Saved result
Submitted arguments
{
"filer": "hca-medical-city"
}Native result and verdict
{
"does_not_license": {
"reading": "NO READING IS ATTACHED. These fields carry no interpretation, and this world does not supply one unasked. To obtain a reading, and to find out whether what you are about to write is supported, call judgement_check with receipt \"r886\" and the claims you intend to make. A claim written without checking it is a claim this world did not license.\n\nIf you can write your claim as an expression over the figures above, for example {\"op\": \"<=\", \"left\": \"code_types.MS-DRG.code_count\", \"right\": 12, \"scope\": \"marginal\"}, call claim_check with the same receipt instead. It decides by entailment over those figures and never reads your sentence, so it answers the same way in any language. The two checks are alternatives and neither falls back to the other."
},
"kind": "served",
"licence": {
"coverage_database": "Contains information from the CMS Medicare Coverage Database.",
"cpt": "CPT is copyright the American Medical Association. Served under a distribution licence.",
"linkback": "https://hammer.ai/coverage-intelligence",
"terms": "https://hammer.ai/terms"
},
"payload": {
"basis": "measured",
"basis_caveat": "A filed charge is what the hospital published under 45 CFR 180, not an adjudicated claim. No row behind these figures records that anything was paid, that a patient was billed this, or that a payer agreed to it in the period this file covers. The corpus holds no adjudications at all, so a payment rate, a denial rate and a realised price are not derivable from it by any method, including arithmetic done by hand.",
"basis_detail": {
"inputs": [
{
"basis": "measured",
"name": "filer_inventory",
"source": "hospital-filed standard charges, extract cut 2026-08-29 from the corpus at vintage 2026-08-26"
}
],
"rule": "A figure computed from inputs of more than one basis carries the weakest basis among them, strongest to weakest being measured, then cited, then assumption, so a figure is never stronger than the weakest input that produced it."
},
"code_types": {
"MS-DRG": {
"code_count": 19,
"codes": [
"056",
"057",
"876",
"880",
"881",
"882",
"883",
"884",
"885",
"886",
"887",
"894",
"895",
"896",
"897",
"917",
"918",
"947",
"948"
],
"corpus_codes": 19,
"corpus_rows": 3503,
"counterparty_cells": 32,
"deidentified_cells": 38,
"filed_as": "MS-DRG",
"severity_required": false
}
},
"corpus_vintage": "2026-08-26",
"counterparties": [
{
"cells": 13,
"code_count": 13,
"filed_payer_text": "BCBS",
"methodologies": [
"per diem"
]
},
{
"cells": 19,
"code_count": 19,
"filed_payer_text": "WellCare",
"methodologies": [
"fee schedule"
]
}
],
"counterparty_note": "These are the counterparty strings this filer wrote for itself, reproduced exactly as filed: not case folded, not trimmed, not mapped and not deduplicated across spellings. They are not the payer slugs the rate tools use, and the same company can appear here under several spellings while a single spelling can name a third-party administrator renting a network, or a government programme, rather than an insurer. CANNOT BE CONCLUDED: that any string here names the same contracting party as any slug in the rate extract, or as any other string here, by any method, including reading the two names and judging them alike. A crosswalk is an assumption somebody has to own and declare; this world has not licensed one.",
"counterparty_vocabulary_note": "Every string in the counterparty vocabulary is a payer name exactly as one hospital filed it, and none of them is a payer slug. This world does not resolve them, map them or fold their case, and a caller who does is inventing a payer. At least three kinds of party appear in this column and this world classifies none of them: an insurer, a third-party administrator on a network it rents, and a government programme. Measured on this corpus, `Medicare` is a filed payer name in its own right and `Multiplan` is a rental network rather than an insurer. has_a_hyphen_separator and ends_with_a_filer_payer_id are properties of the string that a reader can check against the string, not a judgement about the party: the hyphen form is how the third-party administrators appear, and a trailing bracketed number is the filing hospital's own internal payer id. The vocabulary is a SELECT DISTINCT over the table rather than a list frozen beside it, so a string the corpus gains is a string this world serves rather than one it refuses.",
"edition_note": "A code absent from one filer and present at another may be a code that filer's MS-DRG edition does not define, rather than an admission it does not price. Measured on this corpus: nyu-langone carries the restructured fusion codes 426 through 430 and 447, 448, 450 and 451, and carries no 453 to 455 and no 459 or 460 at all, while northwell files 453 and 460. Neither schema_version nor file_vintage names the grouper edition, so this file cannot settle which edition a filer used and this world does not settle it either. CANNOT BE CONCLUDED: that a filer absent from a code does not perform or does not price that admission.",
