Native analyses

Descriptions below are excerpts. Open an analysis for its full method, exact input contract and execution links.

• rate_position

  What payers publish that they pay for a code, positioned against the public fee schedule that anchors it. A what-does-X-pay question is this tool, not a search. Codes: the `code` enum on this tool is the closed vocabulary and the only place it is written; it holds every billing code this vintage's extract carries and nothing else. Read it rather than guessing a plausible-looking code, and note that a bare `0689` is not a code while `0689T` is: t

• rate_comparison

  Two to four codes' payer rates set side by side under one receipt. THIS TOOL COMPARES NOTHING: it serves both figures under one receipt so that claim_check can decide a claim about them. Two separate rate_position calls cannot be checked that way, because a claim is decided against one receipt and neither of those two covers both sides. No ratio, no difference, no percentage and no comparability flag is served, deliberately. A served ratio would

• stack_check

  Whether two or more codes may be billed together, per the NCCI PTP edit file. A can-we-bill-X-with-Y question is this tool, not a search. Codes: 91200, 0689T, 0690T, 76981, 76982, 76983, 76705, 0002M, 0003M, 81517, 0014M, 80076. `0689` is not a code, `0689T` is. A pair with no edit row on file is UNKNOWN, never permission: Category III codes routinely carry no row, and 0689T against 91200 is exactly that case in the shipped file. A lapsed edit,

• screening_yield

  How many people a screening programme would find positive and negative, true and false, and what its positive and negative predictive value would be, given a declared test sensitivity, specificity, prevalence and population size. A what-would-this-test-find or what-is-its-predictive-value question is this tool, not a search; this is straight Bayes over exactly the numbers supplied, and this world ships no default for any of the four inputs. Voca

• deferred_versus_avoided

  Whether a set of declared cost events are truly eliminated by an intervention or merely moved to a different time, reported as two separate totals that are never added together. A does-this-savings-model-actually-save or is-this-cost-avoided-or-just-deferred question is this tool, not arithmetic you should do yourself: an intervention that changes whether a cost is incurred at all is different from one that changes only when it is incurred, and s

• interpretation_requirement

  Who a payer's own policy says may FURNISH and INTERPRET a code. Use this for "can a primary care doctor bill this", "does this need a radiologist", or "who has to read the study". This is not a coverage question: a code whose interpretation is unrestricted may still be refused, and this tool says nothing about whether anyone pays. Codes: 91200, 0689T, 0690T, 76981, 76982, 76983, 76705, 0002M, 0003M, 81517, 0014M, 80076. `0689` is not a code, `06

• cohort_figures

  How many people, and by what means somebody counted. Enrollment counts, modelled prevalence rates and member months, each served separately with its measurement class, its case definition and its source attached. Use this for "how many Medicare beneficiaries are in this county", "what is the modelled diabetes prevalence here" or "how many member months did this state's Medicaid programme run". THIS TOOL MULTIPLIES NOTHING, and that is the design

• comorbidity_edges

  MOVED. The MASLD/NAFLD edges are served by `prognostic_association`, with the same figures, the same one-record-per-answer rule and the same refusal against multiplying a stratified estimate by an unstratified one. Call that tool with `exposure` as "masld_nafld" plus the `outcome` and `stratum` you would have passed here. THIS NAME ANSWERS NOTHING. It is kept so a caller holding it is told where the corpus went rather than getting an error that

• fee_components

  What the Medicare Physician Fee Schedule states for a code: whether the code splits into a professional component and a technical component, and the relative values it states for each. Which part of this code is the reading and which part is the equipment is this tool. Codes: 91200, 0689T, 0690T, 76981, 76982, 76983, 76705, 0002M, 0003M, 81517, 0014M, 80076. `0689` is not a code, `0689T` is. Three rows, three different things. The row with no m

• filed_charge

  What a hospital filed as its own standard charge for an inpatient admission, under 45 CFR 180. What-does-an-admission-cost is this tool. What a payer pays for a procedure is rate_position, and the two are separate tools on purpose. Groupers: MS-DRG and APR-DRG. code_type is required, because five DRG groupers sit in one column of this corpus and a code number means a different admission under each: one filer files 453 as a spinal fusion under MS

• hypertension_evidence

  MOVED, and it split in two, because it held two kinds of number. The OBSERVATIONAL records, a pressure predicting an outcome in people nobody randomised, are served by `prognostic_association`: call it with `exposure` as "clinic_bp" or "nighttime_ambulatory_bp" plus the `outcome` you would have passed here, in place of `measurement`. The RANDOMISED records are served by `treatment_effect`, with the same `arm` and `outcome` this tool took. THIS N

• prognostic_association

  One published observational association per answer: this exposure predicted that outcome, in people nobody randomised. Carries its own interval, the population it was measured in, its PMID and the authors' own sentence. NAME THE `exposure`. This world holds more than one observational corpus and ships no default, because a lookup naming no exposure would answer about nothing in particular. `masld_nafld` is the fatty-liver review and takes a `str

