Non‐suicidal self‐injury and its co‐occurrence with suicidal behavior: An epidemiological‐study among adolescents and young adults

Non‐suicidal self‐injury (NSSI) comprising thoughts and behaviors is common and often co‐occurring with suicidal behavior like ideation, plan, and attempt. As limited data are available for adolescents and young adults, this study aims to present prevalence estimates for lifetime NSSI, its co‐occurrence with suicidal behavior, conditional probabilities and their association with socio‐demographic characteristics, severity characteristics of suicidal behavior, and health service utilization.


Introduction
Non-suicidal self-injury (NSSI) and suicidal behaviors are two distinct behaviors demonstrated by their differences in frequency, used methods, severity, and function (1)(2)(3)(4). The use of different terms and definitions impede the understanding of these behaviors and their associations (4,5). Here, NSSI can be understood as a thought or behavior (i.e. action) to do deliberate, self-inflicted destructions of body tissue without a suicidal intent (4), and suicidal behavior can be understood as suicidal ideation, plan or attempt with the intend to die (6). On the one hand, etiological theories regarding suicidal behavior include NSSI as a risk factor for suicidal behavior (6,7), which was supported by empirical evidence in several meta-analyses (8)(9)(10)(11). On the other hand, these behaviors have shared risk factors like mental disorders (e.g., depression (8,12)) or early childhood traumatic events (4,8), and a similar sensitive period for the first onset in adolescence and young adulthood (5,9,(13)(14)(15).
Increasing prevalence estimates of NSSI were found in recent years (16)(17)(18). In a study from England, an increase in the prevalence of NSSI behaviors was found in an adult sample aged 16-74 years from 2000 to 2014, with the highest increase (6.5%-19.7%) in females aged 16-24 years (16). Overall, lifetime prevalence estimates vary widely for NSSI behavior from 1.5% to 54.8% with a mean lifetime prevalence of 18.0%-26.4% in males and up to 33.8% in females (19)(20)(21). Comparing different age groups, adolescent samples show the highest lifetime prevalence estimates compared to young adults (20). In contrast to NSSI behaviors, less is known about NSSI thoughts. A study in college students aged 18-35 years found a lifetime prevalence of 22.6% and a 12-month prevalence of 8.8% (22). A higher estimate of 24.1% for the past 6 months was found in a study of 15 year olds (23). Compared to NSSI, lifetime prevalence estimates of suicidal behavior are lower, but they also show a wide range between 1.5 and 37.9% (5,15,24).
An increasing number of studies examined the co-occurrence of NSSI and suicidal behavior. However, only a few studies examined the association with all types of suicidal behavior including ideation, plan, and attempt in adolescents and young adults from the general population and results vary depending on the direction of the analysis (25)(26)(27). So far, research demonstrated high numbers of co-occurrence of NSSI and suicidal behaviors in adolescents and young adults (25)(26)(27). Moreover, this co-occurrence seems to be associated with an impaired health over time, for example, more depressive symptoms, and lowered selfesteem (8,28,29). However, previous research lacks the differentiation between NSSI thoughts and behaviors as well as their associations with all types and course characteristics of suicidal behavior like age at onset and frequency in adolescence and young adulthood. Such data are crucial to understand the burden and relevance of NSSI thoughts and behaviors in this critical developmental period. Moreover, knowledge is scarce regarding sex, age, and education status differences as well as mental healthcare utilization.

Aims of the study
The aims of the present study are (1) to present lifetime prevalence estimates for any NSSI including thoughts (NSSI-T) and behaviors (NSSI-B) and to examine associations with socio-demographic characteristics; (2) to describe the co-occurrence of NSSI and lifetime suicidal behavior including ideation, plan and attempt, and to examine conditional probabilities and their associations with socio-demographic characteristics and characteristics of suicidal behavior (i.e., age at onset and number of episodes); and (3) to report frequencies of lifetime mental healthcare utilization in those with NSSI with and without any suicidal behavior.

Study design and procedures
Cross-sectional data come from the Behavior and Mind Health (BeMIND) study, an epidemiological cohort study among adolescents and young adults from Dresden, Germany (30). An age-and sexstratified random sample of 14-21 year olds was drawn from the population registry, followed by a maximum of three written invitation letters to participate in the study. Between November 2015 and December 2016, N = 1180 subjects participated in the study (response/participation rate 21.7%, cooperation rate 42.8% 31) comprising of two assessment days around one week apart in the Center for Longitudinal and Epidemiological Studies (CELOS) at the Technische Universit€ at Dresden. The two assessment days at the study center included a standardized diagnostic interview, questionnaires, cognitive paradigms, and bio-sampling (blood and hair samples). In between, smartphone-based ecological momentary assessments over four days and an online questionnaire examining risk and protective factors were issued. All participants provided written informed consent or assent, and of those under 18 years of age, written informed consent of all legal guardians were gathered. Participants received 50 Euro as incentive. The BeMIND study protocol has been accepted by the ethics committee of the Technische Universit€ at Dresden, Germany (EK381102014). For a comprehensive overview of the BeMIND study program, recruitment and detailed sample characteristics, see Beesdo-Baum, Voss, Venz, Hoyer, Berwanger, Kische, Ollmann, Pieper (30). The present study follows the Strengthening the Reporting of Observational Studies (STROBE) guidelines (32).

Measures
At the first assessment day, an updated version (33) of the fully standardized computer-assisted Munich-Composite International Diagnostic Interview (DIA-X/M-CIDI; 34,35-38) was conducted face-to-face by trained clinical interviewers accompanied by tablet-based self-administered questionnaires. During the interview, all participants answered questions about socio-demographic characteristics, suicidal behavior, and service use, and filled out a self-administered questionnaire for NSSI and course characteristics of suicidal behavior.
Lifetime non-suicidal self-injury. To examine NSSI, a questionnaire was used based on two dichotomous items of the self-injurious thoughts and behaviors interview for self-harming behavior (SITBI-G; 39,40) after the interview section for depression. One item assessed NSSI thoughts (NSSI-T): 'Have you ever had thoughts of purposely hurting yourself without wanting to die? (e.g., cutting, scratching, or burning)'; the second item assessed NSSI behaviors (NSSI-B): 'Have you ever purposely hurt yourself without wanting to die?' Lifetime suicidal behavior. Lifetime suicidal behavior was assessed in all participants during the interview section for depression using the following questions: Ideation: 'Have you ever thought over a period of days or weeks about killing yourself, that is, to attempt suicide?' Plan (if ideation): 'Have you ever made a specific plan for killing yourself?' Attempt (all participants): 'Have you ever attempted suicide?' Participants endorsing lifetime suicidal behavior were additionally asked for age at onset and the number of episodes of ideation, plan, and number of attempts using a self-administered questionnaire.
Mental health service utilization. Service utilization was assessed in section Q of the interview by asking the following gate question: 'Have you ever visited/contacted any of the healthcare institutions [as listed in the respondent's booklet] because of mental health, psychosomatic, or substance use problems, either by yourself, or by advise of others, for example, medical doctors, relatives, or your partner?'. If the individuals endorsed the gate question, healthcare use of different institutions was coded separately comprising inpatient (i.e., psychiatric hospital or home for children), outpatient (i.e., general practitioner (if reason was mental health problem), psychotherapist), and complementary healthcare services (i.e., counseling centers for students, telephone counseling).

Statistical analysis
All analyses were weighted to ensure a comparable distribution with the whole population of adolescents and young adults aged 14-21 years in Dresden regarding age and sex. Robust standard errors were calculated by using the first-order Taylor-series linearization method to obtain correct 95% confidence intervals (95% CI) for prevalence estimations (41). The prevalence of the co-occurrence was examined in the total sample and the conditional probabilities (42) were determined by calculating the frequency of individuals with suicidal behavior among those with NSSI and vice versa. Logistic regression models (odds ratios (OR) with 95% CIs) were used to quantify the associations; negative binominal regression (Incidence Rate Ratios, IRR) was used for count data (frequency); and exact Fisher's test was used for associations with mental healthcare use if there were five or less cases in one group. To ensure stability of the models, the most frequent category of a variable was used as the reference group. Tests were done at the 2-sided a = 0.05 level. Regarding course characteristics for suicidal behavior, a dichotomous item was calculated differentiating early and late onset based on the median and categorical variables were built for number of episodes both using previous findings (15). Missing data were conservatively counted as no occurrence (four individuals for suicidal behavior and seven for NSSI and help-seeking behavior). Stata software package, release 14.2, was used to conduct the analyses (43).

Sample characteristics
In the following, the reported percentages are based on weighted data. The weighted mean age of the N = 1180 participants was M = 17.9, SD = 2.3, and 48.3% were female (n = 685). Most participants reported a German citizenship (97.1%) and living with their parents (65.1%; 17.1%, 12.5%, and 5.4% lived with other people, that is, roommates, alone, or with a partner, respectively). Three quarter (76.5%) reported currently a high education, that is, A-Level, a fifth (18.5%) a middle education, that is, secondary school, and 5.0% another (e.g., private schools) or low education status (e.g., lower secondary school levels). More information about the sample characteristics can be found in detail elsewhere (30).

Prevalence of any NSSI, NSSI-T and NSSI-B
Any lifetime NSSI was reported by 19 .0]) than males (lower OR CI at least 1.9). No age group differences were found. The current education status was associated with NSSI-T and NSSI-B indicating more NSSI in the group with a middle, low or other compared to high education status. Among those with an average compared to a good or very good subjective financial status, a significant higher prevalence was found for all types of NSSI. Growing up with a single or no parent was associated with all types of NSSI.
Regarding the co-occurrence of NSSI-T and NSSI-B, both were reported by 12 [95% CI 0.7-3.0]). Significant sex differences emerged indicating that those with both compared to none were more likely females (OR = 4.1, 95% CI 2.6-6.5, P < 0.001). No age group differences were found. Reporting both (OR = 1.7, 95% CI 1.1-2.4, P = 0.011) was more likely in the group with a middle, low or other compared to high education status.

Co-occurrence of NSSI and suicidal behavior
Prevalence and associations. Lifetime prevalence estimates of suicidal behavior in the present sample are published elsewhere (15). Table 2 and Figure 1 depict the frequencies of the co-occurrence of any NSSI, NSSI-T and NSSI-B with suicidal ideation, plan, attempt, and any suicidal behavior for the total sample, by sex, age cohort, and current education status. Overall, 23 The lifetime co-occurrence of any NSSI with any suicidal behavior was reported by 7.7% (95% CI 6.2-9.4) of the adolescents and young adults. Results for NSSI-T and NSSI-B are presented in Figure 1. Those with NSSI and co-occurring suicidal behavior had 3.3-to 8.8-fold odds to be female than male (Appendix Table S1), which is also illustrated in Figure 1. There was no significant difference with regard to age. For the current education status, there was an association indicating that those with NSSI and co-occurring suicidal behavior had more likely a middle, low or other than a high education status.
Regarding non co-occurrence, only suicidal behavior was reported by 3.9% (95% CI 2.8-5.4) and only NSSI by 11.7% (95% CI 9.9-13.7) of the adolescents and young adults ( Table 2). Significant sex differences emerged indicating that those reporting both and NSSI only were more likely females and those with suicidal behavior only more likely males (Appendix Table S1). No age group differences were found. Reporting both was positively associated with a middle, low, or other compared to high education status. Results were similar for any NSSI, NSSI-T, and NSSI-B.
Association with characteristics of suicidal behaviors. Table 3 depicts the characteristics of suicidal behavior and its association with the co-occurrence of any NSSI, NSSI-T, and NSSI-B. Any NSSI, NSSI-T, and NSSI-B were positively associated Data are weighted to refer to the age and sex distribution in the general population of individuals aged 14-21 years in Dresden, Germany; numbers of participants (n) are reported unweighted. Percentages are row percentages. ‡ Associations were determined by logistic regression analyses with any NSSI, NSSI-T, and NSSI-B as dependent variables and socio-demographic variables as independent variable, adjusting for sex and age except for sex and age group.  with the number of episodes of suicidal ideation, and NSSI-B was associated with more episodes of a suicide plan. An early age at onset of plan was associated with NSSI-B.

Mental healthcare utilization
As shown in Table 4, 42.3% of those with any NSSI reported to have used any healthcare service for mental health problems at some point during their lifetime (NSSI-T: 42.7%; NSSI-B: 46.0%, Appendix Table S4). In those with any NSSI, most common was the use of an outpatient service (36.1%) including mental health specialist (32.3%) and/or psychotherapist (27.7%). Moreover, inpatient care services were used by 13.6%, for example, inpatient clinic (9.6%), but also complementary care services were used (10.6%) like professional counseling centers (7.0%). There was no significant difference in any healthcare use for the distinct groups of NSSI-T only, NSSI-B only and both. In those with any NSSI, there were no significant sex (OR = 1.4, 95% CI 0.7-2.8, P = 0.28) and age cohort (OR = 1.4, 95% CI 0.8-2.4, P = 0.20) differences in any mental healthcare utilization. Though, those with a middle, low, or other compared to high education status (OR = 2.6, 95% CI 1.4-4.7, P = 0.002) were more likely to use healthcare services. Similar results were found for NSSI-T and NSSI-B.

Discussion
Non-suicidal self-injury and suicidal behavior are common behaviors during adolescence and young adulthood, though a detailed description of their lifetime co-occurrence is lacking taking types of NSSI (thoughts and behaviors) and suicidal behaviors, that is, ideation, plan, and attempt into consideration. Using a random-community sample of adolescents and young adults aged 14-21 years from Dresden, Germany, results showed (1) high lifetime prevalence estimates for non-suicidal selfinjury including thoughts (18.0%) and behaviors (13.6%)-especially in females; (2) high co-occurrence of non-suicidal self-injury and suicidal behaviors with differences in females and males and an association of NSSI with the number of episodes of suicidal ideation and also plan for NSSI-B; and (3) insufficient mental healthcare utilization in those with non-suicidal self-injury with and without suicidal behavior.

NSSI prevalence and its socio-demographic correlates
Regarding the prevalence estimates, the results of the present study for NSSI behaviors fall in the lower range of prevalence estimates from previous national and international studies in adolescents and young adults (19,20). Though, prevalence estimates were higher compared to a study in adults from Germany (3.1%; 44). Nearly every fifth adolescent and young adult reported thinking about deliberately hurting themselves (18.0%) and 13.6% actually hurt themselves at least ones during their life. For NSSI thoughts, the present study found similar results to a study in college students aged 18-35 years (22.6%; 22), but lower rates compared to a regional school-based study in aged 15 year olds in Germany (24.1% past 6 months; 23). Females showed a prevalence estimate more than twice as high than males, which confirms previous results of a meta-analysis showing that females report 1.5-times more NSSI behaviors (21). More than every fourth girl/women (25.9%) thought about deliberately hurting herself and more than every fifth girl/women (20.6%) actually hurt herself at least ones during their life. Several explanations exist why NSSI is more common in females than males (21), even though the function as an emotion regulation approach seems to be equal (1). So far, the primary methodological/descriptive explanation is the use of different methods in females and males, similar to the findings of different method use for suicide attempts.
While females seems to more likely use cutting or scratching methods, males seems to more likely punch or burn themselves, or bang their heads (45,46), though results regarding punching and burning were not significant for males compared to females in a meta-analysis (21). In the present study, the examples 'cutting, scratching, or burning' were used, whereby males might have had less Data are weighted to refer to the age and sex distribution in the general population of individuals aged 14 to 21 years in Dresden, Germany. The numbers of participants are unweighted, and percentages are column percentages. Participants could endorse several response options. ‡ Associations between service use as the dependent variables and suicidal behavior as independent variable were determined in those with NSSI by logistic regression analyses, adjusting for sex and age. § If there were five or less cases in one group, the exact Fisher's test was used to test associations. chance to endorse the items resulting in lower prevalence estimates for males. Interestingly, the estimates for those showing NSSI-T only, and NSSI-B only did not differ by sex, while reporting both was more often in females than males. Current education status, subjective financial situation, growing up with a single parent and number of siblings were associated with NSSI. No association was found between citizenship, living arrangement, subjective social class, and urbanization. Previous studies found similar results regarding income (25), citizenship, education status and family compensation, though results stayed not significant in multivariate analyses (26). However, no overview exists so far combining the results regarding sociodemographic correlates of NSSI of previous studies. Like for other mental health problems during childhood and adolescence (47), lower socio-economic status measured by different indicators might be a general risk factor for NSSI, similar to findings for suicide behavior (48).

Co-Occurrence of NSSI and suicidal behavior
The present study assessed the lifetime co-occurrence of NSSI and suicidal behavior in different ways and directions. First, 23.2% reported any NSSI and/or suicidal behavior during life with 19.4% for NSSI-B and/or suicidal behavior, which shows that around every fifth individual in this age group is affected. These findings highlight the need for the assessment of suicidal behavior in individuals with any NSSI-regarding both thoughts and behaviors. The result for the prevalence of NSSI-B and/or suicidal behavior is marginally lower to a previous finding in a random sample of university students aged 18-24 from the US (24.9%; 27). Considering the distinct groups for NSSI-B, 5.7% reported suicidal behavior only, 7.8% NSSI-B only, and 5.8% reported both, which also differs only slightly from the above mentioned study in university students. Here, 7.9% reported suicidal behavior only, 10.2% reported NSSI-B only, and 6.9% reported both (27). Regarding the conditional probabilities, the current study showed that 39.6%-42.7% of those with NSSI reported also any suicidal behavior (ideation: 39.3%-42.0%, plan: 22.4%-24.9%, attempt: 11.8%-14.2%) and illustrated that suicidal ideation, plan, and attempt were significantly more common in those with NSSI compared to those without NSSI. In those with suicidal behavior, more than half of the individuals reported also NSSI. Moreover, 86.2% of those with a suicide plan reported NSSI thoughts. Results are comparable to results from a schoolbased sample of 15 year olds from Germany, where more suicidal behavior was reported in those with occasional and repetitive compared to no NSSI-B for suicidal ideation (43.3%, 81.9%, 7.6%), plan (20.5%, 55.4%, 2.5%), and attempt (26.1%, 55.0%, 3.3%), respectively (26). In a community sample of young adults ages 19-26 years from Mexico City, 48.6%, 18.7%, 25.2% of those with NSSI-B reported suicidal ideation, plan, or attempt, respectively (25). Overall, previous studies found higher rates for suicide attempts in those with NSSI-B compared to the results of the present study. Considering the other direction, previous studies found also higher rates for NSSI-B in those with suicidal behavior. NSSI-B was reported by 40.7% of those with suicidal ideation and 63.6% of those with attempt in the past 12 months (49) as well as by 56.6% of those with lifetime ideation, 61.9% plan, and 68.1% attempt (25). Different theoretical considerations exist how NSSI and suicidal behavior are linked with each other, for example, the Gateway Theory, the Third Variable Theory, or Joiner's Interpersonal Theory of Suicide (28). Mostly, etiological theories regarding suicidal behavior include NSSI as a risk factor for suicidal behavior (6,7). Beyond research confirming the role of NSSI as a risk factor for suicidal behavior (8)(9)(10)(11), there is also research indicating suicidal behavior as a risk factor for NSSI (9,50). Interestingly, besides sex and current education status none of the other examined socio-demographic variables were associated with the co-occurrence of both behaviors. Future studies need to address which risk factors contribute to a co-occurrence of both behaviors in contrast to those which lead to either of these behaviors.
So far, most studies examined the association of NSSI severity indicators and suicidal behavior. The present study adds valuable information about NSSI and its association with severity characteristics of suicidal behavior including age at onset and number of episodes or attempts. Here, number of episodes of suicidal ideation was associated with NSSI illustrating a higher number of episodes in those with NSSI compared to no NSSI. In addition, NSSI-B was associated with number of episodes of plan. Looking at age at onset, only an earlier onset of plan was more common in those with NSSI-B compared to those without NSSI-B. Results are in line with some previous studies showing higher levels of suicidal ideation in those with NSSI (51). No association was found with the number of suicide attempts in general. Previous studies found that frequency of NSSI was associated with the number of attempts, which was not analyzed in the present study (26,52), though results are mixed (27,53). Others found associations with the methods of NSSI and number of attempts (54).

Mental healthcare utilization
About 42% of those with any NSSI reported any lifetime healthcare utilization because of mental health problems. Rates were higher for those with any suicidal behavior compared to none (62.3% vs. 29.1%), especially for using a psychiatrist (21.0% vs. 2.8%). Similar rates were found for NSSI-T and NSSI-B. These findings are similar to previous studies assessing help-seeking behavior in adolescents with self-harming behavior (55,56). There are only a few studies comparing those with and without suicidal behavior showing more helpseeking in those with suicidal behavior (57) and more online help-seeking behavior (58). Several barriers for mental healthcare utilization were discussed so far like stigmatization, fear of confidentiality, or other negative consequences like hospitalization (55,56). Although a large proportion of those with NSSI identify their need for help (59), the present study underpins the insufficient use of mental healthcare services in this critical age group even if individuals have free access to mental health services like in the present sample.

Limitations
The present study assessed adolescents and young adults up to age 21. Therefore, all results are restricted to this age. Findings are also confined to a region (major city and capital of Saxony) in the Eastern part of Germany with a lowered proportion of migrants, but a similar population density and mean age to most larger cities in Germany (30). Sample weights were used to adjust for sex and age differences because of a higher participation rate in females. Overall, the participation rate was relatively low, which is a general trend in epidemiological studies (60-64) and might not necessary lower the validity of the findings (62).The BeMIND study was not specifically designed to address the questions of this analysis. Yet, full information regarding the reliability of results can always be drawn from the given confidence intervals.
An underestimation of the prevalence estimates might be the case in the present study because of a smaller number of participants with a lower education status, and the exclusion of currently institutionalized individuals (e.g., hospitalization), because previous studies indicated more NSSI in those with lower education status (26) and in nonresponders (65,66). In addition, a recall bias might have lowered the prevalence estimates because of the retrospective assessment of NSSI and suicidal behavior including its onset (38,(67)(68)(69). As the age range was 14 to 21, data are right censored indicating that some individuals will develop either of these behaviors until age 21 which leads to an underestimation of the prevalence. Also, face-toface assessment might have resulted in fewer reports of suicidal behavior (24,70). The assessment of NSSI with single items also resulted in lower prevalence estimates in previous studies (19) and the binary assessment showed differences in the association with risk factors compared to a dimensional assessment (50). However, in the present study, NSSI needed to be presented only ones during life which might have overestimated the group of individuals in need for help. So far, it is unknown, when NSSI is starting to be harmful (50). Considering the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) research criteria for NSSI, a frequency of NSSI on five or more days in the past 12-months is understood as relevant (71), data driven approaches found significant difference in severity between five and six (72), while others found even higher cut-offs (73). Others found curvilinear associations between frequency of NSSI and co-occurring suicidal behavior indicating no clear picture (74). The items assessing NSSI are from a validated questionnaire, though only two items were used in the present study. Regarding mental healthcare utilization, participants were asked for any lifetime service use because of mental health problems but not particularly because of or following NSSI or suicidal behavior, which has to be considered when interpreting the results. The present study only looked at cross-sectional associations between NSSI and suicidal behavior; therefore, no causal inferences can be drawn.
To conclude, the lifetime co-occurrence of NSSI and suicidal behavior in adolescence and young adulthood is frequent with females being more commonly affected by both behaviors than males. To guide targeted measures for early detection and intervention, future studies should examine temporal associations and interactions of both behaviors. Real life studies, for example, by using smartphone-based ecological momentary approaches, appear as valuable avenue to gain more insights into the short-term dynamics of co-occurrence as well as its predictors. The identification of underlying (overlapping) risk factors and mechanisms driving these behaviors like genetic variation (75), reward seeking (76,77), or stress (78,79) is crucial to improve etiopathogenetic models and to derive distal and proximal targets for prevention.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1 Association of the co-occurrence of NSSI and any suicidal behavior with sex, age group, current education status (N = 1180). Table S2 Conditional probabilities of NSSI and suicidal behavior. Table S3 Prevalence of any suicidal behavior by any NSSI, NSSI-T and NSSI-B and sociodemographic characteristics. Table S4 Mental health care utilization in those with NSSI-T and NSSI-B and co-occurring suicidal behavior.