MIMIC-II v2.6

Changelog

• Merge the last year of the ISM database

The last year of ISM, through September 2008, has been added, including indications of matches to waveforms.

• New dateshifts

We have given all subjects a new, random dateshift.

• Add overall SOFA score as total of each organ system’s score

The individual-system SOFA scores are in CHARTEVENTS, with itemids 20002 through 20007; a new itemid, 20009, has been added, and it s for the total score. The relevant columns in ICUSTAY_DETAIL (sofa_first, sofa_min, sofa_max) indicate this total score.

• Add a column to ICUSTAY_DETAIL indicating number of matched waveforms

A new column, matched_waveforms_num, has been added to ICUSTAY_DETAIL; it gives the number of waveforms matched to the patient.

• Add waveform statistics

Basic waveform statistics and metadata have been added to the database; these are essentially the organized output of what one would see by running the WFDB utility wfdbdesc (over all of the waveforms).

• Add LOINC codes for hospital labs

D_LABITEMS contains two new columns, loinc_code and loinc_description, giving the LOINC codes and descriptions respectively.

• Add diagnosis-related groups (DRGs)

A new table, DRGEVENTS, maps subject IDs and hospital-admission IDs to DRG itemids, which are defined in D_CODEDITEMS.

• Add procedures

PROCEDUREEVENTS contains indications of procedures performed on patients, for given hospital-admission IDs. The itemids for procedures are defined in D_CODEDITEMS. Exact times for procedures are not given; however, their dates are given in the proc_dt column, and the order in which a patient’s procedures are performed, for a given day, are indicated by consecutive numbers (integers) in the sequence_num column.

• Add demographic information

DEMOGRAPHICEVENTS contains indications of marital status, type and source of admission, ethnicity, overall payor group, and religion as recorded for each patient upon admission to the hospital. The definitions of the itemids are given in D_CODEDITEMS.

• Add microbiology information

MICROBIOLOGYEVENTS contains information on microbiology tests performed on patients: specimen, organism, and antibacterium are indicate in the spec_itemid, org_itemid, and ab_itemid columns respectively; these itemids are defined in D_CODEDITEMS.

• Remove duplicate mappings of waveform case IDs to subject IDs in PARAMETER_MAPPING

• Fix the icustay_expire_flg in ICUSTAY_DETAIL

icustay_expire_flg indicates whether the given patient died during the given ICU stay. For some patients, this column was ‘Y’ for all of his/her ICU stays, rather than just the last; this has been fixed.

• Fix case of hospital admission ID not unique to subject ID

Hospital admission 15449 was mapped to two different patients; this has been fixed.

• Add Exlihauser comorbidity scores

The Exlihauser comorbidity scores (calculated from DRG and ICD-9 codes) can be found in COMORBIDITY_SCORES

Known limitations, bugs.

• Missing data

Newly added data: DRGs, Demographics, Microbiology tests and procedures were not available for all subjects. Therefore, you will find that this data is missing for some subjects.

We were unable to complete the import of some of the text data. Radiology, Echo and ECG reports are missing from NOTEEVENTS.

• Waveform dates

The waveform metadata contains dates from the current waveforms available on PhysioNet:

https://archive.physionet.org/physiobank/database/mimic2db/

The dates found here will not match those found elsewhere in the database.

MIMIC-II v2.5

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.

Database statistics (adult records summary)

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

MIMIC-II v2.4

Changelog

• Bug #5: Fix problems with icustay generation
• Bug #13: Fix in “Unknown” service type in icustay_detail
• Bug #21: Rename ORDER column in PARAMETER_MAPPING table
• Bug #26: Make event date fields consistent
• Bug #27: Rename d_labitems.lab_category to category for consistency with other item tables
• Bug #40: Modify SAPS I calculation to correctly account for ventilated patients
• Feature #11: Delete the materialized view VITAL_SIGNS_RAW, because this was just a temporal data to be used by some statistics calculations.
• Feature #12: Add indices on cgid, itemid to chartevents
• Feature #14: Add AKI materialized view
• Feature #15: Add organ system failure materialized view
• Feature #16: Restore the Physionet 2009 challenge patients
• Feature #34: Re-generate SAPS scores after fixing Icustay problems
• Change request #36: Make the “Length Of Stay” columns consistent in minutes
• Change request #37: Replace ICUSTAY materialized view with ICUSTAYEVENTS table
• Change request #38: Make consistent some column names in ICUSTAY_DETAIL

Database statistics (adult records summary)

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)
SAPS I - day 2 ***	13,495		11.8 ±4.6	11.00	6 (9,15)
SAPS I - day 3 ***	9,070		12.2 ±4.5	12.00	6 (9,15)
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. 2nd and 3rd day data comes from version 2.5 of the database.