How ProcedureRadar Works
From raw hospital files to the prices you see. Every step, explained.
Last updated
Methodology in one paragraph
Our methodology covers how we collect, parse, and normalize hospital pricing data. We download machine-readable files from roughly 3,500 hospital websites, required under the federal Hospital Price Transparency Rule (45 CFR Part 180), then map raw billing codes to plain-English procedure names, flag suspect rows, and compute distributions so prices are comparable across facilities.
Founder of ProcedureRadar, former NFL running back (Philadelphia Eagles, Minnesota Vikings), former offensive assistant coach (New York Giants), corporate finance and technology entrepreneur, Villanova University alumnus.
Reviewed by Kevin Monangai on .
Methodology is where we show our work. Every price on the site traces back to a hospital Machine-Readable File required under 45 CFR Part 180. We download, parse, normalize, and quality-flag roughly 3,500 hospitals every month, then publish the prices that clear our checks, which today covers more than 2,000 hospitals. If you want to know exactly how a number ended up on our site, this page walks through every step.
Our Pipeline
Every price on ProcedureRadar passes through a six-step pipeline. Each step is automated, logged, and auditable.
Hospitals publish Machine-Readable Files (MRFs) as required by the CMS Hospital Price Transparency Rule (45 CFR Part 180). These files contain every price the hospital has negotiated with every insurer, plus cash and self-pay rates. The rule took effect January 1, 2021, with enforcement penalties increasing through 2026.
We maintain a registry of MRF URLs for 3,500 hospital MRFs across 100 US metros. URLs are refreshed monthly using automated scrapers, the CMS hospital registry, and a CCN-EIN bridge that cross-references hospital identifiers across federal databases. When a hospital changes its file location, our discovery system adapts.
MRF files arrive in three formats, each requiring a dedicated parser. We handle encoding issues, malformed rows, schema variations, and format inconsistencies automatically. Files range from a few kilobytes to multiple gigabytes.
Raw billing codes (CPT, HCPCS, DRG) are mapped to plain-English procedure names using CMS crosswalk files. HCPCS code "G0121" becomes "Colonoscopy screening." Payer names are normalized across hospitals: "UNITED HEALTH CARE," "UHC," and "United Healthcare" all resolve to the same entity. This normalization is what makes cross-hospital comparison possible.
Every pricing record is validated against expected ranges. $0 prices, $999,999 placeholders, and duplicate entries are flagged and quarantined. Each hospital receives a transparency score based on data completeness, freshness, and consistency. Records that pass validation receive a confidence rating visible on every pricing page.
See confidence badges in actionValidated, normalized pricing data is loaded into our database and published to procedureradar.com. Pages regenerate weekly via Incremental Static Regeneration (ISR). The raw source file URL is preserved on every record so you can verify any data point by downloading the original hospital file yourself.
See a live example: MRI of the Knee in New YorkData Sources
Every price on ProcedureRadar comes from a file that federal law requires someone to publish. Almost all of it is hospital Machine-Readable Files. We also load rate files that payers publish themselves, which we cover further down this section.
The Hospital Price Transparency Rule requires all hospitals operating in the United States to publish their standard charges for all items and services in a machine-readable format. This includes gross charges, discounted cash prices, payer-specific negotiated rates, and de-identified minimum and maximum negotiated charges. Non-compliant hospitals face penalties of up to $2 million per year as of 2024.
What's in an MRF File
File Formats Vary
CMS does not mandate a single file format. Hospitals publish in CSV (tall or wide layout) or JSON. Each format requires a dedicated parser, and column naming conventions vary across hospitals. Our pipeline handles all three.
// Actual MRF data rows, simplified for illustration billing_code,description,payer,charge_type,price "G0121","COLONOSCOPY SCREENING, NOT HIGH RISK","CASH","cash_discount",1050.00 "G0121","COLONOSCOPY SCREENING, NOT HIGH RISK","AETNA PPO","negotiated",2340.00 "G0121","COLONOSCOPY SCREENING, NOT HIGH RISK","UNITED HC","negotiated",1875.00 "470","MAJOR JOINT REPLACEMENT W/O MCC","CASH","cash_discount",28500.00
Payer-Published Rates
Hospital MRFs are not our only source. We also load rate files that payers publish directly. Right now that is 19 payers covering 1,311 hospitals across 42 states, and a few things are worth knowing about that data specifically, because these payers publish in genuinely different shapes and we load very different amounts of each one.
