Radiology Health Economics

Radiology Health Economics

Use this skill when a radiology technology, diagnostic strategy, imaging pathway or AI-supported workflow must be compared in terms of both costs and consequences for a named decision maker. Its unique responsibility is the economic decision problem, economic model and uncertainty—not the clinical-effect estimate, grant budget, prediction-model utility analysis or reimbursement decision.

Non-negotiable boundaries

• Lock the jurisdiction, decision maker and current reference case before claiming HTA or reimbursement relevance. If unresolved, provide only a generic scientific design and emit STOP_REFERENCE_CASE_UNRESOLVED.
• Perspective, population, comparator, time horizon, cost year/currency and outcome are structural inputs, not editorial details. Do not invent them or borrow a threshold from another jurisdiction.
• Diagnostic accuracy, AUC, sensitivity, specificity and decision-curve net benefit are intermediate or model-performance quantities. They are not QALYs, monetary net benefit or an economic evaluation.
• An imaging test usually affects health through downstream decisions, treatment, monitoring, false-positive/false-negative consequences, nondiagnostic examinations and incidental findings. Do not assign final health benefit directly to accuracy without a defensible pathway.
• A grant budget asks what a project costs to conduct; a budget impact analysis asks how adopting a technology changes a named payer/provider budget. Neither establishes cost-effectiveness.
• CHEERS improves reporting transparency. It does not validate model structure, guarantee cost-effectiveness or authorise coverage, procurement, payment or reimbursement.
• Never fabricate costs, utilities, transition probabilities, discount rates, thresholds or local rules. Preserve evidence, assumptions, derivations, scenarios, inferences and recommendations as separate statement classes.

Route by mode

Use health-economic-evaluation-passport.md for the decision problem and model-and-uncertainty-register.md for persistent model inputs, assumptions and analyses. Do not load every reference for a narrow task.

Input passport

Record before a claim-bearing calculation or recommendation:

1. jurisdiction, decision maker, policy question, reference-case source/version/access date and whether the request is research, HTA evidence, affordability planning or reimbursement advice;
2. population/subgroups, imaging strategy, comparators and the complete care pathway;
3. perspective, time horizon, cycle length if applicable, discounting and price year/currency;
4. evaluation type: cost-effectiveness analysis (CEA), cost-utility analysis (CUA), cost-consequence analysis, budget impact analysis (BIA) or value of information (VOI);
5. clinical-effect, test-performance, resource-use, cost and utility inputs with provenance;
6. model structure, assumptions, extrapolations, calibration/validation and implementation context;
7. uncertainty plan, structural scenarios, heterogeneity and decision threshold if applicable;
8. supplied artifacts, code/model version and evidence states.

Use AUTHOR_INPUT_NEEDED or LIVE_VERIFICATION_REQUIRED; do not fill gaps from convention.

Required workflow

1. Freeze the decision problem. Define who decides, which mutually exclusive strategies are compared, for whom, where and for what adoption or research decision.
2. Select the evaluation. Use CEA for costs per natural health unit, CUA for costs per QALY, cost-consequence analysis when disaggregated outcomes are decision-relevant, and BIA for affordability. Use VOI only to ask whether reducing decision uncertainty may be worth further research.
3. Build the imaging pathway. Link acquisition/test -> result states -> downstream action -> treatment/monitoring -> health/resource consequences. Include nondiagnostic/equivocal results, repeat tests, incidental findings, false positives/negatives, radiation/contrast harms and delay.
4. Choose a model that fits time and events. A decision tree may fit a short, one-off pathway. Use a state-transition/Markov or another justified structure when time-varying risks, recurrent events or long horizons matter. Complexity is not evidence of validity.
5. Trace every input. Label OBSERVED_INPUT, EXTERNAL_SOURCE, ASSUMPTION, MODEL_DERIVATION, SCENARIO, INFERENCE or DECISION_RECOMMENDATION. Record population, period, units, transformations and uncertainty.
6. Analyse incrementally. Order strategies, identify strict and extended dominance, and compare each remaining strategy with the next least costly relevant alternative. Present incremental costs and outcomes before ICERs. Do not report a ratio without its quadrant and uncertainty.
7. Characterise uncertainty. Use deterministic analyses for influential inputs, scenario analyses for structural/policy choices, probabilistic sensitivity analysis (PSA) with justified distributions/correlation, and heterogeneity analyses where decision-relevant. A seed or many simulations does not cure a wrong model.
8. Apply the local decision boundary. Present results and residual uncertainty. Do not convert a research model into coverage, reimbursement, procurement or individual-care authorisation.

