survivalContour: visualizing predicted survival via colored contour plots

Area of Science:

• Biostatistics
• Data Visualization
• Machine Learning

Background:

• Survival analysis models offer flexibility but lack tools to visualize continuous covariate effects on survival outcomes.
• Existing methods often fail to adequately illustrate the impact of continuous predictors on time-to-event data.
• There is a need for intuitive graphical tools to interpret complex survival models.

Purpose of the Study:

• To introduce a novel colored contour plot for visualizing predicted survival probabilities over time.
• To demonstrate the utility of this visualization technique with both traditional and advanced survival models.
• To provide an accessible implementation of the proposed visualization tool.

Main Methods:

• Development of a colored contour plot to represent survival probabilities.
• Application of the contour plot to conventional survival models (Cox, Fine-Gray).
• Integration of the contour plot with machine learning models (random survival forests, deep neural networks).

Main Results:

• The colored contour plot effectively illustrates the influence of continuous covariates on survival outcomes.
• The visualization method is compatible with a range of survival analysis techniques.
• The tool provides enhanced interpretability for complex survival models, particularly those using machine learning.

Conclusions:

• The proposed colored contour plot is a valuable tool for enhancing the interpretability of survival analysis.
• This visualization method bridges the gap in illustrating continuous covariate effects in survival prediction.
• The associated R package and Shiny app facilitate the practical application of this novel technique.