{
  "dataset": "payer-coverage-requirements",
  "kind": "SAMPLE",
  "vintage": "2026-08-14",
  "generated": "2026-08-21",
  "the_slice": "One drug - infliximab - read across 12 insurers. Chosen as a segment rather than a random subset: the question this dataset answers is what it takes to get a drug approved and how much that differs by insurer, and you cannot see that in a scattering of unrelated rows.",
  "grain": {
    "cells": "payer_slug x drug x lob",
    "rules": "rule_id - one row per rule"
  },
  "scope": {
    "drug": "infliximab",
    "insurers": [
      "bcbs_fep",
      "bcbs_la",
      "bcbs_ma",
      "bcbs_tn",
      "blueshield_ca",
      "capital_blue",
      "centene",
      "excellus",
      "hmhs",
      "horizon",
      "kaiser",
      "wellmark"
    ],
    "cells": 12,
    "rules": 466,
    "prior_auth_required": 11,
    "cells_with_policy_date": 10
  },
  "rule_types_present": {
    "clinical_prerequisite": 228,
    "covered_indication": 76,
    "step_therapy": 28,
    "combination_rule": 26,
    "dosing": 24,
    "quantity_limit": 19,
    "reauthorization": 16,
    "prior_authorization": 14,
    "billing_coding": 13,
    "contraindication_exclusion": 9,
    "preferred_product": 7,
    "site_of_care": 5,
    "benefit_applicability_note": 1
  },
  "reading_it": {
    "two files": "cells is one row per insurer per drug - the summary. rules is the itemised requirements behind it, and is where the detail lives.",
    "start here": "Sort rules by payer_slug and read one insurer end to end, then compare it to the next. The differences between insurers are the product.",
    "rule_id": "The primary key. A single drug-and-insurer pair routinely carries many rules of the same type - one policy here states twenty separate clinical prerequisites - so payer/drug/rule_type/indication/hcpcs is NOT a key and joining on it fans rows out.",
    "citation_quote": "The insurer's own sentence, never reworded. Every rule carries one, and source_url points at the document it came from.",
    "hcpcs": "The billing code a rule attaches to, and the join to the rate datasets. Present on roughly 43% of rules; the rest are written about the drug rather than about a code, and we do not attach a code the policy did not name.",
    "attested_absent": "Requirements we looked for and confirmed are NOT in the policy. An empty list means we found everything we checked for; it does not mean we did not look.",
    "JSON columns": "products, covered_dx, site_of_care, citations, signals and value are JSON. They are nested in the source and are shipped that way rather than flattened lossily."
  },
  "things_that_will_look_odd": [
    "Every row is commercial. Medicare Advantage, Medicaid and exchange use the same column and are not populated in this build.",
    "confidence is high or medium, not a number. Nothing below medium ships.",
    "Most rules have no hcpcs. That is the policy's doing, not a gap in the read.",
    "indication is often empty for the same reason: the rule applies to the drug, not to one diagnosis.",
    "Some cells have no effective_date. The insurer did not print one; 85% of the full corpus does."
  ],
  "not_in_this_sample": [
    "The other 103 drugs. The full build carries 104.",
    "The 18 other insurers that also cover infliximab, and every insurer's rules for every other drug.",
    "Any negotiated rate. Coverage rules and rates are separate datasets."
  ],
  "full_build": {
    "cells": 2423,
    "drugs": 104,
    "insurers": 30,
    "documents_read": 2634
  },
  "files": [
    {
      "path": "carecost_payer_coverage_SAMPLE_2026-08-14_cells.csv",
      "bytes": 74110,
      "sha256": "dab7f5370bc270b3511db0914e41c3175c027d504c72e27188fd0bb04df72697"
    },
    {
      "path": "carecost_payer_coverage_SAMPLE_2026-08-14_rules.csv",
      "bytes": 213980,
      "sha256": "925b48addccd1b4681a8fceab37ef7c70768577bf336d3cfb8f86c296f2ee2a8"
    }
  ]
}