Radiology Consensus and Guideline Development
Use this skill to design, run-plan or audit a consensus statement, appropriateness project, practice parameter, technical standard or clinical practice guideline. It owns the panel/process, evidence-to-decision record, consensus protocol, recommendation ledger and lifecycle plan. It does not perform the underlying systematic review, confer society endorsement, set a legal standard of care or let expert voting substitute for evidence.
Choose the product and mode
First declare the product: clinical-practice-guideline, appropriateness-criteria,
practice-parameter, technical-standard, expert-consensus-statement or
research-priority/definition-consensus. Do not use these labels interchangeably.
| Mode | Use when | Required result |
|---|---|---|
scoping | product, users, questions or authority are unsettled | scope, product decision, stakeholders, evidence needs and stop conditions |
protocol | a versioned development plan is needed before evidence/panel decisions | panel, COI, evidence, consensus, consultation and update protocol |
evidence-to-decision | a frozen evidence synthesis must become recommendation judgments | GRADE/compatible EtD ledger with explicit uncertainty |
formal-consensus | Delphi, RAND/UCLA or nominal group work is planned or audited | prespecified rounds, thresholds, denominator, feedback, dissent and reporting record |
audit-update-retire | an existing statement/guideline is reviewed for trustworthiness or currency | method audit, surveillance result, update/retirement and supersession map |
Read references/evidence-to-recommendation.md for the systematic-review handoff, GRADE/EtD and recommendation wording. Read references/formal-consensus-methods.md only for a formal consensus exercise. Read references/radiology-guidance-lifecycle.md for appropriateness criteria, practice parameters, technical standards, implementation, updating and retirement. Verify methods and authority through references/source-registry.md.
Input passport
Record:
product and mode | sponsor/developer and actual adoption authority | jurisdiction/organization | scope, population, modality/test/intervention/comparator/outcomes/setting | intended users and affected people | panel roles, patient/public members and COI records | frozen systematic-review handoff/version/currency | certainty framework | consensus method and prespecified rule | draft and consultation artifacts | implementation context | publication/update/retirement owner | user authority for edits or external communication.
Missing adoption authority, panel decisions, COI management, patient/public participation, evidence
certainty or official status remains NOT_VERIFIED. Advisory drafting is not an endorsed guideline.
Choose the evidence lane for the declared product. A clinical recommendation presented as
evidence-based requires the relevant systematic-review/EtD handoff. A research-priority or
definition-consensus exercise instead requires transparent item generation, search/mapping or
stakeholder evidence, panel judgments and limits; use NOT_APPLICABLE_GRADE when it makes no
clinical-effect or recommendation-strength claim. A generic Delphi protocol does not require an
invented effect synthesis or GRADE rating.
Core evidence and recommendation states
- Evidence handoff:
EVIDENCE_READY,EVIDENCE_PARTIAL,EVIDENCE_ABSENTorEVIDENCE_STALE. - Recommendation:
DRAFT,CONDITIONAL,STRONG,NO_RECOMMENDATION,RESEARCH_ONLY,SUPERSEDEDorRETIREDonly when defined by the selected framework and authorized process. - Consensus result:
CONSENSUS_REACHED,CONSENSUS_NOT_REACHED,STABLE_DISAGREEMENT,ATTRITION_LIMITEDorNOT_ASSESSABLE.
Evidence certainty, recommendation strength and consensus percentage are different fields. A high vote does not upgrade low/absent evidence, and a low-certainty body does not automatically prohibit a strong recommendation when a transparent framework supports an exceptional case.
Required workflow
- Lock the product and authority. Define who will use the output, for which decision, and who can officially adopt/endorse it. Use “draft” or “advisory” until a verified process confers status.
- Scope the decision. Specify population, imaging scenario, modality/test, comparator, downstream action, patient-important outcomes, setting, exclusions and equity/access questions.
- Constitute the panel. Include relevant clinical disciplines, radiologists, referrers, technologists/medical physicists when technical decisions are material, methodologists, implementers and patients/public or other affected representatives. Record balance, recruitment, roles, COI assessment, management and recusals.
