TAMING THE TORNADO CHART- SIMPLIFYING ONE-WAY SENSITIVITY ANALYSES FOR LARGE-SCALE MULTI-PARAMETER MODELS
Author(s)
O'Day K1, Mezzio D2
1Xcenda, LLC, Palm Harbor, FL, USA, 2Xcenda, LLC, Pleasant Hill, CA, USA
Presentation Documents
OBJECTIVES: One-way sensitivity analyses (OWSAs) quantify parameter uncertainty in health economic models by assessing the impact of individual model parameters on model outcomes and present the results in the form of a tornado chart. However, for models with numerous parameters, OWSAs can become cumbersome for model developers and difficult to interpret by end users. We present an alternative method to simplify OWSAs, where key intermediate calculated values are varied in the OWSA instead of the multiple primary inputs that are used to arrive at those calculated values. METHODS: A pharmacy budget impact model in Microsoft Excel, with an OWSA developed in Visual Basic for Applications (VBA), was used to compare the 2 methods of conducting an OWSA. The traditional OWSA includes all input parameters (eg, plan size, epidemiologic parameters, drug utilization by strength, DACON, drug unit costs, and market shares). The alternative simplified OWSA includes key intermediate calculated values, such as the target population size and net drug cost per month. The intermediate calculated values selected for the simplified OWSA should represent the main components of the model that are integral to calculating the final model outcomes. This method can be utilized with both budget impact and cost-effectiveness analyses. A worked example is provided, along with sample VBA code for conducting the simplified analysis and sample tornado charts to compare the 2 methods. CONCLUSIONS: The simplified OWSA, utilizing intermediate calculated values, offers an improved approach to assessing parameter uncertainty in large-scale models where testing all model inputs with the traditional OWSA approach would be unwieldy. Selecting key intermediate model parameters, in place of the primary inputs that feed into those calculated values, results in a more concise OWSA whose interpretation may be more accessible and relevant to healthcare decision makers.
Conference/Value in Health Info
2018-11, ISPOR Europe 2018, Barcelona, Spain
Value in Health, Vol. 21, S3 (October 2018)
Code
PCP47
Topic
Methodological & Statistical Research
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference
Disease
Multiple Diseases