Network analysis of the relationship between depression, emotion regulation, and non-suicidal self-injury among adolescents and gender differences

Abstract

Background

Non-suicidal self-injury (NSSI) poses a serious threat to adolescents’ physical and mental health. This study aims to systematically examine the network structure of the relationship between adolescent depression, emotion regulation, and NSSI, as well as gender differences, from a network analysis perspective.

Methods

The Adolescent Self-Injury Questionnaire, the Center for Disease Control Depression Scale, and the Emotion Regulation Questionnaire were administered to 690 adolescents from June to October 2024. Regularized partial correlation networks were constructed using Gaussian graph models. Centrality metrics such as node strength and betweenness were calculated. Network comparison tests were conducted to examine differences in network structure, global stability, and edge weights between male and female networks.

Results

Network analysis revealed that adolescent depressive symptoms, emotion regulation, and NSSI behaviors are closely interrelated. Within the network of depressive symptoms, depressive affect is most closely linked to somatic symptoms (r = 0.58). Depressive affect intensity and anticipated impact exhibited the highest centrality. Positive affect exhibited the highest mediating centrality. The strongest edge connection between the emotional regulation network and the depressive symptoms network was positive affect - cognitive reappraisal (r = 0.34). The strongest edge connection between the NSSI and depression symptom network was interpersonal relationships-NSSI (r = 0.12). In the network model of NSSI and emotion regulation, the edge weights for NSSI–expressive suppression and NSSI–cognitive reappraisal were r = 0.08 and r = −0.07, respectively. Notably, expressive inhibition served as a key node connecting the “depressive affect-NSSI” network. Furthermore, comparing the network structures between males and females revealed no significant differences in overall network structure or total strength between the two groups (p > 0.05). Gender differences in edge weights were observed for NSSI-positive affect, NSSI-depressive affect, NSSI-physical symptoms, and physical symptoms-depressive affect (p < 0.05).

Conclusions

Adolescent depressive symptoms, emotion regulation, and NSSI behaviors form a network association through specific symptoms, with depressive affect as the core symptom and positive affect as the key hub. Targeted interventions can be implemented for these nodes in the future. Notably, gender differences exist within these networks, suggesting that universal prevention strategies should incorporate gender-specific adaptations to prevent and reduce adolescent depression and NSSI more effectively.

Clinical trial number

Not applicable.

Supplementary information

The online version contains supplementary material available at 10.1186/s12888-026-07998-y.

Introduction

Non-suicidal self-injury (NSSI) refers to the deliberate and intentional act of harming one’s own body tissue without suicidal intent [1]. It has become a serious and widespread issue among adolescents, severely jeopardizing their physical and mental health [2]. Research indicates that globally, the prevalence of adolescent NSSI is estimated at 18%-23% [3]. A meta-analysis revealed a detection rate of 22.37% for NSSI among Chinese adolescents. In clinical adolescent samples, this proportion reaches as high as 50% [4]. Related research indicates that non-suicidal self-injury not only inflicts painful physical harm on adolescents but also impacts their subsequent cognitive, emotional, and psychological developmental trajectories, perpetuating a cycle of maladaptive behavior [2].

A growing body of research has conceptualized NSSI as a behavioral response to intense negative emotions. According to the functional model of NSSI [5], such behaviors are primarily maintained by negative reinforcement, meaning that adolescents engage in NSSI to escape or reduce aversive emotional states. Depression, as a common negative emotional experience, plays a crucial role in the development and persistence of NSSI [6]. Related research indicates that each additional depressive episode increases the risk of NSSI by 18% [7]. Studies also reveal that depression can result from non-suicidal self-injury [8]. NSSI also serves as a precursor that exacerbates emotional deterioration and contributes to depression. Klonsky et al. [9] found that individuals engaging in NSSI frequently report feelings of shame, guilt, and anger following the act, which in turn may intensify depressive symptoms. These findings highlight a reciprocal and reinforcing relationship between depression and NSSI.

