Eliciting priors to characterize uncertainties in decision analytic models
(English)Manuscript (preprint) (Other academic)
Background: Expert opinions are often used in decision models when evidence is scarce. This study describes the details of a formal elicitation exercise to estimate parameter values and their associated uncertainty and compares the results in term of cost-effectiveness and value of information with results from only eliciting mean values from experts.
Methods: Elicited distributions for 11 unknown parameters where incorporated into a previously developed cost-effectiveness model for prosthetic knees for amputees. The original model included elicited mean values for the missing values, thus ignoring any uncertainty across experts’ beliefs.
Results: The incremental cost-effective ratio (ICER) for the analysis based on the current elicited distributions was substantially higher (€13 625) than the ICER in the original analysis (€3 258). Even decision uncertainty, at a €35 000 threshold, increased significantly, increasing the value of further research from €355 100 in the original analysis, to €5 987 444 for the current elicited values.
Conclusions: Failing to account for the individual expert’s uncertainty might have a considerable impact on the result of cost-effectiveness analyses. Formal expert elicitation offers a plausible method to generate prior distributions representing the experts’ uncertainty and thereby more appropriately account for the true uncertainty of the decision.
Economics and Business
IdentifiersURN: urn:nbn:se:liu:diva-56593OAI: oai:DiVA.org:liu-56593DiVA: diva2:320603