NSSI is associated with various mental disorders. A meta-analysis showed that individuals with mood disorders (anxiety, depression, and bipolar disorder) have a higher risk of exhibiting NSSI behaviors than did the normal group (7). Therefore, it is critical to study the interaction mechanisms between the emotions and NSSI behavior of adolescents with mood disorders. For example, numerous studies suggest that female sex, depression, and anxiety are risk factors for predicting NSSI behavior (8–11). Fox et al. (12) conducted a meta-analysis of NSSI risk factors and showed that depressive disorder is a risk factor for NSSI. Depression and NSSI often co-occur, with the presence of depression suggesting a significant increase in the incidence of NSSI behavior (13, 14). Clinically, the comorbidity rate of depression and anxiety among adolescents is as high as 15%-75% (15). Previous reports also showed that patients with anxious depression had more frequent episodes of major depression and a higher risk of suicidal ideation and previous suicide attempts than those with non-anxious depression (16). However, how anxiety affects adolescents’ NSSI behavior has not been fully studied.

NSSI functions refer to the motives or reinforcers of NSSI behavior (17). There have been many theoretical models proposed on the functions and purposes of NSSI behavior, including physiological models such as the homeostasis model proposed by Stanley et al. (18), psychological and social factor models such as the four-function model proposed by Nock and Prinstein (19), the developmental psychopathology framework proposed by Yates (20) the experiential avoidance model proposed by Chapman et al. (21), and the integrated motivational-volitional model proposed by O’Connor et al. (22). In their study of NSSI behavior in Chinese adolescents patients, Chinese researchers Qu et al. (23) proposed a three-factor model of NSSI functions that included emotion regulation, attention seeking, and social avoidance. On this basis, the Chinese version of Functional Assessment of Self-Mutilation (C-FASM) was formulated. The C-FASM has good content, structural validity, and reliability. The instrument can be helpful to Chinese adolescents as a comprehensive measure of NSSI behaviors. This model is applicable to both clinical and non-clinical populations within the Chinese cultural context. Emotion regulation refers to the behavior that patients exhibit to reduce or terminate unwanted negative emotional experiences (including anxiety, anger, and depression) through NSSI, so as to re-establish a sense of control. A number of studies have shown a significant positive correlation between negative cognitive emotion regulation and NSSI behaviors (24). Attention seeking means that adolescents engage in NSSI behaviors to seek social support and to gain attention and help from others. A relevant study (25) found that NSSI behaviors are closely related to positive social reinforcement and social influence. Among adolescents, the NSSI behavior of males, compared with females, is more often motivated by attention seeking and social influence. Social avoidance refers to adolescents’ avoidance of social needs, emphasizing that individuals use NSSI as a strategy to avoid adversity and social demands. Research on NSSI functions considers the repeated occurrence of NSSI as an outcome of operant conditioning (26): the specific situations before the behavior and behavioral outcomes exert (positive/negative/self) reinforcement on the individual. Once the individual encounters similar situations in the future, the same behavior will occur again. Therefore, if NSSI functions can be identified and the generation of behavioral outcomes associated with these functions can be controlled, an individual’s NSSI behavior will terminate.

Sex is a key factor affecting all aspects of adolescents’ physical and mental health. Universal sex differences exist in pathological conditions including anxiety and depression, with females being more than twice as likely to develop the disease as males (27), which may be related to differences in various aspects such as brain structure and function, sex hormones, and social expectations (28). Moreover, there are distinct sex differences in NSSI behaviors between males and females. Researchers have confirmed (29, 30) that the incidence of NSSI is higher in females than in males, indicating that females may be more inclined than males to resort to NSSI behaviors to alleviate overwhelming distress. A study has shown (31) that adolescent females are more emotionally sensitive than are adolescent males, suggesting that they are more susceptible to perceiving and absorbing negative emotions. Therefore, this study considers the impact of anxiety in adolescent patients of different sexes on NSSI behaviors.

Although previous studies have confirmed that anxiety, NSSI functions, and sex are all associated with NSSI, the interaction mechanisms between anxiety and NSSI functions in the process of triggering NSSI and whether there are sex differences in these mechanisms remain unclear. Additionally, previous research findings have mostly focused on non-clinical samples, mainly through large-scale epidemiological surveys of middle school students; there are relatively few studies that include clinical samples. Therefore, in the present study, a sample of Chinese adolescents with depressive disorders is used to investigate the mediating effects of NSSI functions in the process by which anxiety triggers NSSI behaviors and to examine the sex difference in these effects (see Figure 1 for the hypotheses).

