Background: Suicidal behavior (SB) and non-suicidal self-injury (NSSI) are significant public health concerns among adolescents. The COVID-19 pandemic may have exacerbated these issues. Methods: This retrospective cohort study analyzed data from 341 adolescents (aged 6–18 years) presenting to a Romanian pediatric psychiatry emergency department during the years 2019 (pre-pandemic) and 2022 (post-pandemic). All participants underwent a thorough psychiatric assessment, and, together with their caregivers, were questioned on a wide range of potentially relevant issues, such as family, social, school, and life history factors. Logistic regression and random forest models were used to identify predictors of SB and NSSI. Results: SB was significantly predicted in regression models based on a prior suicidal ideation (OR = 68.410;
p < 0.001), having a parent living abroad (OR = 11.438;
p = 0.020), depression (OR = 6.803;
p < 0.001), and conflicts with peers (OR = 0.325,
p = 0.042), teachers (OR = 0.119,
p = 0.024), or both (OR = 0.166,
p = 0.012). The random forest model featured a slightly different order of the main predictors and highlighted the importance of additional predictors, such as prior suicide attempts, gender, and past non-suicidal self-injury. NSSI was mainly predicted by a history of self-harm (OR = 52.437;
p < 0.001), the number of comorbid psychiatric disorders (OR = 1.709;
p = 0.003), and conduct disorder (OR = 0.184;
p < 0.001), to which are added, according to random forest models, new predictors, such as borderline personality disorder, suicidal ideation, and school performance. Post-pandemic increases were observed in depression, suicidal ideation, and possible psycho-traumatic negative life event exposure. Conclusions: This study underscores the complex interplay of individual, familial, and societal factors influencing adolescent self-harm. Comprehensive interventions are needed, with early intervention crucial for those with a history of self-harm. Further research using prospective designs is recommended.
Full article