Dashboard

AI-assisted medical coding workflow

Encounters Today
0
No activity yet
Awaiting Review
0
Coder action
AI Confidence
—
No data
Documentation Flags
0
None

Recent Encounters

EncounterSettingAI OutputStatus
No encounters yet. Analyze one to populate this.

Workflow

Paste a note under New Encounter to start.

Engine status

Checking…

New Encounter

Enter documentation and let AI analyze the coding requirements.

01 · Encounter Configuration
02 · Clinical Documentation
Paste a note, then analyze.
Upload Supporting Document 🔒 Premium
🔒
PDF & document upload
Available on Premium
No separate Sources field is needed. The note and uploaded documentation are the source input.
03 · Coding Inputs
Anything left on Automatic is decided by the rules engine. Anything you set here overrides it and is recorded as a coder override.

Review Queue

Every analyzed encounter and its coder decision.

EncounterTypePDXServicesConfidenceStatus
Nothing in the queue yet.

Encounter Review

—

Awaiting Coder Decision

✓ FINAL CODING OUTPUT —

AI Coding Analysis

Billing Details

ServiceCPT/HCPCSRevenueModifierUnitsStatus

Rules Engine Checks

Reports & Quality

Monitor AI coding quality and coder overrides.

First-pass acceptance
—
no decisions yet
Coder override
—
no decisions yet
Documentation flags
—
no decisions yet

Override log

Every encounter a coder corrected, with the reason. This is the error-analysis input for improving the prompt and the rules table.

EncounterAI outputCoder outputReason
No overrides recorded.

Encounter types analyzed

Settings

Configure automatic detection, coding checks and human review.

Automatic Detection

New/established, billing type, care setting and specialty are detected from the note unless overridden on the New Encounter form.

Compliance Checks

Modifier 25, 59 and 26/TC applicability, medical-necessity linkage and documentation gaps run deterministically. NCCI and MUE tables are not loaded yet.

Coder-in-the-loop

The AI recommends; the coder validates, edits or flags. Every override is logged to Reports & Quality.

Architecture

The language model interprets documentation and reports facts. It never selects a code or an E/M level — a deterministic rules engine does that from the facts. This keeps the code set out of the model and makes every recommendation auditable.

Guidelines applied as rules

Uncertain diagnoses excluded; integral symptoms suppressed; first-listed condition selected; coexisting and resolved conditions filtered; specificity and laterality challengedICD-10-CM Official Guidelines · Sections I.B.13, I.B.18, I.C.21, IV.A, IV.D, IV.E, IV.H, IV.J, IV.K
MDM 2-of-3 table → 99202–99215; time-based psychotherapy selection; add-on codes required to carry a primaryCPT E/M Guidelines 2021+ · Appendix A · Appendix D
J-code selection with units derived from the documented dose, and a separate administration line per routeHCPCS Level II · CMS Claims Processing Manual Ch.17
Modifier 25, 59, 26/TC applicability; revenue codes on the UB-04 only; routine foot-care class findingsCMS Claims Processing Manual Ch.13, Ch.25 · Benefit Policy Manual Ch.15 §290
Laboratory CPT/HCPCS lookup (0 entries) and office-procedure lookup

Datasets not loaded

NCCI procedure-to-procedure pairs — no bundling check is performedNCCI PTP
MUE unit limits — only implausible unit counts are flaggedNCCI Policy Manual · MUE
Local and National Coverage Determinations — no coverage or frequency checkLCD · NCD
Radiology CPT selection — depends on study, views and contrast
ED and inpatient E/M families (99281–99285, 99221–99239)

Laboratory code table

MatchesCPT / HCPCS

A starter set covering common panels. Verify every code against the current code set before submission.

Stored data

Checking…