Independent contribution of emotion dysregulation to suicide attempts in adolescents beyond the role of depressive symptoms Authors: Massimo Apicella, Maria Pontillo, Gino Maglio, Alessandra Minutolo, Giulia Serra, Giorgia Della Santa, Elisa Andracchio, Giulia D’Amario, Milena Labonia, Roberto Averna, Stefano Vicari Published: 2026 ABSTRACT AbstractEmotion dysregulation (ED) is increasingly recognized as a transdiagnostic factor in adolescent psychopathology and suicidal ideation and behavior. This study examined the association between ED and recent suicide attempts (SAs) in a clinically referred adolescent sample (n = 182; mean age = 15.70 ± 1.44 years; 80.8% female). We compared adolescents with recent SAs (n = 41) to psychiatric controls (n = 141) on diagnostic and dimensional variables, including the RIPoSt–Y, a novel measure of ED. Adolescents with SAs had higher rates of major depressive episodes (MDEs), depressive symptoms, and significantly elevated RIPoSt–Y interpersonal sensitivity and affective instability. Sequential logistic regressions revealed that both MDE diagnosis and interpersonal sensitivity independently predicted recent SAs, with the latter showing incremental predictive value beyond depressive symptoms. Mediation analysis using structural equation modeling indicated that interpersonal sensitivity had both a significant direct effect and a smaller indirect effect mediated by MDE and depressive symptoms. The total model explained 32.7% of the variance. Findings support the role of interpersonal sensitivity as a dispositional vulnerability of adolescent suicidal behavior, partially mediated by depression. This is the first study to demonstrate the predictive value of RIPoSt–Y dimensions for SAs in youth, underscoring the clinical importance of ED assessment in suicide risk evaluation.Keywords: Adolescents, Suicide attempts, Emotion dysregulation, RIPoSt-Y, Depression, Interpersonal sensitivitySubject terms: Diseases, Health care, Medical research, Psychology, Psychology Introduction Suicide represents a major cause of mortality in youth and is a priority public health issue1. Suicidal ideation and behaviors are common in adolescent populations2 and are associated with significant morbidity and adverse mental health outcomes in adulthood3. Following the recent COVID-19 pandemic, there has been a significant increase in access to emergency departments for suicide attempts (SAs)4,5. Research suggests that the majority of individuals who die by suicide have a co-occurring psychiatric diagnosis6,7. Studies on clinical populations of children and adolescents who died by suicide indicate that depression was the most frequent diagnosis (approximately 40% of cases) followed by oppositional defiant disorder/conduct disorder, attention-deficit/hyperactivity disorder, and bipolar disorder8. Affective disorders are the most frequent diagnosis associated with suicidal behavior9. Beyond psychiatric diagnoses, a complex interplay of individual and social factors must be considered to explain suicidal behaviors. At an individual level, a complex interaction between distal (predisposing), developmental (mediating), and proximal (precipitating) factors is considered crucial for understanding the heterogeneous and complex phenomenon of death by suicide. Psychopathology, particularly depressed mood and hopelessness, is among the most important proximal factors10. Developmental factors, conceptualized as cognitive styles and temperamental/personality traits that mediate between distal and proximal factors10, are of utmost importance for child and adolescent psychiatrists, given that these factors are relevant in developmental age. Among these, a combination of impulsive-aggressive behaviors and anxiety is commonly implicated as a significant factor for suicide risk11, yet it is still unsatisfactorily captured by current diagnostic systems. Severe clinical presentations in children and adolescents with marked irritability, outbursts, and mood swings may variably and more or less satisfactorily fit into diagnostic categories such as unipolar/bipolar mood disorders, disruptive behavior disorders, Attention Deficit Hyperactivity Disorder (ADHD), and Disruptive Mood Dysregulation Disorder (DMDD)12,13. In this context, research on the transdiagnostic construct of Emotion Dysregulation (ED) is of increasing importance and has been a focus of research for decades14. A widely used definition of ED describes it as an impaired ability to regulate emotions at different levels, encompassing the awareness and acceptance of emotions, the ability to control impulsive behaviors and act in accordance with desired goals when experiencing negative emotions, and the flexible use of situationally appropriate emotion regulation strategies. The Difficulties in Emotion Regulation Scale (DERS)15, widely used in psychiatry, was developed for adults based on this formulation. However, despite a growing consensus regarding the importance of ED in psychopathology, the field has not yet reached a unified definition of the construct16. ED is a transdiagnostic factor, present across various Diagnostic and Statistical Manual of Mental Disorders (DSM) categories, including both internalizing and externalizing disorders17. It is a multidimensional construct, encompassing emotional reactivity and impulsivity, affective intensity and polarity, and behavioral self-control—concepts that are not always comprehensively and completely captured by the current literature on youth16,18, thus underscoring the importance of further research on its conceptualization and correlation with outcomes. A novel and promising instrument in this field, recently validated for the assessment of ED in adolescents, is the Reactivity, Intensity, Polarity and Stability questionnaire–youth version (RIPoSt–Y)19. RIPoSt–Y is based on Hantouche’s20 description of ED as an extension of specific affective temperaments. It is consistent with current understanding of early-onset mood disorders and suicidal behaviors by conceptualizing a neurobiological basis (e.g., dysregulation of the hypothalamic-pituitary-adrenal axis and imbalances in the prefrontal-amygdala circuitry involved in emotional top-down control)21 that may influence psychopathology from developmental stages. It has shown promise in assessing a wide range of ED components in adolescents, demonstrating a good capacity to discriminate between adolescents with ADHD, mood disorders, ADHD-mood disorder comorbidity, and healthy controls19. ED has already been shown to be implicated in suicidal and non-suicidal self-injurious (NSSI) behaviors22–24. However, not all studies agree on the importance of the role of ED in suicidal behaviors, as several have indicated that its predictive power becomes less significant when controlling for other psychological variables25, including depression26. This highlights a relevant gap