Hospital prices vs Medicare: median gross charges ran 3.3 to 5.6 times Medicare at six outpatient codes

How do published prices compare with Medicare's amount?

At six common outpatient codes, the median hospital's gross charge was 3.34 to 5.55 times its own Medicare amount. The discounted cash price ran 1.65 to 2.76 times. Negotiated rates ran 1.22 to 1.56 times with all listed payers and 2.00 to 2.53 times for payers not on our government-plan name list.

The six codes are the emergency department visit (99284) and five pain and orthopedic procedures: 64483, 64493, 62323, 20610 and 25600. Each figure uses outpatient lines (not marked as a physician charge) for that exact code. Medicare here is the hospital outpatient amount, not the physician's fee. The emergency visit and the four imaging codes carry payment conditions, given under Table 1 and in the imaging section.

What do the three kinds of published figure mean?

The federal price transparency rule requires hospitals to publish five kinds of standard charge, and this study uses three of them. A gross charge is the amount on the hospital's own price list. The discounted cash price is the offer to a patient paying without insurance. A negotiated rate is an amount a payer and hospital agreed for a plan.

A published figure is not a payment. A ratio above 1.00 says a published figure is higher than Medicare's amount, not what any patient or payer paid, and it is not a judgment of any hospital.

Why do we show two negotiated-rate figures?

The conclusion depends on which figure you use. With all listed payers, six-code negotiated medians ran 1.22 to 1.56 times Medicare. For negotiated rates from payers not on our government-plan name list, they ran 2.00 to 2.53. Neither figure is the price, so both appear wherever a negotiated rate does.

Why does the all-listed-payers figure sit closer to 1.00?

Government plans such as Medicare Advantage and Medicaid supply part of the pooled dollar entries. For the six headline codes, 83,813 entries were pooled, and 28.39 percent belong to payers we classify as government by name. Their median ratio to the hospital's Medicare amount is 1.00, which pulls the pooled median toward 1.00. The other entries have a median ratio of 1.72.

Government entries are centered on Medicare but spread on both sides. Of them, 25.59 percent sit within 2 percent of it, 44.14 percent are at least 2 percent below and 30.26 percent at least 2 percent above. The other entries sit mostly above. Figures count entries, not hospitals, and are unweighted by volume.

How do we classify government plans?

The classification is by name: an entry counts as government when the payer or plan name matches Medicare, Medicaid, TRICARE, CHAMPVA, CHIP, Medi-Cal or the abbreviations MCR and MCD. Some government plans carry none of those words, and a few commercial plans may, so it can be wrong in both directions.

We first saw the pattern on a partial test of our script, on 30 of the 194 files. The all-listed-payers definition had been written on 4 October 2026, before any ratio was seen, and we kept it as the headline definition for negotiated rates. The second figure was added beside it, and the entry-level figures after the independent review.

How does this compare with studies of what insurers pay?

RAND's Round 5.1 hospital price study measures a different thing: what employers and private insurers actually paid on claims, relative to what Medicare would have paid for the same services at the same facilities. For 2022, it reports that outpatient hospital facility services averaged 279 percent of Medicare, and all hospital inpatient and outpatient services, professional claims included, 254 percent (RAND report RRA1144-2, page vii, and the journal article on the same study). We measure posted prices: what a hospital's file lists, divided by that hospital's own Medicare amount, code by code. A posted negotiated rate is not a paid claim, and our all-listed-payers figure includes government plans, so the two sets of figures are not comparable.

How do you work out one hospital's ratio to Medicare?

Divide the hospital's published amount for the code by Medicare's outpatient amount for that hospital and code. Take the published amount from an outpatient line not marked as a physician charge, and the Medicare amount from our free hospital outpatient payment rate lookup. A ratio of 1.00 means the two are equal.

