Changelog:
Major changes since version 2.4 :
- Date of death after hospital discharge
Patients have been matched against the Social Security death tapes, and the date of death has been imported into MIMIC II. For patients who died in the hospital, their respective dates of discharge were used as their dates of death. Two additional columns have been added to the D_PATIENTS table:
DOD = Date of Death
HOSPITAL_EXPIRE_FLG = Whether or not the patient deid in the hospital
The column EXPIRE_FLG in the ADMISSION table has been removed.
- New Waveform case ID matches
A first round of waveforms with up to 8 simultaneous channels have been matched to SUBJECT_IDs. The matches can be found in the PARAMETER_MAPPING table where CATEGORY is "SUBJECT_ID_TO_CASE_ID", the columns PARAM1_NUM (for the SUBJECT_ID) and PARAM2_STR (for the waveform case ID).
- Merge NOTEEVENTS into REPORTEVENTS
REPORTEVENTS has been merged into NOTEEVENTS. Categories have been extracted into D_REPORTITEMS, with the ITEMID column as the foreign key, but not completely; the D_* tables will be combined into a complete dictionary in a later release.
- New ICUSTAY_DAYS materialized view
The ICUSTAY_DAYS materialized view was added to divide each ICU stays into daily periods. This greatly helps computation of cumulative queries.
- Daily SOFA scores
New entries in D_CHARTITEMS were added to represent SOFA scores (itemid = 20002), and corresponding scores were added in CHARTEVENTS.
- Daily SAPS I calculations
For each ICU stay, a daily SAPS I was added into the CHARTEVENTS table, with ITEMID 20001.
- Add a view for summarizing the D_CHARTITEMS table
It was somewhat troublesome for users of the database to collect statistics for the parameters (ITEMIDs) in D_CHARTITEMS. Therefore, a summarizing materialized view, D_CHARTITEMS_DETAIL, was added as a first introduction of all charted parameters; it describes how many patients each parameter was charted from, and various statistics such as min., max., and average.
- Clean "value1" and "value2" columns in CHARTEVENTS for numeric itemids
Some ITEMIDs in the CHARTEVENTS table that are supposed to be only numeric (such as ABP, HR) had repeated values in columns VALUE1/VALUE2 (which are VARCHAR columns) and VALUE1NUM/VALUE2NUM (NUMBER columns); VALUE1 and VALUE2 were cleaned accordingly.
- Reorganized ICUSTAY_DETAIL materialized view
The ICUSTAY_DETAIL view has been used as a key component in many of computation for several researches. We added/changed some columns to reflect:
EXPIRE_FLG = Flag to indicate if the paitent is dead, either in the hospital or after discharge
HOSPITAL_EXPIRE_FLG = If the patient deid in the hospital
ICUSTAY_EXPIRE_FLG = If the patient deid in the ICU
HEIGHT = Patient height
WEIGHT_FIRST = The first measured weight for the patient in the ICU stay
WEIGHT_MIN = The minimum weight for the ICU stay
WEIGHT_MAX = The maximum weight for the ICU stay
SAPSI_FIRST = The first measured SAPS-I score for the patient in the ICU stay
SAPSI_MIN = The minimum SAPS-I score for the ICU stay
SAPSI_MAX = The maximum SAPS-I score for the ICU stay
SOFA_FIRST = The first measured SOFA score for the patient in the ICU stay
SOFA_MIN = The minimum SOFA score for the ICU stay
SOFA_MAX = The maximum SOFA score for the ICU stay
- Add an ICUSTAY_ID column to all event tables
To facilitate queries and other computations, the ICUSTAY_ID column has been added to event tables: CHARTEVENTS, IOEVENTS, MEDEVENTS, LABEVENTS and TOTALBALEVENTS.
- Clean duplicates in D_CAREUNITS
There were duplicate names in the table D_CAREUNITS that needed to be merged, and their foreign keys updated.
- Remove entries from item tables which contain no records in event tables
292 unused rows have been removed from D_MEDITEMS; 3544 from D_CHARTITEMS.
- Fix problems with duplicate hospital admissions
Six patients had hospital admissions with overlapping admission-to-discharge date intervals. These were fixed by examining the discharge summaries.
- Add missing foreign key for POE_ORDER
For POE_ORDER, columns SUBJECT_ID and HADM_ID were properly made to be foreign keys for the tables D_PATIENTS and ADMISSIONS, respectively.
Parameter listing:
For a list of the parameters in MIMIC , please download the following file ( mimic2v25-parameter-list.pdf [PDF])
Database statistics:
MIMIC II Adult records summary ( Download the full report [PDF] and the SQL source code ) |
|||||
| N* | #cases** (%unit) | mean ±std | median | IQR (Q1,Q3) | |
| Hospital admissions | 22,870 | ||||
| ICU Stays | 25,328 | ||||
| Age (years) | 25,328 | 63.3 ±17.7 | 65.20 | 26.1 (51.3,77.4) | |
| Gender (male) | 25,328 | 14,259 (56.3%) | |||
| ICU Los (days) | 25,328 | 4.7 ±11 | 2.20 | 3.3 (1.1,4.4) | |
| Hospital Los (days) | 22,870 | 10.7 ±13 | 7.00 | 9 (4,13) | |
| SAPS I - day 1 *** | 20,142 | 13.4 ±5.3 | 13.00 | 7 (10,17) | |
| Mechanical ventilation | 25,328 | 11,969 (47.3%) | |||
| Invasive Swan-Ganz hemodynamic monitoring | 25,328 | 5,637 (22.3%) | |||
| Invasive arterial blood pressure monitoring | 25,328 | 14,054 (55.5%) | |||
| Use of vaso-active medications | 25,328 | 8,693 (34.3%) | |||
| Hospital mortality | 22,870 | 2,666 (11.7%) | |||
* N = The number of ICUStays or Hospitalizations. that were included in the calculation for the given ICU Unit. ** #cases = The number of ICUStays orHospitalizations that match the category criteria. For example: In Gender, #cases will be the number of male ICUStays in a unit. *** Only ICUStays where they have all 14 SAPS parameters were included |
|||||