Teenage patients were recruited from 16 hospitals’ psychiatric outpatient clinics or inpatient departments in nine provinces of China using convenience sampling between December 2020 and December 2021. Inclusion criteria for the study were as follows: (1) adolescents between the ages of 12–18 years old; (2) a diagnosis of depressive episode made by experienced clinical psychiatrists in accordance with the ICD-10; (3) at least one instance of NSSI behavior in the past year; and (4) having finished primary education. The exclusion criteria were as follows: (1) experiencing acute episode of obvious psychotic symptoms, and (2) presenting with cognitive or intellectual impairments.

This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Institutional Review Board (IRB) of Shenzhen Kangning Hospital (020-k021-02). Written informed consent was obtained from the participants and their legal guardians.

Participants completed the questionnaire independently, with a trained research assistant available to answer any questions they might have had. To ensure data quality, research assistants would check the questionnaires after completion, and any unusual answers would be confirmed.

A total of 1,101 participants were included, with 199 (18.07%) males and 902 (81.93%) females, and a mean age of 14.7 (SD = 1.63). Nearly half of the participants (44.96%) had experienced their first episode of depression. More than half (64.12%) lived in cities. Only about a quarter (25.16%) were only-child, whereas 74.84% had siblings. Nearly 20% had experienced being left-behind, which is a common social phenomenon in China, especially in the underdeveloped regions. It refers to children who are left behind in their hometowns while their parents work and earn money in larger cities. In this study, participants who had experienced being left-behind for over 1 year were classified as left-behind children. Most children (77.38%) had parents in marriage, while 19.62% had parents divorced or separated, and 3% had one or both deceased. In terms of educational status, the majority of adolescents with depression (79.75%) were enrolled in school, 18.17% were suspended, and 2.09% had dropped out of school. The demographic and clinical information of the participants has been shown in Table 1.

The measurement of NSSI versatility was based on the number of methods reported by participants, and the sum of NSSI frequency under each method was used to calculate the total NSSI behavior frequency. NSSI duration was defined as the difference between the current and first NSSI age and there were 64 cases with missing data in the first NSSI age were excluded from duration calculation and related data analyses. The scores of the AR, SNR, and SPR subscales were calculated as the mean of their respective items due to differences in the number of items included. The skewness of all metric variables (except NSSI frequency) was within an acceptable range [<2 (30),]. To address non-normality in NSSI frequency, a natural logarithmic transformation was applied, resulting in approximately normal distribution of the transformed frequency data for use in all relevant analyses.

Descriptive statistics, including percentages, means, and standard deviations, were used to provide a concise report of the basic characteristics of NSSI in this sample. The prevalence of NSSI functions was determined through the calculation of endorsement rates for each function. Gender differences were examined using Mann–Whitney U tests for every single function and t-tests for function categories. Linear regression was conducted to analyze NSSI frequency. As for the two count variables, negative binomial regression was used for NSSI versatility and Poisson regression was for NSSI duration. It was due to the overdispersion observed in NSSI versatility (but not in NSSI duration), with its variance (6.40) exceeding the mean (4.30) and the dispersion parameter (1.11) was over 1. And binary logistic regression was utilized to explore suicide attempts (yes/no). In all regression analyses, the three function categories were first entered to construct category models. And then, models of the individual AR functions were taken for further examinations. Control variables, including age, gender, and hospital site, were included in all regression analyses to eliminate potential confounding effects. All analyses were performed using SPSS version 26.0. Statistical significance was set at a two-tailed p < 0.05 (as the default value), and Bonferroni correction may be implemented in the event of necessity.

According to the criteria of Muehlenkamp et al. (31), patients were stratified into three groups based on the NSSI frequency they reported: low-frequency (<5), medium-frequency (5–25), and high-frequency (≥25), corresponding to proportions of 9.72%, 28.07%, and 62.22%, respectively. The mean number of methods employed was 4.30 (SD = 2.53), with the most common being “cut or crave on skin” (87.83%) and “hit self” (52.59%). The mean age of onset for NSSI was 13.31 (SD = 1.71) years, with an average duration of 1.41 (SD = 1.21) years. More than half of adolescents with depression (51.50%) engaged in the behavior without hesitation, while some (26.61%) hesitated for only a few minutes, and the majority (88.28%) did not hesitate for more than 1 h. Regarding the physical pain experienced during NSSI, more than half (51.23%) reported mild pain, nearly 30% did not feel any pain, 17.71% reported moderate pain, and only 2.45% reported severe pain.

The most frequently reported functions were “to stop bad feelings” and “to feel relaxed,” followed by “to relieve feeling numb or empty,” all of which fell under the AR category. The most commonly reported functions of SNR were “to avoid doing something unpleasant you do not want” and “to avoid being with people.” The function of “to let others know how desperate you were” and “to get your parents to understand or notice you” received the highest endorsement in SPR. Further elaboration on the prevalence of NSSI functions can be observed in Table 2.

Only 15.99% of participants endorsed a single function category, which was AR (15.44%) in most cases. Furthermore, 29.43% of participants endorsed two function categories, including AR + SNR (17.08%) and AR + SPR (11.99%). More than half of the participants (54.59%) endorsed all three function categories.

A significant difference was observed in the degree of recognition of the three function categories (F = 650.73, p = 0.000***, η_p² = 0.28), with post hoc analysis indicating that all pairs significantly differed from each other (p = 0.000***). And the results indicated that the most commonly endorsed function category of NSSI in adolescents with depression was AR, followed by SNR, and then SPR.

However, there were some differences between males and females in the prevalence of NSSI functions. Specifically, females endorsed AR > SNR > SPR (F = 611.07, p = 0.000***, η_p² = 0.31; p = 0.000***). Males endorsed AR > SPR and SNR (F = 59.47, p = 0.000***, η_p² = 0.17) and no significant difference was noted between SPR and SNR (p = 0.54). Females endorsed each function of AR at a higher rate than males, with differences ranging from 1.34% to 12.61%. The mean score of AR category in females was significantly higher than in males, with a moderate effect size. The Mann–Whitney U test showed that females had higher endorsement in the function of “to punish yourself,” “to feel relaxed,” and “to stop bad feelings” than males. Males endorsed each function of SPR higher than females, especially “to try to get a reaction from someone, even negative,” for which the rate was 20.01% lower in females, indicating a significant difference. This was where the greatest gender difference was observed. Males also gave more endorsement to the function of “to receive more attention from your parents or friends” and “to get attention” in SPR than females, and the score of SPR category was significantly higher in females than in males. However, no significant difference existed in SPR between males and females. Age did not correlate with social function and had a significant but weak negative correlation with AR (r = −0.07, p = 0.018*).

Age, gender, hospital site, and the three function categories were inputted simultaneously (model 1) to conduct the linear regression analysis of NSSI frequency. Result revealed that except age, only AR was significant after the Bonferroni correction. Although the p-value of SNR was below 0.05, it did not meet the Bonferroni correction and the beta weight was small. R² of the model was 0.23, which mean that the model accounted for approximately 23% variation of NSSI frequency. In Model 2, which focused on the three specific AR functions, affect regulation, anti-dissociation, and self-punishment were all independently associated with NSSI frequency, see Table 3.

In the negative binomial regression analysis for NSSI versatility, another indicator of NSSI severity, AR was also the only significant function category. And in model 2, only anti-dissociation and self-punishment survived the Bonferroni correction.

The same procedure was conducted for the Poisson regression of NSSI duration, and likewise only AR was significantly associated with enduring NSSI. However, only anti-dissociation survived in model 2, while self-punishment did not meet the Bonferroni correction and affect regulation failed to reach statistical significance totally.

Among adolescents with depression and NSSI, the vast majority (97.09%) reported having suicidal thoughts, and nearly half (49.59%) reported lifetime suicide attempts. Additionally, over 30% experienced both NSSI and suicide attempts in the recent 1 year.

