Clear the queue.
Trust the answer.

The deduplication engine your immunization registry has been waiting for

Intelligent matching
Automated patient and vaccine deduplication tuned to IIS data realities.
Every decision, explained
Full transparency on every determination — score, rule, and reasoning.
Local, on-premises
Deployed entirely within your environment. PHI never leaves your network.

Built for the messy reality
of immunization data.

State immunization registries handle millions of records from hundreds of sources — pharmacies, clinics, schools, EHRs. Until now, the choice has been between aggressive automated tools that create false merges, and conservative manual review queues that can't keep up. CohesiveIZ resolves that tension.

Phase 1 — Patient identity

Patient identity dedup, end to end.

Married names, adoption, foster placements, naturalization, twins, common-name collisions, and gender-conflict records — all handled by a single coherent pipeline that knows when to merge, when to defer, and why.

  • Algorithmic scorer combines Fellegi-Sunter, Jaro-Winkler, and population frequency
  • Guard rails catch structural patterns the score alone can't see
  • LLM reasoning with score breakdown, key factors, and identified risks
  • Calibrated uncertainty — every verdict is decisive, deferred, or rail-overridden
CohesiveIZ patient identity dedup interface
Phase 2 — Immunization events

Vaccine event dedup, audit-ready.

Combo-component reconciliation, ACIP interval violations, live virus conflicts, series overcounting, source-authority differentials — each with a clear, explainable verdict. Built around CDC IIS Functional Standards, ACIP Best Practice Guidelines, and CVX/MVX code sets.

  • Combo / component overlap detection (e.g. Pediarix vs. its DTaP component)
  • ACIP-aware interval and live-virus conflict checks
  • Source authority differentials between administered and claim records
  • Per-decision scoring breakdown across CVX, date, lot, provider, source, dose
CohesiveIZ vaccination event dedup interface
Operational visibility

The dashboard every IIS manager wants.

Queue burndown, match and reject rates, false merge counts, average confidence, processing time. Override and defer rates that decline over time as the system learns. Confidence distribution by decision type — the metrics auditors and program leads actually ask for.

  • Queue burndown — cumulative resolved vs. estimated remaining
  • Month-over-month volume, match rate, defer rate, and confidence
  • System learning visible — override and defer rates declining over time
  • Confidence distribution by decision type (Match / No Match / Needs Review)
CohesiveIZ analytics dashboard

A layered architecture,
working in concert.

Deterministic rails handle structural patterns. AI reasoning handles ambiguous cases. The two work together, not in conflict — clearing 95%+ of pairs automatically and surfacing only what genuinely needs a human.

Your IIS
Inbound HL7messages from clinics, pharmacies, schools
Inline matcherresolves 95%+ on the wire
Pending review queue2–5% residual · grows daily
CohesiveIZ ringCohesiveIZ Deduplication Appliance
1
Algorithmic scorerFellegi-Sunter · Jaro-Winkler · population frequency
2
Guard railscomplex scenario definitions & structural guidance
3
LLM reasoningdecision · confidence · key factors · risks
Back to IIS
MATCHmerge per MIROW
NO MATCHdismiss pair
REVIEWanalyst decides

Zero false merges as a design constraint —
not an aspiration.

Zero false merges by design

Validated 100% sensitivity on a 350-pair benchmark. Every false positive in testing traces to a single documented test-data limitation.

Layered defense

Deterministic rails handle structural patterns. AI reasoning handles ambiguous cases. The two work together, not in conflict.

Every decision, explained

Score breakdown by dimension. Which rail fired and why. The full reasoning the AI used. No black boxes.

100%
sensitivity in benchmark testing
32
dedup scenarios validated across patient + vaccine
98.6%
of pairs resolved without human review
0
PHI to third parties — local-only
Compliance & security

Built for the way IIS actually operates.

HIPAA-compliant by architecture, not by promise. Deploys in your environment in days, not quarters — on commodity hardware, with a small open-weights AI model running entirely inside your network.

Local, on-premises, no cloud

Your data never leaves your network. Runs entirely on commodity hardware with a small open-weights AI model.

No PHI exposure

No telemetry. No third-party API calls. HIPAA-compliant by architecture, not by policy promise.

CDC-aligned

Built around CDC 2013/2025 IIS Functional Standards, ACIP General Best Practice Guidelines, and CVX/MVX code sets.

Audit-ready trail

Every decision logged with score, rail, reasoning, and key factors — exactly what state IIS managers need at audit time.

Commodity hardware

No GPU farms. No specialized infrastructure. The whole appliance runs on hardware your team already manages.

Days, not quarters

Deploys in your environment in days. Configurable to your jurisdictional rules, MIROW conventions, and source authorities.

In partnership with
AMCI — Atlantic Management Center, Inc.

Ready to clear the queue?

We'd love to walk through your data, your environment, and exactly how CohesiveIZ would fit. No slideware — a working demo on real records.