Using Machine Learning to Identify Infant and Child Environmental and Biological Predictors of Callous-Unemotional Traits.

Callous-unemotional (CU) traits (i.e., blunted affect, low guilt) develop through the interplay of neurophysiological and environmental factors. However, critical gaps remain in understanding the relative importance of different physiological systems and early experiences over time. The goal of the current study is to identify which early-life biological and environmental features at which time points most strongly predict later CU traits in middle childhood. Using prospective longitudinal data from a rural community sample (N = 725; 48% female; 34% Black) and machine learning models, this study examined the relative predictive influence of biological stress systems (Sympathetic Nervous System (SNS), Hypothalamic-Pituitary-Adrenal (HPA) axis) and adversity indices (exposure to intimate partner violence (IPV), economic hardship, and lack of socially and cognitively stimulating toys or activities in the home) between 6 and 48 months on CU traits and conduct disorder (CD) at age 7. Models explained 7.3% of the variance in CU traits at 7 years. Exposure to IPV and sociocognitive resources across several time points, particularly in toddlerhood, emerged as influential predictors of later CU traits. Additionally, SNS functioning (i.e., salivary alpha-amylase) in early childhood was the most influential physiological predictor of CU traits. Prediction for CD was limited, with the final model explaining only 3.4% of the variability. Findings highlight the role of sympathetic regulation and early life experiences in shaping CU traits, providing important insights for the development of targeted interventions.