Available online at www.sciencedirect.com

ScienceDirect Comprehensive Psychiatry xx (2015) xxx – xxx www.elsevier.com/locate/comppsych

Non-suicidal self-injury and suicidal ideation as predictors of suicide attempts in adolescent girls: A multi-wave prospective study Lori N. Scott a,⁎, Paul A. Pilkonis a , Alison E. Hipwell a , Kate Keenan b , Stephanie D. Stepp a a University of Pittsburgh School of Medicine, Department of Psychiatry, 3811 O’Hara Street, Pittsburgh, PA 15213, USA University of Chicago, Department of Psychiatry Behavioral Neuroscience, 5841 South Maryland Avenue, Chicago, IL 60637, USA

b

Abstract Although both suicide ideation (SI) and non-suicidal self-injury (NSSI) are known risk factors for suicidal behavior, few longitudinal studies have examined whether having a history of one or both of these factors prospectively predicts increased risk for suicide attempts. According to the theory of acquired capability for suicide, engagement in NSSI may reduce inhibitions around self-inflicted violence, imparting greater risk for suicide attempts among those with SI than would be observed in those with SI who do not have a history of NSSI. We used prospective data from the Pittsburgh Girls Study, a large community sample, to compare groups of girls reporting no SI or NSSI, SI only, or both NSSI and SI between early to late adolescence on any lifetime or recent suicide attempts in late adolescence and early adulthood. As compared to girls with no SI or NSSI history and those with only an SI history, girls with a history of both NSSI and SI were significantly more likely to subsequently report both lifetime and recent suicide attempts. Results are consistent with the acquired capability theory for suicide and suggest that adolescent girls who have engaged in NSSI and also report SI represent a particularly high-risk group in need of prevention and intervention efforts. © 2014 Elsevier Inc. All rights reserved.

Suicide is the third leading cause of death among adolescents and young adults [1]. Although rates of completed suicide are consistently higher in males [2], females attempt suicide two to three times more often than males [3,4]. Suicide ideation (SI), defined as thinking about or planning to engage in behaviors with the intent to end one’s life, strongly predicts suicide attempts [5]. SI tends to first emerge during adolescence, and the prevalence of SI is higher among adolescents compared to all other age groups [6], with especially high rates of SI among girls [5]. In addition, SI has been shown to significantly predict suicide attempts among psychiatrically hospitalized adolescent girls but not boys [7]. Thus, the prevention of SI may lead to significant decreases in morbidity and mortality for This research and the efforts of the authors were supported by grants from the National Institute of Mental Health (R01 MH101088, PI: Stepp; R01 MH56630, PI: Loeber; R01 MH056888, PI: Pilkonis; K01 MH101289, PI: Scott), the National Institute on Drug Abuse (R01 DA012237, PI: Chung), and by funding from the Office of Juvenile Justice and Delinquency Prevention, the FISA Foundation and the Falk Fund. ⁎ Corresponding author at: Western Psychiatric Institute and Clinic, 3811 O’Hara St., Pittsburgh, PA 15213. Tel.: +1 412 383 5016; fax: +1 412 383 5068. E-mail address: [email protected] (L.N. Scott). http://dx.doi.org/10.1016/j.comppsych.2014.12.011 0010-440X/© 2014 Elsevier Inc. All rights reserved.

adolescent girls. Data regarding SI and suicide attempts in atrisk female community samples, however, are scarce, but are crucial for identifying those who are at highest risk. Furthermore, although an estimated 30% of adolescents report having thought of suicide at some point, only a small fraction of those who engage in SI ever attempt suicide [5]. Hence, identifying factors that increase the likelihood of the progression from ideation to suicide attempts is of utmost importance to prevention efforts. Recent theoretical and empirical work suggests that individuals with SI who engage in non-suicidal self-injury (NSSI), defined as direct and deliberate destruction of one’s own body tissue without suicidal intent [8], are at increased risk for suicide attempts [9–12]. NSSI occurs in as many as 13%–21% of community-based adolescents [13–15] and is a strong predictor of suicide attempts above and beyond other risk factors, including SI, depression, borderline personality disorder, family problems, and trauma or abuse exposure [9,16]. The age of onset of NSSI in community samples is in early adolescence, between the ages of 11 and 15 [15,17,18]. Some evidence suggests that NSSI is more prevalent and has an earlier age of onset among adolescent girls than among boys [16].

2

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

Given the developmental emergence of NSSI and SI, it is not surprising that the two often co-occur during adolescence [19–21], and overlap between the two likely increases the risk for suicidal behavior. According to the theory of acquired capability for suicide [22], one must not only desire to end one’s life but must also overcome the fear and pain associated with self-inflicted harm to increase one’s capability for engaging in suicide-related behaviors. NSSI may increase the capability for self-inflicted violence via desensitization to the fear and pain associated with these behaviors. Thus, those with a history of both NSSI and SI may be at greatest risk for suicidal behavior as compared to those with neither or only one of these risk factors. Evidence supporting the acquired capability theory has recently begun to emerge [16]. For example, higher frequency of NSSI is associated with more lethal suicide attempts [23], and a longer duration of engagement in NSSI is associated with more frequent suicide attempts [21]. Studies have also shown that adolescents with a history of both attempted suicide and NSSI report less fear about engaging in suicidal behavior than those with suicide attempts and no history of NSSI [24,25], suggesting that NSSI may desensitize individuals to the fear associated with suicide [16]. In addition, results of a recent meta-analysis examining multiple risk factors for suicide attempts found that a combination of SI and NSSI was strongly associated with history of suicide attempts [26]. These data, however, were based mostly on retrospective studies of individuals with histories of NSSI and/or past suicidal behavior, and we could not find any published studies that have prospectively examined both NSSI and SI in predicting future suicidal behavior. To aid in the identification of youth who are at greatest risk for future suicide attempts and most in need of early intervention, it is important to examine the degree to which the co-occurrence of NSSI with SI in adolescence prospectively predicts risk for suicide attempts. Accordingly, the current study utilizes prospective data from a large and diverse community sample of adolescent girls to examine the combination of NSSI and SI during early to late adolescence as a correlate of lifetime suicidal behavior and a predictor of subsequent suicide attempts. Girls with “no SI or NSSI”, and those with histories of “SI only” or “both SI and NSSI” from age 10 through adolescence were compared in their propensity for reporting lifetime and recent (past year) suicide attempts in late adolescence and early adulthood. In accordance with previous studies showing high rates of SI among those who self-injure [19,27], there was not sufficient representation of “NSSI only” (i.e., NSSI without any history of SI; n = 21) in this sample to examine this subgroup. To determine the unique risk for future suicidal behavior associated with combined SI and NSSI histories, we controlled for several other known risk factors for suicide attempts as demonstrated in previous studies [16], including various types of family adversity and trauma, depression, and borderline personality disorder (BPD) symptoms. We hypothesized that although those with “SI only” would be

