Find discharge risk hidden in clinical notes.
Marqi converts discharge summaries into structured, auditable risk signals for post-discharge complications, care gaps, and readmission-risk workflows.
Critical risk signals are buried in narrative text.
The most important discharge-risk clues are often buried in free text: pending tests, unresolved acute issues, new mobility limitations, medication changes, social barriers, and follow-up gaps.
These signals are hard to extract consistently across thousands of patients. Manual chart review does not scale. Keyword searches miss context. Marqi solves this with reliability-gated clinical AI.
What gets missed:
- Pending lab or imaging results
- New functional limitations
- Unresolved acute problems
- High-risk medication changes
- Social determinants barriers
- Incomplete follow-up arrangements
Structured risk signals from unstructured notes.
Marqi turns discharge documentation into structured clinical features that can power care-transition dashboards, quality review, population health workflows, and risk modeling.
Pending Workup at Discharge
Tests ordered but results not yet available at time of discharge
"CT chest ordered day of discharge, results pending"
Functional Status Limitation
New mobility or ADL limitations documented this admission
"Now requires 2-person assist for transfers, previously independent"
Unresolved Acute Issue
Active clinical problems without clear resolution pathway
"Ongoing chest pain, cardiology follow-up arranged"
Anticoagulation Initiated
New anticoagulation started during this admission
"Started on apixaban 5mg BID for new AFib"
Wound Care Visits Arranged
Post-discharge wound care or dressing changes required
"Home health arranged for daily wound packing"
IV Antibiotic Treatment
Infection treated with IV antibiotics requiring outpatient completion
"PICC line placed for 4-week IV vancomycin course"
How Marqi extracts risk signals.
A five-step reliability-gated pipeline ensures every extracted signal meets clinical validation standards.
Ingest Discharge Summary
Secure intake of discharge documentation via HL7, FHIR, or direct EHR integration.
Extract Concept-Level Signals
Rubric-driven extraction of specific risk concepts with evidence quotes from the source text.
Model Validation
The Marqi Risk Prediction Model extracts and validates signals through reliability-gated workflows.
Agreement Gating
Concepts meeting Krippendorff alpha thresholds pass automatically. Ambiguous cases route to review.
Export Structured Features
Validated risk signals exported with evidence quotes, confidence scores, and audit trails.
Methodology
Reliability-gated clinical AI.
Every signal is extracted through a rubric-driven workflow and evaluated by the Marqi Risk Prediction Model. Concepts that do not meet reliability thresholds are routed to refinement or adjudication before production use.
Marqi Risk Prediction Model
Our proprietary model processes each document with reliability-gated extraction, eliminating inconsistent outputs.
Rubric-Based Rules
Explicit clinical definitions for each concept ensure consistent extraction across documents.
Agreement Scoring
Krippendorff alpha thresholds gate which concepts can be used in production workflows.
Human Adjudication
Ambiguous cases route to clinical reviewers for ground-truth labeling and model refinement.
Audit Trails
Every extraction includes source evidence quotes and decision provenance for compliance.
Structured Output
Patient-level risk profiles exported with confidence scores and feature-level validation.
Explore the data.
Interactive charts and calculators to understand the impact of discharge risk intelligence.
Risk Signal Distribution
Sample cohort of 150 discharge summaries
Extraction Reliability Metrics
Krippendorff alpha-gated validation
Processing Volume Simulator
See the impact at your scale
Use Case Impact Calculator
Explore the impact across different workflows
Care Transitions
Enhanced post-discharge follow-up for high-risk patients
Structured outputs for clinical workflows.
Marqi delivers patient-level risk profiles with evidence quotes, confidence scores, and complete audit trails.
Structured Risk Features
| Feature | Present | Confidence |
|---|---|---|
| Pending workup | 94% | |
| Functional decline | 89% | |
| New anticoagulation | 97% | |
| Wound care needed | 91% | |
| IV antibiotics | 95% | |
| Social barriers | 86% |
Evidence Quotes
“CT chest with contrast ordered on day of discharge. Results to be followed up by outpatient pulmonology.”
“Patient now requires rolling walker for ambulation. Previously independent without assistive device.”
“Patient lives alone. Daughter unable to provide daily assistance. Home health evaluation recommended.”
Built for care transition workflows.
Discharge risk intelligence integrates into existing clinical and operational systems.
Care Transitions
Surface high-risk patients for enhanced post-discharge follow-up and care coordination.
Population Health
Stratify populations by discharge risk burden for targeted intervention programs.
Quality Improvement
Identify documentation gaps and care process variations across providers and units.
Readmission Research
Generate validated features for readmission prediction models and outcome studies.
Discharge Planning
Support discharge coordinators with structured risk intelligence at the point of care.
Payer Review
Enable risk-adjusted care management and utilization review workflows.
Turn discharge notes into transition-risk intelligence.
Marqi is a reliability-gated clinical AI system that turns discharge summaries into structured post-discharge risk signals with evidence quotes, audit trails, and concept-level validation.