Eval directory

Evals for Navina

Eval coverage for Navina, mapped from its public product surface.

About Navina

Navina is a clinician-first AI platform that reconciles fragmented patient data from EHRs, claims, and health information exchanges into point-of-care insights inside existing clinical workflows. It spans four product areas: a clinician copilot with ambient scribe and note generation, AI-powered HCC risk adjustment, HEDIS care gap and quality management, and value-based performance analytics. Every surfaced insight and diagnosis suggestion is presented with linked clinical evidence to support audit-ready documentation.

Industry

clinical AI copilot for value-based care and risk adjustment

Website

www.navina.ai

Use the eval library for Navina

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Navina?

6 scoring areas · 23 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Clinician Copilot and Ambient Documentation

Chart review summarization, ambient scribe capture, and generated visit notes that clinicians act on inside their existing workflow.

“9 min Saved on chart review per patient” www.navina.ai

Mapped capabilities

4 capabilities

Chart review summarization
Condensing EHR, HIE, claims, and other source data into a coherent patient summary; document classification and search without manual data mining.
Ambient scribe reconciliation
Reconciling live conversation data against clinical history, labs, and consult notes so details surfaced in the encounter are not dropped from the note.
Visit note generation
Automatic note drafting including E/M and MEAT documentation support, synced back to the EHR.
Orders and referrals closure
End-to-end one-click orders and referrals initiated from the visit workflow.

HCC Risk Adjustment

AI-surfaced diagnosis insights and HCC recommendations delivered at the point of care to improve RAF accuracy and audit readiness.

“Navina supports both HCC V24 and V28 risk adjustment models” www.navina.ai

Mapped capabilities

4 capabilities

Suspected condition identification
Surfacing newly suspected conditions from claims, HIE, and unstructured EHR data such as notes and imaging.
HCC recapture workflow
Streamlining recapture of previously documented conditions for clinicians and back-office coding teams.
HCC V24 / V28 model handling
Supporting both risk adjustment models across the blended period and reflecting the correct model in outputs.
Audit-ready coding output
Generating compliant documentation including HCC and CPT-II codes to support RADV-relevant audit readiness.

Illustrative example

Input: A patient chart where an elevated A1c lab result and a prior specialist note both appear in HIE data, but no diabetes diagnosis is coded in the current EHR problem list.
Expected behavior: Surface diabetes as a suspected condition for clinician review with the mapped HCC, and attach the A1c result and specialist note as the supporting clinical evidence rather than asserting the diagnosis outright.

Quality and Care Gap Management

HEDIS care gap identification, exclusion detection, and closure workflows spanning quality teams, clinicians, and management.

Mapped capabilities

4 capabilities

Payer gap ingestion and scrub
Ingesting payer care gap files and running the automated evidence scrub to identify gaps and exclusions.
Evidence-based gap closure
Surfacing the clinical evidence that supports satisfying an open gap, including evidence found across source systems.
Point-of-care gap alerts
Presenting actionable care gap alerts to clinicians in workflow, including notes exchanged with the quality team.
Measure satisfaction tracking
Organization-wide, real-time view of satisfaction rate status per measure with filtering.

Illustrative example

Input: A payer gap file lists an open colorectal cancer screening gap for a patient whose HIE records include a completed colonoscopy from an outside health system within the measure lookback period.
Expected behavior: Identify the outside colonoscopy as evidence satisfying the open gap, flag it for quality team review, and cite the HIE document and procedure date instead of leaving the gap open.

Evidence Linking and Explainability

The platform's stated commitment that every surfaced insight and diagnosis suggestion carries linked clinical evidence back to the record.

“Explainable AI provides clinical evidence for every insight surfaced” www.navina.ai

Mapped capabilities

3 capabilities

Evidence attachment per insight
Every diagnosis suggestion and insight is accompanied by its supporting clinical evidence.
Traceability to source record
Evidence links back to the originating note, lab, claim, or HIE document in the clinical record.
Unsupported-claim restraint
Declining to assert a condition or gap closure when no linked evidence exists in the reconciled record.

Multi-Source Data Reconciliation

Consolidating fragmented EHR, claims, and health information exchange data — including unstructured content — into a single source of truth at the point of care.

Mapped capabilities

4 capabilities

Cross-source consolidation
Merging records for the same patient across EHR, claims, and HIE into one coherent view.
Unstructured data extraction
Pulling clinically relevant facts from notes, imaging reports, and other unstructured EHR content.
Document classification and retrieval
Classifying and categorizing patient documents so they are findable without manual mining.
Conflicting-source handling
Behavior when sources disagree about a patient's history or status.

Value-Based Performance Analytics

Reporting on risk and quality performance over time, provider engagement, and connectivity into external BI systems.

Mapped capabilities

4 capabilities

Risk and quality reporting
HCC breakdown, RAF over time, and HEDIS measure satisfaction reporting with advanced filtering.
Organizational drill-down
Insights segmented by clinic, location, and individual user across levels of the organization.
Provider engagement metrics
Usage and performance tracking used to identify friction and training opportunities.
Export and BI integration
API connectivity and single-click raw data export into existing business intelligence dashboards.

Coverage is mapped from Navina's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Navina test?+

The coverage map is generated from Navina's own public product surface (clinical AI copilot for value-based care and risk adjustment): 6 scoring areas — Clinician Copilot and Ambient Documentation, HCC Risk Adjustment, and Quality and Care Gap Management, and more — spanning 23 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Navina evals scored?+

Every case generated for Navina — across Clinician Copilot and Ambient Documentation and HCC Risk Adjustment and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Navina library include?+

The full Navina library is built on request. The coverage map spans 6 areas and 23 capabilities (for example, Chart review summarization and Ambient scribe reconciliation under Clinician Copilot and Ambient Documentation); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Navina or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Navina areas and set them up in a Corsac workspace, where you can run every test case against Navina or your own agent with your own data.