Attending-level inpatient clinical intelligence

One verified patient state.
Every inpatient decision.

MED AI Inpatient IQ reconstructs the clinical state behind the chart—what changed, what proves it, what needs attention, and what still blocks safe care. MediNote is its documentation layer, not the whole product.

Request a governed pilot Explore the IQ system

Read-only firstLeast-privilege integration pathway

Source-attributedEvidence and chronology remain visible

Physician-controlledNo autonomous chart decisions

Public GPT: synthetic or properly de-identified cases only · Live PHI and Epic connectivity require a separately contracted, institution-approved deployment

INPATIENT IQ · CURRENT STATE

What matters today

Physician review

Hospital day 43 active problems2 pending1 barrier

01Acute hypoxemic respiratory failureImproving

Oxygen 6 L → 2 L; current objective trajectory remains attached to the problem instead of copied prose.

CHANGEDRenal function improvingMedication implications require reconciliation.

REVIEWHeld RAAS therapyRestart criteria and discharge intent are unclear.

BARRIERHome oxygen not confirmedQualification and arrangement status remain pending.

✓

Audit-ready stateFacts, suggestions, uncertainty, contradictions, and completion status remain distinct.

Clinical state Quantitative reasoning Evidence provenance Medication trajectory Physician control

One coordinated platform

The intelligence comes first.
The note comes later.

The final system preserves every prior MediNote layer while assigning each one a clear role in a single clinical operating model.

MED AI

Clinical intelligence core

Interprets clinical meaning, ranks active risk, preserves uncertainty, and links evidence to clinician-review considerations.

MCIF 2.0

Patient-state framework

Tracks source, time, state, confidence, trajectory, medication status, contradictions, and unresolved data.

Quantitative IQ

Statistical reasoning layer

Runs time-aware multimodal models in parallel, quantifies uncertainty, and estimates change, risk, and decision value without turning correlation into fact.

Inpatient IQ

Clinician workspace

Organizes what changed, what matters today, what remains pending, and what prevents safe transition of care.

MediNote

Workflow output layer

Turns the verified state into concise notes, handoffs, discharge courses, medication reviews, and Epic-ready text.

Audit Layer

Verification and governance

Checks support, chronology, conflicts, medication transitions, CDI risk, disposition consistency, and approval state.

Chartowns facts→MED AI / MCIForganize, reason, suggest, detect→Auditverifies→Physiciandecides and approves→Epicholds the signed record

Always-on quantitative intelligence

Not another score.
A statistical reasoning system.

Quantitative IQ runs behind the clinical modules in parallel. A governed method router selects the simplest defensible model for the question, fuses multiple data types over time, and returns an estimate only with its uncertainty, provenance, assumptions, and action threshold.

Model-appropriate, not model-maximalMore complexity is used only when it improves calibration, clinical net benefit, transportability, and human-team performance under independent validation.

TEMPORAL

Patient-state estimation

Bayesian state-space, trend, change-point, and measurement-error methods distinguish true change from noise across irregularly sampled data.

MULTIMODAL

Evidence fusion

Labs, vitals, medications, orders, imaging, and note-derived assertions are joined by time and provenance; copied or correlated evidence is not double-counted.

DYNAMIC RISK

Competing-outcome modeling

Dynamic survival and competing-risk models estimate time-dependent outcomes such as deterioration, ICU transfer, discharge, and readmission with explicit horizons.

DECISIONS

Causal and utility analysis

Causal models are used only when assumptions are defensible. Decision curves and expected utility test whether an estimate would improve a real clinical choice.

UNCERTAINTY

Calibration and abstention

Every estimate carries calibration, uncertainty, applicability, and out-of-distribution checks so the system can defer when the input is unreliable.

MONITORING

Learning-system controls

Site-level validation, subgroup performance, drift, alert burden, clinician overrides, and patient outcomes are monitored across the total product lifecycle.

Every quantitative result exposesQuestionInputs & time windowMethod & versionEstimate & uncertaintyCalibrationAssumptionsDecision implication

The complete IQ layer

Built around the patient—not a document template.

