Decision Panel Threshold Calculator (p

Not yet released in meddecide. This article documents decisionpanel, which is not part of the meddecide 1.0.4 release. It is documented here ahead of a future release, so the options, defaults and output described below are subject to change and the analysis will not appear in your jamovi menu yet. For the analyses that ship today see the meddecide articles index.

Purpose (Clinician-Friendly)

Use your own harm/benefit assumptions to calculate the threshold probability p_t at which you would switch from “no treatment” to “treat”. Then read the Decision Curve at that p_t.

Formula: p_t = harm_false_positive / (harm_false_positive + benefit_true_positive)

Tips: Only the ratio matters; you can measure “harm” and “benefit” in any consistent units (risk, cost, discomfort, etc.).

Quick Calculator

# Enter your assumptions
harm_false_positive   <- 1   # e.g., biopsy harm units
benefit_true_positive <- 4   # e.g., benefit units of early treatment

# Basic validation
stopifnot(harm_false_positive >= 0)
stopifnot(benefit_true_positive > 0)

# Threshold probability p_t = harm / (harm + benefit)
p_t <- harm_false_positive / (harm_false_positive + benefit_true_positive)

cat(sprintf("Recommended threshold p_t = %.2f (%.0f%%)\n", p_t, 100 * p_t))

Sensitivity Check (Optional)

# Explore how p_t changes if benefit varies
benefit_seq <- c(2, 3, 4, 6, 8)
pt_seq <- harm_false_positive / (harm_false_positive + benefit_seq)

result <- data.frame(benefit_true_positive = benefit_seq,
                     p_t = round(pt_seq, 2),
                     p_t_pct = paste0(round(100*pt_seq), "%"))
result

Next Steps

In jamovi, open the meddecide menu > Decision Panel > Decision Curve, set a threshold range that includes your p_t (e.g., 0.20), and pick the strategy with the highest net benefit at that p_t.
See also: decisionpanel_clinician_guide.md and START-HERE.md.

Visualizing p_t vs Benefit

benefit_seq <- seq(1, 10, by = 1)
pt_seq <- harm_false_positive / (harm_false_positive + benefit_seq)
df <- data.frame(benefit_true_positive = benefit_seq,
                 p_t = pt_seq)

if (requireNamespace("ggplot2", quietly = TRUE)) {
  library(ggplot2)
  ggplot(df, aes(x = benefit_true_positive, y = p_t)) +
    geom_line(color = "#2C7FB8", linewidth = 1) +
    geom_point(color = "#2C7FB8") +
    scale_y_continuous(labels = scales::percent_format(accuracy = 1)) +
    labs(x = "Benefit of a True Positive (units)",
         y = "Threshold probability p_t",
         title = "Threshold p_t decreases as benefit increases",
         subtitle = "p_t = harm_false_positive / (harm_false_positive + benefit_true_positive)") +
    theme_minimal(base_size = 12)
}

Visualizing p_t vs Harm (benefit fixed)

# Fix the benefit and vary harm
benefit_true_positive <- if (!exists("benefit_true_positive")) 4 else benefit_true_positive
harm_seq <- seq(0, 10, by = 1)
pt_seq_harm <- harm_seq / (harm_seq + benefit_true_positive)
df_harm <- data.frame(harm_false_positive = harm_seq,
                      p_t = pt_seq_harm)

if (requireNamespace("ggplot2", quietly = TRUE)) {
  library(ggplot2)
  ggplot(df_harm, aes(x = harm_false_positive, y = p_t)) +
    geom_line(color = "#D95F02", linewidth = 1) +
    geom_point(color = "#D95F02") +
    scale_y_continuous(labels = scales::percent_format(accuracy = 1)) +
    labs(x = "Harm of a False Positive (units)",
         y = "Threshold probability p_t",
         title = "Threshold p_t increases as harm increases",
         subtitle = paste0("Benefit fixed at ", benefit_true_positive, " units")) +
    theme_minimal(base_size = 12)
}

Decision Panel Threshold Calculator (p_t)