PubMed Central Patient Summaries With Relevance Annotations

Patient summaries extracted from 167,000 case reports in PubMed Central. The dataset includes 3.1 million patient-article relevance annotations and 293,000 patient-patient similarity annotations. It was created by the aisc-team-b1 for an AISC class at Harvard Medical School and was last updated in March 2024.

Use Cases

• Train models to predict patient-article relevance scores using patient summary text and article metadata.
• Develop patient similarity search algorithms based on the 293,000 annotated patient-patient pairs.
• Fine-tune language models for generating or summarizing clinical case reports from the 167,000 patient summaries.
• Benchmark retrieval-augmented generation systems using the patient-article relevance annotations as ground truth.

Strengths

• Contains 167,000 patient summaries, providing substantial text data.
• Includes 3.1 million relevance annotations, enabling large-scale training for retrieval tasks.

Limitations

• Annotations are derived from PubMed citations, which may not capture direct clinical relevance.
• The dataset is derived from published case reports, which may have a publication bias towards rare or complex cases.
• Specific column names and data structure are not provided in the input.

Provenance

Source

PubMed Central (PMC) case reports.

Collection Method

Extracted from case reports and annotated based on PubMed citation relationships.

Time Range

null

Freshness

Last updated in March 2024.

Geography

null