1. Find the file. Our guide to hospital price files shows where a hospital's file lives.
2. Pick the lines. Keep lines for the exact code whose setting is outpatient or both, and leave out any line with a professional billing class or modifier 26. Where a file lists several lines or payers for the code, we took the median of them for each kind of figure.
3. Get the Medicare amount. In the lookup, enter the code and the date of service, then pick the hospital. Read the amount shown as paid at this hospital.
4. Divide, kind by kind. Divide each kind of published figure by that amount, and never mix a gross charge, a cash price and a negotiated rate in one comparison.

To match this study, use 30 September 2026 as the date of service. Our Medicare amounts come from the same engine and the same CMS July 2026 release files as the lookup, so for that date the lookup shows the amount we used. A later date can fall under a later CMS release and give a different amount. For the emergency visit and the imaging codes, the lookup also shows the payment condition, and for imaging it labels the amount "if wage adjusted", the reading our tables lead with.

How many hospitals publish a price file we could read?

Of 240 sampled hospitals, 194 had a price file link that answered and 192 parsed completely. A usable outpatient line (not marked as a physician charge) for at least one of ten codes appeared in 176. A file we could not find or read is not evidence that a hospital publishes nothing.

Forty-six hospitals stopped before a file answered. Bot challenges or 403 errors blocked 30, a limit of our collection because we did not try to get around a block. The rest returned a 404 or a web page instead of a pointer file, had a price file link that failed, or sat in a pointer file that did not list them.

How loose are our matching rules?

Looser than a strict exact-name rule. Of the 194 hospitals whose file answered, 88 matched by exact normalized name and 29 sit in a single-entry file. Another 69 matched by partial-name rules we added after seeing ambiguous cases, and 8 we matched by hand. Some homepages were blocked, so 35 domains rest on search evidence alone. Every request used one client, curl with a browser user agent. A later section reruns the ratios on strict matches alone.

What do six outpatient codes show?

At all six codes, among the hospitals that published each kind, the gross-charge median is the highest of the four kinds and the all-listed-payers negotiated median is the lowest. Gross medians run 3.34 to 5.55 times Medicare. Table 5 gives each code's median, quartile range and hospital count.

How wide is the spread across hospitals?

At the 75th percentile, gross-charge ratios ran from 4.62 for 25600 to 8.27 for 99284. For 64483, the gross ratio went from 2.44 at the 25th percentile to 4.89 at the 75th. Few hospitals published a gross charge below Medicare, from 0 to 6.99 percent across the six codes, against 17.05 to 25.53 percent for the cash price.

Why do hospital counts differ by kind?

For 64483, 110 hospitals published a gross charge and 144 a negotiated dollar rate. Many JSON-format files leave the gross and cash fields blank on lines that carry negotiated rates. For 64483, that holds for 26 of 52 JSON-format hospitals, against 9 of 69 tall-CSV and 2 of 27 wide-CSV hospitals. We checked the raw JSON of two such files directly: the hospital leaves those fields null while listing negotiated rates, so the extractor is right. Gross highest and negotiated lowest therefore compares different sets of hospitals.

Do larger hospitals post higher prices relative to Medicare?

Pooled across the six headline codes, larger hospitals have higher median ratios for gross charges, discounted cash prices and negotiated rates from payers not on our government-plan name list. The all-listed-payers negotiated median is about the same in every size band. The ranges overlap heavily, so the differences may be sampling variation.

Gross charges ran 3.60 times Medicare at hospitals under 100 beds (40 hospitals), 4.09 at 100 to 299 beds (75) and 4.46 at 300 or more (47). Negotiated rates with all listed payers ran 1.35 (43), 1.27 (73) and 1.31 (50). Each hospital counts once: it gets the median of its own ratios across the six codes it lists for that kind, and each band's median is taken over those figures. The bands use the same bed counts as Table 4.

Is the size pattern the same at every code?

No. Code by code, the bands hold fewer hospitals and the pattern is not uniform: for the emergency visit (99284), the gross median was 6.18 at 100 to 299 beds (73 hospitals) and 5.58 at 300 or more (46). Table 7 gives every code. No interval is computed for the band figures, and bed count is a size measure only: it says nothing about why a hospital prices as it does.

How precise are the medians?

