{"schemaNamespace":"https://mc.chai.org/v0.2/schema.xsd","schemaVersion":"v0.2","schemaPath":"docs/chai/applied-model-card-v0.2.xsd","voluntarySentence":"ARKA is not a certified Health IT Module and is not required to publish these attributes. Here they are.","noCertificationSentence":"This crosswalk does not claim or imply certification, accreditation, or endorsement by ONC, CHAI, or the Joint Commission.","BasicInfo":{"ModelName":"ARKA clinical decision support (appropriateness at order entry)","ModelDeveloper":"ARKA Health","DeveloperContact":"https://arka.health/trust · trust@arka.health"},"ReleaseInfo":{"ReleaseStage":"FDA Non-Device CDS under FD&C Act §520(o)(1)(E); not a certified Health IT Module","ReleaseDate":"23 August 2026","ReleaseVersion":"1.2.0","GlobalAvailability":"United States (voluntary publication; customer deployments by agreement)","RegulatoryApproval":"ARKA is not a certified Health IT Module and is not required to publish these attributes. Here they are. This page does not claim or imply certification, conformance, or endorsement by ONC, the Joint Commission, or CHAI."},"ModelSummary":{"Summary":"Transparency Commitments for Software That Influences Clinical Ordering, v1.2 — published by ARKA Health (v1.2.0) constrains ARKA's clinical decision support: GRADE-certainty ratings with published distribution, zero payer-authored criteria on the appropriateness path, never auto-deny, locked pre-period, public corrections, justification firewall, determinism with no LLM clinical prose, open research artefacts, published interrupt burden, independent evidence panel (committed), calibration and drift publishing (committed), and published abstention. The deterministic knowledge matrix and coverage runtime are evidence-based decision support. The pretest/post-test probability model in lib/aiie-v3/ is predictive. Each attribute below names which component(s) it applies to.","Keywords":{"Keyword":["clinical decision support","appropriateness","GRADE","HTI-1","non-device CDS"]}},"UsesAndDirections":{"IntendedUseAndWorkflow":"Surfaces patient-specific imaging and related appropriateness context at order entry from structured FHIR to support clinician judgment. Does not analyze medical images or physiological signals. Does not block, cancel, or place orders.","PrimaryIntendedUsers":"Ordering clinicians and health-system AI / utilisation governance committees.","HowToUse":"CDS Hooks card at order-select; clinician retains final decision. Every card ends with the FDA Non-Device CDS disclosure.","TargetedPatientPopulation":"Adults in ambulatory and inpatient settings where advanced imaging or related orders are placed (see published measure and coverage scopes).","CautionedOutOfScopeSettings":"Not for automatic denial, utilisation targets, image diagnosis, or any use that replaces clinician judgment. Not a certified Health IT Module source-attribute filing."},"Warnings":{"KnownRisksAndLimitations":"Recommendations carry GRADE certainty including low and very-low cells; the system publishes when it abstains. Low-certainty ratings are scheduled for independent panel review. See /evidence/methodology and /docs/model-limitations.","KnownBiasesOrEthicalConsiderations":"Payer-authored criteria are firewalled from clinical appropriateness. Justification text is never joined to performance measures. Certainty census and abstention rate are published, including the thin end.","ClinicalRiskLevel":"FDA Non-Device CDS (supports, does not replace, clinical judgment). Not a medical device."},"TrustIngredients":{"AISystemFacts":{"OutcomesAndOutputs":"Appropriateness / concordance signals with GRADE certainty and COI class; optional pretest/post-test probability from the predictive component. Human-readable card copy; no LLM-authored clinical prose.","ModelType":"Deterministic evidence-based knowledge matrix and coverage runtime, plus a glass-box pretest/post-test probability model. Identical inputs regenerate identical outputs.","FoundationModels":"None. No foundation model writes clinical or patient-facing prose.","InputDataSource":"Structured FHIR resources supplied by the ordering environment; synthetic fixtures for public demos. No PHI in public artefacts.","OutputAndInputDataTypes":"Inputs: structured clinical elements with provenance. Outputs: CDS cards, peer packets (aggregates), and published monitoring artefacts. Money as integer cents; rates as numerator/denominator.","DevelopmentDataCharacterization":"Evidence ratings cite published guidelines and literature registered in the Source Register. Predictive component characterisation is published under the HTI-1 source-attribute sheet; unanswered attributes are marked unanswered with reasons.","BiasMitigationApproaches":"Appropriateness firewall (zero payer-authored anchors on the clinical path); COI class on every rating; small-cell suppression (n < 11); no clinician leaderboard; published certainty distribution.","OngoingMaintenance":"Versioned Standard and DSI attribute sheet; public corrections with fixing commit; committed calibration/drift and alert-burden reports. Schema: docs/chai/applied-model-card-v0.2.xsd","Security":"HIPAA program; PHI hashed at ingestion boundaries for TCOC; no justification text stored by ARKA-AJ. See /security.","Transparency":"ARKA is not a certified Health IT Module and is not required to publish these attributes. Here they are. Full HTI-1 source attributes at /docs/model-card#dsi-source-attributes; Standard at /trust#arka-standard."