Predicting Maltreatment in Adolescents with Mentally Ill Parents: A Random Forest Tree Analysis

Children of mentally ill parents are at a higher risk of developing mental health disorders and experiencing maltreatment. Identifying predictors of maltreatment in this risk group may help provide support at an earlier stage. A random forest classifier was applied to a sample of psychiatric inpatients (n = 330, 59.70% female, M _Age = 41.15 years, SD _Age = 7.12) and their children (n = 95, 60.00% female, M _Age = 15.03 years, SD _Age = 2.14) to examine whether child maltreatment could be predicted based on reports from both parents and children. Maltreatment symptoms were assessed using a cut-off score from the Childhood Trauma Questionnaire Short-Form. Model 1, which predicted children's trauma scores as estimated by the parents, achieved 76.62% accuracy with an area under the curve (AUC) of .85. Model 2, which used only parental data to predict children's self-assessed maltreatment scores, showed a slightly lower accuracy of 68.42% and an AUC of .69. Model 3, which predicted children's maltreatment scores based solely on children's self-reports, showed an accuracy of 73.68% and an AUC of .84. These findings indicate moderate-to-good predictability of children's maltreatment scores and provide initial insights into the risk-assessment of children's maltreatment in families with a mentally ill parent.This study was preregistered as a clinical trial (28.04.2017, DRKS00011533) at the Deutsches Register Klinischer Studien (DRKS).