in the field and it has been recommended future studies on the relationship between ED and suicide consider the impact of key confounders like depression25. The present study seeks to bridge this gap by using a new, developmentally validated measure of ED, the RIPoSt–Y, designed for adolescent population. This instrument explores dimensions of ED that may be particularly relevant to risk factors for SAs in adolescents, offering a complementary and more nuanced perspective. Our research is also grounded in a developmental perspective, treating RIPoSt–Y measured ED as a dispositional risk factor that precedes and influences more proximal states like current depression and recent suicide behavior. Given that previous longitudinal studies have demonstrated that ED can predict the development of depressive symptoms in adolescence27, our conceptual model positions ED as a predisposing developmental risk factor for suicidal behaviors, with current depression/depressive symptoms acting as mediating variables. In line with a developmental psychopathology framework28, we conceptualize ED as a developmental risk factor that precedes formal psychiatric onset (e.g. current depressive symptoms). We performed the present study with the following specific aims: Aim 1: To measure which dimensions of ED, as assessed by the RIPoSt–Y, significantly differ between adolescents with recent suicidal behaviors and other severe psychiatric patients within a large clinical sample, identifying specific aspects of ED that may be particularly relevant to suicidal risk. Aim 1: To measure which dimensions of ED, as assessed by the RIPoSt–Y, significantly differ between adolescents with recent suicidal behaviors and other severe psychiatric patients within a large clinical sample, identifying specific aspects of ED that may be particularly relevant to suicidal risk. Aim 2: To assess an independent effect of ED on suicidal behavior controlling for current DSM diagnoses, including depression, and for depressive symptoms in multivariate analyses. We also aim to measure incremental predictive value of dimensional ED symptoms (as measured by the RIPoSt–Y) on recent suicidal behaviors, over and above the predictive power of categorical DSM diagnoses and dimensional depressive symptoms (measured by the children's depression inventory-2, CDI-2). Aim 2: To assess an independent effect of ED on suicidal behavior controlling for current DSM diagnoses, including depression, and for depressive symptoms in multivariate analyses. We also aim to measure incremental predictive value of dimensional ED symptoms (as measured by the RIPoSt–Y) on recent suicidal behaviors, over and above the predictive power of categorical DSM diagnoses and dimensional depressive symptoms (measured by the children's depression inventory-2, CDI-2). Aim 3: To examine a mediational model in which we assess the direct effect of ED on suicidal behaviors, and its indirect effect mediated by current depression and depressive symptoms. Consistent with the aforementioned theoretical background, we hypothesized that the dispositional vulnerability represented by ED would predict SAs, both directly and indirectly, by exacerbating acute depressive states. Aim 3: To examine a mediational model in which we assess the direct effect of ED on suicidal behaviors, and its indirect effect mediated by current depression and depressive symptoms. Consistent with the aforementioned theoretical background, we hypothesized that the dispositional vulnerability represented by ED would predict SAs, both directly and indirectly, by exacerbating acute depressive states. Methods Participants The sample comprised consecutive adolescents referred for the first time to the Child and Adolescent Neuropsychiatry Unit of the Bambino Gesù Children’s Hospital IRCCS between May 1st and November 30th, 2024. Participants were adolescents referred to a tertiary referral center due to the severity of their clinical presentation. To obtain a sample enriched with a high prevalence of recent suicide attempts (SAs) and symptom severity, we consecutively enrolled patients accessing the psychiatry inpatient ward and an outpatient clinic which provides high-intensity psychiatric care to patients with mood disorders and other severe psychopathology requiring intensive treatment (including severe mood or behavioral disorders requiring timely stabilization, severe suicidal ideation, acute psychomotor agitation, severe non-suicidal self-injury, or acute psychotic symptoms). Patients primarily admitted for medical stabilization due to eating disorders were treated in a separate medical service and were not included in the present sample. Inclusion criteria were intentionally broad to allow for an estimation of the transdiagnostic value of ED across a wide range of diagnoses, reflecting current clinical practice with a sample of adolescents at high risk of suicidal behavior. We included (a) adolescents aged 12 to 17 years (inclusive) and (b) those with adequate comprehension of the Italian language to allow for the administration of self-report tests and who completed the assessment with the measures considered in the study. We excluded patients (a) with an intelligence quotient (IQ) below 70, and (b) those with autism spectrum disorder or schizophrenia according to Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR) criteria. These conditions were excluded to ensure the validity of self-report measures, as formal thought disorders or neurodevelopmental social-communication deficits could confound the assessment of ED and interpersonal sensitivity (RIPoSt-Y was developed and validated to assess ED in affective and behavioral disorders). Two authors (AM and GM) independently reviewed the clinical records and consulted a third author (MA) to resolve minor discrepancies; patients were included only after reaching diagnostic agreement. This study received approval from our Institutional Review Board (Comitato Etico Ospedale Pediatrico Bambino Gesù IRCCS) with practice number 3195. All data were processed in accordance with relevant Italian data protection laws. Parents/legal guardians provided written informed consent for anonymous reports of data in aggregate form, in compliance with research ethics. Assessment All participants were evaluated by an experienced child and adolescent psychiatrist and an experienced psychologist using the Kiddie-Schedule for Affective Disorders and Schizophrenia for School-aged Children, Present and Lifetime version (K-SADS-PL)29. The K-SADS-PL is a semi-structured, clinician-administered diagnostic interview conducted with both the adolescents and their caregivers to assess current and past psychopathological features and psychiatric disorders according to the DSM-5 criteria. A current episode was defined based on the period of maximum symptom severity within the month preceding the interview. The Clinical