Artificial Intelligence in Health Economic Modeling and Systematic Reviews: Opportunities, Challenges, and Future Directions
Author(s)
Jahangir Nabi Mir, M. Pharm1, Manne Mithun Chakrawarthy, M. Pharm1, Yogesh Suresh Punekar, PhD2.
1IQVIA, Gurugram, India, 2IQVIA, London, United Kingdom.
1IQVIA, Gurugram, India, 2IQVIA, London, United Kingdom.
OBJECTIVES: Artificial intelligence (AI) is increasingly investigated in health economics and outcomes research (HEOR) to enhance efficiency, decision-making and minimize operational burden. However, evidence of its economic value and real-world application remains scarce. This review assesses the existing evidence base of AI use in HEOR, focusing on AI-enabled economic modelling and automation of systematic literature reviews (SLRs).
METHODS: PubMed, Google Scholar, and grey literature were searched to identify English-language studies published from January 2015 to May 2025. Studies were included if they evaluated AI’s application in economic modelling or evidence synthesis relevant to HEOR.
RESULTS: AI use was associated with substantial projected healthcare savings, estimated at US$200 to 360 billion annually in the United States (5-10% of the national health spending), with estimated cost reductions of 7-9% for private payers and 3-8% for physician groups. These were, however, majorly model projections and had no real-world validation. In economic evaluations, AI tools assist in semi-automated generation of decision trees and Markov models, with generative AI (e.g., GPT-4) used for health state definitions and ICER estimations. Despite these advancements, concerns persist regarding hallucination risks, black-box models, prompt sensitivity, and absence of standardized validation. The proposed CHEERS-AI reporting extension aims to address these gaps in reporting. In SLRs, AI tools have been applied to literature screening, data extraction, and risk-of-bias assessments, with consequent time savings. For example, open-source platforms like ASReview and proprietary systems based on machine learning algorithms have decreased human review efforts by 50% in some workflows. Nonetheless, adoption is constrained by concerns over reproducibility, algorithmic bias, and the lack of regulatory frameworks.
CONCLUSIONS: AI offers promise in HEOR workflows, but broader implementation requires transparent modelling, AI-specific reporting standards, and robust regulatory frameworks to ensure reliability and real-world impact. Recent Cochrane and NICE guidance reinforce responsible and ethical AI integration in evidence generation and reporting.
METHODS: PubMed, Google Scholar, and grey literature were searched to identify English-language studies published from January 2015 to May 2025. Studies were included if they evaluated AI’s application in economic modelling or evidence synthesis relevant to HEOR.
RESULTS: AI use was associated with substantial projected healthcare savings, estimated at US$200 to 360 billion annually in the United States (5-10% of the national health spending), with estimated cost reductions of 7-9% for private payers and 3-8% for physician groups. These were, however, majorly model projections and had no real-world validation. In economic evaluations, AI tools assist in semi-automated generation of decision trees and Markov models, with generative AI (e.g., GPT-4) used for health state definitions and ICER estimations. Despite these advancements, concerns persist regarding hallucination risks, black-box models, prompt sensitivity, and absence of standardized validation. The proposed CHEERS-AI reporting extension aims to address these gaps in reporting. In SLRs, AI tools have been applied to literature screening, data extraction, and risk-of-bias assessments, with consequent time savings. For example, open-source platforms like ASReview and proprietary systems based on machine learning algorithms have decreased human review efforts by 50% in some workflows. Nonetheless, adoption is constrained by concerns over reproducibility, algorithmic bias, and the lack of regulatory frameworks.
CONCLUSIONS: AI offers promise in HEOR workflows, but broader implementation requires transparent modelling, AI-specific reporting standards, and robust regulatory frameworks to ensure reliability and real-world impact. Recent Cochrane and NICE guidance reinforce responsible and ethical AI integration in evidence generation and reporting.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
MSR35
Topic
Health Policy & Regulatory, Methodological & Statistical Research, Real World Data & Information Systems
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas