2.1. Developing the digital self-harm inventory(DSI-8) and the procedure for item generation
Previous studies have shown that a large number of adolescents with traditional self-harm are more likely to have digital self-harm[7].Therefore, from May 2022 to September 2022, we interviewed patients with digital self-harm in two psychiatric clinics in Changchun, Jilin, China. The interviewers were 3 psychiatrists and 3 psychological researchers. According to the definition of online NSSI in previous studies, we set the interview content as: "What would you post online that would cause harm to you?""What would you like to express when you post that?""Are there any other people around you who post something that hurts you? What do they post?" In the end, a total of 215 (69 males, 146 females) questionnaire samples were collected,, among whome 38 are from single-parent families, 177 non-single-parent families, 130 are the only child, and 85 non-only children, with an average age of 15.29 ± 2.23 years (range,12–19 years).
After collecting the interview data, three psychological researchers in our team coded and classified the survey content, deleting words that did not conform to the research theme. After coding, areas where coding categories were inconsistent or meaning unit attribution was unclear were further discussed to reach consensus. Finally, a consistent two-factor expected structure of adolescent data NSSI was formed, namely self-image harm and self-emotional harm. Self-image harm refers to publishing, sending or other ways to share content harmful to self-image online, and self-emotional harm refers to publishing, sending or other ways to express negative emotions about oneself online.
After analyzing the structure, three professors and three psychological researchers in my team developed questions based on the coding results. In the process of compiling questions, the two-item digital injury Survey and the Ottawa self-injury inventory (OSI) were also referred to[7, 13].
Finally, the words with ambiguity, repeated expression and vague statement were deleted and revised, and 17 items were retained. The score was scored with Likert scale 5 points (1 never, 2 rarely, 3 occasionally, 4 often, and 5 always). The sum of the scores of each item was the total score.
2.2 Study design and participants
The inclusion criteria were as follows: aged 12–24, no mental disorder, no disability, no cognitive impairment, and able to fully understand the meaning of the items in the scale. Voluntary participation. Sample size should be at least 300[14]. Three surveys were conducted for this purpose.
Sample 1 was mainly used for project analysis and exploratory factor analysis. In October 2022,we fully communicated with 3 middle schools in Changchun City, Jilin Province, and randomly selected 2 or 3 classes in each grade, taking all middle school students in the class as the survey objects. A Chinese online survey platform named Questionnaire Star was used to collect questionnaires. Invalid questionnaires, such as consistent responses, inconsistent responses, and too short response time, were excluded. A total of 1651 valid questionnaires were collected. The respondents had been informed of the purpose of the research, the principle of confidentiality, and the use of data before conducting the research, and informed consent was obtained from the respondents.
Sample 2 was mainly used for confirmatory factor analysis, aggregate validity, discriminant validity, internal consistency reliability, and split half reliability testingand it was conducted in December 2022 using the same sampling method as sample 1 (the survey subjects were different ). A total of 843 valid questionnaires were collected.
Sample 3 is mainly used for retesting reliability.In sample 2, the subjects were tested again after 3-month-according to the voluntary principle. Finally, 200 valid questionnaires were obtained. Sample characteristics are displayed in Table 1.
For this study, all students will be asked to complete items in an app called “Wenjuanxing”, which can be used on any mobile device.Before completing all projects, we need to obtain consent from their parents or their legal guardians.The statement is as follows “I have been informed of the background, purpose, procedures, risks, and possible benefits of this study,I have discussed this research project with my child, my child and me agree to participate in this study and may request to withdraw from this study at any time and for any reason.”This study can be started when the option: “I am aware of this and agree to the study.”was selected. The study was terminated when the option:“I have been informed of this and do not consent to the study” was selected.Our study have written informed consent was obtained from a parent and/or legal guardian
Table 1
Sociodemographic characteristics of the participants
Variable | Sample1(N = 1651) | Sample 2(N = 843) | Sample 3(N = 200) |
Gender | | | |
Males | 833 | 172 | 30 |
Females | 818 | 671 | 1275 |
Family status | | | |
single-parent families | 146 | 119 | 25 |
non-single-parent families | 1505 | 724 | 175 |
Whether the only children or not | | | |
only children | 1167 | 287 | 68 |
non-only children | 484 | 556 | 132 |
Age | 14.25 ± 1.48 | 14.83 ± 0.95 | 14.72 ± 0.93 |
2.3 Calibration Measures
Previous studies have analyzed the causes of digital self-harm according to the theoretical framework of general strain theory (GST)[15]. GST suggests that tension comes from the negative relationships we have with other people. The three negative relationships we have with other people are that they prevent us from achieving those important and worthwhile goals; Others deprive or threaten to deprive us of the positive stimuli we already have; We can divide stress into three types: the inability to achieve a desired goal, the removal of a positive stimulus, and the appearance of a negative stimulus. Therefore, individuals intend to achieve their goals, but also strive to avoid those painful and aversive situations in the pursuit of the goal. In adolescents' lives, these negative stimuli may include stressful events of negative relationships with parents, teachers and classmates, such as abuse and neglect, insult by teachers, and bullying by classmates, etc. When experiencing these stressful events, adolescents will naturally have negative emotions such as uneasiness, anxiety, and depression. When the negative impact of stressful events exceeds adolescents' ability to cope with or the negative emotions generated cannot be disposed of and transferred, they may adopt direct or indirect aggressive methods to resist the influence of negative stimuli and negative emotions[16].
