IDSA/ATS Score Distribution: NLP vs. Human Risk Labels

A 5.2 KB dataset by Alejandro Hernández-Arango, last updated April 30, 2026, compares NLP-extracted and human-assigned IDSA/ATS clinical risk scores. Score distributions are presented under both labeling scenarios, along with the percentage of patients reclassified to a higher or lower risk stratum.

Use Cases

• Evaluating NLP model performance for clinical risk scoring based on the comparison of extracted and human labels.
• Analyzing the impact of automated labeling on patient risk stratification based on reclassification percentages.
• Benchmarking natural language processing tools for extracting structured scores from medical text.

Strengths

• Dataset is openly licensed under CC-BY-4.0.
• Directly compares NLP-extracted labels with a human ground truth, a key evaluation feature.

Limitations

• Dataset is very small at 5.2 KB, indicating limited scope.
• Row count and column-level documentation are unknown, requiring manual inspection after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

figshare

Freshness

Last updated 2026-04-30 17:49:47; freshness should be verified.