Nursing-Home CMS-2567 Deficiency Findings: an open dataset (Dark Health Data)

Dark Health Data: CMS-2567 nursing-home Statements of Deficiency, extracted via Claude claude-haiku-4-5 (OCR for scanned pages) with a verification layer. This release (v0.4.1) is an expanded national crawl of states that post 2567 PDFs publicly: 17,826 records from 920 source documents (705 facilities) across 30 states: 10,733 cited deficiencies, 7,093 plans of correction. Many 2567s are scanned images read by OCR (tesseract), so trust scores are lower and more rows are flagged than in the born-digital datasets. The crawl is ongoing; later versions add more facilities. Every record carries full provenance and a quality/trust score; nothing is imputed or dropped. AI-extracted (with OCR), not yet independently validated — preliminary; filter on the trust score. Public-record documents only; no PHI. Code (Apache-2.0): github.com/sanjaybasu/dark-health-data. CC0-1.0.