Potholes in the road to prediction

The task of making predictions constantly challenges those of us who treat and study the mentally ill. On the clinical front, we are required to predict which patients will respond to various therapies, which patients will need hospitalization, which will kill themselves or harm others. Fortunately, experience, as quantified in empiric base rates, helps with these predictions if we maintain a clear head and go with the base rates rather than rely on clinical hunches. It is a fact that actuarial prediction is nearly always better than clinical judgment (Grove, 1986; Meehl, 1954). The clinician's task becomes one of evaluating the patient so as to decide which base rates apply. For many years we have admired the clear and critical thinking that Norman Garmezy (1968, 1974) has brought to our task; most often we agreed with his formulations and have every reason to believe that the processes of influence have been mutual.

The treatment of mental illness, though still leaving much to be desired, has made great strides, and scientists’ efforts are turning toward new goals. In the last decade, much research in mental illness has focused, for the first time, on trying to predict mental illness in individuals not yet affected. Achieving the power of prediction may point the way to the eventual goal of designing rational interventions that will prevent future mental illness in those people identified to be at risk. Success will depend, in part, on the wisdom and endurance of scientists engaged in this brand of research. Success will also depend, in large part, on the nature of the illness.