Why CMS's shoppable-services lists don't help consumers

The federal hospital price-transparency rule was sold as a consumer tool. The phrase "shoppable services" — CMS's own term — implied patients could now compare hospital prices the way they compare flights. Five years after the rule took effect, that promise has not been kept. Hospital prices are public, but they are not shoppable in any consumer sense. Most patients still get a bill they can neither predict nor verify.

This is the gap between the policy and the reader. It is also the reason this site exists.

What the law promised

The CMS rule (covered in detail in the regulations explainer) requires every Medicare-participating hospital to publish two artifacts: a comprehensive machine-readable file of every standard charge, and a consumer-friendly display of at least 300 "shoppable services" — services a patient can plan in advance, where price comparison is meaningful. The intent was a price-transparency analog to a search engine: any patient could compare hospitals, narrow to a procedure, and pick.

On paper, the architecture works. Hospital A publishes its prices, hospital B publishes its prices, the patient compares them. In practice, the architecture only works if a patient can find both files, can read both files, and can decide whether the rates apply to their plan. Each of those is a non-trivial barrier on its own.

What patients got

What actually shipped: each hospital publishes its own files on its own website, in one of four CMS-approved formats. No two hospitals format the data the same way. There is no central registry, no aggregator, no search interface — and CMS does not provide one. A patient who wants to compare three hospitals for an MRI must visit three websites, find three buried links, download three multi-megabyte files, open them in three different tools, and reconcile three different schemas.

This is not what consumer comparison shopping looks like in any other industry. When the airline-fare-disclosure rule was modernised, search engines were the obvious downstream layer. The price-transparency rule has no equivalent — the disclosure landed, but the discovery and comparison surface didn't follow.

The six pieces of friction

The work an unaided patient has to do to compare two hospitals on price:

1. Decide which hospitals to check. No price-transparency disclosure includes geographic search. A patient must already have a candidate list — usually from their insurer's directory or a personal recommendation — before any comparison can begin.
2. Find each hospital's MRF. The link is supposed to be on the homepage. In practice it is buried, behind redirect chains, or broken in the days after a republish. (How we keep the link graph current is in the discovery piece.)
3. Open the file. A typical CSV is 50 to 200 megabytes; a typical JSON is several gigabytes. Spreadsheet tools choke on these sizes. Most patients don't have an analytical tool that scales.
4. Find the right code. The MRF lists procedures by CPT, HCPCS, MS-DRG, APC, or NDC code. Patients usually don't know the code for the procedure they're shopping for, and an MRI knee can be coded multiple ways depending on whether contrast was used.
5. Decide whether the rates apply to the patient's plan. A payer name in an MRF is one signal — not proof of in-network status for any specific plan today. Plan names vary. Bundling differs. Two readers with the same insurer can land on different rates.
6. Interpret the gross / cash / negotiated distinction. The chargemaster (gross), self-pay cash, payer-specific negotiated, and de-identified min-max rates are all in the same row, all mean different things, and all matter for different scenarios. None of them is the patient bill.

Any one of those is enough to defeat a casual reader. All six together are why patient-facing price-comparison adoption of MRF data has been close to zero five years in.

A real Colorado example

Five Colorado hospitals, the same MRI knee (CPT 73721), the same Colorado data refresh — what their published median commercial negotiated rate, plus their median cash price, looks like side by side:

• · Boulder Community Health Foothills Hospital (Boulder): $353 negotiated · $2,543 cash
• · National Jewish Health (Denver): $827 negotiated · $1,555 cash
• · UCHealth Longs Peak Hospital (Longmont): $849 negotiated · $2,238 cash
• · HCA HealthONE Aurora Medical Center (Aurora): $7,076 negotiated · $10,109 cash
• · HCA HealthONE Sky Ridge Medical Center (Lone Tree): $12,405 negotiated · $19,383 cash

Five hospitals, all on the Front Range, all publishing their MRF as required. The median commercial negotiated rate ranges from $353 to $12,405 — a 35x spread for the same CPT code in the same metro area. Across all 39 Colorado hospitals reporting an MRI knee in the most recent refresh, the 25th-to-75th percentile band sits at $611 to $2,747, with the cheapest CO hospital at $227 (UCHealth Estes Valley) and the most expensive at $12,405 (Sky Ridge).

This is the data the rule put on the public record. It is also the data that a patient would never see laid out this way under the rule alone — every number above is from a different file, in a different format, on a different hospital website, requiring six independent steps to reach.

The data is there. Aggregation is what's missing.

The information asymmetry the rule was supposed to fix has not been fixed. The CMS validator and enforcement bulletins ensure the data exists. They do not ensure anyone can read it. The single most important thing the rule did not require — and the single most expensive thing for any individual patient to do — is aggregation.

There are roughly 6,000 hospitals subject to the rule in the United States. Their MRFs are 6,000 separate files on 6,000 separate websites. Until something pulls them together, the price disclosure is closer to a regulatory compliance artifact than to a consumer tool.

What aggregation looks like

Each procedure page on this site folds the same procedure across every Colorado hospital that publishes it into one view:

• · A percentile band — the 25th to 75th percentile of per-hospital median commercial negotiated rates. Half the rates fall in that window.
• · A Colorado median — the 50th percentile of the same distribution.
• · Median cash and gross — what self-pay and chargemaster look like across the state.
• · A per-hospital ranking — every reporting Colorado hospital, with its own rate, its percentile rank, and its in-hospital negotiated band.
• · A payer count per hospital — how many distinct insurers each hospital has rates posted for, as one signal of disclosure depth.

Same for every other shoppable service in our catalog. The transformation isn't analytical magic; it is just the work the rule didn't require any single party to do — read every MRF, normalise the schema, exclude government fee-schedule rates, compute medians, and present them somewhere a reader can land.

The next missing piece

Aggregation across hospitals isn't enough on its own. A reader on the Front Range comparing an MRI doesn't usually want every Colorado hospital — they want the ones in their metro, on their plan, that they can reasonably get to. The current procedure pages compare across all of Colorado uniformly. The next layer is geographic, plan-aware, and quality-overlaid filtering — what a real consumer comparison surface looks like, and what the rule never delivered.

That is the topic of the next piece in this series — what cross-hospital price discovery should actually look like, and the honest constraints involved in building it.

Where to start, if you're a Colorado reader

Three procedure pages cover the highest-volume shoppable services in our catalog and are the most useful starting points if you have an upcoming MRI, screening, or lab visit:

• · MRI of the Knee, Without Contrast (CPT 73721) — the example used in this piece.
• · Colonoscopy, Diagnostic (CPT 45378) — most-asked-about screening, also in the CMS-shoppable set.
• · Screening Mammography, Bilateral (CPT 77067) — annual screening; price varies less than imaging but still spans 3-4x.
• · Lipid Panel (CPT 80061) — the basic cardiovascular-risk blood test; prices range from under $20 to over $200.
• · Comprehensive Metabolic Panel (CPT 80053) — common chemistry panel; another lab where price spread is striking relative to clinical complexity.