This site has been optimized to work with modern browsers and does not fully support your version of Internet Explorer.

What we learned at ISPOR Europe 2026

Author

AI can speed up health economic modelling, but speed is not the same as rigour. That was the message we took away from ISPOR Europe 2026 in Vienna, where AI in health technology assessment (HTA) was a prominent theme. Sessions explored how it could support systematic literature reviews, real-world evidence analysis and health economic modelling. Health Innovation Oxford and Thames Valley (HIOTV) also presented four posters on early evidence generation, which means building the evidence an innovation needs before it has full clinical or economic data. Here are our key takeaways:

  • AI can improve the efficiency of HTA, but methodological rigour remains critical.
  • Early economic evaluation should be proportionate to the maturity of the technology and available evidence.
  • Health economics can inform development and adoption not just reimbursement.

AI in HTA: where human expertise still matters

ISPOR sessions demonstrated how AI could accelerate literature reviews, real-world evidence analysis, health economic model development and report writing. Although these applications may improve productivity, panellists emphasised that any efficiency gains must be accompanied by appropriate human oversight, transparency, accountability and methodological rigour.

The overarching message, however, was that AI should augment rather than replace expert judgement. Greater speed should not be assumed to deliver equivalent analytical rigour. Human oversight remains essential, particularly for primary review, methodological decision-making and interpretation. Any model developed using AI, regardless of how quickly it is produced, must undergo transparent and rigorous validation by an experienced health economist before it can reliably inform decisions.

A session on AI agents for health economic modelling* demonstrated an end-to-end process. The process began with mapping the decision problem and assessing the existing evidence landscape. Based on this assessment, a model specification is developed and used to guide an AI agent in constructing the health economic model. The modeller then performs a series of quality assurance checks to ensure the model’s accuracy, robustness and methodological integrity. Once the model has been validated, AI is used to generate a draft technical report. The modeller subsequently reviews and validates the report to ensure accuracy, transparency, completeness and appropriate referencing.  Although time savings were reported across several stages, their value depends on how the modeller’s released capacity is used and the level of validation required.

HIOTV believes validation of outputs is particularly important because each innovation, population and care pathway presents distinct methodological challenges. Errors or omissions in the initial model specification may propagate throughout the analysis, potentially creating additional validation work and reducing, or even eliminating, the anticipated efficiency gains. In some circumstances, developing the model manually may therefore remain more proportionate.

Further questions arise in early economic evaluation, where evidence may be limited, uncertain or heavily reliant on assumptions and expert elicitation. As large language models are trained primarily on existing information, their ability to support the assessment of genuinely novel technologies requires careful scrutiny. In these circumstances, continued expert oversight is essential. Credibility depends on transparent assumptions, appropriate validation and clear reporting of where and how AI has been used.

We believe AI-supported HTA should therefore complement, rather than replace, methodological expertise. Any application must preserve the principle that each assessment is tailored to the innovation and its intended pathway. Innovators should also retain a clear understanding of the model’s structure, assumptions, limitations and findings, particularly when using the analysis to inform business cases or adoption decisions.

Ultimately, AI should be evaluated against the same standards of transparency, validity and reproducibility as the technologies it is used to assess. Hence, AI-supported analyses must continue to meet relevant HTA requirements, reporting standards and data-governance expectations. These considerations are particularly important in early evidence generation, where uncertainty is often substantial and analytical choices may materially affect the findings. The four posters presented by HIOTV at ISPOR Europe 2026 (see below) illustrate how proportionate methods can be applied in practice.

* Content relates to the ISPOR session “Can Agentic AI Deliver HTA-Ready Health Economic Models? Governance, Validation, and Trust.” The session was moderated by Andrew Briggs, DPhil, of the London School of Hygiene & Tropical Medicine. Speakers included Jag Chhatwal, PhD, of Harvard Medical School and Massachusetts General Hospital; Andre Verhoek, MSc, of AstraZeneca; and Xiaoyan Wang, PhD, of Tulane University.

Our ISPOR posters demonstrated early evidence-generation methods

The four posters presented by HIOTV demonstrated how early evidence-generation methods can be tailored to a technology’s maturity, available evidence and intended care pathway. The work ranged from stakeholder research and pathway mapping to resource-use modelling and exploratory cost-effectiveness analysis.

  • Poster 1: Lean Assessment Process (LAP) – This poster described the LAP methodology involving different steps within a mixed-methods approach and described how it was applied to better understand the value proposition of a real-time patient feedback platform. Based on engaging with healthcare professionals and patient representatives across eight NHS trusts, the LAP methodology helped clarify the value proposition and level of acceptance of the platform and identified influential stakeholders, workflow implications and organisational barriers that could affect adoption.
  • Poster 2: Early cost-consequence analysis of 1View – This poster examined the impact of this 3D imaging tool for stroke care on planning time, procedural duration, staff and resource use, radiation exposure and NHS costs in the mechanical thrombectomy pathway, indicating where efficiency gains might arise while making clear that prospective evidence is required to validate the modelled benefits.
  • Posters 3 and 4: Early cost-effectiveness analysis of mutTNF on brain metastases in melanoma and lung cancer patients: These posters were developed building on an earlier analysis in breast cancer brain metastases, whose abstract was ranked among the top 5% of submissions to ISPOR Europe 2025. The modelling framework was adapted to reflect indication-specific pathways and treatment patterns, using scenario, subgroup, probabilistic sensitivity and value-of-information analyses to explore the potential cost-effectiveness of mutTNF, a new way to help cancer drugs reach the brain, and identify priorities for further research. Read more in our case study.

Across these projects, HIOTV’s support extended beyond the production of preliminary economic estimates. Innovators received structured evidence to refine their value propositions, identify potential adoption barriers, prioritise data collection and inform clinical studies, pricing considerations, funding applications and future discussions about NHS adoption.

How we can support you

Innovators need to know what evidence decision-makers will expect, how to show value credibly, and which uncertainties to tackle first. Our health economists and methodologists work with innovators, SMEs, academics and NHS teams to build evidence that is proportionate to the stage of the technology, the data available and the intended NHS pathway. Support can include:

  • Value proposition development and stakeholder studies to understand need, acceptance and adoption barriers.
  • Pathway and resource-use mapping.
  • Early economic modelling, including cost-consequence, cost-effectiveness, scenario, sensitivity and value-of-information analysis.
  • Real-world health economics studies, using real-world data to assess costs and outcomes in practice.
  • Evidence-generation planning, including health economics input for funding applications such as NIHR, Innovate UK, SBRI and UKRI.

Early evidence is not only about reimbursement. It can help you shape the product, plan your studies and prepare for NHS adoption long before a pricing conversation. You can come away knowing which assumptions matter most, what data to collect next, and how to make a stronger case to NHS decision-makers, funders and partners. As the technology matures, this can build towards budget impact analysis and HTA-ready evidence. All outputs are validated by experienced health economists, whether or not AI has supported the work. Download our guide to see how early health economic evaluation could support your innovation, or get in touch to talk through what stage you are at.

About the author

News Categories: Blog