A 95 percent bootstrap interval shows how far a median could move with a different draw of hospitals. The intervals are wide: for the 64483 gross charge, 2.92 to 4.16 around a median of 3.34. Where intervals for two codes overlap, this study does not claim their medians differ. Table 8 lists them all.

We resampled the hospitals' ratios 2,000 times with seed 20261005 and took the median of each resample. The interval describes sampling variation among the hospitals whose files we could read. It does not cover the 30 blocked hospitals, the matching rules or non-response. The gross-charge intervals for 64483, 64493 and 20610 overlap, so this study reads nothing into the small gaps between their medians of 3.34, 3.41 and 3.35. The endpoints are approximate: two separate runs of 2,000 resamples differ by up to 0.12 at the six headline codes and up to 0.50 at the four imaging codes.

What do the four imaging codes show under two readings of Medicare's amount?

For CT of the head or brain (70450) and three MRI codes, median gross charges ran 13.88 to 23.34 times Medicare with the wage-adjusted reading and 14.21 to 22.85 without the wage adjustment. Both readings are published because the engine says it is not proven that CMS wage-adjusts these rates, which carry status indicator Q3 (may be paid through a composite APC).

An imaging code may instead be paid through a composite payment, in which case the composite rate is paid and the code's own rate is not. Our amounts are the engine's amounts for each code standing alone. The open question and the payment condition leave these ratios with more doubt than the six headline codes.

How much do the two readings differ?

The wage-adjusted Medicare amount divided by the unadjusted one ranges from 0.844 to 1.415 across 760 hospital-and-code pairs (0.726 to 1.415 if the four excluded Puerto Rico hospitals are counted). Median ratios differ between the readings by less than half a point in every cell, and no statement here changes between readings.

The share of hospitals below Medicare moves more. For 70450 with all listed payers, 5.42 percent of 166 hospitals sat below under the wage-adjusted reading and 9.64 percent under the unadjusted one. For payers not on our government-plan name list, 3.68 percent of 163 sat below under the first reading and 1.84 percent under the second. Imaging ratios run higher than the headline codes at every kind of figure and are never combined into one number.

What is the Medicare amount, and which date does it use?

The Medicare amount is the engine's final amount for that hospital and code: Medicare's outpatient rate with the hospital's wage index and, where they apply, the rural sole-community, quality-reporting and non-drug-reduction steps. It is priced as of 30 September 2026 under the July 2026 release and excludes the physician's professional fee.

We computed it with the engine behind our free hospital outpatient lookup, which follows the CMS July 2026 OPPS addenda, with no formula of our own. The engine states two payment conditions. Medicare may pay the emergency visit through a comprehensive payment. The visit is packaged into that payment whenever a comprehensive-APC service is billed on the same claim, and paid separately only when none is. An imaging code may instead be paid through a composite payment, in which case the composite rate is paid and the code's own rate is not.

Why 30 September, and why leave out Puerto Rico?

We priced Medicare as of 30 September, not the collection date. CMS normally issues an October quarterly update, and our archive holds nothing after the July 2026 release. September 30 is the last day before the quarter turns over, so the release in force is certain. We chose the date before comparing any price with any Medicare amount.

Four Puerto Rico hospitals are excluded from every ratio. The frame leaves out the engine's territory kind (American Samoa, Guam, the Virgin Islands and the Northern Mariana Islands). Puerto Rico is inside the outpatient system, so four of its hospitals were drawn. CMS's own files disagree about their wage index, so the Medicare amount differs by 17.6 to 18.7 percent between two readings. We chose to exclude, count and disclose them, leaving 190 analyzed, so the results say nothing about Puerto Rico.

Do the corrections and stricter rules change the result?

No. On the original basis, on a strict-match subset and with one questionable file left out, the ordering of the four kinds of published figure and the conclusion stay the same. The exact medians do move, some by more than 0.05. Each test is below.

Does the billing-class correction change the result?