"facilities": [
{
"filed_facility_name": "MEDICAL CITY MH FRISCO",
"filer": "hca-medical-city",
"rows": 41402,
"source_files": 1,
"state": "TX"
}
],
"facility_identifier_note": "A facility here is the tuple a filer supplied, and no single column of it is present everywhere. npi is a National Provider Identifier, served as a string because it is an identifier: no arithmetic on one means anything and a leading zero would be lost the moment it became a number. filed_state_facility_id is the column the corpus calls `ein`, renamed here because it does not hold an employer identification number: its values are New York state facility permit numbers such as 7002053H, Connecticut facility codes such as CT0053, and one filer's own zero-padded internal ids such as 0000000027, and an EIN is nine digits. The two are complements rather than duplicates. cleveland-clinic files an npi on every row and a state facility id on none; vanderbilthealth-com does the exact reverse; northwell files an npi on 2.1% of its rows and a state facility id on all of them, naming eighteen hospitals where the npi column names six. So a facility list keyed on npi alone would report six hospitals for a filer that names eighteen and none for a filer that names three. filed_facility_name is the only one of the three every filer populates and it is served exactly as filed, trailing spaces and all, so one hospital may appear under more than one spelling. hca-medical-city files one name and neither identifier.",
"facility_note": "A filer here is a health system, and these are the facilities inside this one, identified by the tuple exactly as filed rather than by any single column. No one column covers the corpus: some filers file no npi at all, one files it on a small fraction of its rows while naming every hospital in the column the corpus calls `ein`, and one does the exact reverse. An identifier a filer left empty is an ABSENT KEY here rather than a null, and identifier_coverage says, per filer, how many rows carry each identifier, so an absent key can be read as the column this filer does not file rather than as a fact about the hospital. CANNOT BE CONCLUDED: that a facility absent from this list files no standard charge, performs no admission or has no price; and CANNOT BE CONCLUDED: any ordering of these facilities by anything, since they carry no figure here at all.",
"filer": "hca-medical-city",
"groupers_filed_and_not_answered": [],
"identifier_coverage": {
"facility_names": 1,
"npis": 0,
"rows": 41402,
"rows_with_npi": 0,
"rows_with_state_facility_id": 0,
"state_facility_ids": 0
},
"inventory_note": "This is what one filer carries in this extract, and it is a listing rather than a measurement. code_count is the number of codes this extract can answer for this filer under that grouper, and it is the length of the list beside it rather than a figure computed apart from it: nothing here is truncated, so a code absent from a list is a code this extract holds none of for this filer. counterparty_cells and deidentified_cells count CELLS in this document, one per code and counterparty and setting and money kind, not admissions, not patients, not claims and not payments. corpus_rows and corpus_codes are a different quantity again: they count the filer's rows in the corpus this extract was cut from, before the parse and the quality filter this document declares, so corpus_codes is at least code_count and the gap between them is what this extract could not carry. CANNOT BE CONCLUDED: from any count here, that a filer performs, does not perform, prices or does not price any admission.",
"methodologies": [
{
"cells": 19,
"code_count": 19,
"methodology": "fee schedule"
},
{
"cells": 13,
"code_count": 13,
"methodology": "per diem"
}
],
"methodology_distribution_note": "This counts CELLS carrying each methodology at this filer, so it says what shape this filer's filings take rather than what any of them says. A filer whose cells are almost all one methodology and a filer split across several are different contracting pictures, and that is the question this list answers. It weights nothing, and a cell is not an admission: cells sit under codes of very different volumes and this document holds no volume to weight them by.",
"methodology_note": "methodology is how the filer said the figure was arrived at, and it decides what the figure is. A dollar filed under `percent of total billed charges` is a number the hospital computed from its own gross charge rather than a price observed anywhere, and 268,559,804 rows of this corpus carry that methodology. A dollar under `fee schedule` and a dollar under `case rate` are different objects again: a case rate is one payment for a whole admission, so no per-day or per-code figure follows from one, and a per-diem row is priced by the day while this corpus holds no length of stay. Cells are never pooled across methodologies here, and pooling them elsewhere would be one number over several objects.",
