Journal of Affective Disorders 295 (2021) 63–71

Contents lists available at ScienceDirect

Journal of Affective Disorders

journal homepage: www.elsevier.com/locate/jad

Research paper

Different profiles with multiple risk factors of nonsuicidal self-injury and

their transitions during adolescence: A person-centered analysis
Qianqian Gao a, Jiaqi Guo a, Huijiao Wu a, Jialing Huang a, Nini Wu b, Jianing You a, *
a
Center for Studies of Psychological Application, Guangdong Key Laboratory of Mental Health and Cognitive Science, Key Laboratory of Brain, Cognition and Education
Sciences (South China Normal University), Ministry of Education, & School of Psychology, South China Normal University, Guangzhou, 510631, PR China
b
Department of Psychology, School of Education, Guangdong University of Education, No.351 XinGangZhong Road, 510303 Guangzhou, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: Background: Nonsuicidal self-injury (NSSI) is a serious public health concern among adolescents. However, ad­
Nonsuicidal self-injury olescents with different combinations of NSSI risk factors may constitute different subgroups that vary in sub­
Risk factors sequent risks of NSSI engagement, and the membership in these subgroups may change during early adolescence.
Profiles and transitions
Gender differences related to risk profiles and transition patterns may also exist. This study thus aimed to address
Gender differences
these three questions.
Person-centered approaches
Chinese early adolescents Methods: A total of 2,942 participants ranging from 12 to 17 years old (55.2% females; Mage = 14.60, SD = 1.21)
completed a series of questionnaires about NSSI and its risk factors on three occasions, every six months. Latent
profile and latent transition analyses were used to examine NSSI risk profiles (subgroups), their transition pat­
terns, and gender differences.
Results: Three distinct risk profiles were identified: Low-risk profile, Medium-risk profile, and High-risk profile.
The profile membership was fairly stable over one year, and the shift across profiles represented a “resilient”
pattern. Regarding gender differences, boys were less likely to be in either the Medium- or the High-risk groups,
but more likely to transition from the Medium- to the Low-risk group, relative to girls.
Limitations: This study relied upon self-report measures. Additional research is needed to investigate the
generalizability of the results to other populations, and to examine more predictors of class membership and
transitions.
Conclusions: The results highlight the importance of considering subgroup differences in understanding NSSI
risks. Differentiated interventions with multi-pronged approaches tailoring to different risk classes and gender
groups may therefore prove more efficient and effective.

1. Introduction 2018). In light of these challenges, a better understanding of NSSI risk

factors is of great importance to inform prevention and treatment of such
Nonsuicidal self-injury (NSSI), defined as the deliberate, direct, and behaviors.
culturally sanctioned destruction of body tissue in the absence of sui­ Previous studies have identified independent risk factors of NSSI
cidal intent (Nock, 2010), has become a widespread public health (Brown and Plener, 2017). These factors include distal, general predis­
problem. It can not only incur personal, economic, and social costs, but posing, and specific vulnerability factors. Yet, little is known about how
also increase the likelihood of future suicide attempts (You and Lin, all of those risks work together to impact NSSI and how their effects
2015). Notably, adolescence is a high-risk period for NSSI engagement might vary across different subpopulations. Furthermore, given the
(Dahl et al., 2018), and the lifetime prevalence of NSSI among com­ occurrence of a broad range of psychosocial, biological, and environ­
munity adolescents is 17.2% worldwide (Swannell et al., 2014). In mental changes in early adolescence, this period may be characterized
China, NSSI is also a prevalent phenomenon among adolescents, with by significant fluctuations in psychopathological symptoms (Sontag
the 12-month incidence rates ranging from 17.0% to 29.2% (Tang et al., et al., 2011). Thus, it is important to examine whether the profiles or

* Corresponding author at: School of Psychology, South China Normal University, Guangzhou, 510631, PR China.
E-mail addresses: [email protected] (Q. Gao), [email protected] (J. Guo), [email protected] (H. Wu), [email protected] (J. Huang), nini713@126.
com (N. Wu), [email protected] (J. You).

https://doi.org/10.1016/j.jad.2021.08.004
Received 4 June 2021; Received in revised form 2 August 2021; Accepted 3 August 2021
Available online 8 August 2021
0165-0327/© 2021 Elsevier B.V. All rights reserved.
Q. Gao et al. Journal of Affective Disorders 295 (2021) 63–71

subgroups observed vary during this critical developmental period. The framework of self-injurious behavior, Yates (2004) proposed that
aim of this study, therefore, is to explore the potentially distinct risk adverse childhood experience (e.g., parental hostility) undermines
profiles contributing to NSSI and the transitions of the profiles over time positive adaption at emotional, attitudinal, and relational levels of
among Chinese early adolescents. While both girls and boys are exposed competence. Specifically, an invalidating environment may deprive the
to the same risk factors, NSSI appears to be a more severe behavior in child’s opportunity of learning adaptive emotion regulation methods
girls characterized by higher engagement in NSSI and higher longitu­ and render her/him internalize a negative self-image. Moreover, an
dinal stability. Thus, this study will also investigate how gender may be invalidating environment may also compromise the child’s relational
related to the risk profiles and their transitions. competence by distorting the sense of self–other boundaries. Thus, it
seems necessary to investigate the combined effect of these multiple risk
1.1. The multiple risk factors of NSSI factors in understanding adolescent NSSI.
To data, only a very limited number of studies considered the
According to the integrated theoretical model proposed by Nock simultaneous effects of multiple risk factors on NSSI, and they usually
(2010), NSSI is caused by the complex interplay of multiple risk factors. used the cumulative risk index approach (Evans et al., 2013). For
Distal risks, general predisposing risks, and specific vulnerability risks instance, Bedi et al. (2014) calculated a cumulative risk index by sum­
jointly affect the onset and continuity of NSSI. Distal risk factors include ming up the scores of six internal risk factors, and this index predicted
familial hostility and genetic predisposition for high emotional reac­ the frequency of deliberate self-harm. Such studies, however, still suf­
tivity. The link between the invalidating relationships with parents in fered the limitation of relying primarily on the variable-centered ap­
childhood and later self-injury has been extensively supported (Bureau proaches, assuming homogeneity in adolescents (Howard and Hoffman,
et al., 2010; Martin et al., 2011). For example, adolescents with a history 2018). But in fact, adolescents with multiple risk factors are heteroge­
of NSSI reported perceiving their parents as being more intrusive and neous (Lanza et al., 2010).
expressing more criticism compared to those without NSSI (Ammerman In such a case, scholars have called for studies using the person-
and Brown, 2018). Beyond the familial risks, emotional reactivity is also centered approach to examine the multiple risk profiles of psychopa­
a strong distal predictor of NSSI in adolescence (Dawkins et al., 2019). thology (Bergman and Lundh, 2015; Bogat et al., 2016). A commonly
General predisposing factors for NSSI comprise both intrapersonal applied person-centered approach, the latent profile analysis (LPA),
and interpersonal factors, just mapping onto the intrapersonal and presents an opportunity to address some limitations inherent to the
interpersonal functions of this behavior. On the one hand, previous work variable-centered approach. LPA is designed to identify mutually
has demonstrated that intrapersonal functions, mainly referring to the exclusive latent profiles derived from a set of observed continuous in­
emotion-regulation functions, are endorsed by the majority of adoles­ dicators (Peugh and Fan, 2013). Using this method, Li et al. (2017)
cents with NSSI (Klonsky et al., 2015). Negative emotions have been identified four distinct risk profiles (i.e., Low Risk, Medium Risk, High
identified as particularly salient triggers of NSSI (Boyes et al., 2020). So Risk, and Peer Risk) based on 14 risk factors (e.g., low parental moni­
does the ability to tolerate the distress arising from the negative toring) across family, school, peer, and individual domains, and found
emotional experiences (Slabbert et al., 2018). On the other hand, that these risk profiles predicted differential individual propensities of
interpersonal functions, such as escaping from social situations and internet addiction. Similarly, Newcomb-Anjo et al. (2017) identified
eliciting attention, are also endorsed by a substantial portion of ado­ four patterns of risk (i.e., Low Risk, Low Social Support Risk, Financial
lescents (Turner et al., 2016). Thus, relationship disturbance is also Risk, and Multiple Risk) for compromised wellbeing. Several studies of
regarded a precursor to NSSI. Substantial studies have shown that NSSI have also used this approach (Case et al., 2020; Somer et al., 2015),
relationship disturbance is negatively associated with NSSI both but they derived the profiles from the manifestations of NSSI itself, and
concurrently and longitudinally (Brickman et al., 2014; You et al., thus are not helpful for the prediction of at-risk adolescents. To address
2012). this significant gap, this study employed the LPA to identify the distinct
With regard to the NSSI-specific vulnerability factors, self-criticism subgroups of adolescents with multiple NSSI risk factors, and to examine
and impulsivity, stemming from the self-punishment and pragmatic how these identified risk profiles related to actual NSSI engagement.
hypotheses respectively, are especially important. Self-critical cogni­ In addition, we would like to explore whether the risk profiles vary
tions are concomitantly and prospectively associated with NSSI (Burke over time. Early adolescence is an important transition period when
et al., 2021; Xavier et al., 2016), and self-punishment was among the adolescents experience frequent psychosocial, biological, and environ­
primary reasons those who had NSSI gave for engaging in this behavior mental changes. All these changes may contribute to changes in mental
(Nock and Prinstein, 2004). Besides, NSSI has also long been considered health outcomes (Blum et al., 2014). The previous studies based on both
an impulse control problem. Individuals who had NSSI are more likely to negative indicators (e.g., emotion distress) and positive indicators (e.g.
engage in other impulsive behaviors, such as aggressive behavior (Claes life satisfaction) had identified four mental health subgroups in ado­
et al., 2015). A meta-analysis has revealed that individuals who engaged lescents: Flourishing (i.e., low psychopathology and high well-being),
in NSSI self-reported greater impulsivity than those who did not, and Vulnerable (i.e., low psychopathology and low well-being), Symptom­
these differences were most pronounced for negative urgency, a facet of atic but content (i.e., high psychopathology and high well-being), and
impulsivity (Hamza et al., 2015). Troubled (i.e., high psychopathology and low well-being) groups
Taken together, these studies indicate the key role distal risks (i.e., (Moore et al., 2019; Zhou et al., 2020). These studies examined the
parental hostility and emotional reactivity), general predisposing risks stability and transitions of the mental health groups and suggested that
(i.e., negative emotions, distress intolerance, and relationship distur­ patterns of instability were evident among all groups, with the Troubled
bance), and specific vulnerability risks (i.e., self-criticism and negative group exhibiting the lowest levels of stability over one year. Thus, we
urgency) play in NSSI development, and these risk factors would be would use the latent transition analysis – a longitudinal extension of LPA
included in the present study. –to examine the instability of membership in the identified risk profiles.