Previous studies have identified emotion regulation difficulties as a core issue in the development of NSSI behaviors [10]. Among these, adolescents’ self-directed emotion regulation is particularly crucial. Emotion regulation refers to the internal processes and external behaviors that monitor and adjust emotions to adapt to environmental demands and interpersonal interactions [11]. Gross’s [11] influential process model of emotion regulation distinguishes between two major regulatory strategies: cognitive reappraisal and expressive suppression. When adolescents encounter negative emotions such as sadness, they are prone to impulsive self-harm behaviors and employ ineffective emotional regulation strategies. A longitudinal study found that difficulties in cognitive reappraisal and expression suppression strategies were closely associated with the frequency, duration, and severity of NSSI behaviors [12], suggesting that emotional dysregulation may increase the risk of NSSI. Emotion regulation develops throughout an individual’s lifespan [13] and plays a crucial role in the mental health development of depressed adolescents. For these adolescents, a lack of effective emotion regulation strategies may lead to worsening depressive symptoms [14]. According to the emotional cascade theory, depression—as a negative emotional experience—can further increase the risk of NSSI when other emotion regulation strategies are unavailable or ineffective [10]. Research indicates adolescents engaging in NSSI exhibit higher levels of depressive symptoms and employ more maladaptive cognitive-emotional regulation strategies [15].

In summary, there exists an interrelated relationship among adolescent depression, emotion regulation, and non-suicidal self-injury (NSSI). Considering that NSSI may function as a distinct form of emotion regulation and that poor emotion regulation may be associated with depressive disorders, previous studies have separately examined adolescent depression and NSSI [15] and the relationship between emotion regulation and non-suicidal self-injury [16], respectively. However, the interactive mechanisms among these three factors have not been thoroughly explored. Furthermore, these findings may not be generalizable to non-clinical samples, and the precise logical relationship among them remains unclear.

Furthermore, traditional research has focused on variable-level or total-score associations, with relatively little attention to how specific symptom domains and regulatory strategies interact at a granular level. Gender differences further complicate these associations: although some studies report higher rates of NSSI and depression among girls [17–19], others find no significant gender differences in NSSI [7]. To date, few studies have systematically investigated gender-specific network structures linking depressive symptom domains, emotion regulation strategies, and NSSI.

To address these gaps, the present study employs network analysis—a methodological approach grounded in dynamic systems theory that visualizes complex interactions among variables as networks of interconnected nodes [20]. This method places symptoms within a network of interconnected nodes and describes the relationships between them. By identifying core symptoms (those with high centrality) based on network centrality metrics (strength and expected influence), this approach offers significant practical value for determining intervention focal points and enhancing intervention effectiveness. Network analysis can quantify and visually represent the relationships between NSSI and depression and emotion regulation at the symptom level, deepening our understanding of comorbidity mechanisms [21]. In this study, interactive mechanisms refer to the dynamic relationships through which symptoms and regulatory processes mutually influence one another. Symptom clusters are defined as groups of two or more co-occurring and interrelated symptoms that function as cohesive units within the broader psychopathological system [21]. Understanding these interactions at the symptom-cluster level has important implications.

Accordingly, the present study aims to address the following research questions:

1. What is the network structure of depressive symptoms (four CES-D dimensions), emotion regulation strategies (two ERQ dimensions), and NSSI among Chinese adolescents?

2. Which nodes are most central in the network, based on strength and expected influence centrality?

3. What are the bridge connections linking emotion regulation strategies, depressive symptom domains, and NSSI?

4. Are there significant gender differences in the network structure and centrality of these constructs?

Methods

Participants

This study is a cross-sectional investigation conducted from June to October 2024. Students were recruited from three secondary schools in Nanyang City and Zhoukou City, Henan Province. A two-stage cluster sampling method was used: first, three schools (two junior high schools and one senior high school) were selected; second, classes within each school were chosen using simple random sampling, with the class serving as the unit for questionnaire administration. The detailed sampling results are as follows: nine classes (approximately 40 students per class) were selected from the first junior high school, three classes (around 50 students per class) from the second junior high school, and seven classes (about 50 students per class) from the senior high school. Inclusion criteria were: ① Age between 10 and 19 years; ② No cognitive impairment, with the ability to understand the questionnaire content and willingness to cooperate; ③ Obtaining informed consent from the participant or their guardian. Exclusion criteria were: ① Current diagnosis of other mental disorders, such as schizophrenia or bipolar disorder; ② Presence of severe physical illness or serious neurological disease. Consequently, a total of 860 eligible students participated in this study.