Global Assessment Scale (C-GAS)30, a clinician-rated tool evaluating overall psychosocial functioning over the preceding two weeks, was used to assess the patients’ functional impairment. The K-SADS-PL and C-GAS are considered gold standards for diagnostic assessment in youth, with inter-rater reliability often exceeding 0.70 in naturalistic clinical settings31,32. Suicidal ideation and suicidal behaviors were systematically evaluated using the Columbia Suicide Severity Rating Scale (C-SSRS) screening module, followed by the complete present/lifetime version if two or more items were positive on the screening module33. The C-SSRS has demonstrated robust psychometric properties in international adolescent populations and meta-analytic reviews, with synthesized Cronbach’s alpha values ranging from 0.76 to 0.88 for its subscales and overall scores, and excellent inter-rater reliability (Kappa coefficients ranging from 0.67 to 1.00)34,35. Patients were categorized into the “recent SA” group if they presented with a suicide attempt or an interrupted attempt within the last month; the remaining patients were referred to as Psychiatric Controls (PC). A suicide attempt, according to Posner et al.33, is defined as any self-harming behavior resulting in any form of physical damage with a non-zero intent to die, as reported by the patient or evident from documented circumstances. An interrupted attempt was defined as a potentially self-injurious act that was initiated but stopped by another person before any physical injury could occur. ED was assessed using the Reactivity, Intensity, Polarity and Stability questionnaire–youth version (RIPoSt–Y)This instrument has been validated in community and clinical samples of Italian adolescents19. The RIPoSt-Y is a 31-item self-report questionnaire designed to assess stable dispositional dimensions of ED in adolescents over a general time frame (e.g., ‘how you usually feel’). It includes four subscales: Affective Instability (e.g., ‘I have the impression that my emotions follow cycles with ups and downs’), Emotional Reactivity (e.g., ‘I have quick, brutal, emotional reactions, almost impulsive ones’), and Interpersonal Sensitivity (e.g., ‘I am very sensitive to criticism’). Each item is scored on a 6-point Likert scale from 1 (never) to 6 (always). The RIPoSt-Y has shown excellent psychometric properties in its original validation, with a Cronbach’s alpha value higher than 0.8 for each subscale; each subscale also showed significant convergent validity with the Youth Self Report for ages 11 to 18 years (YSR-11/18) Dysregulation Profile19. The RIPoSt-Y was selected for its capacity to capture relatively stable emotional reactivity patterns, as evidenced by its high test-retest reliability even over extended periods19. This aligns with its strong correlation with temperamental measures like the Cyclothymic-Hypersensitive Temperament Questionnaire (CHT-Q)19. Depressive symptoms were assessed using the Children’s Depression Inventory 2 (CDI-2)36, a widely used self-administered test for depressive symptoms in adolescent populations. The CDI-2 is a 27-item self-report instrument that evaluates the frequency of depressive symptoms over the past two weeks. Items are presented in a three-option multiple-choice format, with scores ranging from 0 to 2 for each response. In the present analysis, the total T-score was used as a measure of depressive symptom intensity. It has high internal consistency, typically ranging from 0.82 to 0.91 in clinical samples of adolescents37. In contrast to the RIPoSt-Y, the CDI-2 assesses clinical symptoms occurring in the past 2 weeks. History of childhood trauma and maltreatment was assessed through clinical records, focusing on Adverse Childhood Experiences (ACEs), including abuse, neglect and household dysfunction38. Current Non-Suicidal Self Injury – Disorder (NSSI-D) diagnosis has been defined according to DSM-5 section III criteria. Statistical analysis Categorical variables were presented as counts and percentages. Continuous variables were presented as means and standard deviations (SD). Differences in the frequencies of categorical variables between the recent SA group and the remaining sample were analyzed using Pearson’s chi-square test. For our aim 1, mean differences in continuous variables between the two groups were analyzed using Student’s t-test for independent samples. Cohen’s d for mean differences and Cramer’s V for differences in frequencies were reported as measures of effect size (ES). For our aim 2, based on the results of preliminary bivariate analyses, categorical diagnoses (current DSM-5 diagnoses according to the K-SADS-PL) and dimensional symptoms (CDI-2 for depression and RIPoSt–Y subscales for ED) that showed significant differences between the recent SA and PC groups were selected for inclusion in subsequent regression models. Sequential multivariable logistic regressions were performed to examine the association between categorical diagnoses, depressive symptoms, ED, and membership in the recent SA group. The effects of diagnoses, symptoms, and transdiagnostic dimensions on recent SA were examined by steps. In Step 1, a model was specified to examine the main effects of selected diagnoses on recent SA. In Step 2, the independent effects of selected diagnoses and depressive symptoms on recent SA were tested. In subsequent steps, selected RIPoSt–Y subscales were added to the model in a stepwise forward manner. For each predictor in the models, the estimated coefficient (β), the odds ratio (OR), and the corresponding p-value are presented. Model fit was assessed using the − 2 log-likelihood (-2LL) and changes in -2LL (Δ − 2LL), with the significance of model improvement evaluated using the chi-square test (p Δχ²). Nagelkerke’s R² was reported for each model as a measure of explained variance. Multicollinearity was assessed using Variance Inflation Factors (VIFs). For our aim 3, the effect of relevant ED dimensions, selected from previous steps for having significant independent effects on recent SA, was further studied using mediation analysis. Based on the theoretical framework of developmental and proximal risk factors for suicide risk9, selected ED subscales were entered as exogenous variables, and their direct effect on recent SA group belonging and indirect effect through selected categorical diagnoses and depressive symptoms (entered as mediators) were tested39. The mediation model was constructed positioning ED as the distal factor and recent psychopathology as the mediator (proximal factor). This directionality was chosen to reflect the theoretical background and also for psychometric consistency (the RIPoSt-Y measuring ED asks for general and usual experiences and has high test-retest reliability after months whereas the CDI-2 and K-SADS-PL measure current clinical dimensions). Structural equation modeling (SEM) was