Through binary logistic regression analysis, only affect regulation (AR) was found to be associated with an increased risk of suicide attempts. Specifically, an increase in the endorsement of self-punishment was independently relevant to concurrent suicide attempts (refer to Table 4 for detailed results).

The most prevalent function was affect regulation (“to stop bad feelings” and “to feel relaxed”), which is in line with the prevailing view that NSSI is primarily used to alleviate negative emotions, reduce arousal levels, and simultaneously bring about positive feelings, such as relaxation or calmness (8, 32). The nature of NSSI, as a maladaptive emotion regulation strategy, has also been demonstrated in adolescents with depression (33). Anti-dissociation (e.g., “to relieve feeling numb or empty”) was the second most common function, which was somewhat at odds with general findings. Meta-analyses (14, 34) and a study on BPD adolescents (16) reported that anti-dissociation was the relatively less endorsed function. However, our sample revealed a higher prevalence of it, which may be attributed to the association of depression with feelings of numbness and emptiness. NSSI can provide a person with physical and emotional sensations that facilitate a sense of authenticity and self-awareness, which could be the compensation and particularly reinforce the recognition of the anti-dissociation function in depression (6). In addition, a study using pathway analysis found that the relationship between post-traumatic stress, depression, and self-harm is mediated through dissociation, suggesting that anti-dissociation may be the crucial factor in the high co-occurrence of NSSI and depression (35).

Pronounced trait of avoidance was observed in NSSI function. Consistent with the hypothesis proposed by Nock and Prinstein (7), NSSI functions involve both decrease in negative outcomes and increase in positive outcomes, which often occur simultaneously and intertwine (36, 37). However, our results revealed a greater endorsement of negative aspects compared to positive ones (e.g., “to relieve feeling numb or empty” exceeded “to feel something, even if it was pain”). Similar findings have been reported in other studies (28, 38). These results suggested that avoiding unwanted internal states may be more desirable than inducing positive states or that negative reinforcement may be more potent in NSSI, which supports the experience avoidance model of NSSI (39). Individuals who engage in NSSI may be experiencing severe psychological distress and require urgent professional assistance.

The majority of adolescents with depression in this study had both automatic and social functions, and more than half recognized all three functional categories, which demonstrates the cross-functionality of NSSI. AR was consistently more prevalent than SNR and SPR, regardless of gender. However, females endorsed SNR more than SPR, while this difference was not observed in males. Males had significantly higher SPR scores than females, indicating a greater likelihood of self-harm to gain positive social feedback, such as attention or response. One theoretical perspective suggests that NSSI serves as a powerful form of communication to signal psychological distress and avoid abandonment or seek understanding (40, 41). (42) conducted a diary study and found that disclosing NSSI behaviors significantly increased perceived social support. NSSI may be a means of influencing the environment, reflecting the adaptability of this behavior, particularly in males who may experience difficulty expressing their emotions or needs. Additionally, we found that females had significantly higher AR scores than males, who might be more likely to engage in NSSI as a means of self-punishment or affect regulation. This finding is consistent with previous research (43) and may be attributed to the greater prevalence of severe internal problems among females (44).

AR was associated with more severe and persistent NSSI in this study, which is in line with the results of previous studies (42, 46). Although affect regulation was the most popular function, it no longer played the dominant role in the association between NSSI function and worse behavioral outcomes. Our study further confirmed that the three individual automatic functions, anti-dissociation, affect regulation, and self-punishment were associated with NSSI frequency, and anti-dissociation together with self-punishment were associated with NSSI versatility. Only anti-dissociation was independently associated with NSSI duration. Moreover, anti-dissociation beyond affect regulation showed the strongest overall association, which has also been partially reflected in Reinhardt’s study (46) that reported correlation coefficients between them. And our study further confirmed these relationships through regression analysis.

