The Appropriate Elicitation of Expert Opinion in Economic Models

Decision-analytic modelling has become an indispensible method to inform healthcare resource allocation decision making. Prospective randomized economic trials can produce internally robust cost and outcome data. However, such prospective data are typically unavailable or insufficient for the purpose of guiding national-level resource allocation decisions because results are not generalizable outside the trial study population or design. Using an economic model allows a decision analyst to carefully structure the research question and assimilate all available data. Additionally, outcome data from trials with short timeframes can be extrapolated to long-term estimates. Modelling also has a key role in prioritizing and planning future trials and research. Ideally, a model should be structured to represent all clinical pathways relevant to the decision problem. Methods guidelines state that data availability should not be a reason for rejecting a model or even an overriding factor when defining model structure. However, more often than not, the model cannot be populated using published, robust trial or observational data. Alternative data sources are then needed. Expert opinion is a recognised potential source of data and is often used to fill in data gaps and/or supplement the trial and observational data. Data elicited from experts is appropriately viewed as being second best and a potential source of parameter uncertainty. This is in part due to experts’ limited experience and perspective but also, importantly, because the methods used to identify the experts and elicit their opinion is often neither conducted robustly nor reported explicitly. Guidelines for the selection, design and population of economic models are necessary to encourage the transparent and explicit reporting of methods. Previous commentators have already noted that the methods used to elicit expert opinion for health technology assessments should also be reported explicitly. Existing methods guidelines do allude to the need for explicit reporting of expert opinion elicitation methods but do not go into methodological detail about the different elicitation methods available and their application. This commentary describes the potential use of the Delphi method as an elicitation technique. It clarifies that there are different types of Delphi that can be used, with quite distinct research and policy applications and also outlines where other expert elicitation methods might be more appropriate when focussing on the primary data requirements for economic modelling.