What shared fixes could save

Every organisation that moves quality reporting to FHIR meets the same data problems: a code with the wrong label, a result stored as free text, a search that silently returns half the data. Today each one discovers and fixes them on its own. If every ACO, registry and vendor solves the same eClinicalWorks problem separately, the country pays for that fix hundreds of times.

OpenDQM’s idea is simple: solve a problem once, for each EHR product, and share the fix. This model estimates what that is worth. Move the sliders to test your own assumptions.

Who pulls and calculates the data?

Each ACO Registries and vendors One central platform

Without shared fixes–

With OpenDQM–

Net avoided cost–

Problem investigations

AI tokens used

FHIR requests per year

Practices connected

Scope

ACOs

Participating practices (TINs)

Practices on an EHR with FHIR access

Distinct EHR products in use

EHR products per ACO (average)

Quality measures reported

Patients in scope (millions)

Years modelled

Data problems

EHR-wide problems per EHR, per measure

Practice-specific problems per practice, per measure

Hours to solve an EHR-wide problem

Hours to solve a practice-specific problem

Hours to connect one practice

Blended hourly cost (engineering, informatics, QA)

Rework each later year (EHR and measure updates)

AI and data pulls

AI input tokens per problem (millions)

AI output tokens per problem (millions)

Price per million input tokens

Price per million output tokens

FHIR requests per patient, per run

Data pulls per year

Cost per 1,000 FHIR requests

With OpenDQM

Effort to apply a shared fix, vs solving it

Less effort on practice-specific problems

Less effort to connect a practice

Fewer wasted data pulls (paging, retries)

Cost to run the OpenDQM community, per year ($M)

Reset to defaultsCopy link to this scenario

An illustrative scenario, not a forecast

The defaults are assumptions drawn from our own integration work and from public program figures (511 ACOs, 15,353 participant practices and 12.6 million assigned patients in 2026). They have not been measured across the industry. The results are avoided costs across the healthcare system: engineering, informatics, testing and data pulls. They are not savings in program or claims spending.

How the model works

Where the cost comes from

Cost	Without shared fixes	With OpenDQM
Connecting practices	Each practice’s EHR connection is set up from scratch	Shared playbooks per EHR cut the effort
EHR-wide problems	Rediscovered by every organisation, on every EHR it uses	Solved once per EHR product, then applied by everyone else at a fraction of the effort
Practice-specific problems	Each one investigated from scratch	Shared patterns and tools make them faster to solve
Rework in later years	EHR upgrades and yearly measure updates repeat a share of the work	The same share, of a much smaller base
Data pulls	Every run pays for paging mistakes and retries	Known paging and search fixes cut wasted pulls
OpenDQM community	—	Governance, review, signing and maintenance, counted as a cost

The key difference: per EHR, not per practice

An EHR-wide problem, such as how one EHR labels a code, is the same for every practice on that EHR. Without sharing, it is solved once per organisation per EHR:

ACOs × EHR products per ACO × measures × problems per EHR and measure

With OpenDQM it is solved once per EHR product:

EHR products × measures × problems per EHR and measure

Every other organisation only has to apply and check the shared fix.

Who pulls the data

• Each ACO: every ACO builds its own connections to every EHR its practices use. Sharing saves the most here.
• Registries and vendors: roughly one platform for every twelve ACOs, each covering up to 20 EHR products. There is less duplication to remove, but still a lot.
• One central platform: all problems are solved in one place, so sharing changes little. The model shows this honestly; the open library still lets any organisation check the results.

AI tokens

Each problem investigation uses an assumed number of AI tokens for searching, testing and retrying. At today’s prices that is a few dollars a problem, compared with hundreds or thousands of dollars of people’s time. The saving does not come from cheaper AI. It comes from not running the same investigation again.

What would make the numbers stronger

• Counting distinct EHR products and connection points across ACO practices, not just practices.
• Measuring the hours actually spent on each data problem, and how often the same problem recurs across organisations.
• Measuring what share of problems can be fixed safely with a shared, reviewed rule.
• A pilot that tracks real costs, with and without the shared library, across a group of ACOs.