"no_figure_note": "This answer carries no dollar figure at any depth, by design rather than by omission. An inventory is where a caller finds out what exists; filed_charge is where a figure lives, with the methodology, the two row counts, the reporting floor and the refusals that decide what the figure means. Take a code named here, with its grouper and, where the grouper has one, a severity subclass, and call filed_charge with it. CANNOT BE CONCLUDED: any price, bracket, median or spread from anything on this payload, by any method, including counting what is listed here.",
"parse_note": "Where a filer wrote the grouper into the code itself, this extract reads it rather than guessing at it, and every cell carries filed_code, the string exactly as filed, so the parse can be checked from the answer. THE TWO GROUPERS DO NOT HAVE THE SAME EVIDENCE. For MS-DRG the description is a genuine second witness: it names the grouper and the number, so it can contradict the prefix, and measured over 1,158,471 rows it never does. For APR-DRG the description CANNOT confirm the severity: it names the base's condition and the four subclasses of one base share it word for word, and the words severity, subclass and SOI appear in none of the 2,142,812 rows. The severity's evidence is the digit's range, 1 to 4 and nothing else at any length or filer, together with each stem carrying exactly four siblings that share one description; a code failing that test is served to nobody. THAT DESCRIPTION COUNT IS TAKEN OVER THE TEXT WITH A LEADING COPY OF THE ROW'S OWN FILED CODE REMOVED, and it has to be: nyu-langone writes the severity-bearing code into the front of the description, so APR001-1 LIVER TRANSPLANT &/OR INTESTINAL TRANSPLANT and APR001-4 LIVER TRANSPLANT &/OR INTESTINAL TRANSPLANT are two strings for one admission. Read raw, that count was 4 on 318 stems of 318 at that filer and 644,844 rows reached the test and none passed it. The removal is anchored to the row's OWN code: a description beginning with some other code, or with a digit that is part of the procedure's name, is left exactly as filed. THE THRESHOLD IS STILL ONE AND NOT FOUR, and four stems say why. hss stems 02, 30, 31 and 32, 352 rows, carry two, seven, four and three siblings whose descriptions stay distinct after the removal: stem 31 is four siblings and four different procedures, FOOT & TOE PROCEDURES,MINOR against TENDON, MUSCLE & OTHER SOFT TISSUE PROCEDURES,MINOR, which is one number over two admissions rather than one admission at two severities. A threshold of four or fewer admits those; a normalisation refuses them, and they stay served to nobody. Measured 2026-08-29 by probe_apr_drg_gate.py, Cloud Run execution apr-drg-gate-mp6tv, at corpus vintage 2026-08-26. The witnesses table reports both halves separately, and apr_severity_confirmed_by_description is zero on purpose.",
"population_note": "This corpus holds seven filers and each is a health system rather than a hospital, spanning up to eighteen employer identification numbers and two states. They are not a sample of anything: hss alone is roughly half the corpus and the smallest filer is a ten-thousandth of it, so any figure taken across them would be a figure about hss wearing a wider name. This tool serves each filer's own filing and computes nothing across them, at any n.",
"quality_filter": "code_type IN ('MS-DRG', 'APR-DRG', 'DRG'), read as the grouper the code itself spells where the label is the bare DRG that holds two of them; codes_on_item = 1; standard_charge is not null; rows deduplicated on the full natural key before counting, because 22.1% of MS-DRG rows are exact duplicates and 72.9% belong to a cell arriving under more than one source_file_hash. The facility grain is deduplicated on that same key WITH the facility tuple added, which is a different key leaving a different count: see facility_grain_note",
"receipt": "r886",
"severity_note": "An APR-DRG code is a base and a severity subclass, and the subclass is a price rather than a label. Measured on this corpus: one description is shared across all four subclasses of a base once a leading copy of the row's own filed code is removed from it, so the text cannot tell them apart. Removing that code is not a detail: nyu-langone writes the severity-bearing code into the front of the description, so read raw its four subclasses carry four distinct strings for one admission, and the words severity, subclass and SOI still appear in none of them. See parse_note. The four carry four different medians, running 20,194 to 86,167 on base 023 and 66,513 to 182,684 on base 001. So a figure for a base without a subclass would be one of four answers spanning more than a factor of four, chosen on the caller's behalf. Subclasses are never pooled here and no figure spans them.",
"vintage": "2026-08-29",
"world_version": "hammer-coverage-intelligence:2026-10-01+2"
},
"receipt": "r886",
"session": "49ec6090f32b112cd7fc2d39a72e0974",
"tool": "filer_inventory",
"world": {
"build": 2,
"description": "What payers publish that they pay, code by code, and what may be said with it.",
"name": "hammer-coverage-intelligence",
"vintage": "2026-10-01"
}
}Permanent saved result
Read this evidence receipt
claim_check
judgement_check