MIMIC-IV Benchmark Tasks

Standardized ICU benchmarks on MIMIC-IV — mortality, LOS, sepsis, AKI, drug dosing.

Composite
89.4
Experimental validation
Clinical
Stages
Phase IIIClinical DevelopmentPost-market / RWE
Modalities
cross-modality
Task types
clinical-outcome
Size
patients: 299,712
tasks: 14
License
PhysioNet Credentialed
First release
2020
Last updated
2025-02
Official site
→ project page
Leaderboard
→ leaderboard
Dataset
→ dataset
Code / GitHub
→ repository
HuggingFace
→ HF
Paper
MIMIC-IV, a freely accessible electronic health record dataset · Johnson AEW, Bulgarelli L, Shen L, et al. · 2023 · paper · doi:10.1038/s41597-022-01899-x · 620 citations
Flags
none
Experts
Alistair Johnson, Leo Anthony Celi, Roger G. Mark
Groups
MIT Lab for Computational Physiology
Hosted by
MIMIC-IV / eICU
Related benchmarks
CT-Outcome (TrialBench v2)

Rubric (7-criterion)

rigor
5
coverage
4
maintenance
5
adoption
5
quality
5
accessibility
3
industry_relevance
4

Notes

Canonical clinical ML benchmark. Credentialed access limits casual use.

← Back to all benchmarks

Compare:
Open comparison →