6

LABEVENTS data is from the actual hospital wide laboratory data management system. CHARTEVENTS data is from the ICU database that is used as an electronic chart for the patient at the bedside. Since the chart needs to contain lab values, as they are relevant to the patient's health, the laboratory data is pulled from the hospital wide laboratory database ...


6

It is important to be aware that the ICD-9 codes are assigned to hospital stays primarily for billing purposes. The codes are recorded by a coding team upon patient discharge, following a review of the health records. No specific time or date is associated with the codes. Two relevant timestamps are the admission and discharge times associated with the ...


5

The Metavision workstation allows physicians to add notes relating to patients. These notes are created using template forms, which include a 'Set as Error' button as shown in the following screenshot (captured from the training notes highlighted below): A '1' in the ISERROR column of the noteevents table indicates that a physician has selected the 'Set as ...


5

The official citation for the MIMIC-III database is: MIMIC-III, a freely accessible critical care database. Johnson AEW, Pollard TJ, Shen L, Lehman L, Feng M, Ghassemi M, Moody B, Szolovits P, Celi LA, and Mark RG. Scientific Data (2016). DOI: 10.1038/sdata.2016.35. Available at: http://www.nature.com/articles/sdata201635 The following URL includes ...


5

The ICUSTAY_DETAIL table was shared by researchers at the Laboratory for Computational Physiology to simplify certain queries on the MIMIC-II database. It is derived from other tables within the database and so does not form part of the 'core' dataset. ICU_EXP_FLG is not included in MIMIC-III database (as of version 1.2) because we have not yet attempted to ...


5

You can use this query (tested on MIMIC-III v1.1, took around 80 seconds to run on my computer): -- ------------------------------------------------------------------ -- Title: Extract height and weight for ICUSTAY_IDs -- Description: This query gets the first, minimum, and maximum weight and height -- for a single ICUSTAY_ID. It extracts data from ...


5

There is a known issue with the PPG signals in the MIMIC II/III waveform database, although I don't know whether this is the same issue as what you're seeing. The issue we've seen is that the PPG seems to run "slow" compared to the other signals. For example, if you measure the pulse transit time (the difference in time between the start of the QRS complex ...


5

Patients may move between various care units over a single hospital admission. MIMIC-III provides significantly more detail than MIMIC-II about these movements thanks to the addition of the transfers table. As noted in the MIMIC documentation, icustay_id is an identifier that has been generated using the transfers table for convenience of analysis. The ...


5

Negative cultures were added to the microbiologyevents table in version 1.4 of MIMIC-III: http://mimic.physionet.org/about/releasenotes/ #188 - Negative microbiology cultures are now recorded in MICROBIOLOGYEVENTS - negative cultures can be identified by a NULL value for the organism name (ORG_NAME). Negative cultures were not available in prior ...


4

Adding more severity scores is in the todo list: APACHE score or similar #28: We plan to release code for calculating severity scores as soon as we have time to! On the list are SAPS, SOFA, APS III, and SAPS II. The code to compute the Elixhauser comorbidity measure is available for MySQL and Postgres.


4

Here is my take: is it appropriate to simply ask for “diabetic” patients? Would the method of identifying the cohort vary wildly depending on the type of diabetes of interest? A: It would not be appropriate to simply ask for "diabetic" patients. Patients with diabetes are usually classified by the type (Type 1 or Type 2, which roughly correlates to whether ...


4

Code for deriving an Elixhauser comorbidity table for the MIMIC-III Critical Care Database has been developed by Alistair Johnson and shared via the MIMIC Code Repository. The repository can be found by following the "Code" link on the front page of the MIMIC website.


4

Two different ICU data management systems were in operation over the period of data collection: Philips Carevue (pre ~2008) and IMDSoft Metavision (post ~2008). The structure of the data collected by these systems is fairly different, so we have provided two separate tables rather than a single merged one. The tables are inputevents_cv (Carevue) and ...


4

I am new to SQL and am trying to explore the MIMIC-II dataset using the query builder so the logical step seemed to be follow the tutorials within the introduction document [1]. The current version of the MIMIC database is MIMIC-III. MIMIC-II is an old version that is no longer supported by the MIT Laboratory for Computational Physiology, so I would not ...


4

You're correct - the category type "PROCEDURE" in the mimic2v26.d_codeditems table refers to ICD-9 procedure codes. These have been more explicitly delineated in MIMIC-III: the equivalent definitions are in D_ICD_PROCEDURES, and the actual codes for patients are in PROCEDURES_ICD.