1.2. Profiles of NSSI risks and transitions related to the risk profiles 1.3. Gender differences in risk profiles and transitions

While ample research have established the relationships of these risk Gender differences exist in relation to NSSI. In general, girls are more
factors to the engagement in NSSI, the majority of them have only likely to engage in NSSI than boys (Sornberger et al., 2012). The
focused on the effects of a single risk factor or at the most two-factor engagement in NSSI is also more stable in girls than in boys (Wang et al.,
interactions. However, these multiple risks may be interrelated and 2017). These gender differences may be accounted for by two reasons.
work together in NSSI. In the developmental psychopathology One explanation may involve the hormone differences between genders.

64
Q. Gao et al. Journal of Affective Disorders 295 (2021) 63–71

For example, estradiol increases at pubertal onset appears to increase following behaviors to deliberately harm yourself but without suicidal
girls’ vulnerability to the development of internalizing problems (Balzer intent?” At T2 and T3, the period for assessing NSSI was the past 6
et al., 2015; Rutter et al., 2003). Another explanation may involve the months. All seven items were rated on a 4-point scale ranging from 1 =
gender differences in emotion regulation. Specifically, girls tend to use never to 4 = six times or more. To facilitate interpretation, this variable
more self-focused thinking (e.g., rumination) and more passive re­ was re-coded into a dichotomous variable that classified participants as
sponses to their emotions (e.g., NSSI) than boys (Nolen-Hoeksema, having engaged in NSSI or not. The Cronbach’s alphas were 0.72 and
2012). However, little is known about how the subgroups based on 0.75 for the T2 and T3 data, respectively.
multiple risks and the membership transitions across groups vary be­
tween girls and boys. Thus, this study will investigate the gender dif­ 2.2.2. Parental hostility
ferences related to the risk profiles and transitions of the profiles. The 10-item Hostile/Coercive subscale of Parent Behavior Inventory
(PBI; Lovejoy et al., 1999) was used to assess perceived parental hostility
1.4. The current study (e.g., “My parents spank or use physical punishment with me”). Re­
sponses were made on a 6-point scale ranging from 1 = never to 6 =
This study has three aims. The first aim is to identify distinct risk always. The total score was calculated with higher scores indicating
profiles based on a combination of distal, general predisposing, and more perceived parental hostility. This scale has demonstrated good
specific vulnerability risks using the person-centered approach, and to reliability and validity among Chinese adolescents (You et al., 2017).
examine whether these risk profiles differentially relate to later NSSI. The Cronbach’s alpha value was 0.88 for all three waves.
We expect to identify at least two profiles: one characterized by rela­
tively low levels of all risk factors and the other characterized by rela­ 2.2.3. Emotional reactivity
tively high levels of all risk factors. We also hypothesize that adolescents Emotional reactivity was assessed by the 21-item Emotion Reactivity
within the high-risk profile report higher probability for NSSI engage­ Scale (ERS; Nock et al., 2008), which has demonstrated good reliability
ment than those in the low-risk profile. The second aim is to characterize and validity with Chinese adolescents (Liu et al., 2020). It has three
the development of the identified risk profiles. We hypothesize that subscales: sensitivity (10 items; e.g., “I tend to get emotional very
variability will exist in all identified risk profiles, and the high-risk easily”), intensity (7 items; e.g., “When I experience emotions, I feel
profile will be more unstable. The third aim is to explore the gender them strongly”), and persistence (4 items; e.g., “When I am upset, it
differences in the profiles and transitions. We hypothesize that the high- takes me much longer than most people to calm down”). Responses were
risk group will include a higher probability of girls, and that girls will be made on a 5-point Likert scale, ranging from 1 = not at all like me to 5 =
less likely than boys to transition from the high- to the low-risk group. completely like me. A total score was calculated with higher scores indi­
cating higher levels of emotional reactivity. The Cronbach’s alphas were
2. Method 0.96, 0.96 and 0.97 at T1, T2, and T3, respectively.