The tension caused by stressful events (such as bullying) can cause people to have negative emotions such as anxiety and depression. As an important mediating variable between stress and deviant behavior, these negative emotions increase the individual's sense of harm, resulting in digital self-harm behavior and even self-injury[8, 17], accompanied by other physiological abnormalities (e.g. Decreased duration of sleep)[18].In conclusion, this study took the following issues into account: 1)adolescents may experience cyber bullying and physical bullying in real-life stressful events;2) they may experience negative emotions such as depression and anxiety, 3) they may have problem behaviors such as NSSI; and4) they may have problems with sleep duration. At the same time, given the high correlation between NSSI and borderline personality disorder, this study also included borderline personality disorder as a criterion[19].
The socio-demographic characteristics investigated in this study and the basic information of the relevant calibration tools are as follows:
Socio-demographic information: gender, age, single parent family, only child or not.
Revised Cyber Bullying Inventory (RCBI): A total of 14 items, using a 4-point scale, according to their own situation from low to high. On each item, participants rated how often they had engaged in or experienced the behavior in the past 6 months on a 4-point scale, with 1 indicating "never engaged in" or "never encountered" and 2 indicating "once", 3 means "2–3 times" and 4 means "more than 3 times"[20]. The Cronbachs'a coefficient of online bullying in this study was 0.958, which had good reliability.
Traditional Bullying Questionnaire: This scale was developed by Topcu in 2008 to measure traditional bullying and traditional bullying. In this paper, only the questionnaire items of the scale on traditional bullying are used, with a total of 5 items. A 3-point Likert scale is used, ranging from 1 (never) to 3 (often). Higher scores indicated more traditional bullying behaviors. The internal consistency reliability coefficient of the traditional bullying scale in this study was 0.672[21].
21-Item Depression Anxiety and Stress Scale (DASS-21) : This scale contains three parts: Depression, anxiety and stress, each with 7 items, were scored by Likert-4 scale, with 0 indicating not consistent,1 indicating sometimes consistent, 2 indicating often consistent, and 3 indicating always consistent. The higher the total score, the higher the degree of depression-anxiety-stress. In this study, the internal consistency coefficient of the scale was 0.912, the internal consistency coefficient of the depression subscale was 0.823, the internal consistency coefficient of anxiety was 0.754, and the internal consistency coefficient of stress was 0.796[22].
Ottawa self-injury inventory (OSI):The 28-part scale was used to assess the thoughts of non-suicidal self-injury, the frequency of actual self-injury, the mode and location of self-injury, the effect of self-injury on releasing negative emotions, and the way of resisting self-injury. The questionnaire adopts the first part of the scale, that is, people who have committed self-harm but do not want to commit suicide. 0 means never, 1 means at least once, 2 once a week, and 3 every day. The score is scored on a dichotomous scale, with a choice of 0 indicating no self-harm behavior and a choice of 1–3 indicating the presence of self-harm behavior[23].
Sleep duration: Based on previous studies, participants were surveyed on their sleep duration using a single question, with options including, 4 hours or less, 5 hours, 6 hours, 7 hours, 8 hours, 9 hours, and 10 hours or more[24]
The McLean Screening Instrument for Borderline Personality Disorder (MSI-BPD): This scale has 10 items, each answered "yes" and "no", Each item was scored as 1 and 0, respectively. The scale contains four dimensions: emotional disturbance, cognitive system disorder, impulsive behavior disorder and interpersonal disorder. The higher the score indicates the more severe borderline personality tendencies. The internal consistency coefficient of the scale in this study was 0.725[25].
2.4 Statistical analysis
Item analysis and exploratory factor analysis were performed in sample 1. Among them, the items with correlation coefficient less than 0.4 in item analysis was deleted, and the Keiser-Meyer-Olkin (KMO) and Bartlett sphericity test χ2 were was used to explain the suitability of the scale as an exploratory factor analysis. KMO (greater than 0.7) and P value of Bartlett sphericity test χ (less than 0.05) indicated that the items of the scale were2 suitable for exploratory factor analysis[26].Confirmatory factor analysis, convergent validity, discriminant validity, internal consistency reliability and split-half reliability were tested in sample 2. The indicators in confirmatory factor analysis included: chi-2square/degree of freedom ratio (χ2 /df), comparative fit index (CFI), Tucker-Lewis index (TLI), standardized root mean (standardized root mean) were used in confirmatory factor analysis square residual (SRMR), root mean square root of approximation (RMSEA), wherein, χ2 /df < 5, CFI > 0.9, TLI > 0.9, SRMR should be close to zero;, RMSEA < 0.08[27–29]. The evaluation criteria of convergent validity and discriminant validity were as follows: the average extracted variance value (AVE) of each factor was greater than 0.5, and the combined reliability was greater than 0.7 as the evaluation criteria of convergent validity. The discriminant validity required that the square root value of AVE corresponding to each factor was greater than the correlation coefficient between the factor and other factors. The Cronbach's α coefficient was used to test the internal consistency reliability, and the α coefficient should be greater than 0.8. Split-half reliability was calculated using the Spearman-Brown correction formula, and the split-half reliability was at least 0.8. Test-retest reliability was tested in sample 3, and the test-retest reliability was required to be at least above 0.7.
The item analysis, exploratory factor analysis, convergent validity, discriminant validity, internal consistency reliability, split-half reliability and test-retest reliability were tested by SPSS 24.0, and the confirmatory factor analysis was tested by Amos 21.0.