A cross-sectional study published in Child and Adolescent Psychiatry and Mental Health identifies posttraumatic stress symptoms as the primary psychological predictor of non-suicidal self-injury among adolescents with a history of sexual trauma. The research, conducted in Iran, utilized both traditional statistical methods and machine learning techniques to analyze the complex interplay between various mental health factors and self-harm behaviors. The study involved 448 adolescents aged 14 to 18, recruited from psychology clinics in Tehran, providing a focused look at a vulnerable population where mental health challenges are particularly acute.
The investigators examined the individual and synergistic roles of posttraumatic stress symptoms, depression, anxiety, and sleep quality in predicting non-suicidal self-injury. Among the 448 participants, 39.7 percent, or 179 adolescents, reported engaging in non-suicidal self-injury. The sample was predominantly female, with 60.7 percent identifying as such, and all participants had confirmed histories of sexual abuse.
To assess these relationships, the researchers employed hierarchical logistic regression to determine the incremental contribution of each predictor to the likelihood of self-injury. The analysis revealed that posttraumatic stress symptoms were the strongest individual predictor, with an odds ratio of 1.07 and a p-value less than 0.001. This factor accounted for 17.9 percent of the variance in self-injury outcomes.
In addition to regression analysis, the study implemented a Random Forest classifier, a machine learning algorithm, to evaluate predictive accuracy and identify key risk factors. The model was trained on 70 percent of the data, comprising 313 participants, and tested on the remaining 30 percent, or 135 participants. The Random Forest classifier achieved an accuracy of 74.07 percent on the held-out test set, indicating a moderate level of predictive capability for identifying adolescents at risk for non-suicidal self-injury based on the measured psychological variables.
To enhance the interpretability of the machine learning model, the researchers used Permutation Importance and SHAP analyses. The analysis identified overall posttraumatic stress symptoms as the leading contributor to classification, with a mean accuracy decrease of 0.103 when the feature was permuted. Avoidance and numbing symptoms also showed significant importance, with a mean accuracy decrease of 0.080. In contrast, hyperarousal symptoms showed negligible importance, with a mean accuracy decrease of -0.002, suggesting that this specific symptom cluster was less useful for distinguishing between those who engaged in self-injury and those who did not.
SHAP analysis provided further insight into the magnitude of model predictions, revealing that sleep difficulties exerted a strong influence on the individual predictions made by the model. The mean absolute SHAP value for sleep difficulties was 0.076, indicating that while sleep quality may not be the primary driver of overall classification accuracy, it significantly impacts the probability assigned to individual cases.
The study’s findings emphasize the strong association between posttraumatic stress symptoms and non-suicidal self-injury in sexually traumatized Iranian adolescents. The results support the implementation of trauma-focused interventions that specifically address posttraumatic stress symptoms, potentially supplemented by sleep-related therapies given the significant influence of sleep difficulties on model predictions.
However, the cross-sectional design of the study limits the ability to establish causal relationships or temporal sequences between the psychological factors and self-injury. Furthermore, the study was conducted in a specific cultural and geographic context, Iran, which may limit the generalizability of the findings to other populations. Future longitudinal research is needed to clarify the temporal relationships between these psychological factors and non-suicidal self-injury, as well as to optimize culturally sensitive treatments for this vulnerable population.
Subject of Research: Psychological predictors of non-suicidal self-injury in sexually traumatized adolescents
Article Title: Psychological risk factors for non-suicidal self-injury in sexually traumatized adolescents: a cross-sectional analysis with random forest classification
Article References: Rasouli, A., Salehi, E., Raissi, M., & Mohajerin, B. (2026). Psychological risk factors for non-suicidal self-injury in sexually traumatized adolescents: a cross-sectional analysis with random forest classification. Child and Adolescent Psychiatry and Mental Health. https://doi.org/10.1186/s13034-026-01175-5
Image Credits: AI Generated
DOI: 10.1186/s13034-026-01175-5
Keywords: non-suicidal self-injury, posttraumatic stress, adolescent mental health, sexual trauma, random forest, machine learning, cross-sectional study, sleep quality, Iran, risk factors, Psychological, risk
News Source: Glenn Wilkins. (November 8, 2026). PTSS Key Predictor of Self-Harm in Traumatized Teens. Scienmag.



