Our Methodology
How We Calculate Health Worth
We use available public evidence to show procedure prices, possible financial effects, and possible health effects as separate measures with visible limitations.
Transparent. Source-linked. Clearly qualified.
Available assumptions and sources are shown alongside confidence and limitations.
From Care Delayed to Total Cost
Financial estimates, health effects, and optional economic valuations answer different questions and are shown separately.
Care Delayed
Future Medical Costs
Additional treatments, hospitalizations, and care that may be needed later. Source basis: CMS and condition-specific literature.
Lost Productivity
Workdays lost, reduced performance, and time needed for recovery. Source basis: Bureau of Labor Statistics.
Quality of Life
Impact on daily living and overall well-being, measured in quality-adjusted life years. Source basis: utility studies and ICER.
Health Outcomes
Probability-weighted risk of complications, disability, or premature death. Source basis: condition-specific clinical literature.
Five-Year Journey Total
A probability-weighted condition estimate using one five-year window; not calculated from a procedure price alone.
The Result
Total Cost of Delay
The true value of getting care now versus paying more later.
Our Principles
Transparent
Every assumption is visible and explainable.
Evidence-Based
We use published research and trusted public data sources.
Conservative
We avoid inflating costs and use defensible assumptions.
Educational
Estimates help you understand tradeoffs - not predict individual outcomes.
Plain Language
We believe everyone should understand how these numbers are calculated.
- No jargon
- No fine print
- Just clear explanations
Want the technical details?
Confidence Framework
Every estimate is assigned a confidence rating based on the strength and availability of evidence.
Health Worth is built by Trisha Rawlings, a Certified Health Education Specialist with a doctorate in Public Health Leadership. Her expertise guides the work; these evidence labels describe the underlying data and assumptions, not an independent clinical or health-economic review of each cost model.
High
Direct, recent, well-matched public data or consistent high-quality evidence.
Moderate
Relevant evidence exists but requires modeling or includes material variability.
Low
Limited evidence, indirect evidence, or substantial assumptions.
Review not documented
A separate review of this specific cost model has not been documented. Sources and assumptions remain available for inspection.
Evidence labels
Sources & Evidence →
We pull from best-in-class, peer-reviewed research and public datasets.
FAIR Health
Patient cost benchmarks
CMS Transparency in Coverage
(45 CFR 147.211) shoppable services
Bureau of Labor Statistics
Median daily earnings
ICER
Value-per-QALY thresholds ($100k-$150k)
Peer-Reviewed Literature
Condition-specific outcomes research
Utility Studies
QALY decrement research
Federal data cross-checks
These figures come directly from federal datasets and carry a fetch or publication date. They are cached and refreshed on a schedule - never called live on page load - and are kept separate from estimates still pending advisor validation.
Medicare national average payments
Medicare Physician & Other Practitioners - by Geography and Service · CMS Data API · fetched 2026-08-16 · refreshed annually. Used as a federal cross-check alongside hospital-published cash prices, state price data, and clearly labeled modeled planning bands.
- Total knee arthroplasty · HCPCS 27447$1,177.11
- Diagnostic/screening colonoscopy · HCPCS 45378$187.98
- Polysomnography with CPAP titration · HCPCS 95811$241.43
- Laparoscopic cholecystectomy · HCPCS 47562$470.55
- Screening mammography, bilateral · HCPCS 77067$65.2
- Open inguinal hernia repair · HCPCS 49505$450.39
Future-cost trend factor
Our future medical cost figures are grown at 5.8% per year over a 5-year window, using the CMS-projected national health expenditure growth rate rather than an internal assumption.
- Projection window
- 2024-2033
- Published
- 2025
- Last verified
- 2026-08-16
- Next annual review
- 2027-08-01
Interoperability & Data Roadmap
Today, public datasets and imported pricing files are normalized into traceable models. The architecture is designed with future compatibility in mind; production FHIR, USCDI+, CARIN, SMART App Launch, and TEFCA connections are not implemented.
Where AI Is—and Is Not—Used
Prices, arithmetic, horizons, and probability weighting are deterministic. AI may turn cited research into plain-language education; it does not invent prices or silently change formulas. New narratives require source review before publication.
Built to Scale
A shared condition schema separates evidence, price inputs, insurance assumptions, and presentation. That lets additional conditions, populations, and data partners use the same five-year rules without rebuilding the product.
Example: Migraine journey (modeled from cited sources)
Medical care
Visits and treatment
Varies by journey stage and coverage
Other visible dimensions
- Work and incomeModeled
- Transportation and caregivingModeled
- Daily functionLiterature-derived
- Quality of lifeLiterature-derived
Evidence status
Research-based estimate
Sources, formulas, and limitations remain visible
Combining federal benchmarks with external pricing data
CMS data can tell us what Medicare allows for a billing code. That is a federal reimbursement benchmark, not a price anyone is charged. Hospital price files and Turquoise Health can add what hospitals actually publish for the same code.
These measure different things, so Health Worth does not mix them together. Each source is kept separate, checked on its own, and named on the result it produced. A rate observed in the market is never presented as what you will owe, and Turquoise Health is never presented as a government source.
When there is not enough local data to describe a market, the result falls back to the broader federal benchmark and carries a lower confidence rating. Observations that Turquoise flags as outliers are excluded before anything is calculated, and a distribution that is too thin, too scattered, or too stale is rated insufficient and is not shown at all. Whichever source answered is listed in the evidence view for that result.
1. Procedure and billing code
The search resolves to one service in the registry and the billing codes mapped to it.
2. Available price inputs
CMS reference amounts for those codes, Health Worth curated hospital price files where a market is loaded, and cached Turquoise Health hospital rates where they exist for the state.
3. Validation and normalization
Outlier-flagged and unpriced rows are dropped, duplicate provider/payer/plan rows collapse to the most recent rate, and provider-eligibility rules for the service are applied.
4. Geographic matching
State-level matching, with a national federal benchmark as the fallback when no state data qualifies.
5. Calculation
Median with 10th and 90th percentiles for observed rates; for CMS, the median allowed amount with a published 1.0x-2.6x spread.
6. Confidence and limitations
A confidence rating from observation count, provider coverage, payer diversity, spread, and freshness. Anything rated insufficient is not shown.
7. Displayed planning range
The range, plus a plain statement of which source produced it and what it does and does not include.
| Capability | Status |
|---|---|
| Federal benchmark pricing (CMS reference amounts) | Implemented |
| Hospital price-file ingestion (Health Worth curated markets) | Implemented — limited markets |
| Turquoise Health pricing integration | Connected — coverage varies |
| Source named on every displayed price | Implemented |
| Nationwide external price coverage | Not yet implemented |
Availability varies by procedure, facility, geography, and what the external dataset covers. This is a pricing-data integration only: it is not a FHIR connection, not a payer connection, and it carries no patient records.
Example of external pricing data in use
Verified example — not a guaranteed result for every location
- Procedure
- MRI of a lower-extremity joint (knee or ankle) without dye
- Billing code
- CPT 73721
- ZIP entered
- 37203 (Nashville, Tennessee)
- Geographic scope actually used
- Tennessee statewide hospital-price observations. No metro-level or distance filtering was applied, because a sufficiently reliable Nashville-specific sample was not available.
- Why Tennessee data was selected
- The ZIP resolves to Tennessee, and Tennessee is one of the states where mapped observations for this code pass the confidence checks. States without a passing distribution never reach this step.
- Original publishers
- Hospitals and facilities publishing machine-readable price files under the federal hospital price-transparency requirement.
- Turquoise Health's role
- Ingestion and normalization only. Turquoise collects those hospital files into a dataset Health Worth imports on a schedule. Turquoise did not publish the underlying rates and has not endorsed this result.
- Health Worth's role
- Selecting the eligible Tennessee observations for the mapped code, removing outlier-flagged and unpriced rows, collapsing duplicate provider/payer/plan rows to the most recent rate, applying provider-eligibility rules, aggregating, and rating confidence.
- Amount type
- Payer-negotiated rates, estimated allowed amounts, and discounted cash prices.
- Observations used
- 1,449 eligible observations across 47 facilities and 20 payers.
- Data retrieval date
- 2026-08-31
- Displayed result
- $219 (10th percentile) to $903 (90th percentile), typical $308 (median). The typical value is a median, not the midpoint of the range.
- Why it is not a personalized quote
- These are prices hospitals published for a billing code, spread across an entire state. They do not apply your plan, your deductible, or the facility you will actually use.
- Coverage today
- Coverage varies by procedure, facility, payer, and geography, and external observations are currently limited to Tennessee, New York, and Alabama. Everywhere else, and anywhere the confidence checks fail, the result falls back to the CMS federal benchmark and says so.
FHIR Interoperability: Current Demonstration and Roadmap
Synthetic FHIR ingestion demonstrated. Health Worth does not currently connect to a real payer, provider, EHR, or patient record. Production connectivity will require security, privacy, legal, consent, logging, token-handling, and vendor review.
What is demonstrated today
Working demonstrations and future work are listed separately. Roadmap items are not current integrations or capabilities.
Connection capability states
| Connection | State | What that means |
|---|---|---|
| Medicare — CMS Blue Button 2.0 | SANDBOX VERIFIED · NOT IN PRODUCTION | Authorization, synthetic Coverage and ExplanationOfBenefit retrieval, normalization and disconnect have been demonstrated against CMS's sandbox, which serves synthetic beneficiaries only. No real Medicare data is retrieved; production access has not been granted. |
| Aetna Patient Access | CODE IMPLEMENTED · SANDBOX RETRIEVAL NOT YET VERIFIED | Sign-in, token exchange, record reading and normalization are built from Aetna's published Patient Access specification and its sandbox capability statement. No authenticated retrieval of synthetic records has been exercised yet. Not in production. |
| Other health plans | PLANNED | No other payer is connected. The offer is shown as unavailable rather than pretending to work. |
| What a connection supplies | CLAIMS AND COVERAGE ONLY | Connections retrieve claims and Coverage. When Coverage explicitly supplies supported individual USD deductible, coinsurance or out-of-pocket maximum terms, those payer-reported inputs can refine the illustrative estimate without replacing your edits. Most connections may not supply these terms; copays, deductible met, network status and state remain editable. Sandbox data is synthetic, and a plan term is neither a claim nor a guarantee of future payment. |
| Estimate without any connection | LIVE | The full personalized estimate works with no connection at all. A connection only refines it after the person confirms each record. |
Working today
- • Anonymous public-data estimates and condition journeys without login.
- • Evidence, confidence, formula, limitation, and source disclosures.
- • Local synthetic FHIR ExplanationOfBenefit import, validation, normalization, and field-level provenance.
- • CMS sandbox authorization and synthetic Coverage and ExplanationOfBenefit retrieval.
- • Protected, encrypted, shared temporary sessions with automatic deletion.
- • Exact code-to-service and code-to-journey mapping where a verified mapping exists.
- • Screening-to-journey relevance without inferring a diagnosis.
- • Optional health-data connection offered inside the personalized-estimate questionnaire, never required.
- • Provider-agnostic connector layer, so a future payer connection needs no change to the questionnaire or the cost model.
- • Patient confirmation before any connected record is attributed to a condition.
- • Four-way provenance on every estimate component: connected claim, patient-reported, Health Worth calculation, population benchmark.
Not yet implemented
- ROADMAP — NOT YET IMPLEMENTEDProduction payer authorization
- ROADMAP — NOT YET IMPLEMENTEDRetrieval of real patient data
- ROADMAP — NOT YET IMPLEMENTEDSMART App Launch authorization
- ROADMAP — NOT YET IMPLEMENTEDProduction CARIN connectivity
- ROADMAP — NOT YET IMPLEMENTEDProvider/EHR clinical-data retrieval
- ROADMAP — NOT YET IMPLEMENTEDUS Core clinical-resource import
- ROADMAP — NOT YET IMPLEMENTEDTEFCA-enabled exchange
- ROADMAP — NOT YET IMPLEMENTEDPersistent longitudinal patient accounts
- ROADMAP — NOT YET IMPLEMENTEDClinically validated personalization from claims
| Capability | Status |
|---|---|
| Anonymous public-data estimate | Available |
| Synthetic FHIR ExplanationOfBenefit import | Demonstrated |
| FHIR field validation and mapping | Demonstrated |
| Field-level provenance | Demonstrated |
| Session-only synthetic-data handling | Demonstrated |
| Shared, auto-expiring server session store | Implemented |
| CMS sandbox authorization and synthetic-record retrieval | Demonstrated — synthetic records only |
| Production payer authorization | Roadmap — not yet implemented |
| Retrieval of real patient data | Roadmap — not yet implemented |
| SMART App Launch connection | Roadmap — not yet implemented |
| Provider/EHR clinical-data import | Roadmap — not yet implemented |
| Production CARIN Blue Button implementation | Roadmap — not yet implemented |
| TEFCA exchange | Roadmap — not yet implemented |
| Persistent Health Worth patient account | Not part of the initial model |
| Optional connection offer inside the questionnaire | Implemented — Medicare sandbox and Aetna sandbox |
| Provider-agnostic connector abstraction | Implemented |
| Aetna Patient Access (sandbox) connection | Code implemented — sandbox retrieval not yet verified |
| Other health-plan (Patient Access API) connections | Planned — not implemented |
| Patient-confirmed condition attribution | Demonstrated |
| Provenance-labelled estimate components | Demonstrated |
One framework, five different conditions
The shared framework currently powers 33 published condition journeys. 5 journeys have structured public lived-experience review records; additional validation records are being added as documentation becomes available.
See the cross-condition policy demonstration
| Condition | Distinct burden represented |
|---|---|
| Asthma | Recurring treatment, medication and episodic escalation |
| Depression | Behavioral healthcare, work effects and quality of life |
| Type 2 Diabetes | Continuous management and long-term complications |
| Migraine | Episodic disability, treatment and work disruption |
| Throat Cancer | Intensive treatment, recovery, transportation and caregiving |
What every journey reuses
- Journey-stage structure
- Time-horizon logic
- Medical and patient-cost categories
- Work and income categories
- Transportation and caregiving categories
- Daily-function and quality-of-life categories
- Evidence classifications
- Confidence framework
- Source-lineage structure
- Validation-status structure
- Financial-assistance framework
- Sharing and downloading components
What remains condition-specific
- Care stages
- Treatment pathways
- Billing and clinical codes
- Treatment frequency
- Clinical probabilities
- Complication risks
- Utility effects
- Evidence sources
- Cost inputs
- Validation status
Reuse does not mean that clinical assumptions transfer automatically between conditions. Each condition requires its own evidence and review.
Reuse, in implementation terms
- All five journeys are defined as data against one shared TypeScript journey schema; none has its own model type.
- All five use the same journey-stage renderer, the same stage-cost helpers, and the same five-year horizon logic.
- All five use the same evidence-status badges, confidence badges and evidence drawer.
- All five read validation status from one centralized validation record, not from page copy.
- All five use the same sharing and PDF-download components and the same financial-assistance finder.
- No condition has its own renderer, its own calculation function, or its own page component.
No reuse percentage is published, because no reproducible method for calculating one has been documented. Implementing a journey does not mean that journey has completed review: review status is tracked separately for each condition.
Demonstrated reuse vs validation coverage
Demonstrated framework reuse
- Shared journey schema
- Shared journey renderer
- Shared evidence system
- Shared confidence system
- Shared cost categories
- Shared financial-assistance framework
- Shared sharing and download components
Validation coverage
- 5 condition-specific lived-experience review records, out of 33 published journeys
- Other journeys may reuse the framework without having completed condition-specific patient review
- Clinical and health-economic review remains separately tracked and undocumented
Framework-level review does not automatically validate every condition-specific model.
Ready to explore a condition journey?
See costs, evidence, assumptions, and confidence across a five-year journey.
Explore a Condition Journey