Skip to content
← All research

Healthcare economics / Research prototype

A reimbursement calculation you can replay.

Versioned payer assumptions, integer arithmetic and explicit prerequisites make a small payment model inspectable.

A diagnosis code is one input

An ICD-10-CM code alone cannot determine what a provider will be paid. A reimbursement workflow also depends on the service, payer, coverage, setting, applicable dates and other claim facts. A transparent model should expose those dependencies before showing an amount.

This repository formalizes a small synthetic payer scenario. It shows how a rule revision and its prerequisites can control a calculation, and how a missing prerequisite can prevent that calculation from being presented as an applicable result.

Make arithmetic and applicability separate

Integer cents keep the fixture’s money arithmetic exact. The encoded rule states how inputs produce an example amount. Applicability checks establish whether the calculation should run for this synthetic claim at all. Dates and coverage assumptions belong in that check, not in a footnote after the answer.

The fixture uses invented amounts and simplified rules. Its purpose is to test the structure of an inspectable adjudication process. Real payer rules require sourced contracts, service definitions, modifiers and a much larger set of exceptions than this first model contains.

Useful failure cases

A claim with missing coverage evidence should not receive an authoritative-looking payment number. The same applies when an authorization prerequisite is unresolved or the policy revision does not cover the claim date. These cases are part of the executable examples.

The proofs concern the implemented gates and calculation. They do not certify a payer’s acceptance of a claim. A reproducible estimate under explicit assumptions and an actual remittance are different observations that should have distinct records.

Evaluation before scale

Synthetic cases test arithmetic and boundary behavior. A future study should replay de-identified claims against versioned source rules and compare modeled results with adjudication outcomes. Differences need classification: missing source facts, policy gaps, timing, rule interpretation or arithmetic.

The same frozen claim and rule snapshot should reproduce the same output. That reproducibility helps an analyst investigate a disagreement. It does not eliminate disagreements about the facts or which rule applies.

Where this connects

The clinical-coding repository asks whether a candidate has the required documentation. This repository asks whether a small payment rule applies and how its amount follows. They can share provenance records, while retaining separate acceptance conditions and separate evidence about their performance.