more likely to attempt suicide than those without SI or NSSI, girls with histories of both SI and NSSI would demonstrate the highest rates of reported lifetime and recent suicide attempts. In other words, we expected that the combination of NSSI with SI would impart significantly greater risk for suicide attempts than would SI alone or having no history of SI or NSSI. 2. Method 2.1. Participants and sample description The Pittsburgh Girls Study (PGS; N = 2450) is a highrisk urban community sample of four cohorts of girls, ages 5–8 at the first assessment year, and their primary caretaker, who have completed annual assessments according to an accelerated longitudinal design. The study sample was identified by oversampling low income neighborhoods, such that neighborhoods in which at least 25% of families were living at or below poverty level were fully enumerated and a random selection of 50% of households in all other neighborhoods was enumerated [28,29]. The current analyses include all four age cohorts, covering ages 10–21 years (see Table 1). NSSI and SI data were collected from early adolescence, starting with age 10 for SI and age 13 for NSSI, through the 12th assessment wave (when girls were ages 16– 19). Suicide attempts were assessed in the 13th and 14th annual assessment waves (when girls were ages 17–20 and 18–21, respectively). All data for this study were collected in calendar years 2003 through 2014. We included girls who had sufficient data to run the analyses, which required having completed the wave 12 assessment (when the covariates, depression and BPD symptoms, were assessed), at least one assessment of SI and NSSI at any time from ages 10 through wave 12, and at least one assessment of suicide-related behavior in either wave 13 or 14 (N = 1971). There are no known completed suicides in the PGS. From this sample, and as we expected based on prior studies [19,27], only a very small percentage (1.1%; n = 21) endorsed a history of NSSI without endorsing any history of SI. To minimize noise that may be potentially introduced by these outliers, these cases were excluded from the final analyses (final N = 1950). However, all substantive results and conclusions remained unchanged if these cases were included. As compared to those who were excluded (total n = 500) due to missing data (n = 479) or being in the NSSI only group (n = 21), those who were included in the analyses (n = 1950) were more likely to identify as minority race (χ 2 = 23.08, df = 1, p b .001) and to report having received public assistance (χ 2 = 11.98, df = 1, p b .001) and living in a single parent household (χ 2 = 7.028, df = 1, p = .008) in the first assessment wave. Those who were included (n = 1950) did not differ from those who were excluded but provided suicide attempt data in waves 13 or 14 (n = 248) in likelihood of reporting lifetime (χ 2 = 0.44, df = 1, ns) or recent (χ 2 = 1.40, df = 1, ns) suicide attempts. The majority

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

3

Table 1 Age (in years) by assessment wave and cohort in the analyzed sample (N = 1950). Assessment wave (w) Age cohort (n)

w1

w2

w3

w4

w5

w6

w7

w8

w9

w10

w11

w12

w13

w14

Cohort 5 (n Cohort 6 (n Cohort 7 (n Cohort 8 (n

5 6 7 8

6 7 8 9

7 8 9 10

8 9 10 11

9 10 11 12

10 11 12 13

11 12 13 14

12 13 14 15

13 14 15 16

14 15 16 17

15 16 17 18

16 17 18 19

17 18 19 20

18 19 20 21

= = = =

473) 505) 491) 481)

All predictors were assessed prior to and including wave 12, and outcomes (suicide attempts) were assessed at waves 13 and 14. The early adversity variable includes measures administered from wave 1 (when participants were ages 5–8 years) through age 16. Suicide ideation (SI) and non-suicidal self-injury (NSSI) were assessed beginning at ages 10 and 13, respectively, through wave 12 (when participants were ages 16–19 years). Borderline personality disorder (BPD) symptoms and depression severity were assessed in wave 12.

of the analyzed sample identified as African American (55.7%, n = 1086) followed by Caucasian (38.8%, n = 757), multi-racial (4.9%, n = 96), and Asian (0.6%, n = 11). At the wave 1 assessment, the majority of caretakers were female (93.2%); 57.3% were cohabiting with a spouse or domestic partner; 52.5% completed more than 12 years of education; and 40.6% of families received public assistance in the form of WIC, food stamps, and/or welfare. 2.2. Measures 2.2.1. Suicidal ideation and non-suicidal self-injury SI in the past year was assessed using girls’ reports on the Child Symptom Inventory-4th edition (CSI-4) [30] starting when girls were 10 years old (i.e., “Have you thought about death or suicide?”). The item is scored on a four-point scale (0 = never; 1 = sometimes; 2 = a lot; 3 = all the time). When girls reached adolescence (age 12) and adulthood (age 18), SI was assessed using the same item from the Adolescent Symptom Inventory-4th edition (ASI-4) [31] and the Adult Self-Report Inventory-4 (ASRI-4) [32], respectively. In order to determine a history of SI during adolescence, we calculated SI as present if girls endorsed having SI at least “sometimes” in the past year during any assessment year from age 10 through the wave 12 assessment, when girls were 16 (cohort 5) to 19 (cohort 8) years old. A total of 779 girls (39.9%) in the analyzed sample reported engaging in SI at some time by age 19, with the majority (66.1%) having onset prior to age 13. Most girls who reported SI only did so for a total of one (57.6%) or two years (22.8%) across the examined waves. To examine group differences in SI severity, we also calculated the individual mean of SI severity across all assessments. Girls reported on the occurrence of NSSI in the past year (i.e., “Did you hurt yourself on purpose, like cutting or burning, even if you did not mean to die?”) in response to items adapted from the Structured Clinical Interview for DSM Disorders, Research Version, Non-patient Edition (SCID-I) [33]. At the first annual assessment of NSSI (age 13), lifetime prevalence was also assessed. Girls who endorsed NSSI in the past year (0 = no and 1 = yes) were asked to describe the method they used to engage in the behavior (“In the past year, what kinds of things have you

