Abstract
BACKGROUND: Risk assessment is crucial in preventing child maltreatment as it can identify high-risk cases in need of child protection intervention. Despite this importance, there have been no validated risk assessment instruments available in the Netherlands for assessing the risk of child maltreatment. Therefore, the predictive validity of the California Family Risk Assessment (CFRA) was examined in Dutch families who received family support. In addition, the added value of a number of experimental items was examined. Finally, it was examined whether the predictive value of the instrument could be improved by modifying the scoring procedure. METHODS: Dutch families who experienced parenting and/or child developmental problems and were referred by the Centres for Youth and Family for family support between July 2009 and March 2011 were included. This led to a sample of 491 families. The predictive validity of the CFRA and the added value of the experimental items were examined by calculating AUC values. A CHAID analysis was performed to examine whether the scoring procedure could be improved. RESULTS: About half of the individual CFRA items were not related to future reports of child maltreatment. The predictive validity of the CFRA in predicting future reports of child maltreatment was found to be modest (AUC = .693). The addition of some of the experimental items and the modification of the scoring procedure by including only items that were significantly associated with future maltreatment reports resulted in a 'high' predictive validity (AUC = .795). CONCLUSIONS: This new set of items might be a valuable instrument that also saves time because only variables that uniquely contribute to the prediction of future reports of child maltreatment are included. Furthermore, items that are perceived as difficult to assess by professionals, such as parental mental health problems or parents' history of abuse/neglect, could be omitted without compromising predictive validity. However, it is important to examine the psychometric properties of this new set of items in a new dataset.