Ethical considerations

This study complies with the Declaration of Helsinki and the Ethical Review Measures for Biomedical Research Involving Human Subjects. It was approved by the Local Ethics Committee of Zhengzhou University (Ethics No. ZZUIRB2023-302), and written or electronic informed consent was obtained from the parents of all participants. All data underwent rigorous anonymization to protect participant privacy, with personal identifiers removed. Data were securely stored and accessible only to the research team.

Measures

General demographic information questionnaire

Demographic information for all participants, including gender, age, grade level, place of residence, whether they are only children, whether they have experienced being left behind, whether parents raised them during childhood, and the closeness of their relationship with their father and mother.

Center for Epidemiological Studies Depression Scale (CES-D)

Developed by Radloff [22], this scale primarily assesses the frequency of depressive symptoms or feelings experienced by an individual over a week, focusing on depressive mood or affect. The scale comprises four dimensions: depressive affect, positive affect, somatic symptoms, and interpersonal relationships. It consists of 20 items: 8 items for depressive affect, 4 items for positive affect, 6 items for somatic symptoms, and 2 items for interpersonal relationships. A 4-point rating scale (0–3 points) is used. Higher scores indicate a greater risk of depression. In this study, the Cronbach’s α for this scale was 0.897.

When Radloff [23] developed the CES-D, he proposed a cutoff score of 16 to indicate depression. Research by Li and Hicks [24] demonstrated that a cutoff of 16 points yielded 100% sensitivity and 76% specificity. Therefore, in this study, a CES-D total score of ≥ 16 was considered indicative of depressive affect. Following Radloff’s [22] recommendation, a cutoff of 28 points is adopted for screening depression in adolescent populations. Although CES-D is a well-validated screening tool, it should be noted that it does not provide a clinical diagnosis.

Emotion Regulation Questionnaire (ERQ)

Developed by Gross and John [25], the Chinese version was revised by Wang and Li [26]. It comprises 10 items, with 4 items measuring the expression suppression dimension and 6 items measuring the cognitive reappraisal dimension. The scale employs a 7-point Likert scale ranging from 1 (“Strongly Disagree”) to 7 (“Strongly Agree”). Higher scores indicate a greater tendency to use emotion regulation strategies. In this study, the Cronbach’s α for the scale was 0.806, with the expression suppression dimension yielding a Cronbach’s α of 0.753 and the cognitive reappraisal dimension yielding a Cronbach’s α of 0.865.

Adolescent Self-harm Scale

Feng Yu et al. [27] revised and developed the Adolescent Self-harm Scale, which is the most widely used instrument in China for assessing the severity of self-harm among adolescent patients. The scale comprises 19 items, including 18 methods of intentional self-harm and one open-ended question. Each item consists of two sections: frequency of self-harm and self-rated severity of injury. The frequency section offers four levels: 0 times, 1 time, 2–4 times, and 5+ times. The severity section provides five levels: none, mild, moderate, severe, and extremely severe. Using a 5-point Likert scale, the severity of self-harm for each item is represented by the product of frequency and self-rated severity. The total score is the sum of all item scores, with higher scores indicating greater severity of self-harm. The criterion for assessing the occurrence of NSSI is whether the final score is “0.” In this study, the Cronbach’s α for this scale was 0.947.

In this study, the screening of depression and NSSI in adolescents mainly relied on adolescent self-reports.

Statistical methods

Statistical analysis was performed using SPSS 25.0 and R 4.2.1 software. For quantitative data, results are presented as mean ± standard deviation if normally distributed; if not, results are shown as median (P25, P75). Count data were expressed as frequency and percentage. In this study, the normality of continuous variables was assessed using the Shapiro-Wilk test, combined with a thorough evaluation of skewness and kurtosis. It is generally considered that when the absolute value of skewness is less than 3, and the absolute value of kurtosis is less than 10, the data can be regarded as normally distributed, meaning they meet the assumption of normality. Since Gaussian Graphical Models were used in the network analysis, normality was a requirement for all included variables [28]. Variables that did not meet the normality assumption were log10-transformed before network estimation. All network analyses were performed using these transformed variables. Missing data in valid questionnaires were imputed. For numeric missing values, missing data were imputed using the mean of all other subjects. For non-numeric missing values, the mode (i.e., the most frequent value) was used for imputation. in the statistical information.

Network estimation

Recent methodological literature has demonstrated that dimensional-level network analysis is a valid and valuable approach in psychopathology research [29, 30]. For instance, Blasco-Belled and Alsinet [29] explicitly examined network structures at both item and dimension levels, showing that dimension-level networks can reveal higher-order organizational patterns that may be obscured at the item level. Similarly, the phased framework proposed by recent research incorporates dimension extraction as a legitimate first step before network characterization [30]. Therefore， in our study, the network was constructed at the dimensional level rather than the item level. For depressive symptoms, we used the four established CES-D subscales (depressive affect, somatic symptoms, positive affect, and interpersonal problems) based on the scale’s validated four-factor structure [22]. For emotion regulation, we used the two ERQ dimensions (cognitive reappraisal and expressive suppression) [11, 25]. NSSI was included as a single node.

This network includes seven nodes, requiring estimation of 28 parameters [7 threshold parameters and 21 (7 × 6/2) pairwise correlation parameters]. Thus, the study sample size is sufficient for statistical analysis [20]. In the constructed network model, correlations between nodes were calculated, and the Gaussian Graphical Model (GGM) was evaluated. In the visualized network, blue lines represent negative edges, while red lines denote positive edges. Thicker lines indicate stronger connections.

Network centrality

The R package “qgraph” was used to compute centrality metrics for the network, including strength, closeness, and betweenness. Strength is the sum of the weighted values of all edges connected to a node, measuring its importance within the network. Closeness is the reciprocal of the sum of the shortest path distances from all other nodes to that node. Betweenness is the frequency with which a node appears on the shortest path between any two other nodes, reflecting the node’s bridging role in the network [28].

Estimation of network accuracy and stability

Use the R package ‘Bootnet’ to estimate accuracy and stability. Perform 1500 iterations of the network using the bootstrap method to calculate the accuracy of edge estimates and plot the 95% nonparametric bootstrap confidence intervals (CI) for each edge, where narrower intervals indicate higher accuracy. Conduct 1500 bootstrap iterations to assess the stability of centrality measures. According to Epskamp et al. [20], the CS coefficient should not be lower than 0.25 and ideally should be above 0.5.

Network comparison

Finally, the R Network Comparison Test package was employed to conduct permutation tests for the significance of network invariance across genders, invariance in global strength (GS), and invariance in centrality. To ensure test quality, 1500 permutations were performed.

Results

Descriptive analysis

This study identified a total of 860 eligible students who participated in this study. Among them, 836 students completed the questionnaire, while 24 students did not due to school suspension, leave of absence, or absence on the day of the survey, yielding a valid response rate of 97.2%. After data cleaning, 146 participants were excluded due to patterned responses (e.g., straight-lining, invariant answers) or incomplete data (missing values > 20%). The final valid sample consisted of 690 adolescents. The average age of respondents was 15 years, with 47.68% identifying as male. The percentage of participants reporting NSSI was 31.6%, and the prevalence rate of depression was 26.96%. Demographic characteristics of adolescents and gender differences are supplemented in Additional File 1 Table S1. Table 1 presents scores across various dimensions of each scale, as well as gender differences. Results indicate that among depressive symptoms, females scored higher than males on somatic symptoms, interpersonal relationships, and depressive affect, with the latter difference being statistically significant (t = −2.187, p = 0.029). No statistically significant differences were found between males and females in emotion regulation or NSSI.

Table 1.

Descriptive statistics of scale dimensions and gender differences

Variable	Total（690）	Male（329）	Female（361）	t*/Z	P
CES-D (M±SD) (score)	0.91 ± 0.50	0.88 ± 0.50	0.92 ± 0.51	−1.173	0.241
Depressive affect	0.84 ± 0.61	0.79 ± 0.58	0.89 ± 0.62	−2.187	0.029
Positive affect	2.14 ± 0.68	2.03 ± 0.69	2.05 ± 0.67	−0.388	0.698
Somatic symptoms	1.08 ± 0.60	1.07 ± 0.59	1.09 ± 0.61	−0.316	0.752
Interpersonal relationships	0.51 ± 0.64	0.49 ± 0.63	0.53 ± 0.66	−0.837	0.403
ERQ (M±SD) (score)	4.17 ± 0.92	4.21 ± 0.98	4.14 ± 0.87	0.981	0.327
Expressive suppression	3.64 ± 1.20	3.64 ± 1.24	3.63 ± 1.17	0.109	0.917
cognitive reappraisal	4.53 ± 1.15	4.59 ± 1.21	4.48 ± 1.10	1.234	0.217
NSSI M(P25,P75)	0.00(0.00,1.00)	0.00(0.00,1.00)	0.00(0.00,2.00)	−1.466	0.143

