Compare four pricing models for autonomous medical coding and understand the cost assumptions health systems should assess before purchasing.
No vendor in this category publishes a rate card, so every buyer arrives at the same disadvantage: three quotes, three billable units, and no way to line them up. Autonomous medical coding cost is not one number. It is a fee basis multiplied by a volume you have to predict, plus the coder time the engine does not remove, plus a set of one-time and annual costs that sit outside the quote entirely.
This page does not publish rates, because the ones in circulation are unverifiable. It publishes the logic instead: the four fee structures in use, the arithmetic that converts any of them into an effective cost per chart, and the questions that make two quotes comparable.
The four pricing models
| Model | What you are billed for | What it rewards the vendor to do | Where volume risk sits | What to negotiate |
|---|---|---|---|---|
| Per chart | Each chart, either submitted to the engine or cleared without coder touch | Depends entirely on which of those two it is | With you, and it scales cleanly both ways | The definition of a billable chart, in writing |
| Subscription or platform fee | A period, usually annual, sometimes banded by bed count, encounter volume or facility count | Grow the account at renewal, not per chart | With the vendor inside the band, with you outside it | Band width, true-up mechanic, and what happens if volume falls |
| Percentage of collections | A share of what the coded claims collect | Raise collected dollars, which is not the same objective as coding correctly | With the vendor, which is why it is attractive | Whether the arrangement survives compliance review at all, before anything else |
| FTE-equivalent or managed service | Coder capacity replaced or supplied, priced as headcount | Meet a throughput commitment | Shared, usually with a floor | What happens to the commitment when your volume is seasonal |
Hybrids are common, most often a platform fee plus a per-chart rate above a threshold. A hybrid is not a fifth model, it is two of the above with two sets of terms to read. Fee structures vary considerably more than the underlying products do.
The headline rate is not the cost
The number that matters is the effective cost per chart across your whole volume, not the rate on the quote. It has four parts:
Effective cost per chart = vendor fee ÷ charts the engine clears + coder cost on the charts it routes back + amortised one-time costs + annual revalidation cost
The second term is the one that decides most comparisons, and it is governed by a rate the vendor states and you should verify independently.
A worked illustration. The inputs below are arbitrary round numbers chosen to show the arithmetic. They are not benchmarks, not survey data, and not MEDIQCODE's rates.
Assume 100,000 outpatient charts a year, a loaded coder cost of $45 an hour, and 12 minutes of coder time on each chart the engine routes back, so $9 per routed chart.
| Quote A | Quote B | |
|---|---|---|
| Fee basis | $1.50 per chart submitted | $2.75 per chart cleared |
| Share cleared without coder touch | 55% | 70% |
| Vendor cost | 100,000 × $1.50 = $150,000 | 70,000 × $2.75 = $192,500 |
| Coder cost on routed charts | 45,000 × $9 = $405,000 | 30,000 × $9 = $270,000 |
| Total | $555,000 | $462,500 |
| Effective cost per chart | $5.55 | $4.63 |
Quote B's unit price is 83% higher than Quote A's. Its effective cost is 17% lower. Two things drive that. The clear rate moves 15,000 charts out of the coder queue, and the fee basis means Quote A bills you for 45,000 charts that a coder codes anyway. You pay twice on those charts, once to the engine and once to the coder.
One refinement worth insisting on: coder time on a routed chart is review time, not coding-from-scratch time, but only if the chart arrives with suggested codes and the supporting documentation highlighted. If it arrives blank, the routed chart costs you full coding time and the $9 assumption above is wrong in the direction that hurts. Ask which it is, and ask to see it.
The third and fourth terms of the formula are where term length quietly changes the answer. Implementation, interface build and pre-go-live validation are one-time costs that spread across however many charts the contract covers, so a three-year term amortises them over three times the volume a one-year term does. That is the real reason vendors push long terms, and it is a legitimate reason rather than a trick. The trade is that a long term prices in a clear rate you have not yet verified on your own chart mix. Annual revalidation runs the other way, because it recurs every code set cycle for as long as the contract does, so it belongs in the per-chart figure rather than in a one-time implementation line.
Before you model anything, confirm what the engine actually clears on your own chart mix rather than a reference population.
Percentage of collections carries a question the others do not
Contingency pricing is structurally appealing. There is no capital outlay, the cost tracks cash rather than volume, and the vendor carries the downside if nothing collects.
It also sits in a place regulators have commented on for nearly thirty years. In its 1998 Compliance Program Guidance for Third-Party Medical Billing Companies, published in the Federal Register, the HHS Office of Inspector General recorded a longstanding concern that percentage billing arrangements may raise the risk of abusive billing practices. The guidance is not binding and percentage arrangements are not prohibited under federal law, but the concern is on the record, and a number of states restrict percentage-based arrangements under fee-splitting rules. Your counsel, not your vendor, should tell you where your arrangement lands.
