Presentation (CTI)

OBJECTIVES: The excitement around artificial intelligence (AI) in health economics and outcomes research (HEOR) is growing rapidly. However, there is increasing concern that many tools and processes marketed as “AI” are, in fact, traditional automation. This conflation can lead to misunderstanding of capabilities, misplaced investments, and reduced credibility. This study attempts to compare the nature and claims of AI applications in HEOR to clarify the boundary between traditional automation and true artificial intelligence.
METHODS: A desk review was conducted to examine HEOR publications from 2023-2024 that claimed to apply AI. Each identified paper was screened and categorized into three groups:
A) Automation only - rule-based tools, scripts, or macros with no learning capability (e.g., automated review tools, batch analyses).
B) Automation with AI elements - systems that incorporate machine learning components (e.g., predictive models) but lack adaptive behavior, generative capabilities, or closed-loop feedback. These systems typically run deterministic code where outputs do not change based on evolving data.
C) True AI - pipelines that employ AI models integrated into workflows with minimal human intervention. Rule-based logic may be used to connect and orchestrate different components, but critical operations are driven by AI. These systems exhibit adaptability, reasoning, or autonomous decision-making.
RESULTS: Initial findings revealed that the majority of HEOR papers reviewed fell under Category A, comprising automation tasks repackaged as AI. A smaller proportion demonstrated Category B characteristics, often involving predictive algorithms used without updates or feedback loops. Only a few studies showed semblances of Category C, with limited autonomous or learning-based components.
CONCLUSIONS: This analysis highlights a critical gap between perceived and actual use of AI in HEOR. Mislabelling automated processes as AI risks inflating expectations and undermining scientific rigor. As AI becomes more embedded in decision-making, HEOR must adopt clearer definitions and reporting standards to differentiate true AI from automation.