ricu: R's interface to intensive care data.

Area of Science:

• Critical Care Medicine
• Biomedical Informatics
• Data Science

Background:

• Analyzing data from multiple intensive care unit (ICU) databases is challenging due to data heterogeneity.
• Existing ICU datasets are often siloed, requiring separate processing and harmonization efforts for each.
• A unified framework is needed to facilitate large-scale, reproducible research using diverse ICU data sources.

Purpose of the Study:

• To develop a unified computational framework for analyzing data from five large, publicly available intensive care unit (ICU) datasets.
• To create an R-package that simplifies data access, harmonization, and analysis across different ICU databases.
• To promote data sharing and collaborative research in critical care medicine.

Main Methods:

• Integrated data from three American (MIMIC-III, MIMIC-IV, eICU) and two European (AmsterdamUMCdb, HiTRUCS) ICU databases.
• Developed a mapping to a common set of clinical concepts, utilizing the Observational Medical Outcomes Partnership (OMOP) Vocabulary.
• Synchronized units of measurement and data types, and implemented a unified Application Programming Interface (API) for data loading.
• Created the `ricu` R-package to provide computational infrastructure for handling these harmonized ICU datasets.

Main Results:

• The `ricu` R-package enables simultaneous analysis of data from five major ICU databases.
• The package allows users to load 119 existing clinical concepts from the harmonized data sources.
• The developed framework standardizes data representation and facilitates efficient data retrieval and setup.

Conclusions:

• The `ricu` R-package is the first tool enabling simultaneous analysis of multiple publicly available ICU datasets.
• This unified interface significantly saves researcher time and improves the reproducibility of ICU data studies.
• Future work aims to expand the concept dictionary for a more comprehensive data harmonization, fostering a community-wide effort.