BRENNEN

Industries / Health / Healthcare / Pharmacy

Insurance claims

Insurance claims are how a pharmacy gets paid for each prescription by a health plan, a benefit manager or the state: the pharmacy sends a claim, and the payer checks and prices it, usually within seconds, while the patient waits.

7

Milestones achieved, from 1948 to 2024.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1948

    A national list of medicines

    Australia's Pharmaceutical Benefits Scheme (PBS) began in 1948 as a limited scheme, with free medicines for pensioners and a list of 139 "life-saving and disease preventing" medicines free of charge for everyone else. A published list of covered medicines, called a formulary, is how drug plans everywhere decide what they pay for.

    pbs.gov.au

    Credit

    Australian Government

  2. 1988

    Claims answered in seconds

    NCPDP (National Council for Prescription Drug Programs), the pharmacy standards body, was formed in 1977 to standardize insurance forms; its paper Universal Claim Form, adopted in 1978, replaced many plans' own forms. In 1988 it released version 1.0 of its Telecommunication Standard for online, real-time claims: the pharmacy sends the claim and learns within seconds whether the patient is covered, what the plan pays and what the patient owes. Version 5.1 was named in the HIPAA regulations as the US format.

    ehcca.comncpdp.org

    Credit

    NCPDP

  3. 2016

    Paid what the pharmacy paid

    US Medicaid rules had priced a drug's ingredient cost at an "estimated acquisition cost". A CMS rule effective 1 April 2016 replaced it with "actual acquisition cost", the prices pharmacies actually pay, and required states to pay an adequate professional dispensing fee on top, with state plans due by 2017. To measure those prices, a CMS contractor surveys a random sample of independent and chain community pharmacies in all 50 states and Washington, DC, every month; the result is published weekly as the National Average Drug Acquisition Cost (NADAC), with files from November 2013.

    federalregister.govdata.medicaid.gov

    Credit

    Centers for Medicare & Medicaid Services (CMS)

  4. 2017

    Generic prices jump in England

    During 2017 the prices pharmacies paid for certain generic medicines supplied to the NHS rose unexpectedly, some more than tenfold. The National Audit Office estimated net NHS spending on the affected medicines at £315 million in 2017-18, seven times the year before; the NHS spent £4.3 billion on all generic medicines in 2016-17.

    nao.org.uk

    Credit

    National Audit Office

  5. 2018

    Ohio measures the spread

    In spread pricing, a pharmacy benefit manager (PBM), the company a plan hires to run its drug benefit, charges the plan more for a prescription than it pays the pharmacy and keeps the difference. On 16 August 2018 Ohio's Auditor of State, Dave Yost, reported that PBMs in Ohio Medicaid's managed care took a spread of nearly 9% across all drugs and 31% on generics from April 2017 to March 2018: about $225 million of $2.5 billion, $208 million of it on generics. Ohio Medicaid told its five managed care plans to end spread-pricing contracts by 1 January 2019 and move to pass-through pricing, a flat fee plus what the pharmacy is actually paid. CVS Health, one of the PBMs, disputed the findings.

    fiercehealthcare.com

    Credit

    Ohio Auditor of State

  6. 2024

    Fees moved to the counter

    Many Part D plans, the private plans that run Medicare's drug benefit, paid pharmacies less after the sale if they missed set criteria, so the price reported at the counter, on which patients' cost-sharing is based, was often higher than what the pharmacy was finally paid. A CMS rule of April 2022 redefined the negotiated price as the lowest possible payment to the pharmacy and required plans to apply all such pharmacy price concessions at the point of sale from 1 January 2024.

    cms.gov

    Credit

    CMS

  7. 2024

    One hack stops claims

    On 21 February 2024 a ransomware group called ALPHV, or BlackCat, attacked Change Healthcare, a UnitedHealth Group company that Military Times called the nation's largest commercial prescription processor. Change disconnected its systems, and pharmacies that relied on it could not process claims. UnitedHealth's chief executive told the US Senate Finance Committee on 1 May 2024 that 99% of the pharmacies it had served could process claims again by 7 March. Military pharmacies filled prescriptions by hand and returned to full operations about six weeks after the attack.

    Lessonone processor's outage reached thousands of pharmacies at once.

    finance.senate.govmilitarytimes.com

    Credit

    Military Times; UnitedHealth Group testimony

Watch for

  1. PBM pay cut from prices

    The Consolidated Appropriations Act, 2026, signed on 3 February 2026, bars PBMs serving Medicare Part D plans from earning income tied to drug prices or rebates and limits them to a flat "bona fide service fee" from 1 January 2028. It also requires PBMs to pass all rebates through to employer health plans.

