Regret graph calculator

Explanation

Scenario 1: No test available
You are considering a Treat/No Treat decision. In the absence of a test, the best decision depends on probability of disease and the "regret" associated with treatment decision errors. Regret is the difference in value between the action we took and the best action we could, in retrospect, have taken.

The relevant parameters are:
     C = regret associated with treating someone without the disease unnecessarily
     B = regret associated with failing to treat someone with the disease

The treatment threshold probability P_tt is where the expected regret associated with No Treat reaches the expected regret associated with Treat.

Scenario 2: Dichotomous test available
Add the possibility of doing a dichotomous test. Now you can

• withhold treatment without testing (No Treat)
• do the test and treat if positive (Test)
• treat without testing (Treat).

Two probability thresholds, the No Treat/Test and the Test/Treat threshold, divide the entire range of pre-test probabilities (0.0 to 1.0) into the 3 regions where each of these strategies is optimal. In addition to C and B above, the threshold probabilities depend on the cost and accuracy of the test and on the "regret" associated with treatment decision errors. Specifically, the additional parameters are:
     T = the cost or regret associated with doing the test
     Sensitivity = Probability of a positive test in disease P(T+|D+)
     Specificity = Probability of negative test in non-disease P(T-|D-).

For more details, see Chapter 2 of Evidence-Based Diagnosis.

View the derivation of the probability thresholds