Development and Pilot Testing of a Standardized Simplified Modeling Tool for Health Technology Assessment Scoping

OBJECTIVES: After a topic has been chosen for assessment, early project scoping phases comprise literature review, stakeholder input, and PICOTS identification. We aimed to develop and pilot a semi-flexible, simplistic modeling tool for use during early scoping phases of health technology assessments.

METHODS: We developed a 3-state Markov model tool in Microsoft Excel^® accommodating various disease process pathways: episodic disease, progressive disease, and partitioned survival. Key model inputs included baseline demographics, transition probabilities, treatment effectiveness, health state utilities, and costs. Durations over which to apply cost and effectiveness benefits were included. Pilot testing included internal validation, feasibility testing via dependent reference model replication and de novo model building, performance testing via independent model replication, and assessment piloting. Development and testing spanned 5 primary models.

RESULTS: All models demonstrated acceptable internal validity. Dependent model replication generated cost-effectiveness spreads of ≤25% versus reference models in oncology and multiple sclerosis. De novo model development for two case studies revealed domains where greater flexibility is needed in a comprehensive model, namely treatment discontinuation and time-varying transition probabilities. Independent replication demonstrated cost-effectiveness spreads of <20% versus reference models, and suggested better reproducibility when replicating partitioned survival models (e.g., muscular dystrophy). Rapid exploratory modeling supporting formal technology assessment is planned in 2023.

CONCLUSIONS: Feasibility and performance testing suggested complex reference models could be reasonably collapsed into 3-state state Markov models without significant loss of fidelity for rapid exploratory purposes. Feasibility testing revealed novel insights surrounding structure conceptualization when building de novo rapid models where no reference model existed. Evidence supported increased complexity in downstream comprehensive model builds, such as treatment discontinuation, flexible parametric modifiers of treatment effects, and health state expansion. The tool demonstrated insufficient rigor to supplant a comprehensive economic evaluation. However, it may support rigorous full model builds and cross-collaborative teams conducting technology assessments.