Details of the issues surrounding data collection and annotation can be found in (5) (11) and (12), and details related to de-identification can be found in (6). Although the database described in this document is large and detailed, caution should be taken when analyzing the data, since the clinical systems for monitoring and archiving the data are not perfect and some data may be incorrect.
Moreover, we were unable to capture all of the data associated with a patient's stay, since we did not have data collection facilities in the OR, ER, or step-down wards. Neither were we able to obtain data from other hospitals for any given individual, so visits to outpatient care, or to other hospitals will not be reflected in a patient's stay. Finally, no patient data before the start of the project is available, other than the pre-existing history recorded at a particular stay. Despite these limitations, there is a wealth of data to be found in MIMIC II.
Note that this document does not provide detailed information on how to extract data from the WFDB flat-file versions of the data found in the MIMIC II database. More information on this can be found here: