Under the Hood

You're not short on data. You're short on a harness.

Every system already captures its piece. They don't talk — so the group ends up with three partial dashboards and no answer to the one question that matters.

The silos · what each system already holds

RCM / billing

· charges
· payments
· denials
· payer mix

[p1] [p2] [p3] [p4]

PowerScribe / mPower

· reports
· findings
· follow-ups

[p6]

PACS / worklist

· volume
· turnaround
· timestamps

[p7]

siloed — each answers only its own slice

One harness → one fact table → one dashboard

What does ED coverage actually cost us — in wRVUs and dollars, by site of service?

Credit · foundation in place

You built the foundation — we set the keystone.

Every tool the group already runs — mPower, the RCM portal, PACS analytics — does its job well and makes this faster to build. None of that spend is wasted; the harness sits on top of it.

The one thing no single vendor can sell is the layer that joins billing and reports and worklist by site of service. No vendor owns all three domains. That is market structure, not a judgment on what you bought.

This isn't AI. It's flat code joining data you already generate.

The engine · answers in, spec out

Eleven answers determine the entire build plan.

The rules are flat lookup. Same answers in, same spec out. No model is reasoning about your group — the spec follows directly from the rules.

Questionnaire · 11 inputs

Billing / RCM

Who runs professional-fee billing?

Claim history depth?

Reporting

Reporting platform?

mPower analytics?

RIS / PACS

Where do they read?

PACS timestamp export?

Data capability

Existing BI?

Data warehouse?

Analyst on staff or contract?

Governance

Vendor BAA process?

OK with a de-identified extract?

Live build plan

Foundation already in place

0 / 4

Confirm tools to credit

Sources

Billing / RCM
blocked · lead deferred
Unknown — answer rcm_owner
access: TBD
Reporting / clinical
blocked · lead deferred
Unknown — answer mpower and reporting
access: TBD
PACS timestamps
blocked · lead deferred
Unknown — answer pacs_ts
access: TBD

Build tier

Standard

Stand up thin stack (warehouse + Metabase/Looker Studio) end to end.

Analyst availability unknown — assume implementation support until confirmed.

Compliance

Not cleared — illustrative only until gates close.

· Establish BAA before any real extract; build on synthetic/illustrative until signed.
· Compliance review of de-identified / limited-data-set approach before extract.

Panels live

0 / 8

Collection rate by sitepending — source
Payer mix by sitepending — source
Uncompensated coverage costpending — source
Net revenue per wRVU by sitepending — source
Denial rate by reason / indicationpending — source
Negative-read rate by indication (the 'fall' panel)pending — source
Turnaround (STAT / routine)pending — source
Past → current → projected trendpending — source

Timeline

~2–3 weeks from access provisioning.

The dashboard · engine truth, rendered

The same panels, eight of them, flip as the questionnaire is answered.

Panel-level state comes from the engine. A whole panel flips when its domain is ready. Compliance trumps. Money panels read the shared model — a number here matches the Sandbox and the Story win-row to the dollar.

Foundation meter

0 of 9 panels wired from data you already own

Illustrative · sample data

★ count + $ · [p1][p7] · two-domain join

Lost-study reconciliation · found money

PENDING — needs source

—

unbilled reads / yr

—

at $38.96/wRVU · blended

Completed reads (worklist) − billed reads (billing). Work already done, charge never dropped. ~100% margin. Default ≈ 1.0% of 150,000 reads — illustrative; 0.5–1.5% typically slips until the join reveals it.

% · [p1]

Net collection rate

ASSUMED · benchmark

92.4%

collections ÷ allowed

$/wRVU · [p1][p5]

Net revenue per wRVU

ASSUMED · benchmark

$38.96/wRVU

blended · across payers

% shares · [p1]

Payer mix

ASSUMED · benchmark

Commercial33%

Medicare22%

Medicaid30%

Self-pay15%

wRVU + $ · [p1][p5]

Coverage gap vs Medicare

ASSUMED · benchmark

$1.11M

33,210 wRVU · no-pay + shortfall

% by CARC · [p2]

Denials

ASSUMED · benchmark

7.2%

CO-50 · not medically necessary · 38%
CO-97 · bundled / inclusive · 22%
CO-16 · missing info · 14%

days · [p1]

Days in A/R

ASSUMED · benchmark

38 days

AR ÷ avg daily charges

% · [p6]

Negative-read rate · fall

ASSUMED · benchmark

55%

fall · clean reads ÷ reads

minutes · [p7]

Turnaround time

ASSUMED · benchmark

28 min

STAT · signed − complete

Provenance · endnotes

p1 — 837/835 claim and remittance feeds (billing-domain truth).
p2 — CARC/RARC denial codes from 835.
p5 — CMS public RVU file (CPT → wRVU).
p6 — Reporting platform (mPower export or NLP over PowerScribe text).
p7 — PACS timestamps (exam_complete → report_signed).