Rate Type Categories
Every rate in Clear Rates output belongs to one of four categories reflecting data defensibility.
Four Categories
| Category | Source | Description | Typical canonical_rate_type prefix |
|---|---|---|---|
| Posted Rates | Payer MRF, Hospital MRF | Fixed dollar amounts from MRF files. No estimation. Most defensible. | raw: payer_*, raw: hospital_*_dollar, impute: msdrg_mrf_base_rate_* |
| Real-World Rates | Payer MRF, Hospital MRF, Komodo | Variable-rate provisions where all components come from reported data (e.g. 90% of charges × hospital's own reported charge). | raw: hospital_*_allowed_amount, transform: hosp_per_diem_mult_glos, transform: *_gc_hosp_perc_to_dol |
| Enhanced Rates | MRF provision + estimated benchmark | MRF reports one component but not the charge. Turquoise fills in the missing component from Komodo or geographic averages. | transform: *_gc_komodo_perc_to_dol, *_gc_hosp_cbsa_perc_to_dol, most impute:* |
| Benchmark Rates | Medicare reference | Entirely derived from reference data — no reliable MRF data available. Last resort. | benchmark_* columns |
Tie-Breaking Priority
When multiple rates share the same integer accuracy score, the category hierarchy determines which wins:
| Priority | Category | Rationale |
|---|---|---|
| 1 | Posted Rates | All components from reported data, no estimation |
| 2 | Real-World Rates | All components from reported data, but rate varies by claim |
| 3 | Enhanced Rates | Provision reported, but one component estimated from benchmarks |
| 4 | Benchmark Rates | Entirely benchmark-derived — most speculative |
Where Rate Types Appear
canonical_rate_type records the exact column name with a category prefix (e.g., "raw: payer_negotiated_rate", "transform: hospital_perc_gc_hosp_perc_to_dol", "impute: rc_family_gc_hosp_perc_to_dol"). Users can filter by prefix to understand what fraction of coverage comes from each data quality tier.
When the orchestrator merges multiple months, a validated rate (score 7) from an older sub-version beats an unvalidated rate (score 6) from a newer one. Same-score ties across sub-versions use the rate type hierarchy: Posted beats Real-World beats Enhanced beats Benchmark.