Methodology

What every number means, and what it doesn't

This is the reference we would want a skeptical colleague to read first. For each axis we state the career attribute we are trying to approximate, the public-data proxy we actually use, how it is computed, the biases it carries, and what a physician, coder, or health economist still needs to verify.

Modeled estimate

Modeled estimate. Not a salary survey. See methodology.

Not a salary survey. Values are modeled ranges, external benchmark references, or user-entered assumptions with limitations.

How to read these pages

Proxies, not paychecks

The Medicare Part B activity proxy reflects Medicare-billed activity, not actual income. It is not salary, collections, take-home pay, or a practice's total revenue. Every figure is a labeled proxy or a reasoned index. A stand-in for something we cannot observe directly, never a stated salary.

Confidence is part of the number

Let the confidence grade guide interpretation. A high score carrying a C or D grade is a real but weak signal, not a firm conclusion; treat it as a hypothesis to pressure-test, not a fact.

Sectors are inferred

Sector classification is inferred from procedure mix and self-reported taxonomy, which is imperfect. Surgeons whose practice spans sectors are attributed approximately, and boundary cases (for example, spinal cord stimulation between spine and functional) are judgment calls.

Aggregates only

Individual physicians are never ranked, scored, or named. V1 reports sector-level aggregates only; any provider-level data exists solely as a suppressed input and is never displayed. This is a decision-support and market-intelligence tool, written to inform judgment rather than to rank people.

Relative, not absolute

Index scores are normalized within these five sectors: the strongest sector sits near 100. They rank the sectors against each other; they do not measure an absolute level, and a 100 is not a claim of excellence.

The confidence model

Each cell receives five sub-scores from 1 to 5, blended by the weights below into a 0–5 score and mapped to an A–E grade. The breakdown is always inspectable on the value itself, so confidence is an auditable object rather than a vibe. Two caps keep us honest: reasoned indices cannot exceed grade C, and suppressed small cells cannot exceed grade D.

Sub-score weights

  • Proxy validity30%
  • Coverage25%
  • Mapping certainty20%
  • Sample adequacy15%
  • Recency10%

Grade bands (on 0–5)

  • AHigh4.3
  • BGood3.6
  • CModerate2.8
  • DLow2.0
  • EVery low0.0

What the sub-scores ask

Does the source actually cover this sector's economic reality (coverage)? How current is it (recency)? How sure are we of the code/condition crosswalk or the assumption (mapping)? Are the cells large enough (sample)? And how tightly does the proxy track the real attribute (proxy validity)?

Reading the Open Payments signal

Transfers of value, not income, and never a leaderboard

Open Payments figures are reported industry transfers of value - consulting, royalties and licenses, travel, food, ownership - not necessarily net personal income. Their presence is neither an endorsement nor evidence of wrongdoing. We report it only as sector-level distributions (median, P90, P99) and category frequencies, with royalty/license separated from consulting and food/travel excluded from the KOL total. Research payments. Frequently directed to institutions rather than the physician. Are excluded as well. Individual physicians are never ranked or named.

See the sector-level Open Payments views

Jump to a metric

Medicare Clinical Revenue Proxy

Data-derived proxyFree
Proxy· Higher is better

What it is

Medicare fee-for-service clinical activity over sector-attributed codes, weighted by work RVUs and shown alongside Medicare allowed amounts.

What it is not

Not income, salary, or total collections. Medicare FFS is one slice of a practice and excludes commercial, Medicaid, facility fees, and salaried support.

Latent target (what we wish we could measure)
Clinical professional revenue capacity.
Proxy (what we actually use)
wRVU-weighted Medicare FFS volume and Medicare allowed-amount sum over sector-attributed codes, shown side by side.
Computation
Aggregate per-NPI service counts x wRVU (and allowed amounts) over the verified code crosswalk, roll up to sector x geography, suppress small cells.
Confidence inputs
Coverage is low where Medicare exposure is sparse; mapping certainty is gated by the unverified CPT crosswalk.

Known biases

  • Medicare FFS Part B only; excludes Medicare Advantage, commercial payers, facility/technical fees, and salaried structures.
  • Only the Medicare subset of each surgeon's volume is observed; total volume requires an explicit payer-mix assumption.
  • Systematically understates sectors with younger / commercially insured patients (especially pediatrics).

Requires verification

Requires verification
  • CPT/HCPCS -> sector crosswalk (licensed coder).
  • Target-year conversion factor and wRVU file release.
  • Whether to apply any payer-mix scaling (default: none).

Industry / Royalty (Open Payments) Signal

Data-derived proxyFree
Proxy· Higher is better

What it is

Reported industry transfers of value by category, with the royalty/license bucket isolated as a noisy signal of monetizable IP.

What it is not

Not net personal income, and not a measure of merit, endorsement, or impropriety. Disputed and misattributed records exist.

Latent target (what we wish we could measure)
Monetizable industry-relationship and IP-royalty potential.
Proxy (what we actually use)
Open Payments aggregate by category (general / consulting / royalty-license / ownership), with the royalty-license bucket isolated as a noisy IP signal.
Computation
Sector/geo aggregate of payment categories via the company->sector map; royalty-license isolated.
Confidence inputs
Proxy validity is weak (payments != income); company->sector mapping is interpretive.

Known biases

  • Disputed and attribution-error records exist.
  • Presence of payments is not merit, endorsement, or income.
  • Conflates marketing relationships with genuine innovation income.

Requires verification

Requires verification
  • Payment-category definitions and de-duplication rules.
  • Company -> sector attribution and multi-sector weighting.

Device Innovation Density

Data-derived proxyPremium report preview
Index (0–100)· Higher is better

What it is

A composite of FDA clearances, active trials, and aggregate patent activity mapped to the sector's device ecosystem.

What it is not

Not a measure of any individual's inventiveness, and clearances and patents are not commercial value.

Latent target (what we wish we could measure)
How innovation-rich and IP-fertile the sector's device ecosystem is.
Proxy (what we actually use)
Composite of FDA clearances (510(k)/PMA/De Novo) by sector product codes + active trials + aggregate patent activity, normalized and weighted.
Computation
Normalize each component, combine with documented weights, express per geography (or per provider) density. No individual attribution.
Confidence inputs
Mapping certainty limited by product-code and patent-class crosswalks.

Known biases

  • Product-code -> sector mapping is interpretive.
  • Clearances and patents are not commercial value.
  • Inventor disambiguation is unreliable; aggregate only.

Requires verification

Requires verification
  • FDA product-code -> sector map.
  • Patent-class -> sector map.
  • Component weighting in the composite.

Clinical Trial Activity

Data-derived proxyFree
Count-based· Higher is better

What it is

The volume of registered clinical trials mapped to the sector by condition - a proxy for research and trial-access momentum.

What it is not

Not a quality measure; a sponsor or site is not an individual, and registration practices vary.

Latent target (what we wish we could measure)
Academic / research momentum and trial access in the sector.
Proxy (what we actually use)
ClinicalTrials.gov counts by condition->sector map, optionally by phase/status and geography.
Computation
Sector/geo trial counts and active-trial share via the condition->sector map.
Confidence inputs
Mapping certainty gated by the condition->sector map.

Known biases

  • Condition tagging is interpretive.
  • Sponsor/site is not an individual.
  • Registration practices vary across sponsors.

Requires verification

Requires verification
  • Condition -> sector map.

Ownership / Facility Potential

Structured assumptionPremium report preview
Index (0–100)· Higher is better· capped at grade C

What it is

A reasoned index of how readily the sector's work supports equity through ASCs, surgery centers, and imaging.

What it is not

Not a measured return. ASC ownership economics are not directly observable, so this encodes qualitative assumptions.

Latent target (what we wish we could measure)
Ability to build equity via ASCs / surgery centers / imaging.
Proxy (what we actually use)
Structured index informed by ASC-migration suitability, Provider-of-Services/HCRIS facility presence (context), and ASC-economics literature.
Computation
Weighted index over sector ASC-suitability assumptions; facility presence is context, not a claim.
Confidence inputs
Capped at C: this is a reasoned index, not measured.

Known biases

  • ASC ownership economics are not directly observable.
  • Reasoned estimate, not a measured dollar figure.

Requires verification

Requires verification
  • Sector ASC-suitability assumptions and weights.

Lifestyle / Sleep / Call Burden

Structured assumptionFree
Index (0–100)· Lower is better· capped at grade C

What it is

A structured index of call intensity, emergency exposure, and schedule predictability, where a higher value means a heavier burden.

What it is not

Not derived from any claims dataset. It rests on qualitative assumptions and varies widely by practice setting and group.

Latent target (what we wish we could measure)
Predictability, call intensity, and sleep disruption (higher index = heavier burden).
Proxy (what we actually use)
Structured index from emergency-driven vs elective case mix, 24/7 coverage needs (e.g., stroke), and published workforce/lifestyle literature.
Computation
User-adjustable weighted index; explicitly not derivable from claims data.
Confidence inputs
Capped at C; coverage and sample adequacy are intrinsically low.

Known biases

  • Not observable in any public quantitative source.
  • Highly dependent on practice setting and group structure.

Requires verification

Requires verification
  • Case-mix and coverage-burden assumptions per sector.

Sources

None. A structured assumption with no quantitative source.Full registry ↓

Geographic Portability

Structured assumptionPremium report preview
Index (0–100)· Higher is better· capped at grade C

What it is

A judgment-weighted index of how freely the subspecialty travels across markets given its infrastructure dependence.

What it is not

Not a measured mobility statistic; the infrastructure-dependence weighting is an explicit assumption.

Latent target (what we wish we could measure)
How freely the surgeon can practice the subspecialty across markets.
Proxy (what we actually use)
Structured index of infrastructure dependence (e.g., comprehensive stroke center / children's hospital concentration) vs ubiquity of demand (e.g., degenerative spine), with NPPES geographic spread as context.
Computation
Weighted index; NPPES spread informs but does not determine the score.
Confidence inputs
Capped at C; reasoned index.

Known biases

  • Infrastructure-dependence weighting is a judgment call.

Requires verification

Requires verification
  • Infrastructure-dependence assumptions per sector.

Fellowship ROI

Structured assumptionPremium report preview
Index (0–100)· Higher is better· capped at grade C

What it is

A transparent, user-modeled estimate of a fellowship's incremental career-economic value, using your own inputs.

What it is not

Not a predicted dollar payoff. You supply the income assumptions; the output is an explicitly relative index.

Latent target (what we wish we could measure)
Incremental lifetime career-economic value of the fellowship vs not.
Proxy (what we actually use)
Transparent user-modeled formula: income delta x horizon - opportunity cost + option value. No claimed dollar payoff; user sets the inputs.
Computation
See fellowship_roi.placeholder.json for the model, inputs, and per-sector illustrative outputs.
Confidence inputs
Capped at C; entirely model/assumption based.

Known biases

  • Defaults are illustrative; the user must supply income assumptions.
  • Pure-income ROI can be negative for some paths (e.g., pediatrics) while non-financial value is high.

Requires verification

Requires verification
  • All default model parameters.

Sources

None. A structured assumption with no quantitative source.Full registry ↓

Wealth Compatibility

Structured assumptionPremium report preview
Index (0–100)· Higher is better· capped at grade C

What it is

A composite index of how well a sector's structure supports building wealth beyond W-2 salary.

What it is not

Not a dollar amount. It inherits the limitations of every component and depends on user-adjustable weights.

Latent target (what we wish we could measure)
How well the sector's structure supports wealth-building beyond W2 salary.
Proxy (what we actually use)
Composite index over ownership potential, royalty/IP signal, schedulability, and (qualitative) commercial-payer exposure, with transparent user-adjustable weights.
Computation
Weighted composite of components (#5, #2/#3, #6, commercial exposure). Index, not a dollar amount.
Confidence inputs
Capped at C; composite of reasoned and proxy inputs.

Known biases

  • Inherits the limitations of every component.
  • Commercial-payer exposure is qualitative and requires verification.

Requires verification

Requires verification
  • Component weights and commercial-exposure assumptions.

Sources

None. A structured assumption with no quantitative source.Full registry ↓

Data sources & provenance

V1 uses public sources only. Proprietary compensation surveys (MGMA, SullivanCotter, AAMC, full Doximity reports) are never republished or used as numeric inputs; where they are relevant we label them as external benchmark references and verify independently.

Two caveats deserve emphasis. The Medicare picture is partial: CMS Medicare data miss commercial insurance, facility and technical distributions, Medicaid, employed-salary support, and private contracts. They therefore undercount most practices and systematically understate sectors that skew younger or commercially insured - pediatrics most of all. And CPT (HCPCS Level I) descriptors are AMA-licensed and are never reproduced here - the crosswalk stores codes and a short, non-copyright basis only, and requires verification by a licensed coder.

CMS Physician Fee Schedule / RVU filesPublic

Conversion factor requires verification for target year.

2023

CMS Open Payments (Sunshine Act)Public

Disputed records exist; payments != income.

2023

NPPES / NPI Registry + NUCC taxonomyPublic

Taxonomy is self-reported and coarse.

not yet ingested (placeholder)

HCPCS Level IIPublic

Public; does not cover CPT (Level I).

not yet ingested (placeholder)

CPT (HCPCS Level I) descriptorsLicensed

AMA copyright; descriptor text must not be republished.

not ingested; license required

ClinicalTrials.gov (API v2)Public

Condition->sector tagging is interpretive.

not yet ingested (placeholder)

openFDA 510(k)Public

Product-code->sector mapping is interpretive.

2023

openFDA PMAPublic

Clearance != commercial success.

2023

openFDA De NovoPublic

Small counts; interpret with care.

2023

USPTO PatentsViewPublic

Inventor disambiguation is hard; aggregate only.

not yet ingested (placeholder)

CMS Provider of Services / HCRIS cost reportsPublic

ASC ownership economics not directly observable.

2023