Health Claims Corpus (HCC) v1.0: 55.5 Million Automatically Extracted Disease–Factor Records, 2010–2026, with Linked PubMed Metadata

A research dataset containing 55,545,265 automatically extracted disease–factor records associated with 25,981,473 source document identifiers. Five linked tables provide claim labels, document metadata, disease–factor pair counts, and disease and factor term summaries. Labels include relation, direction, epistemic stance, inferred study category and assigned evidence category. Approximately 59% of document identifiers carry a PubMed identifier; associated publication-type metadata is included where matched. These automated labels have not been independently validated as clinical evidence or sentence-level annotations. Terms are free text and may include extraction errors or non-disease concepts. Dated records span 2010–2026; undated records are provided separately and 2026 is incomplete. Source coverage changes over time. Intended uses include extraction evaluation, network analysis and exploratory research with appropriate validation. Produced by Fliar BI. See the codebook and methodological limitations before analysis.

Produced by Fliar BI. Automatically extracted annotations are provided for research reuse, with a codebook and explicit validation limitations. This release is also under Harvard Dataverse review (reserved identifier doi:10.7910/DVN/CICSTQ); that submission is not yet a published release.