Session (CTI)

PURPOSE: Generative AI, particularly large language models (LLMs), can transform evidence synthesis, health economic modeling, and real-world evidence (RWE) generation. To realize this potential HEOR professionals need to master prompt engineering—the art and science of crafting instructions that guide LLMs to produce precise, contextually relevant outputs. Offered by members of the ISPOR Generative AI Working Group, this session will teach participants how to optimize their prompts to tasks such as data extraction, health economic modeling, and generating insights from real-world datasets.

DESCRIPTION: This session begins with a primer on LLMs, focusing on their underlying mechanics and how thoughtful prompt design improves output quality. We will highlight the wide range of HEOR applications for LLMs.Participants will explore key prompt engineering techniques, including Zero-shot, Few-shot, Chain of Thought, Tree of Thoughts, and persona prompting, with examples tailored to evidence synthesis, HEOR, and RWE tasks. Practical applications, such as extracting data, synthesizing insights from real-world datasets, and health economic modeling, will be demonstrated.To ensure engagement, the session features open discussions where participants collaboratively explore prompt engineering challenges and solutions. Attendees will also discuss LLM capabilities and limitations, such as contextual retention and accuracy issues, and learn how effective prompting mitigates these challenges.This interactive format allows participants to share experiences, ask questions, and develop practical strategies for integrating LLMs into HEOR and RWE tasks. While optional, participants are encouraged to experiment with these techniques using tablets or mobile devices.