Introduction and instructions of MIMIC-IV, Intensive care medicine database

Tech 2023-07-07 17:19:29 8 Source: Network

The MIMIC database is a powerful tool that can provide clinical data for clinical researchers. The database was jointly established by the Computational Physiology Laboratory of Massachusetts Institute of Technology, Beth Israel Deaconess Medical Center of Harvard Medical School and Philips Medical Company in 2003 with the support of the National Institutes of Health

The MIMIC database is a powerful tool that can provide clinical data for clinical researchers. The database was jointly established by the Computational Physiology Laboratory of Massachusetts Institute of Technology, Beth Israel Deaconess Medical Center of Harvard Medical School and Philips Medical Company in 2003 with the support of the National Institutes of Health.

Researchers can screen clinical information of interested Patients based on certain inclusion criteria, and use this information to conduct subsequent data analysis and write articles. Data collection and analysis can serve as an important basis for publishing SCI papers. In addition, MIMIC is a public database, where all Patient information is desensitized and does not require clinical ethical review for publication.

The following types of diseases are involved:

The MIMIC database has now produced three versions: MIMIC II, III, and IV. The MIMIC database contains data from all BIDMC internal and external ICU Patients. To protect Patient privacy, the data team has de labeled Patient information and made it available for free to researchers worldwide.

MIMIC IIIThe database collected data from 53423 adult Patients admitted to the ICU by BIDMC from June 2001 to October 2012, and 7870 critically ill newborns admitted from 2001 to 2008.

MIMIC IVMIMIC III 20082019BIDMC1945

MIMIC IV

The first category is clinical data extracted from EHR, including Demographics, disease diagnosis, laboratory testing, drug treatment, vital signs, etc.
The second type is waveform data, vital signs, fluid management, and event recording collected by ICU bedside monitoring equipment, mainly from the IMDSoftMetaVision system.
The third type is death follow-up data, which is important for studying the prognosis of Patients by obtaining the date of death outside the hospital through the social Insurance system as a component of the MIMIC database.

Patient Admission information is recorded on a per Admission basis, with each record having a separate HADM_ ID, host_ Expire_ Flag only checked whether there was an in hospital death during the hospitalization, and some Patients who died in the hospital did not have a death time, which may be due to database issues

Name	Datatype	Description
Subject_ ID	INTEGER	Patient ID
Hadm_ ID	INTEGER	InPatient ID
Admittime	TIMESTAMP (0)	Admission time
Dischtime	TIMESTAMP (0)	Discharge time
Deathtime	TIMESTAMP (0)	Time of death
Admission_ Type	VARCHAR (40)	Divided into 9 types based on urgency of Admission
Admission_ Location	VARCHAR (60)	Admission source
Discharge_ Location	VARCHAR (60)	Discharge destination
Insurance	VARCHAR (255)	insurance
Language	VARCHAR (10)	language
Marital_ Status	VARCHAR (80)	marriage
Ethnicity	VARCHAR (80	race
Edregtime	TIMESTAMP (0)	Time of emergency treatment
Edouttime	TIMESTAMP (0)	Emergency response time
Hospital_ Expire_ Flag	SMALLINT	Did you die during hospitalization? 1 is 0 is not

Patient definition:

Subject_ ID Subject_ ID
Hadm_ ID Hadm_ ID
Transfer_ ID Patients have a unique transfer for each ward change_ ID
Stay_ If the ID is transferred within the same type of ward, a transfer will be updated_ ID, but there will be the same stay_ ID, for example, if one ward in the ICU is transferred to another ward, then stay_ ID remains unchanged, transfer_ ID update. The allocation of all IDs is random and independent of time sequence.

Date and time

The suffix 'date' has a minimum resolution of days; Fields with the suffix 'timed' have a minimum resolution of minutes.

Chart time and storage time: They are the recording time and storage time of the measurement. Usually named d after charttime_ The beginning is the encoding table. D_ The ICD starts with the ICD encoding table. The table at the end of the ICD is a record table that uses ICD encoding.

Tables

Divided into six modules, Core, Hosp, ICU, ED, CXR, Note

Core

The Core module contains Patient tracking data. This describes population statistics, Admission information, and inPatient ward Transfers

Admission Hadm_ ID

PatientPatient information

TransfersWard transfer information

Hosp

The information covered by the Hosp module includes laboratory measurements, microbiology, drug management, and paid diagnosis, among others

D_ Icd_ Diagnostics contains diagnostic codes from the 9th and 10th editions of the International Classification of Diseases (ICD). Icd_ The code must be read as a string, as starting with 0 is meaningful. If it is read as a numerical value, there will be duplicates, such as 01622 and 1622 representing different diseases.

