- We are using the eICU database v2.0. This database requires credential access. Sign up here
- Instructions to get eICU postgres locally
- Official paper explaining dataset
- Database schema
- Table information
- Features for mortality prediction and length of stay prediction are replicated according to this work
- Different from the authors, we are operating on eICU database v2.0
- Run
extract-features.ipynbwith eICU credentials to downloadrawdata.csvfile - Summary of features and tables where the feature was extracted from are as follows:
| Feature | Table location | Data type |
|---|---|---|
| Heart rate | apacheapsvar | Numerical |
| Mean arterial pressure | formula | Numerical |
| Diastolic blood pressure | vitalperiodic | Numerical |
| Systolic blood pressure | vitalperiodic | Numerical |
| O2 | lab | Numerical |
| Respiratory rate | apacheapsvar | Numerical |
| Temperature | apacheapsvar | Numerical |
| Glucose | apacheapsvar | Numerical |
| FiO2 | apacheapsvar | Numerical |
| pH | apacheapsvar | Numerical |
| Height | patient | Numerical |
| Weight | patient | Numerical |
| Age | patient | Numerical |
| Admission diagnosis | patient | Categorical |
| Ethnicity | patient | Categorical |
| Gender | patient | Categorical |
| Glasgow Coma Score Total | formula | Numerical |
| Glasgow Coma Score Eyes | apacheapsvar | Numerical |
| Glasgow Coma Score Motor | apacheapsvar | Numerical |
| Glasgow Coma Score Verbal | apacheapsvar | Numerical |
mean arterial pressure = SBP + 2(DBP) /3
GCS total = GCS sum of eyes, motor, verbal