Predicting relapse of problematic child-rearing situations

• Structured clinical judgement of child maltreatment did not perform better than chance.
• An actuarial risk classification was developed by means of a CHAID analysis.
• This classification might be promising because it performed better than chance.

BackgroundThe development and evaluation of risk assessment instruments for child maltreatment is still in its infancy, both in the Netherlands and internationally. The aim of this study was to examine the predictive validity of a structured clinical judgement instrument – the Check List of Child Safety (CLCS) – that is widely used in the Netherlands. The second aim was to examine the predictive validity of a newly developed actuarial risk classification that is based on variables measured with the CLCS.MethodThe sample consisted of 3963 Dutch families who were under the supervision of the Dutch Child Welfare Agency (CWA) between 2011 and 2013 because of problematic child-rearing situations. Relapse was defined as restarting treatment by the CWA because of newly substantiated problematic child-rearing situations. The actuarial risk classification was developed by means of a CHAID analysis. The predictive validity of the CLCS and the actuarial risk classification were examined by calculating several performance indicators (sensitivity, specificity, false positives, false negatives and AUC values).ResultsThe predictive validity of the CLCS was poor, with a non-significant AUC of .530, meaning that the CLCS performed no better than chance. The predictive validity of the actuarial risk classification was moderate, with a significant AUC of .630.ConclusionThe actuarial risk classification not only outperformed the CLCS, but is also time-saving in practice since it comprises only variables that are significantly related to relapse.