Each capability reads from the same reconciled state, so a medication decision, consultant recommendation, and discharge barrier do not drift across separate outputs.

Patient-state reconstruction

Current and historical facts are separated, timestamped, deduplicated, and reconciled into one traceable clinical state.

Quantitative clinical reasoning

A governed model router combines temporal, multimodal, Bayesian, survival, causal, and decision-analytic methods only when the question and data support them.

Acuity & trajectory

Active problems are ranked by immediate risk with improving, worsening, stable, resolved, new, or unclear trajectories.

Diagnostic intelligence

Supported possibilities are linked to supporting and opposing evidence, uncertainty, and the smallest missing discriminators.

Medication & order intelligence

Home, ordered, administered, held, stopped, resumed, inpatient-only, and discharge-intent states remain distinct.

Consultant & ownership alignment

Recommendations, implementation status, changed conditions, inter-service conflicts, pending work, and ownership stay visible.

Safety & contradiction audit

Meaningful chart conflicts, unsupported claims, copied-forward errors, missing critical data, and transition risks are surfaced for review.

CDI, ICD & medical necessity

Supported specificity, query opportunities, severity, causal links, and remaining inpatient need are clarified without upcoding.

Disposition & discharge intelligence

Mobility, cognition, support, equipment, access, placement, pending tests, medication feasibility, and barriers are reconciled.

Documentation & handoff

The same verified state produces H&P, progress, signout, med rec, discharge, pre-op, audit, and structured-data outputs.

End-to-end workflow

One state engine.
Many clinical workflows.

The system selects the narrowest useful workflow instead of mechanically generating every product.

01Chart sourcesEpic, records, orders, MAR, results, consultants

→

02MCIF stateNormalize, timestamp, deduplicate, reconcile

→

03MED AI + Quantitative IQClinical meaning, time-aware models, risk, uncertainty, decision value

→

04Safety auditSupport, conflicts, gaps, completion status

→

05Clinician actionReview, decide, order, communicate, document

Patient overviewRounds briefH&PDaily progressMedication reviewConsult updateClinical handoffCDI auditDisposition planDischarge coursePre-op reviewStructured extraction

Epic integration pathway

Designed to work in the clinical workflow.

The enterprise edition is designed as a clinician-facing SMART on FHIR application. Epic supplies authorized patient, encounter, and user context; MED AI builds the reconciled state; the physician controls every clinical decision and final chart action.

BUILD STATUSSandbox connector and institutional pathway in developmentNo claim of current production Epic access. Every health system must approve, configure, test, and authorize its own deployment.

SMART on FHIR EHR launch

OAuth 2.0, least-privilege scopes, and current patient/encounter/clinician context.

FHIR R4 clinical data

Conditions, observations, medications, administrations, diagnostics, procedures, care plans, requests, practitioners, and location as approved.

Event and backend options

Authorized backend services, CDS Hooks, HL7v2, or local interfaces only when the customer workflow requires them.

Controlled output

Physician-reviewed text or selective write-back only through enabled endpoints; no autonomous ordering, discharge, signing, or silent chart changes.

PatientEncounterConditionObservationMedicationRequestMedicationAdministrationDiagnosticReportDocumentReference

Governed evidence intelligence

Current medicine, with provenance attached.

In an authorized deployment, the evidence layer retrieves the most authoritative available source, records its publication or version date and retrieval date, identifies conflicts, and keeps external evidence separate from chart facts, patient orders, and physician judgment.

Privacy ruleReference queries use the minimum necessary de-identified clinical concepts. Patient names, MRNs, dates of birth, and other direct identifiers are not sent to external reference services.

LICENSED

UpToDate / Lexidrug

Institution-authorized Epic links or HL7 Infobutton integration; no scraping or implied access without a license.

LITERATURE

PubMed / NCBI / PMC

Current peer-reviewed literature, abstracts, open-access full text where permitted, and dated search provenance.

MEDICATIONS

FDA DailyMed / RxNorm

Current structured product labels, label history, standardized drug concepts, and medication-class normalization.

