Clear the queue.
Trust the answer.

The deduplication engine your immunization registry has been waiting for — layered architecture, complex reasoning, and zero false merges by design.

Request a Demo
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. Until now, the choice has been between aggressive automated tools that create false merges, and conservative manual review queues that can't keep up. Cohesive clears the queue.

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 — math: similar name, DOB, address, vaccines, etc.
  • Guard rails — domain knowledge: structural patterns the math alone misses
  • LLM reasoning — judgment: weighs evidence, explains every verdict
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 and ACIP guidance.

  • Combo / component overlap detection (e.g. Pediarix vs. its DTaP component)
  • ACIP-aware interval and live-virus conflict checks
  • Per-decision scoring 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 with your existing IIS.

Cohesive picks up where your existing IIS matcher leaves off — resolving the candidate pairs your blocking and probabilistic scoring couldn't resolve, and returning a verdict with full reasoning. Your IIS records the results.

Existing IIS
Pending duplicate review queueHundreds of thousands to millions of potential duplicate records … and growing daily
CohesiveIZ ringCohesive IZ 3-Layer Architecture
1
Algorithmic scorerFellegi-Sunter · Jaro-Winkler · population frequency
2
Guard rails(20+) — solution / scenario rules · ACIP intervals · 30+ patterns
3
LLM reasoningweighs evidence · explains decisions · calibrate confidence
Back to IIS
MATCHmerge records
NO MATCHdismiss pairs
REVIEWanalyst decides

Zero false merges — by design.

Not a wrapper around an LLM

Open-weights models are commodity. The architecture around them isn't. Twenty-plus guard rails. A calibrated scorer. Benchmarks across 17 failure modes.

Layered defense

Deterministic rails handle structural patterns. AI reasoning handles ambiguous cases. The two work together, not in conflict — covering what either alone would miss.

Every verdict explained

Score breakdown by dimension. Which rail fired and why. The full reasoning the AI used. Every decision logged and auditable. 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 — by architecture.

Local, on-premises, no cloud

Your data never leaves your network. Runs entirely on Cohesive supplied or commodity hardware with a frontier open-weight local AI LLM model.

HIPAA-compliant by design

No PHI exposure. No record-level data leaves your network. Compliance is architectural — not a policy promise.

CDC-aligned & audit-ready

Built around CDC 2013/2025 IIS Functional Standards and ACIP best practices. Every decision logged with score, rail, and reasoning.

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.