Abstract:Background: A clinically meaningful comparative assessment of medical decision support in emergency care is challenging due to a lack of appropriate datasets with multimodal input modalities and comprehensive prediction task. This hampers measurable progress in the field. Results: We introduce a dataset based on MIMIC-IV, a benchmarking protocol, and initial results for evaluating multimodal decision support in the emergency department (ED). We use diverse data modalities from the first 1.5 hours after patient arrival, including demographics, biometrics, vital signs, lab values, and electrocardiogram waveforms. We analyze 1443 clinical labels across two contexts: predicting diagnoses and patient deterioration. Our diagnostic model achieves an AUROC score over 0.8 in a statistically significant manner for 609 out of 1428 conditions, including cardiac conditions like myocardial infarction and non-cardiac conditions such as renal disease and diabetes. The deterioration model scores above 0.8 in a statistically significant manner for 14 out of 15 targets, including critical events like cardiac arrest, mechanical ventilation, intensive care unit admission, as well as short- and long-term mortality. Furthermore, we provide one of the first robust demonstrations of the significant impact of raw waveform input data on model performance. Conclusions: This study highlights the proposed dataset as a unique resource to foster progress towards measurable progress in the domain of algorithmic decision support in emergency care. The presented multimodal baseline models showcase the potential of diagnostic decision support in the field and provide strong incentives for including raw waveform data.

MIMIC-IV, a freely accessible electronic health record dataset

MIMIC-III, a freely accessible critical care database

Benchmarking with MIMIC-IV, an irregular, spare clinical time series dataset

MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports

The MIMIC Code Repository: enabling reproducibility in critical care research

Accessing Critical Care Big Data: a Step by Step Approach.

Mimic-IV-ICD: A new benchmark for eXtreme MultiLabel Classification

The eICU Collaborative Research Database, a freely available multi-center database for critical care research

MIMIC-CXR-JPG, a large publicly available database of labeled chest radiographs

PIC, a Paediatric-Specific Intensive Care Database

Reproducing a Prospective Clinical Study as a Computational Retrospective Study in MIMIC-II

Experimental Evaluation and Development of a Silver-Standard for the MIMIC-III Clinical Coding Dataset

Salzburg Intensive Care database (SICdb): a detailed exploration and comparative analysis with MIMIC-IV

MDS-ED: Multimodal Decision Support in the Emergency Department -- a Benchmark Dataset for Diagnoses and Deterioration Prediction in Emergency Medicine

Synthetic Acute Hypotension and Sepsis Datasets Based on MIMIC-III and Published as Part of the Health Gym Project

A Multidatabase ExTRaction PipEline (METRE) for Facile Cross Validation in Critical Care Research

Benchmarking emergency department prediction models with machine learning and public electronic health records

Automated Medical Coding on MIMIC-III and MIMIC-IV: A Critical Review and Replicability Study

Copycats: the many lives of a publicly available medical imaging dataset

A Corpus for Detecting High-Context Medical Conditions in Intensive Care Patient Notes Focusing on Frequently Readmitted Patients

eICU Collaborative Research Database application and data extraction process