2.1. Participants 2.3.4. Negative emotions

The Chinese version of the short Depression Anxiety Stress Scale
Participants were recruited from eight secondary schools in Hong (DASS21; Taouk et al., 2001) was used to assess the three common
Kong and were surveyed semiannually for three times (for the full negative emotions: depression (7 items; e.g., “I couldn’t seem to expe­
recruitment details see You et al., 2016). All testing materials and the rience any positive feelings at all”), anxiety (7 items; e.g., “I find myself
procedures were approved by the ethical board of the Chinese University getting agitated”), and stress (7 items; e.g., “I find it difficult to relax”).
of Hong Kong and the participating schools’ authorities. At Time 1 (T1), Participants responded on a 4-point scale ranging from 0 = do not apply
5,423 students aged between 12 and 17 years (52.7% females; Mage = to me at all to 3 = applies to me very much or most of time. The sum of the
14.63 years, SD = 1.25) participated. Overall student participation rates 21 items was taken, with a higher score indicating higher levels of
were close to 99 % in all schools. We re-assessed the Time 1 sample after negative emotions. The DASS-21 has demonstrated high internal con­
a 6-month (Time 2) and a 12-month (Time 3) interval, respectively. At sistency and adequate test-retest reliability (Zhang et al., 2017). In this
Time 2 (T2), 6,911 adolescents (52.6% females) participated, and 3,999 study, the Cronbach’s alphas were 0.93, 0.94, and 0.94 for the T1, T2,
(55.4% females) of them were retained from the T1 sample. At Time 3 and T3 data, respectively.
(T3), 6,831 adolescents (52.6% females) were included. Overall, we
successfully followed 3,600 adolescents over the three waves. Attrition 2.3.5. Distress intolerance
was mainly due to students transferring to other schools or being absent Distress intolerance was measured by the 15-item Distress Tolerance
from school on the day of assessment. The Missing Completely at Scale (DTS; Simons and Gaher, 2005). The scale included four subscales:
Random test (Little, 1988) was conducted, and results (a normed χ 2/df tolerance (3 items; e.g., “Feeling distressed or upset is unbearable to
of 1.09) indicated that the missing of data was a random pattern (Bollen, me”), absorption (3 items; e.g., “My feelings of distress are so intense
1989). that they completely take over”), appraisal (6 items; e.g., “My feelings of
We removed 658 adolescents who had engaged in NSSI at T1, so that distress or being upset scare me”), and regulation (3 items; e.g., I’ll do
the prospective relationship between the risk profiles and the occurrence anything to avoid feeling distressed or upset”). Participants responded
of NSSI could be examined. Finally, 2,942 adolescents (55.2% females; on a 5-point Likert scale, ranging from 1 = strongly disagree to 5 =
Mage = 14.60 years, SD = 1.21) were included in the analyses. strongly agree. The sum of the 15 items was taken, with item 6 reverse
scored. This scale has proved to be suitable for Chinese adolescents
2.2. Measures (Kang et al., 2018). The Cronbach’s alphas were 0.91, 0.93 and 0.94 at
T1, T2, and T3, respectively.
2.2.1. Nonsuicidal self-injury (NSSI)
Seven NSSI behaviors, i.e., self-cutting, burning, biting, punching, 2.3.6. Relationship disturbance
scratching skin, inserting objects into the nail or skin, and banging the A 7-item scale extracted from the Revised Diagnostic Interview for
head or other parts of the body against the wall, were assessed in this Borderlines (DIB-R; Zanarini et al., 1989) was used to assess relationship
study. These behaviors were selected because they were found to be disturbance (e.g., “My relationship with other people are very unsta­
relatively common among adolescents (Nock, 2010) and displayed good ble”). Participants indicated their responses to each item on a 4-point
psychometric properties in previous studies (You and Lin, 2015). At T1, scale, ranging from 1 = strongly disagree to 4 = strongly agree. A total
participants were asked, “In the past 12 months, have you engaged in the score was calculated with higher scores indicating higher levels of

65
Q. Gao et al. Journal of Affective Disorders 295 (2021) 63–71

relationship disturbance. This scale has demonstrated good validity

23.35
among Chinese adolescents (Ren et al., 2018). In this study, its Cron­

6.33
21
bach’s alphas were 0.85, 0.84, and 0.86 at T1, T2, and T3, respectively.

–
29.88
10.29
0.48
2.3.7. Self-criticism

20

–
The self-critical thinking style was measured by the 9-item Self-
Criticism subscale of the Depressive Experiences Questionnaire (DEQ;

13.02
0.60
0.55

4.12
Blatt et al., 1976). A sample item is “I often find that I don’t live up to my

19

–
own standards or ideals”. Participants responded on a 7-point Likert

34.58
10.61
scale, ranging from 1 = not at all like me to 7 = like me very much. A total

0.59
0.60
0.58
18
score was calculated with item 9 reverse scored. Demonstrating good

–
reliability and validity, this scale has been widely used in studies with

0.54
0.48
0.61
0.46
8.50
9.03
Chinese adolescents (Gong et al., 2019). The Cronbach’s alphas were

17

–
0.86, 0.88, and 0.89 at T1, T2, and T3, respectively.

44.81
17.43
0.60
0.71
0.65
0.63
0.57
16
2.3.8. Negative urgency

–
The 12-item Negative Urgency subscale of the UPPS Impulsive

18.25
0.27
0.40
0.25
0.26
0.34
0.20

7.64
Behavior Scale (Whiteside and Lynam, 2001) was administrated to the

15

–
participants. A sample item is “I have trouble controlling my impulses”.
Responses were made on a 4-point scale ranging from 1 = strongly

23.06
0.20
0.46
0.38
0.45
0.45
0.42
0.53

6.45
disagree to 4 = strongly agree. The sum scores, with higher scores indi­

14

–
cating higher levels of negative urgency, were used in the analyses.

30.51
10.21
Previous studies have supported its reliability and validity within Chi­

0.54
0.25
0.48
0.46
0.45
0.44
0.63
0.37
13
nese adolescents (You et al., 2016). In this study, the Cronbach’s alphas

–
were 0.85, 0.89, and 0.94 at T1, T2, and T3, respectively.

12.17
0.56
0.58
0.25
0.46
0.40
0.40
0.54
0.42
0.39

4.05
12

–
2.3. Analytic strategy

35.12
11.53
0.59
0.58
0.56
0.25
0.48
0.44
0.52
0.42
0.44
0.38
11

Data analyses using Mplus version 7.2 (Muthén and Muthén, 2017)

–
were proceeded in three stages corresponding to the three research aims.

10.66
In the first stage, a series of latent profile models were run separately for

0.53
0.51
0.63
0.53
0.26
0.48
0.60
0.44
0.41
0.51
0.39

9.98
10

–
each wave’s data. In order to facilitate interpretation, all raw scores of
indicator variables were converted into Z scores. To determine the

38.04
15.84
0.69
0.57
0.57
0.58
0.57
0.27
0.59
0.54
0.46
0.46
0.47
0.42
optimal number of latent profiles, we considered both multiple com­
9

plementary fit indices and substantive interpretation of the profiles –

22.21
(Collins and Lanza, 2009). Evaluation criteria includes: a) Low Bayesian
0.37
0.37
0.31
0.31
0.36
0.27
0.58
0.24
0.26
0.22
0.25
0.29
0.18

8.35
Information Criterion (BIC) and Sample-Size Adjusted BIC (aBIC); b)
8

significant Lo-Mendell-Rubin Likelihood Ratio Test (LMR-LRT); c) high

24.85
0.14
0.35
0.32
0.35
0.34
0.31
0.47
0.12
0.38
0.27
0.36
0.34
0.31
0.47

5.60
Entropy (> .80); and d) appropriate class proportions (> 5.0%). After
7

identifying the best latent profile solution for each wave’s data, we used
the Block-Croon-Hagenaars (BCH) approach to examine the association

33.28
0.46
0.22
0.39
0.43
0.37
0.34
0.53
0.36
0.19
0.36
0.32
0.34
0.33
0.50
0.32

9.31
between profiles and the probabilities of adolescents engaging in NSSI at
6

T2 and T3 (Asparouhov and Muthén, 2014).

