What your hospitals actually get paid, and 10 questions for your plan’s network
Every network says it buys better than the others. Since January 2026, hospitals have had to publish what they were actually paid, not only what the contract says. Nine Virginia hospitals, five insurers and six services, out of the hospitals’ own files.
This page carries the chart from Mockingbirds, a prompt that rebuilds it out of your facilities rather than mine, and the questions that sit underneath a network discount. Everything runs in your browser. Nothing is uploaded, nothing is saved, and nothing asks for your email.
Two numbers, and only one of them is the price
Hospitals have had to publish their negotiated prices since January 1, 2021, under 45 CFR 180.50. The second number is newer. A negotiated rate is what a contract specifies. A median allowed amount is what the hospital reports it was actually paid, required since January 1, 2026 wherever a contract is priced as a percentage or an algorithm, drawn from the hospital's own remittance data over a twelve to fifteen month lookback [1].
Those are different numbers and they are not interchangeable. Fauquier's sepsis contracts run $35,363 to $50,041 on the rate on file. Its own published median for the same admission is $9,139 to $14,587. Case mix, length of stay, DRG weighting and outlier rules all sit between a contracted rate and a remittance. Quote what was actually paid, not the rate on file.
Ask a hospital what something costs and the answer is almost always the same: it depends on what your insurance pays. That is not evasion, it is the accurate answer, and it is the whole problem in one sentence. The price was never a property of the service. It was a property of the arrangement, and until this year the arrangement was not something a buyer could read.
The chart below opens on what was actually paid and shows all nine filers on every service, including the ones that publish nothing, because a blank lane is itself a finding. Switch the number, pick a service, filter to one carrier.
The commercial rates on file, one service at a time
Each dot is one payer contract. Position is the rate the contract specifies, which is not the same as what the plan actually paid. For that, see the last tab. The axis is logarithmic, because the range inside a single market spans more than one order of magnitude. Eight lanes rather than nine. Inova Fairfax is held for the sister-hospital tab, because it files the same rows as Inova Fair Oaks and a second identical lane would only pad this one.
How the codes and carriers were chosen, and what the dots do not include
What the plans actually paid
Same chart, different number. Since January 2026 a hospital that prices a contract as a percentage or a formula has also had to publish the median amount it was actually allowed on that service, taken from its own remittance data over a twelve to fifteen month lookback. Each dot here is one of those published medians. Nothing on this tab is calculated.
Four systems, four sister pairs, four different answers
A quaternary referral center and a community hospital twenty miles apart file the same 21,427 rows in the same order under the same tax ID. Compared cell by cell across all 320 columns, 98.4575% of cells match exactly. The differences are not evenly spread: the gross charge column differs on only 24 rows out of 21,427, while the Anthem and UnitedHealthcare negotiated-dollar columns differ on several thousand each. The table below shows where. Inova Fairfax appears here rather than on the ladder, since a second lane of the same file would only pad the comparison.
Paired lanes show the same service at both hospitals in a system. Where a dot appears in one lane and not the other, only one facility publishes that contract.
Where the two Inova files actually diverge
| Column | Rows differing | Share of file |
|---|
The Bon Secours pair behaves the opposite way. St. Francis and Rappahannock General file separately, at different row counts, and neither publishes a commercial dollar figure for any outpatient CPT code on this chart. There is nothing to pair on the outpatient side. The Sentara pair sits between the two: separate files, separate rates, and both publishing real dollars on inpatient and outpatient alike. The UVA pair is the fourth answer. Haymarket and the Charlottesville academic center file separately, at 482,614 and 2,241,494 rows, and both publish a dollar on nearly every commercial line, which no other pair on this chart does.
The hospital that is being sold, and the system buying it
Fauquier Health is under contract to join Bon Secours Mercy Health. These are the two filers side by side, on the codes where a comparison is possible at all.
What each one charges before any contract applies
| Service | Fauquier gross | St. Francis gross | Difference |
|---|
Gross charge is the chargemaster, the number a percent‑of‑charges contract multiplies. On outpatient services Bon Secours charges less than Fauquier does.
