EVIDENCE-CENTERED DECISION INTELLIGENCE FOR HTA: A CLOUD-BASED MILP-MCDA OPTIMIZATION AND FRAGILITY FRAMEWORK FOR HIGH-VALUE THERAPIES
Author(s)
Athanassios Vozikis, Professor1, Thanasis Chalikias, M.Sc.2.
1Laboratory of Health Economics and Management (LabHEM), University of Piraeus, Economics Dept., Piraeus, Greece, 2Laboratory of Health Economics and Management (LabHEM), University of Piraeus, Economics Dept.,, Piraeus, Greece.
1Laboratory of Health Economics and Management (LabHEM), University of Piraeus, Economics Dept., Piraeus, Greece, 2Laboratory of Health Economics and Management (LabHEM), University of Piraeus, Economics Dept.,, Piraeus, Greece.
OBJECTIVES: As European health systems enter the EU Joint Clinical Assessment environment; payers need transparent tools to assess whether high-value medicines can be funded within fixed budgets. This study developed a cloud-based MILP-MCDA decision framework to evaluate patient access, budget headroom, and decision fragility for a hypothetical Greek Innovation Fund under a Traditional Cost Method.
METHODS: A four-year model was developed for 2026-2029, including phased entry of 15 oncology, rare disease, and advanced therapy products. Manufacturer prices were converted to net costs through sequential deductions: 8.74% statutory deduction, 70% rebate, and 5% invoice discount, yielding 26.01% of manufacturer price. A deterministic MILP maximized MCDA-weighted clinical utility subject to a €50.0 million statutory cap, €47.5 million effective cap, and 30% product-level cap. Fragility was assessed using 5,000-run Monte Carlo simulation, deterministic sensitivity analysis, threshold analysis, and EVPI/EVPPI screening.
RESULTS: The model funded all projected eligible patients and remained below both caps across all years. Optimized budget impact increased from €3.29 million in 2026 to €14.77 million in 2027, €29.99 million in 2028, and €32.51 million in 2029. Patient access increased from 42 to 904 patients. In 2028, 841 patients were funded across 15 therapies, leaving €17.51 million headroom versus the effective cap. Monte Carlo analysis showed strong stability, with 2028 mean budget impact of €31.01 million, CVaR95 of €32.73 million, and 0% probability of exceeding €50.0 million. Value-of-information screening estimated an EVPI of 50.62 net portfolio benefit units, with uncertainty uptake accounting for nearly all partial value of information (EVPPI = 48.77), compared with approximately 0.00 for cost and outcome uncertainty.
CONCLUSIONS: A cloud-based MILP-MCDA framework can provide an auditable and reproducible approach for HTA-oriented resource allocation and budget-governance analysis. Under heavy discount assumptions, affordability was confirmed and the main policy challenge shifted to patient-volume forecasting, uptake monitoring, and registry governance.
METHODS: A four-year model was developed for 2026-2029, including phased entry of 15 oncology, rare disease, and advanced therapy products. Manufacturer prices were converted to net costs through sequential deductions: 8.74% statutory deduction, 70% rebate, and 5% invoice discount, yielding 26.01% of manufacturer price. A deterministic MILP maximized MCDA-weighted clinical utility subject to a €50.0 million statutory cap, €47.5 million effective cap, and 30% product-level cap. Fragility was assessed using 5,000-run Monte Carlo simulation, deterministic sensitivity analysis, threshold analysis, and EVPI/EVPPI screening.
RESULTS: The model funded all projected eligible patients and remained below both caps across all years. Optimized budget impact increased from €3.29 million in 2026 to €14.77 million in 2027, €29.99 million in 2028, and €32.51 million in 2029. Patient access increased from 42 to 904 patients. In 2028, 841 patients were funded across 15 therapies, leaving €17.51 million headroom versus the effective cap. Monte Carlo analysis showed strong stability, with 2028 mean budget impact of €31.01 million, CVaR95 of €32.73 million, and 0% probability of exceeding €50.0 million. Value-of-information screening estimated an EVPI of 50.62 net portfolio benefit units, with uncertainty uptake accounting for nearly all partial value of information (EVPPI = 48.77), compared with approximately 0.00 for cost and outcome uncertainty.
CONCLUSIONS: A cloud-based MILP-MCDA framework can provide an auditable and reproducible approach for HTA-oriented resource allocation and budget-governance analysis. Under heavy discount assumptions, affordability was confirmed and the main policy challenge shifted to patient-volume forecasting, uptake monitoring, and registry governance.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
HTA116
Topic
Health Policy & Regulatory, Health Technology Assessment, Real World Data & Information Systems
Topic Subcategory
Decision & Deliberative Processes
Disease
Oncology, Personalized & Precision Medicine, Rare & Orphan Diseases