Modelling downstream effects in the presence of technological change.

Downstream effects are typically evaluated given current technology and current practice patterns rather than for technology and practice patterns that will be available at the time when downstream effects accrue. Where a relatively short time horizon can be expected to capture all relevant costs and effects, the current approach is unlikely to introduce substantial error into estimates of the costs and benefits attributed to an intervention; the estimates will remain valid so long as the context to which estimates relate remains unchanged. However, for longer time horizons, the magnitude of error associated with the current approach might be substantial. This paper describes three strategies for incorporating uncertainty associated with technological change into modeled economic evaluations: (i) discounting; (ii) within-trial analysis; and (iii) threshold/sensitivity analysis with horizon scanning. The appropriateness of each strategy for handling uncertainty associated with technological change is then considered under various possible situations defined over the characteristics of technological change (pace and whether technological change produces interventions that are dominant, cost increasing or cost saving) and the characteristics of downstream effects (proximity and the sensitivity of policy recommendations to their inclusion/exclusion). Selecting the appropriate strategy (or strategies) for the situation should permit estimation of more realistic upper and lower bounds around base-case estimates.