Notice: t* refers to the independent samples t-test. Z refers to the Mann-Whitney U test

Network analysis

The network structure depicting the relationships among depression, emotion regulation, and non-suicidal self-injury (NSSI) in the full adolescent sample is shown in Fig. 1 (The corresponding partial correlation matrix is supplemented in Additional File 1 Table S2). Among depressive symptoms, somatic symptoms exhibited the strongest edge connection with depressive affect (r = 0.58), followed by interpersonal relationships and depressive affect (r = 0.37).In the NSSI and depressive symptoms network model, the edge between NSSI and interpersonal relationships (r = 0.12) represented the strongest association between NSSI and depressive symptoms, followed by the edge between NSSI and depressive affect (r = 0.10) and that between NSSI and somatic symptoms (r = 0.09).In the network model of NSSI and emotion regulation, the edge weights for NSSI–expressive suppression and NSSI–cognitive reappraisal were r = 0.08 and r = −0.07, respectively. In the depression-emotion regulation network model, positive affect and cognitive reappraisal (r = 0.34) exhibited the strongest edge connection. This indicates that adolescent depression, emotion regulation, and NSSI are interconnected through distinct nodes.

Fig. 1.

Full-sample network architecture

After grouping by gender, the network structures for the male and female groups are shown in Fig. 2 (corresponding edge weight matrices are supplemented in Additional File 1 Table S4 and Table S5). Within the male group network, the strongest edge weight among depressive symptoms was between depressive affect and somatic symptoms (r = 0.51). In the network model linking NSSI and depressive symptoms, the strongest edge connection was between NSSI and interpersonal relationships (r = 0.15). In the network model of NSSI and emotion regulation, the edge weights for NSSI–expressive suppression and NSSI–cognitive reappraisal were r = 0.07 and r = −0.05, respectively. In the female group network, the strongest edge connection among depressive symptoms was between depressive affect and somatic symptoms (r = 0.63). In the network model linking NSSI and depressive symptoms, the NSSI-somatic symptoms connection was the strongest (r = 0.19). In the NSSI-emotion regulation network model, the edge connection weights for NSSI-expressive suppression and NSSI-cognitive reappraisal were r = 0.07 and r = −0.08, respectively.

Fig. 2.

Network structures of different samples (The left panel shows the male network diagram, while the right panel depicts the female network diagram). Note: positive correlation:red;negative correlation:blueLabels: Pink: depressive symptoms;Green: emotion regulation strategies; Blue: NSSI

Centrality measures analysis

Figure 3 displays standardized estimates of node centrality measures across the entire sample. In our study, we focused on expected influence and strength as the primary centrality indices. Depressive affect showed the strongest expected influence and strength, followed by somatic symptoms. Positive affect exhibited the highest betweenness centrality, suggesting that it serves as a bridge node in the network.

Fig. 3.

Standardized estimates of node centrality measures across the entire sample

Standardized estimates of node centrality by gender are presented in Fig. 4. Among males, depressive affect exhibited the strongest expected influence and strength. Among females, depressive affect showed the strongest strength, while somatic symptoms showed the strongest expected influence.

Fig. 4.

Standardized estimates of centrality measures for different sample nodes

Figure 5 Bootstrap sample tests indicate correlation stability (CS) of 0.751 for node strength and 0.751 for expected influence. CS values exceeding 0.5 indicate good node stability.

Fig. 5.

Node stability test. Note: CES-D1 positive affect CES-D2 depressive affect CES-D3 somatic symptoms CES-D4 interpersonal relationships ERQ1: expressive suppression ERQ2: cognitive reappraisal NSSI: NSSI

Network comparison

Network structural invariance analysis revealed no significant differences between male and female groups (Male: 2.784, Female: 3.006, p = 0.161). Tests for global network strength invariance indicated no significant differences between genders (p = 0.32).

The network centrality invariance test revealed significant differences in expected influence and strength of somatic symptoms between the male and female networks. (See Additional File 1 Table S6).