• treatment_effect

  One published RANDOMISED effect per answer: being assigned this arm changed that outcome's rate in a trial. Carries its own interval, the trial population, its PMID and the authors' own sentence. SEPARATE FROM `prognostic_association` ON PURPOSE. A hazard ratio for a pressure predicting an outcome is not the effect of lowering that pressure: the two are measured against different comparators, adjustment is not randomisation, and no arithmetic ov

• regression_to_mean_expectation

  What Galton's arithmetic predicts a cohort selected for being extreme will measure next period, before anybody intervened. Route here whenever an operator penalised, flagged or targeted the worst performers on a noisy measure, measured them again, and found them improved: a cohort selected for sitting at the extreme of a noisy measure sits closer to the mean at the next measurement whether or not anything was done to it, and this says by how much

• competing_event_coupling

  How much of a readmission rate's movement between two periods is the survivor denominator moving underneath it. A patient who dies in hospital cannot be readmitted, so death is a competing event: it removes people from the denominator a readmission rate is measured over rather than contributing a zero to it. The crude rate an operator reports therefore moves when mortality moves, with nothing about readmission having changed at all, and the direc

• withhold_earnback

  What a measure result is worth under a capitation withhold. A state holds back a share of capitation from a managed care organisation and releases it on measure performance; declare the capitation, the withhold share and the menu of measures with their earn-back shares and their results, and this says what came back and what did not. Route here whenever somebody is asking what passing or missing a measure cost them under a withhold arrangement.

• tournament_position

  Where a declared score sits against declared cut points, and how far the FIELD would have to move to cross each one with that score held still. Route here whenever a rating, a star, a band or a tier is being read as a standard. It is not one. Cut points set from the year's own distribution move with the distribution, so a plan can improve absolutely and lose a star because the field improved more, and that sentence is what this tool makes computa

• measure_reliability

  How much of a measured rate is the provider and how much is the sample it was measured on. Route here whenever a rate is about to be compared, ranked, penalised or rewarded, because a rate computed on a small denominator moves on its own and a comparison that does not know by how much is reading noise as performance. Declare `numerator` and `denominator`, the counts the rate was computed from, and `between_provider_variance`, the variance of the

• declared_inputs

  Bring a customer's own aggregate figures into this world under a declared basis, and get back a handle the calculators read. Route here when a figure a question needs belongs to the caller rather than to this world: their book's population, their prevalence, their cost per affected member per month, their procedure volume. Nothing that enters here becomes a measurement. This world reserves `measured` for a figure it read out of its own register,

• cost_of_care_exposure

  Size an exposure over a declared population: what a condition accounts for across a book, over a declared period, on figures the caller declares. Route here when the question is how big something is on a plan's book, and read the last paragraph before quoting anything downstream of the answer. Declare EITHER an `affected_members` count OR a `population` with a `prevalence`, never both routes at once, together with `monthly_cost_per_affected_memb

• savings_methodology

  Classify a declared savings mechanism as hard or soft against the buyer's own definitions, and say what evidence a case built on it is held to. Route here when the question is what kind of saving is being claimed: the kind that shows up in adjudicated paid claims, or the kind that exists in a model. Declare `mechanism`, one of unit_cost_reduction, utilisation_reduction, demand_redirection, avoidance_of_future_cost, overpayment_recovery and adher

• hospital_variance

  How far apart providers actually are on one published quality measure, read off a public measure corpus rather than declared by whoever needed the number. measure_reliability requires between_provider_variance and ships no default, because the defaulted input is the one that decides the answer. This tool reads that input instead, for one named measure, out of a corpus this world holds. measure_id is required and there is no default. A variance

• filer_inventory

  What one hospital filer carries, so a call that prices something is a call that answers. One filer in, one answer out, and no figure anywhere in it. The loop this replaces is guessing a code, being refused, and reading the edition note in the refusal. That loop works and it costs a call per guess, and it teaches nothing about the codes nobody thought to guess. Back from this tool come the groupers that filer files, the codes this build can answe

• candidate_screen

  The screen that runs before the pitch rather than after it: the caller's own declared candidate programmes, put through an ordered routing of steps, under one receipt. The steps are ordered and the order is not a preference, because each one decides whether the next is worth running at all. At your volume, is the thing you want to compare measurable. Does a published effect for this mechanism exist, and does its interval cross no effect. Does th

• frontier

  Where each part of a goal would be settled, before you spend a call on any of it. The only tool here that takes a goal rather than a claim or a call. Everything else in this world answers a question you already know how to ask. A goal is not that: it is a thing you want to be true, usually unfalsifiable as written, and a verdict on it is not the useful answer. Put a goal to the checkers and they are each right and their sum is a dead end, becaus

• judgement_check

  Check a claim you intend to write against the answer you were given, and receive the reading of that answer. Call this before stating any conclusion drawn from this world. Pass the receipt from the answer and one entry per claim. State the sentence you intend to write, and separately the quantities and the entities that sentence rests on, so the check is a comparison against the payload rather than a reading of your prose. Declare the shape of ea

