The following information should be cited as: Bentley, K. H., Franklin, J. C., Ribeiro, J. D., Kleiman, E. M., Fox, K. R., & Nock, M. K. (in press). Anxiety and its disorders as risk factors for suicidal thoughts and behaviors: A meta-analytic review. Clinical Psychology Review.
Anxiety and its disorders are listed as important risk factors for suicide by leading national organizations, including the American Association of Suicidology (AAS, 2015), the American Foundation for Suicide Prevention (AFSP, 2015), and the National Suicide Prevention Hotline (2015). The idea that heightened anxiety plays an important role in suicidal thoughts and behaviors (SITBs) is also consistent with widely discussed theories of suicide (e.g., Baumeister, 1990; Fawcett, 2001; Riskind, 1997; Shneidman, 1993; Wenzel & Beck, 2008). Recent large-scale studies have also found that anxiety and its disorders (defined here as any anxiety, obsessive-compulsive, trauma or stressor-related, and somatic symptom disorders) are statistically significant predictors of future SITBs, even when controlling for co-occurring diagnoses like depression (e.g., Bolton et al., 2008; Borges et al., 2008; Nock et al., 2009, 2010). However, some earlier research studies showed that anxiety disorders are not related to future SITBs (e.g., Hornig & McNally, 1995; Warshaw et al., 1995). In addition, whether anxiety is a useful indicator of elevated suicide risk for practicing clinicians who need to determine whether a patient is going to attempt suicide in the near future (such as over the next several days or week) is unclear.
Given these inconsistencies and uncertainties: How well does anxiety predict future SITBs? Our meta-analysis addresses this question and includes three primary aims:
First, we assess what the empirical literature on anxiety and suicide looks like.
Second, we examine whether anxiety is a statistically significant predictor of future suicide ideation, attempts, and death.
Third, we explore whether sample age, sample severity, and study follow-up length moderate the effect of anxiety on suicide ideation, attempts, and death.
Fourth, we consider the clinical utility of anxiety as a risk factor for suicide ideation, attempts, and death, quantified as: improvement in diagnostic accuracy above chance and odds ratios in terms of absolute risk of suicide ideation, attempts, and death.
What does the literature look like?
Answer: The first study that tested anxiety as a predictor of future suicidal SITBs was completed in 1982. Since then, a total of 65 qualifying articles (using just over 852,000 participants) have been published. Together, these studies yielded 180 instances in which anxiety symptoms or an anxiety or related diagnosis (such as posttraumatic stress disorder [PTSD]) was used to predict suicide ideation, attempts, or deaths. The majority of these studies used very long follow-up periods (most commonly 5 years), with only 4 studies involving a follow-up period shorter than a year. Just over half of these studies used clinical samples (meaning psychiatric patients), whereas just over a quarter of studies used general population samples and just under a quarter used participants with a history of prior SITBs. About 70% of studies involved adult samples. The most common outcome was suicide attempt (with 61% of studies including at least one instance of anxiety predicting suicide attempt), followed by death (38%) and ideation (13%).
does anxiety predict sitbs?
Answer: Anxiety (symptoms and diagnoses) overall appeared to be statistically significant risk factors for suicide ideation and attempts; however, the effects were weak. Overall weighted mean odds ratios for ideation and attempts were near 1.5 and adjusting for publication bias further reduced estimates. Anxiety was not a statistically significant predictor of suicide death.
We also evaluated the effects of specific categories of anxiety, including individual disorders (such as panic disorder, social anxiety, PTSD, etc.) and anxiety symptoms on future SITBs.
- Generalized anxiety disorder, specific phobia, and social anxiety disorder were significant predictors of future suicide ideation.
- PTSD, panic disorder, and social anxiety disorder were significant predictors of future suicide attempts.
- Adjustment disorder was a significant predictor of fewer suicide attempts and deaths (in other words, a protective factor for suicidal behavior).
- Anxiety symptoms were significant predictors of suicide ideation and attempts, but not death.
Note. These graphs present overall odds ratios (ORs) and corresponding 95% confidence intervals for any anxiety construct (meaning symptoms or diagnosis), any anxiety symptoms, and any anxiety diagnosis predicting suicide ideation, suicide attempt, and suicide death. An OR must exceed 1.0 to be statistically significant.
what moderated the effect of anxiety on sitbs?
Answer: The type of sample, age of sample, and length of follow-up period all moderated the effect of anxiety (symptoms and diagnoses) on future SITBs. Specifically, we found that:
- Studies using general samples observed stronger effects of anxiety on suicide attempts than studies using psychiatric patients or individuals with a SITB history. This was not the case for suicide ideation or death.
- Studies using adult-only samples observed stronger effects of anxiety on suicide attempts than those using adolescent-only or mixed samples. This was not the case for suicide ideation or death.
- Studies with shorter follow-up periods found stronger effects of anxiety on suicide ideation and attempts. This was not the case for suicide death.
However, anxiety was still a relatively weak predictor of future SITBs, even in general and adult-only samples, and over shorter follow-up periods.
Is anxiety a clinically useful predictor of sitbs?
Answer: We evaluated clinical utility on two domains.
- First, we looked at improvement in diagnostic accuracy statistics, and found that anxiety offered only slight improvement above chance prediction.
- Second, we looked at the magnitude of findings in terms of absolute risk of suicide ideation, attempts, and death. Let's consider an example:
- The likelihood of death by suicide in the United States in a given year is 12.5/100,000 (i.e., 0.000125); attempts are estimated to be 25 times more likely (i.e., 0.0031).
- The strongest predictor in this meta-analysis was a PTSD diagnosis (weighted mean OR = 2.25) predicting suicide attempts.
- Based on this estimate, the presence of a PTSD diagnosis would only increase the odds of an attempt to 0.007 – a figure still nearly zero.
- This is also based on prediction of suicide attempts over a one-year period, whereas most clinicians have to determine suicide risk over the next several days or week.
- These findings suggest that, in terms of absolute (rather than relative) odds, anxiety is a risk factor for future suicide ideation, attempts, and death, at least as these constructs have been traditionally studied in the literature to date.
however, this does not necessarily mean that anxiety is not an important risk factor for suicidal thoughts and behaviors. instead, we believe that it has not yet been studied under the conditions necessary to detect such a relationship.
Where do we go from here?
We suggest several methodological changes for future research on risk factors for SITBs:
Instead of assessing anxiety in terms of only diagnoses or scores on measures of broad, trait-like symptoms, studies should use more dynamic (including physiological, implicit, behavioral) measures of anxiety.
Instead of only examining anxiety in isolation from other risk factors, researchers should examine the interactions of anxiety with other risk factors, and how several variables may combine to increase suicide risk (for example, pervasive depression + past suicidal behavior + recent negative life event + acute anxiety + access to lethal means).
- Instead of letting months or years go by between time points, studies should assess risk factors (such as anxiety) and SITBs more frequently and over shorter periods of time (such as minutes, hours, days).
- Instead of conducting this research in relatively small samples where it is difficult to observe effects, researchers must include very large numbers of participants in their studies to detect a sufficient number of suicidal behaviors during follow-up.