Symmetron | ISPOR EU 2024

Large Language Models for Outcomes Research

Monday, 18 November, 4:00-7:00 PM

A targeted literature review examining the application of large language models for health outcomes research. Use cases, recommendations for implementation and future research, and barriers to implementation are summarised.
Future research examining approaches to manually validate automated actions, for example using sampling, could help mitigate variations in LLM performance and improve efficiency gains while maintaining research quality.

Modelling Approaches for Digital Health Technology in NICE's Early Value Assessment

Tuesday, 19 November, 10:30 AM-1:30 PM

A targeted literature review of early value assessments (EVA) conducted by NICE to appraise digital health technologies. EVAs often face evidence constraints, therefore requiring decisions on how to robustly undertake economic analysis.
This research aims to investigate the model structures and approaches used to evaluate the cost-effectiveness of the appraised technologies despite identified evidence gaps.

The Role of RWE in NICE Submissions for Rare Diseases

Tuesday, 19 November, 10:30 AM-1:30 PM

The Highly specialised technologies (HST) process by NICE evaluates treatments for rare diseases, where real-world evidence (RWE) is crucial due to limited RCT feasibility.
A review of 10 HST submissions (June 2022–May 2024) found RWE used in 9/10 cases to support findings. Though generally accepted, concerns over RWE quality and uncertainty were noted, with additional data collection recommended in 5 cases.

Large Language Models for Health Economics and HTA Submissions

Tuesday, 19 November, 10:30 AM-1:30 PM

A targeted literature review examining the application of large language models for health economics research as well as understanding health technology assessment body decision precedent. Research recommendations and barriers to implementation identified by the studies are discussed.
Future research should continue to explore the feasibility of integrating LLMs into common HE tasks and develop effective workflows for validating model responses.

Considerations for Modelling in Crowded Treatment Landscapes

Tuesday, 19 November, 4:00-7:00 PM

This poster examines the unique considerations of health economic modelling in disease areas with multiple available treatments, drawing on insights from recent health technology appraisals in the UK, Canada, Australia, and the US across three disease areas.
Building on previous approaches and feedback from HTA bodies, the poster provides recommendations and guidance for optimising model development in crowded landscapes.

A Simulation Study Assessing a Stopping Rule to Reduce Screening Burden

Tuesday, 19 November, 4:00-7:00 PM

A simulation study demonstrating a substantial reduction in total records screened during title and abstract screening is possible with a stopping rule.
The rule relies on an initial set of randomly sampled records to certify that a desired percentage of relevant records have been discovered during prioritised screening.
The appropriateness of implementing the rule for different review types is discussed.

How to Improve Credibility through All Phases of Model Development

Wednesday, 20 November, 9:00-11:30 AM

A targeted literature review which explored the processes involved in validating a health economic model. The 21 unique approaches identified included model tools and conceptual frameworks aimed at developers, users, and external decision-makers.
The evidence was synthesised to highlight key methodologies across five key stages of model development: conceptualisation, development, finalisation, reporting, and post-completion.