The probability of treatment success, failure and duration—what can be learned from empirical data to support decision making in clinical practice?

Abstract Empirical methods have been found to be superior to clinical judgment for the purpose of correctly identifying patients at risk for treatment failure and, hence, to enhance psychotherapy outcomes. The development and evaluation of an empirical approach aimed at supporting clinical decisions during the course of psychotherapy is described. The tool provides predictions based on a patient‐specific sampling strategy called the nearest neighbors method and on growth curve approaches to model an expected treatment course for each patient. Using session‐by‐session data from an outpatient center in the US ( N = 4365), this new empirically derived decision model was evaluated and compared with a clinically based approach loosely based on an adaptation of clinically significant change concepts. The empirically derived decision system was found to be superior to the rational clinically based one in almost all measures of prediction accuracy, indicating its potential to identify patients at risk for treatment failure. Copyright © 2006 John Wiley & Sons, Ltd.