Evaluating medical interview process components. Null correlations with outcomes may be misleading.

A standard criterion for judging the value of a medical interview process component, such as physician information giving, is the degree to which it correlates with positive outcomes, such as patient satisfaction, compliance, or symptom relief. There is a serious problem, however, with this use of correlations. Physicians and patients, like any human beings, give and respond to signals as to what each requires in the interaction. Any such appropriate responsiveness tends to attenuate (and may even reverse) the process-outcome correlation. Under optimum conditions, process components covary with patient requirements but not with outcomes. Thus, for expertly conducted interviews, the expected correlation of process components with outcomes approaches zero, even for components that are causally related to those outcomes.