Introduction
Non-suicidal self-injury (NSSI) is defined as the deliberate and repeated infliction of direct injury to one’s own bodily tissues without suicidal intent (Klonsky, Reference Klonsky2007). Adolescents are at a critical stage of physical and mental development (Best & Ban, Reference Best and Ban2021), which are particularly vulnerable to NSSI (Xiao et al., Reference Xiao, Song, Huang, Hou and Huang2022). The study found that adolescents with NSSI, regardless of its frequency, experienced more stress and mood disorders in early adulthood compared to those without NSSI (Daukantaitė et al., Reference Daukantaitė, Lundh, Wångby-Lundh, Claréus, Bjärehed, Zhou and Liljedahl2021). NSSI is not only commonly associated with mental disorders (Hirot et al., Reference Hirot, Ali, Blanchet, Grandclerc, Gicquel, Godart, Berthoz, Lalanne, Duclos, Mattar, Roux, Thiébaud, Vibert, Hubert, Courty, Ringuenet, Benoit, Blanchet and Godart2025; Kaufman et al., Reference Kaufman, Coon, Shabalin, Monson, Chen, Staley, Keeshin, Docherty, Bakian and DiBlasi2024; Poudel et al., Reference Poudel, Lamichhane, Magar and Khanal2022; Serra et al., Reference Serra, Presicci, Quaranta, Caputo, Achille, Margari, Croce, Marzulli and Margari2022), but it may also lead to serious consequences such as suicide (Hawton et al., Reference Hawton, Bergen, Kapur, Cooper, Steeg, Ness and Waters2012; Whitlock et al., Reference Whitlock, Muehlenkamp, Eckenrode, Purington, Baral Abrams, Barreira and Kress2013). In conclusion, adolescent NSSI seriously undermines their own health and has become a major challenge to global public health. Therefore, it is necessary to further investigate the development trends and influencing factors.
Trends in NSSI over time
It is generally considered that NSSI is prevalent among adolescents (Xiao et al., Reference Xiao, Song, Huang, Hou and Huang2022). A meta-analysis revealed that the global lifetime prevalence of NSSI among adolescents was 22.1% (Lim et al., Reference Lim, Wong, McIntyre, Wang, Zhang, Tran, Tan, Ho and Ho2019). The prevalence varies in different countries and regions: 27.6% in Europe (Brunner et al., Reference Brunner, Kaess, Parzer, Fischer, Carli, Hoven, Wasserman, Sarchiapone, Resch, Apter, Balazs, Barzilay, Bobes, Corcoran, Cosmanm, Haring, Iosuec, Kahn, Keeley and Wasserman2014); 6.4%–30.8% in the United States (Monto et al., Reference Monto, McRee and Deryck2018); 24.7% in China (Qu et al., Reference Qu, Wen, Liu, Zhang, He, Chen, Duan, Yu, Liu, Zhang, Ou, Zhou, Cui, An, Wang, Zhou, Yuan, Tang, Yue and Chen2023). Despite these variations, there seems to be a consistent upward trajectory in global NSSI prevalence across recent years (Cipriano et al., Reference Cipriano, Cella and Cotrufo2017; Turner et al., Reference Turner, Robillard, Ames and Craig2022).
Longitudinal trends in NSSI severity (continuous variable) compared to NSSI prevalence (binary variable) remain inconclusive. Most findings indicate that NSSI severity tends to increase over time. For instance, a longitudinal study involving 6,092 Chinese adolescents revealed an upward trend in NSSI over two and a half years of follow-up (Xiong et al., Reference Xiong, Liao, Luo, Luo, Tong and Li2023). Similar results have been reported in longitudinal studies from Canada (Daly & Willoughby, Reference Daly and Willoughby2019), the United States (Glenn et al., Reference Glenn, Kleiman, Cha, Nock and Prinstein2016; Jewett et al., Reference Jewett, Jewett, Borowsky, Mathiason, Taliaferro and Areba2023), Spain (Faura-Garcia et al., Reference Faura-Garcia, Calvete and Orue2024), and Australia (Scott et al., Reference Scott, McGrath, Beard, Chislett, Baldwin, Nehme, Lubman and Ogeil2024). However, some follow-up studies have observed a decreasing trend in NSSI severity among high school students (Guerry & Prinstein, Reference Guerry and Prinstein2009). The discrepancies in these findings may stem from complex factors, including subject characteristics, social environments, and measurement instruments (Muehlenkamp et al., Reference Muehlenkamp, Claes, Havertape and Plener2012). These inconsistencies suggest that NSSI may be influenced by regional distribution, a hypothesis that will be further explored in this study.