4

Code for severity scores and other related clinical concepts can be found in the mimic-code repository: https://github.com/MIT-LCP/mimic-code


4

On MIMIC-III v1.1 and v1.3: SELECT DEATHTIME FROM mimiciii.admissions WHERE DEATHTIME IS NOT NULL ORDER BY DEATHTIME ASC LIMIT 500 ; Maybe you use MIMIC-II.


4

While hospitals charging for data access is a disappointing trend, I don't think MIMIC paid for access in the sense implied by the question. The research and development of MIMIC is the product of collaboration among MIT, Philips Healthcare, and Beth Israel Deaconess Medical Center sponsored by National Institutes of Health grants: NIH-R01-EB017205, NIH-R01-...


3

dose_val_rx is a varchar column that contains non-numerical values (for example, sometimes a range is specified such as '1-2' tablets). It isn't clear how you would sum these values, so most database systems will return a type error. Postgres, for example, returns: HINT: No function matches the given name and argument types. You might need to add explicit ...


3

It looks like you are trying to load MIMIC-III (v1.1) data using the MIMIC-III (v1.2) build scripts. edregtime is a new column that was added between v1.1 and v1.2. Major issues addressed, including additional data made available: ... #151 - The time of emergency department registration and exit has been added to the admissions table, where ...


3

That's a typo in the documentation page. The DIAGNOSIS column in the ADMISSIONS table is the diagnosis on hospital admission. It should read 58,976 admissions. It has been corrected now, and reads: The DIAGNOSIS column provides a free text diagnosis for the patient on hospital admission. The diagnosis does not use a systematic ontology: as of MIMIC-III ...


3

For around 61046 admissions out of the 61532 admissions (~99.2% admissions), the ICU intime occurs after the hospital admission time. SELECT i.intime - a.admittime as tdiff, a.subject_id, a.hadm_id, a.admittime, i.intime, a.dischtime, i.outtime FROM mimiciii.admissions a INNER JOIN mimiciii.icustays i ON a.hadm_id = i.hadm_id ORDER BY i....


3

When was 1.3 officially released? When was 1.2 officially released? MIMIC-III v1.3 was released on 10 December 2015. MIMIC-III v1.2 was released on 20 November 2015. What is the release history for MIMIC-III? Are there other versions? The first version of MIMIC-III v1.0 was made available on 25 August 2015. For an outline of the release history for ...


3

The SUBJECT_IDs in MIMIC-II are the same as SUBJECT_IDs in MIMIC-III. HADM_IDs are different because MIMIC-II used an ICU database to define hospital admission and discharge (and only had dates), whereas MIMIC-III uses a hospital database (and has dates and times). You could create an approximate mapping, but since there will be erroneous entries in both ...


3

For Q1 see diagnoses_icd.sequence in MIMIC-III: does the order matter aside from the primary diagnosis? and Priority sequence column in diagnosis_icd. (#156 - The SEQUENCE column in DIAGNOSES_ICD and PROCEDURES_ICD has been renamed SEQ_NUM in MIMIC-III v1.2) For Q2: see https://github.com/MIT-LCP/mimic-code/issues/24: Question: I see 47 ...


3

I will assume you are using the MIMIC-II v3 Waveform database. If you look at section 5.4 of the MIMIC-II user guide (warning, 2.3 MB PDF), you will see the following: Inter-waveform alignment problems. The method used for MIMIC waveform data extraction was not designed for inter-waveform analysis. The waveform data contain unspecified/unknown ...


3

The data files conform to the RFC 4180 specification, which is as close as CSVs get to a formal standard. Loading a RFC 4180 formatted CSV is trivial for many database systems. For example, the load command in Postgres is: \copy ADMISSIONS FROM 'ADMISSIONS.csv' DELIMITER ',' CSV HEADER Using a different delimiter is an option, but there would need to be a ...


3

The chartevents table of MIMIC-III is populated with data that has been archived by Metavision ICU, a clinical information system that integrates charted data from various sources, for example heart rates from bedside monitors. Data from the bedside monitors must be validated by a caregiver in order for it to be stored as part of a patient record. If a ...


3

You've actually found a bug in MIMIC-III v1.3. Some of the CD counts were muddled - actual counts (#/uL) were merged with percentages, and vice versa. If you look at the LOINC code for percent CD8 (8137-2), you can see the same issue: loinc | valueuom | count ------ | -------- | ------ 8137-2 | #/uL | 1580 8137-2 | % | 566 edit: This has ...


2

When they type their progress notes, some nurses use all caps and some don't! Just personal style!


Only top voted, non community-wiki answers of a minimum length are eligible