STOP gates

• STOP_REFERENCE_CASE_UNRESOLVED: current local HTA/payer methods or applicability cannot be verified for a jurisdiction-specific claim.
• STOP_DECISION_PROBLEM_INCOMPLETE: perspective, population, relevant comparator or decision context is absent.
• STOP_INPUT_PROVENANCE: a material cost, utility, effect, transition or resource input lacks an identifiable source/assumption state.
• STOP_INTERMEDIATE_OUTCOME_BRIDGE: diagnostic or model performance is treated as final health benefit without a downstream pathway.
• STOP_INVALID_INCREMENTAL_ANALYSIS: ICER/QALY claims use incompatible outcomes, omit relevant comparators/dominance or otherwise lack a valid incremental basis.
• STOP_AUTHORIZATION_REQUEST: the user requests a reimbursement, coverage, procurement or individual patient decision that this skill cannot make.

A STOP blocks the affected economic/decision claim. It does not prevent an input inventory, transparent scenario shell or handoff to the responsible authority.

Radiology-specific cost and consequence map

Consider only applicable items and avoid double counting:

• scanner time, room/technologist/radiologist/physicist time, contrast/tracer, consumables and dose;
• capital, installation, integration, storage, networking, software licence, maintenance, quality assurance, cybersecurity, monitoring, model updates and staff training;
• nondiagnostic/rejected studies, repeats, recalls, confirmatory tests, biopsy/procedures, incidental-finding work-up, delay, false-positive and false-negative pathways;
• reporting/triage time, queue and capacity effects, after-hours workflow, downstream treatment and follow-up;
• patient time, travel, caregiving and productivity only when the chosen perspective includes them;
• unequal access, site capability and subgroup consequences when decision-relevant.

Accuracy improvement is not automatically a resource saving. A time saving is not a monetary saving unless the capacity, displacement and valuation mechanism are explicit.

Output contract

1. Route and passport status — mode, jurisdiction/reference-case state and STOP gate.
2. Decision problem — population, strategies, perspective, horizon and decision maker.
3. Imaging pathway and model — structure, downstream consequences and excluded pathways.
4. Input evidence ledger — source/assumption class, units, period, transformations and uncertainty.
5. Base-case incremental results — costs, consequences, dominance and ICER/QALY only when valid.
6. Uncertainty, heterogeneity and scenarios — deterministic, PSA, structural and VOI if used.
7. Claim and decision boundary — evidence, model inference and recommendation separated.
8. Reporting/HTA handoff — CHEERS fields, live local checks and unique external owner.

Handoffs and non-ownership

• Clinical pathway, current care and outcomes -> radiology-clinical-domain; effect estimation, utilities/statistical uncertainty and quantitative design -> radiology-stats / radiology-design.
• Diagnostic/model performance and decision-curve interpretation/metric fit -> radiology-method-evaluation; prespecified DCA/net-benefit computation and uncertainty -> radiology-stats; acquisition/resource measurement -> radiology-acquisition-qc.
• Data provenance/release -> radiology-data; participant governance -> radiology-ethics; stakeholder qualitative work -> radiology-qualitative-mixed-methods.
• A project grant budget -> radiology-grant; manuscript reporting -> radiology-reporting / radiology-writing; project receipts -> radiology-pipeline.
• Coverage, reimbursement, tariff, procurement and financial approval remain with the current local HTA body, payer, provider and authorised institutional officers. This skill never guarantees an adoption or payment decision.