A machine learning approach to identifying non-parental caregivers' risk for harsh caregiving towards infants in daycare centers

Background: Harsh Caregiving behavior amongst daycare providers (i.e., non-parental Harsh Caregiving) negatively impacts children's development across a variety of domains. As prevalences of non-parental Harsh Caregiving appear to increase worldwide, identifying its predictors is crucial for screening and intervention. Objective: The goal of this study was to identify a set of indicators and predictive rules that may accurately predict women's risk for Harsh Caregiving behavior in daycare environments. Participants and Setting: The study recruited 75 female non-parental caregivers, from the general population, who work with infants aged 0-1. Caregivers filled out self-report questionnaires including a Harsh Caregiving measure as well as a broad variety of potential predictors. Methods: To elucidate combinations of input variables that are predictive of non-parental Harsh Caregiving, we used machine learning Decision Three Inference and CHAID algorithms. Results: Study results revealed a predictive model including 27 questions and four different prediction paths. For example, the first path indicated that women who reported low levels of attention deficit and hyperactivity problems and low levels of rigid-negative caregiving philosophies, had 100 % chance to report low levels of Harsh Caregiving behavior. Overall classification accuracy for "High Harsh Caregiving behavior" was 95.2 %. Conclusions: After replication in larger samples, the model can be used as a screening tool for women expressing their wish to work with infants. Women at risk can either be declined employment or alternatively receive targeted supervision throughout their work with small infants.