Improving treatments for mental disorders using computational models

Abstract

Progress in the treatment of psychopathology has slowed and much remains unknown about how treatments achieve their beneficial effects. We propose that computational models can be used to provide new insights into how treatments may work and how they may be improved. We argue that treatments can be understood as interventions on systems of interacting components, and that computational models are needed if we are to accurately and precisely determine the effect an intervention will have on this system. We demonstrate this approach by using a computational model of panic disorder to conduct an in silico dismantling study of cognitive behavioral therapy. This simulated trial allows us to: identify a common source of treatment failure; propose a revised treatment protocol that mitigates this source of failure; and demonstrate that, if the model is accurate, this revised protocol will lead to improved treatment outcomes for 10% of patients. We conclude with a discussion of the promise and challenges of using computational models for treatment research.