done to hurt yourself?”). Descriptions of NSSI methods included cutting with a knife or razor (more than 85%) and other severe forms of NSSI such as burning, head banging, and choking. In order to categorize girls according to a reported history of NSSI during adolescence, we calculated NSSI as present if girls endorsed having engaged in NSSI (0 = no and 1 = yes) at any time in their lifetime at the age 13 lifetime assessment or in any assessment year from ages 13 through the wave 12 assessment, when girls were 16 (cohort 5) to 19 (cohort 8) years old. After excluding the 21 girls who engaged in NSSI only with no history of SI, a total of 170 girls (8.7% of the analyzed sample) were included who reported engaging in NSSI at some time by age 19, with the majority (74.1%) having onset by age 15. Most girls who reported NSSI only did so for a total of one (75.3%) or two years (16.5%) across the examined waves. A history of NSSI was significantly correlated with history of SI (r = .38, p b .001) across the entire examined trajectory, with correlations in any given year ranging from r = .26 to .42 (all p’s b .001). Based on the above-described measures of lifetime history of SI and NSSI through wave 12, we created an “SI group” variable as follows: 1) girls who had no lifetime history of reported SI or NSSI (No SI or NSSI, n = 1171); 2) girls who reported SI at any time and no NSSI (SI only, n = 609); and 3) girls who reported both NSSI and SI (SI + NSSI, n = 170). As noted previously, girls who reported only NSSI without history of SI were rare in this sample (1.1%; n = 21) and were therefore not included in the final analyses. 2.2.2. Suicide attempts Suicide attempts were assessed using items from the Suicide Behaviors Questionnaire-Revised (SBQR) [34], which was administered in the waves 13 and 14 assessments when girls were 17–21 years old. Both waves 13 and 14 were included due to the low base rates of suicide attempts in any one assessment year. Lifetime suicide attempts were assessed based on responses from one item rated on a sixpoint scale, with scores of 5 (I have attempted to kill myself, but did not want to die) and 6 (I have attempted to kill myself and really hoped to die) indicating having attempted suicide. We also included an item to assess the frequency of suicide

4

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

attempts in the last year on a four-point scale (1 =never; 4 = often (3–4 times)). The majority of girls in the analyzed sample completed the SBQR in both waves (95.2%; n = 1877). For the remaining girls who only completed one of the two SBQR assessments (4.8%; n =94), we simply used their item scores from the completed assessment. Twenty-six girls reported making at least one suicide attempt in the past year in the wave 13 assessment, and 17 girls reported a suicide attempt in the past year in wave 14. Among girls who provided data in both waves 13 and 14, five reported having attempted suicide in both years. Due to the low rates of suicide attempts in any one assessment year, however, we were unable to examine models in which wave 14 suicide attempts were predicted while controlling for wave 13 suicide attempts. To make our analyses feasible, we combined suicide attempts across waves 13 and 14 into a single variable to summarize suicide attempts during both years. We created a dichotomous variable for lifetime suicide attempts reported by wave 14, as well as a dichotomous variable for recent suicide attempts based on the report of at least one past year suicide attempt in either wave 13 or 14 (0 = no attempt; 1 = 1 or more attempts). 2.2.3. Depression severity Past-year depression severity was assessed with girls' reports at wave 12 using the ASI-4 [31] for girls under age 18 (cohorts 5 and 6) and the ASRI-4 [32] for girls over 18 (cohorts 7 and 8). The ASI-4 and ASRI-4 include DSM-IV (American Psychiatric Association, 1994) symptoms of depression scored on a four-point scale (0 = never to 3 = very often). Adequate concurrent validity, sensitivity, and specificity of depression severity scores to clinicians’ diagnoses have been reported [30–32]. A depression severity score was calculated based on the sum of depression items. To reduce overlap between depression symptoms and SI history, we excluded one item corresponding to suicidal ideation from calculation of depression severity scores. The depression severity scores had good internal consistency values (α = .87 (10 items) for the ASI-4, and α = .88 (9 items) for the ASRI-4). 2.2.4. Borderline personality disorder symptom severity Past-year BPD symptoms were assessed with girls’ reports at wave 12 using questions from the screening questionnaire of the International Personality Disorders Examination (IPDE-BOR) [35]. Although the IPDE screening questionnaire was originally developed for adults, this questionnaire has been validated for use in adolescent samples [36–38] and found to have adequate concurrent validity, sensitivity, and specificity to clinicians’ diagnosis in a sample of youth, with a score of 4 or greater considered to be clinically significant [39]. The IPDE-BOR consists of nine items (e.g., “I get into very intense relationships that don’t last”) rated either true or false (scored 1 or 0, respectively). An additional item measuring identity confusion was added: “Feelings about myself change frequently.”

Items were summed to yield a dimensional BPD symptom severity score. To avoid overlap with other variables in the analysis, the item referring to suicidality or self-injury, “I’ve never threatened suicide or injured myself on purpose,” was excluded from the calculation of the total BPD severity score. One additional item, “I show my feelings for all to see,” reduced the internal consistency alpha to .66, and was therefore dropped from calculation of BPD symptom totals. The internal consistency for the BPD symptom score was adequate, α = .71 (8 items).

2.2.5. Early adversity A composite measure of early adversity comprised prospectively gathered annual parent and child reports from wave 1 (when girls were ages 5–8) through age 16. Measures used to compute early adversity included caregivers’ reports on the Difficult Life Circumstances Scale (DLC) [40], single parent status (0 = cohabitating; 1 = single parent household), and family poverty (0 = not receiving public assistance; 1 = receiving public assistance) in each assessment year from wave 1 through age 16, as well as child reports of traumatic experiences from ages 10 to 16, and parent and child reports of sexual assault or abuse of the child from ages 12 to 16. The DLC is a 28-item questionnaire designed to assess life stressors and chronic family problems (e.g., regular arguments or conflicts with family members, partner absence or incarceration, lack of privacy or crowding in the home). Each item was coded yes (1) or no (0), and items were summed to yield an overall scale score for each assessment year. DLC scores in each year were recoded as 1 if the participant was in the top quartile of DLC scores in that year, and as 0 if the participant was not in the top quartile. To assess traumatic experiences, girls were administered the Child PTSD Symptom Scale (CPSS) [41] at ages 10–16 to assess whether they experienced a traumatic event that was “really terrifying and scary” such as “you or someone you love nearly killed or badly hurt”; “being in a fire, flood or tornado”; or “seeing someone killed, murdered, or badly beaten”. To assess experiences of sexual abuse, parents were asked one yes/no question regarding whether their child experienced any type of sexual assault or abuse in the last year, and girls were asked four yes/no questions regarding whether they experienced different forms of sexual abuse in the last year from ages 12 to 16. If any of these five questions were endorsed, the girl received a “1”, and if no items were endorsed, the girl received a “0” in that assessment year. The overall early adversity index was calculated based on the average of these categorical items (i.e., single parent status, family poverty, DLC top quartile, experienced trauma, and reported sexual abuse in each year) across time (Wave 1, or earliest year of assessment of a given construct, through age 16) for each participant in order to capture their overall level of exposure to cumulative adversity on a scale ranging from 0 to 1 (α = 0.90). Similar approaches to quantifying cumulative risk have been used in previous studies to

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

combine multiple risk factors into a single construct that is parsimonious and statistically sensitive [42]. 2.3. Procedure Annual interviews were conducted in the home separately for both the girl and caretaker by trained interviewers using a laptop computer. All study procedures were approved by the University Institutional Review Board. Parents provided consent and girls provided assent for study participation. Families were compensated for their participation. At the outset of the interview, interviewers informed the girl that all information would be kept confidential unless reports of harm to self or others were endorsed. In the event of reported SI or NSSI, interviewers talked with the caregiver and the on-call psychologist who assessed for seriousness and made referrals when appropriate. 2.4. Data analytic plan First, we calculated descriptive statistics and examined group differences in each of the variables using univariate analysis of variance (ANOVA) for continuous variables and chi-square tests for categorical variables. Next, we conducted two binary logistic regressions with “lifetime suicide attempt” reported by wave 14 and “recent suicide attempt” reported in the past year during wave 13 or 14 as the dependent variables, covariates (race, adversity, depression, and BPD symptoms) entered in Step 1, and SI group entered in Step 2. The SI group variable was specified as categorical with the “no SI or NSSI” group first specified as the reference group for simple contrasts. To examine potential differences in odds ratios between the “SI only” and “SI + NSSI” groups, logistic regressions were then repeated with the “SI + NSSI” group specified as the reference group. 3. Results Descriptive statistics and preliminary analyses are presented in Table 2. A total of 63 participants (3.2%)

5

reported a lifetime suicide attempt by wave 14, and 36 of the 63 (57.1%) participants reported a recent suicide attempt (in waves 13 or 14). In preliminary ANOVA models, Tamhane’s T2 post hoc tests (equal variances not assumed) demonstrated that the “SI only” group reported more adversity than did the “no SI or NSSI” group. In addition, the “SI only” and “SI + NSSI” groups both reported significantly more depression severity, BPD symptoms, and SI severity than the “no SI or NSSI” group, and the “SI + NSSI” group reported more depression severity, BPD symptoms, and SI severity than the “SI only” group. Due to shared variance between SI severity and SI group membership, we conducted the primary analyses without SI severity included as a covariate, and then repeated analyses with this variable as a covariate to determine if its inclusion substantially changed the results. Further, chi-square tests revealed that girls who reported “SI only” were significantly more likely to identify as minority race than Caucasian, whereas, those who reported “No SI or NSSI” and “SI + NSSI” were more likely to identify as Caucasian than minority race. Preliminary analyses also demonstrated that girls in the “no SI or NSSI” group were significantly less likely to report lifetime or recent suicide attempts. Conversely, those in the “SI + NSSI” group were significantly more likely than not to report a lifetime or recent suicide attempt. These group differences are further explored in the primary logistic regression models described below. 3.1. Logistic regression predicting lifetime suicide attempts As hypothesized, both “SI only” and “SI + NSSI” groups had greater odds of reporting a lifetime suicide attempt than those in the “no SI or NSSI” group, with the “SI + NSSI” group showing the greatest odds of a suicide attempt (see Table 3). Specifically, as compared to girls in the “no SI or NSSI group”, those in the “SI only” group were about three times more likely to make a lifetime suicide attempt by wave 14, whereas, those in the “SI + NSSI” group were about 12.6 times more likely than those with “no SI or NSSI” to make a

Table 2 Descriptive statistics for analyzed sample (N = 1950). SI Group Observed Range Continuous Variables Early Adversity Index (wave 1, age 16) Depression Severity (wave 12) BPD Severity (wave 12) SI Severity (mean SI severity, ages 10 through wave 12) Dichotomous Variables (0 = no, 1 = yes) Minority Race Lifetime suicide attempts (by wave 14) Recent (past year) suicide attempts (reported in waves 13 or 14)

0–0.75 6–33 0–8 1.00–2.90 0/1 0/1 0/1

no SI or NSSI (n = 1171)

SI only (n = 609)

M 0.26a 11.36a 1.20a 1.00a N 686 11 10

M 0.31b 13.78b 2.22b 1.21b N 422 24 15

SD 0.18 4.55 1.64 0.00 % 58.6 0.9 0.9

F/χ 2 test statistic (df)

SI+NSSI (n = 170) SD 0.18 5.14 1.99 0.15 % 69.3 3.9 2.5

M 0.29 15.61c 2.84c 1.35c N 85 28 11

SD 0.19 5.43 2.05 0.26 % 50.0 16.5 6.5

F (2,1947) 17.74*** 89.16*** 70.30*** 1125.29*** χ 2 [2] 29.16*** 115.97*** 27.70***

Row means for continuous variables with different subscripts are significantly different from each other at p b .05 or less using Tamhane's T2 post-hoc comparisons. Significant χ 2 values for dichotomous variables indicate differences in likelihood of identifying as minority race or reporting suicide attempts for those in each group (results further described in the text). *p b .05. **p b .01. ***p b .001.

6

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

Table 3 Logistic regression analysis predicting girls’ reported lifetime suicide attempts by wave 14. Variable Step 1 Minority Race Early Adversity Index Depression Severity BPD Severity Step 2 Minority Race Early Adversity Index Depression Severity BPD Severity SI Group SI only (n = 609) SI + NSSI (n = 170)

SE B

Wald χ 2

df

p

0.14 0.68 0.07 0.33

0.31 0.80 0.03 0.07

0.20 0.73 6.70 22.87

1 1 1 1

.66 .39 .01 b.001

1.15 1.98 1.07 1.39

0.63–2.08 0.41–9.48 1.02–1.12 1.21–1.59

−0.03 0.49 0.03 0.27

0.32 0.81 0.03 0.07

.92 .55 .27 b.001 b.001 .003 b.001

0.52–1.79 0.33–7.99 0.98–1.09 1.14–1.51

0.38 0.39

1 1 1 1 2 1 1

0.97 1.63 1.03 1.31

1.11 2.54

0.01 0.37 1.20 14.08 47.19 8.59 42.84

3.04 12.63

1.45–6.38 5.91–27.00

B

Odds Ratio

95% CI

N = 1950. Reference for SI group comparisons is the “No SI or NSSI” group (n = 1171). Supplemental analysis with “SI + NSSI” as the reference group revealed that those in the “SI only” group were significantly less likely to report a lifetime suicide attempt than were those in the “SI + NSSI” group, B = −1.43, SE = .31, Wald χ 2 [1] = 21.69, p b .001, OR = 0.24, 95% CI [.13, .44].

lifetime suicide attempt by wave 14. To compare the “SI only” and “SI + NSSI” groups, we re-ran the same analysis with the “SI + NSSI” group specified as the reference group, which revealed that those in the “SI only” group were significantly less likely to report a lifetime suicide attempt than were those in the “SI + NSSI” group, B = −1.43, SE = .31, Wald χ 2 [1] = 21.69, p b .001, OR = 0.24, 95% CI [.13, .44]. Although both depression and BPD symptoms predicted a greater likelihood of a lifetime suicide attempt in Step 1 after controlling for race and adversity, only BPD symptoms remained significant in Step 2 after controlling for race, adversity, depression severity, and histories of SI and NSSI.

