1. Introduction
Among highly socially discouraged behaviors, non-suicidal self-injury (NSSI) demands special attention due to its high prevalence among adolescents and young adults, and due to its adverse mental long-term consequences for the person with NSSI [
1]. More specifically, NSSI is associated with a higher risk of suicide behavior [
2]. Non-suicidal self-injury (NSSI) refers to intentional self-inflicted destruction of body tissue without suicidal intention [
3,
4]. Behavior of NSSI may include cutting, pricking, sandpapering, biting, self-punching, spraying acid, hitting, scratching, and breaking and burning oneself [
5], along with the result of the destruction or disfigurement of a part of the body. Further, such behaviors often are repetitive, intentional, and direct, and they are not in line with social expectations and norms [
6]. The prevalence rates of NSSI range between 5.5% and 17% in the general population, between 5.5% and 17% among adolescents, 13.4% among young adults, and 5.5% among adults [
7]. It follows that the prevalence rate of NSSI is higher during late adolescence and early adulthood, that is to say that NSSIs dramatically increase during the psychosexually, psychosocially, and economically demanding developmental stage of early adulthood. As such, NSSIs are observed in the most economically active population, and the population with the highest pressure to prevail and assert for successful mating [
8,
9,
10].
To explain the emergence and occurrence of NSSIs, at least six highly intertwined dimensions are identified, which for clarity are reported separately. (1) Unfavorable personality traits refer to traits such as low coping skills [
1,
11], low emotion regulation [
1,
12,
13], high emotion reactivity [
14], unfavorable shame-coping [
15], and low self-efficacy [
14,
16]. (2) Psychiatric issues refer for instance to symptoms of panic disorders [
13], post-traumatic stress disorders [
13], symptoms of anxiety and depression [
17,
18,
19], alexithymia [
20,
21], and substance use disorders [
19], including risky drinking behavior [
12]. (3) Past and current social interactions refer to the lack of reliable and long-term relationships (also as a proxy of poor attachment to a reliable person [
22]), and the occurrence of childhood maltreatment [
1,
19], sexual abuse [
22,
23], and negative life events [
24]. (4) Sociodemographic information: younger age [
15,
19] and female gender [
12,
19] were associated with higher risks of NSSIs. (5) Ethnic groups refer to the observation that black females are at higher risk, and Asian males are at lower risk to report NSSIs [
11] (6) Poor sleep: a recent review [
25] identified 16 publications on the relation between poor sleep and the risk of NSSIs. The overall pattern was as follows: poor sleep was associated with NSSIs both cross-sectionally and longitudinally. Further, three out of the 16 publications reported in the review had a longitudinal design: among 392 adolescents, sleep problems at the age of 12–14 years predicted the occurrence of NSSIs two to three years later [
26]. Among 881 adolescents, poor sleep predicted the occurrence of NSSI among female adolescents, but not male adolescents [
27]. Among 72 university students, those reporting nightmares at baseline reported the occurrence of NSSIs later; negative effects mediated the relationship between nightmares and self-harmful thoughts and behaviors [
28].
A very similar pattern was observed among 1973 children crossing into adolescence: Participants’ sleep trajectories were observed when children were approximately seven to nine years old, and approximately six years later; three characteristic patterns were observed: those participants who continued to sleeping > 8/night, those moderately decreasing the sleep duration, and those rapidly decreasing the sleep duration. Compared to those continuing to sleep > 8 h/night, those moderately decreasing the sleep duration had a 2.58-fold risk to report NSSIs, and those reporting a rapidly decreasing sleep duration had a 4.86-fold risk to report NSSIs [
29]. Given this background, the first aim of the present study was to investigate the associations between subjective sleep patterns and past and current NSSI behaviors.
Next, while there is a growing body of research on the association between poor sleep and suicidal thoughts and suicidal behavior [
30,
31], this is not the case regarding the associations between NSSI and nightmares: Among adolescents, a higher prevalence of nightmares was associated with more NSSI and suicidal ideations, both cross-sectionally [
32] and longitudinally [
33]. We also note that the direction of influence was from nightmares to NSSI and suicidal ideation. However, also the opposite direction is conceivable: a person with NSSI could be impaired in their cognitive-emotional information processing, such to trigger nightmares as a strategy for coping with their concerns and serious mental health issues. With this assumption in mind, the second aim of the present study was to investigate if remembered NSSI occurred 12 months ago predicted the occurrence of nightmares later on. Similarly, the third aim was to associate the occurrence of nightmares with NSSI.
