gem-hunter

Jun 28th, 2021: The repo updates to MIMIC-IV V4.0. As there is a major update of MIMIC-IV from version 0.4 to 1.0, I modify some codes to accomodate these changes.

Jun 2nd, 2021: MIMIC-IV-ED is released. Shout out to the authors. If you don't need the ED data, just skip the steps numbered with "ED". If you have already installed MIMIC-IV using this repo, you only need to run 1.ED, 4.ED, 5.ED.

In light of there is no offcial guide of deploying MIMIC-IV to Postgres server, this repo will help you navigate this process. We made necessary modification based on the offcial guide of MIMIC-III to adapt to MIMIC-IV.

Complete the PhysioNet Credentialing and download MIMIC-IV data. Decompress all .csv.gz file using "gzip -d *.gz". You will have such directory structure:

BASE_DIR  
│
└───core
│   │   admissions.csv
│   │   patients.csv
│   │   transfers.csv
└───hosp
│   │   d_hcpcs.csv
│   │   ...
└───icu
│   │   d_items.csv
│   │   ...

Complete the PhysioNet Credentialing and download [MIMIC-IV-ED data][MIMIC-IV_ED]. Decompress all .csv.gz file using "gzip -d *.gz". Move those decompressed file to the same base directory. You will have such directory structure:

BASE_DIR  
│
└───core
│   │   admissions.csv
│   │   patients.csv
│   │   transfers.csv
└───hosp
│   │   d_hcpcs.csv
│   │   ...
└───icu
│   │   d_items.csv
│   │   ...
└───ed
│   │   edstays.csv
│   │   ...

Please refer here for postgresql download and installation

In my case, I create a user named mimicuser, a new database named mimic4 containing a new schema named mimiciv holding all data by:

createuser -P -s -e -d mimicuser             #creat user
psql -U mimicuser                            #launch psql
CREATE DATABASE mimic4 OWNER mimicuser;      #create new database
\c mimic4;                                   #enter the new database
CREATE SCHEMA mimiciv;                       #create new schema
\q                                           #quit psql

psql 'dbname=mimic4 user=mimicuser options=--search_path=mimiciv' -f create_tables.sql

psql 'dbname=mimic4 user=mimicuser options=--search_path=mimiciv' -f create_ed_tables.sql

Change /YOUR/BASE/DIR/ to your BASE_DIR in the line 3,11,27 and run:

psql 'dbname=mimic4 user=mimicuser options=--search_path=mimiciv' -f load_data.sql

Change /YOUR/BASE/DIR/ to your BASE_DIR in the line 3 and run:

psql 'dbname=mimic4 user=mimicuser options=--search_path=mimiciv' -f load_eddata.sql

• The data description and real data are inconsistent in data format, varchar length, etc. This repo is implemented based on the real data. Some comments are left in create_tables.sql to mark down the difference.
• We follow the installation of MIMIC-III that creating 10 subtables for labevents.csv and chartevents.csv, respectively. Chartevents table is divided by itemid to 10-fold, while labevents table is divided by labevents_id.
• The size of MIMIC-IV after de-compression may exceed 100GB. Therefore, loading data is time-consuming. You may expect several hours to run through the process.
• This repo is based on the latest data release before Jun 28, 2021. Should there be any data update, incompatibiltiy may observed.
• Leave an issue, if you encounter any error.