Results indicated significant differences in edge weights between the two networks. Statistically significant differences were observed in the edge weights of NSSI-positive affect, NSSI-depressive affect, NSSI-somatic symptoms, and somatic symptoms-depressive affect (p < 0.05, see Additional File 1 Table S7).

Discussion

In this study, the prevalence of NSSI among adolescents was 31.6%. This rate is higher than the national average of approximately 22–25% reported in meta-analyses of Chinese adolescents [31], but lower than the 47.70% reported among rural Central Chinese adolescents [32]. The relatively high prevalence may reflect the specific characteristics of our sample, including the inclusion of both junior and senior high school students from regions in Henan. Previous research has documented substantial regional variation in NSSI prevalence across China, with higher rates often observed in the Central and Western regions [31]. Additionally, differences in measurement instruments, recall periods (lifetime vs. past-year), and scoring methods may contribute to variations across studies. Future research should employ standardized instruments and assessment periods to facilitate cross-study comparisons.

Against this background, the present study further employed network analysis to explore the interrelationships between depression, emotion regulation, and NSSI in adolescents, as well as gender differences, while identifying the central nodes in the network. We found that depression symptoms, emotion regulation, and NSSI were closely interconnected in the network structure, suggesting that these factors interact with each other rather than functioning in isolation.

The network analysis results of this study indicate that the connection weight between somatic symptoms and depressive affect is highest within the depressive symptom cluster, suggesting a synergistic reinforcing effect between the two. This aligns with findings from Bohman et al. [33], which demonstrated that somatic symptoms during adolescence are associated with more severe adult psychiatric disorders, such as depression. The underlying mechanisms may involve pathophysiological pathways. First, inflammatory mechanisms, where depression is characterized by elevated levels of cytokines such as interleukin-1, interleukin-6 (IL-1, IL-6), and tumor necrosis factor (TNF). These factors can directly induce somatic symptoms, for example, fatigue and pain [34, 35]. Second, neuroendocrine mechanisms: chronic stress and depressive mood activate the HPA axis, leading to persistently elevated stress hormones like cortisol, which in turn trigger sleep disturbances, fatigue, and other physical problems [36]. Network comparison results further revealed higher edge weights linking somatic symptoms and depressive affect in the female group compared to the male group. This indicates a stronger association between depressive affect and somatic symptoms in adolescent females, consistent with findings by Zhao and Zhou [18] and Barrocas et al. [37]. This may be attributed to the regulatory effect of estrogen on the HPA axis during adolescence, making females more prone to somatization responses under stress [38]. Hormonal fluctuations associated with the menstrual cycle may amplify the connection between depressive affect and physical discomfort. The present findings also indicate a strong association between depressive affect and interpersonal relationships within the depressive symptom cluster. This aligns with the study by Liu et al. [39]. The potential explanation may be that adolescents experiencing depressive affect during adolescence may suffer from more frequent physical complaints and lower self-efficacy, thereby reducing life satisfaction and affecting interpersonal relationships [40].

Previous studies have demonstrated a strong association between NSSI and depressive symptoms [6]. Our research confirms this finding, revealing connections between NSSI and depressive affect, interpersonal relationships, and somatic symptoms within the depressive symptom cluster. The positive association between depressive affect and NSSI in this study suggests that heightened depressive affect may increase the risk of NSSI behavior. This aligns with Nock’s [1] functional model of non-suicidal self-injury, which posits that NSSI may serve as a negative reinforcement mechanism for adolescents to escape negative emotions such as depression. This study also found that interpersonal relationships showed the strongest marginal association with NSSI. A possible explanation lies in the structural linkage of depressive affect: interpersonal issues may trigger or exacerbate depressive mood, thereby increasing NSSI risk [41, 42]. Network comparison results further revealed gender differences, with females exhibiting higher edge weights in both depression-NSSI and somatic symptoms-NSSI pathways than males. This may relate to differing emotional processing styles among adolescent males and females, as females tend to be more susceptible to the physical and psychological impacts of negative emotional events [43]. They exhibit heightened sensitivity to negative emotions and are more prone to rumination [44], which amplifies distress, intensifies physiological stress responses, and exacerbates somatic symptoms [36]. Consequently, they are more likely to resort to NSSI to cope with emotional distress, exhibiting higher rates and severity of self-harm behaviors [17]. This disparity reflects adolescents’ tendency to internalize emotional problems, whereas males are more inclined toward externalizing behaviors like aggression [45, 46].

