OBJECTIVES: It is common for the uncertainty in an economic evaluation to be ‘tested’ using one-way sensitivity analysis. A common method for reporting such analysis is to present a ‘tornado diagram’. This research explores how useful this method is, and whether alternative approaches should be recommended.

METHODS: We examined the results sections of recent technology appraisals from the National Institute for Health and Care Excellence (NICE) in the United Kingdom. We recorded whether one-way sensitivity analysis was undertaken and, if so, how it was reported.

RESULTS: One-way sensitivity analysis was undertaken in all appraisals. In the vast majority of cases, the results were reported using tornado diagrams. In most cases, the ranges used were based on observed uncertainty for the parameter inputs (e.g. standard errors or confidence intervals) and in a small proportion, they were based on arbitrary ranges. However, in most cases, the diagram or table only showed the key output (e.g. incremental cost-effectiveness ratio or net health benefit) for the extreme values in the range, and no information was provided for intermediate values.

CONCLUSIONS: One-way sensitivity analysis is a tool to help decision makers understand the degree of uncertainty associated with model inputs, and how that uncertainty affects the model’s outcomes. Showing the results only for the extreme values deprives the decision maker of useful information. For example, where a 95% confidence interval is used, the decision maker is only able to see the impact of the changes for outcomes that have a 2.5% likelihood of occurring (i.e. at each end of the range). Instead, tornado diagrams could take a number of alternative approaches to show the impact of different values along the plausible ranges. This could include, for example, decile values, inter-quartile ranges and threshold points. This would allow decision makers to have more constructive discussions around the impact of uncertainty.