On the Iatrogenic Risk of Assessing Suicidality: A Meta-Analysis
Abstract
Previous studies have failed to detect an iatrogenic effect of assessing suicidality. However, the perception that asking about suicide may induce suicidality persists. This meta-analysis quantitatively synthesized research concerning the iatrogenic risks of assessing suicidality. This review included studies that explicitly evaluated the iatrogenic effects of assessing suicidality via prospective research methods. Thirteen articles were identified that met inclusion criteria. Evaluation of the pooled effect of assessing suicidality with regard to negative outcomes did not demonstrate significant iatrogenic effects. Our findings support the appropriateness of universal screening for suicidality, and should allay fears that assessing suicidality is harmful.
Attempted and completed suicide are compelling concerns with significant consequences for individuals, families, and communities (World Health Organization [WHO], 2014). A recent nationwide survey estimated that approximately 9.4 million adults seriously considered suicide, and 1.1 million adults attempted suicide during the year prior to data collection (Lipari, Piscopo, Kroutil, & Miller, 2014). Previous research has also emphasized the importance of assessing suicidality effectively within the context of clinical practice and scientific inquiry (Linehan, Comtois, & Ward-Ciesielski, 2012). Despite extant scholarship that has demonstrated the benefits associated with universal screening for suicidality (Osman et al., 2001), previous literature has also noted the ways in which suicide often goes unaddressed within the context of clinical and research encounters (Lang et al., 2016; Linehan et al., 2012; Hickey, Hawton, Fagg, & Weitzel, 2001; Suokas & Lonnqvist, 1991). For example, a recent study of trauma centers found that less than half of level I and level II trauma centers surveyed routinely screened for suicidality (Love & Zatzick, 2014).
There are several reasons underpinning the relatively low rate of suicide screening observed in the setting of trauma centers and emergency departments (Betz et al., 2016), including providers’ perception that assessing suicidality may induce suicidal thoughts and behavior in patients. This perception has been noted in previous studies (Bajaj et al., 2008; Bocquier et al., 2013), which have described how providers’ attitudes toward suicide may limit their willingness to address suicidality directly. For example, in a study of French general practitioners, nearly 10% of physicians endorsed the belief that asking about suicide may induce suicidal behavior (Bocquier et al., 2013). A similar study of 103 physicians found that more than one third of physicians reported that screening for suicidal thoughts in depressed patients could induce self-harm (Bajaj et al., 2008).
Concerns related to the potential iatrogenic effects of assessing suicidality are not limited to physicians and medical settings and may contribute, in part, to the limited rates of suicide risk assessment that have been observed in other settings, including schools (Hayden & Lauer, 2000), correctional institutions (Humber, Webb, Piper, Appleby, & Shaw, 2013), and social work encounters (Osteen, Jacobson, & Sharpe, 2014). For example, Eckert, Miller, Dupaul, and Riley-Tillman (2003) noted how the widespread implementation of school-based suicide prevention programs may be limited by perceptions of screening and assessment efforts as stigmatizing and thus potentially harmful to students. Furthermore, perceptions of suicide-related research and assessment as potentially harmful may prevent researchers from including explicit assessments of suicidality within research protocols (Lakeman & Fitzgerald, 2009), as well as limit the inclusion of suicidal participants within clinical research studies (Linehan et al., 2012). As was observed by Lakeman and Fitzgerald (2009) in their study of ethics committee members’ perceptions of suicide research protocols, most (i.e., 65%) perceived asking about suicidality as potentially harmful, particularly in the absence of effective crisis intervention and referral to treatment for distressed or high-risk participants.
Indeed, concerns related to the potential harms of assessing suicidality represent a ubiquitous barrier to the effective assessment and treatment for those suffering thoughts of suicide (e.g., Joiner, 2009). Despite the long-standing concerns for iatrogenic risks, several studies have suggested that suicide assessment is acceptable to patients within hospital settings (Ballard et al., 2013), effective for promoting access to follow-up care with mental health providers (Barr, Leitner, & Thomas, 2004), and potentially therapeutic for patients suffering suicidal thoughts (Smith, Poindexter, & Cukrowicz, 2010).
