Last week, the annual R for Health Technology Assessment (HTA) conference – a gathering to ‘present interesting and enlightening presentations on the use of R … [for] those working in the field of health technology assessment and related analysis’ – was held over the course of three days at the University of Exeter. The conference brought together those from a range of backgrounds within the HEOR industry, with attendance from those in academia, External Assessment Groups and strong representation from industry and consultancy perspectives. Alex Mclean, a Senior Health Economist at Costello Medical, was in attendance – discover his key takeaways.
I had the pleasure of being one of the first presenters at R for HTA, talking through an Excel and R hybrid cost-effectiveness model we have spent the last year developing at Costello Medical in collaboration with one of our clients. The presentation demonstrated our work taking an existing VBA and Excel based individual patient simulation cost-effectiveness model, and adding in fully automated R functionality. Patient simulations are offloaded to R, drastically reducing model run times of up to 90% for the more complex analyses. Engagement with the presentation was encouragingly positive, with the audience showing a keen interest and having plenty of questions.
Excel/R hybrid models seem to be an emerging trend in the industry, with another presentation that day showing a similar approach to utilising the computational efficiency of R without sacrificing the familiar user interface of Excel. Fully R integrated models undoubtably have their benefits, offering a more streamlined workflow, greater AI compatibility and ways to host these models online for ease of distribution among stakeholders and affiliates. Despite this, I think the overwhelming takeaway from the hybrid model presentations were the need for a hybrid approach to offer the industry a stepping stone towards more fully integrated R modelling methods.
Parts of the HEOR industry lack the required skills to conduct detailed modelled adaptations in R. This limits the success of rolling out fully R integrated models if local affiliates cannot easily adapt model inputs or HTA bodies cannot easily test preferred modelling scenarios. Despite not being quite as streamlined as fully R integrated models, hybrid approaches offer fantastic efficiency gains – our model offering up to 90% reductions in run time for some of the most complex analyses. If hybrid models offer (almost) all the efficiency gains as fully R integrated models, with the additional benefit of a familiar Excel user interface allowing easy engagement with affiliates and stakeholders, what degree of benefit do fully integrated R models need to offer to outweigh this? This is not to dispute the benefits of fully R integrated models, but simply whether the trade-off with lack of engagement and usability makes them currently worthwhile for HEOR professionals.
A similar, broader takeaway from the conference as a whole was the trade-off between complexity and necessity. I am certainly guilty myself of approaching difficult problems in a complex way in a bid to develop the most technically accurate solution. While in an academic setting, this complex, technically accurate approach is the standard; in an HTA setting, where the real point of a cost-effectiveness model is to provide accurate enough evidence to allow for correct decision making, the level of technical complexity only needs to go as far as to ensure correct decisions are made. Pragmatic solutions for technical modelling problems in an HTA setting, which often offer easier engagement with users and simpler validation, shouldn’t be discounted provided overall decision making isn’t affected.
Day 1 of R for HTA 2026 came to a close with an open discussion hosted by Gabriel Rogers and Catherine McGuire from Manchester Centre for Health Economics and University of Manchester Research Software Engineering. Here Gabriel Rogers, a classically ‘Excel-trained’ health economic modeller, presented his work converting a verbose Excel model into a streamlined R model. A common theme limiting uptake of R is health economic modellers not having the required skills, something openly discussed in the room at the close of their presentations. The following stood out to me from the presentation:
As someone who has come from a physics further education background, where coding languages were actively taught alongside high level maths and problem-solving skills, all of which are extremely relevant to the HEOR world, I wondered if there is a potential need for broader hiring within the industry. While health economics was not a field I was familiar with during my studies, I have enjoyed having been able to utilise these technical skills within the HEOR world. If technical programming and other complex modelling techniques are becoming more mainstay, advertising the industry to a wider range of technical backgrounds could help address skill shortages.
The other standout highlight from R for HTA 2026 was the issue panel held at the conclusion of day 2: Using R to leverage the power of gen AI in HTA. It is clear AI will (and already has in many cases) transform the typical health economist workflows over the next few years. One of the standout benefits of economic modelling using coding languages like R is the compatibility with AI. For the time being, AI can write, analyse and validate code to a far more accurate degree than it can with Excel, indicating that as AI develops, so too will the uptake of R modelling. The logical next step is that health economist workflows will go from building models (which may well be dominated by AI), to validating AI outputs or guiding AI tools to fine tune models.
During the issue panel, the common concerns regarding AI were voiced: How can we trust the outputs of AI written models? How can we ensure sensitive data remains confidential? How will we ensure HTA bodies keep up with the pace of AI innovation in industry? Many of these questions remain unanswered, however panellists consistently emphasised AI is a tool to enhance workflows, and not a dangerous crutch. Sure, AI can have inherent bias, but is that any different to humans sitting on a committee panel? Provided responsibility for outputs remain with the human and are not offloaded to AI, a lot of concerns over this new technology seem to dwindle.
As increasingly complex treatments and their pathways which require detailed models become more common in combination with AI dominating health economist workflows, it seems the increased uptake of R modelling is not only necessary but inevitable. Excel will certainly remain as the dominant modelling programme in HEOR for now, still being the standout tool for the vast majority of modelling tasks that do not require the computational power of programming languages. I provided overall decision making is not impacted. It is also evident that upskilling across sectors of the HEOR industry, through collaboration with other industries and changes to health economics higher education courses, are required to keep ahead of these trends.
If you would like any further information or advice on the themes presented above, please get in touch, or visit our Health Economics page to learn how our expertise can benefit you. Alex Mclean (Senior Health Economist), created this article on behalf of Costello Medical. The views/opinions expressed are their own and do not necessarily reflect those of Costello Medical’s clients or affiliated partners.
