Presentation (CTI)

OBJECTIVES: Uncertainty in the cost-effectiveness of a new intervention is often evaluated in a one-way sensitivity analysis (OWSA) and displayed with a tornado diagram. This approach, however, does not work well when the resulting ICER bounds for at least one parameter (1) are both lower (or greater) than the base case ICER and/or (2) include both a positive and negative value. In these situations, interpretation of the tornado diagram can be difficult, and identification of the most influential parameters may be misleading. We propose a new graphical display - the CE Star - to supplement the traditional tornado diagram in the OWSA.
METHODS: The CE Star is formed by plotting the (ΔQALYs, ΔCost) points to the upper and lower bound values for each parameter in the standard OWSA and then connecting each pair of points from the same parameter. The relative importance of a parameter is determined by the Euclidean distance between these two points, with greater distance associated with greater importance. We compare the CE Star to the tornado diagram in a hypothetical example and from a published cost-effectiveness model.
RESULTS: In the hypothetical case, the tornado diagram and CE Star produced different rank orders of influential parameters. The CE Star was uniquely able to identify the importance of those parameters whose lower and upper bounds led to opposite ICER interpretations (e.g., cost-effective versus not cost-effective intervention). Further, the CE Star preserved the rank ordering of influential parameters under fixed relative changes to the parameter variability, while the tornado diagram did not. Application of the CE Star and tornado diagram in a published model revealed differences in the list of influential parameters.
CONCLUSIONS: When conducting an OWSA, the CE Star should be created to supplement the traditional tornado diagram to ensure appropriate identification and interpretation of influential parameters.