In brief: Health technology assessment (HTA) is a multidisciplinary process that uses explicit methods to determine the value of a health technology at different points in its lifecycle, in order to inform decisions that promote an equitable, efficient and high-quality health system. Artificial intelligence meets HTA in two ways: as a tool that helps produce HTA evidence, and as a technology that HTA itself has to evaluate.
Health technology assessment and artificial intelligence: a general introduction
What HTA is
The definition above was agreed in May 2020 by the main international HTA networks, including INAHTA, HTAi, ISPOR, EUnetHTA, RedETSA and HTAsiaLink (O’Rourke et al., 2020). Four notes accompany it. A health technology is any intervention developed to prevent, diagnose or treat medical conditions, promote health, provide rehabilitation or organise healthcare delivery: a test, device, medicine, vaccine, procedure, programme or system. The process is formal, systematic and transparent. Value is multidimensional, typically covering clinical effectiveness, safety, costs and economic implications, ethical, social, cultural and legal issues, organisational and environmental aspects, and wider effects on patients, carers and the population. And HTA can be applied at any point in the lifecycle, from pre-market through to disinvestment.
Assessment and appraisal
Most HTA systems separate two steps. Assessment is the scientific part: systematic reviews of clinical evidence, comparison with existing alternatives, and economic evaluation, often a cost-effectiveness model that expresses health gains in quality-adjusted life years (QALYs) and summarises the result as an incremental cost-effectiveness ratio (ICER). Appraisal is the decision part: a committee weighs the evidence together with uncertainty, equity, budget impact and social values, and recommends whether and how the technology should be funded or used.
| Question | Typical HTA method |
|---|---|
| Does it work, and for whom? | Systematic review and meta-analysis of clinical trials and real-world evidence |
| Is it safe? | Review of adverse events, post-market surveillance data |
| Is it worth the money? | Cost-effectiveness or cost-utility analysis, ICER, budget impact analysis |
| Can the system deliver it? | Organisational analysis: workforce, training, infrastructure, pathway changes |
| Is it fair and acceptable? | Ethical, legal, social and patient-perspective analysis |
A moving policy landscape
HTA began as a policy instrument in the 1970s and spread through national agencies such as England’s NICE. In Europe, Regulation (EU) 2021/2282 has applied since 12 January 2025. It introduces joint clinical assessments at EU level, first for new cancer medicines and advanced therapy medicinal products, extending to orphan medicines in 2028 and to all new centrally authorised medicines in 2030; economic evaluation and pricing remain national decisions (Regulation (EU) 2021/2282).
Where AI and HTA meet
| AI as a tool for HTA | HTA of AI technologies | |
|---|---|---|
| Core question | Can AI make evidence generation faster or better? | Does an AI-enabled device or software deliver value in practice? |
| Typical uses or objects | Automated search, screening and data extraction for systematic reviews; extraction of real-world data; support for building economic models | Diagnostic algorithms, triage tools, clinical decision support, predictive models embedded in care |
| Main risks | Algorithmic bias, cybersecurity, reduced human oversight and transparency, “black box” reporting | Weak clinical evidence, poor generalisability, algorithms that change after approval, unclear accountability |
| Current guidance | NICE (2024): AI should augment, not replace, human involvement; use for comparative effect estimation is higher risk and needs sensitivity analysis and triangulation; report with tools such as PALISADE and TRIPOD+AI | Evidence standards for digital health technologies; the NICE statement explicitly excludes AI-enabled technologies, which are evaluated through other routes |
The first EU joint clinical assessments
The EU HTA Regulation became operational on 12 January 2025. In its first year it supported ten joint clinical assessments (JCAs) and seven joint scientific consultations. The 2026 work programme foresees a large expansion: around 50 JCAs of medicines, about 35 of them oncology medicines with new active substances and 15 advanced therapies, plus around five JCAs of high-risk medical devices starting from June 2026 (Becaris Publishing, 2026).
The first JCA report, on tovorafenib for relapsed or refractory paediatric low-grade glioma, was endorsed by the HTA Coordination Group on 30 April 2026 and published by the European Commission on 9 June 2026 (European Commission, 2026). Its substance was that the evidence was insufficient to show added benefit: the medicine had been approved on uncontrolled data, comparative evidence was missing for most of the questions asked, and an under-powered indirect comparison was heavily discounted (Ramagopalan & Pannelay, 2026). The case shows how strictly EU assessors read single-arm trials and unanchored comparisons, and it is a useful warning for developers of AI-based devices, whose evidence is often similar in form.
Frameworks for assessing AI technologies
| Framework | Origin | What it adds |
|---|---|---|
| MAS-AI (Model for ASsessing the value of AI) | Odense University Hospital, Denmark, 2022 | An HTA model for AI in medical imaging with nine domains in two steps: first the patients, the development of the AI model and initial ethical and legal issues; then safety, clinical, economic, organisational and patient aspects (Fasterholdt et al., 2022). Its validity and transferability were later tested in workshops in Denmark, Italy and Canada (Fasterholdt et al., 2026). |
| CHEERS-AI | International Delphi study, 2024 | A reporting guideline for economic evaluations of interventions that use AI: the 28 CHEERS 2022 items plus 10 AI-specific items, with 8 original items elaborated (Elvidge et al., 2024). |
| NICE early value assessment | England | Conditional recommendations while evidence is generated. In January 2025, for example, NICE allowed four AI technologies that help detect fractures on X-rays in urgent care to be used in the NHS during a two-year evidence generation period (National Institute for Health and Care Excellence [NICE], 2025). |
Regulation is moving on a different clock
AI-enabled devices are reaching the market quickly. The US Food and Drug Administration reports that it had authorised more than 1,600 AI-enabled medical devices by September 2026 (U.S. Food and Drug Administration, 2026). In the EU, the AI Act treats most AI that is, or is part of, a medical device as high-risk, but the Digital Omnibus agreed in 2026 postponed the high-risk obligations for AI embedded in regulated products such as medical devices to 2 August 2028 (CMS, 2026). For HTA bodies the implication is that market authorisation and value assessment run on different timelines: a tool may be approved, and in use, long before its comparative effectiveness and cost-effectiveness are known.