The chargemaster is the price
Fauquier prices 143,325 of its commercial rate rows as a percent of total billed charges, against 20,048 on a fee schedule and 5,645 on a case rate. Each contract carries a small number of percentages. At Fauquier, Cigna’s HMO and PPO both sit at 82.5% of billed charges across settings, Aetna files 58.3% on one plan and 70.8% on another, and UnitedHealthcare’s all‑payer line is 44%. Whatever the chargemaster says, multiply.
| Commercial contract | Percent of billed charges | Applied to this service |
|---|
Head to head, where both publish something
| Service | Filer | Low | Median | High | Contracts |
|---|
The same rule produced five different disclosures
All nine files claim compliance with 45 CFR 180.50 in the attestation row. Here is what a reader can actually take out of them for the six codes tested.
| Filer | System | Rows in file | Commercial rates recoverable | Coverage of the six codes |
|---|
Commercial payer rows carrying an actual dollar amount
At Fauquier, across the whole file rather than just the tested codes, 169,018 commercial payer rows carry a rate. 5,761 of them state a dollar, 3.4%. The other 163,257 state a percentage and stop there. UVA contracts the same way and files the opposite: 366,358 commercial rate rows at Haymarket and 1,934,070 at the Medical Center, of which 91.5% and 93.8% state a dollar, even though roughly two thirds to nine tenths of them are priced as a percent of billed charges. The rule asks for a dollar where a dollar can be expressed. Both hospitals attest that they complied. One did the arithmetic for the reader; the other did not.
Method, scope and known defects
Which carriers
Five are shown: Aetna, Anthem and its Blue Cross affiliates, Cigna, UnitedHealthcare and Sentara. Optima is filed under Sentara, since Optima Health was renamed Sentara Health Plans effective January 1, 2024 and is the same company. Smaller networks and rental PPOs that appear in only one or two of these files are excluded, so the comparison holds across all of them.
What the badge on each chip means
It is how many of the nine filers publish a commercial rate for that code. The three inpatient MS‑DRGs reach 9 of 9; the three outpatient CPT codes reach 7 of 9. That is not a coincidence. Across both Bon Secours files, every single commercial dollar amount on record sits on an inpatient DRG row, and neither hospital publishes a commercial rate for any outpatient CPT code at all. Inova and Sentara publish both.
What a hollow dot is
The hospital’s gross charge multiplied by the percentage of billed charges its contract names, and only where the file’s own methodology field says that is what the percentage means. Fauquier prices roughly 85% of its commercial rows this way. A hollow dot is a calculation, not a published price, and a hospital’s actual allowed amount may differ.
Scope
Facility rates only. The radiologist, surgeon, obstetrician and anesthesia components are billed separately and appear in none of these files. Where a hospital lists more than one gross charge for a code, the higher is used, so derived figures are an upper bound. Plan names are reproduced as filed. Government and Medicare Advantage lines are excluded throughout.
Two known defects in the source files
A negotiated rate exceeds the hospital’s own gross charge on 32,073 UVA Haymarket rows and 41,059 UVA Medical Center rows, at a median of 2.1 times the gross charge. A DRG case rate of $101,963.91 sits against a gross charge of $53,372.31 on the same line. No other filer here does it on a single row. And Fauquier records percent‑of‑Medicare figures above 100% in the same column as percent‑of‑charges figures, on 13,832 commercial rows, so only rows whose methodology field reads “Percent of total billed charges” were converted here.
Fringe Theory is independent and unaffiliated. Views expressed are my own and do not represent those of my employer. Nothing here is legal, tax, medical, or investment advice. Figures described as modeled or derived are my own arithmetic from the public sources listed, not disclosed data.
Nine hospital machine-readable files posted under 45 CFR 180.50, CMS schema v3.0.0, captured August 20 and 21, 2026 from files the hospitals dated between March 24 and July 14, 2026. Group commercial plans only. Facility rates only. Figures derived from a percent of billed charges are marked as derived and are an upper bound.
What this tool does not account for
- It is a dated snapshot, not a live feed. These files are republished and they change. Nothing here should be quoted without its capture date attached.
- Facility only. No professional fees, no separately billed drugs, no behavioral carve-outs. The surgeon, the radiologist and the anesthetist are in none of it.
- Small counts are suppressed. The CMS data dictionary instructs a hospital to encode a remittance count between one and ten as "1 through 10" rather than the figure, and most of the allowed amounts behind this chart are reported that way. The count sits in the tooltip on every dot, and a suppressed one should not anchor a conclusion.