12.79
In the second stage, an LTA was conducted with the data of all three
0.48
0.46
0.21
0.43
0.40
0.39
0.51
0.39
0.42
0.17
0.42
0.34
0.35
0.47
0.36
0.37

4.15
waves combined, examining if and how membership of each group
5

changed across waves (Nylund-Gibson et al., 2014). The latent profile 37.52
10.39
0.43
0.47
0.46
0.17
0.35
0.32
0.50
0.34
0.34
0.34
0.15
0.36
0.30
0.41
0.31
0.29
0.28

variables (C) were regressed onto the same variables from the previous
4

time points (i.e., C3 on C2, C3 on C1, and C2 on C1), and the transition
Descriptive statistics for all indicator variables (N = 2,942).

probabilities were estimated to describe the degree of stability of NSSI

12.01
10.08
0.43
0.50
0.55
0.44
0.22
0.46
0.56
0.38
0.33
0.41
0.38
0.17
0.38
0.47
0.36
0.32
0.39
0.34

risk over time.

3

Note. All correlations are significant at p < .01 level.

In the third stage, we examined how gender was related to the latent
41.58
16.73

profiles at T1, T2, and T3. We also examined how gender was related to
0.65
0.44
0.46
0.50
0.50
0.22
0.56
0.46
0.40
0.38
0.38
0.42
0.18
0.49
0.41
0.38
0.37
0.37
0.36

the probabilities of transitioning between the latent profiles in T1 and

2

those in T2 and T3, and between the profile in T2 and that in T3. The
23.61
0.30
0.30
0.23
0.30
0.31
0.19
0.57
0.24
0.24
0.20
0.23
0.24
0.20
0.47
0.19
0.23
0.17
0.20
0.26
0.15

8.90

inclusion of gender in LCA and LTA models was conducted using logistic
1

regression analysis.
12. Relationship Disturbance (T2)

19. Relationship Disturbance (T3)

5. Relationship Disturbance (T1)

3. Results
11. Distress Intolerance (T2)

18. Distress Intolerance (T3)

16. Emotion Reactivity (T3)
10. Negative Emotions (T2)

17. Negative Emotions (T3)

4. Distress Intolerance (T1)
2. Emotion Reactivity (T1)

9. Emotion Reactivity (T2)

14. Negative Urgency (T2)

21. Negative Urgency (T3)

3. Negative Emotions (T1)

15. Parental Hostility (T3)

7. Negative Urgency (T1)
1. Parental Hostility (T1)

8. Parental Hostility (T2)

Descriptive statistics, including correlation matrix, means (M) and

13. Self-criticism (T2)

20. Self-criticism (T3)

6. Self-criticism (T1)

standard deviations (SD), for all indicator variables were presented in

Table 1. In the current sample, the proportion of adolescents reporting
having engaged in NSSI was 6.7% (n = 196) at T2 and 3.9% (n = 116) at
T3. Among the participants who have engaged in NSSI, the most prev­
Variable
Table 1

alent method was biting at T2 (n = 69, 2.3%) and cutting at T3 (n = 52,

SD
M

1.8%). In addition, 2.8 % (n = 83) of adolescents at T2 and 1.4% (n = 41)

66
Q. Gao et al. Journal of Affective Disorders 295 (2021) 63–71

of adolescents at T3 used more than one method to injure themselves. classes maintained a substantively similar pattern across the three
waves. Profile 1 reflected a Low-risk profile, consisting of individuals
with the lowest means on parental hostility, emotional reactivity,
3.1. Latent profile solution negative emotions, distress intolerance, relationship disturbance, self-
criticism, and negative urgency. This profile increased in size from T1
As shown in Table S1 (see Supplementary Material), we estimated to T2, and then decreased from T2 to T3 (35.8%, 41.8%, and 33.9% at
solutions including 2-5 profiles at each time point. Results indicated that T1, T2, and T3, respectively). In contrast, Profile 2 reflected a High-risk
the 3-class solution was the most appropriate. At T1, although the values profile, consisting of individuals with the highest means on all risk fac­
of BIC and aBIC decreased with each additional class, a relative leveling tors, especially on negative emotions. This profile included 15.4%,
off occurs after the three-profile model. Furthermore, the LMR-LRT p- 14.1%, and 17.0% of the participants at T1, T2, and T3, respectively.
values for the 4- and 5- class models were not significant (p > 0.01), and Profile 3 reflected a Medium-risk profile that displayed moderate levels of
the entropy of the 3-class model was greater than those of the 4- and 5- all risk factors. This profile was the largest across waves, comprising
class models. Similarly, at T2 and T3, the 3-class model was also the 48.8%, 44.1%, and 49.1% of the samples at T1, T2, and T3, respectively.
most appropriate.
Fig. 1 presents the estimated mean plots for the final 3-class solu­ 3.2. Risk profiles and NSSI
tions. Although varying slightly in indicator means over time, the three
We examined the associations between the identified risk profiles
and adolescent NSSI. As shown in Table 2, there were significant dif­
ferences in the probability of NSSI across the multiple risk profiles for
each time point. Post-hoc tests demonstrated that the High-risk groups
showed higher probability of NSSI than the Medium- and the Low-risk
groups; and the Medium-risk groups showed higher probability of
NSSI than the Low-risk groups.

3.3. Changes in profile membership over time

The three-wave first-order and second-order LTA was conducted to

investigate changes in adolescents’ profile membership. Table 3 presents
the three transition probability matrices generated by this procedure:
one for the T1 to T2 transition, one for the T1 to T3 transition, and one
for the T2 to T3 transition. As shown, from T1 to T2 and T1 to T3, the
Low-risk group showed the highest stability, followed by the Medium-
risk group and the High-risk group. From T2 to T3, the Medium-risk
group showed the highest stability, followed by the Low-risk group
and the High-risk group.
Of those who exhibited instability from T1 to T2 and T1 to T3, ad­
olescents initially fitting a Low-risk group all transitioned to the
Medium-risk group. Adolescents fitting a Medium-risk group were more
likely to transition to the Low-risk group than to the High-risk group.
Among adolescents who transitioned out of the High-risk group, most of
them transitioned into the Medium-risk group, followed by the Low-risk
group.
Regarding the instability from T2 to T3, similarly, adolescents in the
Low-risk group all transitioned to the Medium-risk group. Also, ado­
lescents in the High-risk group were more likely to transition to the
Medium-risk group than to the Low-risk group. However, if adolescents
transitioned out of the Medium-risk group, they were more likely to
transition to the High-risk than to the Low-risk group.

Table 2
Proportion of adolescents within each groups reporting NSSI at T2 and T3.
Latent status T2 T3
a
T1 Low-risk 0.02 0.02a
Medium-risk 0.07b 0.03a
High-risk 0.16c 0.10b
T2 Low-risk 0.02a 0.02a
Medium-risk 0.06b 0.04b
High-risk 0.22c 0.11c
T3 Low-risk – 0.01a
Medium-risk – 0.04b
High-risk – 0.11c
Fig. 1. Latent profiles of NSSI risk by wave. PH = perceived parental hostility;
ER= emotional reactivity; NE = negative emotions; DI = distress intolerance; Note. Values with different superscripts are significantly different at the p < 0.05
RD = relationship disturbance; SC = self-criticism; NU = negative urgency. level.