In this study, we found that emotion regulation strategies showed a strong association with both depression and NSSI, and functioned as a critical bridge node linking depression and NSSI in the network structure. This aligns with studies by Qian et al. [15], Sun et al. [7], and Yen et al. [47], indicating that depression is strongly associated with NSSI and may be closely interconnected with NSSI via an emotional cascade pattern. This cascade exacerbates emotional regulation difficulties via cognitive biases and expressive suppression. Network analysis revealed that while the direct association between expression suppression and NSSI was relatively weak, the positive correlation between expression suppression and depressive affect suggests that excessive suppression of emotional expression may exacerbate depressive symptoms, indirectly increasing the occurrence of NSSI behaviors. This pathway aligns with Gross’s [11] process model of emotion regulation, where adolescents employing ineffective coping strategies like expression suppression to manage depressive affect may develop impulsive behaviors (e.g., NSSI). Findings indicate that cognitive reappraisal strategies (e.g., reframing situational meaning) in emotion regulation negatively correlate with depressive affect, suggesting adolescents can reduce depression risk by restructuring cognitive frameworks around negative events [48].

The centrality results indicate that depressive affect exhibits the strongest expected influence and intensity, making it a crucial node in the network linking adolescent depression, emotional regulation, and NSSI behavior. Depressive affect is a hallmark feature of depression [49], referring to a range of negative emotional experiences such as sadness, feeling “down,” depressed mood, fear, loneliness, crying, and perceiving oneself as a failure. First, other nodes within the depressive symptom cluster (e.g., somatic symptoms, interpersonal relationships) were found to connect to NSSI through the depressive affect node, indicating that depressive affect and other symptoms co-occur to form a symptom cluster associated with greater depressive severity and higher NSSI vulnerability in the network. Second, depressive affect showed strong associations with increased expressive inhibition and decreased cognitive reappraisal, highlighting a structural pathway where emotion regulation difficulties are closely intertwined with depressive affect and NSSI. Consistent with studies by Campbell and Osborn [50] and Kenny et al. [51], depressive affect and other negative emotions demonstrated strong structural connectivity with other symptoms (e.g., somatic symptoms, interpersonal relationships, NSSI) across multiple network pathways. Therefore, early identification of adolescents experiencing depressive affect is crucial to provide targeted interventions that reduce the occurrence of NSSI.

An unexpected finding in our study was that positive affect exhibited the strongest bridge centrality among all node centrality measures in the full sample. Although positive affect was not directly associated with NSSI, it showed strong structural connections with depressive affect via the cognitive reappraisal pathway, suggesting a concurrent structural linkage among these variables in the network. Cognitive reappraisal is conceptualized as altering interpretations of emotional experiences, thoughts, and events to reshape the meaning of situations, which is associated with adaptive emotional responses [11]. Positive psychology theory similarly emphasizes actively identifying and nurturing personal strengths to reframe life experiences, which is linked to greater positive emotions, well-being, and life satisfaction. Therefore, strengthening adolescents’ capacity for positive emotion regulation may represent a promising entry point to weaken the co-occurring linkage between depression and NSSI, which may help reduce depressive symptoms and the likelihood of NSSI in adolescents.

This study holds significant practical implications for the intervention and treatment of adolescent depression and NSSI. In prevention and intervention efforts targeting adolescent depression and NSSI, early identification of physical symptom manifestations, assistance in managing interpersonal relationships, enhancing positive emotional levels and promoting appropriate expression, and teaching effective emotion regulation strategies are essential to reduce the occurrence of depression and NSSI.

The identification of robust connections—particularly the central role of depressive affect—has important clinical implications. It suggests that interventions targeting depressive affect may yield downstream effects on both emotion regulation and broader interpersonal functioning. This aligns with the comprehensive adolescent suicide prevention framework proposed by Baldini et al., which emphasizes the urgent need for multi-sectoral strategies integrating early detection, evidence-based interventions, and improved access to mental health care [52]. Therefore, Björgvinsson et al. [53] found that cognitive-behavioral therapy (CBT) strategies—such as behavioral activation and cognitive restructuring—may help reduce depressive symptoms, thereby breaking the vicious cycle of “depression-somatization-NSSI.” Zieff et al. found that mindfulness-based stress reduction training enables patients to observe their inner experiences objectively, evaluate themselves, and cultivate self-acceptance [54]. This fosters positive coping mechanisms, thereby reducing the somatization of depressive emotions. Prioritizing the management of depressive emotions may also have cascading effects on alleviating NSSI and improving emotional regulation.

