Increasing the information provided by probabilistic sensitivity analysis: The relative density plot

Area of Science:

• Health economics
• Decision science
• Statistical modeling

Background:

• Probabilistic sensitivity analysis (PSA) scatterplots suffer from overplotting and limited density insight.
• Traditional PSA visualizations hinder a comprehensive understanding of model uncertainty.

Purpose of the Study:

• Introduce the Relative Density plot (PSA-ReD) to overcome limitations of traditional PSA scatterplots.
• Enhance visualization and quantification of PSA results for improved model interpretation.

Main Methods:

• The PSA-ReD integrates density and contour plots to visualize PSA results.
• Relative density is mapped to cumulative probability, with contours highlighting specific probability regions.
• Demonstrated utility through two real-world case studies.

Main Results:

• PSA-ReD proves feasible and provides additional visual information beyond standard scatterplots.
• Color-coding identifies high-density areas, and contour plots quantify PSA iterations within the cost-effectiveness plane.
• The plot reveals model non-linearities and clarifies uncertainty distribution.

Conclusions:

• PSA-ReD is easily implemented and offers richer information from PSA data, preventing misinterpretation.
• Provides enhanced insight into model functioning and uncertainty surrounding cost-effectiveness estimates.