Reading focus
Syllabus: How Should Psychotherapy Adapt to the Person and Context?
Track the difference between a generally favorable prognosis and a comparative treatment advantage.
Why you’re reading it. Build the concepts needed to evaluate a proposed matching rule.
A question to consider. A characteristic predicts improvement equally in two therapies. Why does that fail to identify which therapy to choose?
Why this reading
Knowing who is likely to improve is different from knowing which treatment would help them more. Cohen and DeRubeis explain that distinction and the evidence needed to turn a prediction into a useful decision.
Before you read
A prognostic factor predicts outcome; a treatment-effect moderator identifies a difference in the comparison between treatments. Overfitting occurs when a model captures peculiarities of its development data that do not travel to new cases.
Read for
- Distinguish prediction within one treatment from prediction of a comparative advantage.
- Follow the requirements for independent validation, including decisions made while selecting predictors.
Consider afterwards
- If a model predicts poor outcomes under both treatments, what has it established about which to choose?
- What would a prospective evaluation add beyond accurate prediction in an existing dataset?
Editorial context
This 2018 review concerns depression and a developing research program. It does not establish a routinely available optimal-treatment calculator. Predicted alternatives cannot both be observed for the same person, and performance can change across services. The authors treat decision tools as contributions to informed choice, with uncertainty and practical costs still to consider.