used, estimated using maximum likelihood. Binary outcomes were modeled using probit links to account for their dichotomous nature, and ORs are provided in the text to facilitate ES interpretation for binary outcome relationships. To account for potential non-normality in the data and to obtain robust estimates of the indirect effects and their confidence intervals (CIs), a bias-corrected and accelerated (BCa) bootstrap procedure with 5000 resamples was employed. Unstandardized coefficients for each path are presented, along with their BCa 95% CIs. The significance of the direct and indirect effects was determined by examining whether the 95% BCa bootstrap CI excluded zero. All statistical analyses were performed using StataNow software (version 18.5, [https://www.stata.com]; StataCorp LLC, College Station, TX, USA). Statistical significance was set at p < 0.05. Results Bivariate analyses on psychiatric diagnoses and symptoms Between May 1st and November 30th, 2024, 182 consecutive adolescents meeting inclusion criteria (mean age = 15.70 ± 1.44 years; 80.8% females) accessed our inpatient/high-intensity outpatient service and completed the psychological assessment. For subsequent analyses, 22.5% of the sample (n = 41) were classified into the “recent SA” group due to reported recent actual suicide attempts or interrupted attempts, while the remaining participants were classified into the “psychiatric controls (PC)” group. The mean age at assessment was slightly higher in the recent SA group compared to the PC group (16.54 ± 1.05 years vs. 15.45 ± 1.45 years, p < 0.001); there were no significant differences in the proportion of females between the two groups. Patients in the recent SA group required inpatient treatment more frequently than those in the PC group (80.5% vs. 31.9%, p < 0.001) and exhibited significantly greater impairment in general functioning (C-GAS: 49.34 ± 6.77 vs. 55.85 ± 8.77, p < 0.001) compared to PCs. Among DSM-5 psychiatric diagnoses and comorbidities, a current Major Depressive Episode (MDE) was the only diagnosis significantly associated with recent SA (70.7% vs. 35.5%; p < 0.001), with no significant differences observed between bipolar and unipolar MDEs. Other diagnoses included Generalized Anxiety Disorder (13.7%), ADHD (13.7%), adjustment disorder (9.3%), social anxiety disorder (6.0%), DMDD (5.5%), oppositional defiant or conduct disorder, Post-Traumatic Stress Disorder (PTSD), Anorexia Nervosa (4.4% each), which had similar rates among groups (p > 0.05). No significant differences were found between adolescents with and without SAs regarding the prevalence of ACEs (41.5% vs 47.5%; p = 0.494). Regarding pharmacological treatment at the time of first contact with our services and evaluation, the SA group showed a higher prevalence of antidepressants use (36.6% vs. 12.1%; p < 0.001) and second-generation antipsychotics use (73.2% vs. 47.5%; p = 0.004). No significant differences were found for lithium, mood stabilizers, or benzodiazepines. Regarding transversal symptoms, patients with recent SA presented significantly higher levels of ED, as measured by the RIPoSt–Y affective instability subscale (71.56 ± 13.53 vs. 57.40 ± 21.90, p < 0.001) and the RIPoSt–Y interpersonal sensitivity subscale (34.93 ± 6.80 vs. 28.16 ± 8.13, p < 0.001). They also exhibited higher depressive symptoms (CDI-2 T-score: 76.03 ± 11.85 vs. 67.20 ± 13.42, p < 0.001) compared to PCs. For further details on psychiatric diagnoses and symptoms in the recent SA and PC groups, please refer to Table 1. Table 1Description of the sample. Legend: SA: suicide attempt; PC: psychiatric controls; ACE: Adverse Childhood Events; BD: bipolar disorder; MDE: Major Depressive Episode; NSSI-D: Non Suicidal Self-Injury Disorder; RIPoSt–Y: Reactivity, Intensity, Polarity and Stability questionnaire–youth version; CDI-2: Children’s Depression Inventory 2Recent SA (n=41)PC(n=141)Total sample (n=182)p (statistic)Effect Size Age at intake16.54 ± 1.0515.45 ± 1.4515.70 ± 1.44<0.001(t=-4.47)0.84Females85.4% (n=35)79.4% (n=112)80.8% (n=147)0.502(χ2=0.72)0.06Psychiatric inpatient treatment80.5% (n=33)31.9% (n=45)42.9% (n=78)<0.001 (χ2=30.60)0.41C-GAS49.34 ± 6.7755.85 ± 8.7754.38 ± 8.78<0.001 (t=4.22)0.83Lifetime ACE41.5% (n=17)47.5% (n=67)46.2% (n=84)0.494(χ2=0.47)0.05Pharmacologic treatment at admissionAny 80.5% (n=33)58.2% (n=82)63.2% (n=115)0.009 (χ2=6.81)0.19Antidepressants 36.6% (n=15)12.1% (n=17)17.6% (n=32)<0.001 (χ2=13.20)0.27First Generation Antipsychothics2.4% (n=1)3.5% (n=5)3.3% (n=6)0.727 (χ2=0.12)-0.03Second Generation Antipsychotics73.2% (n=30)47.5% (n=67)53.3% (n=97)0.004 (χ2=8.40)0.22Benzodiazepines29.3% (n=12)17.7% (n=25)20.3% (n=37)0.106 (χ2=2.61)0.12Other mood stabilizing drugs9.8% (n=4)7.1% (n=10)7.7% (n=14)0.573 (χ2=0.32)0.04Lithium34.1% (n=14)23.4% (n=33)25.8% (n=47)0.167 (χ2=1.91)0.10Methylphenidate2.4% (n=1)2.8% (n=4)2.7% (n=5)0.891 (χ2=0.02)-0.01DSM-5 DiagnosesMajor Depressive Episode70.7% (n=29)35.5% (n=50)43.4% (n=79)<0.001 (χ2=16.09)0.30Of which with BD diagnosis7.7% (n=15)12.4% (n=15)11.2% (n=18)0.419 (χ2=0.65)0.06Of which MDE With Mixed Features12.2% (n=5)16.3% (n=23)15.4% (n=28)0.628 (χ2=0.41)0.05Manic Episode0.0% (n=0)0.7% (n=1)0.5% (n=1)1.000 (χ2=0.29)0.04Substance Use Disorder4.9% (n=2)2.1% (n=3)2.7% (n=5)0.343 (χ2=0.90)0.07Current NSSI-D17.1% (n=7)18.4% (n=26)18.1% (n=33)0.842 (χ2=0.04)-0.02Symptom ScalesRIPoST-Y Affective Instability71.56 ± 13.5357.40 ± 21.9060.59 ± 21.13<0.001(t=-3.92)0.77RIPoST-Y Emotional Reactivity25.98 ± 9.0824.09 ± 9.5324.52 ± 9.440.262(t=-1.13)0.20RIPoST-Y Interpersonal Sensitivity34.93 ± 6.8028.16 ± 8.1329.68 ± 8.33<0.001(t=-4.86)0.90CDI-2 total T-score76.03 ± 11.8567.20 ± 13.4269.06 ± 13.56<0.001(t=-3.64)0.69 Description of the sample. Legend: SA: suicide attempt; PC: psychiatric controls; ACE: Adverse Childhood Events; BD: bipolar disorder; MDE: Major Depressive Episode; NSSI-D: Non Suicidal Self-Injury Disorder; RIPoSt–Y: Reactivity, Intensity, Polarity and Stability questionnaire–youth version; CDI-2: Children’s Depression Inventory 2 Multivariate analysis of psychiatric diagnosis and symptoms for prediction of recent SA Based on the preceding bivariate analyses, which indicated that, among diagnoses, a MDE was associated with recent SA, and, among dimensional symptoms, depressive symptoms (CDI-2 total T-score) and two dimensions of ED (RIPoSt–Y interpersonal sensitivity and affective instability) were significantly associated with recent SA, we performed multivariate analyses for our aim 2. Specifically, to examine the incremental predictive value of psychological scales beyond the diagnosis of current MDE in predicting recent SA, we conducted sequential logistic regressions. The initial model (Model 1), which included only the diagnosis of MDE as a predictor, was significant (β = 1.34, OR = 4.23, p < 0.001). Subsequently, in Model 2, we added quantitative depressive symptoms (CDI-2 total T-score) to evaluate its additional predictive contribution. Both MDE diagnosis (β = 1.07, OR = 3.30, p = 0.009) and CDI-2 total T-score (β = 0.05, OR = 1.05, p = 0.007) were significant predictors. Finally, Model 3 included, in addition to the diagnosis of MDE and depressive symptoms, the Interpersonal Sensitivity scale of the RIPoSt–Y. Both MDE diagnosis (β = 0.94, OR = 2.55, p = 0.026) and RIPoSt–Y Interpersonal Sensitivity (β = 0.08, OR = 1.08, p = 0.012) were significant predictors. The initial model (Model 1) showed a -2 log-likelihood of 168.55 and a Nagelkerke’s R² of 0.106. In Model 2, the addition of quantitative depressive symptoms resulted in a significant improvement in model fit (Δ -2 Log-likelihood = 8.38, df = 1, p = 0.004) and an increase in Nagelkerke’s R² to 0.174, indicating an enhanced proportion of variance explained. Finally, in Model 3, adding the Interpersonal Sensitivity scale of the RIPoSt–Y as a measure of ED further improved the model fit (Δ -2 Log-likelihood = 6.75, df = 1, p = 0.009), with Nagelkerke’s R² increasing to 0.226. The subsequent addition of the Affective Instability scale of the RIPoSt–Y in Model 4 (the full results of which are not presented in the table for conciseness), despite being significantly associated with recent SA in previous bivariate analyses, did not lead to a significant improvement in the overall model fit compared to Model 3 (Δ -2 Log-likelihood = 0.223, p = 0.637). Multicollinearity was not a concern, with Variance Inflation Factors (VIFs) of 2.01 or lower (2.01 for RIPoSt–Y Affective Instability, 1.97 for RIPoSt–Y Interpersonal Sensitivity, 1.59 for CDI-2 total T-score, and 1.12 for MDE diagnosis). For further details on Models 1, 2, and 3, please refer to Table 2. Table 2Sequential logistic regression study of the independent effect of diagnoses and symptoms on recent SA. Note: Model 1: R²= 0.106, -2 Log-Likelihood = 168.55. Model 2: R²= 0.174, -2 Log-Likelihood = 160.17, Δ -2 Log-Likelihood = 8.38, p (Δ χ²) = 0.004. Model 3: R²= 0.226, -2 Log-Likelihood = 153.42, Δ -2 Log-Likelihood = 6.75, p (Δ χ²) = 0.009. Legend: SA: suicide attempt; MDE: major depressive episode; ED: emotion dysregulation; OR: odds ratio; RIPoSt–Y: Reactivity, Intensity, Polarity and Stability questionnaire–youth version; CDI-2: Children’s Depression Inventory 2.Independent variable: Recent SAModel 1: MDEModel 2: MDE and depressive symptomsModel 3: MDE, depressive symptoms and EDPredictorsβORpβORpβORp MDE diagnosis 1.344.23< 0.0011.073.300.0090.942.550.026 CDI-2 total T-score -0.051.050.0070.031.030.168 RIPoST-Y Interpersonal sensitivity --0.081.080.012 Sequential logistic regression study of the independent effect of diagnoses and symptoms on recent SA. Note: Model 1: R²= 0.106, -2 Log-Likelihood = 168.55. Model 2: R²= 0.174, -2 Log-Likelihood = 160.17, Δ -2 Log-Likelihood = 8.38, p (Δ χ²) = 0.004. Model 3: R²= 0.226, -2 Log-Likelihood = 153.42, Δ -2 Log-Likelihood = 6.75, p (Δ χ²) = 0.009. Legend: SA: suicide attempt; MDE: major depressive episode; ED: emotion dysregulation; OR: odds ratio; RIPoSt–Y: Reactivity, Intensity, Polarity and Stability questionnaire–youth version; CDI-2: Children’s Depression Inventory 2. After confirming the significant and independent effect of MDE diagnosis and RIPoST-Y Interpersonal sensitivity on recent SA, we controlled for the potential effect of age and sex in a successive logistic regression model, using MDE diagnosis, RIPoST-Y Interpersonal sensitivity, age and sex (with male set as reference) and their interaction as predictors of recent SA. RIPoST-Y Interpersonal sensitivity (β = 0.098; OR = 1.103; 95%CI 1.035–1.175; p = 0.003), MDE diagnosis (β = 1.073; OR = 2.925; 95%CI 1.248–6.858; p = 0.014), remained significant and independent predictors. Age was a significant predictor too (β = 1.750; OR = 5.755; 95%CI 1.109–29.86; p = 0.037), while female sex (p = 0.158) and age-sex interaction (p = 0.153) had non-significant effects. Additionally, we tested a parsimonious model including only Interpersonal Sensitivity as a predictor. In this unadjusted model, the subscale remained strongly associated with recent SAs (OR = 1.12; 95% CI 1.06–1.19; p < 0.001), indicating that this trait carries a robust risk signal even when considered independently of the clinical contribution of depression. To evaluate the robustness of our primary findings, we conducted a sensitivity analysis by simultaneously entering additional potential confounders into the logistic regression model, including current NSSI-D, history of ACEs, current substance use disorder (SUD), and previous pharmacotherapy. In this full model, RIPoSt-Y Interpersonal Sensitivity consistently remained a significant and independent predictor of recent SAs (OR = 1.09; 95% CI [1.02–1.17]; p = 0.009). Similarly, the presence of a current MDE diagnosis maintained its significance (OR = 2.36; 95% CI [1.01–5.49]; p = 0.046), whereas CDI-2 total scores (OR = 1.03; p = 0.133) and all other added variables—NSSI-D (OR = 0.37; p = 0.084), past pharmacotherapy (OR = 1.79; p = 0.222), history of ACEs (OR = 0.71; p = 0.423), and current SUD (OR = 2.02; p = 0.528)—did not reach statistical significance. Mediation analysis For aim 3 of the present study, we performed mediation analysis using SEM to estimate the direct effects of the relevant ED dimension (interpersonal sensitivity) on recent SA, and indirect effects mediated through the relevant diagnosis (MDE). Given that current quantitative depressive symptoms (CDI-2 total T-score) also emerged as having a relevant effect on recent SA and are logically connected to an MDE diagnosis, we also included CDI-2 total T-score in the model as a mediator, assuming a shared covariance with current MDE diagnosis based on theoretical grounds (Fig. 1). Direct and indirect effects of RIPoSt-Y interpersonal sensitivity on SA trough major depressive episode diagnosis/depressive symptoms. The SEM demonstrated a reasonable fit, explaining 32.7% of the combined variance. The CDI-2 total T-score had the highest explained variance (R² = 28.8%), followed by recent SA (R² = 14.0%) and MDE diagnosis (R² = 7.4%). The model revealed significant mediation pathways: The total effect of RIPoSt–Y interpersonal sensitivity on recent SA was statistically significant (β = 0.31, 95% BCa CI [0.21, 0.41], p < 0.001). The total effect of RIPoSt–Y interpersonal sensitivity on recent SA was statistically significant (β = 0.31, 95% BCa CI [0.21, 0.41], p < 0.001). The direct effect of RIPoSt–Y interpersonal sensitivity on recent SA, after accounting for the mediators, was significant (β = 0.21, 95% BCa CI [0.04, 0.38], p = 0.021). An OR of 1.09 (95% CI [1.01, 1.17]) was calculated for this effect, indicating that each one SD increase in interpersonal sensitivity scores (corresponding to 8.33 points) is associated with a direct 9% greater odds of recent SA. The direct effect of RIPoSt–Y interpersonal sensitivity on recent SA, after accounting for the mediators, was significant (β = 0.21, 95% BCa CI [0.04, 0.38], p = 0.021). An OR of 1.09 (95% CI [1.01, 1.17]) was calculated for this effect, indicating that each one SD increase in interpersonal sensitivity scores (corresponding to 8.33 points) is associated with a direct 9% greater odds of recent SA. The indirect effect of RIPoSt–Y interpersonal sensitivity on recent SA was significant (β = 0.11, 95% BCa CI [0.02, 0.19], p = 0.038). The indirect effect of RIPoSt–Y interpersonal sensitivity on recent SA was significant (β = 0.11, 95% BCa CI [0.02, 0.19], p = 0.038). The model supported partial mediation, with approximately two-thirds of the effect of ED on recent SA operating through direct pathways and one-third mediated indirectly. As expected, the assumed error covariance between MDE diagnosis and CDI-2 total T-score was also significant (β = 0.17, p = 0.017), indicating shared unmeasured factors (psychopathology specific to mood disorders) influencing both. The direct effect of CDI-2 total T-score on recent SA was not significant in the model, while the effect of MDE diagnosis was significant. An OR of 1.16 (95% CI [1.008, 1.314]) can be calculated for this effect, indicating that an MDE diagnosis increases the odds of recent SA by 16%. For further details on the model, please refer to Fig. 1. Discussion This study examined ED, measured by the RIPoSt–Y, in a large clinical sample of adolescents referred to a tertiary care center. We aimed to determine the transdiagnostic impact of ED on SA, considering DSM diagnoses and dimensional depressive symptoms. Our key findings were: both affective instability and interpersonal sensitivity, two ED dimensions, were significantly higher in patients with recent SA than in PCs, with large effect sizes (0.77 and 0.90, respectively). As expected9, MDE and depressive symptoms were also more prevalent in adolescents with recent SA. Importantly, RIPoSt–Y interpersonal sensitivity independently predicted recent SA, even after controlling for MDE and depressive symptoms, and improved SA prediction more than depressive symptoms alone. In Logistic Regression Model 3, MDE diagnosis had an unstandardized coefficient of 0.94 (OR = 2.56), indicating increased odds of recent SA. Interpersonal sensitivity had a smaller but significant unstandardized coefficient of 0.08 (OR = 1.08); considering the scale’s parameters in our sample (mean = 29.7 ± 8.3, range = 5–46), if we consider the effect of an increase of one SD (8.33 points), the associated OR would be approximately 1.90, indicating that adolescents scoring one SD above the mean in interpersonal sensitivity have a 90% higher likelihood of having attempted suicide. Furthermore, approximately a quarter of the variance in recent SA was explained by the combination of categorical diagnosis and dimensional symptoms, highlighting their relevance. Consistent with the literature40, SA peak age in adolescents is after 15 years, so age significantly predicted recent SA in the logistic regression in our sample, but MDE and ED retained independent effects. We found no significant sex differences, however the high proportion of females in our high-intensity care services, which may have limited our power to detect such effects. Finally, mediation analysis supported a significant direct effect of ED on recent SA, with a smaller but significant indirect effect mediated by depression. Our findings suggest a robust association of ED with an increased likelihood of SAs, highlighting the potential role of interpersonal sensitivity. Although psychopathology, particularly mood disorder diagnoses9,41, is a critical factor in understanding suicide, relying solely on specific DSM-5 diagnoses cannot fully capture its complexity. Current research increasingly highlights transdiagnostic dimensions as significant copredictors of suicidal behaviors. Depressive symptoms and ED are distinct yet related constructs. For instance, feelings of non-acceptance or unawareness of emotions are shared features42. Moreover, ED may serve as a diathesis, increasing vulnerability to heightened depressive states due to the intense and prolonged negative emotional experiences associated with ED43. Prior studies have also shown that maladaptive emotion regulation strategies like avoidance, rumination, and suppression are significantly correlated with depressive symptoms44. In the present study, depressive symptoms were assessed using the CDI-2, which focuses on symptoms within the past two weeks, and current DSM-5 diagnoses were considered. In contrast, the RIPoSt–Y has demonstrated the capacity to measure more stable emotional, cognitive, and behavioral styles, exhibiting high test-retest reliability over extended periods20. Based on the instruments used and the current theoretical framework that posits suicidal behaviors as the result of distal factors (personality traits and cognitive styles) interacting with more proximal factors, including current psychopathology, to precipitate suicidal behavior11,38, we chose to examine MDE/current depressive symptoms as potential mediators between RIPoSt–Y and SA. Our decision to model ED as a predictor rather than a mediator of depression is grounded in a diathesis-stress framework10. According to this ontogenic perspective45, ED acts as a diathesis rooted in temperamental influences and neurobiological reactivity and shaped by the interaction between biological vulnerability and environmental stressors during development. This structural reactivity remains relatively invariant over time and increases susceptibility to MDEs. In this framework, our mediation analysis consistently shows ED as a developmental vulnerability which translates into suicidal action both by intensifying current psychopathology and partly directly. The established role of current diagnosis as a proximal determinant of an individual’s response to stressors that can trigger suicidal behavior is also consistent with the conceptualization of dysregulation and suicide in children and adolescents proposed by Benton et al.24. Our findings demonstrate that ED, as measured by the RIPoSt–Y, is robustly associated with recent SAs, in a population with significant and varied psychopathology. This predictive power remains independent of both categorical current MDE diagnosis and dimensional current depressive symptoms. Furthermore, our mediation model revealed that while the effect of ED on recent SA is partly mediated by its influence on current depression/depressive symptoms, the majority (approximately two-thirds) is a direct and independent effect. This aligns with and potentially expands upon the work of Orri et al.42, who identified irritability (reactive aggression and temper tantrums), a construct related to ED but conceptually distinct (and potentially complementary) from our