Standardized Multicenter Critical Care Database Integrating Minute-Level Vital Signs, Laboratory Tests, Interventions, and Outcomes: Profile of the OneICU Database
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Objective
To create a standardized, multicenter intensive care unit (ICU) database with minute-level data collection, enabling enhanced patient monitoring, early detection of clinical deterioration, and data-driven clinical decision support.
Design
Retrospective, multicenter observational study.
Setting
Tertiary care ICUs in Japan participating in the OneICU database project.
Patients
Critically ill patients admitted to ICUs at 8 tertiary care hospitals from 2013 to 2024.
Interventions
None.
Measurements and Main Results
The established database, OneICU, employs an extract-load-transform workflow to standardize diverse electronic medical record data into static, point time-series, and interval time-series tables. OneICU currently includes 118,272 ICU admissions from 97,718 unique patients, including minute-level recordings of vital signs, laboratory values, interventions, and diagnoses mapped to ICD codes. Compared with two benchmark databases— Medical Information Mart for Intensive Care (MIMIC)-IV and eICU—OneICU captures more frequent vital signs (minute-level vs. hourly in MIMIC-IV and every five minutes in eICU) and broader availability of the Sequential Organ Failure Assessment (SOFA) score components. In particular, the respiratory component is available for 64.1 % of admissions in OneICU versus 37.8 % in MIMIC-IV and 30.9 % in eICU, and the liver component for 86.6 % versus 45.2 % and 41.3 %, respectively. We compared the area under the receiver operating characteristic curve (AUROC) of machine-learning models predicting next-hour hypotensive events, which were defined as a median invasive mean arterial pressure < 65 mmHg or vasopressor initiation, across the three databases. Test-set AUROC was highest with OneICU (0.977) compared with MIMIC-IV (0.837) and eICU (0.950), attributable to OneICU’s minute-level vital signs and more complete covariate availability.
Conclusions
A high-resolution, multicenter ICU database integrating minute-level vital sign recordings with comprehensive SOFA score coverage is feasible and was associated with superior hypotension-prediction performance. OneICU enables detailed analyses of ICU trajectories and addresses the current scarcity of large-scale ICU data from Asian populations.
Key Points
Question
Does integrating high frequency physiological data and comprehensive clinical variables within a standardized multicenter ICU database improve machine-learning accuracy in predicting hypotensive events?
Findings
In this retrospective multicenter cohort study, we developed the OneICU database comprising 118,272 ICU stays from 97,718 patients across eight tertiary care hospitals in Japan. Using identical pipelines, models trained on OneICU achieved the highest test-set AUROC, attributable to minute-level vital signs and more complete covariate measurements.
Meaning
The OneICU database provides a high-resolution, generalizable resource for critical care research and addresses the current lack of large-scale ICU data from Asian populations.