67
Q. Gao et al. Journal of Affective Disorders 295 (2021) 63–71

Table 3 scores on all indicators). The finding echoes those of prior studies
Unconditioned latent transition probability. showing that multiple psychosocial risk factors generally co-occur
T2 T3 (Buehler and Gerard, 2013; Copeland et al., 2009).
Low- Medium- High- Low- Medium- High-
risk risk risk risk risk risk 4.2. Risk profiles and NSSI
T1 Low-risk 0.81 0.19 0.00 0.68 0.32 0.00
Medium- 0.23 0.66 0.11 0.18 0.66 0.16
In order to validate the meaningfulness of the different risk profiles,
risk we examined how these risk profiles related to adolescents’ actual
High-risk 0.08 0.35 0.57 0.07 0.36 0.57 engagement in NSSI. In line with our hypothesis, adolescent NSSI can be
T2 Low-risk – – – 0.69 0.31 0.00 meaningfully predicted by their risk profile membership. Specifically,
Medium- 0.12 0.72 0.16
adolescents within the High-risk profile did report the highest proba­
– – –
risk
High-risk – – – 0.03 0.30 0.67 bility of NSSI at present or at a later time point. This finding provided
support for a cumulative risk perspective that high co-occurrence of risk
factors contributes to more negative outcomes relative to low occur­
3.4. Gender differences in profiles and transitions rence of risk factors (Evans et al., 2013). Notably, despite the fact that
adolescents within the Medium-risk profile did not display any obvious
Once the optimal class model had been identified for each time point, risk characteristic, they had relatively high probability of engaging in
this study examined whether the proportion of boys and girls differed NSSI. This finding is particular important because such groups might
across classes. As shown in Table 4, the logistic regression analysis have been overlooked in traditional problem-focused assessment sys­
revealed that boys were less likely to be in the Medium-risk group than tems in which individuals in these groups may be classified as “low risk”.
in the Low-risk group at T1, T2, or T3 (OR = 0.77, 0.65, and 0.71). Also, Taken together, these results highlight the usefulness of latent profile
boys were less likely than girls to be in the High-risk group at T2 or T3 analysis in understanding the patterns of risk and identifying adoles­
(OR = 0.78 and 0.69). cents who are most in need of targeted interventions (Lanza and Cooper,
In the final LTA model, gender was added as a covariate into the 2016).
logistic regression analysis. As shown in Table 5, gender differences only
occurred in the transition from T1 to T2 and T3. Boys in the Medium-risk 4.3. Transition patterns of risk profiles over time
profile at T1 were more likely than girls in the same profile to transition
to the low-risk profiles at T2 and T3 (OR = 2.41 and 1.70). Due to the dynamic nature of development during early adolescence,
it was expected that instability existed among all risk groups. Out of our
4. Discussion expectation, there was a high probability that adolescents remained in
the same groups over time. One explanation may be that some risk in­
Using the person-centered approaches with a longitudinal design dicators were relatively stable. For example, previous study had pro­
among Chinese early adolescents, the present study identified three posed that distress tolerance is a trait-like, temporally stable
heterogeneous groups (i.e., Low-risk group, Medium-risk group, and characteristic (Cummings et al., 2013). Another explanation may be that
High-risk group) of adolescents based on multiple risk factors of NSSI. one year is too short to capture the developmental changes. Additional
These groups had differential stabilities and unique transition patterns. research with much longer followed-up period is required to replicate
This study also found that genders were both differentially associated the results of the current study.
with profile membership and transitions among profiles. Partially in line with our hypothesis, however, significant transitions
among groups still occurred. Specifically, the High-risk group showed
the highest instability, with 33.0%-43.0% of adolescents transitioned
4.1. Multiple risk profiles out of this group. Such results were consistent with the previous studies
revealing the lowest over time stability of the Troubled group of mental
Overall, the latent profile analyses revealed the heterogeneity of health (characterized by high psychopathology and low well-being;
early adolescents based on multiple risks of NSSI. In particular, three Moore et al., 2019; Zhou et al., 2020). Notably, 30.0%-36.0% of ado­
distinct risk profiles were identified at each time point: Low-risk profile, lescents within the High-risk group transitioned into the Medium-risk
Medium-risk profile, and High-risk profile. Similar to those obtained group, and 3.0%-8.0% of them transitioned into the Low-risk group.
from a previous study (Li et al., 2017), this study found that the Moreover, 12.0%-23.0% of adolescents within the Medium-risk group
Medium-risk profile emerged as the most common profile. More transitioned into the Low-risk group. Such transition patterns may
importantly, results showed great consistency at levels of all risk factors represent a “resilient” developmental tendency (Clinton, 2008). At the
within the same profile (e.g., High-risk group had the highest mean same time, it should be cautioned that still a few adolescents
(11.0%-16.0%) in the Medium-risk groups transitioned into the
Table 4 High-risk group. Taken together, these findings offered continued sup­
Gender distribution within each risk profiles at T1, T2 and T3. port for the developmental heterogeneity of adolescents with NSSI risk.
a
Latent Status Number of Girls Number of Boys OR p

T1 Low-risk b
555 505 – –
4.4. Gender differences in risk profiles and transitions
Medium-risk 826 619 0.77 <0.001
High-risk 246 210 0.94 0.59 Regarding gender differences, girls were more likely to be in the
T2 Low-risk b 618 620 – – Medium- and High-risk groups and had higher stability. These results
Medium-risk 774 533 0.65 <0.001
provide further support for the finding that NSSI shows a marked female
High-risk 235 181 0.78 <0.05
T3 Low-risk b 500 511 – – preponderance (Bresin and Schoenleber, 2015). Several factors may
Medium-risk 838 612 0.71 <0.01 contribute to the gender differences. First, estrogen and progesterone
High-risk 289 211 0.69 <0.001 may delay girls’ ability to recover from stress (Zahn-Waxler et al., 2008).
Note. a Odd ratios reflect the gender differences on profiles distribution; OR < 1 Second, girls tend to experience more affect intensity and reactivity than
indicates boys have lower odds to be in the Medium- or High-risk groups than boys, because they are more aware of and attentive to their emotions
girls. (Nolen-Hoeksema and Aldao, 2011). Third, girls prefer to use more
b
Low-risk groups are served as the reference groups in logistic regression. internally focused responses to their emotions (Nolen-Hoeksema, 2012).

68
Q. Gao et al. Journal of Affective Disorders 295 (2021) 63–71

Table 5
Gender differences in the probabilities of transition between groups.
T2 T3

Low-risk Medium-risk High-risk Low-risk Medium-risk High-risk

T1 Low-risk – 1.11 N/A – 0.74 N/A

Medium-risk 2.41*** – 1.44 1.70** – 0.91
High-risk 1.76 1.34 – 1.57 1.16 –
T2 Low-risk – 0.80 N/A
Medium-risk 1.29 – 1.12
High-risk 1.87 1.32 –

Note. The values in the table are odd ratios (OR); OR > 1 indicates boys have higher odds of transitioning from the original group to that particular group than girls, and
OR <1 indicates boys have the lower odds of transitioning from the original group to that particular group than for girls.
The odds of remaining in the same group served as reference in logistic regression.
**p < 0.01, ***p < 0.001.
N/A = not available, because the proportion of adolescents who transitioned from Low- to High-risk group is too small to be analyzed.

Taken together, these biological, cognitive, and social-emotional char­ Thus, future researchers are advised to investigate more individual and
acteristics combined in adolescence leave girls more vulnerable than environmental characteristics that confer risk and resilience beyond
boys to NSSI. gender.