However, existing longitudinal studies typically span only a few years (Buelens et al., Reference Buelens, Luyckx, Gandhi, Kiekens and Claes2019; Marshall et al., Reference Marshall, Tilton-weaver and Stattin2013; Wei et al., Reference Wei, Liu, An and Wang2025), which limits their ability to reflect long-term changes in NSSI severity among adolescents. Given that frequency serves as a clinical severity marker for NSSI (Ammerman et al., Reference Ammerman, Jacobucci and McCloskey2018) and can effectively predict its continuation (Brausch & Boone, Reference Brausch and Boone2015), further investigation into trends in NSSI frequency is warranted.
Influencing factors of NSSI
Individual characteristics are frequently studied as NSSI influences. Longitudinal research indicates that NSSI development may fluctuate during puberty (De Luca et al., Reference De Luca, Pastore, Palladino, Reime, Warth and Menesini2023). Additionally, females may exhibit higher NSSI reporting than males due to hormonal differences and gender-based variations in emotion regulation (Balzer et al., Reference Balzer, Duke, Hawke and Steinbeck2015; Gao et al., Reference Gao, Guo, Wu, Huang, Wu and You2021; Nolen-Hoeksema & Aldao, Reference Nolen-Hoeksema and Aldao2011). Moreover, compared to non-clinical samples, adolescents with diagnosed Non-Suicidal Self-Injury Disorder (NSSID) typically display a higher frequency and greater variety of NSSI methods (Washburn et al., Reference Washburn, Potthoff, Juzwin and Styer2015).
Based on the modified integration model proposed by Jacobson and Batejan (Reference Jacobson, Batejan and Nock2014), NSSI among adolescents involves not only individual psychological and physiological mechanisms but is also closely linked to environmental factors. However, prior studies have predominantly focused on micro-level environments, such as adverse childhood experiences and bullying (Wang et al., Reference Wang, Luo, Hong, Yang, Zhao and Jia2022), with limited exploration of macro-level factors. While some studies have considered social factors, their data rely predominantly on self-reports. To the best of our knowledge, no study has systematically examined the relationship between these macro-level social factors and NSSI. Therefore, this study draws on and integrates social indicators from prior researches (Miranti & Mendez, Reference Miranti and Mendez2023; Pratama & Al-Shaikh, Reference Pratama and Al-Shaikh2012; Rogerson, Reference Rogerson2013; Titisari & Santoso, Reference Titisari and Santoso2025) to analyze the macro-social factors affecting NSSI among adolescent. Specifically, GDP per capita, the Human Development Index, and Internet penetration rate are selected as proxies for social development, the Gini coefficient as a proxy for social equity, and the happiness index as a proxy for social well-being in the study.
First, social development may be inversely related to adolescent NSSI. Data from 171 countries reveal significant disparities in mental health resource allocation and accessibility across nations with varying income levels (Lora et al., Reference Lora, Hanna and Chisholm2020). This suggests that high-income countries, with their more robust psychological service systems, may better mitigate NSSI through specialized interventions such as outpatient care (Lora et al., Reference Lora, Hanna and Chisholm2020) and psychological hotlines (Matthews et al., Reference Matthews, Cantor, Brooks Holliday, Eberhart, Breslau, Bialas and McBain2023). A meta-analysis also found that NSSI prevalence is lower in developed than in developing countries (Deng et al., Reference Deng, Zhang, Zhang, Yan, Zhuang, Liu, Li, Xue and Wang2023).
However, the Internet plays a dual role in the lives of adolescents, who constitute the demographic with the highest Internet usage rates (Corcoran & Andover, Reference Corcoran and Andover2020). While the Internet offers a platform for anonymous communication, potentially reducing social isolation (Cho, Reference Cho2015), thereby possibly decreasing NSSI, a growing body of research highlights its negative impacts. Adolescents may be affected by social contagion when they are exposed to NSSI related content on the Internet (Brown & Plener, Reference Brown and Plener2017). Moreover, the Internet can reinforce NSSI behaviors, provoke self-injurious impulses, and stigmatize NSSI (Brown et al., Reference Brown, Fischer, Goldwich, Keller, Young and Plener2018).
Second, social equity may be inversely related to adolescent NSSI. According to relative deprivation theory, individuals who perceive themselves as disadvantaged may experience intense negative emotions (Smith et al., Reference Smith, Pettigrew, Pippin and Bialosiewicz2012) and may engage in risky behaviors (Balsa et al., Reference Balsa, French and Regan2014; Liao et al., Reference Liao, Wang, Ni and Yu2024). A meta-analysis also indicated that economically disadvantaged adolescents report more mental health problems (Kim & Hagquist, Reference Kim and Hagquist2018).
