VALUE OF INFORMATION ANALYSIS FOR MANAGEMENT FOR INTERMITTENT CLAUDICATION: LLM-BASED RESULTS
Author(s)
Chen Dun, PhD, MS1, Andrew Wu, MD2, Caitlin Hicks, MD, MS2, Harold Lehmann, MD, PhD2;
1Baltimore, MD, USA, 2Johns Hopkins University, Baltimore, MD, USA
1Baltimore, MD, USA, 2Johns Hopkins University, Baltimore, MD, USA
OBJECTIVES: To demonstrate the utility of Value of Information (VOI) analysis in prioritizing research for intermittent claudication management and to assess the feasibility of using Large Language Models (LLMs) to accelerate decision model development. We applied these methods to evaluate the cost-effectiveness of Early Peripheral Vascular Intervention (PVI) versus conservative management (Supervised Exercise Therapy [SET], Medical Therapy, or Observation).
METHODS: A decision-analytic model was developed to simulate disease progression over a 24-month time horizon. Two LLMs were used to create the model from a graphic of the model plus tables of parameter definitions. The model compared four strategies: Early PVI, SET, Medical Therapy, and Observation. Transition probabilities, utilities, and costs were derived from literature, with uncertainties modeled using PERT (for probabilities and utilities) and Gamma (for costs) distributions A Probabilistic Sensitivity Analysis (PSA) with 2,000 Monte Carlo simulations was performed. The primary outcomes were Incremental Cost-Effectiveness Ratios (ICERs), Net Monetary Benefit (NMB), Expected Value of Perfect Information (EVPI) and Expected Values of Partial Perfect Information (EVPPI) to assess the value of reducing parameter uncertainty.
RESULTS: LLM-generated Python and R code produced identical validation results. In PSA, the Observation and Medical therapy strategy resulted in the highest expected NMB. The population-level EVPI was approximately $50 per patient. EVPPI analysis identified that uncertainty regarding natural history transition probabilities (progression from mild to moderate claudication) and PVI efficacy were the primary drivers of decision uncertainty. Conversely, uncertainty surrounding costs and baseline utilities contributed minimally to decision risk.
CONCLUSIONS: Early intervention for claudication is unlikely to be cost-effective compared to conservative management. The VOI analysis suggests that future research resources should be targeted specifically toward better defining natural disease progression rates rather than broad comparative effectiveness trials of interventions. This study further demonstrates that LLMs can significantly accelerate the execution of complex VOI analyses for medical decision making.
METHODS: A decision-analytic model was developed to simulate disease progression over a 24-month time horizon. Two LLMs were used to create the model from a graphic of the model plus tables of parameter definitions. The model compared four strategies: Early PVI, SET, Medical Therapy, and Observation. Transition probabilities, utilities, and costs were derived from literature, with uncertainties modeled using PERT (for probabilities and utilities) and Gamma (for costs) distributions A Probabilistic Sensitivity Analysis (PSA) with 2,000 Monte Carlo simulations was performed. The primary outcomes were Incremental Cost-Effectiveness Ratios (ICERs), Net Monetary Benefit (NMB), Expected Value of Perfect Information (EVPI) and Expected Values of Partial Perfect Information (EVPPI) to assess the value of reducing parameter uncertainty.
RESULTS: LLM-generated Python and R code produced identical validation results. In PSA, the Observation and Medical therapy strategy resulted in the highest expected NMB. The population-level EVPI was approximately $50 per patient. EVPPI analysis identified that uncertainty regarding natural history transition probabilities (progression from mild to moderate claudication) and PVI efficacy were the primary drivers of decision uncertainty. Conversely, uncertainty surrounding costs and baseline utilities contributed minimally to decision risk.
CONCLUSIONS: Early intervention for claudication is unlikely to be cost-effective compared to conservative management. The VOI analysis suggests that future research resources should be targeted specifically toward better defining natural disease progression rates rather than broad comparative effectiveness trials of interventions. This study further demonstrates that LLMs can significantly accelerate the execution of complex VOI analyses for medical decision making.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
EE81
Topic
Economic Evaluation
Topic Subcategory
Thresholds & Opportunity Cost, Value of Information
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory), STA: Surgery