Real prices for real procedures, pulled directly from two Charlotte hospitals' own federally-required pricing files. This build is a feasibility test, not a finished product — it shows exactly what the underlying data can and can't support yet.
What actually showed up when we pulled the real files for these two hospitals — the part that matters most for a feasibility call.
Knee MRI comparison works. Both hospitals publish a gross charge, a self-pay cash price, and CPT-matched line items for all three MRI variants. Atrium's cash price runs 19–26% higher than Novant's across all three — a genuine, checkable finding a patient could act on.
Colonoscopy comparison does not work — for either hospital, for different reasons. Novant Charlotte Orthopedic is a specialty orthopedic facility and simply doesn't perform or bill colonoscopies. Atrium Cabarrus lists the CPT codes but with no gross charge attached in its published file. Neither is a parsing bug — it's a real coverage gap. A production version needs a facility-selection step that only offers procedures a hospital actually has priced.
"Negotiated rate" is often a range, not a number. Atrium lists Aetna under one payer name covering several distinct plans with no plan-level detail in this export — for knee MRI without contrast, the rate ranged from $237.95 to $1,945.37 across 9 rows, an 8x spread. The site shows the median with a visible flag rather than picking one silently; that flag needs to survive into any production UI.
File freshness differs by almost a year. Novant's file is dated 1/31/2025; Atrium's is 12/1/2025. A real product needs to surface this per hospital, not imply both prices are equally current.
The core mechanic — CPT-code-matched, side-by-side hospital pricing — is real and works with two live Charlotte hospitals' actual published files. That's the hard technical risk from the original notes, and it clears for at least one procedure category.
What it doesn't yet prove: coverage across a wider set of shoppable procedures, a resolution rule for payer/plan ambiguity that a lawyer or your own judgment is comfortable standing behind, and whether "median of 9 differing Aetna rates" is something you're willing to show a consumer as an actionable number. Those are product and legal decisions, not data-availability ones — the data itself is usable.