Medicaid Managed-Care External Quality Review (EQR) Findings: a national open dataset (Hidden Health Data)

Hidden Health Data: findings from U.S. state Medicaid and CHIP managed-care External Quality Review (EQR) annual technical reports (42 CFR 438.350-438.364), extracted by a large language model (Anthropic Claude claude-haiku-4-5) into three tables. v0.4.1 is a documentation update of v0.4.0; all data files are byte-identical. It contains 25,190 records from 49 reports (44 states and jurisdictions; 43 with records): 11,733 quality measures, 2,653 performance improvement projects, and 10,804 compliance findings. Report years 2023-2026; 96.3% of quality-measure rows refer to measurement years 2022-2025. Checks. Five consistency rules check each quality-measure record; qa_status and trust_score (1.00 pass, 0.83 warn, 0.00 fail) reflect these rules only. Every record links to its source report and page. Human validation (quality-measure records). Field accuracy 92.3% (95% CI 88.6%-94.8%; 284 records); recall 83.1% (95% CI 78.1%-87.2%; 217 of 261 values), with most misses in earlier-year columns of multi-year tables. Improvement-project and compliance records were not validated. See VALIDATION_STATUS.md. Public-record documents only; no PHI. CC0-1.0. Code (Apache-2.0): github.com/sanjaybasu/dark-health-data.