Third, social well-being may be inversely associated with adolescent NSSI. High social well-being, such as high levels of education and happiness, is positively linked to better mental health (Van Lente et al., Reference Van Lente, Barry, Molcho, Morgan, Watson, Harrington and McGee2012). Individuals with high well-being typically possess strong social relationships and the ability to respond adaptively to life events (Diener & Seligman, Reference Diener and Seligman2002), which may reduce the risk of NSSI.
Current study
Cross-temporal meta-analysis can uncover trends in variables by integrating isolated studies chronologically (Twenge, Reference Twenge1997; Xin & Chi, Reference Xin and Chi2008). Researchers often include studies using the same measurement tool (Karazsia et al., Reference Karazsia, Murnen and Tylka2017) or different versions of the same scale such as revised or shortened versions (Buecker et al., Reference Buecker, Mund, Chwastek, Sostmann and Luhmann2021) to maintain the conceptual consistency of psychological variables across studies (Wang et al., Reference Wang, Xin and Hou2024).
The Deliberate Self-Harm Inventory (DSHI), a widely used measure of non-suicidal self-injurious behaviors globally, was employed in this study (Fliege et al., Reference Fliege, Kocalevent, Walter, Beck, Gratz, Gutierrez and Klapp2006; Latimer et al., Reference Latimer, Meade and Tennant2013; Vigfusdottir et al., Reference Vigfusdottir, Dale, Gratz, Klonsky, Jonsbu and Høidal2022). The scale was originally developed by Gratz (Reference Gratz2001) and comprises 17 items. Over the past two decades, various revisions and short versions of the DSHI have been created. This meta-analysis included studies utilizing the original DSHI-17 and its derivatives. These versions, while differing in length, retain the core behavioral definition and assessment approach of NSSI from the original scale, ensuring conceptual comparability. For example, the DSHI-simplified version (DSHI-s) developed by Lundh et al. (Reference Lundh, Karim and Quilisch2007) improves response efficiency and is more suitable for adolescents. The DSHI-9, a subsequent revision, has also demonstrated good internal consistency and retest reliability in adolescent samples (Bjärehed & Lundh, Reference Bjärehed and Lundh2008). Some researchers, based on their specific research backgrounds and previous studies (Nock, Reference Nock2010; You & Lin, Reference You and Lin2015), have excluded items less common in adolescent NSSI behaviors. By incorporating different DSHI versions, this study aims to cover diverse geographic regions, reduce version-related biases, and comprehensively explore chronological changes in adolescent NSSI (Buecker et al., Reference Buecker, Mund, Chwastek, Sostmann and Luhmann2021). The DSHI is scored by summing the item scores, with higher scores indicating greater NSSI frequency. The included studies show good internal consistency (α = 0.70–0.97).
This study aims to explore the trends in adolescent NSSI severity over time and the influence of individual and social factors. The following hypotheses were formulated: (1) NSSI among adolescents increases over time (Hypothesis 1). (2) Demographic variables may influence the trend of NSSI among adolescents. Specifically, female adolescents, older adolescents, and those with clinical samples may exhibit a more pronounced increase in NSSI (Hypothesis 2). (3) Social indicators may be associated with NSSI among adolescents. Specifically, higher GDP per capita, human development index, and happiness index may correlate with reduced NSSI severity, while higher Internet penetration and Gini coefficient may correlate with increased NSSI severity (Hypothesis 3). This study offers quantitative evidence on changes in adolescent NSSI severity, contributing to inform its prevention and intervention.
Methods
This meta-analysis study is pre-registered with PROSPERO (registration number: CRD420250656170).
Literature search
The following criteria guided literature inclusion: (a) Studies using the DSHI scale as a measurement tool. (b) Empirical studies reporting descriptive statistics (sample size, mean, and standard deviation). (c) Studies focusing on adolescents aged 10–19 (Sawyer et al., Reference Sawyer, Azzopardi, Wickremarathne and Patton2018). (d) For studies using follow-up data, only the first measurement was included. (e) For duplicate articles by the same author using the same dataset, only the earliest publication was selected. (f) Studies written in Chinese or English.
We searched Chinese and foreign language database such as CNKI, Wanfang, VIP, Web of Science, PubMed, Scopus, ProQuest, Wiley. Search terms included ‘adolescents’, ‘non-suicidal self-injury’, and ‘deliberate self-harm inventory’ along with their synonyms in both English and Chinese (see Supplementary Material Table A1). Literature collection concluded on December 19, 2024, yielding 70 eligible publications. These articles were published between 2008 and 2024, including 96,382 adolescent subjects. In the case of journal articles where the specific data collection years are not indicated, the data collection year (hereinafter referred to as the ‘year’) are estimated by subtracting two years from the publication year. In the case of theses, the year are estimated by subtracting one year from the actual publication year (Oliver & Hyde, Reference Oliver and Hyde1993; Twenge, Reference Twenge2001; Wang et al., Reference Wang, Xin and Hou2024). Our study’s data spanned from 2007 to 2023.
