Healthcare and AI, in plain language.
Care, provider trust, shared governance and healthcare AI — explained in plain language.
The protocol
- Codex
- VANTHEON’s proposed system for qualified healthcare knowledge, care decisions, coordination and learning. Participants propose, evaluate and vote on improvements; authorized changes are monitored and can be revised or rolled back.
- Provider Trust IndexPTI
- A service-specific evaluation of human and machine providers. Better validated decisions and outcomes are intended to support better incentives and greater voting weight on relevant Codex proposals. Patient representation is separate.
- Attribution
- Connecting care work to its contributors, patient context and outcome evidence, including uncertainty and corrections. These links support learning and contribution credit under agreed rules; they do not by themselves prove what caused an outcome.
- Learning health system
- A system that uses experience from care to evaluate and adopt improvements, then checks their effects on people. VANTHEON connects that cycle to governance, provider incentives and authorized model research.
- Trust-Time
- A proposed commitment to a defined service within a time window. Capacity, scope, authority and the provider’s obligations must all be specified.
- Treasury
- Proposed financial governance and execution for authorized payments, reconciliation and remedies. Restricted network funds support development and shared research; care funds and stablecoin backing retain separate obligations.
- VVAL and VHTH
- The proposed payment stablecoin and network token. Neither has been issued. Holding either does not itself confer clinical authority.
- Verifiable healthcare
- Care whose need, authority, work and observed results can be checked against defined evidence. Verification does not by itself establish that an intervention caused better health.
- Verified event
- A record of specified care work and its supporting evidence, applicable rules, permissions, observed results and responsible participants, checked within a defined scope. Identifying missing care also requires knowing what was due and who was eligible.
- Computed trust
- A service-specific estimate from qualified evidence, with uncertainty. Risk adjustment and calibration must be validated for the population and use.
- Instant authorization
- A proposed approval path when guideline, evidence, trust and authority checks all hold. Other cases go to human review; only a person can deny care.
- The pool
- The proposed funding arrangement for qualified care obligations, with distinct rules for capital, costs, reserves and distributions.Members, depositors and producers of care participate under defined terms for each role.
- Surplus
- Value potentially available after care liabilities, costs and reserve requirements are accounted for. Distribution depends on adopted terms and can increase future shortfall risk.
- Zero-knowledge proof
- A way to prove a specified statement, such as a computed score meeting a threshold, without revealing the private data behind it. The proof does not establish clinical trustworthiness.
- Producer of care
- A clinician, clinic, AI agent or robot whose care work is supported by evidence and recognized under the relevant rules.
Care & coverage
- Prior authorization
- Approval a health plan requires before some tests, procedures or medicines, so that it agrees to pay.An MRI may need approval before it is scheduled.
- Health plan (payer)
- The insurer or program that pays for care — a company, Medicare, or Medicaid.
- Electronic health recordEHR
- The system a hospital or clinic uses to keep a patient’s visits, results and orders.
- Quality measure
- A standard check of whether patients got recommended care.A yearly eye exam for people with diabetes.
- HEDIS
- A widely used set of quality measures for health plans, maintained by NCQA.
- Value-based care
- Paying for health results, such as fewer complications, instead of only for each visit or test.
- Care gap
- Recommended care a patient still needs to receive.
- CMS
- The U.S. Centers for Medicare & Medicaid Services — the federal agency that runs Medicare and oversees Medicaid.
- FQHC
- A federally qualified health center: a community clinic that serves patients regardless of ability to pay.
Health data
- FHIR
- HL7’s standards framework for exchanging healthcare information, used internationally. It gives systems shared structures for records such as lab results.Pronounced “fire.”
- Interoperability
- The ability of different health systems to exchange records and understand them.
- TEFCA
- A U.S. framework that lets approved networks exchange health records nationwide.
- Individual Access ServicesIAS
- Under TEFCA, services that let a person request their own records from across the network.
- Consent
- A person’s permission for their information to be used — for a stated purpose and time.
- Source record
- Where a piece of information came from, who produced it and when.
- BAA
- A business associate agreement: a contract defining how protected health information may be handled when a HIPAA business-associate relationship requires one.
AI & verification
- THEON
- VANTHEON’s research program for healthcare foundation models and collaborating agents. Its first model and research-procedure studies are specified, not yet executed.
- Recursive self-improvement
- Research in which a system proposes and tests changes to its own research procedures. In THEON, independent evaluation, fixed budgets and human release authority constrain this process.
- VANTHEON Agent
- An agent product in development for care teams: carrying authorized context across tasks, building reusable skills from reviewed experience and coordinating work. It is distinct from THEON’s foundation-model research.
- AI model
- Software trained on large amounts of data to recognize patterns, answer questions or draft text.
- AI agent
- Software that uses AI models, tools and context to work toward a task under defined permissions—for example, finding sources or preparing a request.In VANTHEON’s design, each task has a purpose, limits and an accountable owner.
- Who decides
- Authority to take a particular action. People and machines have defined roles and permissions; consequential decisions retain an accountable owner and a path to human review.
- Human in the loop
- A person reviews or approves before an AI output becomes a decision.
- Verification
- Checking specified evidence, rules, authority and observed results. What a check establishes depends on its scope, sources and assumptions.
- Hallucination
- When an AI model states something that sounds right but is not supported by its sources. Verification exists to catch it.
- Settlement
- Determining and carrying out authorized obligations, including payment and reconciliation. A calculated amount, a payment instruction and confirmed funds movement are different states.
A term we missed?
Tell us what was unclear. We add to this glossary as we learn what people need explained.