OBJECTIVES

Temporal uncertainty in cost-effectiveness modelling arises when the analysis time horizon exceeds the observed evidence end point. Quantifying the consequences of such uncertainty is often problematic. This study uses structured elicitation of experts’ beliefs to characterise temporal uncertainty in the treatment effect, using a case study of a multifaceted podiatry intervention designed to reduce the rate of falls and fractures in the elderly.

METHODS

An elicitation exercise was designed and conducted to inform the temporal change in the effect of the podiatry intervention on the rate of falls and fractures. The elicited change in treatment effect was applied to the results of a randomised trial, REFORM, that reported at 1-year. The exercise was delivered using an online tool, designed to be completed remotely and independently to encourage experts’ participation.

RESULTS

The study recruited 38 experts (geriatricians, general practitioners, physiotherapists, nurses, and academics) from across the UK. The elicited priors implied that uncertainty around the treatment effect, and thus the predicted outcomes, increased over time - the 95% confidence interval around the predicted rate of falls was 1.4-1.6 falls one year after starting treatment, compared to 0.7-1.6 falls after five years. The majority of experts believed that the treatment effect would depreciate over time, although beliefs varied across professions - geriatricians were least certain that the treatment effect would change, while nurses were the only experts who believed the treatment effect could potentiate. The differences could reflect the heterogeneity in their role in treating patients at risk of falling.

CONCLUSIONS

The elicitation task was somewhat difficult for experts to complete and did rely on assumptions in order to make the exercise deliverable under the time and budget constraints. Nevertheless, the study demonstrates expert elicitation as a plausible method for informing uncertainty around the temporal change in the treatment effect, and illustrates the associated methodological challenges.