Protege Discovery Platform
Viewing as USER

MIMIC-IV Demo MEDS 0.0.1 EVENT_STREAM patient-level

PhysioNet · ODC-BY-1.0 (PhysioNet Open Data)
Request access

The same MIMIC-IV cohort reshaped into the MEDS event-stream standard with ready-made train/tuning/held_out splits. Ideal when you want temporal event sequences per patient rather than relational tables.

Potential training uses
  • Sequence models over patient event streams
  • Foundation-model pretraining on clinical events
  • Benchmarking with predefined data splits
Description
MEDS event-stream representation of MIMIC-IV Demo as Parquet train/tuning/held_out splits plus code and subject-split metadata.
Source
https://github.com/Medical-Event-Data-Standard
Tags
clinicalevent-streammedspatient-levelparquet

Request access

Listing details and schema are open for evaluation. Request access to unlock representative previews (sample rows and image thumbnails).

Snapshot & provenance

Snapshot kind
TRUSTED_LOCAL read-only reference (not copied)
State
PUBLISHED
Files
10
Package SHA-256
— (trusted local seed)
Published
2026-07-11T08:58:35.603193+00:00

Quality overview

0 error · 0 warning · 0 info

Assets (10)

DOCUMENT · 4

NameSource pathFilesRowsQuality
LICENSE.txt LICENSE.txt 1 OK
README.txt README.txt 1 OK
SHA256.txt SHA256.txt 1 OK
SHA256SUMS.txt SHA256SUMS.txt 1 OK

JSON_SIDECAR_COLLECTION · 1

NameSource pathFilesRowsQuality
metadata (json sidecars) metadata 1 1 OK

PARQUET_FILE · 2

NameSource pathFilesRowsQuality
codes.parquet metadata/codes.parquet 1 2661 OK
subject_splits.parquet metadata/subject_splits.parquet 1 100 OK

PARQUET_TABLE · 3

NameSource pathFilesRowsQuality
held_out data/held_out 1 42554 OK
train data/train 1 803992 OK
tuning data/tuning 1 69620 OK

Annotations

HUMAN_CURATED summary · APPROVED
MEDS event-stream form of MIMIC-IV, derived from the relational demo. Splits (train / tuning / held_out) support ML workflows.
HUMAN_CURATED customer_summary · APPROVED
The same MIMIC-IV cohort reshaped into the MEDS event-stream standard with ready-made train/tuning/held_out splits. Ideal when you want temporal event sequences per patient rather than relational tables.
HUMAN_CURATED training_uses · APPROVED
Sequence models over patient event streams Foundation-model pretraining on clinical events Benchmarking with predefined data splits