Prediction of the severity of child abuse using nationwide survey data from Child Guidance Centers in Japan: focus on infancy and preschool age

The present study conducted a secondary data analysis of a comprehensive survey from Child Guidance Centers in Japan to identify factors that are associated with child abuse severity in infancy (0–3 years, 1,868 cases) and preschool age (4–6 years, 1,529 cases). A predictive model for abuse severity was developed.The data originated from a nationwide survey that was conducted in April 2013, consisting of details of abuse cases, including child characteristics, abuser attributes, and family situation. Abuse severity was assessed on a five-level scale (suspected, mild, moderate, severe, and life-threatening) that was converted into a binary outcome. Logistic regression analysis was used to create a predictive model using two-thirds of the data, which was validated with the remaining third of the data.As a result, in infancy, risks of severity increased with younger age of the abused child, physical abuse, neglect, witnessing domestic violence, and the involvement of Child Guidance Centers or hospitals in detection. The abuser's mental problems and cumulative child damage contributed to severity. For preschool age, similar factors applied, with additional risks that included abuse overlap and guardian separation. Cumulative abuser issues and child physical damage impacted severity. Validation yielded moderate prediction accuracy (areas under the curve: 0.703 and 0.714).