Expert elicitation for long-term survival outcomes
This Technical Support Document (TSD) discusses how expert knowledge and uncertainty about long-term survival outcomes should be obtained and reported. This report will support health technology assessments in cases where appropriate data are lacking: where extrapolation is needed beyond the observed data. TSD 14 and TSD 21 have addressed model-based extrapolations, where it was noted that these can result in diverging long-term survival estimates, with significant implications for cost-effectiveness results. Both TSDs identified the role of expert judgement to support survival extrapolation. Here, we discuss how to obtain and use expert judgement.
The main methodological approach recommended is to elicit probability distributions from clinical experts, so that expert uncertainty is quantified, and experts are not merely asked to provide ‘best estimates’ or approve the clinical validity of pre-selected model-based survival extrapolations. There are various established protocols for eliciting probability distributions from experts in a structured manner, and such a protocol should form the basis of how the elicitation exercise is conducted. There are, however, aspects of the expert judgement task that are particular to survival extrapolation: the availability of data from which to extrapolate, and the way qualitative knowledge can be incorporated via the relationship between survivor and hazard functions. We discuss and illustrate how to modify a standard elicitation protocol accordingly. Software to support the elicitation process is also discussed.
We review the use of expert elicitation for long-term survival in health technology assessment as well as the broader literature and observe that the majority of NICE technology appraisals have not used structured expert elicitation. We set out recommendations for best practice and discuss future research directions for this methodology.
This Technical Support Document (TSD) explores how expert knowledge and uncertainty about long-term survival outcomes should be obtained and reported to support health technology assessments when data is insufficient for extrapolation.
Downloads
TSD 26: Expert elicitation for long-term survival outcomes