Our findings partly confirmed and refined the findings of a preliminary study by Yen et al. (29) that it’s anti-dissociation rather than affect regulation that was involved in the maintenance of NSSI. This can be understood in terms of stability. A follow-up study (47) found that affect regulation was the most variable function. Negative emotions easily subside as the internal or external environment improves, at which point there appears to be no reason to continue NSSI.

AR was identified as a significant factor associated with suicide attempts, consistent with prior research (42). However, our study revealed that only an increased in the self-punishment function was independently associated with an elevated risk of suicide attempt. Chapman et al. (39) posited that NSSI as a form of self-punishment could alleviate negative self-cognition, shame, and self-hatred. And it has been supported by the findings of experimental studies (48). However, unlike other functions, self-harm motivated by self-punishment implicates anger and aggression directed toward oneself. It can run through the continuum from NSSI to suicidal self-injury and escalate to the latter (49). That is, when the motivation to punish oneself expands to some extent, the most extreme form, suicide, may be the chosen course of action.

It is generally believed that intense psychological distress and despair are the primary drivers of suicide (50). Surprisingly, we did not observe a role of affect regulation in the co-occurrence of NSSI and suicide. This could be because affect regulation was already effectively achieved through NSSI. As reported by Saraff et al. (21), the majority of participants (94%) recognized the role of NSSI in emotional relief. Brausch and Muehlenkamp (47) also found that affect regulation was the most effective function in self-reporting. These findings demonstrate the value of NSSI as a coping strategy, as it may mitigate the risk of more serious consequences, even if doing so incurs costs.

Our findings highlight the significance of conducting a comprehensive assessment of NSSI function. A high endorsement of automatic functions could indicate severe NSSI and underlying suicide attempts, which necessitate close monitoring and intensive treatment. In addition, given the multifacetedness NSSI function, mental health practitioners and clinicians should comprehend the NSSI functions that patients endorse and devise appropriate treatment plans. Currently, the primary therapy for NSSI is dialectical behavioral therapy [DBT (51),], which concentrates on affect regulation and enhancement of emotion regulation ability and distress tolerance. However, given our findings that functions of anti-dissociation and self-punishment might be the more important risky factors, we recommend that treatments targeting dissociation and negative self-cognition should be considered, such as behavioral activation (52), mindfulness (53), and cognitive behavioral therapy (CBT), particularly in individuals experiencing depression.

This study has some limitations that should be considered when interpreting the findings. Firstly, the data were cross-sectional, which means that only associations between NSSI functions and behavioral outcomes were examined, and directionality and causality could not be clarified. Future longitudinal studies should be considered to verify these findings and explore the predictive role of NSSI functions in treatment effects or the transition from NSSI to suicide. Secondly, participants were only asked to report on NSSI that occurred within the past year. The lifetime data was unknown, which represents the overall level and may potentially impacting the results. Thirdly, self-reported data may be subject and exist bias. The use of ecological momentary assessment techniques could be considered to obtain more objective and impartial behavioral and physiological data. Finally, NSSI function is not necessarily equivalent to reinforcement, as sometimes the functions are not effectively satisfied. We suggest that future studies could include measures of function validity to gain a deeper understanding of associations between NSSI functions and behavioral outcomes from these subtle but essential perspectives.

In this study, we shed light on the significance of understanding the functions of NSSI in identifying prevalent motivations and potential risk factors associated with severe behaviors. Based on a large sample, we identified affect regulation as the most prevalent function, and found an increase in the endorsement of anti-dissociation function in adolescents with depression. We also found gender difference that females recognized AR more frequently than males, while males had a higher prevalence of SPR. AR was not only prevalent but also the risky function category linked to severe outcomes. However, function of affect regulation no longer played the primary role in associations between NSSI function and behavioral consequences as it did in the prevalence aspect. Instead, anti-dissociation and self-punishment seemed to be more risky factors. Indicators of these functions should be given more attention to risk evaluation. We recommend that more attention should be given to these functions in risk evaluation of NSSI, and the targeted intervention or treatment should be developed accordingly in a timely manner.