Governing the moment between detection and action.
ThresholdSignal Inc. builds the Verification-Action Readiness Engine (VARE), an evidence-centred decision-governance framework that tells public health authorities when the evidence is sufficient to act, ensures that action is equitable, and leaves an auditable record of why.
Detection systems are increasingly good at telling us what might be happening. Emerging equity frameworks tell us who should benefit when we act. VARE governs the decision in between: when the evidence is sufficient to act, and how to ensure that action is fair.
Surveillance tells you something is happening. It rarely tells you whether to act, or who should benefit when you do.
Evidence-weighted scoring
VARE assigns confidence scores to incoming signal clusters across multiple independent evidence streams, and works even without laboratory confirmation.
Auditable decision trail
Every gate assessment (the evidence bundle, the threshold, the recommendation) is recorded immutably, giving health authorities a defensible record of why they acted.
Equitable & sovereign by design
VARE runs inside the host institution’s own infrastructure and governance boundaries, so no data leaves the system it was generated in, and it links escalation to equitable benefit-sharing obligations.
Closing the gap between evidence and equitable action.
ThresholdSignal Inc. researches, develops, and shares decision-support tools and governance frameworks that strengthen early detection of, and equitable, accountable response to, infectious disease outbreaks and other public health emergencies.
We are organized exclusively for charitable, educational, and scientific purposes. No part of our work is designed to benefit any private individual, and our methodology is built to be shared, not locked away.
Two compounding gaps
Modern epidemic intelligence (the WHO’s EIOS, AI-driven forecasters, and large-scale genomic and environmental monitoring) is increasingly effective at answering one question: is something happening? It is far less effective at the two questions that follow, and those two gaps compound one another.
The decision-governance gap
Outbreak after-action reviews, from COVID-19 to mpox to recent filovirus events, describe the same failure: credible signals are visible weeks before formal action, not because evidence is missing, but because no structured mechanism converts mixed-reliability signals into a defensible, threshold-crossing decision. The result is either over-reaction to weak signals, or paralysis for fear of acting without justification.
The equity-enforcement gap
Countries that detect and report novel pathogens earliest have historically been punished with travel bans and trade restrictions, while wealthier nations secured disproportionate access to the countermeasures derived from shared data. Frameworks meant to correct this rely on voluntary participation, with no binding way to ensure that action, once taken, benefits everyone fairly.
That gap costs time and trust. In an outbreak, time is the resource that cannot be recovered, and equity is the condition on which future data-sharing depends.
How we work
What we build
A decision-governance layer that sits above existing biosurveillance and data-integration systems. It is not a replacement for them, but the missing layer that turns integrated signals into accountable, equitable action, with a full audit trail.
In the open, with the field
Methodology and code developed under grant-funded work are published under an open license and made freely available to national and regional public health authorities. Decisions stay human-led; VARE does not automate attribution, mandate action, or replace expert judgement.
The Verification-Action Readiness Engine
VARE is a confidence-scoring and decision-governance framework for outbreak response. It does not replace surveillance; it governs what happens after a signal appears.
Four evidence streams, one score
VARE ingests outputs from existing national and international biosurveillance systems and fuses them into a single, evidence-weighted confidence score for each active monitoring context.
The Evidence Graph
At the core of VARE is an Evidence Graph, a continuously updated decision case file. Think of it as a courtroom docket, where claims are logged, evidence is attached, credibility is assessed, and uncertainty is made explicit; or as a forensic audit, where data is checked for provenance and manipulation before conclusions are drawn. Each claim is evaluated on four dimensions, and confidence ranges narrow or widen as corroborating or conflicting evidence arrives (probabilistic, Bayesian-style updating rather than false precision).
Evidence quality
Source reliability, timeliness, and corroboration across independent reports.
Provenance
How and where data was generated and transmitted, the chain of custody behind each signal.
Manipulation risk
Indicators of synthetic, misleading, or adversarially injected data.
Bounded uncertainty
Confidence expressed as probability ranges, not single point estimates.
