MIMIC-III, a freely accessible critical care database

Johnson, Alistair E. W.; Pollard, Tom; Shen, Lu; Lehman, Li Wei H.; Feng, Mengling; Ghassemi, Mohammad M.; Moody, Benjamin; Szolovits, Peter; Celi, Leo Anthony; Mark, Roger G.

doi:10.1038/sdata.2016.35

Cited by 5,317 publications

(3,794 citation statements)

References 13 publications

Supporting

Mentioning

3,432

Contrasting

Unclassified

Order By: Relevance

“…The data set we used is from MIMIC-III, a freely accessible critical care database [8]. We have two sets of experiments.…”

Section: Data Processingmentioning

confidence: 99%

“…Since our two 'local' hospital sites both came from splitting MIMIC III [8], technically the number of corresponding anchor pairs can account for 90% of all events. To create more realistic simulations, we tried different smaller fractions of all possible corresponding anchor pairs, and changed the rest of the pairs artificially to be labeled differently so no events could be recognized by the other site except for the corresponding events.…”

Section: Data Processingmentioning

confidence: 99%

“…However, permissions to access other's data in a central warehouse are still cumbersome to obtain and de-identification efforts are either costly, error prone, or ineffective [6,7]. Human-based de-identification effort costs over 5,000 hours and US$500,000 on Medical Information Mart for Intensive Care III (MIMIC-III) dataset [8], which contains only about 50,000 patient visits and 100 million words [9]. and produces error recall ranging from 0.63 to 0.94 [10].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Privacy-Preserving Predictive Modeling: Harmonization of Contextual Embeddings From Different Sources (Preprint)

Huang¹,

Lee²,

Wang³

et al. 2017

Preprint

View full text Add to dashboard Cite

Background: Data sharing has been a big challenge in biomedical informatics due to privacy concerns. Contextual embedding models have demonstrated very strong representative capability to describe medical concepts (and their context), and they have shown promise as an alternative way to support deep learning applications without the need to disclose original data. However, contextual embedding models acquired from individual hospitals cannot be directly combined because their embedding spaces are different and naive pooling renders combined embeddings useless.Objective: We present a novel approach to address these issues to promote sharing representation without sharing data. We can build a global model from representations learned from local private data without sacrificing privacy and synchronize information from multiple sources. Methods:We propose a methodology that harmonizes different local contextual embeddings into a global model. We use Word2Vec to generate contextual embeddings from each source and Procrustes to fuse different vector models into one common space by using a list of corresponding pairs as anchor points. With harmonized embeddings, we performed prediction analysis. Results:We used sequential medical events extracted from the Medical Information Mart for Intensive Care III database to evaluate the proposed methodology in predicting the next likely diagnosis of a new patient using either structured data or unstructured data. Under different experimental scenarios, we confirmed that the global model built from harmonized local models achieves more accurate prediction than local models and global model built from naive pooling.Conclusions: Such aggregation of local models using our unique harmonization can serve as the proxy for a global model, combining information from a wide range of institutions and information sources. It allows information unique to a certain hospital to become available to other sites, increasing the fluidity of information flow in health care.

show abstract

“…The data set we used is from MIMIC-III, a freely accessible critical care database [8]. We have two sets of experiments.…”

Section: Data Processingmentioning

confidence: 99%

Section: Data Processingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Privacy-Preserving Predictive Modeling: Harmonization of Contextual Embeddings From Different Sources (Preprint)

Huang¹,

Lee²,

Wang³

et al. 2017

Preprint

View full text Add to dashboard Cite

show abstract

“…Besides, changing the name from "Multiparameter Intelligent Monitoring in Intensive Care" to "Medical Information Mart for Intensive Care" [10], the database has also been augmented with collected data between 2008-2012. In addition, many data elements have been regenerated from the raw data in a more robust manner to improve the quality of the underlying data.…”

Section: Introductionmentioning

confidence: 99%

MIMIC III Database: A Descriptive Epidemiology of Severe Cholangitis Patient Cohort

Hedjoudje¹

2018

BJSTR

View full text Add to dashboard Cite

Objective: Electronic medical records include detailed information on clinical care. Besides its clinical utility, they afford researchers to evaluate impact of diagnostic and therapeutic decisions on patient outcomes. However, these dataset are not shared mainly for confidential reasons and remain consequently underused. MIMIC-III ('Medical Information Mart for Intensive Care') is a large, single-center database comprising information relating to patients admitted to critical care units at a large tertiary care hospital has been made publicly available. The objective of this study is to investigate and introduce the database content though the descriptive epidemiology of a severe cholangitis patient cohort. Method:We conducted a retrospective study of patients with severe cholangitis admitted to the intensive care unit of the Beth Israel Deaconess Medical Center using the MIMIC-III v1.4 ('Medical Information Mart for Intensive Care'). We investigated type of data available in the dataset such as patient characteristics, mortality, drug prescription, microbiology data and fluid balance.Result: MIMIC-III v1.4 stores 26 different tables with an overall number of 46,520 patients accounting for 58,976 hospitalizations. Records are stored in 21 tables and 5 tables are dictionary tables. The largest file contains more than 330 millions monitoring values data. We identified 125 patients with a severe cholangitis associated with a septic shock. Mean age was 70.75 (12.57) years. 52% of patients were male. Of the 125 patients, 27 died at the hospital leading to a in-hospital mortality of 21, 77%. Patients stayed 7.1 (9.35) in intensive care unit on average. Mean fluid balance was 111. 34 (73.15), 48.34 (55.94), 34.94 (50.83), 24.25 (48.79), 27.42 (38.92), 27.73 (61.59) and 26.36 (55.66) at 24, 48, 72, 96, 120, 144 and 168 hours. The most frequent microorganism found was E. Coli on 23, 41% followed by K. pneumonia in 6,12% of specimens. Blood culture was the microbiological test most widely prescribed to check for infectious pathogens.Conclusion: MIMIC-III v1.4 is a large, single-center database that contains a very large number of patients and hospital admissions. As illustrated by the septic shock associated cholangitis patients, a very broad type of data is stored.

show abstract

“…The increased adoption of high resolution longitudinal EHRs has created novel opportunities for researchers, clinicians and data scientists to access large, enriched patient databases (Harrison, Brady, and Rowan 2004) (Johnson et al 2016). The purpose of cleanEHR is to enable researchers to answer clinical questions that are important to patients.…”

mentioning

confidence: 99%