To summarize, a broad variety of psychological, psychosocial, psychiatric, cognitive-emotional, and sleep-related dimensions may explain the emergence and maintenance of NSSIs. In the present study, we focused on the relation between NSSIs and sleep. While there is already extant literature in this regard, we expanded upon the current research in the following four ways. First, we assessed a larger sample of people during early adulthood. As mentioned, this period of psychosocial development as particularly demanding as this age-range reflects the most economically active population, and the population with the highest pressure to prevail and assert for successful mating [
8,
9,
10]. Second, we assessed a community sample, while previous study prevalently focused on high-school or university students. Third, unlike previous studies, which investigated the predictive values of sleep patterns on NSSI, we asked if and to what extent NSSIs could predict sleep issues. Fourth, we designed a quasi-longitudinal study; to this end, participants reported both current NSSIs and NSSIs 12 months before.
Based on the current research, the following five hypotheses and five research questions were formulated. First, following others [
25], we predicted that poor sleep and NSSIs were associated with one another. Second, based on previous research, we predicted that younger age [
15,
19] was associated with higher NSSIs. We also predicted that compared to their male counterparts, female participants would report more sleep-related issues [
34,
35] (third hypothesis) and more NSSRIs [
12,
19] (fourth hypothesis). Next, we assumed that NSSI and the occurrence of nightmares would be associated [
32,
33] (fifth hypothesis). The explorative research questions were as follows: first, we explored if NSSIs 12 months ago could predict the occurrence of nightmares. Second, we investigated if past NSSIs predicted current NSSIs (second research question), suicide attempts (third research question), and current poor sleep (fourth research question). Last, we explored which factors could predict current sleep quality (Pittsburgh Sleep Quality Index [PSQI], insomnia: Insomnia Severity Index [ISI]) and current NSSIs. To this end, we calculated a series of equations to test the direct and indirect effects of such factors.
Overall, we claim that the present results are of clinical and practical importance, since data have the potential to identify psychological risk factors of young adults at risk of reporting both past and current NSSIs.
4. Discussion
The key findings of the present study on non-suicidal self-injury (NSSI) behavior among a larger sample of young adults were as follows: first, past NSSIs predicted current NSSIs. Second, current sleep patterns had a modest impact on the association between past and current NSSIs. Third, compared to male participants, female participants did not report more sleep complaints or more current NSSIs, but did report more past NSSIs. Fourth, past NSSIs predicted the occurrences of nightmares and suicide attempts. The present results expand upon previous results in the following four ways. First, while previous studies focused on the association and influence of sleep on NSSIs, here, in a quasi-longitudinal design, we showed that current sleep patterns were modestly associated with current NSSIs, when past NSSIs were introduced as a further factor. Second, data were extracted from a larger non-clinical sample of young adults, for whom, by definition, the pressure to achieve on the economic and mating markets is particularly high. Third, against expectations, compared to male participants, female participants neither reported more sleep issues, nor more current NSSIs, but did report more NSSIs twelve months ago. Fourth, the occurrence of nightmares appeared to be important. Overall, the present pattern of results supports the notion of past and current NSSIs, poor sleep, nightmares, and suicide attempts as proxies of distressed mental health and deteriorated emotion regulation.
Five hypotheses and five research questions were formulated, and each of these is considered now in turn.
First, following others [
25] we predicted that poor sleep and past and current NSSIs were associated, and data did support this assumption. However, as shown in
Table 2, correlation coefficients were small to medium, suggesting that further unassessed, and latent factors might have had a stronger impact on such associations (see also below). The novelty of the present results is three-fold: first, data were carried out from a larger and non-clinical sample of young adults. Second, we introduced the recall of NSSIs occurred twelve months ago. Third, while in previous longitudinal studies [
26,
27,
28,
29] sleep patterns were considered as predictors and NSSIs as dependent variable, here, the study design was such to introduce recalled past NSSI as predictor, and sleep patterns as dependent variables. As such, we claim that this methodological change of paradigm expanded upon the current knowledge in the field in an important fashion.
With the second hypothesis we assumed that younger age was associated with higher NSSIs, and data did confirm this. Thus, the present results replicated and confirmed what has already been reported elsewhere [
15,
19]. The quality of the data does not allow a deeper understanding of this association. We know from imaging and neurophysiological studies on morphological brain changes among adolescents and individuals crossing into (emerging) adulthood that younger age appeared to be associated with higher impulsivity [
51,
52,
53,
54]. In the same vein, impulsive and compulsive symptoms become apparent during young adulthood, which, by definition, is a critical time for brain development and the establishments of life goals [
55]. From late childhood to young adulthood, brain maturation occurs in brain regions associated with cognitive control and goal-directional behavior, including working memory, social cognition, and inhibitory control [
56]. Not surprising, a bidirectional relationship between impulsivity and NSSIs was observed among 782 adults within a time lapse of three years [
57]. To conclude, we assume that also in the present study ongoing brain maturation, impulsivity and higher non-suicidal self-injury patterns might be associated with younger age. On the flip side (see
Table 2) correlation coefficients between age and past and current NSSIs were modest and trivial. As such, the influence of age should not be overestimated. This statement received further support from developmental studies, in which cognitive control had a larger importance on risk-taking and impulsivity, compared to age [
58].
With the third hypothesis, we assumed that, compared to their male counterparts, female participants would report more sleep-related issues, though, data did not support this assumption (see
Table 3). As such, the present results do not match those of previous studies [
34,
35,
59,
60]. Again, the quality of the present data does not allow a thorough explanation of this zero-result. One explanation might be that other studies did fully rely on
p-values, which, by definition, are sensitive to sample sizes. In contrast, as shown in
Table 3, we relied on effects sizes; as such, effect sizes are more precise to accurately reflect the correct mean differences [
61].
With the fourth hypothesis we assumed that compared to male participants, female participants would report more NSSIs, although this assumption was only partially confirmed. More specifically, the assumption was confirmed about recalled and past NSSIs, but not about current NSSIs (
Table 3). As such, the present results do not match what has been observed before [
12,
19]. To explain this mismatch between the present and previous results, again, the quality of the data does not provide more insight. We advance the following, although admittedly highly speculative assumptions: (1) compared to previous studies the results of the current study were drawn from a larger non-clinical sample of young adults, leading thus to different results; (2) data on past NSSIs have not been investigated so far, thus, comparisons might be biased; (3) different measurements to assess NSSIs were used, which in turn might have yielded incongruent overall results.
With the fifth and last hypothesis, we assumed that NSSIs would be associated with the occurrence of nightmares, and data did confirm this. As such, we were able to replicate what has been observed before [
32,
33]. We claim that this pattern of results is in accord with the assumption that in individuals with NSSIs, the underlying cognitive-emotional information processing appears to reflect a highly distressed psychobiological system and a dysregulated emotion processing.
Next, we also formulated five exploratory research questions.
With the first exploratory question we investigated if NSSIs 12 months ago could predict the occurrence of nightmares, and the answer was yes. As such, this pattern of result is novel, and adds to the current knowledge on the relationship between NSSIs, nightmares, and sleep in a new and important fashion. To explain the possible underlying mechanisms, we refer to the more general discussion below.
Second, we investigated if past NSSIs predicted current NSSIs, and the answer was again yes: past NSSIs predicted current NSSIs. In our opinion, for the following four reasons, this is the most important result of the present study. First, unlike previous studies on the relationship between NSSI and sleep, here, past NSSIs were introduced as independent predictor. Therefore, second, this study design changed the paradigm in this field or research. Third, it appeared that the non-suicidal self-injury-related behavioral pattern remained quite stable over time. However, fourth, we also note that the correlation coefficient was r = 0.552, or conversely: the variance of past NSSIs explained approximately 30% of the variance of the current NSSIs. Thus, approximately 70% of the variance of the current NSSI remained unexplained. One might question about the quality of the present quasi-longitudinal study design, and about participants’ ability to recall their non-suicidal self-injury-related behavior 12 months ago. This objection is justified; however, in order to minimizing the recall bias, we asked explicitly participants to vividly imaging what they were doing 12 months previously. From studies on mood induction via memory recall we know that this technique appears reliable to retrieve specific information from the long-term memory [
62,
63].
The third research question asked was if past NSSI-behavior could predict suicide attempts within the following 12 months, and the answer was yes. As such, in our opinion, this result underscores and justifies the present quasi-longitudinal study design. The result also confirms that NSSIs are not limited to future NSSIs (see also results of the second research question), but to an increased risk to turn NSSIs into suicide attempts [
2,
64].
The fourth research question asked was if past NSSIs predicted current poor sleep, and the answer was again not as straightforward as expected. While the correlation coefficients were trivial to small (
Table 2), a series of regression equations revealed that both past and current NSSIs predicted current poor sleep. As such, again, we claim that the quasi-longitudinal study design allowed to gain further insight into the psychological mechanisms of NSSIs and poor sleep.
Fifth, and last, we explored, which factors could predict current sleep quality (Pittsburgh Sleep Quality Index [PSQI], Insomnia Severity Index [ISI]) and current NSSIs. Note that to answer to these questions, we performed a series of equations to test the direct and indirect effects of possible predictors. Briefly, both past and current NSSI predicted poor sleep both directly and indirectly, while poor sleep had no predictive power on current NSSIs. This pattern of results was unexpected, as it does not match previous concepts [
25,
29]. However, as has already been mentioned, previous studies were either cross-sectional, and as such unable to allow to draw causal effects, or previous studies were longitudinal, but with sleep as predictor and NSSIs as dependent variables. Overall, we claim once again that with the present quasi-longitudinal study design and with past NSSIs as predictors, the entire pattern of the relationship between NSSIs and sleep changed.
Furthermore, the quality of the present study does not allow a deeper understanding of the cognitive-emotional and neurophysiological mechanisms to explain the associations between past and current NSSIs and poor sleep patterns. However, to counterbalance, we refer to previous findings in the field of cognitive-emotional sleep regulation and NSSIs.
First, as mentioned above, we know from imaging and neurophysiological studies on morphological brain changes among adolescents and individuals crossing into (emerging) adulthood [
51,
52,
53,
54] that impulsive and compulsive symptoms become apparent during young adulthood, which, by definition, is a critical time for brain development, for the establishments of life goals [
55], and for successful mating [
8,
9,
10]. Above all, regarding the associations between suicide attempts and successful mating, previous results evidenced that higher risks of suicidal behavior were associated with less success in mating and with low sexual activities within the last 12 months [
65,
66,
67]. Next, from late childhood to young adulthood, brain maturation occurs in brain regions associated with cognitive control and goal-directional behavior, including working memory, social cognition, and inhibitory control [
56], with a bidirectional relationship between impulsivity and NSSIs [
57]. We also note that cognitive control had a larger importance on risk-taking and impulsivity, compared to age [
58]. As such, we claim that also in the present study ongoing brain maturation, impulsivity and higher non-suicidal self-injury patterns were associated in a bi-directional fashion.
Second, there is sufficient evidence that poor sleep, poor executive control, higher impulsivity, and emotion dysregulation are associated. More specifically, and as more extensively described in Khazaie et al. [
25], briefly, four theoretical frameworks are proposed; these frameworks are highly intertwined, and presented separately for clarity and methodological reasons.
First, the concept of cognitive-emotional hyperarousal of poor sleep [
68,
69] asserts that impaired sleep is the result of dysfunctional and over-aroused cognitive-emotional processes.
Second, the psychophysiological hyperarousal model of Riemann et al. [
70] expands upon the purely cognitive-emotional model in which severe sleep disturbances and depression are the endpoint of a bi-directional and deteriorating process between dysfunctional cognitive-emotional processes and sleep-impairing physiological changes.
Third, the cortical hyperarousal model of Fernandez-Mendoza et al. [
71] and Zhao et al. [
72] claims that an increased activity in brain network configurations appeared to be involved in the pathophysiology of insomnia.
Finally, fourth, following Yoo et al. [
73] individuals with insomnia showed an unstable and disrupted interplay and connection between the medial-prefrontal cortex (MPFC) and its inhibitory and top-down control of the amygdala and adjacent mesolimbic structures, leading to less functional emotional responses and to a lower impulse control inhibition.
Overall, we claim that the results on brain maturation, inhibition control, and impaired sleep due to dysfunctional cognitive-emotional and physiological processes and increased NSSIs appear to reflect the same neurophysiological and cognitive-emotional processes.
The novelty of the results should be balanced against the following limitations. First, given the large sample size, we relied on effect size calculations, as effect sizes reflect the true mean differences, while
p-values become “significant” with increasing sample sizes. Second, the study design was quasi-longitudinal; one may claim that completing questionnaire items on the present and past NSSIs might lead to biased results. While we cannot fully rule out this risk, the following points should be considered. First, as mentioned, participants were encouraged to recall their life 12 months ago, and there is evidence that such interventions impact on the retrieval of long-term memory information [
62,
63]. Second, if reporting current NSSIs and recalling past NSSIs were identical or biased, then one would expect a definitively higher correlation coefficient than just r = 0.552. Nevertheless, we are aware that the present quasi-longitudinal study design is unable to replace longitudinal studies with two or more timepoints.
Third, by definition, it is conceivable that latent and unassessed cognitive-emotional processes might have biased two of more dimensions in the same or opposite directions. As described extensively in the Introduction section, NSSIs are related to a broad variety of psychological, psychiatric, social-interactional, and cultural factors. More specifically, one might claim that symptoms of depression might have biased the whole pattern of results. While with the present data we cannot full rule out this claim, we also note that among a larger sample of adolescents the association between suicidal behavior and poor sleep was above and beyond symptoms of depression [
31]. As such, it is highly conceivable that also in the present study, symptoms of depression were not confounders.
In this view, fourth, there is sufficient evidence that sleep complaints among adolescents and young adults might be related to excessive exposure to screens (e.g., tablets, smart phones). [
74,
75,
76,
77]. As such, it is conceivable that exposure to screen might have biased the present pattern of results.