Skip to main content
Version: 3.0

Rate Arrays

Parallel arrays per ROID, with rate values, accuracy scores, and metadata in corresponding positions.

Array Structure

ArrayContent at position iExample
rate_arrayDollar value125.00
rate_score_arrayAccuracy score (0–7.x)7.00000125
source_arrayData source label"payer"
rate_type_arrayColumn name with prefix"raw: payer_negotiated_rate"
methodology_arrayContract methodology"Fee Schedule"
rate_class_array"Raw", "Transform", or "Impute""Raw"

Position alignment is guaranteed: position i in rate_array corresponds to position i in every other array. ARRAY_POSITION on rate_score_array yields the index used to extract all other fields.

Priority Ordering

Array position matters when multiple rates have identical scores. The ordering differs by bill type:

Inpatient:

  1. Hospital raw rates
  2. Hospital untransformed rates
  3. Payer raw rates
  4. Payer untransformed rates
  5. Hospital transforms (provider GC)
  6. Payer transforms (provider GC)
  7. Hospital transforms (CBSA GC)
  8. Payer transforms (CBSA GC)
  9. State-level transforms
  10. Imputations

Non-Inpatient: Similar structure, but untransformed rates get slight priority within each tier, provider-level GC comes before CBSA, and hospital source is placed before payer within each tier.

No-impute variants

*_no_impute array variants track what the canonical rate would be without imputations — for example, best_idx_no_impute records the rate selection result excluding imputed values. These are used internally to measure imputation lift and are available in prod_combined_all.