},"TransparencyInformation":{"FundingSource":"ARKA Health (self-funded product development).","StakeholdersConsulted":"Independent evidence panel charter published; roster and decisions public when the panel is live (Standard clause 10, committed)."}},"KeyMetrics":{"UsefulnessUsabilityEfficacy":{"MetricGoal":"Publish measured effect against a locked pre-period analysis plan.","Results":{"MetricResult":[{"Label":"Not yet published","ValueJson":"{\"kind\":\"empty\",\"reason\":\"Prospective measured effect is not yet published; the analysis plan and baseline lock artefacts are public at /baseline. Empty field preferred to a stretch claim.\"}"}]},"Interpretation":"Prospective measured effect is not yet published; the analysis plan and baseline lock artefacts are public at /baseline. Empty field preferred to a stretch claim.","TestType":"Not yet published","TestingDataDescription":"Prospective measured effect is not yet published; the analysis plan and baseline lock artefacts are public at /baseline. Empty field preferred to a stretch claim.","ValidationProcessAndJustification":"Prospective measured effect is not yet published; the analysis plan and baseline lock artefacts are public at /baseline. Empty field preferred to a stretch claim."},"FairnessEquity":{"MetricGoal":"Quantitative fairness metrics on local and external data.","Results":{"MetricResult":[{"Label":"Not yet published","ValueJson":"{\"kind\":\"empty\",\"reason\":\"HTI-1 attribute 'Quantitative performance measures for fairness in test and external data' is unanswered with a published reason on the DSI sheet. Empty field preferred to a stretch claim.\"}"}]},"Interpretation":"HTI-1 attribute 'Quantitative performance measures for fairness in test and external data' is unanswered with a published reason on the DSI sheet. Empty field preferred to a stretch claim.","TestType":"Not yet published","TestingDataDescription":"HTI-1 attribute 'Quantitative performance measures for fairness in test and external data' is unanswered with a published reason on the DSI sheet. Empty field preferred to a stretch claim.","ValidationProcessAndJustification":"HTI-1 attribute 'Quantitative performance measures for fairness in test and external data' is unanswered with a published reason on the DSI sheet. Empty field preferred to a stretch claim."},"SafetyReliability":{"MetricGoal":"Publish interrupt burden, abstention, and calibration including when they worsen.","Results":{"MetricResult":[{"Label":"Standard clause coverage","ValueJson":"{\"kind\":\"single\",\"value\":12,\"unit\":\"clauses\"}"},{"Label":"FDA disclosure on every card","ValueJson":"{\"kind\":\"text\",\"value\":\"This recommendation is intended to support, not replace, clinical judgment. It is generated by ARKA, software designed t…\"}"}]},"Interpretation":"Safety commitments are encoded as The ARKA Standard with named checks; alert burden, panel, and calibration clauses are committed with target dates where not yet live.","TestType":"Published artefact + automated CHECK-* gates","TestingDataDescription":"CI lints (lint:standard, lint:dsi-attributes, lint:sources) against sealed fixtures.","ValidationProcessAndJustification":"See docs/arka-standard.json and docs/dsi-source-attributes.json."}},"Resources":{"PeerReviewedPublications":"Open research artefacts with resolvable identifiers before revenue — e.g. https://doi.org/10.5281/zenodo.22051812.","PatientConsentOrDisclosure":"This recommendation is intended to support, not replace, clinical judgment. It is generated by ARKA, software designed to meet the four criteria for Non-Device Clinical Decision Support under FD&C Act §520(o)(1)(E) and FDA's final guidance on Clinical Decision Support Software (January 2026). The clinician is responsible for the final decision.","EvaluationReferences":"HTI-1 source attributes: /docs/model-card; Standard: /trust#arka-standard; CHAI schema: docs/chai/applied-model-card-v0.2.xsd"},"Bibliography":"@misc{arka-standard-2026,\n  title = {Transparency Commitments for Software That Influences Clinical Ordering, v1.2 — published by ARKA Health},\n  author = {{ARKA Health}},\n  year = {2026},\n  howpublished = {\\url{https://arka.health/trust\\#arka-standard}},\n  note = {v1.2.0, published 2026-08-23}\n}"}