GUIDANCE

CDC / NIH / FDA

Current federal guidance, safety communications, public-health recommendations, and official evidence resources.

SPECIALTY

Professional societies

Current guidelines from the relevant specialty organization, version-checked at the point of use.

PATIENT

MedlinePlus Connect

Authoritative patient education linked separately from clinician-facing recommendations.

Every evidence result carriesSource Publisher Publication or version date Retrieved date Evidence type Patient applicability

Two deployment profiles

One intelligence core.
Different data boundaries.

The public GPT and enterprise clinical product share the same medical reasoning and governance architecture, but they do not share the same data access. PHI capability is environment-specific, contract-specific, and institution-controlled.

PUBLIC GPT STORE EDITION

Clinical reasoning without PHI

For synthetic or properly de-identified cases, education, product evaluation, and controlled workflow demonstrations.

No real patient identifiers or live charts
No Epic connection or external actions
Complete MED AI reasoning with clinician-review boundaries
No claim of HIPAA compliance or production clinical deployment

ENTERPRISE INPATIENT IQ

Contracted, integrated, auditable

For institution-approved clinical use after the complete data flow, hosting, model service, access controls, and workflow are validated.

Applicable BAA and eligible services
Hospital privacy, security, legal, AI, and Epic approval
Role-based access, tenant isolation, audit, and retention controls
Nonproduction testing before any live PHI or chart write-back

Deployment boundary

A public GPT is not the production Epic connector. The enterprise pathway requires appropriate contracts, safeguards, risk analysis, institutional validation, and approved clinical governance before protected health information enters the system.

Founding hospitalist pilot

Test the full intelligence layer.

A controlled 30-day evaluation measures the patient-state engine, quantitative calibration and net benefit, medication and order reconciliation, clinical usefulness, safety findings, discharge barriers, subgroup reliability, and workflow time—not merely note length. Begin with synthetic or properly de-identified cases; activate any PHI workflow only after the enterprise pathway is approved.

FOUNDING PILOT

^$2,500

flat · 30 days · up to 10 named hospitalists

MED AI Inpatient IQ deployment and onboarding
Structured case validation and physician training
Privacy, security, and Epic-readiness review
Weekly adjudication and Day-30 outcome report

Request pilot review →

After the pilot

Clinical intelligence pricing.
Integration scoped separately.

Proposed pricing scales with the physician group. Epic integration, security, implementation, support, and organizational analytics are scoped to each customer environment.

FOUNDING CUSTOMER

^$199/ physician / month

For early design partners converting after a successful founding pilot.

STANDARD

^$299/ physician / month

Core MED AI Inpatient IQ, physician review, reconciliation, safety audit, and workflow outputs.

ENTERPRISE + EHR

Custom

Epic integration, implementation, support, security, governance, and organizational analytics.

Proposed launch prices, not guaranteed future pricing. Volume agreements available for 10+ physicians.

Pilot scorecard

Claims are earned with adjudicated data.

Success criteria are agreed before the pilot. Actual physician-reviewed outcomes—not broad AI promises—determine whether the system expands.

01Patient-state accuracy

02Trajectory accuracy

03Medication-state accuracy

04Calibration & uncertainty

05Clinical net benefit

06Actionable safety findings

07Disposition-barrier capture

08Physician edit ratio

09Workflow time

010Subgroup reliability

Clinical governance

The system assists.
The physician governs.

The chart owns facts

Current objective data, orders, MAR, results, consultants, function, and documented history remain the factual source.

MED AI owns organization and analysis

It structures, reconciles, prioritizes, links evidence, and runs governed quantitative methods while preserving uncertainty and separating estimates from facts.

The Audit Layer owns verification

It checks support, chronology, contradictions, medication state, missing information, CDI risk, disposition, and completion status.

The physician owns care

Diagnosis, treatment, orders, medications, disposition, discharge, communication, and signature remain under clinician authority.

One final system

MED AI intelligence.
MediNote precision.
Inpatient IQ workflow.

Request the founding pilot →

One verified patient state.Every inpatient decision.