    Achieved whenCMS's rules to carry it out are final and Part D contracts for 2028 pay PBMs flat fees, as shown in the new PBM reports or an outside audit.

    kff.orgsidley.com

    Claimed by

    US Congress, in a law signed February 2026

Open

  1. 01

    A days' supply answer key

    A public set of 500 made-up prescriptions for products whose days' supply is hard to work out (insulin pens, inhalers, eye drops, nasal sprays, creams), each with the correct days' supply under stated rules and the source of every factor used. Pharmacy software, calculators and AI models can then be scored on the number that goes on the claim.

    The brief
    Credit · unclaimed
  2. 02

    A test set for rejects

    A public set of 300 made-up claims, each with the plan's coded reply rejecting it for an administrative reason (patient not covered, prior authorization required, refill too soon, quantity over the plan limit, drug not on the formulary) and the correct next step for the pharmacy team. AI assistants that read rejects can then be scored before anyone trusts them with real claims. Rejects from drug-interaction and other clinical checks are left out.

    The brief
    Credit · unclaimed

Open 01 · the brief

A days' supply answer key

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public set of 500 made-up prescriptions for products whose days' supply is hard to work out (insulin pens, inhalers, eye drops, nasal sprays, creams), each with the correct days' supply under stated rules and the source of every factor used. Pharmacy software, calculators and AI models can then be scored on the number that goes on the claim.

Why it is still open

When a PBM audits a pharmacy it can take back payment, a recoupment, for claims it finds wrong. PAAS National, an audit-defense service, says overbilled quantities and incorrect days' supply "account for a sizable portion of audit chargebacks", and keeps its days' supply charts and calculator for paying members. It also notes that there is no industry standard for drops per mL, so PBMs use their own conversion factors. Recorded values are often wrong: in Québec, the days' supply in claims matched the original prescription for 39.6% of inhaled steroid prescriptions for children and 56% for people aged 12 to 64; in Estonian records it was given correctly for 3.1% of insulin injection prescriptions.

paasnational.compaasnational.compmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.gov

Built with what exists

package sizes and units from the FDA's NDC Directory through the openFDA API; doses per device from DailyMed labels; the conversion rules in PBMs' public provider manuals; Python for the scoring script; AI models such as Claude, Gemini and GPT as test subjects.

  1. First step

    pull 50 insulin pen and inhaler package records from openFDA and write the first 100 prescriptions with worked answers.

  2. Done looks like

    500 items whose answers two reviewers, one of them a licensed pharmacist, set independently and agree on for at least 95% before resolving the rest, plus published exact-match scores for at least three tools.

Open 02 · the brief

A test set for rejects

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public set of 300 made-up claims, each with the plan's coded reply rejecting it for an administrative reason (patient not covered, prior authorization required, refill too soon, quantity over the plan limit, drug not on the formulary) and the correct next step for the pharmacy team. AI assistants that read rejects can then be scored before anyone trusts them with real claims. Rejects from drug-interaction and other clinical checks are left out.

Why it is still open

Rejects are common and change what patients get. Among 297,312 US patients newly prescribed an oral anticoagulant for a blood clot, 9.1% had the first claim rejected, and of those who later filled a prescription, 14.2% first filled a different anticoagulant. Answers exist only as guidance for single programs: in June 2026 CMS published a troubleshooting table for its Medicare GLP-1 Bridge that pairs each reject code with the message the pharmacy sees and how to resolve it. Searches of Europe PMC, arXiv and Hugging Face in November 2026 found no public test set of pharmacy claim rejects with the right responses.

pmc.ncbi.nlm.nih.govcms.gov

Built with what exists

the reject codes and payer sheets that state Medicaid programs and PBMs publish in their pharmacy provider manuals, and CMS's GLP-1 Bridge table as a model answer key; field definitions from NCPDP's Telecommunication Standard, available to NCPDP members; Python to generate the synthetic claims; AI models to test, including services that answer fixed multiple-choice questions with a probability (see the Note "A thousand questions, one pass").

  1. First step

    collect the reject codes and their stated meanings from three state Medicaid pharmacy provider manuals and write the first 50 cases.

  2. Done looks like

    300 cases across at least 12 reject codes, an answer key checked by two certified pharmacy technicians, and published accuracy for at least three AI models, overall and for each reject code.