Data extraction
The dataset for this study was constructed based on prior transect historical research (Wang et al., Reference Wang, Xin and Hou2024) and involved the following steps. Each article was assigned a unique identifier, and basic data (including sample size, mean, and standard deviation), publication year, journal type (1 = core journal, 2 = general journal, 3 = Master’s thesis/Preprint version), year of data collection, mean age of participants, female proportion, and clinical state was coded and entered into the dataset by two psychology graduate students independently following preset coding rules. Specifically, clinical samples were defined as participants recruited from mental health institutions or diagnosed professionally, while non-clinical samples were defined as those recruited from the community or school without known mental disorders. Discrepancies were resolved through discussion.
For studies that only provided sub-study data (e.g. gender subgroups) without overall results, we used weighted synthesis of sub-study outcomes (Wang et al., Reference Wang, Xin and Hou2024). Additionally, to ensure comparability of DSHI scores across versions, we performed a percent of maximum possible (POMP) transformation on the mean DSHI scores and standard deviations (Buecker et al., Reference Buecker, Mund, Chwastek, Sostmann and Luhmann2021; Cohen et al., Reference Cohen, Cohen, Aiken and West1999), employing these scores to assess NSSI severity.
Quality assessment
This study employed the Joanna Briggs Institute (JBI) Critical Appraisal Checklist (Sabilillah, Reference Sabilillah2020) to systematically evaluate the quality of the included literature. Two researchers independently completed the quality coding. The checklist comprises eight assessment items, each scored as 0 or 1. The final total score ranges from 0 to 8, with higher scores indicating better study quality.
Sources of social indicator data
The social indicators are compiled and disseminated by a variety of United Nations-affiliated organizations. The detailed sources and statistical agencies are presented in Table 1. GDP per capita serves as an indicator reflecting the economic scale and the average economic output per inhabitant of a nation or region. It is derived by dividing the gross domestic product (GDP) by the average annual population. The Human Development Index (HDI) encompasses the geometric mean of three dimensions: health, education, and income, functioning as a composite measure of a country’s overall development in these three domains. Internet penetration signifies the proportion of individuals within a population who have accessed the Internet (via fixed or mobile networks) from any location over the past three months. The Gini coefficient is an index quantifying the extent of inequality in the distribution of income or wealth, with 0.4 recognized as the international warning threshold. The Happiness Index denotes the overall satisfaction and emotional experiences of a population with their lives. It is gauged through life assessment inquiries utilizing the Cantril Self-Anchoring Striving Scale (Cantril Ladder), where individuals are requested to rate their lives on a scale of 0–10.
Table 1. The source of social indicator
Meta-analytic procedure
Data were analyzed using R 4.4.1 (https://www.r-project.org). Initially, a linear regression model was utilized to characterize the overall trend of adolescent NSSI over time. Secondly, to better estimate the overall score, the scores were weighted by sample size (Wang et al., Reference Wang, Xin and Hou2024). Then, we also statistically controlled for the effects of gender, age, clinical state, and journal type (Xin & Zhang, Reference Xin and Zhang2009; Wang et al., Reference Wang, Xin and Hou2024). To assess the robustness of the results of this meta-analysis, this study used the Leave-One-Out (LOO) method of sensitivity analysis (Dodell-Feder & Tamir, Reference Dodell-Feder and Tamir2018). All model fitting outcomes were visualized utilizing the ggplot2 package.
Subsequently, the change in adolescent DSHI scores was quantified based on the linear regression model results. To estimate the temporal variation in NSSI, a regression equation (Y = BX + C) was developed to predict NSSI scores, where Y represents the predicted score, B denotes the unstandardized regression coefficient, C signifies the regression intercept, and X corresponds to the year of data collection. This equation was employed to compare predicted NSSI levels between 2007 (the earliest study in the dataset) and 2023 (the most recent study). The change in NSSI was standardized by dividing it by the mean standard deviation reported in the study, with effect sizes expressed in Cohen’s d units. Similar cross-temporal meta-analytic procedures have been employed in prior research (e.g. Buecker et al., Reference Buecker, Mund, Chwastek, Sostmann and Luhmann2021; Curran & Hill, Reference Curran and Hill2019; Twenge, Reference Twenge2000).
