PricePortal: an open multi-state corpus of hospital price transparency machine-readable files for California and Indiana

PricePortal is an open, reproducible corpus harmonizing federal Hospital Price Transparency (HPT) machine-readable files (MRFs) for 528 Medicare-certified hospitals in California and Indiana with Medicare OPPS/MPFS benchmarks and ZIP-level community-risk variables. This deposit contains 16 analysis-grade parquet files plus a data dictionary and SHA-256 manifest (6.42 GB total) covering: (1) gross + cash chargemaster prices (305M rows, 478 hospitals); (2) payer-specific negotiated rates (417M rows, 411 hospitals); (3) Medicare CPT/HCPCS benchmark (9,709 codes, OPPS preferred + MPFS fallback); (4) hospital×code price-to-Medicare ratio panel (1,528,609 rows, four price types per row); (5) state and per-payer summary tables; (6) Wang 2023 within-hospital correlation replication, with a schema-stratified disambiguation test (kaiser_2023, misc_csv, §180 v1.x/v2.x/v3.x) and a discounter-threshold sensitivity analysis (5%, 15%, 30%); (7) Chang & Psek 2024 ZIP-level socioeconomic gradient extension (257 hospital ZIPs; pooled cash-ratio βlog_inc=−0.96, p=0.063); (8) hospital identity crosswalk (CCN ↔ OSHPD ↔ NPI ↔ EIN, 528 hospitals; 95.6% NPI coverage); (9) Indiana ZIP-level ACS 2024 5-year demographics and CDC PLACES health outcomes. Headline findings. Median chargemaster-to-Medicare ratio is 3.92× in California vs 2.19× in Indiana; cash 1.42× vs 0.90×; minimum-negotiated 0.95× vs 0.55×, consistent with Indiana HEA 1004's Medicare-repricing benchmark. Within hospital, the gross–cash log-correlation is ≈1.0 across every MRF schema bucket — including pre-§180 Kaiser legacy chargemasters and custom non-§180 formats — weighing decisively against a v3-template encoding-artifact explanation. See DATA_DICTIONARY.md for full schema, provenance, and caveats per file. File integrity is verifiable via MANIFEST.sha256. Source code: iupui-soic/mrf-pricing-research. Public portal: pricingapp.streamlit.app.PricePortal是一个开放、可复现的语料库，将加利福尼亚州与印第安纳州共528家医疗保险认证（Medicare-certified）医院的联邦医院价格透明度（Hospital Price Transparency, HPT）机器可读文件（machine-readable files, MRF）与医疗保险门诊预期支付系统（Outpatient Prospective Payment System, OPPS）及医师费用表（Medicare Physician Fee Schedule, MPFS）基准，以及邮编级社区风险变量进行了统一整合。本数据集包含16个分析级帕奎特（parquet）格式文件，外加一份数据字典与SHA-256校验清单（总大小6.42GB），涵盖以下内容： (1) 总费用与现金价收费清单（chargemaster）价格（3.05亿行，覆盖478家医院）； (2) 特定付款方协商费率（payer-specific negotiated rates）（4.17亿行，覆盖411家医院）； (3) 医疗保险CPT/HCPCS（Current Procedural Terminology/Healthcare Common Procedure Coding System）基准编码（共9709条，含OPPS首选编码与MPFS备选编码）； (4) 医院×编码的价格-医疗保险费率比面板（1528609行，每行包含4种价格类型）； (5) 州级与各付款方汇总统计表； (6) Wang 2023年院内相关性复现研究数据集，包含分层模式消歧测试（kaiser_2023、misc_csv、§180 v1.x/v2.x/v3.x）以及折扣阈值敏感性分析（阈值设为5%、15%、30%）； (7) Chang与Psek 2024年的邮编级社会经济梯度拓展研究数据集（覆盖257个医院邮编区域；合并现金价比的β_log_inc=−0.96，p=0.063）； (8) 医院身份交叉映射表（医疗保险认证编号（CMS Certification Number, CCN）↔ 加州卫生规划与发展局（Office of Statewide Health Planning and Development, OSHPD）↔ 国家提供者标识符（National Provider Identifier, NPI）↔ 雇主识别号（Employer Identification Number, EIN），覆盖528家医院；NPI覆盖率达95.6%）； (9) 印第安纳州邮编级美国社区调查（American Community Survey, ACS）2024年五年期人口统计数据与疾病控制与预防中心（Centers for Disease Control and Prevention, CDC）PLACES健康结局数据。 ### 核心研究发现 加利福尼亚州的收费清单价格与医疗保险费率比的中位数为3.92倍，印第安纳州则为2.19倍；现金价对应比值分别为1.42倍与0.90倍；最低协商费率比值分别为0.95倍与0.55倍，这与印第安纳州HEA 1004的医疗保险重新定价基准一致。在每家医院内部，所有MRF模式分组（包括§180出台前的Kaiser遗留收费清单与非§180自定义格式）中，总费用与现金价的对数相关性均接近1.0，这一结果有力驳斥了“v3模板存在编码伪影”的解释假说。 完整的数据集结构、来源与使用注意事项详见DATA_DICTIONARY.md。可通过MANIFEST.sha256校验文件完整性。 源代码仓库：iupui-soic/mrf-pricing-research。公开访问门户：pricingapp.streamlit.app。