Moreover, several studies have evaluated the iatrogenic risks of suicide assessment empirically, including several randomized controlled trials and experiments. One widely cited study that examined this effect was a randomized controlled trial that assigned 2,342 high school students to either complete an assessment that included suicide-related items or an assessment that did not include such items. This study found no statistically significant iatrogenic effects (i.e., suicidal ideation or psychological distress) associated with completing an assessment of suicidality (Gould et al., 2005). Several other studies have assessed and not found empirical evidence in support of the potential iatrogenic risks of suicide assessment in other populations, including adult psychiatric inpatients (Eynan et al., 2014), adolescents (Deeley & Love, 2010), patients with borderline personality disorder (Law et al., 2015; Reynolds, Lindenboim, Comtois, Murray, & Linehan, 2006), and undergraduates (Bender, 2013; DeBeurs, Ghoncheh, Geraedts, & Kerkhof, 2016). In addition, previous qualitative reviews and commentaries (e.g., Dazzi, Gribble, Wessely, & Fear, 2014) have asserted that the benefits of assessing suicide outweigh the risks, particularly given the absence of evidence that demonstrates significant iatrogenic effects.
Present Study
Previous research has not quantitatively synthesized the literature to evaluate the potential for iatrogenic effects of assessing suicide across extant studies. In this review we sought to identify studies that examined the iatrogenic risk of assessing suicidality and test the hypothesis that suicide assessment is harmful to respondents. We evaluated the effect of assessing suicidality with regard to potential increases in suicidal ideation, suicidal behavior, and psychological distress.
Method
Search Strategy and Study Selection
The authors searched for empirical studies that assessed the iatrogenic risk of assessing suicidality by reviewing all available records indexed in Academic Search Complete, MedLine, PsycINFO, PubMed, and SCOPUS until December 15, 2016. Search terms were developed to capture a wide range of potential negative outcomes associated with the assessment of suicidality, and included: “Suic*” and (“Assess*” or “Measur*” or “Ask*”) and (“Iatro*” or “Harm*” or “Distress*”). All records indexed in the databases queried were considered for the present meta-analysis, including peer-reviewed journal articles, dissertations, published conference abstracts, and edited volumes. The reference lists of articles meeting criteria for inclusion in this study, and the reference lists of major qualitative reviews concerning the risks of assessing suicide, were also screened to identify articles meeting study criteria (i.e., “root” search). The reference lists of all articles that cited the studies included in this review were also screened to identify additional studies of relevance to this review (i.e., “branch” search). A hand search of three suicidology journals—Suicide and Life-Threatening Behavior, Archives of Suicide Research, and Crisis: The Journal of Crisis Intervention and Suicide Prevention—was also conducted.
Studies were eligible for inclusion in this meta-analytic review if they: (1) explicitly assessed or measured suicidality, including assessment of suicidal ideation and/or history of suicidal behavior; (2) employed a prospective design that allowed for explicit evaluation of the possible iatrogenic effect of completing measures of suicidality via randomized, controlled trials, experimental, or single sample longitudinal designs; and (3) assessed a specific negative outcome(s) or harm(s) that may have been associated with the assessment of suicidality.
Suicidality was broadly defined to include any measure of suicidal ideation and/or past suicidal behavior. However, the effect of completing measures of related psychological constructs (e.g., hopelessness, depression, acquired capability) was excluded from the present review. Potentially negative or harmful outcomes were also broadly defined to include any measurement of suicidal ideation at follow-up (i.e., postassessment), suicidal behavior at follow-up, and/or emotional or psychological distress (e.g., negative affect, symptoms of depression, global distress). Identified articles were independently screened by the study authors for inclusion in the present review, and disagreements in eligibility were argued to consensus. Comparison of the authors’ screening of articles indicated strong agreement with regard to article eligibility for inclusion in this review (93% agreement; κ = .82). A flow diagram detailing article identification is included in Figure 1.
Data Extraction
Quantitative data were extracted according to a structured coding scheme, and data extraction was conducted independently by the authors. Extracted data were reconciled, with any disagreements in data extraction being discussed by the coders until consensus was achieved. Primary outcomes for this study included changes in measures of distress, suicidality, and/or suicidal behavior following an initial assessment or screening for suicidality. Both empirically and clinically oriented studies were included in the present review. Sociodemographic data were also coded to characterize the age and gender of included studies.