Our first rule left out only lines marked modifier 26. The files also carry a billing class field, and about a quarter of the usable lines for the six headline codes were labeled professional by the hospital without that modifier. Comparing a physician-charge line with Medicare's facility amount is a mismatch. An independent review before publication found the gap. We added the billing-class rule and reran everything, keeping lines labeled facility, both, or with no billing class.

Table 10 shows the six-code ranges on both bases. The ordering of the four kinds and the conclusion are the same on both. Of 56 median figures, 43 moved by more than 0.05 in ratio. On the six headline codes the largest move in any median is 0.17 (25600, gross charge), and for 99284 the gross median moved from 5.41 to 5.55.

Does a stricter matching rule change the result?

We reran the ratios on only the 113 analyzed hospitals matched by exact normalized name or sitting in a single-entry file. The six-code median ranges were 3.06 to 5.59 for gross charge and 1.53 to 2.37 for discounted cash price. Negotiated rates ran 1.12 to 1.38 with all listed payers and 1.93 to 2.26 for payers not on our government-plan name list. The ordering of the four kinds is unchanged at every code. Individual medians move by more than 0.05 in many cells because the subset is smaller. The largest move is 2.09, for the cash price of 70450 (CT, head or brain).

Does one questionable file move the results?

One hospital's file names a different hospital than the one sampled. Leaving it out moves no median by more than 0.05 in ratio, at any code, kind or reading.

How did we check the figures, and what did we get wrong?

We checked the extractor against separate readings of eight real files and re-priced eight hospital and code pairs with a separate script, matching to the cent. Two things were wrong in earlier versions: a case-sensitivity bug lost 25 hospitals, and an independent review found a billing-class gap.

The extractor streamed all 194 answering files and kept 154,803 matching lines. On seven of the eight checked files, every kept line equaled an independent reading made with grep, jq or a separate parser. The eighth, a 2 GB file, matched on line counts per code. The Medicare script's output was byte-identical across runs, and the two bundled test suites passed, 89 of 89 and 20 of 20.

What did we get wrong?

First, our initial full aggregation run compared the setting field case-sensitively. Hospitals write "Outpatient," "OUTPATIENT" or "outpatient," so valid lines were silently dropped and 25 hospitals were lost. We found it by checking counts. The rule always meant any capitalization, so the code was wrong, not the rule.

Second, the independent review found that our rule kept lines the hospital labels professional without modifier 26. We added the billing-class rule and published the original basis beside it. Every figure here comes from the fixed script, deterministic on two runs.

We also disclose three smaller changes. The second negotiated figure was added after a partial test of the script. The comparison date moved to 30 September before any price was compared. An early note about stub-sized files was corrected: the small sizes came from our size probe.

What can this analysis not tell you?

The data is a sample of 240 hospitals, not a census, and the figures rest on the hospitals whose files we could read, collected on one date. Published figures are asking prices or negotiated rates, not what anyone paid, and Medicare's amount covers the facility payment only.

• A sample. The frame is hospitals Medicare pays under its outpatient system. Children's, cancer, critical access, Maryland, Indian Health Service and rural emergency hospitals, and the engine's territory kind, are outside it. The draw was a seeded shuffle within state and size cells.
• Who we could read. The files we could find and read skew toward larger hospitals, and a file we could not read says nothing about what that hospital publishes.
• Files and dates. Files carry their own effective dates and can differ from the billing system.
• Conditional Medicare amounts. The emergency visit and the imaging codes carry the payment conditions above.
• Codes and settings. The ten codes were fixed before any file was opened. Lines whose setting was missing or inpatient were left out, up to 4,128 for one code. Each hospital gets one median figure per code and kind.
• Exempt hospitals. The federal rule deems some hospitals compliant, such as federally owned facilities, so findings do not extend to every hospital type.

Where can you read more?

Our guide to hospital price transparency files shows where a hospital's file lives. How Medicare physician amounts for ten codes frequent on PI bills changed is in Medicare fee schedule changes for PI codes, and the Medicare rate lookup prices a physician's fee. The same approach applied to no-fault arbitration is in our analysis of New Jersey arbitration outcomes.