Multimethodological approach in psychiatric predictor research.

The discussion of some methodological problems of predictor studies led to the following conclusions: 1. Different outcome criteria result in a different classification of patients as responders or non-responders. 2. Predictor variables and outcome criteria should be better defined. 3. The clinical relevance of a result can be estimated by calculating the proportion of explained variance. 4. Cross validation of results is necessary. 5. Changes during a test therapy of 10 days result in a relatively reliable prediction of outcome in acute schizophrenia and depression.