OBJECTIVES : Deterministic one-way sensitivity analyses (OWSA) reported graphically in Tornado diagrams are a mainstay in present-day cost-effectiveness analyses. The intent is to vary relevant parameters in a comparable manner in order to rank their relative influence on results. Discount rates for costs and effects are surprisingly often included, but rather than varying them according to a pre-specified distribution comparable to that of other parameters, they are tested at arbitrary lower and upper bounds relative to their base case value.

METHODS : During a targeted literature review, cost-effectiveness publications from 2015 to present were retrieved from PubMed on April 1^st 2019. Requirements for inclusion were presence of a Tornado-OWSA and deployment of a 10-year or longer time horizon. No disease area or condition was pre-specified.

RESULTS : 431 abstracts were retrieved. In total, 303 were excluded (97 at abstract screening) because they did not meet the inclusion criteria. Of the 128 articles included, 53 (41.4%) did not include discount rates in the OWSA, whereas another 14 (10.9%) did include discount rates but not as part of the Tornado analysis. The remaining 61 (47.7%) included discount rates in the Tornado analysis.

Most frequently (26/61), the discount rate was varied to specific lower and upper bounds (zero to five or six percent) and contrasted to the base case value (three percent). A number of similar, alternative bounds were used too, reflecting country-specific guidelines on discounting.

CONCLUSIONS : In nearly all 61 analyses where discount rates were included as part of the Tornado OWSA, they were deemed one of the most influential parameters. This is an expected outcome given the variation of typically ±100% compared to ±20% or similar for other parameters. This leads to an over-estimation of the discount rates, and makes it difficult to compare across studies since this is common, but not standard practice.