Background: The restraining of children is a pervasive but controversial practice that has resulted in the injury and death of children. Despite this threat of harm, little research has explored what factors contribute to the risk of a child becoming injured during a restraint event. Objective: This study examined multiple child and restraint factors to better understand what predicts the injury of a child during a restraint event. Methods: Longitudinal data (794 youth, 13,339 restraint events) from six residential treatment centers in the Midwest were utilized to predict injury associated with restraint using nested hierarchical logistic regression models. Results: Injuries to the child are a frequent outcome following the use of restraints on a child. Findings show that a child being male and older contributed to the likelihood of an injury, but the race of the child did not predict injury. In addition, three physical hold types--prone, supine, and settle--were associated with a greater likelihood of injury. Conclusions: Results suggest those who use restraints on children should reevaluate their behavior management plans to reduce their risk of hurting children by restraints. Eliminating or greatly reducing the use of prone, supine, and settle holds will likely lead to a drop in injuries of children.