- Price is not spend. Nothing here speaks to utilization, which is where a plan actually wins or loses a year. For what a cost sharing change does against the same claims, see what plan design is worth.
- It is not quality, and it is not Medicare. Nothing on this page measures outcomes, and nothing on it speaks to the widest denominator any buyer has.
- The source file is the arbiter, not this page. If a figure here conflicts with one you hold, or with one a hospital or carrier quotes back at you, go to the file and the line it came from.
Build the same chart out of your own plan
Nothing above requires this page. It is testable on any self funded population without buying anything, and it takes about an evening.
It is worth saying who it is for, because it is not for everyone. Three things make it worth the evening. Your spend is concentrated enough that ten facilities cover most of it. Your employees sit in one place, or in a few places, rather than scattered across dozens of small sites. And there is more than one hospital within a reasonable drive of where they actually live.
Where those hold, a price difference between two buildings is a difference you could in principle act on, and the file tells you which building is which. Where they do not, and a good number of employers are honestly in that position, the exercise still tells you what you are paying. That is worth knowing on its own. It just will not hand you a lever. A population spread thin across a state, or one sitting in a county with a single hospital inside forty miles, has the same files available to it and considerably less to do with them.
- Pull your top ten facilities by paid dollars. Not by visit count, by dollars. In most populations that list is short and the top three carry the weight.
- Open each one's file. Under 45 CFR 180.50 every hospital must post a
.txtfile in the root folder of its website naming the location of its machine-readable file, and/cms-hpt.txtis the name CMS guidance uses and the one hospitals generally post. That is the fastest route in. If it is not there, search the domain for the file name pattern<ein>_<hospital-name>_standardcharges. - Read the median allowed amount, not the negotiated rate. Where the median is present, that is what the hospital was paid. Where the count reads "1 through 10," the figure is thin and should not be quoted.
- Compare that against the discount your renewal deck claims, at the facilities that carry your dollars.
That last comparison is the one a repricing exercise cannot make, because it starts from the assumption that your facility mix is fixed. It runs off the hospital's file rather than your carrier's, because the payer side Transparency in Coverage files carry contracted rates and nothing else, with no allowed amounts and no volume [3].
The prompt
If you would rather hand the parsing to an assistant than do it yourself, this is the prompt. It will not reproduce the chart above. It produces yours, out of your facilities, your carriers and your codes, which is the only version that says anything about your plan.
You are helping me check what my health plan's largest hospitals were actually paid, using public files. Work only from files I give you or that you fetch from the hospitals' own websites. Do not estimate, model, or fill gaps. If a figure is not in a file, it does not go on the chart.
Step 0. What I will not paste, and what you must not ask for. This work needs facility names and the dollars my plan paid at each. It does not need claim lines, member names or IDs, dates of birth, dates of service, or diagnosis and procedure codes attached to a person. If I paste anything that identifies an individual, stop and tell me, and do not repeat it back. If a facility's total rests on only a handful of claimants, treat that line as identifying and roll it into an "all other" row rather than naming the facility. My own contracted rates and repricing output are confidential under my network agreement; the hospitals' published files are not. Work only from the public files.
Step 1. My facilities. I will paste my plan's top facilities ranked by paid dollars from my own claims report, not by visit count. If I have not given you that list, ask for it before doing anything else. If my plan is fully insured, say plainly that facility-level paid dollars may not be available to me at all, and tell me what I can still do without them.
Step 2. Find each hospital's file, and confirm it is the right hospital. For each facility, fetch https://[hospital domain]/cms-hpt.txt. Under 45 CFR 180.50 every hospital must post a .txt file in the root folder of its website naming the location of its machine-readable file of standard charges, and cms-hpt.txt is the name CMS guidance uses and the one hospitals generally post. If that path returns nothing, search the domain for a root .txt file, then for the file name pattern <ein>_<hospital-name>_standardcharges, and tell me which route you had to take. The name on my claims report is often not the legal filer name: a hospital may file under a holding company or an old corporate name. Match on EIN or NPI where the file gives one, record the filed legal name next to my name for it, and tell me where you were not certain. A system may post one file per campus or one file for several, so confirm which building you are reading.