• claim_check

  Check a claim by entailment over the figures in the answer you were given. This does not read your sentence: the verdict comes from the arithmetic, so it is the same claim in any language. Pass the receipt and one or more claims in the claim grammar, naming a figure by its path in the payload, for example {"op": "<=", "left": "report.level_noise_sd", "right": 12, "scope": "marginal"}. An element of an array is named either cells[0].median or cell

• deliver

  Hand over the report you are about to deliver, exactly as you mean to deliver it, and receive it back unchanged. Call this last, after every tool call and every check, immediately before you write your answer out. This blocks nothing and rewrites nothing. The text comes back byte for byte, no part of it is read, and no verdict is passed on it: nothing here can stop you delivering whatever you have written. What it records is which of the answer

• session_trace

  Read back what this session has asked this world: every receipt it issued, in the order they were issued, with the tool that produced each one, whether the world answered or refused, and whether you ever submitted a claim against it. Call this before you write your answer, not after. Its purpose is the receipts you took and never checked: an answer taken and never checked is the one thing a check on submitted claims cannot see, because there is

• progress_check

  Declare the items your deliverable needs and read back, off this session's ledger, which of them the world has already answered, which it has refused, and which nothing has reached yet. Each declared item names the tools that would serve it and, optionally, strings its arguments must contain, such as a code and a payer. An item is established when an answered receipt from one of its tools carries every one of those strings as a whole value among

• model_advisory

  Ask the trained models about a sentence before you write it down. This is ADVICE, not a verdict: `claim_check` decides by arithmetic over the figures a world declares, and nothing here changes what it decides. Three models answer. `assessor_gate` flags a sentence naming a construct the world says it cannot answer — the case arithmetic cannot catch, because the sums can be perfect while the construct is refused. `flip_discriminator` judges a time

• knowledge_navigate

  Navigate this world's indexed knowledge. Start from exactly one anchor and then take typed steps through the graph. Everything comes back keyed by canonical node id, so a semantic hit, a table row and a graph node are the same identities and need no join. THE ANCHOR IS EXACTLY ONE OF THREE. `query`, a text search: a query naming a billing code is answered by the code index, keyed on the procedure and MS-DRG nodes of the graph, and comes back anc

• simulate

  Run one of this world's own tools many seeded times, with a declared distribution over each input you want to vary, and receive the share of draws the world answered, the share it refused BROKEN DOWN BY THE KIND OF REFUSAL, and the distribution of every figure the answered draws produced. Use it when your inputs are ranges rather than points: an assumed prevalence, a test's sensitivity, a population you are unsure of. Name the world tool in `tool

• sensitivity

  Which of your declared inputs moves the outcome, ranked over a cohort you declare exactly as you declare it to `simulate`, by two measures that fail differently. `first_order_index` is the share of the response's variance tracked by that input, over ten equal-count bins, computed from the replay with no further estimator calls. `p10_to_p90_swing` pins one input at its own p10 across EVERY draw and again at its p90 and reports the difference, whic

• interval_check

  Whether a claimed value falls inside the interval a simulation produced, for a cohort you declare exactly as you declare it to `simulate`. Verdicts: `inside`, `outside`, `unresolved` when the interval spans the quantity's whole admissible range so any claim would have passed, `out_of_domain` when the claim is not a value the quantity can take, `too_few_answered` below thirty answered draws. There is no tolerance parameter: the interval is the ver

• decision_navigate

  From a refused state to the cheapest DECLARED move set that makes the estimator answer, or the other way for fragility. A move is a name, an input, a delta and a cost, and all four are yours: this tool cannot decide that an input is movable, by how much, or at what price, and the cost is in your own units and is never money. Every candidate is re-run through the estimator, so a path reaches the goal only when the world says so and never because t

• network_infer

  A posterior over one node of a Bayesian network YOU declare, conditioned on evidence, by exact enumeration rather than sampling so it can be replayed. An edge licenses a conditional probability table and says NOTHING about causation: `causal`, `intervention`, `do`, `effect_of`, `counterfactual` and `treatment_effect_of` are refused by name rather than answered with a caveat. A table cell is a bare number, a number with a source, or a `from` refer

• hello_world

  CALL THIS FIRST, BEFORE ANY OTHER TOOL ON THIS SERVER. It is not a demo and not a health check. One call hands you everything you need to use this world correctly: who it is and what a receipt is redeemed against, which question goes to which tool, what it refuses and why, which knowledge anchors actually resolve in this build, and the claim discipline every answer from here has to close with. Every part of it is read from this world's own decla

• say_world

  How to say what this world told you, to one named kind of reader. Call it before you write the answer up, once you know who is going to read it. WHAT THE AUDIENCE CHANGES: vocabulary, emphasis and ordering, and which of this world's own refusals to put in front of this reader. WHAT IT NEVER CHANGES: the claim, the refusals, and the claim report. Every audience carries every one of this world's declared refusals, in the world's own words and in t