Given the observed gender differences in network structure and edge strengths, gender‑tailored interventions targeting adolescent NSSI are warranted. For females, who showed stronger connections between depressive affect, somatic symptoms, and NSSI, interventions may prioritize alleviating somatic complaints, reducing rumination, and enhancing adaptive emotion regulation strategies such as cognitive reappraisal. For males, who exhibited relatively weaker connectivity in these pathways, interventions may focus more on improving interpersonal functioning and reducing externalizing responses to negative emotions.

Conclusions

This study employed network analysis to examine the relationships among adolescent depression, emotion regulation, and non-suicidal self-injury (NSSI), identifying core symptoms within the network. Among depressive symptom clusters, somatic symptoms exhibited the strongest connection to depressive affect. Within the depressive symptoms and NSSI network model, interpersonal relationships demonstrated the most robust edge connections to NSSI. Emotion regulation was identified as a key bridge node linking depression and NSSI in the network structure. Within this network, depressive affect exhibited the strongest expected influence and intensity, while positive affect demonstrated the highest betweenness centrality. These findings indicate that adolescent depression, emotion regulation, and NSSI are closely interrelated. Consequently, future interventions should prioritize regulating depressive affect and cultivating positive affect in adolescents, while also emphasizing the improvement of interpersonal relationships and the promotion of emotional expression. Such approaches may help alleviate adolescent depression and reduce the occurrence of NSSI.

Limitations

Several limitations of this study should be acknowledged. First, this study employed a cross-sectional design, which precludes any causal or temporal inferences between variables. Longitudinal or experimental designs are necessary in future research to investigate the directionality and potential causal relationships between depressive symptoms, emotion regulation, and NSSI. Second, some edge weights between NSSI and depressive symptoms were relatively small (r = 0.05–0.15). Although network analysis can detect subtle connections among psychological symptoms, these weak partial correlations must be interpreted cautiously regarding their clinical significance. Third, assessments of depression and NSSI relied solely on adolescent self-reports, which may be subject to recall bias and social desirability bias. Although validated scales were used, the lack of clinical diagnostic interviews or institutional records means that our findings reflect self-reported symptoms rather than formal clinical diagnoses. Fourth, the sample was restricted to adolescents in two regions of Henan Province, China. Therefore, the generalizability of our findings to other regions, cultures, or clinical populations remains unclear. Fifth, this study conducted network analysis only at the dimensional level. Although this provides a parsimonious macro‑level perspective, aggregating items into dimensions may mask within‑domain symptom heterogeneity. Future item‑level network analyses could help identify more specific symptom patterns.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

NSSI

Non-suicidal self-injury

CES-D

Center for Epidemiological Studies Depression Scale

ERQ

Emotion Regulation Questionnaire

GGM

Gaussian Graphical Model:

CS

Correlation stability

GS

Global strength

IL-6

Interleukin-6

IL-1

Interleukin-1

TNF

Tumor necrosis factor

CBT

Cognitive-behavioral therapy

Author contributions

SL: conceptualization, methodology, software, formal analysis, and writing. ZA: methodology, data collection. JY: data collection. MY: data collection. CS: writing. XH: conceptualization, data collection, and writing. All authors read and approved the final manuscript.

Funding

This research was funded by the following grants: Research on Building Psychological Assistance Capacity in Grassroots Communities in Henan Province from the Perspective of the “Fengqiao Experience,” Henan Provincial Social Science Association Research Project, SKL-2025-420, October 2025; and Evidence-Based Research on Campus Psychological Safety Early Warning Systems and Policy Simulation for Adolescent Externalizing Problem Behaviors, Henan Provincial Education Powerhouse Special Research Project, 2026JYQS501, October 2025.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study complies with the Declaration of Helsinki and the Ethical Review Measures for Biomedical Research Involving Human Subjects. It was approved by the Local Ethics Committee of Zhengzhou University (Ethics No. ZZUIRB2023-302), and written or electronic informed consent was obtained from the parents of all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.