measurement, as a “distal marker” of suicidal ideation and behaviors acting through proximal depressive symptoms. Similarly, our findings resonate with the research of De La Torre-Luque et al.43, who found temper tantrums and difficulty calming down, also understandable as a complementary of ED, to be early antecedents of adolescent suicidal behavior. While a substantial body of primarily cross-sectional research in adults and adolescents has consistently linked ED to suicidal ideation, fewer studies have demonstrated a significant association between ED and SAs after controlling for other psychopathology. Furthermore, the RIPoSt–Y, which has shown sensitivity to ED in adolescents by discriminating between those with ADHD (with or without comorbid bipolar disorder) and controls19, between adolescents with and without sleep/circadian rhythm disorders46, and in response to lithium treatment47, had not been previously investigated in relation to SA. Only one study investigated correlation of suicidal ideation with the scores of the adult version of the RIPoSt, which was significant48. Mirkovic et al.49 employed an instrument based on similar theoretical underpinnings, the Cyclothymic-Hypersensitive Temperament Questionnaire, as part of a composite ED measure in older adolescents with borderline personality disorder, also finding that both MDE and ED had an impact on suicidal behavior. The majority of existing studies have utilized DERS to assess ED and have reported cross-sectional associations with suicidal ideation. Findings from adult studies consistently link DERS-measured ED to suicidal ideation50. Notably, a large longitudinal study on community young adults also found DERS-measured ED to be a significant factor in the development of suicidal ideation51, and Clapham and Brausch52 reported a significant cross-sectional association between DERS-measured ED and suicidal ideation in adults, even after controlling for depressive symptoms. However, other researchers53 found that the association between DERS and suicidal ideation became non-significant when controlling for depressive and other psychological symptoms. Reviews of adult studies using the DERS suggest that the association between ED and SA is less consistent than that between ED and suicidal ideation54. Regarding adolescents, cross-sectional studies in community samples using the DERS have found correlations between ED and suicidal ideation54,55. Hatkevich et al.56 found a significant cross-sectional correlation between DERS-measured ED and suicidal ideation in adolescent inpatients, while Eaddy et al.57 demonstrated that in inpatient adolescents, DERS correlated with both suicidal ideation and “capability for suicide”, even when controlling for depressive symptoms. An 18-month longitudinal study on adolescents also indicated that ED acts as a mediator between trauma and suicidal ideation58. Furthermore, a study on inpatients with a 6-month follow-up found that DERS-measured ED correlated with the persistence of suicidal ideation after discharge, which in turn was associated with more SAs post-discharge59. However, significant data on the direct effect of ED on SAs in adolescent studies are less abundant. A non-recent cross-sectional study correlated ED with SAs in adolescent inpatients60, and a recent community population study showed that DERS-measured ED mediated the relationship between childhood emotional abuse and suicidal behaviors61. To our knowledge, our study is the first to describe and characterize a significant effect of ED, using a measure other than the DERS, on SA in adolescents, while also controlling for psychiatric diagnosis and symptoms. It is crucial to acknowledge that while the importance of studying ED for predicting suicidal ideation and behaviors is widely recognized a definitive consensus on the construct’s definition remains elusive. ED is a complex, multidimensional process with numerous components that current scales capture with limitations and incomplete comprehensiveness. Weinberg and Klonsky16, in their validation study of the DERS in adolescents, highlighted that despite its utility, the DERS has developmental limitations in this age group. Specifically, it demonstrates lower discriminant validity between internalizing and externalizing comorbidities compared to adults. Furthermore, the “awareness of emotion” subscale of the DERS presents developmental challenges, exhibiting lower consistency and correlation with clinical variables, and its language for emotion recognition may be less developmentally appropriate, potentially suboptimal for younger adolescents. We, along with other researchers, share this concern, considering from a developmental psychopathology perspective that a grade of lower of emotional awareness may be normative at certain developmental stages and maladaptive at others62. Moreover, emotional regulation in adolescence can be inherently unstable due to the significant neurodevelopmental and social changes occurring during this period63. Alternative measures of ED frequently used to date, besides the DERS, also present significant limitations. The Child Behavior Checklist (CBCL)64 dysregulation profile, calculated by summing the Anxious/Depressed, Aggressive Behavior, and Attention Problems scale scores, has been extensively used in adolescent populations. This composite measure has shown significant correlations with a wide range of outcomes, including self-harm, suicidal ideation, and suicidal behaviors; current and future bipolar disorder and MDEs; conduct disorder; poor psychosocial outcomes; psychiatric hospitalization; self-aggression; chronic irritability; and more generally with higher severity of current psychiatric symptoms and a greater risk for a chronic course, with more comorbidities and more suicide-related outcomes64,65. Consequently, the CBCL dysregulation profile often functions as a non-specific index of general psychopathology severity and chronicity rather than a specific measure of ED. Given that the aforementioned measures of ED do not comprehensively capture all its facets, and researchers are encouraged to adopt a multidimensional approach with clearly stated working hypotheses66, our study offers a complementary perspective on ED through the use of the RIPoSt–Y. Originating from an extension of research on affective temperaments, which are considered innate and stable individual dispositions, the RIPoSt–Y may be particularly suitable for earlier developmental stages. The interpersonal sensitivity subscale of the RIPoSt–Y emerged as a particularly significant and independent determinant of SAs in our analysis. This subscale aims to capture heightened awareness of others’ behaviors and feelings. This finding underscores the importance of this specific dimension of ED in adolescents, a notion supported by existing literature. Interpersonal conflict may be a more salient