Why English? | Reading CHEERS-AI Health Economic Evaluations Through Comparators, Costs, Learning, Uncertainty and Equity

ORIENTATION

What this guide owns

Current fact check

The primary CHEERS-AI publication reports a three-round Delphi process, an online consensus meeting and patient-group review. The final 38-item checklist retains 28 CHEERS 2022 items, adds eight AI-focused elaborations and includes ten new items numbered AI 1 to AI 10.

Canonical boundary

This article owns reporting of health economic evaluations for interventions that use AI. The GATHER article owns health-estimate reporting, STARE-HI owns health-informatics evaluation reporting, and clinical-effectiveness, safety, regulation and procurement decisions retain separate owners.

Worked example

Every chapter returns to a fictional hospital evaluating an AI-supported chest-image triage service against the existing clinician-led queue. It is fictional, low-stakes and educational; it does not report a real school, clinical decision or guaranteed outcome.

Warm up with eduKate’s guides to explaining scientific evidence and asking better questions in class.

CHAPTER 1 OF 16

1. Name the Evaluation and the AI Function

Back to contents

Why this move matters

Identify the report as an economic evaluation, name every strategy compared and state what the AI component actually does.

Translate the decision, not only the algorithm

Identify the report as an economic evaluation, name every strategy compared and state what the AI component actually does. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: The title states that AI triage is compared with the existing clinician-led queue. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

A product name does not describe its clinical function or economic role. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Name the Evaluation and the AI Function”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 2 OF 16

2. Define the Decision Problem and Intended Users

Back to contents

Why this move matters

State the policy or practice question, target population, setting, decision maker, intended users and reason the comparison matters now.

Translate the decision, not only the algorithm

State the policy or practice question, target population, setting, decision maker, intended users and reason the comparison matters now. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Hospital managers need to know whether triage changes urgent-care delay and resource use. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

A model can answer a numerical question that no real decision maker faces. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Define the Decision Problem and Intended Users”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 3 OF 16

3. Describe the Comparator Fairly

Back to contents

Why this move matters

Explain current care and other relevant alternatives with the same detail as the AI intervention, including workflow, staffing and technology.

Translate the decision, not only the algorithm

Explain current care and other relevant alternatives with the same detail as the AI intervention, including workflow, staffing and technology. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Existing clinician prioritisation is documented rather than treated as an empty baseline. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

An unrealistically weak comparator can make an AI option appear valuable. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Describe the Comparator Fairly”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 4 OF 16

4. Identify the AI Version, Purpose and Interaction

Back to contents

Why this move matters

Report technique, version, risk or purpose classification where relevant, static or adaptive status, users, interaction, extra requirements and expected advantage.

Translate the decision, not only the algorithm

Report technique, version, risk or purpose classification where relevant, static or adaptive status, users, interaction, extra requirements and expected advantage. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: The evaluation identifies a locked model version and how radiographers review its priority score. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

Results for one version may not travel to a later update. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Identify the AI Version, Purpose and Interaction”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 5 OF 16

5. Report User Autonomy and Decision Authority

Back to contents

Why this move matters

State whether the AI directs an action or supports a person who retains the decision, and describe overrides and escalation.

Translate the decision, not only the algorithm

State whether the AI directs an action or supports a person who retains the decision, and describe overrides and escalation. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Clinicians can override the triage score and urgent cases receive a second check. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

Human involvement does not automatically prevent automation bias. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Report User Autonomy and Decision Authority”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 6 OF 16

6. Choose Perspective and Time Horizon

Back to contents

Why this move matters

Name whether the analysis takes a health-system, payer, patient, societal or other perspective and justify how long costs and outcomes are followed.

Translate the decision, not only the algorithm

Name whether the analysis takes a health-system, payer, patient, societal or other perspective and justify how long costs and outcomes are followed. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: The hospital and societal perspectives capture staffing, travel and waiting consequences over five years. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

A short horizon may omit maintenance, retraining or delayed health effects. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Choose Perspective and Time Horizon”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 7 OF 16

7. Connect AI Performance to Health Outcomes

Back to contents

Why this move matters

Explain how accuracy, prioritisation or workflow effects become benefits or harms, and distinguish diagnostic, process and health outcomes.