4.5. Implications CRediT authorship contribution statement

The results have important implications for interventions that aim to Qianqian Gao: Conceptualization, Formal analysis, Writing – orig­
reduce adolescent NSSI. First, the current study argues against the “one inal draft, Writing – review & editing. Jiaqi Guo: Data curation, Su­
size fits all” intervention approach, and encourages “this-approach-fits- pervision. Huijiao Wu: Data curation, Supervision. Jialing Huang:
this-profile” interventions by using information about subgroup-specific Data curation, Supervision. Nini Wu: Writing – review & editing.
effects (Wang and Peck, 2013). Specifically, the adolescents in the Jianing You: Data curation, Writing – review & editing.
High-risk group should receive remedial interventions immediately.
Although some high-risk adolescents are likely to transition out of their Declaration of Competing Interest
risk for NSSI, a significant proportion will not. Given that the highest
level of negative emotions are a major characteristic for the High-risk All authors declare that they have no conflict of interest.
group, dialectical behavior therapy (DBT) that includes
emotion-regulation skills training may be helpful (Fassbinder et al., Role of the funding source
2016). The current results also suggest that a multi-pronged approach
may be more effective than those targeting singular risk factors (Fox This research was supported by the National Natural Science Foun­
et al., 2019). In addition to emotion-regulation skills training, school dation of China (Grant No. 31771228), the National Social Science
professionals and parents should also consider teaching adolescents Foundation of China (Grant No. 19ZDA360), and the National education
support-seeking strategies and modifying their unreasonable cognitions sciences planning project (Grant No. CBA210235). The funders had no
(e.g., self-critical cognition). role in study design, data collection and analysis, decision to publish, or
Second, given that quite a few adolescents transitioned into High-risk preparation of the manuscript.
group over time, screening the NSSI risk only once at entry into school,
as most schools do, is not able to follow the dynamic changes in NSSI Acknowledgments
risk. It is of great necessity to periodically monitor the multiple risk
indicators and determine if these indicators have changed over time. This study was also supported by Research Center for Crisis Inter­
Third, despite that programs directed toward girls may be more efficient vention and Psychological Service of Guangdong Province, South China
in reaching larger groups of individuals who have engaged in NSSI Normal University and the base of psychological services and counseling
(Bresin and Schoenleber, 2015), boys may benefit more from interven­ for “Happiness” in Guangzhou.
tion programs than girls because they are more likely to transition out of
the Medium-risk group and into the Low-risk group. Supplementary materials

4.6. Limitations and directions for further research Supplementary material associated with this article can be found, in
the online version, at doi:10.1016/j.jad.2021.08.004.
The current study has some limitations, leaving opportunities for
future research. First, it should be acknowledged that all the measures in Reference
this study are self-reported and thus may be subject to respondent bias.
However, adolescents themselves may be the most reliable source when Ammerman, B.A., Brown, S., 2018. The Mediating Role of Self-Criticism in the
it comes to reporting on their own experiences and feelings (Wu et al., Relationship between Parental Expressed Emotion and NSSI. Curr. Psychol. 37,
325–333. https://doi.org/10.1007/s12144-016-9516-1.
2021). Second, the person-centered analyses depend to a larger degree Asparouhov, T., Muthén, B., 2014. Auxiliary Variables in Mixture Modeling: Three-Step
on specificities of the sample than variable-based approaches. Thus, Approaches Using Mplus. Struct. Equ. Model. 21, 329–341. https://doi.org/
future studies should replicate the present findings to determine if 10.1080/10705511.2014.915181.
Balzer, B.W.R., Duke, S.A., Hawke, C.I., Steinbeck, K.S., 2015. The effects of estradiol on
similar profiles would emerge in different cultural contexts using more
mood and behavior in human female adolescents: a systematic review. Eur. J.
diverse samples. Third, beyond gender, the survey did not investigate Pediatr. 174, 289–298. https://doi.org/10.1007/s00431-014-2475-3.
other determinants in the profiles and transitions. For example, what Bedi, R., Muller, R.T., Classen, C.C., 2014. Cumulative risk for deliberate self-harm
factors differentiate the adolescents who were in the High-risk group at among treatment-seeking women with histories of childhood abuse. Psychol.
Trauma Theory, Res. Pract. Policy. 6, 600–609. https://doi.org/10.1037/a0033897.
T1 but transitioned to the Low-risk group from those adolescents in the Bergman, L.R., Lundh, L.-G., 2015. Introduction: The Person-oriented approach: Roots
High-risk group at T1 and remained in the same group one year later? and roads to the future. J. Pers. Res. 1, 1–6. https://doi.org/10.17505/jpor.2015.01.

