Evaluation of interventions for children exposed to violence is critical for informing dissemination of effective services nationally, but retention rates in such studies are typically low, jeopardizing intent-to-treat analytic approaches. We examine predictors of retention (demographics, violence exposure, child mental health, caregiver demographics, and engagement in intervention) in a large multisite national study of interventions for children exposed to violence. Our analyses found that families were more likely to be retained in the study at 6 months if caregivers rated their own physical health as poor or fair, the child experienced higher levels of maltreatment, or the family was assigned to the intervention group within the study. In models limited to those families in the study's intervention group, only receipt of services was significantly related to study retention. Methods with which to increase retention in family therapy programs for at-risk families are discussed.