As part of predictive evidence-based medicine, in cooperation with our research partner we investigate patient-specific predictors for the efficacy of drugs in the field of schizophrenia using end-point-oriented cluster analyses. In addition to content issues, we analyze methodological approaches from a medical point of view to increase the quality and stability of prognosis. Our investigations focus on the development of automated methods for stable prognoses, especially in the areas of data transformation, automated variable assessment and the number of clusters used.