Analyses
Cohen's d and associated 95% confidence intervals were calculated to assess the potential iatrogenic effects of assessing or screening for suicide, including mean comparisons for pre–post designs and mean comparisons between experimental and control groups. Cohen's d was also calculated for studies that employed single-sample and single-session designs that included baseline assessments of distress prior to the administration of suicide assessment items. A random effects approach was chosen a priori, given the expectation that there would be meaningful variation in study procedures and measurement paradigms across the diverse dimensions of potential harms included in this review. For the first series of meta-analyses that analyzed iatrogenic effects across different time frames, multiple effect sizes from an individual study within a defined time frame (e.g., within 2 days of assessment) were averaged together. Similarly, for the meta-analysis that examined iatrogenic effects among vulnerable samples/subsamples, effects were averaged together across time points for each outcome examined. All analyses were conducted and forest plots were constructed using Microsoft Excel via the procedure and templates recommended by Neyeloff, Fuchs, and Moreira (2012). Heterogeneity between included studies was assessed using the I2 statistic, which assesses the proportion of variability between studies that is attributable to true differences above and beyond differences attributable to sampling error alone (Borenstein, Hedges, Higgins, & Rothstein, 2009). Due to the relatively small number of studies identified, moderator analyses were not conducted. Given that only three identified studies included a formal assessment of suicidal behavior, and only one of those studies was able to statistically analyze the iatrogenic risk of assessing suicidality; it was not possible to calculate an estimate of pooled iatrogenic risk with regard to postassessment suicidal behaviors across studies.
Findings
Our literature search yielded 4,531 citations. The root and branch search of identified articles and hand search of suicidology journals identified 52 citations. After removal of duplicates, 4,531 abstracts were reviewed for inclusion in this review. Of these, 67 full-text articles were reviewed. Forty articles were excluded due to the absence of an explicit measure of iatrogenic effects, or the absence of an initial assessment that explicitly assessed suicidality. Twelve articles were not empirical papers (e.g., reviews, editorials, book chapters) and thus were excluded. One article (Langhinrichsen-Rohling, Arata, O'Brien, Bowers, & Kilbert, 2006) included several assessment items and outcomes related to multiple sensitive topics (i.e., drug use, physical and sexual victimization history, and suicidal ideation) and was excluded due to the possible confounding influence of other potentially upsetting measurements within a single instrument. One article did not include sufficient data for the calculation of an effect size for inclusion in this review and was also excluded. After excluding the aforementioned papers, 13 studies remained and were included in this meta-analysis. Study characteristics for included papers are summarized in Table 1. Overall, this review included 13 studies that represented 4,406 participants (Bender, 2013; Crawford et al., 2011; DeBeurs et al., 2016; Deeley & Love, 2010; Eynan et al., 2014; Gould et al., 2005; Harris & Goh, 2016; Law et al., 2015; Mathias et al., 2012; Reynolds et al., 2006; Rivlin, Marzano, Hawton, & Fazel, 2012; Robinson et al., 2011; Smith et al., 2010).
| Study | Sample | Design | Outcome Measure(s) | ||
|---|---|---|---|---|---|
| Suicidal ideation/intent | Suicidal behavior | Distress | |||
| Bender (2013) |
147 undergraduates Age: M = 18.8, 67.0% females |
Experimental | S-IAT | POMS | |
| Crawford et al. (2011) |
443 primary care patients Age: M = 48.5, 69.1% females |
RCT | Untitled 4-item scale | Single item | |
| Smith et al. (2010) |
21 adults w/history of SI/SA Age: M = 30.6, 60.0% females |
Longitudinal | Single item | Single item | |
| DeBeurs et al. (2016) |
301 Dutch university students Age: 18 and older |
RCT | PANAS | ||
| Deeley & Love (2010) |
129 adolescents Age: M = 17.0, 76.5% females |
Pre–Post | Mood Monitor (Negative Affect) | ||
| Eynan et al. (2014) |
119 psychiatric inpatients Age: M = 37.5, 47.5% females |
Longitudinal | Urge to self-harm or suicide, sense of control | ||
| Gould et al. (2005) |
2342 HS students Age: M = 14.8, 41.9% females |
RCT | SIQ-JR | POMS | |
| Harris & Goh (2016) |
259 Singaporean adults Age: M = 25.0, 58.3% females |
RCT | PANAS | ||
| Law et al. (2015) |
248 psychiatric outpatients Age: M = 43.9, 67.3% females |
Experimental | ESM Item | ESM Item | BPD symptoms |
| Mathias et al. (2012) |
170 adolescents Age: M = 14.0, 50% females |
Longitudinal | SIQ-JR | ||
| Rivlin et al. (2012) |
240 prisoners Age: 18 and older 50% females |
Case–control | Visual Analogue Scale (Mood) | ||
| Reynolds et al. (2006) |
63 BPD patients Age: M = 28.0, only females |
Longitudinal | 2 items: urge to self-harm, urge to suicide | Single item, level of distress | |
| Robinson et al. (2011) |
272 Australian HS students Age: M = 14–16, only males |
Experimental | Single item: past week suicidal thoughts | GHQ-12 and POMS | |
- S-IAT, Suicide Implicit Association Test; SIQ-JR, Suicidal Ideation Questionnaire-Junior; POMS, Profile of Mood States; PANAS, Positive and Negative Affect Schedule; ESM, experience sampling methodology; BPD, borderline personality disorder; HS, high school; GHQ-12, General Health Questionnaire.
Risk of Increased Suicidal Ideation Postassessment
First, studies that evaluated the iatrogenic risks of increased suicidal ideation following assessment of suicidality were stratified based on the timing of follow-up assessments. Effects were grouped to allow for separate random effects meta-analyses across time periods that included at least three studies. As is demonstrated in Figure 2, there were no significant iatrogenic effects of suicide assessment with regard to increased suicidal ideation postassessment across all time frames assessed. There was high between study heterogeneity (I2 = 64.05%) among studies that assessed suicidal ideation at 2 months to 2 years postassessment, and moderate between study heterogeneity (I2 = 19.38%) among studies that assessed suicidal ideation 2–4 weeks postassessment. Between study heterogeneity for studies that reported suicidal ideation within 2 days of assessment was reported as zero based upon a negative I2 value and may underestimate between study heterogeneity given the small number of studies included (von Hippel, 2015).
Risk of Increased Psychological Distress Postassessment
Next, studies that evaluated postassessment psychological distress within 2 days were included within a random effects meta-analysis. Although one study (i.e., DeBeurs et al., 2016) demonstrated a significant increase in negative affect postassessment, the pooled effect of assessing suicidality was nonsignificant with regard to increased psychological distress (d = −.128, 95% CI: −.332 to .076; see Figure 3), and there was high between study heterogeneity (I2 = 85.91%). Two identified studies (i.e., Deeley & Love, 2010; Law et al., 2015) evaluated psychological distress beyond 2 days after suicidality was assessed. Given that one study assessed psychological distress at 2 and 4 weeks (Law et al., 2015), and the other assessed psychological distress at 2 years after the initial assessment (Deeley & Love, 2010), the two studies were not pooled within a separate meta-analysis. Neither study found significant iatrogenic effects of assessing suicidality with regard to psychological distress at follow-up.
Risk of Increased Suicidal Behavior and Need for Crisis Intervention Postassessment
Only three of the identified studies (Crawford et al., 2011; Law et al., 2015; Smith et al., 2010) formally assessed for increases in suicidal behavior following assessment of suicidality. Law and colleagues (2015) found no statistically significant difference in suicidal behavior when comparing rates of attempted suicide across intensive suicide assessment and control conditions, such that 10.1% of participants who completed intensive suicide assessment and 11.1% of controls attempted suicide, respectively. Although Crawford et al. (2011) assessed for changes in suicidal behavior following an initial assessment of suicidality, they observed only one case of attempted suicide in the control condition, and thus, it was not possible to evaluate the relative risk of suicidal behavior in the suicidality assessment condition relative to controls. Similarly, Smith et al. (2010) observed no positive responses to a single item that assessed for suicidal behavior following an initial assessment.