- Accept the product-appropriate evidence handoff. For an evidence-based clinical
recommendation, require search currency, study-family/effect-row identity, risk-of-bias/
applicability, synthesis and certainty from
radiology-systematic-review. For priorities or definitions, record item sources, mapping/search limits, stakeholder input and traceable revision; a full effect review is required only if the intended claim needs it. Do not relabel selected references or panel memory as a systematic review. - Run the applicable decision process. For clinical recommendations, record benefits/harms, certainty, values/preferences, resources, equity, acceptability and feasibility. Link diagnostic accuracy to downstream management and patient outcomes. For priorities or definitions, report explicit selection criteria, content/semantic judgments and dissent without inventing clinical effects, GRADE certainty or recommendation strength.
- Use formal consensus only for its declared purpose. Prespecify method, panel, rounds, anonymity/interactions, feedback, denominator, threshold, stability, attrition, tie/disagreement handling and stopping. Use templates/consensus-guideline-protocol.md.
- Draft atomic recommendations or consensus items. Each clinical recommendation states population/scenario, action, comparator/conditions, strength, certainty, rationale, harms/burden, implementation conditions, exceptions, dissent and research gaps. Use templates/recommendation-etd-ledger.md. For priorities/definitions, use the item register and formal-consensus result sections instead; include wording, provenance, scope, rationale and status, with clinical EtD fields not applicable.
- Consult and quality-control. Preserve stakeholder comments, responses, panel changes, unresolved disagreement and external review. Reporting-checklist completion does not prove method quality or endorsement.
- Plan implementation and lifecycle. Define dissemination, feasibility tools, monitoring, surveillance, update triggers, versioning, supersession and retirement. Imaging technology, contrast/tracers, dose, reconstruction, AI versions and clinical standards can make guidance stale.
Stop and escalation gates
Return GUIDANCE_STOP_AND_REFRAME when product or adoption authority is misrepresented; scope or
decision scenarios are undefined; the evidence packet is missing/stale for a claim presented as
evidence-based; material panel COI cannot be managed; affected patients/public or essential
disciplines are systematically excluded without justification; consensus thresholds are chosen after
results; attrition, dissent or denominator is hidden; expert votes are used to manufacture evidence;
or a current clinical recommendation could expose patients to unreviewed harm.
Escalate formal endorsement, legal/regulatory interpretation and individual patient decisions to the responsible society, institution, regulator or clinical team. The skill may prepare evidence and drafts but cannot adopt, endorse or implement official guidance.
Output contract
Product, scope and authority passportPanel/stakeholder/patient-public and COI matrixProduct-appropriate evidence handoff and currency verdict— systematic review for the clinical evidence claim, or item-generation/mapping/stakeholder evidence for priorities and definitionsQuestion/outcome and imaging-scenario mapGRADE or compatible evidence-to-decision ledgerwhen applicable; otherwiseNOT_APPLICABLE_GRADEplus the item-selection/definition rationaleConsensus protocol and result— rounds, thresholds, denominator, attrition and dissentRecommendation or consensus-item ledger— strength, certainty and consensus kept separate; clinical strength/certainty is not invented for a terminology or research-priority itemImplementation, equity, monitoring and research-gap planConsultation, external-review and change-response recordVersion, update, supersession and retirement plan with exact non-endorsement boundary
Handoffs and non-ownership
- Search, screening, extraction, risk of bias, synthesis and evidence certainty ->
radiology-systematic-review; current clinical facts ->radiology-clinical-domain. - Quantitative Delphi/RAND agreement or comparative inference ->
radiology-stats; CEA/CUA, budget impact, ICER/QALY, PSA/VOI and economic-model evidence consumed by an EtD ->radiology-health-economics. This skill weighs the supplied economic evidence in a recommendation; it does not build or authorize the economic model. - Reporting guideline/quality checklist ->
radiology-reporting; prose, tables, figures and decks -> their artifact owners. - Society boards, institutions, regulators and clinical teams retain endorsement, adoption, implementation, legal and patient-care authority.
