RWE-bench-MIMICIV: Observational Study Protocols from MIMIC-IV (2012-2019)

54 structured observational study protocols and results extracted from real-world evidence (RWE) studies using the MIMIC-IV database. The dataset, created by rwe-bench, contains JSON files representing individual studies sourced from MIMIC-IV version 2.2, covering data from 2012 to 2019.

Use Cases

• Benchmarking real-world evidence extraction methods based on structured study protocols.
• Analyzing trends in observational study design based on metadata fields like title and abstract.
• Training or evaluating natural language processing models on clinical study documentation.
• Comparing research protocols across a corpus of studies derived from a common clinical database.

Strengths

• Contains 54 distinct study samples, providing a defined corpus for analysis.
• Data is sourced from the authoritative MIMIC-IV v2.2 clinical database.
• Covers a 7-year time range from 2012 to 2019, offering longitudinal context.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale tasks.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

MIMIC-IV v2.2 database

Collection Method

Extracted from real-world evidence studies

Time Range

2012-2019

Freshness

Last updated 2026-05-01 06:41:23; freshness should be verified.

Geography

null