3.2. Logistic regression predicting recent suicide attempts Next, to examine a purely prospective model without potential temporal overlap between predictors and outcomes, we conducted another logistic regression with recent suicide attempt (i.e., at least one past year attempt reported in either wave 13 or 14) as the dependent variable. As in the previous analysis, covariates (race, adversity, depression, and BPD symptoms) were entered in Step 1, and SI group was entered as a categorical variable in Step 2. Results of this analysis are presented in Table 4. Results revealed that those in the “SI + NSSI” group were about 4.4 times more likely to report a recent suicide attempt than those in the “no SI or NSSI” group, whereas, girls with SI only were not significantly more likely than those in the “no SI or NSSI” group to report a recent suicide attempt. To compare the “SI only” and “SI + NSSI” groups, we re-ran the analysis with “SI + NSSI” as the reference group, which revealed that those in the “SI only” group were significantly less likely to report a recent suicide attempt than were those in the “SI + NSSI” group, B = −0.88, SE = .42, Wald χ 2 [1] = 4.27, p = .04, OR = 0.42, 95% CI [.18, .96]. In addition, BPD symptoms predicted greater odds of a recent suicide

attempt after controlling for race, adversity, depression severity, and histories of SI and NSSI. 3.3. Supplemental analyses controlling for SI severity Given that the “SI only” and “SI + NSSI” groups significantly differed from one another in SI severity, we repeated the above analyses with SI severity as a covariate while specifying “SI + NSSI” as the reference group to determine if the same differences between the “SI only” and “SI + NSSI” groups would hold after controlling for SI severity. After controlling for SI group, SI severity significantly predicted lifetime suicide attempts, B = 1.55, SE = .66, Wald χ 2 [1] = 5.53, p = .02, OR = 4.72, 95% CI [1.29, 17.19], but SI severity did not significantly predict recent suicide attempts, B = .38, SE = .85, Wald χ 2 [1] = 0.20, p = .66, OR = 1.46, 95% CI [0.28, 7.73]. In addition, the SI group differences for predicting lifetime suicide attempts remained unchanged; i.e., even after controlling for the significant influence of SI severity, those in the “SI only” group were significantly less likely to report a lifetime suicide attempt than were those in the “SI + NSSI” group, B = −1.23, SE = .32, Wald χ 2 [1] = 14.51, p b .001, OR = 0.29, 95% CI [.16, .55]. The only minor difference in statistical significance of findings emerged in the analysis of recent suicide attempts; i.e., after controlling for SI severity (which did not significantly predict recent suicide attempts), those in the “SI only” group were only marginally less likely to report a recent suicide attempt than were those in the “SI + NSSI” group, B = −0.82, SE = .44, Wald χ 2 [1] = 3.39, p = .07, OR = 0.44, 95% CI [.19, 1.05]. However, the magnitude of the difference between the “SI only” and “SI + NSSI” groups in likelihood of recent suicide attempts was similar to that reported in the above analysis when not controlling for SI severity, and given the nonsignificant relation between SI severity and recent suicide attempts, this minor difference in significance of the group comparison is likely due to reduced power from adding an

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

7

Table 4 Logistic regression analysis predicting girls’ reported recent suicide attempts in waves 13 or 14. Variable Step 1 Minority Race Early Adversity Index Depression Severity BPD Severity Step 2 Minority Race Early Adversity Index Depression Severity BPD Severity SI Group SI only (n = 609) SI + NSSI (n = 170)

B

SE B

Wald χ 2

df

p

Odds Ratio

95% CI

−0.30 0.77 0.04 0.39

0.43 1.05 0.03 0.09

0.48 0.53 1.65 19.30

1 1 1 1

.49 .47 .20 b.001

0.74 2.15 1.04 1.47

0.32–1.72 0.27–17.01 0.98–1.11 1.24–1.75

−0.38 0.68 0.02 0.35

0.43 1.05 0.03 0.09

.38 .52 .55 b.001 .007 .15 .002

0.29–1.60 0.25–15.31 0.95–1.09 1.19–1.69

0.43 0.48

1 1 1 1 2 1 1

0.69 1.97 1.02 1.42

0.62 1.49

0.77 0.42 0.37 15.15 9.88 2.05 9.71

1.85 4.43

0.80–4.29 1.74–11.31

N = 1950. Reference for SI group comparisons is the “No SI or NSSI” group (n = 1171). Supplemental analysis with “SI + NSSI” as the reference group revealed that those in the “SI only” group were significantly less likely to report a recent suicide attempt than were those in the “SI + NSSI” group, B = −0.88, SE = .42, Wald χ 2 (1) = 4.27, p = .04, OR = 0.42, 95% CI [.18, .96].

additional variable to the model rather than to any significant variance in recent suicide attempts being explained by SI severity. Thus, the substantive results and conclusions do not appear to be explained by group differences in SI severity.

4. Discussion The primary goal of this study was to examine histories of SI only and combined NSSI and SI during early to late adolescence as prospective predictors of both lifetime and recent suicide attempts reported in late adolescence and early adulthood. As hypothesized, and after controlling for other known risk factors for suicide (early adversity, depression, and BPD symptoms), girls with a history of SI only were more likely to report having made a lifetime suicide attempt by late adolescence than those without SI or NSSI, but the addition of NSSI to a history of SI significantly elevated girls’ risk for suicide attempts. Specifically, girls with histories of both SI and NSSI were significantly more likely to report lifetime suicide attempts as compared to girls with either SI only or no NSSI or SI. In addition, analysis of recent suicide attempts (i.e., attempts that reportedly occurred during waves 13 or 14 of data collection, in the two years after the final assessment of all predictors) revealed that girls with a history of both SI and NSSI during adolescence were significantly more likely than those with SI only or those with no history of SI or NSSI to make a suicide attempt in late adolescence/early adulthood. In contrast to the girls with a history of both SI and NSSI, those with a history of SI only were not significantly more likely than those without a history of either SI or NSSI to make a suicide attempt in late adolescence/early adulthood. Importantly, this is the first longitudinal study to demonstrate that engagement in NSSI during adolescence appears to heighten risk for future suicidal behavior above and beyond the risks associated with SI and other

environmental and psychiatric risk factors. The differences in results for lifetime versus recent suicide attempts (i.e., past-year attempts that occurred only in waves 13 or 14) highlight the importance of examining factors that increase risk of future suicide attempts using longitudinal designs. Our results indicate that SI only, without a history of NSSI, may be associated with a history of suicide attempts, but is not necessarily a strong predictor of future suicide attempts unless there is a combined history of both NSSI and SI. The finding that a combination of NSSI and SI imparts the greatest risk for suicide attempts is consistent with Joiner’s [22] theory of acquired capability for suicide, which suggests that suicidal behavior results not only from the desire to die but also from desensitization to fear and pain associated with self-inflicted violence. An important step for future studies is to examine whether NSSI episodes precede increases in severity or frequency of suicidal behavior using intensive repeated measures designs, as well as to investigate the putative mechanisms by which NSSI may increase risk for suicide attempts (e.g., whether NSSI reduces inhibitions associated with suicidal behavior, thereby increasing risk for subsequent suicide attempts). A history of SI was relatively common in this sample, as nearly 40% of the sample reported having had at least some thoughts of suicide at some time during adolescence. This estimate is somewhat higher than the point prevalence estimates of 15%–25% and lifetime estimates of up to 30% reported in other adolescent community samples [5,43]. This difference may stem from the fact that, because we were interested in whether any history of SI at all (rather than severity or chronicity of SI) predicted suicide attempts, girls who reported only having had thoughts of suicide “sometimes” as well as girls who may have only reported transient SI during one or two out of as many as nine consecutive assessment waves were included in our assessment of lifetime history of SI. Interestingly, however, in supplemental analyses (not presented here) using stricter cutoff criteria

8

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

(SI endorsed at least “often” in any assessment year, which was the case in 6.2% of the sample), the results for suicide attempts reported here were unchanged. It is possible that a history of more persistent SI alone (i.e., endorsement of SI over the course of multiple years), even without a history of NSSI, may be more strongly predictive of future suicide attempts than more transient SI. However, because most of the girls in this community sample only endorsed transient SI, this question will need to be addressed in future studies with samples endorsing more chronic suicidality than in the current sample. On the other hand, the rates of reported history of NSSI (i.e., 9.7% of the sample when including those with NSSI only and no history of SI) and lifetime suicide attempts (3.2%) by late adolescence/early adulthood in this study are similar to (albeit slightly lower than) estimates reported in other community-based samples of adolescents and adults, which range from 13% to 21% for NSSI [13–15] and 4% to 8% for suicide attempts [6,24]. Previous studies may have slightly higher estimates of NSSI and suicide attempts due to the inclusion of behaviors that are typically less severe and may not be of clinical significance. For instance, in previous studies less severe NSSI (i.e., typically less medically severe behaviors such as hair-pulling, skin-picking, self-hitting, self-biting) was endorsed by approximately 10%–20% of participants [44,45]. When adjusting for these less severe NSSIs, prevalence estimates in previous studies would range from 2.5% to 27.7%, much closer to our estimate. It is also possible that some girls in this study may have been reluctant to report sensitive information, such as NSSI or suicide attempts, during face-to-face interviews. Recent work has shown that differences exist between computerized selfinterviews, face-to-face interviews, and paper and pencil assessments, with computer assisted self-interviews revealing the greatest levels of self-disclosure [46]. Therefore, although the longevity and repeated assessments of the PGS probably instilled an element of trust in participants, the faceto-face interviews conducted with this sample may yield slightly lower estimates of NSSI and suicide attempts. The strikingly low prevalence of NSSI only without a history of SI observed in this sample is consistent with previous studies showing that NSSI tends to overlap with SI [19,27]. The fact that only about 1% of this large community sample of adolescents reported a history of NSSI without any history of SI suggests that the presence of NSSI in adolescence without ever having thought of suicide is quite rare, and very large samples would be necessary to meaningfully examine the clinical significance of this subgroup for predicting low-base-rate behaviors such as suicide attempts. Further, although not the focus of this report, it is also noteworthy that BPD symptoms emerged as the only covariate that significantly predicted greater odds of both lifetime and recent suicide attempts after controlling for early adversity (which included poverty, family stress, trauma, and sexual abuse), depression severity, SI, and NSSI. This finding is consistent with studies demonstrating

the incremental predictive validity of BPD for suicidal behavior in adults [47,48], and one longitudinal study with previously suicidal adolescents found that BPD, and not depression, predicted the persistence of suicidal behavior over time [49]. However, this is the first study to our knowledge to show that BPD symptoms prospectively and uniquely predict suicide attempts in a community sample of adolescents. These findings support the assessment of BPD features in evaluating future risk for suicide attempts among at-risk community adolescents and young adults. Interestingly, the “SI only” group reported more early adversity than did the “no SI or NSSI” group, and was also more likely to identify as minority race than to identify as Caucasian. Conversely, those who engaged in NSSI were more likely to identify as Caucasian than as racial minority. Previous findings regarding the relationship between race, class, and NSSI/SI have been mixed. Muehlenkamp and Gutierrez [24] reported that Caucasian/White females from middle- and upper-middle-class backgrounds were at higher risk compared to other demographic groups. A few studies have also found that Caucasians are more likely to engage in NSSI than minority adolescents in the community [14,44]. However, in a large sample of racially/ethnically diverse girls, there was a comparable prevalence of NSSI and SI across race, ethnicity, and socioeconomic status [50]. Nonetheless, neither minority race nor the early adversity variable, which encompassed family demographic factors, emerged as significant predictors of suicide attempts, indicating that these factors may not be as important as other factors such as BPD symptoms or NSSI history in prospectively predicting risk for suicide in young women. The most notable strength of this study is the longitudinal design in which all predictors were assessed prior to study outcomes, particularly in the analysis of recent suicide attempts that occurred only in waves 13 or 14 of data collection. In addition, this is the first study to our knowledge that has examined whether a history of NSSI and SI in adolescence predicts suicide attempts in early adulthood. We also controlled for the influence of multiple known risk factors for suicide, including BPD symptoms, depression, and family and individual-level adversity (i.e., poverty, trauma, abuse), which strengthens the conclusions that can be drawn regarding the unique influence of NSSI and SI on risk for suicidal behavior. Our assessment of cumulative early adversity, which incorporated prospective annual assessments of family stress and poverty as well as experiences of trauma and sexual abuse, is an improvement over retrospective reports that may be subject to recall bias. Moreover, the racially and socioeconomically diverse community sample of adolescents adds valuable knowledge regarding NSSI and SI as predictors of suicide attempts in a population that, due to socioeconomic adversity and its concomitants, may be at particularly high risk for suicide and in need of prevention efforts. A limitation to the current study is the ambiguity of our assessment of NSSI by using the phrase, "even if you did not

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

mean to die” in the wording of the self-injury assessments, which leaves some uncertainty regarding whether some instances of NSSI as measured in this study may actually have had suicidal intent. However, given the much higher rates of NSSI as compared to suicide attempts in community and clinical samples, and the greater likelihood for NSSI to precede suicide attempts temporally [12], it is highly likely that the vast majority of instances of self-injurious behavior were in fact instances of NSSI rather than actual suicide attempts. Considering the rather large overall sample size, the influence of the possible few cases in which the instances of adolescent self injury were carried out with suicidal intent on the results is likely to be negligible. On a related note, another limitation is the inability to predict future suicide attempts while controlling for past history of attempts. We did not have prospectively gathered data on suicide attempts among all age cohorts until wave 13 and the exact timing of lifetime attempts is unknown. Additionally, the low base rate of suicide attempts in any one assessment year (particularly in wave 14 when only 17 girls reported making a past-year suicide attempt) precluded analysis of wave 14 suicide attempts after controlling for wave 13 attempts. Therefore, we could not control for previous suicide attempts as a covariate, and it is possible that much of the variance in recent suicide attempts might be accounted for by previous attempts, particularly among those with a history of both SI and NSSI. In addition to these limitations, due to the extremely low incidence of NSSI without a history of SI, we could not examine this subgroup. Further, the low base rates of NSSI and SI made it necessary to collapse the multiple longitudinal and dimensional assessments of these constructs into dichotomous variables indicating an overall history of these behaviors, leading to loss of information regarding the severity, timing of onset, and persistence (i.e., rates of change) of SI and NSSI over time. There may also have been limited power to detect significant effects due to the low base rate of suicide attempts, particularly recent (past-year) attempts, in this community sample. Finally, given our community sample of only females, these results may not generalize to clinical, male, or mixed-gender samples.

5. Conclusions Despite these limitations, our results have important implications for clinical assessment and suicide prevention efforts. Results suggest that it is important to carefully assess youth early in adolescence for both past and current NSSI, SI, and BPD symptoms, as these risk factors typically emerge by early adolescence and can predict future suicidal behavior as late as early adulthood. Our findings add to previous studies with the PGS in highlighting the need for early assessment; for example, a recent study found that most girls who engage in NSSI during adolescence can be distinguished based on risk factors measured as early as

9

age 8 [51]. Moreover, the current study findings suggest that adolescent girls with a history of both NSSI and SI, as well as those with BPD symptoms, are at particularly high risk and may be most in need of prevention efforts. As such, it would be wise for clinicians not to focus exclusively on depressive symptoms and SI in the assessment of suicide risk, but to also screen for a history of NSSI and BPD symptoms in atrisk adolescent girls. Investigating the potential mechanisms by which NSSI increases risk for suicide is of utmost importance in future research. References [1] Eaton DK, Kann L, Kinchen S, Shanklin S, Ross J, Hawkins J, et al. Centers for Disease Control and Prevention. Youth risk behavior surveillance—United States, 2009. MMWR Surveill Summ 2010;59 (SS05):1-142. [2] Minino AM, Murphy SL, Xu J, Kochanek KD. Deaths: final data for 2008. Natl Vital Stat Rep 2011;59(10):1-126 [PubMed PMID: 22808755]. [3] Weissman MM, Bland RC, Canino GJ, Greenwald S, Hwu HG, Joyce PR, et al. Prevalence of suicide ideation and suicide attempts in nine countries. Psychol Med 1999;29(1):9-17 [PubMed PMID: 10077289]. [4] Ting SA, Sullivan AF, Boudreaux ED, Miller I, Camargo Jr CA. Trends in US emergency department visits for attempted suicide and self-inflicted injury, 1993–2008. Gen Hosp Psychiatry 2012;34 (5):557-65 [PubMed PMID: 22554432; PubMed Central PMCID: PMC3428496]. [5] Evans E, Hawton K, Rodham K, Psychol C, Deeks J. The prevalence of suicidal phenomena in adolescents: a systematic review of population-based studies. Suicide Life Threat Behav 2005;35 (3):239-50. [6] Nock MK, Borges G, Bromet EJ, Cha CB, Kessler RC, Lee S. Suicide and suicidal behavior. Epidemiol Rev 2008;30:133-54 [PubMed PMID: 18653727; PubMed Central PMCID: PMC2576496.]. [7] King CA, Jiang Q, Czyz EK, Kerr DC. Suicidal ideation of psychiatrically hospitalized adolescents has one-year predictive validity for suicide attempts in girls only. J Abnorm Child Psychol 2014;42(3):467-77. [8] Nock MK. Self-injury. Annu Rev Clin Psychol 2010;6:339-63 [PubMed PMID: 20192787]. [9] Guan K, Fox KR, Prinstein MJ. Nonsuicidal self-injury as a timeinvariant predictor of adolescent suicide ideation and attempts in a diverse community sample. J Consult Clin Psychol 2012;80(5):842-9. [10] Wilkinson P, Kelvin R, Roberts C, Dubicka B, Goodyer I. Clinical and psychosocial predictors of suicide attempts and nonsuicidal selfinjury in the Adolescent Depression Antidepressants and Psychotherapy Trial (ADAPT). Am J Psychiatry 2011;168(5):495-501 [PubMed PMID: 21285141]. [11] Cox LJ, Stanley BH, Melhem NM, Oquendo MA, Birmaher B, Burke A, et al. A longitudinal study of nonsuicidal self-injury in offspring at high risk for mood disorder. J Clin Psychiatry 2012;73(6):821-8 [PubMed PMID: 22687609; PubMed Central PMCI D: PMC3563355.]. [12] Bryan CJ, Bryan AO, May AM, Klonsky ED. Trajectories of suicide ideation, nonsuicidal self-injury, and suicide attempts in a nonclinical sample of military personnel and veterans. Suicide Life Threat Behav 2014 [PMID: 25256126]. [13] Muehlenkamp JJ, Gutierrez PM. An investigation of differences between self-injurious behavior and suicide attempts in a sample of adolescents. Suicide Life Threat Behav 2004;34(1):12-23 [PubMed PMID: WOS:000220498700003]. [14] Ross S, Heath N. A study of the frequency of self-mutilation in a community sample of adolescents. J Youth Adolesc 2002;31(1):67-77.

10

L.N. Scott et al. / Comprehensive Psychiatry xx (2015) xxx–xxx

[15] Jacobson CM, Gould M. The epidemiology and phenomenology of non-suicidal self-injurious behavior among adolescents: a critical review of the literature. Arch Suicide Res 2007;11(2):129-47. [16] Hamza CA, Stewart SL, Willoughby T. Examining the link between nonsuicidal self-injury and suicidal behavior: a review of the literature and an integrated model. Clin Psychol Rev 2012;32(6):482-95. [17] Nock MK. Why do people hurt themselves? New insights into the nature and functions of self-injury. Curr Dir Psychol Sci 2009;18 (2):78-83. [18] Madge N, Hewitt A, Hawton K, Wilde EJD, Corcoran P, Fekete S, et al. Deliberate self-harm within an international community sample of young people: comparative findings from the Child & Adolescent Selfharm in Europe (CASE) Study. J Child Psychol Psychiatry 2008;49 (6):667-77. [19] Nock MK, Prinstein MJ, Sterba SK. Revealing the form and function of self-injurious thoughts and behaviors: a real-time ecological assessment study among adolescents and young adults. J Abnorm Psychol 2009;118(4):816-27. [20] Brunner R, Parzer P, Haffner J. Prevalence and psychological correlates of occasional and repetitive deliberate self-harm in adolescents. Arch Pediatr Adolesc Med 2007;161(7):641-9. [21] Nock MK, Joiner Jr TE, Gordon KH, Lloyd-Richardson E, Prinstein MJ. Non-suicidal self-injury among adolescents: diagnostic correlates and relation to suicide attempts. Psychiatr Res 2006;144(1):65-72 [PubMed PMID: 16887199]. [22] Joiner Jr TE. Why people die by suicide. Cambridge, MA: Harvard University; 2005. [23] Andover MS, Gibb BE. Non-suicidal self-injury, attempted suicide, and suicidal intent among psychiatric inpatients. Psychiatr Res 2010;178(1):101-5. [24] Muehlenkamp JJ, Gutierrez PM. Risk for suicide attempts among adolescents who engage in non-suicidal self-injury. Arch Suicide Res 2007;11(1):69-82. [25] Stanley B, Gameroff MJ, Michalsen V, Mann JJ. Are suicide attempters who self-mutilate a unique population? Am J Psychiatry 2001;158(3):427-32. [26] Victor SE, Klonsky ED. Correlates of suicide attempts among selfinjurers: a meta-analysis. Clin Psychol Rev 2014;34(4):282-97. [27] Taliaferro LA, Muehlenkamp JJ, Borowsky IW, McMorris BJ, Kugler KC. Factors distinguishing youth who report self-injurious behavior: a population-based sample. Acad Pediatr 2012;12(3):205-13. [28] Hipwell AE, Loeber R, Stouthamer-Loeber M, Keenan K, White HR, Kroneman L. Characteristics of girls with early onset disruptive and antisocial behaviour. CBMH 2002;12(1):99-118 [PubMed PMID: 12357260]. [29] Keenan K, Hipwell A, Chung T, Stepp S, Stouthamer-Loeber M, Loeber R, et al. The Pittsburgh Girls Study: overview and initial findings. J Clin Child Adolesc 2010;39(4):506-21 [PubMed PMID: 20589562; PubMed Central PMCID: PMC2946599.]. [30] Gadow K, Sprafkin J. Child Symptom Inventories Manual-4. Stonybrook, NY: Checkmate Plus; 1994. [31] Gadow KD, Sprafkin J. Adolescent Symptoms Inventory-4: screening manual. Stony Brook, NY: Checkmate Plus; 1997. [32] Gadow K, Sprafkin J, Weiss M. Adult Self-Report Inventory 4 manual. Checkmate Plus; 2004. [33] First MB, Gibbon M, Spitzer RL, Williams JBW. Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I). Washington, DC: American Psychiatric Press; 1997.

[34] Osman A. Suicide Behavior Questionnaire-Revised (SBQ-R). University of Northern Iowa, Dept. of Psychology; 2002. [35] Loranger AW, Sartorius N, Andreoli A, Berger P, Buchheim P, Channabasavanna S, et al. The International Personality Disorder Examination. Psychol Assess Resour 1994. [36] Chanen A, Jovev M, Djaja D, McDougall E, Yuen H, Rawlings D, et al. Screening for borderline personality disorder in outpatient youth. J Pers Disord 2008;22(4):353-64 [PubMed PMID: 18684049]. [37] Stepp S, Burke J, Hipwell A, Loeber R. Trajectories of attention deficit hyperactivity disorder and oppositional defiant disorder symptoms as precursors of borderline personality disorder symptoms in adolescent girls. J Abnorm Child Psychol 2012;40(1):7-20. [38] Stepp SD, Pilkonis PA, Hipwell AE, Loeber R, Stouthamer-Loeber M. Stability of borderline personality disorder features in girls. J Pers Disord 2010;24(4):460-72 [PubMed PMID: 20695806; PubMed Central PMCID: PMC2920503]. [39] Smith DJ, Muir WJ, Blackwood DH. Borderline personality disorder characteristics in young adults with recurrent mood disorders: a comparison of bipolar and unipolar depression. J Affect Disord 2005;87(1):17-23 [PubMed PMID: 15967232]. [40] Barnard KE. Difficult Life Circumstances Scale. Seattle, WA: NCAST Publications; 1994. [41] Foa EB, Johnson KM, Feeny NC, Treadwell KR. The Child PTSD Symptom Scale: a preliminary examination of its psychometric properties. J Clin Child Psychol 2001;30(3):376-84. [42] Evans GW, Li D, Whipple SS. Cumulative risk and child development. Psychol Bull 2013;139(6):1342-96 [PubMed PMID: 1324549522; 2013-12151-001]. [43] Bridge JA, Goldstein TR, Brent DA. Adolescent suicide and suicidal behavior. J Child Psychol Psychiatry 2006;47(3–4):372-94. [44] Lloyd-Richardson EE, Perrine N, Dierker L, Kelley ML. Characteristics and functions of non-suicidal self-injury in a community sample of adolescents. Psychol Med 2007;37(8):1183-92 [PubMed PMID: 17349105; PubMed Central PMCID: PMCPmc2538378]. [45] Yates TM, Tracy AJ, Luthar SS. Nonsuicidal self-injury among "privileged" youths: longitudinal and cross-sectional approaches to developmental process. J Consult Clin Psychol 2008;76(1):52-62. [46] Kim J, Dubowitz H, Hudson-Martin E, Lane W. Comparison of 3 data collection methods for gathering sensitive and less sensitive information. Ambul Pediatr 2008;8(4):255-60. [47] Brodsky BS, Malone KM, Ellis SP, Dulit RA, Mann JJ. Characteristics of borderline personality disorder associated with suicidal behavior. Am J Psychiatry 1997;154(12):1715-9 [Epub 1997/12/16. PubMed PMID: 9396951.]. [48] Yen S, Shea MT, Pagano M, Sanislow CA, Grilo CM, McGlashan TH, et al. Axis I and Axis II disorders as predictors of prospective suicide attempts: findings from the Collaborative Longitudinal Personality Disorders study. J Abnorm Psychol 2003;112(3):375-81 [PubMed PMID: 12943016; PubMed Central PMCID: PMC3272767]. [49] Greenfield B, Henry M, Weiss M, Tse SM, Guile J-M, Dougherty G, et al. Previously suicidal adolescents: predictors of six-month outcome. Can Child Adolesc Psychiatr 2008;17(4):197-201. [50] Hilt LM, Cha CB, Nolen-Hoeksema S. Nonsuicidal self-injury in young adolescent girls: moderators of the distress–function relationship. J Consult Clin Psychol 2008;76(1):63-71. [51] Keenan K, Hipwell AE, Stepp SD, Wroblewski K. Testing an equifinality model of nonsuicidal self-injury among early adolescent girls. Dev Psychopathol 2014;26(Special Issue 03):851-62.