Welcome

AmsterdamUMCdb is the first freely accessible European intensive care database. It is endorsed by the European Society of Intensive Care Medicine (ESICM) and its Data Science Section. It contains de-identified health data related to tens of thousands of intensive care unit admissions, including demographics, vital signs, laboratory tests and medications.

Version

The current version of AmsterdamUMCdb is 1.0.2, released in March 2020. This version contains data related to 23,106 intensive care unit and high dependency unit admissions of adult patients from 2003-2016. The first version of AmsterdamUMCdb (1.0) was released in November 2019.

Requesting Access

The database, although de-identified, still contains detailed information regarding the clinical care of patients, so must be treated with appropriate care and respect and cannot be shared without permission. To request access, go to the Amsterdam Medical Data Science website.

Acknowledgement

When using AmsterdamUMCdb in your research, please cite:

Thoral, P. J., Peppink, J. M., Driessen, R. H., Sijbrands, E. J. G., Kompanje, E. J. O., Kaplan, L., Bailey, H., Kesecioglu, J., Cecconi, M., Churpek, M., Clermont, G., van der Schaar, M., Ercole, A., Girbes, A. R. J., Elbers, P. W. G., on behalf of the Amsterdam University Medical Centers Database (AmsterdamUMCdb) Collaborators and the SCCM/ESICM Joint Data Science Task Force (2021). Sharing ICU Patient Data Responsibly Under the Society of Critical Care Medicine/European Society of Intensive Care Medicine Joint Data Science Collaboration: The Amsterdam University Medical Centers Database (AmsterdamUMCdb) Example. Crit Care Med. 2021 Jun 1;49(6):e563-e577. doi: 10.1097/CCM.0000000000004916. PMID: 33625129; PMCID: PMC8132908.

Facts and Figures

The current database contains data from the clinical patient data management system of the department of Intensive Care, a mixed medical-surgical ICU, from Amsterdam University Medical Center. The clinical data contains 23,106 admissions of 20,109 patients admitted from 2003 to 2016 with a total of almost 1.0 billion clinical observations consisting of vitals, clinical scoring systems, device data and lab results data and nearly 5.0 million medication records.

Available tables

The table and field definitions are available from the AmsterdamUMCdb wiki and from Jupyter Notebooks in the tables folder.

amsterdamumcdb Python package

The amsterdamumcdb Python package contains common functions for working with the database.

Available functions

amsterdamumcdb.get_dictionary():

Gets a dictionary of all items in AmsterdamUMCdb with translated medical concepts mapped to LOINC, SNOMED CT or ATC (work in progress).

• Returns:
  • dataframe containing dictionary

amsterdamumcdb.get_fluidbalance(admissionid, from_date, to_date, con):

Calculate the fluid balance (i.e. fluid input - fluid output) for a specific time interval. Typically this will be used to calculate daily fluid balances.

• Arguments:
  • admissionid -- the admissionid of ICU admission
  • from_date -- the start of the interval, expressed as the number of milliseconds from start of this ICU admission
  • to_date -- the end of the interval, expressed as the number of milliseconds from start of this ICU admission
  • con -- psycopg2 connection or pandas-gbq Google BigQuery config

amsterdamumcdb.outliers_histogram(data, z_threshold=4.0, lower=None, upper=None, bins=None, binwidth=None, maxbins=None):

Return a pyplot histogram, where the upper and/or lower outliers are binned together for a more even distribution plot. By default, the histogram will be created with bins that are on boundaries aligned with the pyplot axis ticks.

• Arguments:
  • data -- dataframe to create the histogram for.
  • z_threshold (Optional) -- number of standard deviations from the median to determine outliers
  • lower (Optional) -- lower threshold for binning lower outliers together
  • upper (Optional) -- upper threshold for binning higher outliers together
  • bins (Optional) -- int or sequence or str, allows specifying the number of bins, the actual bins or a binning strategy (see: https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.hist.html)
  • binwidth (Optional) -- forces a specified size of the equally sized bins between the lower and upper threshold
  • maxbins -- the maximum number of equally sized bins to create

Changelog

version 0.2.1

• unified PostgreSQL and BigQuery processing
• Fixed filtering based boolean indexing using np.NaN (pandas default) and pd.NA (used by pandas-gbq since the default is Int64) leading to different results, particularly when finding the complement using negation.
• Added unit tests for the cohorts. Forces pd.Na to be considered False during comparisons.
• By default, uses only validated data for amsterdamumcdb functions for consistent PostgreSQL and BigQuery results.
• Removed aggregation in SQL code of Glasgow Coma Scale (GCS) in line with other functions.
• Update SOFA GCS processing for more accurate scoring
• Added rounding to 1 decimal for conversions (from kPa to mmHg) in SQL statements
• Updated deprecated matplotlib style (seaborn-darkgrid)
• Update engine to SQLAlchemy for official pandas support instead of direct psycopg2 connection
• Moved code from Jupyter reason_for_admission.ipynb notebook into amsterdamumcdb.cohorts module.

version 0.2.0

• added sample cohorts (sepsis-3, mechanical ventilation, shock) in cohorts.py
• added SOFA scoring to scores.py
• util functions for PostgresSQL to BigQuery translation
• move SQL code from notebooks to separate SQL files

version 0.1.5

• dictionary: as part of the ICUnity project in preparation for the The Dutch ICU Data Warehouse a number of commonly used items have been mapped to LOINC , SNOMED CT or ATC with data in the vocabulary_id, vocabulary_concept_code, vocabulary_concept_name columns.
• dictionary: UCUM units added

version 0.1.4

• fixes binning in outliers_histogram() when all values < 1

version 0.1.3

• Improves binning in outliers_histogram()

version 0.1.2

• Allows negative dates in get_fluidbalance()

version 0.1.1

• Fixes Google BigQuery Project must be a string error.

version 0.1.0

• initial release for ESICM Datathon 2021