In this study, three-stage sampling was used to select the research participants. First, nine provinces in China were selected based on their economic conditions [three were good, three were moderate, and three were poor); second, 11 specialized psychiatric hospitals and the psychiatric departments of 9 general hospitals (20 institutions in total)] in these nine provinces were selected (All these institutions were willing to participate in the survey and had professional research team to carried out the study); and finally, outpatient and inpatient adolescents with depressive disorder were recruited by convenience sampling. From the beginning of recruitment in December 2020 to its completion in May 2022, the data of a total of 1773 adolescent patients with depressive disorders were collected. The inclusion criteria were as follows: (1) meeting the diagnostic criteria for major depressive disorders in the Fifth Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5); (2) aged 12-18 years; (3) at least one instance of NSSI behavior in the past; (4) a minimum of six years of education, so they could understand the questions in the questionnaire and complete the survey independently; (5) patients and their families consenting to participate in this project and signing an informed consent form. The exclusion criteria were as follows: (1) patients with severe physical illnesses, infectious diseases, and immune system disorders; (2) patients with a history of traumatic brain injury, epilepsy, or other known severe neurological diseases or organic brain diseases; and (3) patients with a history of severe mental disorders such as schizophrenia or intellectual developmental disorder. This study was approved by the Ethics Committee of Tongji University as well as the Shanghai Pudong New Area Mental Health Center (No.PKJ2023-Y21).

SPSS27.0 was used for statistical analyses. Count data are expressed as the number of cases and percentage (%). The χ² test was used for between-group comparisons of categorical variables, and t-test was performed to compare continuous variables. Regression analysis was used to evaluate the effect of anxiety on NSSI functions and frequency, and the nonparametric Spearman rank test was conducted to evaluate the correlation between NSSI frequency and functions. Model 4 in the SPSS macro developed by Hayes was used to test the mediating effects of NSSI motivation on the association between anxiety and NSSI frequency for the male group and for the female group after controlling for confounding factors such as age and previously diagnosed mental disorders (36).

In this study, the data of a total of 1773 adolescents with depressive episodes and NSSI behaviors, including 316 males (17.82%) and 1457 females (82.18%), were analyzed. The age of male adolescents (15.29 ± 1.519 years) was higher than that of females (14.76 ± 1.652 years); the difference was statistically significant (F=5.245, P<0.001). There was also a statistically significant difference in whether they had been previously diagnosed with mental disorders (F=6.960, P=0.008). However, there were no statistically significant differences between the two groups in terms of place of residence, family income, parental education level, family history of mental disorders, presence of physical illnesses, and types of mood disorders.

The frequency of NSSI behaviors in patients was evaluated using the FASM. The patients were divided into five levels based on the self-reported number of instances of NSSI behaviors in the past 12 months: 0 times (never), 1-13 times (rarely), 14-38 times (occasionally), 39-106 times (often), and >106 times (frequently). Regarding NSSI behavior frequency levels, there was a statistically significant difference between males and females (X² = 3.195, P=0.001, Cramer’s V = 0.115); the proportion of patients at the “often” and “frequently” levels was significantly higher for female patients than for male patients, and females had a higher NSSI frequency than did males ( Table 1 ).

The analysis of the NSSI functions in the FASM revealed that males and females differed in emotion regulation (t=-5.336, P<0.001, Cohen’s d = -0.331) and attention seeking (t=2.380, P=0.017, Cohen’s d = 0.148) and that the NSSI functions of females were more oriented toward emotion regulation while those of males were more oriented to attention seeking ( Table 1 ).

Regarding the GAD-7 results, there was a significant difference in anxiety scale scores between males and females (t=2.724, P=0.007, Cohen’s d = -0.169), with female patients having higher anxiety scores than male patients ( Table 1 ).

