REPRESENTING UNCERTAINTY OF EXPERT OPINION IN HEALTH TECHNOLOGY ASSESSMENT THROUGH THE USE OF SUBJECTIVE PROBABILITIES
Author(s)
Grigore B*;Peters J;Hyde C, Stein K University of Exeter Medical School, Exeter, United Kingdom
OBJECTIVES: Where data is sparse, expert opinion may be used to populate decision analytic models. Elicitation techniques which capture uncertainty associated with expert opinion (i.e. elicit probability distributions) have analytic benefits over obtaining single values from experts but may be difficult to carry out and the methodological implications of different techniques are poorly understood. To assess current use of elicitation techniques which incorporate experts’ uncertainty, we conducted a methodological review of the use of elicitation in economic models. METHODS: The review was systematic in approach. Electronic databases (MEDLINE, EMBASE and CRD) were searched using an explicit strategy. Studies describing their approach to elicitation were included. Data was abstracted on pre-defined aspects of the elicitation technique. The synthesis was narrative. RESULTS: Twelve articles were included. The median number of experts per study was five (range 3-23). In eleven studies (91.7%), experts were asked individually. Only one study reported a group elicitation exercise. The most frequently used approaches to elicitation were the histogram technique and the bisection method. Elicitation was used to inform effectiveness (for treatments), performance (for diagnostic tests) or disease progression parameters, with only one study eliciting resource use. The elicited parameters were in the form of a proportion in seven studies, time to event in four studies, rate in three and relative risk in two studies. Across these studies, there was heterogeneity in both approach and reporting of what we considered to be key aspects of the elicitation process. CONCLUSIONS: Judged by the studies identified in the review, elicitation can be useful and is valued. Efforts should be made to improve reporting of elicitation in health technology assessments, with the aim of not only of improving the description of individual studies, but also to contribute towards establishing accepted good practice in elicitation in the context of decision analytic modeling.
Conference/Value in Health Info
2013-05, ISPOR 2013, New Orleans, LA, USA
Value in Health, Vol. 16, No. 3 (May 2013)
Code
PRM87
Topic
Methodological & Statistical Research
Topic Subcategory
Modeling and simulation
Disease
Multiple Diseases