Neurocognitive Suicide and Homicide Markers in Patients with Schizophrenia Spectrum Disorders: A Systematic Review

Suicide and homicide are considered important problems in public health. This study aims to identify the cognitive performance of suicidal and homicidal behaviors in people with schizophrenia spectrum disorders, as well as examining whether there are shared neuropsychological mechanisms. A systematic review of the recent literature was carried out from September 2012 to June 2022 using the Medline (via PubMed), Scopus, Embase, and Cochrane databases. Among the 870 studies initially identified, 23 were finally selected (15 related to suicidal behaviors and 8 to homicidal behaviors). The results evidenced a relationship between impairment of cognitive performance and homicidal behavior; meanwhile, for suicidal behaviors, no consistent results were found. High neuropsychological performance seems to act as a protective factor against violent behavior in people with schizophrenia spectrum disorders, but not against suicidal behavior; indeed, it can even act as a risk factor for suicidal behavior. To date, there is insufficient evidence that shared neurocognitive mechanisms exist. However, processing speed and visual memory seem to be affected in the presence of both behaviors.


Introduction
Suicide and homicide are considered important problems in public health. Suicide is defined as a self-inflicted act that aims to cause death voluntarily [1], whereas homicide is defined as the act of killing a person without premeditation or another aggravating circumstance. According to the World Health Organization (WHO) [2], suicide has increased considerably in recent decades, reaching truly worrying levels. Approximately one million people die every year due to suicidal behavior (one person every forty seconds) and the number of attempts is estimated to be twenty times higher [3]. Death by suicide is the third leading cause of violent death among people aged from 15 to 44, behind only traffic accidents and homicides [4]. In the case of homicide, the total number of cases worldwide has also increased in recent decades, registering 464,000 homicides in 2017. Europe is the continent with the second lowest homicide rate, with 3 deaths per 100,000 inhabitants (22,009 total), 5% of the world total. In Spain, the average is less than 1% [5].
Biological, genetic, psychological, and environmental markers have been highlighted as factors related to suicidal and homicidal behaviors [6,7]. From a psychological point of view, the most important risk factor is the presence of psychiatric disorders [1,8], especially schizophrenia [9,10]. Schizophrenia is defined as a serious mental disorder characterized by an abnormal perception of reality, disorganized thought, maladaptive behaviors, and negative symptoms that affect the cognitive, emotional, and social domains [11].
(a) What is the relationship between cognitive functions and suicide and homicide behaviors in patients with schizophrenia spectrum disorders? (b) Does the performance of cognitive functions predispose to these behaviors? (c) Are there shared neurocognitive mechanisms?

Materials and Methods
This systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [36].

Eligibility Criteria
The inclusion criteria were as follows: articles that (1) were published in a peerreviewed journal; (2) were published in English; (3) included people with schizophrenia, brief psychotic disorders, not otherwise specified (NOS) psychosis, or schizoaffective, schizophreniform, and delusional spectrum disorders diagnosed as per the DSM-IV, DSM-IV-SCID, or DSM-V criteria; (4) included at least one neuropsychological task; and (5) compared at least two groups of patients, one of which comprised patients with a history of suicide attempts (defined as any act carried out with a certain intent to die or to put one's life in danger [37]), or suicidal ideation (defined as a range of contemplations, wishes, and preoccupations related to death and suicide [38]), or with a history of homicide (when there is a dead person and the cause of death can be attributed to another person [39]), or violence (violent acts committed against others which cause or are intended to cause physical harm to the victim [40]).

Information Sources
A systematic search was carried out using the Medline (Pubmed), Scopus, Embase, and Cochrane databases. Furthermore, to identify the potential additional studies not contained in the electronic databases, we performed a manual search of the bibliographical references of retrieved studies and previous systematic reviews [7,8,23].

Search Strategy and Selection Process
We performed a systematic literature search of all clinical trials, as well as cohort, case-control, and cross-sectional human studies published in English from 1 September 2012 to 1 June 2022. The search processes were undertaken blind by M.T.-F. and M.B.-B. in the Medline (via Pubmed), Scopus, Embase, and Cochrane databases. The MeSH terms "schizophrenia" and "psychotic disorders" were combined with the MeSH terms "violence", "homicide", "suicide", and "attempted suicide" in all fields. Then, these were combined with the MeSH terms and TIAB terms of neurocognitive pathways: "cognition", "neuropsychology", "neuropsychological test", "executive function", "decision making", "problem solving", "prefrontal cortex", "neuropsychological functions", "executive functioning", and "executive performance". Tables 1 and 2 show the full search strategy used in the review.
Each study identified through the literature search was independently evaluated by two researchers (M.T.-F. and M.B.-B.). In the identification phase, the total results were added to the Mendeley platform, and duplicates were removed. Afterwards, in the screening phase, the authors identified articles that met the inclusion criteria by validating their titles and abstracts. Subsequently, following a detailed reading of full-text articles, the authors determined the selections. In the inclusion phase, the chosen articles were identified and prepared for data extraction. The complete selection processes divided into suicide and homicide studies are reflected in Figures 1 and 2.

Included
Reports sought for retrieval (n = 1) Reports not retrieved (n = 0) Figure 1. Flowchart of homicide studies. * it was added one study which appeared in the homicide search strategy but pertained to the suicide criteria.

Figure 2.
Flowchart of homicide studies. * One study was subtracted which pertained to the suicide criteria.

Identification of studies via databases and registers
Identification Screening Included Figure 2. Flowchart of homicide studies. * One study was subtracted which pertained to the suicide criteria. Table 1. Search strategy used for the review.

Data Collection Process
All data were independently extracted by the authors (M.T.-F. and M.B.-B.) and the description and characteristics of each study were considered (authors, year of publication, country, objectives, average age, sample number, and type of study) (Tables 3 and 4), and the design and results (diagnoses, cognitive domains and tasks, and the main results) were obtained (Tables 5 and 6). The statistical measures that were taken into consideration were: p-value and effect size (eta partial square, Cohen's d, and odd ratio).

Risk of Bias Assessment of Individual Studies
The quality of the included studies was assessed based on the Appraisal tool for Cross-Sectional Studies (AXIS) [41], which was developed for use in appraising observational cross-sectional studies. In addition, we used the JBI checklist for longitudinal cohort and case-control studies to assess their methodological quality and to determine the extent to which a study addressed the possibility of bias in its design, conduct, and analysis (Appendices A and B). It was carried out independently by two authors (M.T.-F. and M.B.-B.). In general, the quality of the cross-sectional and longitudinal studies included was quite good (Tables A1-A3). Disagreements were resolved by a third author (M.S.-S.).

Synthesis Methods
The systematic literature review identified a total of 23 articles. The results were grouped based on whether the studies explored neurocognitive functioning in homicidal (8) or suicidal (15) individuals with schizophrenia spectrum disorders.
Regarding the suicide research, we added one study that was identified through the references of a previous systematic review. In addition, there was one study which appeared in the homicide search strategy that pertained to the suicide criteria. That is the reason why one is subtracted and added to the other group.

Certainty Assessment
The studies were assessed for their methodological rigor using criteria established by the Oxford Centre for Evidence-based Medicine levels of evidence for diagnostic studies. A summary of the methodology rigor is calculated based on the grade of level evidence for each study (from level 1 "systematic review, validating cohort studies" to level 5 "expert opinion") [42].

Selection of Studies: Suicide
A total of 599 studies were identified (598 through databases and 1 through another source). After eliminating duplicates, 391 studies remained. First, the titles and abstracts were screened, and 49 studies were excluded according to the different exclusion criteria. Second, 342 full-text articles were assessed for eligibility, and 328 of them were eliminated based on the different exclusion criteria. One study related to suicide was included in the systematic literature search for homicide; for this reason, the study was added to the total number of suicide articles. Finally, 15 articles related to suicide were included in this review with a total sample of 5118 patients (Figure 1).

Selection of Studies: Homicide
A total of 271 studies were identified (all obtained through databases). After eliminating duplicates, 157 studies were left. First, the titles and abstracts were screened and 43 were excluded according to the different exclusion criteria. Then, 114 full-text articles were assessed for eligibility; 105 of them were eliminated based on the different exclusion criteria. The final number of articles was 9; however, one of these was related to suicide, so it was included into the final list of suicide articles. Finally, 8 articles related to homicide were included in this review, with a total sample of 3412 patients ( Figure 2). Tables 3 and 4 show the characteristics of the articles included. Of the suicide studies, 63.9% (n = 3268) of participants were men and 30.1% (n = 1540) were women (N = 5118), while the sex of the participants was not specified in the remaining 6.1% (n = 310) of the sample. The mean age of the participants was 36.9 years old, and age was not specified in the remaining 6.6% (n = 1) of the sample. In the case of homicide studies, 65.6% (n = 2240) of the participants were men and 34.4% (n = 1172) were women (N = 3412). The mean age of the participants was 38.7 years old, and age was not specified in one article, which corresponds to 11.6% (n = 398) of the sample.

Risk of Bias in Studies
Appendices A and B summarized the risk of bias for individual studies. Overall, the quality of studies was rated as medium-high in 15 of the 19 examined studies (78.9%), with only 4 studies rated as medium-low quality or doubtful (21.1%) [13,51,64]. A summary assessment was calculated based on the number of items evaluated as a Yes in the AXIS tool. Values above average were interpreted as medium-high quality and values below average were interpreted as medium-low quality. For longitudinal studies, the risk of bias was rated as low, with average values of 9 in items evaluated with the JBI tool.

Certainty of Evidence
The results of the methodological rigor by using the criteria set established by the Oxford Centre for Evidence-Based Medicine are presented in Figure 3. A limited number of highly methodologically rigorous studies were found, with only two homicide longitudinal studies [55,59] and two suicide longitudinal studies [44,45] classified within level 1b. No level 2 studies were found in either homicide or suicide. Most homicide and suicide studies were classified within the level of evidence 3b (moderate). Of the three levels, 5 homicide studies and 11 suicide studies were designed as non-randomized controlled cohort or local non-random samples without follow-up (Table 3). In addition, one homicide study [64] and two suicide studies [13,51] were found in the four levels as they were considered poor quality prognostic cohort studies. study [64] and two suicide studies [13,51] were found in the four levels as they were considered poor quality prognostic cohort studies.

Neurocognitive Functioning in Relation to Homicidal or Suicidal Behavior in Schizophrenia
The systematic literature review identified a total of 23 articles. Eight of these studies explored neurocognitive functioning in homicidal patients with schizophrenia, and fifteen analyzed neurocognitive performance in suicidal patients with schizophrenia. Tables 6  and 7 show the diagnoses (100% of the included articles used DSM-IV or DSM-IV-SCID as diagnostic criteria), cognitive domains and task used in the investigations, and the main results of the included articles on suicide and homicide.

Neurocognitive Functioning in Relation to Homicidal or Suicidal Behavior in Schizophrenia
The systematic literature review identified a total of 23 articles. Eight of these studies explored neurocognitive functioning in homicidal patients with schizophrenia, and fifteen analyzed neurocognitive performance in suicidal patients with schizophrenia. Tables 6 and 7 show the diagnoses (100% of the included articles used DSM-IV or DSM-IV-SCID as diagnostic criteria), cognitive domains and task used in the investigations, and the main results of the included articles on suicide and homicide.
Just one of the studies showed that homicidal patients outperformed non-homicidal patients in working memory (p = 0.04; Cohen's d = 0.61) (BACS-Digit Sequencing Test) and executive function (p < 0.001; Cohen's d = 1.00) (BACS total score) [61]. Finally, one study did not show statistical differences between the groups [57].

Shared Neuropsychological Impairments in Suicidal and Homicidal Patients
The results showed that there are cognitive domains shared that significantly affect homicidal and suicidal behaviors. In Table 7, it can be seen how a worse performance in processing speed [44,58], and visual memory [45,60] could explain both violent behaviors (homicide and suicide).
This section may be divided by subheadings. It should provide a concise and precise description of the experimental results, their interpretation, as well as the experimental conclusions that can be drawn. To explore predictors of suicidal behavior, adjusting the analyses for a set of sociodemographic, clinical, and neurocognitive variables. Additionally, to examine potential long-term differences in clinical measures and neuro-cognitive functioning between patients who undertook suicidal acts and those who did not over the follow-up period.    There were no significant statistical differences between the suicidal and non-suicidal groups. RBANS (repeatable battery for the assessment of neuropsychological status): immediate memory, list learning, story memory, attention, digit span, coding, language, picture fluency, visuospatial, figure copy, line orientation, delayed memory, list recall, story recall, figure recall, and list recognition There were no significant statistical differences between attempters and non-attempters. Total score of RBANS (repeatable battery for the assessment of neuropsychological status): immediate memory (list learning and story memory tasks); visuospatial/constructional (figure copy and line orientation tasks); language (picture naming and semantic fluency tasks); attention (digit span and coding tasks); delayed memory (list recall; story recall; figure recall; list recognition tasks).
No significant statistical differences were found between those with suicidal ideation and those with no suicidal ideation in terms of performance on the RBANS test (total score and cognitive domains). Schizophrenia (DSM-IV-SCID).
Total score of RBANS (repeatable battery for the assessment of neuropsychological status): immediate memory (list learning and story memory tasks); attention (digits span and coding tasks); language (picture naming and semantic fluency tasks); visuospatial/constructional (figure copy and line orientation tasks); delayed memory (list recall, story recall, figure recall, and list recognition tasks).
Both groups with schizophrenia showed significantly lower cognitive scores on RBANS total, immediate memory, attention, delayed memory (all p < 0.001) and language (p = 0.002), than healthy controls. However, when suicide attempters were compared with non-attempters within the schizophrenia group, attempters performed better only on the attention domain (p = 0.025; η P 2 = 0.49).

Patients with suicidal behaviors presented worse scores in visual memory
Suicide attempters presented lower composite summary scores in executive function (p < 0.05; η P 2 = 0.10), problem solving skills (p < 0.01; η P 2 = 0.14), and decision-making (p < 0.01; η P 2 = 0.19) compared to non-attempters. However, after controlling for the effects of alcohol dependence, only decision-making showed significant differences. The attempters scored significantly better in executive function on both TMT A (p = 0.026) and TMT B (p = 0.012) than those who had never attempted. Processing speed was significantly (p = 0.046) more impaired in attempters. No significant differences were found in the other domains.
Total score of RBANS (repeatable battery for the assessment of neuropsychological status): immediate memory (list learning and story memory tasks); visuospatial/constructional (figure copy and line orientation tasks); language (picture naming and semantic fluency tasks); attention (digit span and coding tasks); delayed memory (list recall; story recall; figure recall; list recognition tasks).
There were no significant statistical differences between the suicide attempters and non-attempters. There were no significant statistical differences between the suicidal and non-suicidal groups.

Ref. Year Sample Groups Diagnoses Cognitive Domains/Task Main Results
The Brief Assessment of Cognition in Schizophrenia (BACS): verbal memory (digit sequencing test); working memory (digit sequencing test); motor speed (token motor test); verbal fluency (symbol coding test); attention; executive functioning (BACS total score)).
The violent group performed significantly better than the control group on working memory (p = 0.047; Cohen's d = 0.61) and executive function (p < 0.001; Cohen's d = 1.00).

Discussion
The present systematic review provides the most recent evidence on cognitive functioning in people with schizophrenia spectrum disorders who have committed suicidal and homicidal acts. Regarding the published studies pertaining to suicidal behaviors, the results were contradictory. Four of them [43][44][45]53] conclude that cognitive performance impairments predispose patients to suicidal ideation and suicide attempts. Among the significant cognitive domains that emerge in this research, three of them (working memory, global cognitive functioning (GCF), and decision making) are related to executive functions. These cognitive deficits, especially those related to working memory, have an impact on executive control, decision making, and higher reasoning skills, which are protective factors against suicidal ideation and attempts [53,65]. Additionally, recent studies on brain dysfunctions have reported reduced neural activity in suicide attempters who were exposed to fMRI (functional magnetic resonance imaging) [66,67].
Despite the heterogeneity shown by the selected articles, one trend that emerges shows cognitive performance to be a significant risk factor for suicidal behavior in patients with mental illness [13,46,51,54,56,62]. Previous studies concur that a higher educational level [68], higher IQs [69], better executive performance [24], and greater levels of insight [70] predispose patients to engage in suicidal behaviors. This fact could be explained by a patient's greater understanding of the illness phase, since this could increase their awareness of their limitations and personal decline; additionally, it may raise their likelihood of suffering from positive symptoms such as depression [56]. In addition, there is a consensus that a better performance on cognitive functions, especially in executive domains such as planning, predispose patients to be better able to formulate plans and initiate goal-directed behavior to suicidal acts [71]. Neuroimaging studies also reaffirm this fact; they show greater activation in the prefrontal zone, an area responsible for the organization, planning, and initiation of behavior [29].
Moreover, in contrast to the results for suicide, homicide studies showed stronger findings. Six of the eight selected articles evidenced a relationship between a cognitive performance impairment and homicidal behavior [55,[58][59][60]63,64]. These findings are supported by a recent meta-analysis across 43 studies [72], which concludes that violent behavior was related to structural and functional deficits in the prefrontal cortex. Alterations in the right orbitofrontal and cingulate anterior areas are associated with a lack of emotional processing, difficulties in social behavior, and poor decision making [73,74]. Deficits in the left dorsolateral cortex revealed a lower performance in attention, flexibility, planning, and impulse control [75][76][77].
Some studies reported that detrimental function could be due to reduced gray [78] and white matter volume [79] in the prefrontal cortex, temporal lobe, and superior temporal gyrus [80], as well as reduced blood oxygenation in the amygdala [81] and significantly lower activation in the frontal basal cortex area [82]. Furthermore, another imaging study examined processing speed impairments and found that this deficit was associated with damage in the cerebellar-thalamo-cortical circuit [83].
In addition, the systematic review showed that the shared mechanisms between both behaviors are not conclusive. Although there are two cognitive domains that appear as risk factors in both groups (processing speed and visual memory), there is another function that acts as a risk and protective factor depending on the behavior (working memory). Future research should compare the cognitive performance of the two groups with the same evaluation protocol, as it would be easier to establish relationships in this way.
There are several limitations of the current study that must be highlighted. First, the populations studied were highly heterogeneous in terms of their sociodemographic and clinical characteristics, especially in the primary diagnosis [44][45][46][47]53] and gender, since some studies only included male patients [43,46,61]. Additionally, there were missing sociodemographic data [13] and some studies included small sample sizes. Second, not only the gender or age of the sampled participants, but also the presence of polysubstance abuse (alcoholism, etc.), the duration of the illness, medication, psychotic symptoms, and hallucinations and other comorbidities (antisocial personality disorder) were variables not considered in some of the studies, despite these factors exerting a significant influence on the results. For example, Bulgari et al. [57] and Adan et al. [43] found significant results between the groups, but after controlling for these variables, the evidence was no longer significant. Third, the definitions of suicidal and homicidal behaviors were also heterogeneous, as well as the neuropsychological assessments used to evaluate the cognitive functions. For example, there were studies which used the WCST to measure abstract reasoning/problem solving [43], while others used it to evaluate executive function [46]. Moreover, some authors used a single neuropsychological test whereas others preferred a specific large battery. Fourth, the methodologies of the studies reviewed also varied (i.e., the tools used to assess suicidal and homicidal behaviors). Finally, this systematic review was not registered as a protocol. It will be considered for future research.

Conclusions
This article has brought to light the recent evidence on the relationship between cognitive functioning and suicidal and homicidal behaviors in people with schizophrenia spectrum disorders. These results could be used to give priority not only to psychological and pharmacological factors, but also to the prevention and treatment of neuropsychological domains. Our results suggest that cognitive function is affected in violent or homicidal patients with schizophrenia, showing a detrimental effect in cognitive functions such as learning, fluid intelligence, planification, visual memory, and processing speed. In particular, the suicide studies showed contradictory results. Therefore, no strong conclusions can be established regarding the influence of brain alterations or neuromarkers in suicide behaviors, and deeper investigations are required. When looking for shared pathways between suicidal and homicidal patients with schizophrenia spectrum disorders, there were statistically significant results with processing speed and visual memory. Nevertheless, future research that directly compares the neurocognitive markers of suicidal and homicidal risk are needed. Finally, the main contribution of this review is that it tries to understand if there are shared cognitive mechanisms which can help to identify and predict risk areas in crime intervention and treatment programs. Health institutions should prioritize not only psychological and pharmacological treatments, but also neuropsychological training.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A
Appraisal tool for Cross-Sectional Studies (AXIS) about neurocognitive suicide and homicide markers in patients with schizophrenia spectrum disorders.

Appendix B
JBI checklist for longitudinal cohort and case-control studies to assess their methodological quality and to determine the extent to which a study addressed the possibility of bias in its design, conduct, and analysis. Items: 1. Were the two groups similar and recruited from the same population? 2. Were the exposures measured similarly to assign people to both the exposed and unexposed groups? 3. Was the exposure measured in a valid and reliable way? 4. Were confounding factors identified? 5. Were strategies to deal with confounding factors stated? 6. Were the groups/participants free of the outcome at the start of the study (or at the moment of exposure)? 7. Were the outcomes measured in a valid and reliable way? 8. Was the follow up time reported and sufficiently long for outcomes to occur? 9. Was the follow-up process completed, and if not, were the reasons that no follow up occurred described and explored? 10. Were strategies to address incomplete follow up utilized? 11. Was appropriate statistical analysis used?