Subgroup analyses were also conducted. Given the geographical, cultural, and social normative differences across regions that may influence study outcomes (Rentfrow, Reference Rentfrow2016; Stankov, Reference Stankov2011), and considering that continents can serve as proxies for these regional differences (Buecker et al., Reference Buecker, Mund, Chwastek, Sostmann and Luhmann2021), this investigation adopted a continental subgroup division to examine temporal changes in adolescent NSSI across different continents.
Finally, correlation and lagged correlation analyses were performed to investigate the associations between social indicators and adolescent NSSI from three years prior, one year prior, and the current year, in that sequence(Du et al., Reference Du, Wu, Yalikun, Li, Jia, Duan, Zhou and Ren2025; S. Xin et al., Reference Xin, Zhang, Sheng, Zhao and Peng2023).
Results
Basic characteristics of the included studies
The literature screening and inclusion process is presented in Figure 1, and the basic characteristics of the included studies are outlined in Table 2. Overall, 71 studies from 70 publications were incorporated, involving a total sample size of 96,382 participants. The average proportion of female participants across all independent samples was 51.14% (SD = 14.03), with a range spanning from 1.51% to 100%. The sample’s mean age was 14.01 years (SD = 1.36) and an overall range from 10.32 to 16.69 years. Clinical samples constituted approximately 11.27% of the total. The majority of the studies were conducted in Asia (k = 53), followed by Europe (k = 12), North America (k = 3), and Oceania (k = 3).
Figure 1. PRISMA diagram showing the results of the literature search.
Table 2. Descriptions of studies included in the overall meta-analysis (71 studies in total)
Note: Journal: 1 = core journal; 2 = general journal; 3 = degree thesis. M = mean is converted into POMP scores. SD = standard deviation is converted into POMP scores.
Among them, * represents the articles included in the meta-analysis.
Overall change in NSSI among adolescents over time
Initially, analyses of the 71 included studies examined temporal trends in adolescent NSSI severity (Figure 2). A linear regression analysis incorporating sample size weighting revealed that year was not a significant predictor of adolescent NSSI (b = 0.22, SE = 0.26, t (69) = 0.41, p = 0.54, 95% CI = [−0.31, 0.75]). Using the statistical parameter correction method with White’s robust standard errors, the test results indicate that the negative predictive effect of age is not significant (b = 0.22, SE Het-Robust = 0.21, 95% CI = [−0.20, 0.64], R 2 = −0.004, p = 0.295). This indicates that there was no statistically significant change in the severity of NSSI among adolescents between 2007 and 2023.
Figure 2. Trend of NSSI among adolescents over time.
Subsequently, journal type was incorporated into the linear regression model as a statistical control variable. The regression results demonstrated that journal type (b = −0.78, SE = 1.68, t (68) = −0.47, p = 0.64, 95% CI = [−4.13, 2.56]) was not a significant predictor. Moreover, year (b = 0.15, SE = 0.32, t (67) = 0.48, p = 0.63, 95% CI = [−0.56, 0.74]) continued to exert no statistically significant influence on the severity of NSSI in adolescents.
Finally, statistical controls were applied to gender, age, and clinical state. The results indicated that neither gender (b = −0.05, SE = 0.03, t (65) = −1.62, p = 0.11, 95% CI = [−0.12, 0.01]) nor age (b = −0.08, SE = 0.39, t (65) = −0.21, p = 0.83, 95% CI = [−0.87, 0.70]) significantly predicted NSSI in adolescents. In contrast, clinical state (b = 24.07, SE = 3.49, t (65) = 6.89, p < 0.001, 95% CI = [17.10, 31.05]) was a significant positive predictor. At this stage, year (b = 0.12, SE = 0.13, t (65) = 0.91, p = 0.37, 95% CI = [−0.14, 0.38]) remained a non-significant positive factor. To conclude, these results suggest that the severity of NSSI among adolescents has not changed significantly from 2007 to 2023.
The sensitivity analysis revealed that excluding individual samples had minimal impact on the results. The b-coefficients and p-values of the model showed no significant changes, indicating the model’s robustness (see Supplementary Material Figure A1). Random-effects pooling revealed substantial between-study heterogeneity (Q (5) =382.46, p < .001; I 2 = 98.9%).