D_ Icd_ Diagnoses table

Icd_ Code must be read as a string, as starting with 0 is meaningful, and reading as a numerical value will result in duplication

Diagnoses_ Diagnostic information table for ICD patients. Seq_ Num represents the priority of diagnosis, and the higher the sequence number, the more important it is, while lower priority sorting is sometimes not accurate

DRGCodes is also a diagnostic related table, using DiagnosisRelatedGroup (DRG) code encoding to correspond to the main diagnoses in the diagnosis table.

D_ Icd_ Procedures
Procedures_ Surgical information of ICD patients during hospitalization, including information during ICU.

D_ Laboratory Items Laboratory Inspection Definition Table

Laboratory events Record of patient laboratory examinations

Prescriptions prescription data, drug data

Pharmacy Pharmacy Data Sheet, in Pharmacy_ The id field is associated with the prescriptions table, which is equivalent to supplementing the prescription information in the prescriptions table. In addition, there are forms of Microbiological culture, charges, orders, etc

ICU

Contains information collected from clinical information systems used within the ICU. The recorded data includes intravenous administration, ventilator settings, and other chart items. Each patient may have multiple ICUs per admission, and each ICU corresponds to a stay_ ID

D_ Items contain the codes of all items that occur within the ICU, and are associated with other tables within the ICU module through the itemid field.

ICUstays' check-in time information in the ICU

Chartevents Chartevents contains all available chart data for a patient, with some lab data being duplicated from the labevents table.

ED

Subject_ IDHadm_ IDHadm_ IDed

The diagnostic table provides a diagnostic list for patients. Determine the diagnosis after discharge from the emergency department.
The main tracking table for visits to the emergency department of edstaystable. It provides the time for patients to enter and leave the emergency department
When medrecontable enters the emergency room, staff will ask the patient what medication they are currently taking. This process is called drug coordination, and the medical checklist stores the survey results of nursing staff
PyXistable provides information on preparing drugs through the Pyxis system.
Triagetable contains vital sign information of the patient during their first triage in the emergency room
The routine vital signs of patients admitted to the vital sign table emergency room take 1-4 hours. These vital signs are stored in the vital sign table
Vitalsign_ Patients admitted to the hl7table emergency department can be monitored through telemetry technology. Every minute of vital signs are transmitted to the central server of the hospital, and these vital signs are recorded here.

CXR

X-ray chest X-ray file, source data in dicom format, but also provided for download in jpg format. Including chest X-ray and imaging reports. It is worth noting that there are cases where patients with imaging do not have hospitalization records.

Cxr record list image list
Cxr-study-list impact report list
Mimic cxr-2.0.0 chexpert labeled 14 labels using CheXpertlabel (non manual labeling based on imaging reports, collaboration between Stanford and MIT), as shown in the table below
Mimic-cxr-2.0.0-split provides reference for partitioning training, validation, and testing sets

Note

NOTPUBLICLYAVAILABLE: All text reports, including discharge, ultrasound, electrocardiogram, imaging, etc

Official introduction can be found on the MIMIC official website: https://mimic.mit.edu/
The download of the MIMIC-IV database can be found on the MIMIC-IV release page: https://physionet.org/content/mimiciv/1.0/
Download data in dicom format for X-ray chest X-ray, see MMIC-CXRDatabase: https://physionet.org/content/mimic-cxr/2.0.0/
X-ray chest X-ray JPG format data can be found in MIMIC-CXR-JPG: https://physionet.org/content/mimic-cxr-jpg/2.0.0/
Emergency data can be found in MIMIC-IV-ED: https://physionet.org/content/mimic-iv-ed/1.0/

Reference

MIMIC-IV, afreelyaccessiblelectronichealthrecorddataset: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9810617/
Python series - MIMIC-IVDemo descriptive statistics: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9810617/

Clinical database mining (MIMIC, NHANES), medical statistics, medical data analysis

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:win-e@163.com)

Mobile advertising space rental

Tag: Introduction and instructions of MIMIC-IV Intensive care medicine database