Translate the decision, not only the algorithm

Explain how accuracy, prioritisation or workflow effects become benefits or harms, and distinguish diagnostic, process and health outcomes. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Faster prioritisation changes time to review, which may affect treatment and health rather than counting speed alone. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

A better technical metric is not automatically a better patient outcome. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Connect AI Performance to Health Outcomes”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 8 OF 16

8. Describe Evidence for the AI Effect

Back to contents

Why this move matters

Identify assessment studies, populations, settings, endpoints, reference standards and assumptions used to estimate the intervention effect.

Translate the decision, not only the algorithm

Identify assessment studies, populations, settings, endpoints, reference standards and assumptions used to estimate the intervention effect. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: External validation and a local workflow study supply different parts of the effect estimate. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

Evidence from retrospective accuracy data may not represent real-time clinical use. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Describe Evidence for the AI Effect”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 9 OF 16

9. Open the Development and Validation Record

Back to contents

Why this move matters

Describe training data, errors, bias checks and validation, or cite accessible reports that do so.

Translate the decision, not only the algorithm

Describe training data, errors, bias checks and validation, or cite accessible reports that do so. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: The report links cohort selection, missing data, subgroup checks and external testing. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

A citation is useful only when readers can match it to the evaluated version. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Open the Development and Validation Record”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 10 OF 16

10. Model Learning, Updating and Drift

Back to contents

Why this move matters

If the intervention learns or changes, explain how individual-level performance may change and how that change enters the economic model.

Translate the decision, not only the algorithm

If the intervention learns or changes, explain how individual-level performance may change and how that change enters the economic model. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Scenarios compare a locked model with scheduled updates and possible drift between reviews. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

Assuming permanent performance can overstate long-term value. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Model Learning, Updating and Drift”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 11 OF 16

11. Measure Implementation, Maintenance and Replacement Costs

Back to contents

Why this move matters

Report purchase, integration, data, infrastructure, training, monitoring, cybersecurity, support, retraining and replacement costs, with price year and currency.

Translate the decision, not only the algorithm

Report purchase, integration, data, infrastructure, training, monitoring, cybersecurity, support, retraining and replacement costs, with price year and currency. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Costs include interface work, staff training, model monitoring and downtime response. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

A licence fee can be a small part of total lifecycle cost. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Measure Implementation, Maintenance and Replacement Costs”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 12 OF 16

12. Compare Training Data With Effect Data

Back to contents

Why this move matters

Describe important differences between development data and the population or data used to estimate real-world effects.

Translate the decision, not only the algorithm

Describe important differences between development data and the population or data used to estimate real-world effects. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: The training set came from tertiary hospitals while the effect study includes smaller sites. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

Performance and value can change when prevalence, equipment or workflow changes. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Compare Training Data With Effect Data”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 13 OF 16

13. Characterise Heterogeneity and Distribution

Back to contents

Why this move matters

Report subgroup variation, priority populations and who gains or bears costs, including unequal access or error.

Translate the decision, not only the algorithm

Report subgroup variation, priority populations and who gains or bears costs, including unequal access or error. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Results are separated by site type, age group and access pathway without assuming equal effects. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

An average favourable result can conceal an unfavourable subgroup. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Characterise Heterogeneity and Distribution”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 14 OF 16

14. Expose Structural, Parameter and AI Uncertainty

Back to contents

Why this move matters

Show uncertainty in inputs, model structure, extrapolation, thresholds, drift, implementation and future versions through appropriate analyses.

Translate the decision, not only the algorithm

Show uncertainty in inputs, model structure, extrapolation, thresholds, drift, implementation and future versions through appropriate analyses. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Scenarios vary uptake, false negatives, staffing, update costs and performance decay. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

A narrow confidence interval around one input does not cover structural uncertainty. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Expose Structural, Parameter and AI Uncertainty”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 15 OF 16

15. Report Results With Components and Trade-offs

Back to contents

Why this move matters

Present costs, outcomes, incremental comparisons and uncertainty separately before any summary ratio or conclusion.

Translate the decision, not only the algorithm

Present costs, outcomes, incremental comparisons and uncertainty separately before any summary ratio or conclusion. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: Readers see workload, costs, health outcomes and error consequences before the final incremental result. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

One ratio can hide which component drives the conclusion. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “Report Results With Components and Trade-offs”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

CHAPTER 16 OF 16

16. State Limits, Engagement, Funding and Decision Boundaries

Back to contents

Why this move matters

Report limitations, transfer conditions, patient and stakeholder involvement, conflicts, funding, model access and what the study cannot decide.

Translate the decision, not only the algorithm

Report limitations, transfer conditions, patient and stakeholder involvement, conflicts, funding, model access and what the study cannot decide. Identify the decision maker, perspective, alternatives, population, setting, outcomes, time horizon and resource consequences before reading the cost-effectiveness result.

Rebuild one economic pathway

Worked example: The report explains patient input, vendor involvement and why adoption still needs local safety review. Trace how the AI changes information, action, health outcomes, workload, costs and uncertainty compared with current care.

Stress-test change over time

Ask what happens when performance drifts, the model learns, software versions change, prices fall, maintenance rises or users override advice. Economic value is a moving relationship, not a permanent sticker.

Keep reporting and appraisal separate

Economic reporting does not replace clinical governance or public accountability. CHEERS-AI can make assumptions and inputs visible, but it does not prove that an AI intervention is clinically safe, fair, affordable or worth adopting.

Student, parent and career transfer

Make one evidence card for “State Limits, Engagement, Funding and Decision Boundaries”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports health economics, health technology assessment, medicine, data science, hospital management, policy, procurement and evidence communication without promising a grade, admission, funding decision or job.

PRACTICE LAB

Build a cheers_ai decision trail

Student route

1. Choose one original report inside this guide’s scope.
2. Copy its objective and underline actor, action, context, evidence and intended decision.
3. Draw the specific reasoning chain used in this article.
4. Mark each step reported, inferred, contested or unknown.
5. Rewrite the conclusion so it preserves the narrowest important limitation.

Parent or teacher route

Ask the learner to explain read the economic value of an ai health intervention as a transparent chain of choices, evidence, costs and uncertainty twice: once to a classmate and once to a method specialist. Praise a precise correction and a well-stated unknown; both show disciplined attention to evidence.

Early-career route

Create a one-page cheers_ai audit with source link, date checked, scope, method logic, decisive wording, contrary evidence, limit and next action. A colleague should be able to continue without guessing.

Continue with eduKate’s guides to reading course and career information and giving useful feedback in group projects.

USEFUL NEXT READING

Connect this narrow owner to its neighbour

Read the RELEVANT real-world-evidence appraisal guide for a related but different reader job. Return here when the question is specifically how to read the economic value of an ai health intervention as a transparent chain of choices, evidence, costs and uncertainty. This boundary keeps one canonical owner per intent.

For English practice across research, school and career communication, visit the English at work and in public life guide.

FREQUENTLY ASKED QUESTIONS

Questions students and parents often ask

What is CHEERS-AI?

Reporting guidance for economic evaluations of interventions that use AI to perform their function.

How many items are in the final checklist?

Thirty-eight: 28 CHEERS 2022 items, with eight AI elaborations, plus ten new AI-specific items.

Does cost-effective mean cheaper?

No. It compares additional costs and consequences against alternatives under a stated decision framework.

Why report the AI version?

Performance, workflow and cost may change when a model or interface changes.

Does CHEERS-AI prove safety or fairness?

No. It improves reporting; safety, fairness, effectiveness and governance need their own evidence.

Why include equity?

Average value may hide who gains, who is excluded and who bears errors or costs.