liveBook · Manning

Treatment threshold

As explained in Timeless Algorithms, The Foundational Ideas

A treatment threshold is the probability of a condition at which treating and not treating have equal expected utility.

A treatment threshold is the disease probability at which treating and not treating have equal expected utility. Above it, treatment is preferred; below it, not treating is preferred. section 7.1.4

Deriving the threshold

With the assigned utilities, treating has expected utility 110p − 10 and not treating has expected utility 100 − 100p. Equating them gives p = 110/210 = 0.5238, or 52.38%. section 7.1.4

Table 7.5 Second test results compared with the treatment threshold.
Second test result
Posterior probability of disease
Treatment threshold
Preferred action
Positive
0.484
0.5238
Do not treat
Negative
0.0055
0.5238
Do not treat

Compare each second-test posterior probability with the 52.38% threshold: neither result makes treatment preferable. section 7.1.4

How Timeless Algorithms, The Foundational Ideas covers it

How deep it goes
Derives the cutoff by equating expected utilities, then applies it to test results to assess whether the preferred action changes.
Examples
Uses a second disease test whose positive and negative results yield posterior probabilities of 48.4% and 0.55%.
What it assumes
Requires posterior disease probabilities and assigned utilities for treating and not treating.
What you can do after
Apply expected-utility comparisons to decide whether a posterior disease probability favors treatment.
How it connects
Connects expected utility, Bayesian updating, and value of information in a diagnostic testing decision.

This explainer was generated by Manning's knowledge graph from the text of Timeless Algorithms, The Foundational Ideas. The section links point to the parts of the book it draws on.