Skip to main content
Version: 2.2

Hospital MRF Data

SchemaMethodology

CLD pulls Hospital MRF data from glue.hospital_data.hospital_rates.

Hospital MRF Data represents the direct ingestion of hospital-posted machine-readable files (MRFs), parsed into a standardized schema for analysis. Turquoise processes over 6,000 hospital MRFs quarterly, plus weekly checks for updates, to maintain a near‑current dataset of payer‑negotiated hospital rates.

Data Ingestion & Refresh Cadence​

  • Quarterly: Priority download order: (1) Customer-requested providers, (2) Large systems, (3) Small/independent facilities.
  • Weekly: Automated URL checks to detect updates via CMS-mandated cms-hpt.txt files. Hospitals lacking these files require manual retrieval.
  • Parsing attempts use schema-compliant v2.2 parsers first; legacy custom parsers are updated as needed. Files failing validation go to a QA queue for manual intervention before ingest.
  • All updated records fully replace previous versions.

Schema & Data Fields​

  • Each ingested file feeds the hospital_rates table, with key fields including:
    • billing_code, billing_code_type (e.g. CPT, HCPCS, DRG variants)
    • negotiated_dollar, negotiated_percentage
    • gross_charge, discounted_cash_rate
    • provider metadata: name, state, city, CBSA, NPI, EIN
    • plan labeling: raw plan name, payer name, payer class, payer product network.
  • File and ingestion metadata tracks original filenames, timestamps, schema compliance, and Turquoise’s transparency scores.

Plan Labeling Logic​

  • Raw plan names are standardized into Payer Class (Commercial, Medicare Advantage, Managed Medicaid, etc.) via a hybrid machine learning + rules-based approach.
  • Payer and Product labels also follow ML-based classification, refined with rule-based corrections for mislabels.
  • Products (e.g. HMO, PPO, EPO, POS) are tagged using both text rules and payer-specific mappings.

Clinically Implausible Filtering​

  • For non‑hospital taxonomies (e.g. outpatient facilities), Turquoise excludes rates for services the provider type would never perform.
  • This filter relies on claims data to define the top 99 % of procedures by volume within CPT and HCPCS for that provider type.

Outlier Management (Not Used in Clear Rates)​

Rates in hospital_rates are flagged with rate_is_outlier = TRUE if they meet any of these criteria:

  1. <50 % or >1000 % of the Medicare reference price (if available)
  2. >$$500 k on outpatient, non‑drug billing codes
  3. >$2.5 M on inpatient, non‑drug billing codes
  4. <10 % or >1000 % of ASP for drug codes
  5. Negotiated dollar \≤0
  6. Negotiated percentage >100% with contract_methodology = 'percent of total billed charges'

An outlier_reason field details which rule was triggered.

On this page: