PURPOSE: Highlight and discuss relationships between model validation and uncertainty analysis. Develop suggestions on whether, and how, to keep these aspects separated.

DESCRIPTION: Model validation and uncertainty analysis are closely related. Both are recommended in Guidelines for the Economic Evaluation of Health Technologies. ISPOR good practice recommendations and tools for implementation are available. However, the interrelationship between both aspects has not been fully explored.

Uncertainty analysis is performed assuming health-economic models and input data are valid. In contrast, model validation cannot be performed without paying attention to model uncertainty. This holds especially true for validating model outcomes against empirical data. An important question is how best to weigh these two aspects into policy decisions and assess how they impact decision (un)certainty.

The workshop starts with an overview of the problem by the moderator, an expert in uncertainty and value-of-information analysis. (5 minutes-CR)

An explanation of common validation elements, e.g. external and face validity, will follow. Various types of uncertainty (structural, methodological, and parameter uncertainty) will be discussed and connected to different elements of validation. This provides a common terminology for the rest of the workshop. (15 minutes-TF)

We point out existing good practice guidelines, focusing on operationalising model validity in relation to parameter and decision uncertainty. This will be highlighted using examples of economic evaluation of anti-cancer medication. (15 minutes-SG)

Finally, we discuss how aspects of model validity and uncertainty have influenced reimbursement decisions in the UK. (15 minutes-ICR)

AUDIENCE INTERACTIVE ELEMENT

Between-presentations questions are permitted (5 minutes). The audience participates in an exercise with the purpose of identifying how model validation and uncertainty can be operationalised and weighted in healthcare decision-making. Speakers present examples of different validation/uncertainty elements expected to influence policy recommendations. The audience will judge these examples before/after the presentations (by e-voting). Results will be discussed (10 minutes-CR moderates).