Associations of Suicidality with Cognitive Ability and Cognitive Insight in Outpatients with Schizophrenia

Abstract

Background

Previous literature suggests that better cognitive ability and insight are associated with greater lifetime risk of suicide attempts in schizophrenia, counter to the direction of association in the general population. However, the conjoint association between distinct cognitive domains, insight, and suicidality has not been assessed.

Method

In a cross-sectional study, 162 adults with schizophrenia or schizoaffective disorder completed cognitive testing via the MATRICS battery, symptom and cognitive insight assessments, along with the Columbia Suicide Severity Rating Scale. We then contrasted participants based on history of suicidality by cognitive domains and insight measures and conducted multivariate analyses.

Results

Although a history of any passive ideation was not associated with cognitive ability or insight, verbal learning was positively associated with a greater history of suicidal attempt and prior ideation with a plan and intent. Higher cognitive insight, and the self-reflectiveness subscale insight, was also associated with history of passive or active suicidal ideation. Cognitive insight and cognitive ability were independent from each other, and there were no moderating influences of insight on the effect of cognitive ability on suicide related history. Exploratory analyses revealed that history of planned attempts were associated with greater verbal learning, whereas histories of aborted attempts were associated with poorer reasoning and problem-solving.

Implications

Although cross-sectional and retrospective, this study provides support that greater cognitive ability, specifically verbal learning, along with self-reflectiveness, may confer elevated risk for more severe suicidal ideation and behavior in an independent fashion. Interestingly, poorer problem-solving was associated with aborted suicide attempts.

1. Introduction

Approximately 5% of people with schizophrenia die by suicide (Palmer et al., 2005) and up to 40% attempt suicide (Pompili et al., 2007), yet the unique risk factors for suicide in schizophrenia are somewhat unclear (Kasckow et al., 2011). Similar to people without schizophrenia, risk factors include history of suicide attempts, social isolation, depressive symptoms, hopelessness, and substance use (Pompili et al., 2007). However, several studies indicate that greater premorbid function and cognitive ability are associated with higher rates of suicide attempts (Kim et al., 2003; Nangle et al., 2006), in contrast to the association between lower cognitive function and suicide attempts in the general population (Kosidou et al., 2014). One hypothesis is that greater cognitive ability may be associated with increased insight into illness and cognition, which may lead to distress related to awareness of the disorder and its implications (Cooke et al., 2007). Given the increased attention to suicide in psychotic disorders (Chesney et al., 2014) and to interventions that enhance cognition (Thorsen et al., 2014), further research is needed to understand: a) which cognitive domains are most associated with lifetime history of suicide attempts and ideation, b) whether insight and cognitive function are independent or inter-related correlates of suicide risk, and c) whether these variables are associated with contextual factors in suicide attempt, such as planning or preparation.

Past research has examined the association of cognitive function with histories of suicidality. For example, Kim et al. (2003) found an association between history of any lifetime suicidality (i.e., thoughts, self-harm acts, plans/threats, or attempts) and better cognitive function on psychomotor speed and attention, verbal working memory, verbal fluency, recall memory, and executive function. Nangle et al. (2006) found greater executive functioning in patients with at least one lifetime suicide attempt, with significantly better performance on attention and verbal fluency. Delaney et al. (2012) found better performance on measures of episodic memory, working memory, and IQ in patients with a history of suicidal ideation or a single attempt (vs. multiple attempters), compared to patients with no past ideation or attempts. Taken together, there is evidence supporting greater rates of suicidal behavior among patients with schizophrenia with better cognitive function, but there is little consistency in which cognitive domains are most associated with past suicidal behavior. One limitation is that standardized measures of suicidality were often not used, particularly ones that systematically address levels of suicidal ideation and the nature and frequency of prior attempts.

One mechanistic hypothesis to explain the association between cognitive functioning and suicidality is that greater cognitive ability may increase cognitive insight, which may influence a patient’s ability to initiate and coordinate suicidal behavior (Nangle et al., 2006). Poor clinical insight (e.g., awareness of clinical state) has been associated with impaired cognitive function (Donohoe et al., 2009), while better clinical insight has been associated with a greater likelihood of suicidal behavior in schizophrenia (Crumlish et al., 2005; Massons et al., 2017). However, no studies have examined the potential for an interactive or mediational association of cognitive ability and cognitive insight and suicidality, such as whether insight mediates or moderates the association between cognitive ability and suicidal thinking and behavior. Cognitive insight refers to the patients’ capacity to evaluate their atypical experiences and misinterpretations of events (Beck et al., 2004).

