The threshold represents a tradeoff
A threshold probability reflects how the consequences of acting unnecessarily compare with the consequences of missing a relevant case. Different decisions can justify different thresholds, even for the same predicted risk.
The clinically relevant range should therefore come from the decision context, not from whichever part of a graph makes a model look best.
A small arithmetic example
At a threshold of 20%, the false-positive weight in a common net-benefit formulation is 0.20 divided by 0.80, or 0.25. With 20 true positives and 20 false positives among 100 people, net benefit is 0.20 minus 0.20 × 0.25, giving 0.15.
This is an illustrative calculation, not a treatment recommendation or a measured health gain.
Compare realistic alternatives
Decision curves often include strategies such as acting for everyone or no one. The appropriate comparator and action must fit the clinical question.
A favorable curve depends on valid predictions, suitable data, and defensible preferences. It does not demonstrate that users follow the recommendation correctly, that an intervention is available, or that deployment improves outcomes. Those require additional evidence. Net benefit adds a decision-oriented perspective without replacing evaluation of the actual care process.