Questions to ask about any AI technology under assessment
- Intended use: which decision does the tool change, for which patients, and who acts on its output?
- Evidence of benefit: is there evidence beyond diagnostic accuracy, such as changes in clinical management, patient outcomes or workload?
- Generalisability: was it validated on data similar to the local population, equipment and workflows?
- Change over time: how will software updates be controlled and performance monitored after adoption?
- Economics: does the model include the full cost of implementation, including IT integration, training and monitoring, and is it reported with CHEERS-AI?
- Equity and accountability: are performance differences across subgroups reported, and who is responsible when the tool is wrong?
Our summary of Ramezani et al. (2025) focuses on the first column, and especially on how AI can strengthen the economic dimension of HTA. Keeping both columns in mind avoids a common confusion in the literature, where “AI in HTA” can mean either.
Related reading on this site: the article summary The Intersection of Artificial Intelligence and Health Technology Assessment, the Economic Evaluation & HTA, Outcome Measures & Health-State Valuation and AI Ethics, Trust & Governance collections.
Sources for this introduction
- Becaris Publishing. (2026). What to expect from the EU HTA Regulation in 2026. https://becarispublishing.com/digital-content/blog-post/expect-eu-hta-regulation-2026
- CMS. (2026). The EU’s Digital Omnibus: Streamlined AI Act compliance rules are now in force. https://cms.law/en/gbr/publication/the-eu-s-digital-omnibus-streamlined-ai-act-compliance-rules-are-now-in-force
- Elvidge, J., Hawksworth, C., Avşar, T. S., Zemplenyi, A., Chalkidou, A., Petrou, S., et al. (2024). Consolidated Health Economic Evaluation Reporting Standards for interventions that use artificial intelligence (CHEERS-AI). Value in Health, 27(9), 1196–1205. https://doi.org/10.1016/j.jval.2024.05.006
- European Commission. (2026, June 9). Joint clinical assessment report on tovorafenib (Ojemda). https://health.ec.europa.eu/publications/joint-clinical-assessment-report-tovorafenib-ojemda_en
- Fasterholdt, I., Kjølhede, T., Naghavi-Behzad, M., Schmidt, T., Rautalammi, Q. T. S., Hildebrandt, M. G., Gerdes, A., Barkler, A., Kidholm, K., Rac, V. E., & Rasmussen, B. S. B. (2022). Model for ASsessing the value of Artificial Intelligence in medical imaging (MAS-AI). International Journal of Technology Assessment in Health Care, 38(1), e74. https://doi.org/10.1017/S0266462322000551
- Fasterholdt, I., Schrøder, J. S., Hansen, L. H., Bowen, J. M., Gerdes, A., Kidholm, K., et al. (2026). Validity and transferability of Model for ASsessing the value of Artificial Intelligence (MAS-AI). International Journal of Medical Informatics, 206. https://doi.org/10.1016/j.ijmedinf.2025.106127
- National Institute for Health and Care Excellence. (2024). Use of AI in evidence generation: NICE position statement (ECD11). https://www.nice.org.uk/corporate/ecd11
- National Institute for Health and Care Excellence. (2025). Artificial intelligence (AI) technologies to help detect fractures on X-rays in urgent care: Early value assessment (HTE20). https://www.nice.org.uk/guidance/hte20
- O’Rourke, B., Oortwijn, W., Schuller, T., & the International Joint Task Group. (2020). The new definition of health technology assessment: A milestone in international collaboration. International Journal of Technology Assessment in Health Care, 36(3), 187–190. https://doi.org/10.1017/S0266462320000215
- Ramagopalan, S. V., & Pannelay, A. J. (2026). Access in all areas? A round-up of developments in market access and health technology assessment: First Joint Clinical Assessment report released. Journal of Comparative Effectiveness Research, 15(7), e260117. https://doi.org/10.57264/cer-2026-0117
- Regulation (EU) 2021/2282 of the European Parliament and of the Council of 15 December 2021 on health technology assessment and amending Directive 2011/24/EU. Official Journal of the European Union, L 458. https://eur-lex.europa.eu/eli/reg/2021/2282/oj
- U.S. Food and Drug Administration. (2026). Artificial intelligence-enabled medical devices. https://www.fda.gov/medical-devices/digital-health-center-excellence/artificial-intelligence-enabled-medical-devices
Cite this page (APA 7)
Kurutkan, M. N. (2026, November 15). Health Technology Assessment and AI: An Introduction (101). The Health Topics. https://healthtopic.org/topics/health-technology-assessment-and-ai-an-introduction-101/
