What cross-hospital price discovery should look like

The procedure pages on this site already do something the federal price-transparency rule does not — fold every Colorado hospital's published rates for the same code into one comparable view. That is real progress against the original problem (covered in the anchor piece on shoppable-services gaps). It is not yet what cross-hospital price discovery should mean for a consumer.

This piece lays out where the atlas is now, where it is going, and the honest constraints between here and there. It is a roadmap, not a launch announcement.

Where we are today

Three views are live across the atlas:

• · Per-procedure comparison across all Colorado hospitals. Each procedure page shows a 25th-to-75th percentile band, a Colorado median, median cash and gross, and a per-hospital ranking from cheapest to most expensive. All commercial; government payers excluded (methodology).
• · Per-hospital pages. Each hospital page shows that hospital's median commercial rate per procedure, its rank against the Colorado distribution, and the procedures where it is cheapest or most expensive.
• · Per-payer pages. Each payer index lists the hospitals where the payer has posted negotiated rates. Caveats apply — MRF presence is not in-network proof, and we say so on every payer page.

The three are sufficient for an interested reader to do meaningful price homework. They are not sufficient for the most common question a real patient asks, which is: "What is the cheapest in-network hospital for this procedure within reasonable distance of where I live?"

Where we are going

Geographic filtering

"MRI within 25 miles of Denver" is the obvious next layer. We have hospital addresses; we don't yet have lat/lon enrichment that would make distance filtering exact. The work is small in absolute terms — geocode every hospital once, store coordinates next to the existing address fields, and add a distance-from-zip query to the procedure pages. The pacing constraint is that the geocoding has to be auditable: every hospital location displayed has to be the address the hospital itself publishes, not an estimated centroid.

Plan-specific filtering

The MRF data already includes payer-and-plan rows for most hospitals. Surfacing them as a clean filter — "show me what Anthem Blue Cross has negotiated for CPT 73721 across all reporting Colorado hospitals" — is mostly a UI question. The honest constraint is that there is no industry-standard plan-name registry. Anthem's commercial PPO might appear in one MRF as `anthem-bcbs-ppo`, in another as `anthem-comm-ppo`, in a third as `BCBSCO-PPO`. Plan-name normalisation is a long tail of edge cases. We will not promise plan-level filtering until the normalisation layer is verifiable per hospital.

Quality overlays

Price without quality is half a comparison. The CMS Hospital Compare program publishes star ratings, readmission rates, patient-experience scores, and process-of-care measures. Linking to those on each hospital page is straightforward; presenting them in line with price (so a reader can compare a $353 MRI knee at a 3-star hospital against a $7,000 MRI knee at a 5-star hospital) is the point of the integration. We will link the data rather than host it — it is not our dataset.

Multi-state expansion

Colorado is the first state because the parser, link graph, editorial layer, and validator integration could all be exercised end to end on one state's data before scaling. The next state is whichever state's MRFs the existing pipeline already parses cleanly enough to backfill — we will publish an explicit "state coverage" page when the second state ships. Expansion is sequenced behind data-quality bars, not behind a marketing schedule.

The honest constraints

Three things will not be fixed on our side, no matter how much engineering we put into the surface:

• · Upstream MRF data quality. Hospitals publish what they publish. When a hospital reports an obviously implausible rate or omits a payer, we surface the gap on the workspace scorecard — but the source has to be the hospital correcting its file, not us substituting a number.
• · Plan-name normalisation. Without an industry-wide plan registry, plan-level filtering will always be best-effort with disclosed caveats. We will keep adding to the normalisation map; we will not pretend it is complete.
• · Bundled vs. unbundled billing. A hospital MRI knee row is the facility component. Radiologist professional fees, anaesthesia, contrast — these are billed by other entities. The site cannot reconcile across entities the patient never sees on a single bill.

Publisher first, product second

The position on the about page is straightforward: this is an editorial atlas, not a feature factory. Aggregation across hospitals is the product. Geographic and plan-aware filtering, quality overlays, and multi-state expansion are presentation layers on top of the aggregation. The order matters — pretty filters on bad data is worse than the federal disclosure was to begin with. The data layer comes first; the comparison surface follows.

What that means in practice: editorial decisions (what to publish, what to disavow, what to caveat) take precedence over feature decisions. A filter we cannot stand behind methodologically will not ship until it can be stood behind. Every figure traces back to a hospital's own MRF, full stop — see the editorial policy for the longer commitment.

How to get notified

There is no newsletter signup form yet — we will add one when there is something material to announce. In the meantime, clear-price-corrections@geninspired.com is the contact for any of the following: "my state isn't here yet," "this hospital's price looks wrong," "my plan should be filterable," or "I'm a hospital pricing team and want workspace access." Every email gets read; corrections feed directly into the next refresh cycle.