators of eICU database did some secondary statistics based on available data, and saved them into these 3 tables. Sometimes it would save you a lot of time from making your own nomenclature.
   • There are multiple places where you can get patient's O2 requirement. apacheAps table has a variable called FiO2, which is the FiO2 at presentation. respiratoryCharting will give the most detailed view.