Bayesian reasoning in medical contexts

This package includes a few functions to plot and help understand Positive and Negative Predictive Values, and their relationship with Sensitivity, Specificity and Prevalence.

• The Positive Predictive Value of a medical test is the probability that a positive result will mean having the disease. Formally p(Disease|+)
• The Negative Predictive Value of a medical test is the probability that a negative result will mean not having the disease. Formally p(Healthy|-)

The BayesianReasoning package has three main functions:

• PPV_heatmap(): Plot heatmaps with PPV or NPV values for the given test and disease parameters.
• PPV_diagnostic_vs_screening(): Plots the difference between the PPV of a test in a diagnostic context (very high prevalence; or a common study sample, e.g. ~50% prevalence) versus a screening context (lower prevalence).
• min_possible_prevalence(): Calculates how high should the prevalence of a disease be to reach a desired PPV given certain test parameters.

If you want to install the package can use: remotes::install_github("gorkang/BayesianReasoning"). Please report any problems you find in the Issues Github page.

There is a shiny app implementation with most of the main features available.

PPV_heatmap(Min_Prevalence = 1,
            Max_Prevalence = 1000, 
            Sensitivity = 100, 
            Max_FP = 2, 
            Language = "en")

PPV_heatmap(PPV_NPV = "NPV",
            Min_Prevalence = 800,
            Max_Prevalence = 1000, 
            Sensitivity = 80, 
            Max_FP = 5, 
            Language = "en")