In addition to these three studies that formally assessed suicidal behavior, four other studies (DeBeurs et al., 2016; Eynan et al., 2014; Reynolds et al., 2006; Robinson et al., 2011) explicitly noted whether or not any participants reported clinically significant distress and/or suicidality that required crisis intervention or referral to treatment. As noted in Table 2, the proportion of participants requiring crisis intervention and/or treatment referrals ranged from zero to approximately 5 percent in samples of university students, and from approximately 5 percent to one quarter of participants among clinical samples. The occurrence/nonoccurrence of adverse events and/or the need for crisis intervention following participation was not explicitly noted in the other six studies included in the present review.
| Study | Nature of sample | Nature of intervention/Follow-up | n (%) |
|---|---|---|---|
| Robinson et al. (2011) | Australian HS students | “…identified as requiring ongoing support…” | 13 (4.8) |
| Eynan et al. (2014) | Psychiatric inpatients | “…referred to the emergency psychiatric service…” | 13 (4.5) |
| DeBeurs et al. (2016) | Dutch university students | “…contacted the specialized psychologist before or after the study.” | 0 (0.0) |
| Reynolds et al. (2006) | BPD patients | “…interventions strategies that range from low-level intensity to high-level intensity.” | 15 (23.8) |
- Note. Only includes studies that explicitly stated whether or not crisis intervention and/or referral procedures were enacted to address acute suicidality. HS, high school; BPD, borderline personality disorder.
Risk of Iatrogenic Effects Among High-Risk/Vulnerable Participants
Two additional random effects meta-analyses were conducted to evaluate the potential for iatrogenic effects among participants who could be considered members of high-risk or vulnerable populations. In regard to suicidal ideation, there was no significant pooled effect of assessing suicidality across studies that utilized high-risk/vulnerable samples or subsamples (d = −.093, 95% CI: −.315 to .129; see Figure 4), and there was high between study heterogeneity (I2 = 43.81%). Similarly, there was no significant pooled effect observed with regard to psychological distress across studies with high-risk vulnerable participants (d = −.052, 95% CI: −.146 to .250; see Figure 4), and there was high between study heterogeneity (I2 = 56.99%). However, there was a statistically significant effect of assessing suicidality noted among a study that compared prisoners who attempted suicide with other prisoners (Rivlin et al., 2012).
Discussion
Taken together, the results of this meta-analytic review do not demonstrate statistically significant iatrogenic effects associated with assessing suicidality. Analysis of potential iatrogenic effects across immediate, short-, and long-term follow-up assessments did not reveal significant harms associated with assessing suicidality, nor was there a significant pooled effect of assessing suicide among high-risk and/or vulnerable study populations. Moreover, no individual studies demonstrated significant increases in suicidal ideation that were attributable to assessing suicidality. In contrast, there were some individual studies identified in this review that suggested how assessing suicidality may cause increased negative affect or decreased mood in some participants.
Specifically, DeBeurs et al. (2016) found that those who completed a brief measure of suicidality reported higher levels of negative affect postassessment relative to those who completed an alternative assessment instrument. These researchers noted how the minority of participants who reported increased negative affect following completion of a self-report measure of suicidal ideation also reported higher levels of psychological distress and negative affect at baseline.
Similarly, Rivlin et al. (2012), demonstrated a significant iatrogenic risk of decreased mood following an assessment of suicidality, such that prisoners who had attempted suicide reported lower ratings on a single-item assessment of postassessment mood compared to matched controls. However, the findings of Rivlin and colleagues demonstrated an overall reduction in psychological distress among the entire sample (i.e., cases and controls). Furthermore, although more than one quarter (n = 21, 17.5%) of prisoners who had made a “near-lethal” suicide attempt (i.e., cases) reported decreased mood postassessment, only two (1.7%) cases stated they were displeased with having participated in the assessment. As noted by Rivlin and colleagues, it may be that increased participant distress among those with histories of suicidal behavior is attributable to the increased detail with which they would respond to items related to suicidality, whereas those without such histories would not have reason to indicate specific details of past behavior in response to assessment items. Nonetheless, the results of this meta-analysis do not demonstrate robust iatrogenic risks associated with assessing suicidality across a diversity of samples and methodologies. Furthermore, the present review found no significant pooled effects with regard to suicidal ideation and psychological distress.
