Profiles of Harsh Caregiving Behavior Among Non-Parental Caregivers in Daycare Centers: Associations with Infant Attachment Security and Machine Learning-Based Identification of Predictors

This study is the first to use Latent Profile Analysis (LPA) to identify profiles of different types of harsh caregiving behaviors among non-parental caregivers and examine their association with infants' attachment security. Additionally, machine learning methods were applied to develop predictive screening models for the different profiles of harsh caregiving. The study recruited 134 female non-parental caregivers, working with infants aged 0–1 year, who completed self-report questionnaires assessing various types of harsh caregiving, infants’ attachment security, and a broad range of potential predictors of harsh caregiving. The results of the study identified three distinct profiles: "Overall Low Harshness"; "High Physical and Emotional Neglect"; and "High Physical and Emotional Roughness." Infants from caregivers in the Overall Low Harshness profile showed significantly higher levels of attachment security compared to infants from caregivers in the other two profiles. Machine learning models predicted women’s risk of belonging to one of the two High Harshness profiles with 89.4% accuracy and distinguished between High Neglect and High Roughness with 83.0% accuracy. These findings highlight the prevalence of both neglectful and rough caregiving types in non-parental childcare, as well as their association with lower levels of infant attachment security. Screening models indicate the opportunity to screen for risk before caregivers are employed, and thereby the potential for guiding both employment decisions as well as after-employment follow-up and assistance.