Deterministic Sensitivity Analysis

Deterministic Sensitivity Analysis (DSA) is a technique used in model-based evaluations to assess how changes in specific input parameters affect the model’s outcomes. This method involves manually altering one or more parameters, typically within a predefined range, and analyzing the impact of these variations on the results. The range of parameter variation often reflects the uncertainty reported in source studies, such as the 95% confidence interval for efficacy data from clinical trials or meta-analyses.

In DSA, two primary approaches are used:

– Univariate Sensitivity Analysis: One parameter is varied at a time to observe its individual effect on the model’s output.

– Multivariate Sensitivity Analysis: Multiple parameters are varied simultaneously to understand their combined effect on the outcomes.

The findings from DSA are commonly presented in visual formats like line graphs or bar charts. One notable representation is the ‘tornado chart,’ which displays a series of bar graphs from univariate sensitivity analyses, ordered by the extent of variation in the model’s output. The bars with the widest range of variation are positioned at the top, resembling the shape of a tornado.

While DSA is effective for examining the influence of a few parameters, it becomes impractical when dealing with more than four or five parameters simultaneously. In such cases, Probabilistic Sensitivity Analysis (PSA) is preferred, as it can handle the simultaneous variation of numerous parameters. Additionally, caution is advised when interpreting univariate sensitivity analyses, especially when input parameters are highly correlated. For instance, varying only one parameter in isolation, like sensitivity without adjusting specificity in diagnostic tests, might not provide an accurate picture of the overall uncertainty.

DSA helps identify the parameters that significantly influence the model’s outcomes, thus highlighting areas where more precise data could improve decision-making and model robustness.

Click HEOR TERMINOLOGY