69
Q. Gao et al. Journal of Affective Disorders 295 (2021) 63–71

Blatt, S.J., D’Afflitti, J.P., Quinlan, D.M., 1976. Experiences of depression in normal Lanza, S.T., Rhoades, B.L., Nix, R.L., Greenberg, M.T., 2010. Modeling the interplay of
young adults. J. Abnorm. Psychol. 85, 383. https://doi.org/10.1037/0021- multilevel risk factors for future academic and behavior problems: A person-centered
843X.85.4.383. approach. Dev. Psychopathol. 22, 313–335. https://doi.org/10.1017/
Blum, R.W., Astone, N.M., Decker, M.R., Mouli, V.C., 2014. A conceptual framework for S0954579410000088.
early adolescence: a platform for research. Int. J. Adolesc. Med. Health. 26, 321–331. Li, D., Li, X., Zhao, L., Zhou, Y., Sun, W., Wang, Y., 2017. Linking multiple risk exposure
https://doi.org/10.1515/ijamh-2013-0327. profiles with adolescent Internet addiction: Insights from the person-centered
Bogat, G.A., von Eye, A., Bergman, L.R., 2016. Person-Oriented Approaches. Dev. Psy approach. Comput. Human Behav. 75, 236–244. https://doi.org/10.1016/j.
1–49. https://doi.org/10.1002/9781119125556.devpsy118. chb.2017.04.063.
Bollen, K.A., 1989. A new incremental fit index for general structural equation models. Little, R.J.A., 1988. A test of missing completely at random for multivariate data with
Sociol. Methods Res. 17, 303–316. https://doi.org/10.1177/ missing values. J. Am. Stat. Assoc. 83 (404), 1198–1202. https://doi.org/10.1080/
0049124189017003004. 01621459.1988.10478722.
Boyes, M.E., Wilmot, A., Hasking, P.A., 2020. Nonsuicidal Self-Injury-Related Differences Liu, S., You, J., Ying, J., Li, X., Shi, Q., 2020. Emotion reactivity, nonsuicidal self-injury,
in the Experience of Negative and Positive Emotion. Suicide Life-Threatening Behav. and regulatory emotional self-efficacy: a moderated mediation model of suicide
50, 437–448. https://doi.org/10.1111/sltb.12599. ideation. J. Affect. Disord. 266, 82–89. https://doi.org/10.1016/j.jad.2020.01.083.
Bresin, K., Schoenleber, M., 2015. Gender differences in the prevalence of nonsuicidal Lovejoy, M.C., Weis, R., O’Hare, E., Rubin, E.C., 1999. Development and initial
self-injury: A meta-analysis. Clin. Psychol. Rev. 38, 55–64. https://doi.org/10.1016/ validation of the Parent Behavior Inventory. Psychol. Assess. 11, 534–545. https://
j.cpr.2015.02.009. doi.org/10.1037/1040-3590.11.4.534.
Brickman, L.J., Ammerman, B.A., Look, A.E., Berman, M.E., McCloskey, M.S., 2014. The Martin, J., Bureau, J.-F., Cloutier, P., Lafontaine, M.-F., 2011. A comparison of
relationship between non-suicidal self-injury and borderline personality disorder invalidating family environment characteristics between university students
symptoms in a college sample. Borderline Personal. Disord. Emot. dysregulation. 1, engaging in self-injurious thoughts & actions and non-self-injuring university
1–8. https://doi.org/10.1186/2051-6673-1-14. students. J. Youth Adolesc. 40, 1477–1488. https://doi.org/10.1007/s10964-011-
Brown, R.C., Plener, P.L., 2017. Non-suicidal self-injury in adolescence. Curr. Psychiatry 9643-9.
Rep. 19, 20. https://doi.org/10.1007/s11920-017-0767-9. Moore, S.A., Dowdy, E., Nylund-Gibson, K., Furlong, M.J., 2019. A latent transition
Buehler, C., Gerard, J.M., 2013. Cumulative family risk predicts increases in adjustment analysis of the longitudinal stability of dual-factor mental health in adolescence.
difficulties across early adolescence. J. Youth Adolesc. 42, 905–920. https://doi.org/ J. Sch. Psychol. 73, 56–73. https://doi.org/10.1016/j.jsp.2019.03.003.
10.1007/s10964-012-9806-3. Muthén, L.K., Muthén, B., 2017. Mplus user’s guide: Statistical analysis with latent
Bureau, J.-F., Martin, J., Freynet, N., Poirier, A.A., Lafontaine, M.-F., Cloutier, P., 2010. variables, user’s guide. Muthén & Muthén.
Perceived dimensions of parenting and non-suicidal self-injury in young adults. Newcomb-Anjo, S.E., Barker, E.T., Howard, A.L., 2017. A Person-Centered Analysis of
J. Youth Adolesc. 39, 484–494. https://doi.org/10.1007/s10964-009-9470-4. Risk Factors that Compromise Wellbeing in Emerging Adulthood. J. Youth Adolesc.
Burke, T.A., Fox, K., Kautz, M., Siegel, D.M., Kleiman, E., Alloy, L.B., 2021. Real-time 46, 867–883. https://doi.org/10.1007/s10964-016-0603-2.
monitoring of the associations between self-critical and self-punishment cognitions Nock, M.K., 2010. Self-Injury. Annu. Rev. Clin. Psychol. 6, 339–363. https://doi.org/
and nonsuicidal self-injury. Behav. Res. Ther. 137, 103775 https://doi.org/10.1016/ 10.1146/annurev.clinpsy.121208.131258.
j.brat.2020.103775. Nock, M.K., Prinstein, M.J., 2004. A functional approach to the assessment of self-
Case, J.A.C., Burke, T.A., Siegel, D.M., Piccirillo, M.L., Alloy, L.B., Olino, T.M., 2020. mutilative behavior. J. Consult. Clin. Psychol. 72, 885–890. https://doi.org/
Functions of non-suicidal self-injury in late adolescence: a latent class analysis. Arch. 10.1037/0022-006X.72.5.885.
suicide Res. 24, S165–S186. https://doi.org/10.1080/13811118.2019.1586607. Nock, M.K., Wedig, M.M., Holmberg, E.B., Hooley, J.M., 2008. The emotion reactivity
Claes, L., Luyckx, K., Baetens, I., Van de Ven, M., Witteman, C., 2015. Bullying and scale: development, evaluation, and relation to self-injurious thoughts and
Victimization, Depressive Mood, and Non-Suicidal Self-Injury in Adolescents: The behaviors. Behav. Ther. 39, 107–116. https://doi.org/10.1016/j.beth.2007.05.005.
Moderating Role of Parental Support. J. Child Fam. Stud. 24, 3363–3371. https:// Nolen-Hoeksema, S., 2012. Emotion regulation and psychopathology: The role of gender.
doi.org/10.1007/s10826-015-0138-2. Annu. Rev. Clin. Psychol. 8, 161–187. https://doi.org/10.1146/annurev-clinpsy-
Clinton, J., 2008. Resilience and recovery. Int. J. Child. Spiritual. 13, 213–222. https:// 032511-143109.
doi.org/10.1080/13644360802236474. Nolen-Hoeksema, S., Aldao, A., 2011. Gender and age differences in emotion regulation
Collins, L.M., Lanza, S.T., 2009. Latent class and latent transition analysis: With strategies and their relationship to depressive symptoms. Pers. Individ. Dif. 51,
applications in the social, behavioral, and health sciences. John Wiley & Sons. 704–708. https://doi.org/10.1016/j.paid.2011.06.012.
https://doi.org/10.1002/9780470567333.ch1. Nylund-Gibson, K., Grimm, R., Quirk, M., Furlong, M., 2014. A Latent Transition Mixture
Copeland, W., Shanahan, L., Jane Costello, E., Angold, A., 2009. Configurations of Model Using the Three-Step Specification. Struct. Equ. Model. 21, 439–454. https://
common childhood psychosocial risk factors. J. Child Psychol. Psychiatry. 50, doi.org/10.1080/10705511.2014.915375.
451–459. https://doi.org/10.1111/j.1469-7610.2008.02005.x. Peugh, J., Fan, X., 2013. Modeling Unobserved Heterogeneity Using Latent Profile
Cummings, J.R., Bornovalova, M.A., Ojanen, T., Hunt, E., MacPherson, L., Lejuez, C., Analysis: A Monte Carlo Simulation. Struct. Equ. Model. 20, 616–639. https://doi.
2013. Time doesn’t change everything: The longitudinal course of distress tolerance org/10.1080/10705511.2013.824780.
and its relationship with externalizing and internalizing symptoms during early Ren, Y., Zhang, X., You, J., Jiang, Y., Lin, M., pei Leung, F., 2018. The reciprocal
adolescence. J. Abnorm. Child Psychol. 41, 735–748. https://doi.org/10.1007/ associations between identity disturbance, relationship disturbance, and suicidal
s10802-012-9704-x. ideation among Chinese adolescents: A three-wave cross-lag study. J. Clin. Psychol.
Dahl, R.E., Allen, N.B., Wilbrecht, L., Suleiman, A.B., 2018. Importance of investing in 74, 1174–1188. https://doi.org/10.1002/jclp.22573.
adolescence from a developmental science perspective. Nature. 554, 441–450. Rutter, M., Caspi, A., Moffitt, T.E., 2003. Using sex differences in psychopathology to
https://doi.org/10.1038/nature25770. study causal mechanisms: Unifying issues and research strategies. J. Child Psychol.
Dawkins, J.C., Hasking, P.A., Boyes, M.E., Greene, D., Passchier, C., 2019. Applying a Psychiatry Allied Discip. 44, 1092–1115. https://doi.org/10.1111/1469-
cognitive-emotional model to nonsuicidal self-injury. Stress Heal 35, 39–48. https:// 7610.00194.
doi.org/10.1002/smi.2837. Simons, J.S., Gaher, R.M., 2005. The distress tolerance scale: Development and
Evans, G.W., Li, D., Whipple, S.S., 2013. Cumulative risk and child development. validation of a self-report measure. Motiv. Emot. 29, 83–102. https://doi.org/
Psychol. Bull. 139, 1342–1396. https://doi.org/10.1037/a0031808. 10.1007/s11031-005-7955-3.
Fassbinder, E., Schweiger, U., Martius, D., Brand-de Wilde, O., Arntz, A., 2016. Emotion Slabbert, A., Hasking, P., Boyes, M., 2018. Riding the emotional roller coaster: The role
regulation in schema therapy and dialectical behavior therapy. Front. Psychol. 7, of distress tolerance in non-suicidal self-injury. Psychiatry Res 269, 309–315.
1373. https://doi.org/10.3389/fpsyg.2016.01373. https://doi.org/10.1016/j.psychres.2018.08.061.
Fox, K.R., Huang, X., Linthicum, K.P., Wang, S.B., Franklin, J.C., Ribeiro, J.D., 2019. Somer, O., Bildik, T., Kabukçu-Başay, B., Güngör, D., Başay, Ö., Farmer, R.F., 2015.
Model complexity improves the prediction of nonsuicidal self-injury. J. Consult. Clin. Prevalence of non-suicidal self-injury and distinct groups of self-injurers in a
Psychol. 87, 684. https://doi.org/10.1037/ccp0000421. community sample of adolescents. Soc. Psychiatry Psychiatr. Epidemiol. 50,
Gong, T., Ren, Y., Wu, J., Jiang, Y., Hu, W., You, J., 2019. The associations among self- 1163–1171. https://doi.org/10.1007/s00127-015-1060-z.
criticism, hopelessness, rumination, and NSSI in adolescents: A moderated mediation Sontag, L.M., Graber, J.A., Clemans, K.H., 2011. The Role of Peer Stress and Pubertal
model. J. Adolesc. 72, 1–9. https://doi.org/10.1016/j.adolescence.2019.01.007. Timing on Symptoms of Psychopathology During Early Adolescence. J. Youth
Hamza, C.A., Willoughby, T., Heffer, T., 2015. Impulsivity and nonsuicidal self-injury: A Adolesc. 40, 1371–1382. https://doi.org/10.1007/s10964-010-9620-8.
review and meta-analysis. Clin. Psychol. Rev. 38, 13–24. https://doi.org/10.1016/j. Sornberger, M.J., Heath, N.L., Toste, J.R., 2012. Nonsuicidal self-injury and gender:
cpr.2015.02.010. Patterns of prevalence, methods, and locations among adolescents. Suicide Life-
Howard, M.C., Hoffman, M.E., 2018. Variable-Centered, Person-Centered, and Person- Threatening Behav. McLouth, R., 42, 266–278. https://doi.org/10.1111/j.1943-
Specific Approaches: Where Theory Meets the Method. Organ. Res. Methods. 21, 278X.2012.0088.x.
846–876. https://doi.org/10.1177/1094428117744021. Swannell, S.V., Martin, G.E., Page, A., Hasking, P., St John, N.J., 2014. Prevalence of
Kang, N., Jiang, Y., Ren, Y., Gong, T., Liu, X., Leung, F., You, J., 2018. Distress nonsuicidal self-injury in nonclinical samples: Systematic review, meta-analysis and
Intolerance Mediates the Relationship between Child Maltreatment and Nonsuicidal meta-regression. Suicide Life-Threatening Behav. 44, 273–303. https://doi.org/
Self-Injury among Chinese Adolescents: A Three-Wave Longitudinal Study. J. Youth 10.1111/sltb.12070.
Adolesc. 47, 2220–2230. https://doi.org/10.1007/s10964-018-0877-7. Tang, J., Li, G., Chen, B., Huang, Z., Zhang, Y., Chang, H., Wu, C., Ma, X., Wang, J.,
Klonsky, E.D., Glenn, C.R., Styer, D.M., Olino, T.M., Washburn, J.J., 2015. The functions Yu, Y., 2018. Prevalence of and risk factors for non-suicidal self-injury in rural China:
of nonsuicidal self-injury: Converging evidence for a two-factor structure. Child Results from a nationwide survey in China. J. Affect. Disord. 226, 188–195. https://
Adolesc. Psychiatry Ment. Health. 9, 1–9. https://doi.org/10.1186/s13034-015- doi.org/10.1016/j.jad.2017.09.051.
0073-4. Taouk, M., Lovibond, P.F., Laube, R., 2001. Psychometric properties of a Chinese version
Lanza, S.T., Cooper, B.R., 2016. Latent Class Analysis for Developmental Research. Child of the short Depression Anxiety Stress Scales (DASS21). Rep. new South Wales
Dev. Perspect. 10, 59–64. https://doi.org/10.1111/cdep.12163. Transcult. Ment. Heal. centre, Cumberl. Hosp. Sydney.

70
Q. Gao et al. Journal of Affective Disorders 295 (2021) 63–71

Turner, B.J., Cobb, R.J., Gratz, K.L., Chapman, A.L., 2016. The role of interpersonal You, J., Jiang, Y., Zhang, M., Du, C., Lin, M.P., Leung, F., 2017. Perceived Parental
conflict and perceived social support in nonsuicidal self-injury in daily life. Control, Self-Criticism, and Nonsuicidal Self-Injury Among Adolescents: Testing the
J. Abnorm. Psychol. 125, 588–598. https://doi.org/10.1037/abn0000141. Reciprocal Relationships by a Three-Wave Cross-Lag Model. Arch. Suicide Res. 21,
Wang, B., You, J., Lin, M.P., Xu, S., Leung, F., 2017. Developmental Trajectories of 379–391. https://doi.org/10.1080/13811118.2016.1199989.
Nonsuicidal Self-Injury in Adolescence and Intrapersonal/Interpersonal Risk Factors. You, J., Leung, F., Lai, C.M., Fu, K., 2012. The associations between non-suicidal self-
J. Res. Adolesc. 27, 392–406. https://doi.org/10.1111/jora.12273. injury and borderline personality disorder features among Chinese adolescents.
Wang, M.-T., Peck, S.C., 2013. Adolescent educational success and mental health vary J. Pers. Disord. 26, 226–237. https://doi.org/10.1521/pedi.2012.26.2.226.
across school engagement profiles. Dev. Psychol. 49, 1266. https://doi.org/ You, J., Lin, M.P., 2015. Predicting suicide attempts by time-varying frequency of
10.1037/a0030028. nonsuicidal self-injury among Chinese community adolescents. J. Consult. Clin.
Whiteside, S.P., Lynam, D.R., 2001. The five factor model and impulsivity: Using a Psychol. 83, 524–533. https://doi.org/10.1037/a0039055.
structural model of personality to understand impulsivity. Pers. Individ. Dif. 30, Zahn-Waxler, C., Shirtcliff, E.A., Marceau, K., 2008. Disorders of childhood and
669–689. https://doi.org/10.1016/S0191-8869(00)00064-7. adolescence: Gender and psychopathology. Annu. Rev. Clin. Psychol. 4, 275–303.
Wu, N., Hou, Y., Zeng, Q., Cai, H., You, J., 2021. Bullying experiences and nonsuicidal https://doi.org/10.1146/annurev.clinpsy.3.022806.091358.
self-injury among Chinese adolescents: a longitudinal moderated mediation model. Zanarini, M.C., Gunderson, J.G., Frankenburg, F.R., Chauncey, D.L., 1989. The revised
J. Youth Adolesc. 50, 753–766. https://doi.org/10.1007/s10964-020-01380-1. diagnostic interview for borderlines: Discriminating BPD from other Axis II
Xavier, A., Pinto Gouveia, J., Cunha, M., 2016. Non-suicidal Self-Injury in Adolescence: disorders. J. Pers. Disord. 3, 10–18. https://doi.org/10.1521/pedi.1989.3.1.10.
The Role of Shame, Self-Criticism and Fear of Self-Compassion. Child Youth Care Zhang, X., Ren, Y., You, J., Huang, C., Jiang, Y., Lin, M.P., Leung, F., 2017.
Forum 45, 571–586. https://doi.org/10.1007/s10566-016-9346-1. Distinguishing Pathways from Negative Emotions to Suicide Ideation and to Suicide
Yates, T.M., 2004. The developmental psychopathology of self-injurious behavior: Attempt: the Differential Mediating Effects of Nonsuicidal Self-Injury. J. Abnorm.
Compensatory regulation in posttraumatic adaptation. Clin. Psychol. Rev. 24, 35–74. Child Psychol. 45, 1609–1619. https://doi.org/10.1007/s10802-017-0266-9.
https://doi.org/10.1016/j.cpr.2003.10.001. Zhou, J., Jiang, S., Zhu, X., Huebner, E.S., Tian, L., 2020. Profiles and Transitions of
You, J., Deng, B., Lin, M.P., Leung, F., 2016. The Interactive Effects of Impulsivity and Dual-Factor Mental Health among Chinese Early Adolescents: The Predictive Roles of
Negative Emotions on Adolescent Nonsuicidal Self-injury: A Latent Growth Curve Perceived Psychological Need Satisfaction and Stress in School. J. Youth Adolesc. 49,
Analysis. Suicide Life-Threatening Behav. 46, 266–283. https://doi.org/10.1111/ 2090–2108. https://doi.org/10.1007/s10964-020-01253-7.
sltb.12192.

71