Here we report the publication of the MIMIC-III database, an update to the MIMIC II database (Data Citation 1). MIMIC-III integrates deified and comprehensive clinical data from patients admitted to Beth Israel Deaconess Medical Center in Boston, Massachusetts, and makes it available to researchers internationally under a data use agreement (Figure 1). The open nature of the data allows for the reproduction and improvement of clinical trials in a way that would not otherwise be possible. This video examines the details of the MIMIC-III data model, the clinical data model we use during this course and the specialization. Mimic is synonymous with Medical Information Marked for Intensive Care and is part of a larger data project called PhysioNet, a large collection of open source physiological and clinical data submitted by many institutions. Mimic, developed by MIT`s LABORATORY of Computer Physiology, is a dataset derived from patients admitted to intensive care at Beth Israel Deaconness Medical Center in Massachusetts, Boston. The complete data set is fairly broad and consists of more than 53,000 hospitalizations over an 11-year period. The dataset includes patients who have been admitted to medical, surgical, pediatric and neonatal intensive care units in Beth Israel. The full data set is available free of charge, but requires specific certifications and licensing agreements to access it.
The MIMIC team has released a group of 100 patients who do not have these additional requirements. Here`s the data set we use for this class. This graph shows the extent of data sources and types of data collected by the mimic team, including structured data elements such as patient demographics, diagnoses, procedures and medications, as well as unstructured clinical indications and physiological signals. The data set is processed to remove all patient identifiers. Dates are replaced by fictitious future data and data are broken down into the MIMIC-III data structure. You can see here the extensive coverage of the patient population in the complete MIMIC-III dataset. Later, we will assess the extent to which the smallest dataset of 100 patients reflects the characteristics of the patient population in the complete MIMIC-III dataset. Details of each MIMIC table are available on the mimicry site, mimic.physionetnet.org. This URL is also available in the courses. In the left navigation bar on the imitider`s homepage, select the MIMIC themed tables. In this sub-menu, all tables available in MIMIC-III are listed in alphabetical order.
The DATEN MIMIC-III model contains 26 tables. If you select a table, a page with general information about the contents of the table, z.B the data source in the table, a brief description of the data, the number of lines in the full set and the link between the table and other tables are displayed.