Patterns of childhood trauma co-occurrence and its predictivity for suicidality: A machine learning approach

Childhood trauma (CT) is a significant public health concern. The relationship between different CT patterns and suicidality remains largely undiscovered. The current study aims to identify distinct CT overlapping patterns and to explore their association with suicidality in a comprehensive multivariable context. Participants were 23,721 young adults with CT experiences. Three CT overlapping patterns were identified including the highest neglect group, highest abuse group, and low CT group with Gaussian mixture model. Then a random forest model was used to predict suicidality with the CT patterns alongside 47 comprehensive predictors. Shapley additive explanation (SHAP) analysis showed that CT overlapping patterns emerged as the sixth most important predictor. Furthermore, the predictivity of CT patterns for suicidality decreased among participants with high levels of non-suicidal self-injury (NSSI), depression, anxiety, post-traumatic stress disorder (PTSD), and obsessive-compulsory disorder (OCD), or with low self-compassion. This research emphasizes studying the relationship between CT patterns and suicidality with comprehensive multivariable approaches.