Why this project
The ECG toolkit answers a clinical question with a model. This project is the other half of the work: taking two messy public datasets, joining them, and answering a question a health system executive actually asks on a Monday morning.
I have spent four years inside a cardiac service line. The questions asked in that room are financial as often as they are clinical. Which payers reimburse well for the tests we run? Are we performing procedures at rates below what comparable hospitals negotiated? Where is the gap between the care we deliver and the revenue it produces?
Those questions became answerable in a way they were not three years ago.
The opening
Since 2021, hospitals have been required to publish their negotiated rates. Since April 2026, they must do so in a standardized machine-readable format, which for the first time makes rates comparable across institutions. Most hospitals now post something; considerably fewer post something usable.
The result is one of the largest pricing datasets in American healthcare, and it is barely analyzed outside of consultancies. Published analysis has found negotiated rates for the same inpatient procedure varying by a ratio of roughly nine to one across the country, and different payers paying materially different amounts for the same procedure at the same hospital.
The problem with rates alone
Price transparency files list what a hospital negotiated. They do not list how often the service was performed. Revenue is rate multiplied by volume, so rates by themselves cannot answer a revenue question — a high rate on a procedure nobody orders is worth nothing.
Data
| Source | Provides |
|---|---|
| Hospital price transparency files | Payer-specific negotiated rates by procedure code, standardized across roughly 5,000 hospitals |
| CMS utilization and payment data | Service volumes and Medicare payment by provider and procedure |
| CMS hospital cost reports (HCRIS) | Facility characteristics, payer mix, and operating margin for context |
The questions
For a given cardiac diagnostic, how does one hospital's negotiated rate compare to the rates its regional peers secured from the same payer?
Which payers reimburse cardiac diagnostics well, and which reimburse poorly relative to the market?
Where volume is high and the negotiated rate sits below the regional median, how much revenue does that gap represent annually?
Do hospitals with weaker operating margins also negotiate weaker rates, or is rate negotiation independent of financial health?
Approach
↓
Match facilities — NPI and CCN crosswalk, manual reconciliation
↓
Filter to cardiac diagnostic codes — echo, stress testing, catheterization
↓
Estimate revenue — rate × volume, by payer and facility
↓
Benchmark against regional peers
↓
Dashboard and written recommendation
Scope, deliberately narrow
Washington State, cardiac diagnostics only. The dataset is national and the temptation is to analyze all of it, but a finished narrow analysis is worth more than an unfinished broad one — and a local scope means I can sanity-check the output against hospitals I know. If the model says a facility is underpaid for stress echocardiography, I can ask whether that is plausible.
What the data cannot support
- Transparency files cover hospital-based services only. Most healthcare encounters happen elsewhere, so this is not a complete picture of a system’s revenue.
- Rates are self-published and hospitals interpret the schema differently. Data quality varies, and some files are compliant in name only.
- Utilization data skews Medicare. Commercial volume must be estimated rather than observed, and that assumption needs stating every time a number is quoted.
- This estimates revenue. It does not estimate cost, so it cannot speak to margin.
Status
In design. Data sources identified and limitations mapped; the facility join is the next problem to solve.