Our decision to focus on hemodynamically unstable patients means that, although the MIMIC Database should be representative of the full range of pathophysiologies that result in sudden blood pressure changes, it does not represent the entire ICU population. As automated decision support systems for the ICU develop and mature, there will be a need for recordings that represent other significant groups of patients whose problems also require rapid assessment and appropriate intervention.
The very long time series now available for study in the MIMIC Database will be of particular interest in investigations of heart rate, blood pressure, and respiratory dynamics and their interactions. We also anticipate that the clinical information accompanying each record will make the MIMIC Database a valuable source of well-characterized case studies for medical education, particularly in the intensive care setting.
Figure 3: Mean ABP and diastolic PAP changes are usually correlated in record 216.
Data collection began in late 1994 and is now nearly complete. As of August 1996, we had recorded roughly 250 patient-days. We will finish the selection of the recordings to be included in the MIMIC Database by the end of 1996, and the database will subsequently be made available to other researchers.
Figure 4: In MIMIC Database record 415, there is an anticorrelation between mean ABP and diastolic PAP.
Development of the MIMIC Database has been made possible with the support of the Hewlett Packard Foundation and of Nihon Kohden America, Inc. Technical support from Hewlett Packard was essential to the project, and was freely and generously given to us; we especially wish to thank Dick Myreck and John Ames of HP, who arranged for the long-term loan of a bedside monitor for bench testing. At Boston's Beth Israel Hospital, we received the cooperation and assistance of many individuals to whom we are deeply grateful. We wish especially to thank the patients who participated in the project and their families; the staffs of the Medical, Surgical, and Cardiac Intensive Care Units; the Biomedical Engineering Research and Arrhythmia Laboratories; and the hospital's electrical safety engineers. We also thank our students, Bryant Lin, Adam Hoytya, and Ruilin Zhao, all of whom contributed to the success of this project. Finally, GBM wishes to thank Edna Moody of the BIH Center for Clinical Computing for her inestimable help in obtaining data from the hospital's on-line medical information systems, and for everything else.