factor in adolescent SAs, whereas other factors, such as harm avoidance and mood disorders, might gain more prominence with increasing age10,38. The link between ED and interpersonal difficulties has also been documented using various other measures. Peer rejection, a significant stressor during adolescence, plays a specific role in the development of depressive symptoms67, and affect sensitivity and peer invalidation are recognized as prospective predictors of suicidal ideation68. Although affective instability was identified as a significant correlate of recent SA in bivariate analyses, it did not remain a significant independent predictor in subsequent multivariate analyses. Interestingly, Yen et al.69, studying a different high-risk population (adults with borderline personality disorder, BPD), found that affective instability (operationalized as presence of the DSM criterion for BPD) was correlated with SAs, even after controlling for depression. However, the same author, in a later study on adolescents70, found no difference in emotional regulation between adolescents with BPD and controls. This experience, among others, hints that the specific “facet” of ED most relevant to a psychiatric outcome can vary depending on the sample characteristics, such as age and primary diagnoses. We must not rule out a potentially significant role for affective instability in SA, which might become evident in different samples (e.g., including less severe PCs or older adolescents/young adults), even if in the present sample, a referred clinical population of adolescents with a mean age of 15.7, interpersonal sensitivity appears to be a more significant predictor than other ED dimensions. Lack of predictive utility of our affective reactivity subscale in our regression models might suggest that a general tendency toward intense emotional responses may be a shared characteristic among adolescents acute affective and behavioral psychopathology. In contrast, the specific vulnerability to social cues, rejection, and perceived criticism—as captured by the Interpersonal Sensitivity subscale—appears to be the critical dimension that characterizes the transition to suicidal behavior. This is consistent with literature suggesting that suicidal crises are often precipitated by interpersonal stressors in individuals with a pre-existing sensitivity to social rejection12,71. Finally, we observed that the SA group had a higher rate of antidepressants and antipsychotics treatment. This likely reflects the higher clinical severity and more complex psychiatric history of adolescents with SAs. However, the fact that our ED measure remained a significant predictor of SAs even in a sample with high rates of medication suggests that these emotional traits represent a core vulnerability that may be only partially addressed by pharmacological interventions in this sample. In conclusion, the present study offers a novel contribution by exploring a distinct, and likely complementary, conceptualization of ED using the RIPoSt–Y in relation to SAs, in contrast to the more commonly used DERS and measures of chronic irritability. We found a robust and significant association of RIPoSt–Y and SAs, even after controlling for clinician-diagnosed depression, self-reported depressive symptoms, sex, and age and other potential confounders, also employing state-of-art mediation analysis techniques. However, several limitations warrant consideration. First, the cross-sectional design inherently limits our ability to draw definitive causal or temporal inferences. While our mediation model is theoretically grounded in established developmental frameworks— [21; 10]— the observed indirect effects should be interpreted as statistical associations rather than causal sequences. Furthermore, although the RIPoSt-Y has shown good test-retest reliability, emotional regulation patterns in adolescents are still maturing. What we define as ‘dispositional’ may undergo developmental changes, and future longitudinal research is essential to confirm the stability of these traits and their causal role in the transition to adulthood. Second, regarding the sample and generalizability, our study focused on a highly selected clinical population from a tertiary-care inpatient unit. This ‘severity bias’ means our findings may specifically characterize adolescents with severe clinical presentations. While this provides valuable insights into acute suicidal crises, the results may not be generalizable to adolescents in community settings or primary care, where the interplay between ED and depression might follow different dynamics. Moreover, the strong female predominance—while reflecting common clinical distributions in adolescent internalizing disorders—restricts the applicability of our findings to males. This limitation, along with the cross sectional design, is shared by the majority of current research on ED25. We finally, note measurement considerations. Measurement invariance by sex for the RIPoSt-Y was not tested in this clinical sample due to the limited number of male participants. Future research with more balanced gender distributions is necessary to confirm that the construct of interpersonal sensitivity operates identically across sexes and to establish sex-specific risk thresholds. Declarations. Author contributions Conceptualization, M.A., M.P., A.M., G.S. and R.A.; methodology, M.A., G.S., G.M.; formal analysis, M.A.; data curation, G.D.S., A.M., M.L. and G.D.A.; writing—original draft preparation, M.A.; writing—review and editing, M.P., G.D.A. and M.P.; supervision, S.V. All authors have read and agreed to the published version of the manuscript. Funding This study was supported by a grant (grant 202405 progetto 5 per mille 2024) from Ospedale Pediatrico Bambino Gesù to MA. This work was supported also by the Italian Ministry of Health with Current Research funds. Funders had no involvement with the design, analysis, or reporting of the study. Data availability The datasets generated and/or analyzed during the current study are not publicly available due ethical reasons, but are available from the corresponding author on reasonable request. Declarations Competing interests M.A. received compensation for consultancies from MGGM LLC, NY, USA, a work which is unrelated to the current manuscript. All other authors declare no competing interests. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Ethics approval This study received approval from our Institutional Review Board (Comitato Etico Ospedale Pediatrico Bambino Gesù IRCCS) with practice number 3195. Consent All data were processed in accordance with relevant Italian data protection laws. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets generated and/or analyzed during the current study are not publicly available due ethical reasons, but are available from the corresponding author on reasonable request.