What does a Cigna price mean here? Cigna runs a lot of different plan networks, and it publishes a separate rate file for many of them. We load a wide set of those files, not just one. So a price credited to Cigna here is one network's rate, not the whole picture, and it may not match what your own Cigna plan actually pays. Worth checking before you compare.
What does an Aetna price mean here? Aetna does it differently: it publishes a separate rate file for every employer group it covers, and we load one large employer group's file. So a price credited to Aetna here is what Aetna pays for members of one large employer's plan. It may not match what your own Aetna plan actually pays, and it is one real data point rather than Aetna's going rate.
What does a Blue Shield of California price mean here? Blue Shield of California publishes a separate rate file for each of its plan networks, and the ones we measured overlap almost completely. We load one of those network files. So a Blue Shield of California price here is one network's rate, and it may not match what your own Blue Shield plan actually pays.
What does a UnitedHealthcare price mean here? UnitedHealthcare publishes a separate rate file for each employer plan it runs, and about two thirds of the plans we sampled point at the same national network file, Choice Plus. We load that one file. So a UnitedHealthcare price here is UHC's Choice Plus rate at a hospital, for the procedures UHC prices in dollars there. We place a rate at a hospital only when the file names that hospital by its own provider ID. A tax ID isn't enough: when we checked, tax ID matches often turned out to be a surgery center or a different hospital in the same system. That leaves 44 hospitals in 13 states. Your own UnitedHealthcare plan may run on a different network and pay differently.
What does a Medical Mutual of Ohio price mean here? Medical Mutual of Ohio publishes a separate rate file for every employer group it covers, and we load those employer files across its published book. So a Medical Mutual price here comes from real employer plans, mostly at Ohio-area hospitals. Your own plan may still pay differently.
What about the regional Blue Cross plans? Thirteen regional plans publish a table of contents that points at their in-network rate files, and we load a prioritized slice of those files rather than every one. Blue Cross and Blue Shield of Texas covers 197 hospitals here, almost all in Texas. Anthem Blue Cross covers 118 California hospitals and Empire BlueCross BlueShield 49 in New York; both are Anthem plans, and California and New York are the only Anthem states we load so far. Florida Blue covers 79 Florida hospitals, Blue Cross and Blue Shield of Illinois 55, Regence 42 across Oregon, Washington, Idaho and Utah, Horizon 40 across New Jersey and its neighbors, Independence Blue Cross 33 in Pennsylvania and New Jersey, Blue Cross Blue Shield of Michigan 30 around Detroit, Premera Blue Cross 25, all but one in Washington, Blue Cross and Blue Shield of Oklahoma 24, and Blue Cross and Blue Shield of New Mexico 3. The four Blue plans run by HCSC (Illinois, Texas, Oklahoma and New Mexico) publish part of their hospital rates as a percentage of the hospital's charge rather than a dollar amount, and we leave those out because there's no dollar figure to show: in our samples that was about 47 percent of Illinois's hospital rates, 36 percent of New Mexico's, 29 percent of Texas's, and almost none of Oklahoma's. UPMC Health Plan publishes a file for each of its provider networks and we load one of them, its PREMIUM network, at 12 hospitals in Pennsylvania and New York. Our hospital list doesn't reach the Pittsburgh area yet, and that's UPMC's home market, so read UPMC here as a thin slice. Each rate is one network's negotiated rate inside that plan's own region, so it describes that region rather than the country.
How much Wellmark data is actually here? Wellmark publishes a separate rate file for each of its plan networks. Our Wellmark coverage is thin right now, and it is worth saying plainly: what we actually carry is one of those network files, and it lands on three preventive services (a flu shot and the two annual wellness visits) at 20 hospitals, all but one in Iowa. Read Wellmark here as a narrow slice, not as a read on what Wellmark pays generally.
What happens when a payer's own files disagree? One more thing worth knowing. Sometimes a payer's own files disagree about the same price at the same hospital. When that happens we do not average them: a priority order we committed in advance decides which file wins. Measured across Cigna's files, where we first saw these disagreements, that rule lands on the higher of the disagreeing prices about 61 percent of the time. Every losing price is kept in our history, nothing is thrown away.