Amount of change in NSSI among adolescents over time
Based on the linear regression model results, the effect size d was expressed as a standardized score (Buecker et al., Reference Buecker, Mund, Chwastek, Sostmann and Luhmann2021). The linear regression analysis yielded an approximate equation of Y = 0.22X − 437.23. Subsequently, the Z-scores for adolescent NSSI were calculated by substituting the years 2007 and 2023 into the regression equation, resulting in values of 4.31 and 7.83 respectively. Finally, the difference between Z2023 and Z2007 was computed and divided by the mean standard deviation of the Z-scores over the 16-year period, which was 9.99, to obtain a d-value of 0.35. These findings indicate that the severity of NSSI among adolescents increased by 0.35 standardized scores from 2007 to 2023. According to established criteria where ‘0.2 represents a small effect size, 0.5 a medium effect size, and 0.8 a large effect size’ (Buecker et al., Reference Buecker, Mund, Chwastek, Sostmann and Luhmann2021; Cohen, Reference Cohen1977; Twenge, Reference Twenge2001), the increase in adolescent NSSI severity from 2007 to 2023 corresponds to a small to medium effect size.
Subgroup analysis
Subgroup analyses were conducted to examine potential continental differences in temporal trends of NSSI. The results indicated that relative to European samples, American samples exhibited a non-significant effect size difference of −6.52 (95% CI [−44.67, 31.62]), Asian samples showed a non-significant difference of 4.14 (95% CI [−3.39, 11.66]), and Oceanian samples demonstrated a non-significant difference of 5.94 (95% CI [−9.11, 20.98]). White’s robust standard error test shows that the age effect remains insignificant (b = 0.14, SE Het-Robust = 0.24, 95% CI = [−0.34, 0.61], p = 0.572). Therefore, NSSI changes with year did not significantly differ across continents. These continental comparisons were exploratory in nature, so no specific hypotheses were predefined.
Correlations between adolescent NSSI and Social Indicators
The correlation and lagged correlation analyses revealed that social development indicators (GDP per capita, HDI, and Internet penetration), social equity (Gini coefficient), and social well-being (Happiness Index) exhibited significant correlations with adolescent NSSI (see Table 3).
Table 3. Correlations between social indicators and adolescent NSSI
Note: *p < 0.05, **p < 0.01; GDP = GDP Per Capita; HDI = Human Development Index; INT = Internet Penetration Rate; Gini = Gini Coefficient; HAP = Happiness Index
Quality assessment
The Spearman correlation coefficient between the two raters was 0.75 (p < 0.01), indicating a strong level of agreement. The quality assessment revealed that the included studies had a mean score of 6.55 ± 0.86, with individual scores ranging from 5 to 8, suggesting that all studies were of moderate to high quality. The specific item scores after resolving the disagreements are presented in Supplementary Material Table A2. Subsequent analyses revealed that the moderating effects of the quality assessment score were not statistically significant (b = −0.05, SE = 0.34, t (67) = −0.16, p = 0.88, 95% CI = [−0.72, 0.62]). Therefore, to maintain the integrity of the evidence, no studies were excluded based on quality scores.
Discussion
This study employed a cross-temporal meta-analysis to examine trends in adolescent NSSI and the influencing factors from 2007 to 2023. The results revealed that the severity of NSSI exhibited a small to moderate upward trend from 2007 to 2023, but this trend was not statistically significant. Subgroup analyses indicated no significant continental variation in NSSI trends. Furthermore, gender, social development, social equity, and social well-being significantly impacted these trends. These findings enhance our understanding of adolescent NSSI and may inform effective intervention strategies.
Change in of NSSI among adolescents over time
This study found that the severity of NSSI among adolescents exhibited a non-significant change from 2007 to 2023, contradicting Hypothesis 1. This may be due to the offsetting effect between the risk factors and protective factors for NSSI. Adolescents face global risk factors like COVID-19, climate crises, and scarce mental health resources (Lass-Hennemann et al., Reference Lass-Hennemann, Sopp, Ruf, Equit, Schäfer, Wirth and Michael2024), as well as individual-specific factors such as parent–child and peer relationships (Brown & Witt, Reference Brown and Witt2019). These factors may elevate the severity of NSSI in adolescents.
However, in recent years, there has been increasing global attention on adolescent mental health, with rising expenditures on related services (Brazel et al., Reference Brazel, Allison, Bastiampillai, Kisely and Looi2023; Fullman et al., Reference Fullman, Yearwood, Abay, Abbafati, Abd-Allah, Abdela, Abdelalim, Abebe, Abebo, Aboyans, Abraha, Abreu, Abu-Raddad, Adane, Adedoyin, Adetokunboh, Adhikari, Afarideh, Afshin and Lozano2018), a trend that is beneficial for improving the mental health of adolescents (Zaneva et al., Reference Zaneva, Guzman-Holst, Reeves and Bowes2022). Adolescents can access mental health services through multiple channels, including schools, hospital, and juvenile justice agencies (Duong et al., Reference Duong, Bruns, Lee, Cox, Coifman, Mayworm and Lyon2021). Countries have also established 24-hour public psychological hotlines or online chat platforms, such as the 988 Suicide & Crisis Lifeline in the United States, New Zealand’s 1737 Need to Talk, and Sweden’s Mind. Notably, during the COVID-19, the workload and utilization of mental crisis hotlines increased significantly, partially meeting mental health needs (Arendt et al., Reference Arendt, Markiewitz, Mestas and Scherr2020; Zhou et al., Reference Zhou, Snoswell, Harding, Bambling, Edirippulige, Bai and Smith2020). All these factors collectively buffer the adolescent NSSI process.
Differences in NSSI among adolescents over time by continent
When examining temporal trends in adolescent NSSI severity, no significant differences were observed across continents, which may suggest that adolescent NSSI constitutes a serious global public health problem. However, this does not imply that other characteristics of adolescent NSSI are uniform across continents. Previous cross-cultural studies have demonstrated that method of NSSI exhibits cultural variation (Gholamrezaei et al., Reference Gholamrezaei, De Stefano and Heath2017). For instance, compared with Belgium, Indian samples reported more head banging and less cutting and scratching (Gandhi et al., Reference Gandhi, Luyckx, Adhikari, Parmar, Desousa, Shah, Maitra and Claes2021). Furthermore, the relatively low number of studies from North America and Oceania may influence the effect size of the present findings and warrants caution when generalizing these results.
Individual characteristics influencing NSSI changes among adolescents
This study indicates that clinical samples adolescents exhibit higher levels of self-injury severity compared to non-clinical samples. This heightened NSSI severity may be related to the co-occurring psychiatric disorders and greater symptom severity of clinical samples. In the included studies, clinical samples frequently presented with one or more additional psychiatric disorders (Morthorst et al., Reference Morthorst, Olsen, Jakobsen, Lindschou, Gluud, Heinrichsen, Møhl, Rubæk, Ojala, Hellner, Bjureberg and Pagsberg2022; Ojala et al., Reference Ojala, Hesser, Gratz, Tull, Hedman-Lagerlöf, Sahlin, Ljótsson, Hellner and Bjureberg2024; Viana et al., Reference Viana, Woodward, Raines, Hanna and Zvolensky2018).
The results indicated no significant gender effect on adolescent NSSI, contradicting Hypothesis 2. However, this finding aligns with previous research (Gratz et al., Reference Gratz, Latzman, Young, Heiden, Damon, Hight and Tull2012; Taliaferro et al., Reference Taliaferro, Heerde, Bailey, Toumbourou and McMorris2023; Zhao et al., Reference Zhao, Zhao, Zhou and Liu2024). While consensus remains elusive regarding gender’s impact on NSSI frequency and incidence, more uniform evidence suggests gender influences NSSI methods (Andrei et al., Reference Andrei, Efrim-Budisteanu, Mihailescu, Buică, Moise and Rad2024; Calvete et al., Reference Calvete, Orue, Aizpuru and Brotherton2015; Lundh et al., Reference Lundh, Karim and Quilisch2007; Moloney et al., Reference Moloney, Amini, Sinyor, Schaffer, Lanctôt and Mitchell2025). Similarly, age showed no significant association with adolescent NSSI. This may suggest a possible complex nonlinear relationship between age and NSSI. Wilkinson et al. (Reference Wilkinson, Qiu, Jesmont, Neufeld, Kaur, Jones and Goodyer2022) found a complex interaction among age, gender, and NSSI, with general distress serving as a partial mediator.
Relationship of social Indicators to NSSI among adolescents
This study examined the associations between three types of societal indicators and adolescent NSSI, thereby expanding understanding of its social etiology. This study revealed that higher levels of social development are associated with increased likelihood of NSSI behaviors among adolescents, a finding that contradicts the hypothesis. Two complementary mechanisms may account for this paradox. Specifically, in societies with advanced economic and overall development, individuals typically undergo more years of schooling. This extended education exposes adolescents to heightened academic stress Sathish & Subramanian, Reference Sathish and Subramanian2021. The fast-paced lifestyle and overwhelming academic workload may lead to sleep deprivation in adolescents, further worsening their emotion regulation capabilities (Mazza et al., Reference Mazza, Royant-Parola, Schroder and Rey2024). The widespread use of the Internet exacerbates this risk, a result consistent with prior studies (Mészáros et al., Reference Mészáros, Győri, Horváth, Szentivanyi and Balazs2022; Yang et al., Reference Yang, Xin, Liu and Böke2020). When using social media, adolescents may encounter NSSI - related content, which may trigger imitation. Moreover, within certain online subcultural communities, NSSI can be symbolized as a marker of identity, potentially prompting adolescents to engage in such behaviors to seek peer recognition and group belonging (Brown et al., Reference Brown, Fischer, Goldwich and Plener2020). On the other hand, these regions typically establish stronger mental-health infrastructure, train more school counsellors, provide more accessible psychological service resources (Ndetei et al., Reference Ndetei, Mutiso and Osborn2023; Wolthusen & Andrä, Reference Wolthusen and Andrä2023). Consequently, it is both contributed to enhanced early identification and reporting capacity for NSSI.