Step 3. Pull only these rows. Keep rows where the payer is one of the carriers I name, on a group commercial product, and the code is one of the services I name. Exclude exchange, marketplace and individual products, federal employee plans, out-of-state Blues, union supplementals, Medicare, Medicare Advantage and Medicaid. Exclude two more that are easy to miss and that move the answer: plans covering a single named employer, and the hospital system's own employee plan. Both are filed alongside real group commercial lines and both are priced differently. Reproduce plan names exactly as filed rather than normalizing them, and do not assume a payer name in the file is the product I actually have.
Step 4. Watch the modifier and the setting. A code carrying a site or facility modifier is not the same service as the bare code, and inpatient, hospital outpatient and off-campus lines are not interchangeable. Keep the modifier and the setting on every row you extract, and exclude rather than blend anything you cannot place.
Step 5. Keep the two numbers separate. A negotiated rate is what the contract specifies. A median allowed amount is what the hospital reports it was actually paid, required since January 1, 2026 wherever a contract is priced as a percentage or an algorithm, over a twelve to fifteen month lookback. Never put them in one column, never compare one against the other, and label every figure with which one it is.
Step 6. Per diem is not a case rate. A per diem is a daily amount and a case rate is an episode amount. Comparing one against the other is not a finding, it is a units error. Use the stated cap where a per diem carries one, exclude open per diems rather than converting them, and say which rows you dropped.
Step 7. Respect the suppression rule. The CMS data dictionary instructs a hospital to encode a remittance count between one and ten as "1 through 10" rather than the figure, and a count of zero as "0". Carry the count next to every median. Flag any median resting on a suppressed count, and do not let one anchor a conclusion.
Step 8. Read the whole file, and do it twice. These files run to millions of rows. Stream or filter them rather than loading a sample, and never quietly truncate: if you could not read all of a file, say so. Then check yourself. Write a second extraction that does not reuse the first one's code or its output, run both, and compare them row by row on hospital, payer, plan, code and modifier. Matching totals prove nothing, because two runs can return the same count and disagree about which rows. Report every disagreement and resolve it against the file before you show me anything.
Step 9. Build the chart. One row per service. One dot per hospital-and-carrier pair. Log scale on the dollar axis, because the range inside a single market spans more than one order of magnitude. Color by hospital. Hollow mark for any figure derived from a percent of billed charges, solid for a published dollar. Show every facility on every service, including the ones publishing nothing, because a blank lane is itself a finding.
Step 10. Show the rows behind it. Give me a table where every figure traces back: hospital, filed legal name, payer, plan as filed, code, modifier, setting, methodology, the figure, the claim count, whether the count was suppressed, the file's own version date and the date you retrieved it. I should be able to open the source file and find any line you used.
Step 11. Tell me what you could not do. List plainly: facilities with no reachable file, facilities you could not confidently match, services with no commercial rate at a facility, medians resting on suppressed counts, and any row where a negotiated rate exceeds the hospital's own gross charge, which is a defect in the source file rather than a finding.
Standing rules. Facility rates only. The surgeon, the radiologist, the obstetrician and anesthesia are billed separately and appear in none of these files, so say so on the output. Where a hospital lists more than one gross charge for a code, use the higher and label the result an upper bound. Convert a percent of billed charges only where the file's own methodology field says that is what the percentage means. Price is not spend, and nothing here speaks to utilization. Attach the capture date to every figure, because these files are republished and they change.
Services. Start with three inpatient MS-DRGs and three outpatient CPT codes that actually carry my dollars. If I have not named them, ask me. Do not default to the codes you have seen used elsewhere.
Step 0 is not boilerplate. A facility list and a paid-dollar total are not protected health information, and that is all this needs. A claims extract usually is, and in a small population a single facility line can describe one person. Run this in a tool your organization has cleared for plan data, and one that is not set to train on what you give it.
What to ask the network, and what to ask about the agreement
The discount is the question with the widest reach and the smallest move, and its move only lands where members already go. These are the questions that sit underneath it. Ask them at renewal, in writing, and keep the answers.
As with the exercise above, these are aimed at a particular reader. An employer large enough to see its own network service agreement, to have someone whose job it is to read it, and to have a renewal conversation where the answers could change something. Below that size the questions are still fair ones, but the leverage to act on the answers usually is not there, and a broker or a purchasing coalition is the more realistic route to asking them.
None of them is an accusation. Almost every one is answerable out of data your carrier already holds or a document you have already signed, which is the point. Where the answer is "we do not report that way," that is also an answer, and it is worth writing down.
