Predictors of Attrition in the Pathways to Desistance Study: Testing for Differential Salience Across Time

Abstract

There is limited research examining predictors of attrition in the Pathways to Desistance study. There is also a paucity of research examining time-varying effects of predictors of attrition. Changes in characteristics across a study period may lead to the emergence of increases/decreases in attrition risk across time. The Pathways to Desistance dataset (11 waves) was analyzed. Cox proportional hazard modeling was used to examine effects of time-stable and time-varying predictors of attrition in four domains: sociodemographic, mental health, behavioral, and psychosocial. The proportional hazards assumption was tested to determine if effects of predictors varied. Results indicated that sociodemographic and psychosocial predictors were most relevant for understanding differential attrition. Effects of age, peers, and gender on attrition risk varied across time.