Interestingly, lower social equity (reflected by a higher Gini coefficient) was associated with reduced adolescent NSSI, contradicting the hypothesis. As Adams’ (Reference Adams1963) equity theory, macro-level equity does not inherently ensure micro-level equity. That is to say, even when the Gini coefficient is low, adolescents may still experience a sense of relative deprivation due to micro-level differences such as parental companionship, and adverse childhood experiences (Tian et al., Reference Tian, Zheng, Tong and He2023; Wang & Meng, Reference Wang and Meng2024), which in turn increases the risk of NSSI.
Similarly, social well-being presents a positive correlation with adolescent NSSI, conflicting with the hypothesis. Unlike prior research that predominantly examined happiness at the individual level (Diener & Seligman, Reference Diener and Seligman2002; Van Lente et al., Reference Van Lente, Barry, Molcho, Morgan, Watson, Harrington and McGee2012), the present study utilized a national-level happiness index. This macro-level indicator may capture broader societal conditions, which could help explain the counterintuitive association. From the perspective of mental health destigmatization, countries with elevated well-being indices may prioritize mental health, fostering social openness and acceptance of psychological issues. Adolescents in these countries might encounter reduced mental health stigma (World Happiness Report, n.d.), prompting more honest self-reports of negative emotions and enhancing help-seeking behaviors (Crockett et al., Reference Crockett, Núñez, Martínez, Borghero, Campos, Langer, Carrasco and Martínez2025). Moreover, this situation could also stem from statistical bias.
Significance and limitations
This study offers significant implications for comprehending and addressing adolescent NSSI. First, this study pioneeringly applied cross-temporal meta-analysis to reveal the temporal trends in NSSI from the perspective of frequency. Unlike previous studies that only focused on NSSI prevalence as a categorical variable, the cross-temporal meta-analysis of NSSI frequency as a continuous variable can more finely reveal its changing trends, deepening the understanding of adolescent NSSI. Secondly, the study examined not only individual characteristics but also emphasized the relationship between macro - social indicators and NSSI. This offers a new perspective on understanding adolescent NSSI and provides a scientific basis for devising targeted social policies and intervention measures to prevent and reduce such behaviors in adolescents. Specifically, it suggests developing differentiated intervention programs tailored to different regions and social factors, and providing more intensive interventions for the clinical samples.
Although enhancing our comprehension of adolescent NSSI trends, this study has certain limitations. Firstly, only literature utilizing the DSHI scale was analyzed. Future research could incorporate other tools like the ISAS, OSI, and FASM for comparison. Secondly, the study focused solely on NSSI frequency trends, leaving other characteristics unexplored. Future research could examine changes in NSSI methods and the emerging phenomenon of digital self-harm (Patchin et al., Reference Patchin, Hinduja and Meldrum2023). Third, the language inclusion criteria (Chinese and English only) and the geographic origins of the samples may limit the generalizability of our results. Specifically, the Asian population was overrepresented, while data from Europe, the Americas, and Oceania were limited. Studies from Africa were absent. Future studies should explore cross-cultural differences through multinational, multicenter follow-up research to clarify how sociocultural factors impact adolescent NSSI.
General Conclusions
This cross-temporal meta-analysis examined trends in adolescent NSSI and its influencing factors. The results indicated: (1) No significant change occurred in the severity of adolescent NSSI from 2007 to 2023 (d = 0.35). (2) No significant differences were found in NSSI trends across Asia, Europe, and the Americas. (3) NSSI severity changes were more pronounced in clinical samples than in non-clinical samples. (4) Social indicators showed a close relationship with adolescent NSSI. Specifically, social development (GDP per capita, HDI, and Internet penetration) and social well-being (happiness index) were significantly positively correlated with adolescent NSSI, whereas lower social equity (higher Gini coefficient) was associated with reduced adolescent NSSI.
Supplementary material
The supplementary material for this article can be http://doi.org/10.1017/S0033291725102134.
Funding statement
Our work was supported by the National Key Research and Development Project of China (grant no. 2024YFC3308400).
Competing interests
None declared.
Open access
