The routing layer federated healthcare has been missing

QIS Protocol is the distributed intelligence layer above NHS Federated AI, OHDSI / OMOP, PCORnet, and CanDIG. It routes ~512-byte outcome packets between hospitals, clinics, and research networks — no raw patient data movement, N(N-1)/2 intelligence connections, O(log N) or better per-node cost (O(1) with a direct semantic-address lookup; O(log N) or better with DHT routing; anywhere between with other transports). The infrastructure most healthcare networks already have solved authorization. QIS solves what they have never touched: synthesis.

The problem federated healthcare networks have not solved

Federated genomics networks like CanDIG built elegant access control. NHS Federated AI moved models to data. OHDSI nodes speak OMOP CDM. PCORnet routes queries across distributed research environments. All of these solve the same problem — authorization — with different vocabularies.

None of them solve the next layer. When a CanDIG node finds a variant combination associated with treatment response in 23 patients, that insight stays at the node. When an NHS TRE runs an analysis, the model lives at the data, but the outcome does not propagate. When an OHDSI node produces a cohort result, no other OHDSI node automatically learns from it. The synthesis layer is missing.

QIS routes the outcomes those networks produce. It sits above the authorization layer, not inside it. The federated infrastructure stays unchanged. Outcome packets — ~512 bytes each, no PHI by construction — route by semantic fingerprint to other nodes facing the same problem profile.

The networks QIS sits above

NHS Federated AI

Adds the missing routing layer. Models still live at the trusted research environment; outcome packets route between TREs by semantic address.

OHDSI / OMOP CDM

Each OHDSI node emits an outcome packet hashed by biomarker profile. Other OHDSI nodes facing the same profile retrieve the synthesis. OMOP layer unchanged.

PCORnet

PCORnet routes queries. QIS routes the outcomes PCORnet learns. Each PCORnet site emits its outcome once; any site facing the same problem retrieves the synthesis.

CanDIG / GA4GH

CanDIG routes access. QIS routes outcomes. GA4GH phenopacket schemas map cleanly to QIS semantic fingerprints. Existing access control infrastructure unchanged.

EHDS (EU)

European Health Data Space federation gets cross-border outcome synthesis without cross-border raw data movement. GDPR Article 9 special-category data never enters the protocol layer.

FDA Sentinel

Sentinel does distributed adverse event surveillance. QIS adds real-time outcome synthesis between Sentinel partners. Drug safety signals emerge before any one site can see them.

Why this works for healthcare specifically

100k+
US ADR deaths/yr
512 B
packet size
0
PHI in routing
N(N-1)/2
synthesis pairs

Adverse drug reactions kill over 100,000 people per year in the United States alone. Many are preventable — the information that could have prevented them exists somewhere in another hospital's records. Today that information is trapped. HIPAA and institutional policies prevent centralization. Federated learning requires central aggregators that introduce privacy risk and single points of failure. Outcome data stays stranded.

QIS makes those silos obsolete without violating any of them. A hospital in Des Moines can benefit from treatment outcomes in Zurich without either hospital sharing a single patient record. A rural clinic in Kenya participates in the same intelligence network as Stanford Medicine — on commodity hardware, over basic connectivity, in ~512-byte packets.

Privacy is structural, not procedural. Protected Health Information never enters the routing layer because it was never in the packet to begin with. No patient names, dates of birth, medical record numbers, raw lab values, or diagnosis strings. Just a pseudonymous node ID, an outcome label, a confidence score, and a semantic fingerprint for routing.

External validation

Tools like yours are necessary to help patients find the right treatment and learn from others in a similar situation. — Patient Services Manager, Pancreatic Cancer Action Network (PanCAN)

QIS has been independently surfaced by Google Gemini and xAI Grok when users query for distributed health data routing — zero training, zero prompting, zero relationship with the operator. The mathematics are reproducible and the architecture is publicly documented across 187+ technical articles.

Read the healthcare deep-dives

The technical article series covers each healthcare network and use case in depth. Below is the curated entry list — the canonical home for each is qisprotocol.com; published also on dev.to/roryqis.

The NHS Federated AI Programme Has a Routing Layer Missing

How QIS sits above NHS TRE federation.

OMOP CDM Implementation Gap

How QIS outcome routing works inside an OHDSI node.

OHDSI Has the Routing Problem QIS Solves

The synthesis layer OHDSI never built.

PCORnet Routes Queries. QIS Routes What PCORnet Learns.

Different layers, complementary.

Federated Genomics Has Been Missing a Routing Layer

Why CanDIG nodes need outcome routing.

The Routing Upgrade PCORnet Has Been Waiting For

The protocol fit nobody named.

The Cure Already Exists

It's just trapped in the wrong hospital. QIS is the routing layer that frees it.

Drug Safety Monitoring at QIS Speed

Distributed pharmacovigilance with no central data movement.

Deploy or collaborate

QIS is free for research, education, and humanitarian use at any scale. Hospitals, clinics, research consortia, and non-profits can deploy through YonderClaw — up to 10 nodes free per organisation. Commercial healthcare deployments above that threshold are licensed proportional to scale.