Prospective Investigation of the Interaction between Social Problems and Neuropsychological Characteristics on the Development of Suicide Ideation

Abstract

Objective:

Negative social interactions are known to contribute to the development of suicide ideation. However, it is unclear how this risk factor interacts with other predisposing risk factors. The purpose of this study was to determine how social and neuropsychological factors interact as a prospective predictor of the emergence of suicide ideation in adolescents.

Method:

Data were collected from adolescents (M age=13.12, SD=1.48) over three years as part of a larger study. Participants completed the MINI-Kid and Youth Self-Report, which were used to assess for suicidality. Negative social interactions were operationalized as the Social Problems scale of the YSR. Additionally, adolescents completed a neuropsychological battery at each wave of data collection.

Results:

Logistic moderation analyses demonstrated a significant interaction between task switching and endorsement of negative social interactions in the prediction of suicide ideation one year later, Wald χ²(1)=4.94, OR=.90, p<.05. Distractibility was a significant predictor, Wald χ²(1)=5.52, OR=3.45, p<.05, but it did not demonstrate an interaction effect. Perseveration failed to reach statistical significance independently and in the interaction.

Conclusions:

The results indicate that certain neuropsychological characteristics can aid in predicting which adolescents will develop suicide ideation in the presence of negative social interactions, which may have significant clinical implications.

Suicide is a major health concern in the United States as rates have risen nearly 30% with 44 states reporting significant increases between 1999 and 2016 (Stone et al., 2018). Adolescents are not exempt from this increase in suicide rates. While it remains the second leading cause of death (CDC, 2018), national statistics have shown a 31% increase in suicide during the 2010s for adolescents aged 13–15, primarily driven by females (Twenge, Joiner, Rogers, & Martin, 2018). Adolescence is a key period for identifying potential risk and protective factors as suicide ideation (SI) typically emerges initially during the transition into adolescence with 12.1% endorsing SI and attempts peaking in mid-adolescence with a prevalence of 4.1% (Kessler, Borges, & Walters, 1999; Lewinsohn, Rohde, Seeley, & Baldwin, 2001; Nock et al., 2013; Shaffer et al., 1996). These data underscore the importance of identifying adolescents who may be at an elevated risk for developing suicidality so that early interventions may be implemented.

Although research on suicide is abundant and risk factors have been identified, prediction is still only slightly better than chance (Franklin et al., 2017). There is a need to incorporate multiple known risk factors into suicide prevention, yet prevention often focuses on a single factor, typically diagnosed mental health conditions, to identify those at risk for suicide (Office of the Surgeon General, 2012). Although this may be a prominent predictor, this may not be an adequate focal point for prevention since 54% of those that died by suicide did not have a known mental health condition at the time of death (Stone et al., 2018). This indicates a need to incorporate other, non-affective risk factors into consideration when implementing suicide prevention programs to capture those that do not hold a psychological diagnosis (Caine, Reed, Hindman, & Quinlan, 2018). For 41–43% of youth suicides, the first indication of a mental health condition was the death (McKean et al., 2018; Rodway et al., 2016), which demonstrates the importance of identifying risk factors for youth suicide other than a mood disorder, particularly factors that may make them susceptible to SI and psychological distress.

Given the rate of death from initial attempts, it is important to ascertain factors that may predispose an adolescent to SI. A review of research (Stewart et al., 2017) on adolescent suicidality suggested that Joiner’s (2005) Interpersonal-Psychological Theory of Suicide (IPTS) is applicable for understanding adolescent suicide. The IPTS states that a desire to die and the acquired capability to self-harm are prerequisites for suicide. Further, the theory notes two interpersonal factors which contribute to the development of SI, or the desire to die: thwarted belongingness and perceived burdensomeness. Thus, peer support, or lack of peer support, may be one factor that can aid in identifying adolescents susceptible to developing SI.

Research has demonstrated sex-specific differences in the effect of peer support on SI in adolescents. In a longitudinal study, Adrian, Miller, McCauley, and Vander Stoep (2016) showed that, for females only, a high suicidality trajectory was predicted by low perceived peer support. Similarly, Farrell, Bolland, and Cockerham (2015) found a significant, negative relationship between suicide risk and peer support. Together, these results suggest that perceived peer support acts as a protective factor against SI during adolescence, particularly for females. In a clinical sample, however, Kerr, Preuss, and King (2006) found that, for males only, peer support was positively related to suicidal ideation, which was posited to be a result of affiliating with depressed or suicidal peers. Findings have been inconsistent across settings (O’Donnell, O’Donnell, Wardlaw, & Stueve, 2004), suggesting that peer support’s effect on SI development may be dependent upon other environmental factors.

There is some evidence to suggest that fluctuating levels of peer support and the quality of that support is a contributing factor to suicide risk. Following psychiatric hospitalization, Czyz et al. (2012) found that female adolescents who felt connected with peers showed a decrease in SI for the 3 months after hospitalization, but peer connectedness was a predictor of severe ideation one year later. One explanation for this counterintuitive finding is that one might experience an increase in connectedness immediately following a suicide attempt, which returns to a baseline level over time and is then interpreted as rejection or loss. Furthermore, there is a need to identify if the relationships are supportive and provide adaptive (e.g., counseling suggestion) or maladaptive (e.g., self-harm support) advice as some research has indicated a positive correlation between peer connectedness and suicide thoughts and behaviors (Kaminski et al., 2010).

Suicide risk has also shown to be predicted by elevated levels of social isolation as a coping mechanism (Spirito, Francis, Overholser, & Frank, 1996) and problematic peer relationships (King, Ruchkin, & Schwab-Stone, 2003; Perkins & Hartless, 2002). Specifically, bullying (22%) and social isolation or withdrawal (29%) are prominent antecedents to youth suicide (Rodway et al., 2016). Prinstein and Aikins (2004) found that peer rejection prospectively predicted depression in adolescents, a common precursor to suicidality.

Overall, research has implicated negative social interactions as contributing to the development of SI, particularly for female adolescents, with some research demonstrating a linear relationship between the perceived frequency of peer rejection and increased likelihood for suicide (Bearman & Moody, 2004; Giletta et al., 2015; Gini & Espelage, 2014; Prinstein et al., 2000). However, the presence of negative peer relationships alone does not sufficiently predict suicide attempts; these relationships may interact with other risk and protective factors. For instance, neuropsychological factors has been shown to interact with social relationships to mitigate or exacerbate an adolescent’s likelihood of developing SI. McCall and Black (2013) proposed an impaired disengagement framework in which difficulties with executive functioning, particularly difficulties with shifting attention, impair decision-making processes and make the individual more susceptible to ruminating on negative self-referent information. This, in turn, could possibly result in increased consideration of suicide as solution to this negative affect.

Executive functioning is the ability of an individual to select context-specific actions while inhibiting the prepotent, competing, and context-inappropriate actions (Logan, Schachar, & Tannock, 1997; Pennington & Ozonoff, 1996). Additionally, executive function allows an individual to adapt to changing, complex cognitive tasks (Barch et al., 2009; Botvinick et al., 2001; Scott, 1962). Executive function is comprised of several individual components which work together to aid in these abilities. These smaller components include task switching, cognitive flexibility, attention shifting, and information updating (Baddeley, 1998; Miller & Cohen, 2001; Miyake et al., 2000).

Research has demonstrated deficits in neuropsychological functioning, particularly components of executive function, in populations of individuals who have attempted suicide. Studies of executive function in adults who have attempted suicide suggest that the characteristics of an individual’s deficits may be utilized to predict which types of suicide behaviors they are more likely to engage in (e.g., impulsive or planned acts) and tailor interventions to that individual’s profile of difficulties (Richard-Devantoy et al., 2013; 2015). Keilp et al. (2013) also investigated the relationship between lethality of attempts and cognitive factors. They found that performance on selective attention tasks could not be fully explained by the severity of suicidal ideation or depression. This would indicate that this facet of executive function acts to an extent as an independent suicide risk factor rather than a result of psychological distress.

For adolescents, cognitive flexibility, or the ability to alter thinking and shift attention away from incorrect responses following negative feedback, has been shown to prospectively predict suicidality (Miranda et al., 2012). Additionally, cognitive abilities have demonstrated the ability to differentially predict young-adult depressive symptoms, SI, and suicide attempts in a longitudinal studies with adolescents (Crandall, Allsop, & Hanson, 2018; Schwartz, Ordaz, Ho, & Gotlib, 2019). Taken together, the literature on neuropsychological correlates of SI for adolescents and adults suggest a need to investigate how these neuropsychological deficits interact with other known risk factors to impact suicidality in adolescence prior to the initial emergence of ideation. The current study will focus on three aspects of executive functioning: cognitive flexibility, task switching, and distractibility.

Few studies have investigated the potential effects of negative social interactions (NSI) along with its interaction with neurocognitive functioning as a predictive factor for SI development in adolescents. Thus, the purpose of the current study was to determine if neuropsychological functioning and the presence of NSI can predict the emergence of SI one year later in adolescents. Due to the complex nature of suicide prediction, it was hypothesized that NSI would significantly predict future suicidality but the neurocognitive factors would only emerge as significant predictors when moderated by NSI. This hypothesis was based on research suggesting that, when not experiencing stress, first-degree relatives of individuals who died by suicide are not differentiated from those without this family history on measures of executive function (McGirr et al., 2010). However, these same individuals did not show the expected practice effects on the neuropsychological measures after a stress-inducing test. This study indicated that those that died by suicide may have shown similar performance as their relatives, which would suggest they did not cope with stressors as expected. Thus, we expected to see differences in neuropsychological characteristics between those endorsing SI and those not endorsing SI when the adolescents were experiencing social stress (i.e., negative social interactions). As this study is examining functioning in adolescents prior to the onset of SI, the findings may elucidate the potential mechanism and conditions through which neuropsychological susceptibility to SI emerges as a risk factor.

Method

Participants

Participants were recruited as part of a larger study examining sleep and health in children ages 8–12 (Wong, Brower, Conroy, Lachance, & Craun, 2018). Data was collected for baseline assessment (Time 1) and at one-year (Time 2) and two-year (Time 3) follow-ups. Data from Time 2 (T2) and Time 3 (T3) were used for the current study as the child participants were closer to the age of typical emergence of suicidality (Nock, Borges, Bromet, & Cha et al., 2008). Prior research has indicated that SI peaks for male adolescents at age 12 and for females at age 13 (Adrian, Miller, McCauley, and Vander Stoep, 2016), which was the mean age of the adolescents in the present study at Time 3. Data collection is ongoing, so a total of 78 of the 245 children that completed Time 1 have also participated in Time 2 and Time 3 of the study.

The participants were recruited through community flyers and advertisements through the local newspapers, radio stations, and social media. Children were excluded from the study if they had any medical or psychological diagnoses that may affect sleep, were taking medications that may impact sleep, or met criteria for certain disorders (e.g., ADHD, psychotic disorders, sleep disorders, fetal alcohol syndrome). The parent study is assessing the relationship between parental alcohol use and youth development so some of the children in the present study have a parent who met criteria for a history of an alcohol use disorder. Since the majority (64.1%) of the child participants in this study had a parent with a history of alcohol use disorders and alcohol use is a significant risk factor for suicide (Alonzo, Thompson, Stohl, & Hasin, 2014), children of alcoholic (CoA) status (i.e., whether or not the parent met criteria for a disorder) was controlled for in all analyses. CoA status was not a significant predictor of T3 SI in any of the analyses.

The participants were 64.1% female, 79.5% Caucasian, and had a mean age of 13.15 (SD=1.48) at Time 3. There was little endorsement of suicidality for Time 1 (1.63%) and Time 2 (2.54%), which may be in part due to the young age (Time 1 M age=10.49, SD=1.51; Time 2 M age=11.83, SD=1.60). However, 16.7% (n=13) of participants endorsed suicidality at Time 3 of the study. This increase in SI is consistent with prior literature showing a sharp increase after age 12 which cannot be fully accounted for by psychological disorders (Glenn et al., 2017; Nock et al., 2013; Nock, Holmberg, Photos, & Michel, 2007). Endorsement of suicide ideation prior to Time 3 was not a significant predictor in any of the models so it was dropped from final analyses.

Procedures

Participants completed a screening interview at each wave of the study during which the child is administered the MINI International Neuropsychiatric Interview for Children and Adolescents (MINI-Kid; Sheehan et al., 2010). In the second session of each wave, the child is administered the Youth Self-Report (Achenbach, 1991) as well as a series of questionnaires assessing sleep. In the third and final session of each wave, the child completes a series of neuropsychological assessments including the Wisconsin Card Sorting Task and the Delis-Kaplan Executive Functioning System Color-Word Interference (D-KEFS; Delis, Kaplan, & Kramer, 2001).

Measures

Suicidality

Mini International Neuropsychiatric Interview for Children and Adolescents (MINI-Kid).

The MINI-Kid is a short, structured diagnostic interview for children ages 6–17 that has been validated for use by clinicians and researchers to aid in diagnosing common psychiatric disorders according to the DSM-IV and ICD-10 (Sheehan et al., 1997; Sheehan et al., 1998). The Suicidality module was used for the purposes of the current study. Questions for modules are listed in a binary yes/no format. Participants were categorized based on the standard scoring and administration of this module such that any child who coded in as having low or more severe suicide risk were included in the suicidality group (n=10). Data from T3 of the study were used for this categorization. The MINI-Kid is considered a gold-standard interview for assessing child psychopathology and has been utilized for accurately assessing suicide risk in prior studies (Chan et al., 2016; Sheehan et al., 2010).

Youth Self-Report (YSR).

The Youth Self-Report (YSR; Achenbach, 1995) is a 112-item self-report measure for externalizing and internalizing behavioral problems in children ages 11–18. This measure has evidenced good reliability and validity (Achenbach & Rescorla, 2001) as well as high temporal stability over the course of a decade (Achenbach, Dumenci, & Rescorla, 2002). The YSR has demonstrated accuracy in capturing SI in conjunction with a structured interview (Dhossche, Ferdinand, van der Ende, Hofstra, & Verhulst, 2002; Giannetta et al., 2012; Van Meter et al., 2018). For suicidality, the participants that endorsed thoughts of killing themselves or attempts to hurt themselves were included in the suicidality group. A total of 7 participants endorsed one of these items, which made for a total of 13 participants reporting suicidality after accounting for overlap across measures. The difference in SI endorsement between the two measures was anticipated as research has indicated a difference in endorsement of SI between self-report (e.g., YSR) and interview-based (e.g., MINI-Kid) measures based on the comfort-level of the participant (Viguera et al., 2015). As a result, we included participants in the suicidality group if they endorsed SI in one or both of the measures to more accurately capture SI.

Negative Social Interactions

For NSI, the YSR includes a subscale of social problems. However, this subscale includes items that are not relevant to the current study (e.g., clumsiness), so a subset of this scale was used which focuses on the content of the interactions: not getting along with other kids, teased, and not liked by other kids. The data was originally severely skewed (z=4.24) with an acceptable kurtosis (z=0.3). However, the inverse data transformation normalized the skew (z=1.86) but increased the kurtosis (z=3.03). It is standard practice to dichotomize a variable when transformations do not effectively normalize the distribution so the variable was dichotomized (Tabachnick & Fidell, 2013). At T2, 32 participants endorsed NSI and a dichotomized version of the variable (0=No NSI, 1=NSI Endorsed) was used for analyses Previous research has dichotomized full and abbreviated scales of the YSR as this measure of behavioral problems has low item endorsement in a non-clinical population (Ivanova et al., 2019; Lexcen, Vincent, & Grisso, 2004; Undheim & Sund, 2010; Verhulst, Dekker, Van der Ende, 1997). Data were first analyzed using the full Social Problems scale and similar results were obtained consistent with prior literature comparing continuous and dichotomized scales of the YSR (Korhonen et al., 2012). The results of the abbreviated scale are presented here to minimize the contribution of non-target social problems to the models.

Task Switching

Delis-Kaplan Executive Function System (D-KEFS).

The D-KEFS (Delis, Kaplan, & Kramer, 2001) Color-Word Interference subtest was used to assess task switching. The D-KEFS has been validated for use with children and adolescents in a nonclinical sample to assess for deficits in executive function (Delis, Kramer, Kaplan, & Holdnack, 2004). Additionally, the D-KEFS has evidenced good reliability for children and adults (Corbett, Constantine, Hendren, Rocke, & Ozonoff, 2009; Delis, Kaplan, Kramer, 2001). The Color-Word Interference Test is a measure of inhibition similar to the Stroop Test (Stroop, 1935) with an added component assessing task switching. There are two baseline conditions in which the children name color patches on a page (Condition 1) or read color names printed in black ink (Condition 2). For the interference task (Condition 3), the children name the color of ink the letters are printed in instead of reading the color word (e.g., the word “red” printed in blue ink). In the final task (Condition 4), the participants switch back and forth between reading the word and naming the ink color. This final condition measures both inhibition, as they need to inhibit reading words at times, and task switching, as they need to switch between naming words and colors. The contrast measure between Condition 3 and Condition 4 was used to partial out the lower-level skills of basic word reading and color naming as well as the higher-level skill of inhibition to isolate the skill of task switching.

Cognitive Flexibility

Wisconsin Card Sorting Task (WCST).

The Wisconsin Card Sorting Task (WCST; Heaton, 1981) assesses the participant’s cognitive flexibility, set-shifting and problem solving abilities through perseverative error responses (i.e., repetition of an incorrect response following negative feedback). For the current study, the computerized version of the measure was utilized to minimize administration errors. The participants are presented with four cards that incorporate three parameters (i.e., color, form, and number) and are asked to match a fifth card to one of the four stimulus cards according to different principles. These principles change throughout testing, so the participants are also required to shift their approach during the administration. The WCST has demonstrated good validity and reliability as a measure of children’s neurocognitive functioning (Welsch, Pennington, & Groisser, 1991).

Distractibility

The WCST was also used as a measure of distractibility. This was operationalized as the number of times the participant chooses a card that does not match the pattern after making at least five consecutive correct choices without completing the category (i.e., failure to maintain set). Figueroa and Youmans (2013) found that failure to maintain set on the WCST was an inverse predictor of signal detection, a measure of vigilance and divided attention. Thus, they posited that this variable measures the opposite concept of distractibility as the participants have already mastered the category rule and applied it to the prior cards.

Plan of Analysis

Data were analyzed with logistic moderation using the PROCESS V3.3 macro (Hayes, 2018) for SPSS V23.0 (IBM Corp, 2015). Moderation models used bootstrapping to account for the small sample size. This allowed the program to create 10,000 random samples with replacement to generate 95% confidence intervals to indicate significance. The neurocognitive variables at T2 (i.e., task switching, cognitive flexibility, and distractibility) were entered into the model separately as the independent variables. The NSI at T2 was entered as the moderating variable for each model. Endorsement of suicide ideation on either the YSR or the MINI-Kid at T3 was used as the dichotomous dependent variable (0=No SI, 1=SI Endorsed). Additionally, child age, gender, ethnicity, and CoA status were entered into each model as covariates to control for the effects of demographic variables on the dependent variable. However, only gender was a significant predictor so the other demographics were dropped from the final analyses.

Results

Data Preparation

Data were assessed for normality and transformed when necessary. Perseverative errors on the WCST showed a substantial positive skew, so the variable was log transformed. The NSI composite was severely positively skewed. The full Social Problems subscale showed a severe positive skew, as well, but the abbreviated scale was used in analyses to isolate the variance attributable to social interactions. Rather than using an inverse transformation, which also resulted in a substantially skewed variable, this variable was dichotomized so that participants who endorsed at least one of the items from the YSR were coded as a 1 (NSI Present) and those that did not endorse any of the YSR items were coded as a 0 (NSI Absent). Thus, 41% of participants who endorsed at least one item were coded as having endorsed NSI during Time 2. All subsequent analyses controlled for the participant’s gender.

Moderation Analyses

NSI

A logistic regression was conducted to determine if NSI at Time 2 predicted the emergence of SI at Time 3 after controlling for the demographic variables. The overall model was significant, χ²(5)=17.417, p=.004. NSI was also a significant independent predictor within the model, Wald χ²(1)=4.240, OR= 4.69, p=.039. This result indicates that the risk for developing SI one year later increases in the presence of NSI.

Task Switching

Using the PROCESS macro, the relationship between T2 task switching and T3 SI as moderated by T2 NSI was examined. The overall model was significant, χ²(4)=19.57, p<.001. The main effects for NSI [Wald χ²(1)=6.32, OR=1268.83, p<.05, 95% CI [1.503,11.436]] and task switching [Wald χ²(1)=4.09, OR=1.09, p<.05, 95% CI [.001,.149]] both reached statistical significance. Additionally, the interaction between NSI and task switching was significant, Wald χ²(1)=4.940, OR=.90, p<.05, 95% CI [−.171,−.007]. This interaction was probed and, as shown in Figures 1 and 2, as task switching decreases by one standard deviation, the predicted odds of SI when NSI is present are 1.66 times the initial odds and 0.15 times the initial odds when NSI is not present. This means that task-switching abilities show a small increase likelihood of endorsing SI, but that this effect is larger when the adolescent is also experiencing negative social relationships. In addition to both factors significantly predicting IS, they interact with each other to increase the odds of endorsing SI one year later. Compared to males, the females were 23.57 times as likely to endorse SI regardless of the presence of NSI.

Figure 1.

Interaction between task switching abilities (TS) and negative social interactions (NSI) on the predicted odds of endorsing suicide ideation one year later for female adolescents.

Figure 2.

Interaction between task switching abilities (TS) and negative social interactions (NSI) on the predicted odds of endorsing suicide ideation one year later for male adolescents.

Cognitive Flexibility

Next, the relationship between cognitive flexibility at T2 and T3 SI as moderated by T2 NSI was examined. The overall model was significant, χ²(4)=17.668, p<.01. However, neither the main effects for NSI (p=.10) and perseveration (p=.75) nor the interaction between the two (p=.78) reached statistical significance. These results indicate that a tendency to perseverate (i.e., cognitive inflexibility) at Time 2 is not a significant risk factor for the development of SI at Time 3 regardless of the presence of NSI.

Distractibility

The relationship between distractibility at T2 and T3 SI as moderated by T2 NSI was examined. The overall model was significant, χ²(4)=19.76, p<.001. The main effects for NSI [Wald χ²(1)=4.48, OR=18.80, p<.05, 95% CI [.469,5.609]] and distractibility [Wald χ²(1)=5.52, OR=4.22, p<.05, 95% CI [.043,2.199]] were significant. However, the interaction failed to reach statistical significance, p=.48. These results indicate that NSI does not affect the relationship between distractibility and suicidality. Additionally, as distractibility increases (i.e., more failures to maintain set on the WCST), the odds of endorsing SI one year later increase.

Discussion

The current study investigated the relationship between negative social interactions and neuropsychological functioning in adolescents and their ability to predict the development of suicide ideation one year later. Results demonstrated that negative social interactions prospectively predicted the emergence of suicidal ideation. Additionally, distractibility was also a significant prospective predictor of later SI. Task switching emerged as a predictor of SI after adding social problems as a moderator. This effect was pronounced for females such that the combination of low task switching abilities and the presence of problematic social interactions increase the odds of developing suicidal ideation. Contrary to hypotheses, cognitive flexibility did not significantly predict SI alone or in the moderated analysis. The results of the current study indicate that certain neuropsychological characteristics that are known deficits in suicide attempters can aid in predicting which adolescents will develop SI, yet others do not emerge as risk factors until the adolescent experiences negative social interactions with peers.

The diathesis-stress model (Ingram & Luxton, 2005; van Heeringen, 2012; Zuckerman, 1999) may help explain why cognitive flexibility did not emerge as a significant predictor. The model states that an interaction between biological and psychological factors makes an individual susceptible to psychopathology and the occurrence of a stressful life event increases the risk for suicidality (Mann, Waternaux, Haas, & Malone, 1999). McGirr et al. (2010) examined the relationship between stress and neuropsychological functioning in relatives of suicide decedents as neuropsychological deficits in suicide decedents are also present in first-degree relatives. They found that the relatives did not differ from controls at baseline. However, after inducing stress with the Trier Social Stress Test, the relatives did not show practice effects on the cognitive tasks, but the control group did show significant improvement. This result indicated that the relatives, and potentially the suicide decedents, experience an inability to respond appropriately to stress which then interacts with neuropsychological functioning to contribute to the increased risk of suicide.

The current study holds importance for the field of suicide prevention. Since research suggests a sensitization process to suicide behaviors, prior attempts have become a focus when trying to identify individuals at risk for future attempts and suicide deaths (Shaffer et al., 1996; Lewinsohn, Rohde, & Seeley, 1994; Beautrais, 2004). However, 60–71.4% of those who died by suicide did so on the initial attempt (Bostwick, Pabbati, Geske, & McKean, 2016; McKean, Pabbati, Geske, & Bostwick, 2018). These findings indicate that suicide prevention programs that aim to identify suicide survivors miss a large portion of the individuals that die by suicide. Thus, research identifying risk factors prior to an initial attempt can aid in identifying this population.

The current study had several limitations. The correlational design of this study does not allow us to experimentally assess for causal relationships. The sample size of the adolescents endorsing suicide ideation was smaller than those with no history of suicide ideation. It is possible that some adolescents did not feel comfortable endorsing SI and the true sample of adolescents with SI is larger than what is presented here. However, we attempted to minimize this by including verbal and written reporting of SI for the adolescent. We attempted to reduce the difference in group sizes by using the older adolescents with Time 2 and Time 3 of the study, but there was still a substantial difference. Future studies could be designed to follow a larger population of adolescents over time with a specific aim of collecting baseline data prior to the emergence of suicide ideation. The larger study for the data presented here was focused on CoAs and thus targeted those adolescents. A similar study could be designed to target children that are at-risk for developing suicide ideation during adolescence based on environmental and familial risk factors. Additionally, the participants in the study were primarily Caucasian and living in a rural area of the western United States so these results may not adequately generalize to other areas of the country or minority populations. With these limitations in mind, the results presented here identify a possible condition (NSI) through which neuropsychological susceptibility could result in the development of suicide ideation.

These findings may have significant implications for suicide prevention programs with adolescents as they help to elucidate how various suicide risk factors interact with each other during the development of SI in adolescence. These neuropsychological characteristics could potentially be identified by clinicians and neuropsychologists, which could then help to target at-risk adolescents before we typically begin to observe psychological distress. Preventive measures could be put into place to teach these adolescents compensatory strategies, develop and improve problem solving abilities, and implement social interventions. The results suggest that certain facets of executive function (i.e., task switching) do not become significant predictors of suicide ideation until the adolescent experiences negative social interaction. Thus, a preventative strategy could be to educate the at-risk adolescents on adaptive coping mechanisms, recognition of healthy relationships, and ways to foster healthy friendships.