The operational objection is separate from the legal one and gets less attention. A percentage fee pays the vendor more when your collections rise for reasons that have nothing to do with coding: a rate increase, a payer mix shift, a new service line, a contract renegotiation. It pays the vendor less when collections fall for reasons equally outside their control. The fee is indexed to something neither party is managing, which makes the arrangement hard to govern in either direction.
If you are considering it, treat compliance review as the first gate rather than a formality after the commercial terms are agreed.
The costs that sit outside the quote
Every model above prices the engine. These are the lines that appear later.
- Interface and integration build. Frequently billed by your EHR vendor rather than the coding vendor, on their timeline and their rate card.
- Implementation and configuration. Service line scoping, edit rules, worklist routing, threshold setting.
- Pre-go-live validation. Dual coding a stratified sample with credentialed coders to establish a baseline you can defend. This is coder time you have to find.
- Annual revalidation. Code sets change on a fixed federal calendar, so the system needs retesting after each cycle. Ask explicitly whether revalidation is included in the fee or billed as a service.
- Code set licensing. CPT requires an American Medical Association licence, and APR-DRG requires a Solventum licence where your payer mix uses it. Confirm whether the vendor's licence covers your use or whether you hold your own.
- Change orders. Adding a service line after go-live is rarely inside the original scope.
- Exit. Data export format, retention of your validation set, and whether you can run parallel during a transition.
None of these are hidden in a dishonest sense. They are simply priced by different parties at different times, which is why a side-by-side comparison built only from vendor quotes understates every option by a similar and unknown amount.
Which model fits which situation
- High, predictable volume in a narrow specialty. Per chart, billed on charts cleared. The fee tracks the benefit directly and the arithmetic above is easy to audit monthly.
- Volatile or seasonal volume. Per chart, not subscription. A subscription priced for your peak is a fixed cost through your trough.
- Flat budget with a hard annual number. Subscription, with a wide volume band and a true-up that works in both directions. Predictability has a price and this is it.
- No coding staff and no capital. Managed service, priced on capacity. Contingency only if counsel clears it first.
- Uncertain clear rate on your chart mix. Any model, but with a paid pilot on your own charts before the multi-year term. The automation tier you choose changes both the cost and the risk profile.
Five questions that make two quotes comparable
- What is a billable chart? Submitted, cleared, or billed? Get the definition in the contract, not the proposal.
- When the engine routes a chart to a coder, what do I pay, and does the chart arrive with suggested codes and highlighted documentation?
- Who pays for revalidation after each code set cycle, and what is the contractual deadline for having it done?
- What is the minimum commitment, how wide is the volume band, and does the true-up work downward as well as upward?
- On exit, what do I get back, in what format, and can I run parallel during transition?
Every one of those has a written answer or it does not exist. A vendor who answers question one with a paragraph rather than a definition has told you something about question two as well.
Autonomous medical coding cost is decided by the fee basis and the clear rate, in that order, and almost never by the headline rate that made one quote look cheaper. Model both, on your own chart mix, before the term length is agreed.
This article is educational. It is not legal advice on any specific contracting arrangement, and it is not a coding determination for any encounter. Have counsel review any percentage-based arrangement against federal and state law before signing.
Frequently asked questions
Is per-chart or subscription pricing cheaper?
Neither, structurally. Per chart is cheaper when volume is volatile or declining, and it keeps the fee attached to the benefit. Subscription is cheaper when volume is high, stable and growing inside the band, and it buys budget predictability. Work the effective cost per chart under both and compare those, not the rates.
Is percentage-of-collections pricing allowed for coding services?
It is not prohibited under federal law, but the HHS Office of Inspector General has recorded a longstanding concern about percentage billing arrangements, and several states restrict them under fee-splitting rules. This is a question for counsel in your state, not for a vendor's sales team or for this page.
What should we budget for implementation?
There is no credible published benchmark, and anyone quoting one is quoting a vendor. Budget it as a list rather than a number: interface build, configuration, pre-go-live dual coding, and coder training time. Timeline drives most of it, and timelines are covered in how long autonomous coding implementation takes.
Does a higher clear rate always lower our cost?
Only if the fee is billed on charts cleared and review time on the remainder does not rise. A higher clear rate can concentrate the harder charts in the coder queue, raising minutes per routed chart even as the count falls. For what to expect on the rate itself, see what counts as a good automation rate.
Sources referenced: HHS Office of Inspector General, Compliance Program Guidance for Third-Party Medical Billing Companies, published in the Federal Register on December 18, 1998; American Medical Association CPT licensing requirements; Solventum APR-DRG licensing terms.