Given the high base rate of suicide in schizophrenia (Pompili et al., 2007), there is a need to refine understanding of associations between cognitive ability, insight, and suicidal thoughts and behavior. In a sample of treated outpatients with schizophrenia, we assessed for cognitive ability, cognitive insight, and history of suicidal ideation and behavior. We hypothesized that greater cognitive ability and insight would be independently associated with higher rates of past suicide attempts. In multivariate models, we examined whether insight mediated the association between cognitive ability and suicidality. We also examined whether cognitive insight moderated these associations (such as by enhancing the association of cognitive ability with suicidality). Lastly, we explored the associations between different domains of cognitive ability and different types of suicide attempts (e.g., planned versus unplanned; aborted).

2. Method

2.1 Participants

Data from this study included baseline data of an ongoing randomized controlled trial evaluating the effectiveness of different mobile-health augmented interventions in serious mental illnesses. For this study, we included data from 162 adults aged 18 and older who were diagnosed with schizophrenia or schizoaffective disorder. Participants were recruited from the public San Diego County Adult and Older Adult Mental Health System. The clinical trial was designed to be inclusive of users of the mental health system and therefore other exclusion criteria were kept minimal.

To enroll in the clinical trial, participants must have met DSM-IV criteria for schizophrenia or schizoaffective disorder. Participants must have had at least a minimum level of impairment on at least one of the target outcomes, defined as a moderate score (≥ 3) on at least one of the BPRS depression, mania, hallucinations, or emotional withdrawal items. Diagnoses were formed based on a combination of the Mini-International Neuropsychiatric Inventory (MINI; Lecrubier et al., 1997) and medical records obtained with participant consent. Participants needed to be able to read and speak English, willing to sign a release of information for their provider, and provide informed consent, as well as pass the University of California, San Diego Brief Assessment of Capacity to Consent (UBACC) test (Jeste et al., 2007) for decision capacity for research. Participants were excluded if they were currently enrolled in psychotherapy or had received CBT within the past 5 years, had been diagnosed with dementia, had experienced head trauma with loss of consciousness for more than 20 minutes, or current participation in a psychosocial/pharmacological clinical trial. This trial was approved by the University of California, San Diego’s Human Research Protections Program.

2.2 Procedures

During the baseline visit, participants completed cognitive testing, symptom assessments, and suicide history assessments in a research facility or in the community, depending upon their preference. Raters were trained in administering interview-based measures and needed to achieve a 0.90 inter-rater reliability kappa with gold-standard raters in order to administer tests.

2.3 Measures

2.3.1 Brief Psychiatric Rating Scale-24

The Brief Psychiatric Rating Scale-24 item expanded version 4.0 (BPRS-24) was used to assess psychopathology symptoms (Ventura et al., 2000). The BPRS-24 assesses 24 psychiatric symptoms including anxiety, depression, mania, delusions/hallucinations, unusual behavior, and suicidality. Presence of symptoms and severity of symptoms are rated on a 1- (not present) to 7-(extremely severe) point Likert scale. Scores range from 24 to 168, with higher scores indicative of greater severity of psychotic symptoms.

2.3.2 Cognitive Ability

The MATRICS (Measurement and Treatment Research to Improve Cognition in Schizophrenia) Consensus Cognitive Battery (MCCB) was used to assess cognitive performance (Nuechterlein et al., 2008). The battery consists of 10 tests measuring 7 cognitive and social cognitive domains, but for the present study, only the measures that generated Verbal Learning (Hopkins Verbal Learning Test), Speed of Processing (Trail Making: Test A; Brief Assessment of Cognition in Schizophrenia: Symbol Coding; Category Fluency: Animal Naming), Working Memory (Wechsler Memory Scale Spatial Span; Letter Number Span), and Reasoning and Problem-Solving (Neuropsychological Assessment Battery: Mazes) were administered. The domain scores and composite score were normed for age and education, yielding 4 T-scores and one summary score.

2.3.3 Beck Cognitive Insight Scale

The Beck Cognitive Insight Scale (BCIS) is a 15-item self-report measure that assesses a person’s self-reflectiveness and confidence in interpreting experiences, yielding self-reflectiveness and self-certainty subscales (Beck et al., 2004). The self-reflectiveness subscale consists of 9 items measuring patients’ objectivity and openness to feedback (e.g., other people can be more objective), and the self-certainty subscale consists of 6 items measuring patients’ certainty of their own beliefs and conclusions (e.g. doing something if it feels right). Participants are asked to rate the degree to which they agree with each statement on a 4-point Likert scale ranging from 0- (do not agree at all) to 3- (agree completely).

2.3.4 Columbia Suicide Severity Rating Scale

The Columbia Suicide Severity Rating Scale (C-SSRS) measures suicidal ideation and behavior (Posner et al., 2011). The C-SSRS ideation and behavior sub-scales showed strong convergent validity with established ideation and behavior scales (Posner et al., 2011). The scale distinguishes suicidal ideation and suicidal behavior domains including measures of any ideation, any history of ideation with plan and intent, suicide attempt, as well as planned vs. unplanned, and aborted attempts. The scale was administered in interview format by trained raters, and ratings were reviewed by the PI when discrepancies were identified. For the purposes of this study, we reduced the number of variables of interest to the broadest and most severe forms of ideation, passive ideation and ideation with plan and intent, as well as history of suicide attempt.

2.4 Statistical Analysis

Normality assumptions were first checked. We assessed for potential demographic and clinical covariates for suicidal ideation and attempt history with a criterion of p < 0.10 for inclusion as a covariate. We then contrasted participants based on history of ideation and attempt by the four MATRICS cognitive domains and BCIS subscales using ANCOVA. We calculated Cohen’s d effect sizes for these contrasts.

For variables that were significant predictors in univariate models, we completed mediating and moderating exploratory analyses. We first evaluated potential for BCIS scores to mediate the effect of MATRICS domains on prior suicidal ideation and attempt history using the Baron and Kenny mediation model (Baron and Kenny, 1986). We then evaluated whether the interaction of BCIS and MATRICS scores was significant to evaluate potential moderating effects. Missing data was minimal so we were able to use a complete case analysis. The alpha-level was set to 0.05 for hypothesis testing, and we used a Bonferroni correction for the exploratory analyses examining different MATRICS cognitive domains (4 domains; p = 0.013).

3. Results

3.1 Sample Characteristics (Table 1)

Table 1.

Sample Characteristics

Variable	Mean (SD) or %	Range
Age	50.6 (10.9)	23 – 77
Gender (% Female)	47.0%
Ethnicity
Caucasian	39.6%
African American	25.0%
Hispanic	26.8%
Asian	7.9%
Native American	0.6%
Education	12.0 (2.1)	2 – 18
Age of Onset	23.1 (7.3)	10 – 35
BRPS Total	43.3 (11.6)	24 – 74
MATRICS Total	37.1 (8.6)	14 – 59
MATRICS Subscales
Verbal Learning	37.5 (8.1)	21 – 58
Speed of Processing	33.4 (14.5)	0 – 64
Working Memory	33.3 (11.9)	3 – 66
Reasoning/Problem Solving	44.0 (9.3)	23 – 72
BCIS Total	5.0 (5.6)	−11 – 20
BCIS Subscales
Self-reflectiveness	11.9 (5.4)	0 – 27
Self-certainty	7.3 (3.8)	0 – 18
CSSRS (Lifetime)
Ever Wish to be dead	65.4%
Ever Active Suicidal Ideation w/Specific Plan and Intent	42.9%
Ever Actual Attempt	41.0%
Ever Aborted Attempt	19.5%
Ever Preparatory Acts	22.5%

On average, the sample was middle-aged, slightly more likely to be male (52.4%), and diverse in regard to ethnicity. The majority of participants had a diagnosis of schizophrenia (71%). The average level of education attained was a high school level. Participants had, on average, experienced psychotic illness for 30 years, and current symptom severity was in the mild-moderate range (Leucht et al., 2005). The average level of cognitive functioning was in the impaired range for each of the subscales, with t-scores indicating that participants were experiencing 1 to 1.5 standard deviations below normal (t = 50) on each of the 4 subscales. As such, the sample is broadly representative of chronic, treated, community dwelling, middle-aged and older adults with schizophrenia.

3.2 Demographic Variation of Suicidality Histories

On the Columbia Suicide Severity Rating Scale, the majority of patients had passive suicidal ideation (65.4%) and active ideation without plan or intent (56.8%). Slightly fewer had experienced suicidal intent with a plan (42.9%) and had attempted suicide (41.0%). For patients with at least 1 prior attempt, the average number of attempts per person was 4.3 (SD = 7.6; median = 2). A minority of participants endorsed histories of aborted attempts (19.5%) or preparatory acts (22.5%). Histories of passive ideation or suicide attempt were not associated with age, education, gender, ethnicity, or diagnosis. History of active ideation was associated with younger age (positive history, 48.4, SD = 10.7 years vs. negative history, 53.2, SD = 10.9 years, t₍₁₆₀₎ = 2.1, p = 0.033). In schizoaffective disorder, the rate of active ideation with specific plan/intent was higher (schizoaffective: 64.6% vs. schizophrenia: 33.0%, χ²₍₁₎ = 13.8, p < 0.001), as well as aborted attempt (schizoaffective: 29.2% vs. schizophrenia: 15.0%, χ²₍₁₎ = 4.3, p = 0.038) and preparatory acts (schizoaffective: 33.3% vs. schizophrenia: 17.5%, χ²₍₁₎ = 4.9, p = 0.027). Histories of passive or active ideation and suicide attempt were associated with higher current BPRS scores.

3.3 Suicidality Histories and MATRICS/BCIS Scores (Table 2)

Table 2.

Lifetime C-SSRS Scores, Cognitive Ability, and Cognitive Insight

	Never Wished to be Dead (N=56)	Ever Wished to be Dead (N=106)	t- value	p- value	Cohen’s d
	M (SD)	M (SD)
MCCB
Total Score	36.1 (9.6)	37.7 (8.0)	1.2	.280	−.174
Verbal Learning	36.4 (8.3)	38.2 (8.0)	1.9	.175	−.224
Speed of Processing	30.9 (15.4)	34.9 (14.0)	2.7	.104	−.266
Working Memory	32.3 (13.0)	34.1 (11.2)	0.8	.377	−.143
Reasoning	44.8 (9.9)	43.5 (8.9)	0.7	.392	.139
BCIS
Total Score	3.0 (5.2)	5.6 (5.7)	7.4	.007	−.462
Self-reflectiveness	10.0 (4.8)	13.0 (5.5)	12.1	.001	−.589
Self-certainty	7.0 (3.9)	7.5 (3.8)	0.6	.452	−.126
	No History of Active SI (N=92)	History of Active SI (N=69)
MCCB
Total Score	35.8 (8.7)	38.9 (8.3)	5.2	.025	−.363
Verbal Learning	36.0 (7.5)	39.8 (8.5)	9.4	.003	−.482
Speed of Processing	31.5 (14.8)	36.3 (14.0)	4.4	.038	−.334
Working Memory	31.5 (11.3)	36.2 (12.2)	6.3	.013	−.397
Reasoning	44.4 (9.7)	43.4 (8.6)	0.5	.497	.109
BCIS
Total Score	3.4 (5.4)	6.6 (5.4)	14.0	<.001	−.603
Self-reflectiveness	10.8 (4.8)	13.7 (5.6)	12.5	.001	−.562
Self-certainty	7.4 (3.7)	7.1 (4.0)	0.2	.620	.080
	No Actual Attempt (N=95)	Attempt (N=66)
MCCB
Total Score	36.1 (8.7)	38.7 (8.4)	3.5	.062	−.302
Verbal Learning	36.0 (7.7)	39.9 (8.2)	9.6	.002	−.493
Speed of Processing	31.7 (15.2)	36.2 (13.3)	3.8	.053	−.316
Working Memory	32.3 (11.2)	35.3 (12.7)	2.5	.113	−.252
Reasoning	44.5 (9.7)	43.3 (8.6)	0.6	.441	.125
BCIS
Total Score	4.1 (5.5)	5.7 (5.8)	3.1	.079	−.284
Self-reflectiveness	11.4 (5.1)	13.0 (5.6)	3.5	.064	−.299
Self-certainty	7.3 (3.9)	7.3 (3.8)	<0.1	.922	.016

Note: BCIS: Beck Cognitive Insight Scale; Active SI refers to any history of Ideation with a Specific Plan with Active Intent

A positive history of Active Suicidal Ideation was associated with higher MATRICS Total Score. Among MATRICS domains, the Verbal Learning T-Scores were higher among participants with lifetime Active Ideation and Suicide Attempt. Working Memory was positively associated with Active Ideation. Moreover, there were significant positive correlations with the lifetime number of Suicide Attempt and Verbal Learning T-Score (rho = 0.286, p < 0.001) and weakly with Speed of Processing (rho = 0.156, p = 0.049). Each of the significant associations remained significant at p < 0.05 after adjusting for BPRS Total Score. Better Verbal Learning was also associated with Preparatory Acts; however, lower Reasoning and Problem-Solving was associated with Aborted Attempts (Table 3, Supplemental Material). BCIS Total Score was associated with history of passive and active suicidal ideation (Table 2) and aborted attempts (Table 3, Supplemental Material). Among BCIS subscales, self-reflectiveness was associated with both history of passive and active suicidal ideation, as well as preparatory acts (Table 3, Supplemental Material). Self-certainty was unrelated.

3.4 Mediating and Moderating Effects of Cognitive Insight

There were no associations between cognitive domains and BCIS subscales (Pearson Rs ranging from −0.017 to 0.112, all p > 0.10), disallowing evaluation of insight as a mediating factor in the association between cognition and suicide. We examined potential interactive effects by examining the interaction between the BCIS score with MATRICS Verbal Learning and Working Memory scores (centered at their means), such that a significant interaction term would indicate that levels of insight would alter the association between cognition and suicide. There were no significant interactions.

4. Discussion

Our study sought to further clarify associations among lifetime suicidal ideation and behavior, cognitive ability, and cognitive insight. Our findings were consistent with prior reports indicating greater cognitive ability is associated with suicidal thoughts and behaviors (Delaney et al., 2012; Kim et al., 2003; Nangle et al., 2006). In particular, we found that better global cognitive ability, and verbal learning and working memory in particular, were associated with higher rates of prior active suicidal ideation, with better verbal learning also associated with the history of and higher number of lifetime suicide attempts. We also found that cognitive insight, and higher self-reflectiveness in particular, was associated with higher rates of lifetime passive and active ideation but not suicide attempts, and self-certainty was not associated with suicidal ideation or behavior. Contrary to our expectations, we found that cognitive ability and insight were largely independent predictors of suicidality, as these constructs were not associated with each other and there were no moderating relationships. Finally, cognitive ability was associated with sub-types of suicide attempts, with better verbal learning associated with higher rates of preparatory acts, and diminished reasoning and problem- solving associated inversely with aborted attempts.

The findings of this study should be interpreted with respect to its limitations. This was a retrospective, cross-sectional study in which current cognitive ability and insight were associated with prior suicidal thoughts and behavior. Although significant associations remained after adjusting for current symptom severity, we cannot infer any causal relationships between cognition, insight, and future risk of suicidal behavior. Additionally, all participants were outpatients at the time of evaluation. Therefore, our findings may not be generalizable to inpatients or untreated patients. Our sample was, on average, middle aged and older and had been diagnosed with schizophrenia for several decades. Recent literature indicates the period during the life course of highest risk of suicide in schizophrenia is during the first years of illness (Ventriglio et al., 2016). As such, these findings may not generalize to younger/first episode patients, and future research should examine variability in the role of cognitive ability and insight in suicidality across the life course. Although the C-SSRS is a validated screening measure for suicidality in the general population, it is not fully clear whether this instrument has different properties in samples with psychosis, and we were unable to cross-validate history of attempts with other sources of data (e.g., medical records related to attempts). While our findings were similar after adjustment for current global psychopathology, we lacked a specific measure of depressive symptoms, which may modify the impact of insight and cognition on suicide ideation and behavior (Kim et al., 2003).

Despite these limitations, our findings suggest potential factors to consider in risk profiles for suicide among people with schizophrenia. First, screening procedures for suicide risk may benefit from including cognitive ability and insight, which are not typically assessed in routine screening and are, indeed, in the opposite direction from that in the general population (Kosidou et al., 2014; Naifeh et al., 2016). Second, interventions for schizophrenia often target insight and cognitive functioning (Thorsen et al., 2014), and it is plausible that risk for suicide might increase as a result of improvements in insight or cognition. Given the high base rate and the plausible potential for an association between improvements in cognition and change in risk for suicide, standardized suicide risk screening in the context of clinical trials for schizophrenia are warranted. Longitudinal research should also address whether insight and cognitive ability are truly independent risks, as baseline correlational analysis may not reveal all of the subtleties of relationships between cognitive functioning and other variables. For example, although social cognition and neurocognition seem minimally correlated in naturalistic studies, cognitive remediation increases the rate with which social cognitive skills are learned in treatment studies (Lindenmayer et al., 2013). Third, predictors of ideation differ in some respects from those of attempts. Therefore, it is important to examine risk factors for ideation and behavior separately (Klonsky and May, 2013). In our sample, cognitive insight was associated with ideation but not attempts. One speculation is that greater cognitive insight creates a vulnerability to suicidal thoughts but that other variables may play a role in the transition to suicidal behavior, consistent with psychological theories of suicide (e.g., interpersonal theory of suicide; Van Orden et al., 2010) in which several constructs are posited to converge to lead to suicidal behavior. Finally, the current research suggests a need to understand mechanisms of the influence of cognitive ability and insight on suicide risk in schizophrenia.