Consistent with the present meta-analytic findings, previous studies have found that participants in assessments of suicidality and other items of relevance to the study and prevention of suicide find such assessments to be meaningful and important, despite the potential that such topics may elicit negative emotions or distress. This includes previous studies that employed qualitative methods in diverse populations, including interviews conducted with Alaska native suicide survivors enrolled in college (DeCou, Skewes, López, & Skanis, 2013) and adult medical inpatients who completed suicide risk assessments (Snyder et al., 2016), in which participants reported assessments of suicidality to be helpful and acceptable despite the negative thoughts and transient emotions that were sometimes evoked by assessment items. Thus, the two examples of significant iatrogenic effects identified in the present review (i.e., DeBeurs et al., 2016; Rivlin et al., 2012), which found significant increases in negative affect and decreased mood in some participants (i.e., suicide attempters; Rivlin et al., 2012), may not outweigh the potential benefits that are often perceived by participants who complete assessment of suicidality and related constructs (Smith et al., 2010).
Thus, future studies should continue to consider effective strategies for conceptualizing the overlapping domains of affective, cognitive, and functional sequelae that may be associated with the assessment of suicidality and explore ways to optimize the benefits of assessing suicidality directly. Several studies identified by this review (Eynan et al., 2014; Mathias et al., 2012; Rivlin et al., 2012; Smith et al., 2010) reflect potential direct benefits that emerged from inquiry concerning the topic of suicidality via decreased distress and suicidal ideation. Notwithstanding these benefits, the present findings suggest that the risks of assessing suicidality do not include marked iatrogenic effects and thus do not outweigh the potential benefits of widespread screening for suicidality and continued research in this area. In addition, future studies should examine potential mediators and moderators of distress experienced by a small minority of participants, which may inform more acceptable and effective assessment procedures for those most likely to become distressed. In particular, the intersection of one's history of suicidal behavior and the context or method of assessing suicidality warrants further investigation as a moderator of iatrogenic effects, particularly given the present findings that identified a significant increase in distress among suicide attempters in prison (i.e., Rivlin et al., 2012). It may be that there are unique aspects of assessing suicidality in carceral settings that contributed to respondent distress, as the other samples included in this review (e.g., students, psychiatric inpatients and outpatients, primary care outpatients) did not demonstrate significant increases in distress following assessment of suicidality. This type of focused research may help to better identify the conditional and indirect effects that sometimes lead to unintended harms following assessment of suicidality in a small minority of respondents.
The present review should be considered in view of several limitations. First, although we attempted an exhaustive review of existing studies, it is possible that relevant studies were not identified. Next, some of the studies in this review were not randomized clinical trials, and thus, the ability to offer causal inference concerning the association between assessing suicidality and negative outcomes is limited. Furthermore, given the small number of identified articles, we did not attempt moderator analyses concerning the conditional influence of gender, ethnicity, and age. These sociodemographic and sociocultural variables may in fact condition the effects of assessing suicidality and warrant further study. Finally, there was substantial between study heterogeneity among the studies identified for this review, which likely reflected the diverse measures, methods, and samples employed across studies and suggests the need for additional research that considers the conditional and indirect effects that may explain the observed heterogeneity between studies. Despite these limitations, the present review is the first to quantitatively synthesize previous studies of the iatrogenic risk of assessing suicidality and raises important implications for future research and practice in this area.
Foremost, this study supports the safety of assessing suicidality directly, given that there was no evidence of a statistically significant iatrogenic risk across studies. That is, the present findings do not support the durable perception that assessing suicidality induces suicidal thoughts and/or behavior. Although there were small minorities of subjects within the included studies who reported increased suicidal ideation postassessment, these cases reflect the need for effective protocols and procedures for managing acute suicidality in concert with universal screening, as well as the need for enhanced safety protocols when conducting research with vulnerable populations who may be at high baseline risk for suicide. For example, Eynan et al. (2014) noted that 3% of participants reported increases in the urge to self-harm or suicide following assessment of suicidality. Thus, there is no evidence supporting a robust iatrogenic effect of assessing suicidality across participants it remains essential for clinicians and researchers to address suicidal crises that may emerge within the context of such assessments. Nonetheless, previous studies have noted the ways in which assessing suicide may be therapeutic for many respondents (e.g., Smith et al., 2010), a benefit which may outweigh the potential distress induced in some vulnerable groups. Taken together, our findings support the continued expansion of clinical practice and research to include universal assessment of suicidality in view of the limited evidence that exists concerning the iatrogenic effects of assessing suicidality.