The participants were divided into two groups by sex, each of which was subjected to ordinal regression. We constructed two ordinal regression models to explore the factors influencing the frequency of NSSI in male and female respectively. In all models, sex, age, family income, and previously diagnosed mental disorders were entered as predictors, NSSI frequency was entered as a dependent variable. The results of the parallel lines test for the male group was χ ² = 13.418, df = 12, p > 0.05, indicating the existence of the proportional advantage hypothesis. The analysis results ( Table 2 ) showed that factors associated with NSSI frequency in males included anxiety and age (p < 0.05). The result of the parallel lines test for the female group was χ ² = 15.923, df = 12, p > 0.05, indicating the existence of the proportional advantage hypothesis. Factors associated with NSSI frequency in females included anxiety, age, and previously diagnosed mental disorders (p < 0.05). Anxiety had a stronger positive predictive effect on the increase in NSSI frequency for females (OR=1.090) than for males (OR=1.064).

The participants were divided into two groups by sex, with 316 males and 1457 females, and regression analysis was performed separately for each group. The results of sex stratified analysis ( Table 3 ) showed that the positive predictive effect of anxiety on the three NSSI functions was present in both the male and female groups (ps <0.001), with the positive predictive effect of anxiety on emotion regulation, attention seeking, and social avoidance being slightly higher in males than in females.

The participants were divided into two groups by sex, with 316 males and 1457 females. Due to non-normal distributions of NSSI frequency and functions for each group, a nonparametric Spearman rank test was conducted separately for the two groups. Statistical analysis showed a significant correlation between frequency and the three functions of NSSI for both males and females ( Table 4 ). Effect sizes were evaluated according to Cohen’s (37) benchmarks: correlations of r = .10 to .30 were considered small, .30 to .50 were considered medium, and over .50 were considered large. The result ( Table 4 ) showed that emotion regulation had a moderate and significant correlation with NSSI frequency. Social avoidance and attention seeking both had a small and significant correlations with NSSI frequency.

The participants were divided into two groups by sex. Using Model 4 in the SPSS macro developed by Hayes (36), the mediating effects of NSSI motivation on the relationship between anxiety and NSSI frequency was tested separately for the male and female groups while controlling for age and previously diagnosed mental disorders. The results for the male group ( Tables 5 , 6 ) showed that anxiety had a significant predictive effect on NSSI frequency (β = 0.206, t = 3.727, p < 0.001). However, after including the mediator variables, the direct predictive effect of anxiety on NSSI frequency was not significant (β = 0.005, t = 0.094, p > 0.05). Anxiety had a significant positive predictive effect on the three types of NSSI functions (ps < 0.001), and the emotion regulation function in NSSI functions had a significant positive predictive effect on NSSI frequency (β = 0.415, t = 6.970, p < 0.001). In addition, both the upper and lower limits of the bootstrap 95% confidence intervals for the mediating effects of the emotion regulation functions excluded 0 ( Table 6 ), indicating that emotion regulation in NSSI functions played a full mediating role in the relationship between anxiety and NSSI frequency and that anxiety predicted NSSI frequency through emotion regulation in NSSI functions. The mediating effect of emotion regulation (0.176) accounted for 82.243% of the total effect (0.214).

The results for the female group ( Tables 5 , 6 ) showed that anxiety had a significant predictive effect on NSSI frequency (β = 0.272, t = 10.700, p < 0.001). Furthermore, after the mediator variable was included, the direct predictive effect of anxiety on NSSI frequency remained significant (β = 0.131, t = 5.126, p < 0.001). Anxiety had a significant positive predictive effect on all three types of NSSI functions (ps < 0.001), and emotion regulation (β = 0.363, t = 13.898, p < 0.001) and social avoidance (β = 0.073, t = 2.681, p < 0.01) in NSSI functions also had significant positive predictive effects on NSSI frequency. In addition, both the upper and lower limits of the bootstrap 95% confidence intervals for the direct effect of anxiety on NISS frequency and the mediating effects of emotion regulation and social avoidance excluded 0 ( Table 6 ), indicating that anxiety could not only directly predict NSSI frequency but also predict NSSI frequency through the mediating effects of emotion regulation and social avoidance in NSSI functions. The direct effect (0.130), the mediating effect of emotion regulation (0.130), and the mediating effect of social avoidance (0.019) accounted for 48.327%, 48.327%, and 7.063% of the total effect (0.269), respectively.