Methods, apparatus, and systems are provided for efficiently and accurately identifying individuals at risk for adverse effects from psychotropic or CNS-active treatment. Changes in a brain activity indicator (e.g. EEG cordance) are used to predict the adverse effects of treatment based on an experimentally derived cutoff value. For example, a reliable biological indicator is provided with high predictive capability for identifying, very early in the course of treatment (e.g. <=48 hours after start of treatment), those individuals who are at greatest risk for worsening suicidality and other adverse effects of antidepressant drugs.