Where do these rates actually appear? One practical note. These payer-published rates are not the prices you see on the hospital pages here. They are served through our API on the Scale plan and above, and the two data sets are kept apart on purpose: no payer rate feeds a hospital percentile, a confidence interval, or any figure shown on a price page.
Does this follow the same monthly refresh? Hospital prices on this site get re-checked every month as part of our regular pipeline. Payer-published rates don't run on that same clock. Each one carries its own source date from the payer, so a payer rate and a hospital price can be months apart. Check the date on any figure before you lean on it.
Coverage
Our coverage spans the top 100 US metros by population, representing approximately 85% of US hospitals, 90% of consumer healthcare search volume, and 95% of the addressable B2B market. We prioritize CMS-designated shoppable services: procedures that consumers can schedule in advance and compare prices for. This includes imaging (MRI, CT, X-ray), common surgeries (knee replacement, cataract surgery, hernia repair), diagnostic procedures (colonoscopy, cardiac catheterization), and preventive screenings.
Expansion continues to additional procedures (500+ target) and deeper payer-specific rate coverage as hospitals improve their MRF compliance.
Last pipeline scan: . Next scheduled refresh: November 1, 2026.
Quality Scoring
Not all hospital data is equally reliable. Every pricing record on ProcedureRadar carries a confidence rating so you know how much weight to give it. We never hide data quality issues. We surface them.
- 5+ payers reporting rates
- File updated within 90 days
- Prices consistent with expected ranges
- Multiple charge types present
- 2 to 4 payers reporting
- File date 90 to 180 days old
- Minor data gaps present
- Some charge types missing
- Single payer or cash only
- File older than 6 months
- Significant data gaps
- Potential outlier values
Data Quality, Explained
How fresh is this data? Every hospital publishes its own pricing file, and each hospital updates that file on its own schedule, not ours. Right now, 67.2% of priced records come from a hospital file updated within the last 180 days, 24.3% come from a file older than that, and 8.5% come from a file the hospital never dated at all. We re-parse every hospital's file monthly on our own cadence, regardless of how old that hospital's own file is. The freshness numbers above describe the hospital's publishing habits, not ours.
How is a price linked to a procedure? We link prices to procedures two ways. Most of the time, 63.1% of priced records, we match on the exact billing code the hospital listed. The rest of the time, 36.9%, we match on the procedure's written description because the hospital didn't list a workable code. Code matching is the stronger link. Description matching is a real match, but a looser one, and we think you should know which kind you're looking at.
What does the quality flag mean? Every price on ProcedureRadar carries a quality flag behind the scenes: our own read on how much to trust that specific number, based on things like how it was matched and how recently the source file was updated. When a hospital or procedure doesn't have enough priced records to support a real median or percentile, we don't publish a shaky statistic. We suppress it instead.
Figures above are current as of . See our full methodology white-paper.
Limitations
We believe in being upfront about what our data does and does not include. No healthcare pricing tool can capture every variable that affects your final bill. Here is what falls outside our data.
- Surgeon and anesthesiologist fees. These professionals often bill separately from the hospital facility fee. The MRF price may not include their charges. Ask your provider whether the quoted price is all-inclusive.
- Procedure complexity variants. A "knee replacement" could be total or partial. An "MRI" could be with or without contrast. The same procedure name can encompass different scopes of work at different price points.
- Real-time pricing. Our pipeline scans monthly. Hospital prices can change between scans. The prices you see reflect the most recent monthly snapshot, not a live feed.
- Non-shoppable emergency services. Emergency room visits and unplanned urgent procedures are not typically shoppable. Pricing data for these services exists in MRF files but is less meaningful for pre-visit comparison.
- Non-compliant hospitals. Some hospitals still do not publish MRF files or publish incomplete data despite the federal mandate. We can only show prices for hospitals that publish their data. Non-compliant hospitals are absent from our results.
Common Data Questions
Explore the Data
Now that you know how we collect and validate it, put it to work.
ProcedureRadar's pricing dataset is built from federally mandated hospital Machine-Readable Files.
Disclosure required by 45 CFR Part 180.
Last updated: