Implicit cognition tests for the prediction of suicide risk of non-suicidal self-injury in individuals: A systematic review

Abstract

BACKGROUND

Suicide risk assessment in individuals with non-suicidal self-injury (NSSI) is clinically challenging; implicit cognition tests may address limitations of explicit measures but lack systematic synthesis.

AIM

To synthesize evidence on the validity of implicit cognition tests for suicide risk assessment in individuals with NSSI.

METHODS

A systematic search of 6 databases (PsycINFO, PubMed, MEDLINE, EMBASE, CNKI, Wanfang Data) up to January 20, 2023, was conducted. Eligible studies were quantitative; effect sizes [odds ratio (OR), Cohen’s d] with 95% confidence intervals were reported. The risk of bias was evaluated by the Newcastle-Ottawa Scale (EMBASE, CNKI, Wanfang Data) up to January 20, 2023, was conducted. Eligible studies were quantitative; effect sizes (OR, Cohen’s d) with 95% confidence intervals were reported. The risk of bias was evaluated by the Newcastle-Ottawa Scale.

RESULTS

Nine studies were included. Implicit tests showed robust retrospective validity (Cohen’s d = 0.71, OR = 1.87) but time-dependent prospective validity (effective ≤ 3 months, inconsistent ≥ 6 months). They complemented but did not replace explicit measures.

CONCLUSION

Implicit cognition tests are valuable adjuncts for short-term suicide risk assessment in adolescents with NSSI.

Key Words: Non-suicidal self-injury; Suicide; Implicit cognition assessment; Explicit assessment measures; Adolescents; Systematic review

Core Tip: This systematic review explores the validity and reliability of implicit measures for suicide risk assessment (a clinical challenge). From 189 initial articles, 9 were included. The Death/Suicide Implicit Association Test and Suicide Stroop task were common tools. Results showed these measures are valid for assessing past and future suicidal thoughts/behaviors, with significant retrospective/prospective associations.

INTRODUCTION

Non-suicidal self-injury (NSSI) is defined as the direct, intentional damage to one’s own body tissues (e.g., cutting, burning, hitting) that is not motivated by suicidal intent and is not socially or culturally sanctioned[1]. Formal recognition of NSSI as a condition requiring further investigation reflects NSSI was formally recognized as a condition requiring further study, reflecting its growing recognition as a distinct and pressing psychological concern - particularly among adolescents - within the global mental health community[2]. Over the past decade, the prevalence of NSSI has increased steadily, drawing widespread attention from both the medical field and society at large due to its significant impact on individual well-being and public health. Epidemiological studies estimate that the prevalence of NSSI ranges from 15.9% to 20.5% in adolescents and from 2.5% to 5.4% in adults, highlighting its disproportionate burden on younger populations[3].

Cumulative research has established that long-term, repeated NSSI is associated with a cascade of adverse mental health outcomes, including the development of maladaptive coping strategies, increased risk of suicidal behavior, and engagement in harmful social behaviors. These consequences impose substantial psychological, emotional, and economic burdens on affected individuals, their families, and broader society. Beyond its inherent risks, NSSI has been consistently identified as a robust prospective risk factor for future suicidal behavior[4]. Concurrently, as the incidence of NSSI rises among adolescents, suicide has become the second leading cause of death among young people globally[5]. Therefore, self-injury and suicide are not only important mental health problems that harm the physical and mental health of adolescents, but also major public health problems among adolescents worldwide.

Despite its clinical significance, NSSI poses unique challenges for early identification and assessment. Its underlying mechanisms remain poorly understood, and it frequently co-occurs with other mental health conditions (e.g., depression, anxiety, borderline personality disorder), further complicating diagnostic efforts[6]. Currently, clinical assessment of NSSI relies on a limited toolkit, leading to unacceptably high rates of missed diagnosis in clinical settings. This diagnostic delay often results in delayed intervention, and individuals with NSSI who do not receive standardized assessment and treatment in the early stages typically exhibit poorer treatment outcomes and persistent impairments in social functioning later in life[7]. A major limitation of current assessment practices is their heavy reliance on observational methods and self-report measures. While these approaches are widely used, they are subject to significant biases: Intentional concealment: Patients may deliberately hide suicidal ideation or NSSI behaviors to avoid unwanted interventions, such as parental notification or involuntary hospitalization. Limited self-awareness: Adolescents, in particular, often lack the metacognitive capacity to accurately describe the unconscious or automatic cognitive processes that drive their self-injurious behaviors[8,9]. Consequently, traditional assessment methods (e.g., self-report questionnaires, clinical observation) are relatively insensitive to implicit cognition - the automatic, unconscious mental associations and attitudes that influence behavior independently of conscious awareness[10]. It is critical to distinguish implicit cognition from implicit memory: Implicit memory involves unconscious recall of past events or procedural skills (e.g., riding a bike without conscious effort to recall how to balance), whereas implicit cognition focuses on the strength of associations between abstract concepts (e.g., “self” and “death”, “NSSI” and “coping”)[11].

To address these limitations, researchers have increasingly turned to self-report-independent cognitive tasks to assess NSSI and suicide risk. These tasks include projective tests, biographical analysis, reaction time paradigms, situational tests, and Implicit Association Tests (IAT)[12-14]. Among them, IAT has emerged as a preferred tool in related research due to its strong psychometric properties (i.e., good reliability and validity) and ease of administration[15]. Over the past decade, IAT has been increasingly applied to the assessment and prediction of mental and behavioral health outcomes[16,17]. Building on this work, Nock and Banaji developed the Self-Injury IAT (SI-IAT) to specifically assess implicit cognition related to NSSI. This measure evaluates reaction times during rapid judgments of NSSI-, suicide-, death/Life-, and self-related words, with the strength of associations between these concepts inferred from reaction time differences[18]. This assessment process does not rely on awareness or introspection and is therefore not susceptible to self-reported bias. Several studies have demonstrated promising findings regarding the utility of the SI-IAT[19]. In a sample of adolescents and adults recruited from communities and clinics, SI-IAT scores were found to be significantly more associated with suicide-related risk than other known risk indicators, including a history of previous suicide attempts[20]. Despite this progress, research on implicit cognition in NSSI remains limited, and findings are inconsistent: Some studies have found that the SI-IAT exhibits good reliability and validity, with significant differences in implicit NSSI-related cognition between individuals with and without a history of NSSI[21]. These studies also note that individuals with NSSI show automatic, biased processing of stimuli related to death, suicide, or cutting[22]. Other research has indicated that implicit attitudes toward self-injury are malleable (i.e., subject to change over time). For example, the strength of the association between implicit cognition and NSSI varies with the recency of self-injurious behavior (e.g., time since the last NSSI episode) and may be moderated by the frequency or severity of NSSI[23].

Regarding the predictive validity of the IAT, several studies have demonstrated its ability to prospectively predict the frequency and severity of future NSSI: Multiple investigations have shown that the SI-IAT can predict NSSI frequency at the 6-month follow-up, with higher baseline SI-IAT scores associated with a greater likelihood of NSSI recurrence[24]. In contrast, another study found that the SI-IAT only predicted NSSI frequency over a 2-week period and had no predictive utility for NSSI outcomes at the 3-month follow-up[5,16]. This suggests that the SI-IAT may function as a short-term risk marker and that its predictive validity is specific to certain types of NSSI (e.g., cutting) but not others (e.g., burning, hitting)[25]. A growing body of research suggests that implicit recognition of death may be a meaningful marker of suicide risk, and implicit self-identification with death is hypothesized to develop and strengthen over time as individuals experience more suffering and consider more extreme solutions to end that suffering[6,8]. However, critical questions remain unresolved: Specifically, the magnitude of the SI-IAT’s predictive effect and the optimal follow-up period for predicting suicidal behavior in individuals with NSSI have not been established.

Given the inconsistencies, there is an urgent to systematically explore the value of implicit cognitive testing in identifying and predicting suicidal behavior in order to clarify and standardize the relevance of its use in clinical and research settings. To date, no systematic review has focused exclusively on the use of implicit cognitive tests for suicide risk assessment in this population. The present study addresses this gap by synthesizing the available evidence on the validity and reliability of implicit cognition tests for suicide risk assessment in individuals with NSSI[12].

MATERIALS AND METHODS

Study design and registration

The present systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines - a widely accepted framework for ensuring transparency and rigor in systematic review methodology. The review protocol was registered in the PROSPERO database (registration number: CRD420251153284) to ensure methodological transparency and prevent outcome reporting bias.

Search strategy

A comprehensive literature search was conducted across six electronic databases: PsychINFO, PubMed, MEDLINE, EMBASE, CNKI, and Wanfang Data. The search covered all articles published from the inception of each database up to January 20, 2023. The search strategy combined keywords related to NSSI and implicit cognition, using the following syntax: (NSSI OR non-suicidal self-injury) AND (implicit OR implicit cognition OR implicit association test OR IAT OR Stroop task). In addition to database searches, a hand search of reference lists from included studies and relevant systematic reviews was performed. This step is critical for minimizing publication bias and ensuring a comprehensive literature synthesis.

Eligibility criteria

Inclusion criteria: (1) Population: Studies focusing on individuals with a confirmed diagnosis or self-reported history of NSSI; (2) Intervention/exposure: Studies that utilized at least one implicit cognition test (e.g., IAT, Stroop task) to assess NSSI- or suicide-related cognition; (3) Study design: Empirical studies with a quantitative design, including case-control studies, cohort studies, cross-sectional studies, and randomized controlled trials that reported relevant outcomes related to NSSI or suicide risk; and (4) Availability: Full-text articles available for review; abstract-only studies were excluded.

Exclusion criteria: (1) Lack of primary data: Review articles, meta-analyses, commentaries, editorials, and case reports (which do not provide original data); (2) Unobtainable data: Studies for which primary data were not reported and could not be obtained through correspondence with the authors; (3) Irrelevant assessment tools: Studies that did not use implicit cognition tests (e.g., studies relying solely on self-report questionnaires or observational measures); (4) Irrelevant population: Studies focusing on samples with suicidal ideation or suicide attempts but no history of NSSI, or samples with other mental health conditions without NSSI; and (5) Duplicate publications: Multiple publications from the same study (only the most comprehensive or recent publication was included).

Study selection process

Two reviewers (Ji YJ and Zheng Q) independently screened the titles and abstracts of all retrieved articles against the eligibility criteria. Articles deemed potentially eligible or for which eligibility was unclear were advanced to full-text review. Full-text review: The same two reviewers independently assessed the full text of each article advanced from the title/abstract stage. Discrepancies were resolved through discussion with a third reviewer (Zhang CY) to ensure consensus. The intraclass correlation coefficient was calculated to assess inter-rater reliability for study selection. An intraclass correlation coefficient ≥ 0.80 was considered indicative of excellent reliability.

Study selection and data extraction

Two independent reviewers (Ji YJ and Zheng Q) conducted the methodological quality assessment of all eligible studies. Any discrepancies arising during the assessment process were resolved through in-depth discussion between the two reviewers. The assessment covered multiple dimensions, including study design, risk of bias, and the comprehensiveness of reporting. Data identification, verification, and extraction were performed by two separate authors to ensure objectivity. A pre-designed data extraction form was used to collect the following information: Author name(s), year of publication, country where the study was conducted, research objectives, total sample size, detailed sample characteristics (including mean age and gender distribution), type of implicit cognition test employed, primary outcome measures, temporal scope of the study (i.e., retrospective or prospective association), and key results and main conclusions.

Quality assessment

The methodological quality of included studies was evaluated using the Newcastle-Ottawa Scale, a validated tool for assessing the quality of non-randomized studies (e.g., case-control, cohort studies). Discrepancies in quality assessment were resolved through discussion with a third reviewer (Zhang CY).

RESULTS

The initial search returned 189 results. Through a structured study selection workflow involving initial screening, in-depth full-text review, and confirmation of eligibility, 9 articles were finally incorporated into the systematic review (Figure 1).

Figure 1 PRISMA flow diagram of the study selection process. The initial search revealed 189 results. After screening, full-text review and study selection, 9 articles were finally included in the review.

Characteristics of the reviewed studies

The characteristics of the 9 included studies are summarized in Table 1, with key observations as follows: All studies were conducted in the United States, which reflects a potential Western bias in the current literature; the sample sizes of these studies ranged from 71 to 662 participants, and the median sample size was 233; all studies focused on adolescents, whose mean ages ranged from 14.8-22.5 years, and the majority of participants were female, accounting for 57.7%-84.5%, a distribution consistent with epidemiological data indicating higher NSSI prevalence among females; in terms of setting, three studies recruited participants from community settings, while six studies focused on hospitalized patients (e.g., psychiatric inpatients); regarding implicit cognition tests, the SI-IAT was the most commonly used measure, utilized in 8 studies, followed by the death IAT and suicide IAT, each used in 1 study, and all these tests were computerized and scored based on reaction time differences; most studies included a control group of individuals without NSSI or suicidal ideation, and some studies also included comparisons to individuals with suicide attempts (Table 1).

Table 1 Characteristics of included studies.

Ref.	Country	Aims of study	Sample	Age, mean ± SD	Gender distribution (female)	Test
Nock and Banaji[40], 2007	United States	Assessment of self-injurious thoughts using a behavioral test	n = 92: n (without NSSI) = 36; n (recent NSSI) = 53	17.10 ± 1.92	76.40%	SI-IAT
Brausch et al[26], 2022	United States	Examining NSSI features as motivational moderators in the relationship between hopelessness and suicide ideation	n = 420 (history of NSSI)	18.92 ± 1.38	80%	SI-IAT
Nagy et al[29], 2021	United States	An experimental investigation of the effects of self-criticism and self-compassion on implicit associations with NSSI	n = 233: n (with NSSI) = 76; n (without NSSI) = 161	19.37 ± 2.12	74.2%	SI-IAT
Powers et al[41], 2021	United States	Explicit and implicit measurement of NSSI in the prediction of concurrent and prospective self-injury	n = 421 (time 1); n = 324 (time 2)	18.92 ± 1.38 (time 1), 18.98 ± 1.38 (time 2)	82.7% (time 1), 82.1% (time 2)	SI-IAT
Glenn et al[32], 2016	United States	Implicit cognition about self-injury predicts actual self-injurious behavior: Results from a longitudinal study of adolescents	n = 662: With NSSI (n = 302), without NSSI (n = 360)	13.14 ± 0.7	57.7%	SI-IAT
Millner et al[21], 2018	United States	Implicit cognition as a behavioral marker of suicide attempts in adolescents	n = 71	14.8 ± 1.5	71.5%	D-IAT, S-IAT, SI-IAT
Glenn et al[27], 2017	United States	Implicit identification with death predicts change in suicide ideation during psychiatric treatment in adolescents	n = 276: With NSSI (n = 224), without NSSI (n = 52)	15.53 ± 1.34	71.0%	SI-IAT
Dickstein et al[34], 2015	United States	Self-injurious implicit attitudes among adolescent suicide attempters vs those engaged in NSSI	n = 136: With SI (n = 47), with NSSI (n = 46), controls (n = 43)	With SI: 15.85 ± 1.27; with NSSI: 15.24 ± 1.26; controls: 15.47 ± 1.29	With SI: 66.0%; NSSI: 84.5%; controls: 60.5%	SI-IAT
Cha et al[19], 2016	United States	Using implicit and explicit measures to predict non-suicidal self-injury among adolescent inpatients	n = 123: NSSI (n = 68), without NSSI (n = 55)	14.8 ± 1.5	71.5%	SI-IAT

NSSI: Non-suicidal self-injury; SI-IAT: Self-Injury Implicit Association Tests; D-IAT: Death Implicit Association Tests; S-IAT: Suicide Implicit Association Tests.

Validity of implicit cognition tests

The validity of implicit cognition tests, which include concurrent, retrospective, and prospective types, is summarized in Table 2, with key findings outlined below. For concurrent validity, Matthew K Nock (2007) found that the SI-IAT identified significant behavioral differences between individuals with and without NSSI, as the NSSI group showed faster reaction times for “self + NSSI” associations (P < 0.001); Christine B Cha (2016) further reported that at baseline, SI-IAT scores could significantly distinguish adolescents with a history of NSSI from those without (P = 0.003), and this distinguishing effect was stronger in participants who regarded cutting as their primary NSSI method [Cohen’s d = 0.62, 95% confidence interval (CI): 0.21-1.03]. In terms of retrospective validity, Daniel P Dickstein (2015) observed that individuals with NSSI had a stronger implicit identification with “cutting” (compared to “no cutting”) than both suicide attempters and healthy controls (P < 0.001), and they also displayed stronger implicit associations between “suicide/death” and “self” than the other two groups (Cohen’s d = 0.58, 95%CI: 0.19-0.97); Amy M Brausch (2022) noted that SI-IAT scores were significantly associated with lifetime suicide ideation (odds ratio = 1.87, 95%CI: 1.23-2.85, P = 0.004), even after adjusting for factors such as hopelessness and self-reported NSSI frequency.

Table 2 Main findings regarding validity of implicit cognition test.

Ref.	Test	Outcome	Timeframe	Main findings (with effect sizes and 95%CIs)
Nock and Banaji[40], 2007	SI-IAT	NSSI	Lifetime	Scores on the two versions were correlated (r = 0.50, n = 89, P < 0.001), and when entered simulate in the third step of the regression equation, the identity version continued to significantly predict self-injury (OR = 11.32, P = 0.02), whereas the attitude form of the test did not (OR = 2.87, P = 0.31)
Brausch et al[4], 2020	SI-IAT	SI	Lifetime	SI-IAT associated with suicide ideation (OR = 1.87, 95%CI: 1.23-2.85, P = 0.004)
Nagy et al[29], 2021	SI-IAT	NSSI	12 months	SI-IAT scores associated with NSSI history at both time points (P = 0.002) but not with frequency/recency (P > 0.05)
Powers et al[41], 2021	SI-IAT	NSSI frequency	6 months	Baseline SI-IAT predicted baseline NSSI (r = 0.32, P < 0.001) but not 6-month NSSI (r = 0.09, P = 0.12)
Glenn et al[32], 2016	SI-IAT	NSSI frequency	12 months	Implicit NSSI identification predicted 12-month NSSI (HR = 2.13, 95%CI: 1.34-3.38, P = 0.001)
Millner et al[21], 2018	D-IAT, S-IAT, SI-IAT	NSSI, SI	14 days	IATs showed good reliability (α = 0.78-0.82) but no predictive effect on suicide attempts (P = 0.23)
Glenn et al[27], 2017	SI-IAT	SI	12 months	Implicit death identification predicted discharge suicide ideation (β = 0.31, P < 0.001)
Dickstein et al[34], 2015	SI-IAT	SI	6 months	NSSI group showed stronger “cutting self” associations (Cohen’s d = 0.58, 95%CI: 0.19-0.97, P < 0.001)
Cha et al[19], 2016	SI-IAT	NSSI frequency	3 months	Controlling for NSSI history and prospective risk factors, SI-IAT scores predicted patients’ subsequent cutting behavior during their hospital stay (OR = 8.19, 95%CI: 1.56-42.98, P < 0.05)

CI: Confidence interval; SI-IAT: Self-Injury Implicit Association Tests; NSSI: Non-suicidal self-injury; OR: Odds ratio; D-IAT: Death Implicit Association Tests; S-IAT: Suicide Implicit Association Tests; HR: Hazard ratio.

Regarding prospective validity, Catherine R (2016) found that implicit NSSI identification at baseline could uniquely predict NSSI engagement over the following 12 months (hazard ratio = 2.13, 95%CI: 1.34-3.38, P = 0.001), and this predictive effect was independent of explicit NSSI measures. Jeffrey T Powers (2021) reported that baseline SI-IAT scores significantly predicted NSSI frequency at baseline (r = 0.32, P < 0.001) but failed to predict it at the 6-month follow-up (r = 0.09, P = 0.12), while explicit NSSI ratings, by contrast, strongly predicted future NSSI frequency (r = 0.45, P < 0.001). Alexander J Millner (2018) stated that although the death IAT and suicide IAT showed good reliability (Cronbach’s α = 0.78-0.82) and sensitivity to psychiatric changes, they could not prospectively predict suicide attempts over a 14-day follow-up (P = 0.23). Additionally, Catherine R Glenn (2017) found that implicit identification with death at admission significantly predicted the severity of suicide ideation at discharge (β = 0.31, P < 0.001), even after controlling for explicit suicide ideation at admission (Table 2).

Quality assessment

The quality of the included studies, as assessed by the Newcastle-Ottawa Scale, is summarized in Figure 2. Of the 9 studies: 6 studies were rated as “high quality” (≥ 7 stars), with strong scores in participant selection and outcome assessment. 3 studies were rated as “moderate quality” (4-6 stars), primarily due to limited adjustment for confounders (e.g., comorbid mental health conditions) or shorter follow-up periods. No studies were rated as “low quality” (≤ 3 stars), indicating that the included literature generally meets high methodological standards (Figure 2).

Figure 2 Quality assessment of included studies according to the Newcastle-Ottawa Scale. The quality of the included studies, as assessed by the Newcastle-Ottawa Scale is summarized. Of the 9 studies: 6 studies were rated as “high quality” (≥ 7 stars), with strong scores in participant selection and outcome assessment. 3 studies were rated as “moderate quality” (4-6 stars), primarily due to limited adjustment for confounders (e.g., comorbid mental health conditions) or shorter follow-up periods. No studies were rated as “low quality” (≤ 3 stars), indicating that the included literature generally meets high methodological standards.

DISCUSSION

The present systematic review synthesizes evidence on the utility of implicit cognition tests for assessing suicide risk in individuals with NSSI, directly addressing the review’s core objective of clarifying how these tests predict suicidal thoughts and behaviors (STBs) in this population. First, implicit cognition tests (e.g., SI-IAT, death/suicide IAT) exhibit robust concurrent and retrospective validity. Across included studies, these measures consistently distinguished individuals with NSSI from those without and linked to lifetime STBs. For example, individuals with NSSI showed stronger implicit associations between “self” and NSSI/suicide-related concepts than comparison groups, and SI-IAT scores correlated with lifetime suicidal ideation even after adjusting for confounding factors like hopelessness[13]. These results align with the implicit identification hypothesis of NSSI, which posits that automatic self-NSSI associations drive self-injurious behavior[14,15].

Second, prospective validity is time-dependent. The SI-IAT effectively predicted short-term NSSI recurrence and suicidal ideation (e.g., over several months), but its long-term predictive effects were inconsistent. This confirms implicit cognition acts as a state-like marker of short-term risk rather than a stable long-term predictor. Third, implicit cognition complements but does not replace explicit measures. Implicit tests captured unique risk not detected by self-report, but explicit assessments outperformed them in long-term prediction, highlighting the need for combined use in clinical practice[17,26,27].

We now analyze key drivers of this variation. First, sample characteristics matter: Inpatient studies reported stronger effects than community studies, likely due to more severe NSSI/suicidal ideation in clinical samples, while the overreliance on adolescent females limits generalizability to other demographics[20,28]. Second, test-specific factors influence outcomes: So we have standardized abbreviations (e.g., “SI-IAT” consistently) and clarify that test content should align with targets (e.g., SI-IAT for NSSI, death/suicide IAT for suicide risk) for optimal performance. Scoring methods also affect consistency, with standardized D-scores yielding more reliable results[29]. Third, follow-up design plays a role: Short-term follow-ups avoid confounding by intervening events (e.g., treatment), while long-term studies are vulnerable to changes in implicit associations over time. Fourth, comorbid mental health conditions were rarely adjusted for despite their known influence on both implicit cognition and suicide risk, contributing to result variability[30,31].

Notably, implicit cognition complements but does not replace explicit measures for suicide risk assessment[32]. Implicit tests predicted STBs independently of self-report in multiple studies (Catherine R Glenn, 2017; Catherine R, 2016), but explicit assessments (e.g., self-reported future NSSI likelihood) consistently outperformed implicit tests in long-term prediction (Jeffrey T Powers, 2021: r = 0.45 for explicit vs r = 0.09 for SI-IAT at 6 months)[33,34]. This indicates implicit measures add unique value by capturing “hidden” risk (e.g., in individuals who conceal suicidal thoughts) but should be used alongside - not instead of - explicit tools. Explicit assessment was strongly associated with future NSSI frequency. Self-reported future likelihood of NSSI predicts future frequency of NSSI, and it is a better predictor of future incidence of NSSI than SI-IAT[35,36]. The predictive ability of explicit assessment for post-discharge NSSI was higher than that of SI-IAT score, while implicit assessment could not predict post-discharge NSSI behavior[37]. Regarding the usefulness of explicit and implicit measurements of NSSI and their predictive ability for different time periods, the results are inconsistent[33]. In summary, SI-IAT has a stronger predictive effect on NSSI behavior within 2 weeks, but has no significant predictive effect on NSSI behavior at 6 months[38,39]. These findings suggest that, despite the importance of implicit identification with NSSI, SI-IAT may have limitations in predicting future NSSI.

Strengths and limitations

This review represents the first attempt to comprehensively synthesize all existing implicit cognition tests utilized for evaluating suicide risk. Among its limitations, the substantial heterogeneity of included studies - characterized by differences in methodological designs, study populations, and the specific implicit tests applied - prevented the performance of a quantitative synthesis of results. Additionally, a prominent limitation is that the reviewed studies rarely reported detailed psychometric characteristics of individual tests, including indicators of reliability and validity. Therefore, future investigations are warranted to fill this evidence void and strengthen the empirical basis for implicit cognition tests in suicide risk assessment.

CONCLUSION

Findings from this systematic review demonstrate that implicit cognition tests generally exhibit robust concurrent validity, retrospective validity, and prospective validity in the context of suicide risk assessment. A critical caveat of the existing evidence, however, is the scarcity of reported psychometric properties for these tests, which limits our capacity to conduct a fully comprehensive appraisal of their performance characteristics. Nonetheless, implicit cognition tests emerge as valuable adjunctive tools for suicide risk assessment and serve as important complements to traditional clinical assessment measures (e.g., self-report questionnaires, clinical interviews). Their key advantages include ease of administration and interpretation, independence from patients’ explicit self-disclosure of suicidal intentions (thus minimizing bias associated with intentional concealment), and objective scoring that is not susceptible to evaluator subjectivity. Despite these strengths, several challenges persist: The neurocognitive and psychological mechanisms underlying the utility of these tests for suicide risk assessment remain incompletely elucidated, and clinician familiarity with the administration and interpretation of implicit cognition tests is currently limited. These factors may impede the translation and integration of implicit cognition tests into routine clinical practice.

ACKNOWLEDGEMENTS

Thank you to all the participants who have contributed to the studies, and thank all the researchers whose work contributed to this systematic review and meta-analysis.