From signal to defensible decision
Low-regret vs. high-regret action
VARE’s gate architecture is built on a regret-asymmetry principle: the cost of acting early on a false alarm is weighed explicitly against the cost of acting late on a true positive. Low-regret actions (enhanced surveillance, targeted testing, internal audits) can proceed at a lower confidence threshold (illustratively ~30%), enabling precautionary response without panic. High-regret actions (public alerts, facility closures, attribution) require a higher bar (illustratively ~75%) and mandatory human-in-the-loop review before the gate opens.
The equity gate: benefit-sharing, enforced at escalation
What distinguishes VARE from a standalone decision engine is that equity is not left to goodwill. A governance layer aligned with the WHO Pathogen Access and Benefit-Sharing (PABS) framework can be embedded directly in the escalation workflow: verifying participation, identifying the shared pathogen data that informed a decision, generating structured evidence packages suitable for formal declaration, and logging every step in a tamper-evident audit trail. The principle is simple: participation in equitable benefit-sharing becomes the condition for receiving the most valuable output of the surveillance system, namely timely, verified, evidence-gated intelligence.
Biosecurity, built into the verification layer
VARE integrates safeguards rather than bolting them on: statistical anomaly detection and provenance-chain analysis to flag synthetic or manipulated data, insider-threat and misuse pattern recognition across access logs, and capability gating to prevent synthesis-enabling outputs. VARE does not automate attribution, mandate action, or replace human judgement; human oversight is mandatory for every high-regret decision.
Where VARE fits
VARE does not compete with detection systems; it completes them, adding the escalation-and-equity layer above the data.
Detection systems ask
Is something happening? Where might risk emerge? What biological signals exist? They excel here, and stop here.
VARE asks
Is the evidence sufficient to act, and when we act, will the benefits be shared equitably? Evidence-gated, sovereign, and fully auditable.
Case study: Bundibugyo Virus Disease, Ituri Province, DRC
In May 2026, a Bundibugyo Virus Disease outbreak was declared in Ituri Province, DRC. We selected it deliberately as a hard test case: access to the affected area was constrained, health infrastructure was degraded, and the pathogen strain was initially misclassified, generating exactly the kind of noisy, contested signal environment VARE is designed to operate in.
Our retrospective analysis applied VARE’s scoring and gate logic to the signal data available in the weeks before the outbreak was officially declared. Under those adverse conditions, VARE would have crossed the low-regret action threshold four to six weeks before the 15 May declaration. That is evidence the framework holds up on the hardest cases, not only on clean, textbook outbreaks.
A live surveillance view of this outbreak is maintained at epivare.org.
Built to complement, not replace
VARE is designed to sit on top of existing infrastructure (national surveillance systems, the WHO’s EIOS, regional data repositories such as the Africa CDC Central Data Repository) rather than compete with it. Where those systems focus on detecting and integrating signals, VARE focuses on the decision layer above the data: converting integrated signals into a formal, auditable, threshold-crossing recommendation. It is designed for data-sovereignty compatibility, meaning it can run within a health authority’s own infrastructure and governance boundaries without data ever leaving that system.
How it’s built
VARE is developed on a modern, sovereign-compatible stack: an interactive R Shiny decision dashboard with integrated modules (executive overview, evidence accumulation, manipulation detection, geographic surveillance, action readiness), visualised with Plotly, Leaflet, and DT; a Microsoft SQL Server backend running a Bayesian-inspired confidence engine; and deployment either on-premise or in secure cloud, with federated learning for multi-node coordination. A live demonstrator applying this stack to a current outbreak is maintained at epivare.org.
What a first pilot aims to show
Our 12-month proof-of-concept pilot is scoped around independently evaluable targets, benchmarked against documented decision delays in past outbreak after-action reviews:
We're just getting started.
ThresholdSignal Inc. will continue to share with you our pilots and published methodology as we move forward and detect signals of impending outbreak.
- BVD 2026 retrospective analysis completedProof-of-concept validation of the VARE gate architecture against a real 2026 outbreak.
- First pilot partnershipIn active conversations with public health agencies and research partners.
No published updates yet
Want to be notified the moment we post something real? Reach out and we’ll keep you posted.
Partner with us.
We're a small nonprofit team building VARE in the open. We're especially interested in hearing from the people closest to the problem.