Predicting/preventing child abuse: value of utility maximizing cutting scores

Any standardized method for identifying cases of likely child abuse requires specification of a cutting score (or scores) on a predictor variable. In this paper, we describe two criteria for determining cutting scores--utility maximizing (UtilMax) and error minimizing (ErrMin)--and we demonstrate that UtilMax is often the superior, and never the inferior, criterion. Two types of ErrMin cutting scores, true and artificial, are distinguishable based on whether realistic or artificial base rates are used to find the cutting score. Since studies often compute artificial ErrMin cutting scores, these scores must be modified to produce true ErrMin cutting scores. UtilMax cutting scores are explained and a numerical example is presented to show that maximizing utility is the preferable criterion in that it optimizes the balance between the costs of incorrect decisions and the benefits of correct decisions. The example also illustrates how UtilMax cutting scores help one to decide whether attempting to predict abuse would be worthwhile or not.