Ten questions, in three groups: four on what the network bought, five on what the agreement permits, and one on what reaches a member.
What the network bought
- How does your overall discount compare at our top ten facilities by paid
dollars?
The headline discount is a blended number. Ask for it building by building, at the ten buildings that carry the spend. - Which of these are the most affordable, and which should we steer away
from?
A plain ranking, cheapest to dearest. If the reporting will not produce one at the building level, that is an answer about what the reporting is for. - What does the repricing assume about facility mix?
It almost always holds the mix constant, which assumes our members keep going where they went. That assumption is usually invisible and usually wrong. - What does it assume about contracts renegotiating inside the projection period, and
are any of ours up for negotiation?
The exercise runs on twelve to twenty-four months of history. Ask what happens to the number we were quoted when a contract at one of our largest facilities renews in month seven.
What the agreement permits
These are questions about the document rather than the data, and the document is one we already have. A benefits attorney is the right reader for the answers. The point of asking is to find out which of the four questions above we are actually permitted to pursue.
- What does the agreement permit our plan to audit?
Scope, sample size, auditor of our choosing, look-back period. This one gates most of the others, because you cannot recover what you are not permitted to examine. - Does the agreement let us see rates at the facility level, or only in
aggregate?
Ask it as a document question rather than a reporting question. A reporting limit can be changed by whoever runs the report. A confidentiality term in the agreement cannot. - Which of our facility contracts carry an anti-steering, anti-tiering, all-or-nothing
or most-favored-nation provision, and where in the document can we read it?
These terms decide whether moving volume to a cheaper building is a plan design conversation or a contract amendment. Worth knowing which one we are in before budgeting for a program. - Are there added costs beyond the administrative fee, and can we have a schedule of
all compensation in one document?
Shared savings on out-of-network claims, payment integrity fees taken as a percentage of recoveries, network access fees and pharmacy arrangements are commonly negotiated and disclosed separately. Ask for them together. - What happens if a facility in our top ten leaves the network mid-year?
Notice, continuity of care, what members are told and by whom, and our recourse on a rate quoted against a network that has changed underneath it.
What reaches a member
- You hold cost and quality measures at both ends of the range. What reaches a member
at the moment they are choosing a facility?
In what form, on what screen, and how would we know if it had. An estimator nobody opens is not a program.
There are two more worth asking, on payment integrity results and on eligibility, and they are the two with no public measurement behind them at all. That absence is itself worth knowing before you spend a renewal meeting on the discount.
Why this is here
Democratizing the data is the easy half, and it is mostly done already. These files are public, they are free, and almost nobody opens them. The harder half is the decision. Who makes it, what it actually rests on, what it costs to be wrong, and what it sets running three years downstream. That last part is the tail, and it is usually where the money went.
So this page is not really about nine hospitals in Virginia. It is about whether the question you spent your renewal on was the question worth asking.
Where this came from
The nine files behind the chart, the extracted dataset and the arithmetic behind every figure are in the source manifest. The argument they support is in Mockingbirds.
Sources
- Centers for Medicare and Medicaid Services, CY 2026 Hospital Outpatient Prospective Payment System and Ambulatory Surgical Center Payment System Final Rule (CMS-1834-FC), Hospital Price Transparency policy changes. Median, 10th and 90th percentile allowed amounts and a claim count required where a negotiated charge is a percentage or algorithm, from a lookback of no less than twelve and no more than fifteen months. Effective January 1, 2026, enforcement beginning April 1, 2026.
- Hospital machine-readable files of standard charges posted under 45 CFR 180.50, CMS schema v3.0.0, captured August 20 and 21, 2026: Fauquier Health; Bon Secours St. Francis Medical Center; Bon Secours Rappahannock General; Inova Fair Oaks Hospital; Inova Fairfax Hospital; Sentara Northern Virginia Medical Center; Sentara Martha Jefferson Hospital; University of Virginia Medical Center; University of Virginia Haymarket Medical Center.
- Congressional Research Service, "Technical Challenges with Private Health Insurance Price Transparency Data," R48570, June 13, 2025.
Fringe Theory is independent and unaffiliated. Views expressed are my own and do not represent those of my employer. Nothing here is legal, tax, medical, or investment advice. Figures described as modeled or derived are my own arithmetic from the public sources listed, not disclosed data.