Transforming Regulatory Intelligence With AI Driven Methodologies in HEOR
Author(s)
Devyani Biswal, BSc, PhD, Luk Arbuckle, BSc, MSc;
IQVIA, Ottawa, ON, Canada
IQVIA, Ottawa, ON, Canada
Presentation Documents
OBJECTIVES: Regulatory intelligence involves the collection, analysis, and application of regulatory information to support decision-making and ensure adherence to evolving standards. As health regulations evolve, and privacy, data protection, and AI regulations grow in complexity, managing these actionable insights requires efficient and scalable solutions. This study introduces a methodology leveraging large language models (LLMs) and AI Agents to automate the mapping of diverse regulatory and operational standards to structured frameworks, from quality assurance and patient safety to privacy and AI risk management. The objective is to streamline requirements management while maintaining accuracy, traceability, and scalability.
METHODS: The methodology begins with data preprocessing, including document parsing, normalization, and segmentation of regulatory guidelines and standards. The AI pipeline identifies actionable requirements by mapping text to framework categories using advanced natural language processing. Each compliance action is scored based on risk, relevance, and implementation feasibility, creating an option analysis through score-based thresholding (low score = all possible requirements; high-score = only most needed requirements). A human-in-the-loop refines these mappings, resolves ambiguities, and interprets domain-specific variations. Validation is achieved through benchmarking against pre-established crosswalks and confidence-scoring mechanisms to ensure output reliability.
RESULTS: The AI pipeline processed large volumes of complex documents efficiently, producing accurate mappings within significantly reduced timelines compared to manual approaches, with the scoring confirmed by expert review as reasonable estimation and guide on the alignment of requirements to the relevant guidance or standards processed. The integration of expert review ensured alignment with specific business contexts and usability constraints.
CONCLUSIONS: This scalable, automated approach to regulatory intelligence advances traditional methods by reducing manual effort, enhancing accuracy, and enabling risk-based prioritization. Combining AI capabilities with human expertise creates a robust system that adapts to evolving regulatory landscapes and diverse guidance and standards, providing organizations with actionable and transparent regulatory insights.
METHODS: The methodology begins with data preprocessing, including document parsing, normalization, and segmentation of regulatory guidelines and standards. The AI pipeline identifies actionable requirements by mapping text to framework categories using advanced natural language processing. Each compliance action is scored based on risk, relevance, and implementation feasibility, creating an option analysis through score-based thresholding (low score = all possible requirements; high-score = only most needed requirements). A human-in-the-loop refines these mappings, resolves ambiguities, and interprets domain-specific variations. Validation is achieved through benchmarking against pre-established crosswalks and confidence-scoring mechanisms to ensure output reliability.
RESULTS: The AI pipeline processed large volumes of complex documents efficiently, producing accurate mappings within significantly reduced timelines compared to manual approaches, with the scoring confirmed by expert review as reasonable estimation and guide on the alignment of requirements to the relevant guidance or standards processed. The integration of expert review ensured alignment with specific business contexts and usability constraints.
CONCLUSIONS: This scalable, automated approach to regulatory intelligence advances traditional methods by reducing manual effort, enhancing accuracy, and enabling risk-based prioritization. Combining AI capabilities with human expertise creates a robust system that adapts to evolving regulatory landscapes and diverse guidance and standards, providing organizations with actionable and transparent regulatory insights.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
MSR7
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas