Elizabeth O'Connor, PhD; Bradley N. Gaynes, MD, MPH; Brittany U. Burda, MPH; Clara Soh, MPA; Evelyn P. Whitlock, MD, MPH
This article was published at www.annals.org on 23 April 2013.
Note: Ms. Soh accepted a position at the Pharmaceutical Research and Manufacturers of
America during the finalization of the full report and drafting and final submission of this
manuscript; she contributed most of her intellectual content to this review while she was employed
with the Oregon Evidence-based Practice Center at Kaiser Permanente Center for Health Research.
Disclaimer: The views and opinions expressed in this article are those of the authors
and do not necessarily reflect the views of the Pharmaceutical Research and Manufacturers of America.
The findings and conclusions in this document are those of the authors, who are responsible for its
content, and do not necessarily represent the views of AHRQ. No statement in this report should be
construed as an official position of AHRQ or of the U.S. Department of Health and Human Services.
Acknowledgment: The authors thank the AHRQ staff; the USPSTF; J. Michael Bostwick, MD;
David A. Brent, MD; Gregory Simon, MD, MPH, who provided expert review of the report; Kevin Lutz, MFA;
Daphne Plaut, MLS; and Heather Baird at the Kaiser Permanente Center for Health Research.
Grant Support: By the Oregon Evidence-based Practice Center under contract to AHRQ,
Rockville, Maryland (contract HHS-290-2007-10057-I).
Potential Conflicts of Interest: Dr. O'Connor: Grant (money to
institution): AHRQ; Support for travel to meetings for the study or other purposes
(money to institution): AHRQ; Payment for writing or reviewing the manuscript (money
to institution): AHRQ; Provision of writing assistance, medicines, equipment, or
administrative support (money to institution): AHRQ. Dr. Gaynes: Grant (money to
institution): AHRQ. Ms. Burda: Grant (money to institution): AHRQ;
Consultancy: Oregon Health & Science University; Employment:
Howard Hughes Medical Institute, Vollum Institute (Oregon Health & Science University);
Other: Pennsylvania State University, Howard Hughes Medical Institute. Ms. Soh:
Grant (money to institution): AHRQ; Support for travel to meetings for the
study or other purposes (money to institution): AHRQ; Provision of writing
assistance, medicines, equipment, or administrative support (money to institution): AHRQ. Dr.
Whitlock: Grant (money to institution): AHRQ; Support for travel to meetings
for the study or other purposes (money to institution): AHRQ; Payment for writing or
reviewing the manuscript (money to institution): AHRQ; Provision of writing
assistance, medicines, equipment, or administrative support (money to institution): AHRQ.
Disclosures can also be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum=M12-2880.
Corresponding Author: Elizabeth O'Connor, PhD, Center for Health Research, Kaiser
Permanente NW, 3800 North Interstate Avenue, Portland, OR 97227; e-mail, firstname.lastname@example.org.
Current Author Addresses: Dr. O'Connor, Ms. Burda, and Dr. Whitlock: Center for Health
Research, Kaiser Permanente NW, 3800 North Interstate Avenue, Portland, OR 97227.
Dr. Gaynes: Department of Psychiatry, CB #7160, University of North Carolina School of Medicine,
Chapel Hill, NC 27599.
Ms. Soh: Pharmaceutical Research and Manufacturers of America, 950 F Street NW, Suite 300, Washington,
Author Contributions: Conception and design: E. O'Connor, B.N. Gaynes, C. Soh, E.P.
Analysis and interpretation of the data: E. O'Connor, B.N. Gaynes, C. Soh.
Drafting of the article: E. O'Connor, B.U. Burda, C. Soh.
Critical revision of the article for important intellectual content: B.N. Gaynes, B.U. Burda, C. Soh,
Final approval of the article: B.N. Gaynes, B.U. Burda, C. Soh, E.P. Whitlock.
Provision of study materials or patients: B.N. Gaynes.
Statistical expertise: E. O'Connor.
Obtaining of funding: E.P. Whitlock.
Administrative, technical, or logistic support: B.U. Burda.
Collection and assembly of data: E. O'Connor, B.N. Gaynes, B.U. Burda, C. Soh.
O'Connor E., Gaynes B., Burda B., Soh C., Whitlock E.; Screening for and Treatment of Suicide Risk Relevant to Primary Care: A Systematic Review for
the U.S. Preventive Services Task Force. Ann Intern Med. 2013;158:741-754. doi: 10.7326/0003-4819-158-10-201305210-00642
Download citation file:
Published: Ann Intern Med. 2013;158(10):741-754.
In 2009, suicide accounted for 36 897 deaths in the United States.
To review the accuracy of screening instruments and the efficacy and safety of screening for and
treatment of suicide risk in populations and settings relevant to primary care.
Citations from MEDLINE, PsycINFO, the Cochrane Central Register of Controlled Trials, and CINAHL (2002
to 17 July 2012); gray literature; and a surveillance search of MEDLINE for additional screening
trials (July to December 2012).
Fair- or good-quality English-language studies that assessed the accuracy of screening instruments in
primary care or similar populations and trials of suicide prevention interventions in primary or
mental health care settings.
One investigator abstracted data; a second checked the abstraction. Two investigators rated study
Evidence was insufficient to determine the benefits of screening in primary care populations; very
limited evidence identified no serious harms. Minimal evidence suggested that screening tools can
identify some adults at increased risk for suicide in primary care, but accuracy was lower in studies
of older adults. Minimal evidence limited to high-risk populations suggested poor performance of
screening instruments in adolescents. Trial evidence showed that psychotherapy reduced suicide
attempts in high-risk adults but not adolescents. Most trials were insufficiently powered to detect
effects on deaths.
Treatment evidence was derived from high-risk rather than screening-detected populations. Evidence
relevant to adolescents, older adults, and racial or ethnic minorities was limited.
Primary care–feasible screening tools might help to identify some adults at increased risk for
suicide but have limited ability to detect suicide risk in adolescents. Psychotherapy may reduce
suicide attempts in some high-risk adults, but effective interventions for high-risk adolescents are
not yet proven.
Agency for Healthcare Research and Quality.
Suicide was the 10th leading cause of death in the United States in 2009, accounting for 36 897 deaths
with an age-adjusted rate of 11.8 per 100 000 persons (1,
2). It accounted for more than 1.4 million years of potential life lost before age 85 years (3). Suicide attempts and death rates vary substantially by sex; age;
race; and other risk factors, such as psychiatric disorders (4–7), prior suicide attempts (8–11), a history of nonsuicidal self-harm (7), and a
serious adverse childhood experience (12–15). Individual risk factors have a limited ability to predict suicide in a person at a
particular time, but risk for a suicide attempt and death increases with multiple risk factors (covering
psychosocial, biomedical, and developmental realms) and high levels of distress (16, 17).
Thirty-eight percent of U.S. adults who completed suicide visited their primary care providers in the prior
month; this rate was even higher (50% to 70%) in older adults (18).
Nearly 90% of suicidal youth had primary care visits during the previous 12 months, compared with 70% to 80%
of nonsuicidal youth (19, 20). Screening tools that are
accurate and feasible for use in primary care could represent an important opportunity to identify persons
at increased risk for suicide so that it can be prevented through appropriate treatment.
In 2004, the U.S. Preventive Services Task Force (USPSTF) concluded that evidence was insufficient to
recommend for or against routine screening by primary care clinicians to detect suicide risk in the general
population (I statement). That review found limited evidence that screening tests can reliably detect
suicide risk in primary care populations (21). A large body of
evidence (33 randomized, controlled trials and 2 cohort studies) examined the effects of treatment on
suicide attempts and suicide deaths in adolescents or adults.
Few trials showed benefit of treatment, and many were underpowered for these rare outcomes. Evidence also
showed that nonpharmacologic treatment could reduce depressive symptoms and suicidal ideation in high-risk
older adolescents and adults. Evidence was lacking on harms of screening and treatment.
We developed an analytic framework (Appendix Figure 1) with 6 key
questions to guide our work (Appendix Table 1). Our full report
describes the methods in detail (22).
Analytic framework for suicide risk relevant to primary care.
KQ = key question (see Appendix Table 1). * All studies must
report at least 1 suicide-specific outcome measure.
Appendix Table 1. Key Questions
We considered all studies from the previous review for inclusion. We searched MEDLINE, PsycINFO, CINAHL,
and the Cochrane Central Register of Controlled Trials for studies published between January 2002 and 17
July 2012 to bridge and update from the previous review. We hand-searched bibliographies of relevant
reviews and searched Web sites of government agencies and professional organizations to identify relevant
research published outside of peer-reviewed journals. We also conducted a surveillance search of MEDLINE
through December 2012 to identify additional screening trials.
Two investigators independently reviewed the abstracts and articles against specified inclusion and
exclusion criteria. We resolved disagreements through consultation with the larger project team. We
included English-language studies in general primary care or specialty mental health populations (or
similar populations) of any age. We also included studies limited to patients with depression, substance
misuse, posttraumatic stress disorder, or borderline personality disorder. We excluded studies limited to
patients with other mental health conditions.
For questions related to harms and benefits of screening or treatment, we included randomized and
nonrandomized clinical trials. To address effects of treatment, we included trials of behavior-based or
pharmacologic treatment with a primary aim of reducing suicide deaths, suicide attempts or self-harm, or
suicidal ideation. We included studies of screening instrument accuracy that reported sensitivity,
specificity, or related statistics of brief screening instruments to detect current increased suicide
risk (usually suicidal ideation) relative to a reference standard. The reference standard had to be a
more in-depth assessment of suicide risk by a trained mental health professional or a trained interviewer
using a standardized instrument to determine whether suicide risk was increased. We would have included
suicide attempts in the immediate period after screening (for example, 1 month) as a gold standard if we
had found any studies that did this. We also would have included comparative cohort studies addressing
harms of pharmacologic treatment in suicidal populations if we had found any.
One investigator abstracted data from all included studies into a standard evidence table, and a second
investigator checked the data for accuracy. Two investigators independently assessed the methodological
quality of each study by using predefined design-specific quality criteria based on methods developed by
the USPSTF (23). We supplemented these criteria with the Quality
Assessment of Diagnostic Accuracy Studies tool (24) to evaluate
the quality of diagnostic accuracy (screening) studies, resulting in a rating of good, fair, or poor. We
resolved disagreements in quality assessment through discussion and, if necessary, consultation with a
third reviewer. We excluded poor-quality trials.
For all key questions, we created results tables with important study characteristics. We critically
examined these tables to identify the range of results and potential associations with effect size. We
examined trials limited to adolescents or limited to older adults separately from other adult trials.
For key questions 4 and 5 only, we conducted random-effects meta-analyses to estimate the effect size of
suicide prevention interventions on suicide attempts or self-harm, suicidal ideation, and depression. We
used Stata, version 11.2 (StataCorp, College Station, Texas), for all statistical analyses. Risk ratios
were analyzed for suicide attempts. All trials reported at least 1 suicide attempt or self-harm episode
in each intervention group, so no correction for empty cells was needed. We analyzed standardized mean
differences (SMDs) in change from baseline for the continuous outcomes (suicidal ideation and
depression). We calculated SDs of change from baseline by using a standard formula (25).
We assessed the presence of statistical heterogeneity among the studies by using standard chi-square
tests and the I2 statistic (26). We
applied the Cochrane Collaboration (27) rules of thumb for
interpreting I2 (probably unimportant, <40%; moderate, 30% to 60%; and
substantial, 50% to 90%) and Cohen (28) rules of thumb for
interpreting effect sizes (small, 0.2 to 0.5; medium, 0.5 to 0.8; and large, ≥0.8).
This research was funded by the Agency for Healthcare Research and Quality (AHRQ). The AHRQ staff
provided oversight for the project and assisted in external review of the companion draft evidence
synthesis but had no role in study selection, quality assessment, synthesis, or development of
conclusions. The investigators are solely responsible for the content of the manuscript and the decision
to submit for publication.
We included 56 studies that reported results in 86 publications, from 3925 abstracts and 303 full-text
articles (Appendix Figure 2). We included 7 trials that addressed
screening (key questions 1, 2, and 3): 1 examined short-term benefits of screening (29), 4 examined performance characteristics of screening instruments (30–33), and 3 examined adverse effects of
screening (29, 34, 35). We included 49 trials that
addressed the benefits of treatment (key questions 4 and 5)—36 were conducted in adults or mixed
adolescent and adult populations (36–71), and 13 were done in adolescents (72–84). A
subset of 12 of these trials also reported on adverse events or a (statistically nonsignificant) paradoxical
increase in suicide attempts, which is discussed under key question 6 (36, 40, 41, 49, 51,
55–57, 66, 73, 76, 77).
Summary of evidence search and selection.
* Surveillance search of MEDLINE only from July 2012 through December 2012 for trials related to
screening are not included. No additional trials were identified.
We identified 1 short-term, fair-quality trial (n = 443) that addressed key
question 1. This trial found no clear short-term (that is, within 2 weeks) benefit of screening (29). It was published after the 2004 USPSTF review on this topic and
thus was not included in the previous review.
Appendix Table 2 shows data from the 4 studies that reported
the accuracy of screening instruments for identifying persons at increased risk for suicide (30–33). Two of these trials were conducted in
adult or older adult primary care populations (32, 33).
One study (the only diagnostic accuracy study also included in the 2004 review) examined the clinical
utility of 3 suicide-related items in primary care patients aged 18 years or older with prescheduled
appointments for any reason (n = 1001) (32).
Appendix Table 2. Test Performance Characteristics of Suicide Screening Instruments (Key Question
Items had sensitivities of 83% to 100% and specificities of 81% to 98% relative to a nurse-administered
structured interview on the same day. The positive predictive values were low, ranging from 6% to 30%.
Accuracy was lower in the study of the Geriatric Depression Scale—Suicidal Ideation subscale in
general primary care patients aged 65 years or older (n = 626) (33). The optimal cutoff yielded a sensitivity and specificity of 80%
for suicidal ideation during the previous 2 weeks, and the positive predictive value was fairly low
Two trials reported instrument accuracy in adolescent samples (combined n = 799).
Although these studies used different instruments and approaches to assemble their samples, both
represented high-risk groups that had 22% to 27% prevalence of suicidal ideation or behavior. In these
studies, sensitivity ranged from 52% to 87% and specificity ranged from 60% to 85%. These results are
generalizable only to high-risk populations.
Three trials reported on potential adverse effects of screening (all published since the previous
review). None identified serious adverse effects of screening (29, 34, 35).
Appendix Table 3 (36–112) lists brief population and intervention characteristics of all included trials, and
Appendix Table 4 (36–112) lists all outcomes reported by each trial. The previous review included only 11 of
these 49 trials; almost all of the new trials were published in 2003 or later, well after the end of the
search window of the 2004 review. We organized treatment trials into 3 broad intervention groups:
psychotherapy, enhanced usual care, and medication.
Appendix Table 3. Study Population and Intervention Characteristics of Included Studies (Key
Questions 4 and 5)
Appendix Table 4. Outcomes Reported in Included Trials for Benefits and Harms of Treatment (Key
Questions 4 and 5)
Thirty-one trials investigated a specific psychotherapeutic treatment approach, generally compared
with usual care. Nineteen were conducted in adults (37, 38, 41, 44, 47, 50–52, 54–56, 58–60, 62–65, 67), and 12 were
conducted in adolescents (72–81, 83,
84). Most psychotherapy trials were done in high-risk populations, usually identified because
these persons had a recent suicide attempt or selected mental health disorders (for example,
borderline personality disorder) and a history of suicide attempts. The only study that identified
patients through population-based screening was conducted in Sri Lanka.
In adults, evidence was insufficient to evaluate the effect on suicide deaths, because only 6 of the
19 psychotherapy trials reported suicide deaths. Psychotherapy recipients had a 32% reduction in the
likelihood of a suicide attempt or deliberate self-harm compared with usual care recipients (relative
risk, 0.68 [95% CI, 0.56 to 0.83]; 11 trials; n = 1583; I2 = 16.1%) (Figure 1). However, a single estimate of
absolute benefit would be misleading, given the highly variable rate of suicide attempts or self-harm
(15% to 71% of control group participants had a suicide attempt or self-harm episode at follow-up).
When the trials with the most extreme suicide attempt rates were excluded (38, 50, 54, 55), absolute differences ranged from a low of 46%
in the control group and 39% in the intervention group (65)
to a high of 47% in the control group and 23% in the intervention group (59).
Suicide attempts in psychotherapy trials.
Weights are from random-effects analysis. CB = cognitive behavioral; D = dialectical; DG =
developmental group; DSH = deliberate self-harm; ODT = other therapy, direct; OTND = other
therapy, nondirect; P = psychodynamic; PS = problem solving; SA = suicide attempt.
Psychotherapy did not show greater improvement than usual care for suicidal ideation (SMD, −0.10
[CI, −0.27 to 0.06]; 8 trials; n = 964; I2 =
26.3%) (Figure 2), and most trials reported reduced suicidal
ideation in both intervention and control groups. Psychotherapy had a small beneficial effect on
depression relative to usual care (SMD, −0.37 [CI, −0.55 to −0.19]; 12 trials;
n = 1653; I2 = 60.5%) (Figure 3). Other outcomes were sparsely reported and had mixed results.
Suicidal ideation in psychotherapy trials.
Weights are from random-effects analysis. CB = cognitive behavioral; D = dialectical; DG =
developmental group; OTND = other therapy, nondirect; P = psychodynamic; PS = problem solving;
SMD = standard mean difference.
Depression in psychotherapy trials.
In adolescents, we could not determine the effects of suicide prevention treatment on deaths because
only 1 death occurred in the 3 trials reporting this outcome. Psychotherapy did not reduce suicide
attempts in adolescents at 6 to 18 months (relative risk, 0.99 [CI, 0.75 to 1.31]; 9 trials;
n = 1331; I2 = 49.1%) (Figure 1). The CI of the pooled effect was wide, ranging from a 25% reduction in
risk to a 31% increase in risk for suicide attempts.
The absolute proportion of participants with a suicide attempt or self-harm episode varied greatly
across trials, as did the difference between groups. For example, in the 3 trials reporting suicide
attempts or self-harm in 20% to 23% of control group participants, the proportion of youth in the
intervention groups with suicide attempts or self-harm ranged from 11.4% to 36.1% (72, 78, 84). Given the wide range of results, we
cannot rule out the possibility of harm (or benefit) on the basis of existing evidence.
Psychotherapy had no beneficial effect on suicidal ideation beyond usual care (SMD, −0.22 [CI,
−0.46 to 0.02]; 6 trials; n = 629; I2 =
41.2%) (Figure 2); both the psychotherapy and usual care groups
generally showed substantial improvement. Psychotherapy had a small beneficial effect on depression
(SMD, −0.36 [CI, −0.63 to −0.08]; 6 trials; n = 631;
I2 = 53.6) (Figure 3). Although
statistical heterogeneity was high, all effects suggested that psychotherapy benefited persons with
depression (but most effects were not statistically significant). Other health outcomes were sparsely
reported and rarely showed beneficial effects for the interventions.
Among psychotherapy studies, we found no clear predictors of effect size other than target age (adults
vs. adolescents), where trials in adults more consistently showed larger beneficial effects than those
in adolescents. Among adolescent trials, limited data suggested that interventions targeting parents
and youth, either separately or together, seemed to be more beneficial than those targeting only
We defined trials of “enhanced usual care” as those that attempted to improve the quality
or format of recommended treatment (in primary or specialty care) or patient adherence to usual care
while providing little to no direct therapeutic counseling or specific prescription for psychotherapy.
Treatments varied widely, from mail-only to case management interventions, but most involved
considerably less contact with patients than psychotherapy. One enhanced usual care trial was limited
to adolescents and young adults (aged 15 to 24 years) (82), 2
were limited to older adult primary care patients (42,
71), and the remaining trials included wide age ranges covering primarily adults.
Most trials targeted participants who had an emergency department visit or inpatient stay related to a
suicide attempt or self-harm. However, 3 trials were conducted in primary care settings, including
both trials in older adults. PROSPECT (Prevention of Suicide in Primary Care Elderly: Collaborative
Trial) addressed only depressed older adults (aged 60 to 94 years) and was the sole trial that used
primary care–based screening for depression in the United States to identify eligible
participants (42). The other trial of older adults was a
large cluster randomized trial (patients were randomly assigned at the level of provider) that
included all patients older than 60 years on the panels of participating providers; this trial thus
was not limited to patients who screened positive for suicidal ideation or had known risk factors for
suicide, but was representative of a general Australian primary care population (71).
The third trial with high applicability to primary care was a nonrandomized, population-based
intervention trial that compared intervention and control regions of the county in Hungary with the
highest suicide rates. This trial reported suicide rates per 100 000 persons as its outcome,
rather than following an identified sample (70). It involved
a 5-year provider-education intervention that also offered free consultation and a depression clinic
Although 7 of the 17 enhanced usual care trials reported deaths, only 1 fair-quality trial reported
significant group differences. This fairly large trial (n = 843) sent
participants 24 letters over 5 years expressing concern and encouraging treatment and reported a 49%
reduction in suicide deaths at 2-year follow-up (1.8% in the intervention group vs. 3.5% in the
control group; 1-tailed P = 0.043) (61). The large population-based trial in primary care practices seemed sufficiently powered
to examine suicide deaths and found no reduction in suicide (70). Aside from this population-based trial, very few deaths occurred across the remaining
trials, and deaths were frequently not reported so we could not conclude that suicide deaths
Thirteen of the 17 adult trials of enhanced usual care reported on suicide attempts, and all but 1
(53) found no difference in suicide attempts between 4 and 24
months (relative risk, 0.91 [CI, 0.80 to 1.02]; 13 trials; n = 6592;
I2 = 0.0%) (Figure 4). These
results are consistent with a small to moderate (20% at most) decrease in suicide attempts or no
effect, compared with suicide attempt rates in the control groups ranging from 1% to 30% at 4 to 24
Suicide attempts in enhanced usual care studies.
Weights are from random-effects analysis. DSH = deliberate self-harm; PROSPECT = Prevention of
Suicide in Primary Care Elderly: Collaborative Trial; SA = suicide attempt.
Other health and intermediate outcomes were sparsely reported. The trials in older primary care
patients reduced suicide attempts by 20% to 23% (42,
71), but the results were significant only in the larger trial (71).
We included 1 fair-quality, placebo-controlled trial of lithium to prevent suicide in patients with
depression-spectrum disorders and a recent suicide attempt (167 randomly assigned patients) (57). Retention in this trial was low (only 31% of participants
remained at the final 13-month follow-up). Although all 3 suicide deaths in the study occurred in
placebo recipients, the groups did not differ in cumulative survival without a suicide attempt (hazard
ratio, 0.517; P = 0.21, adjusted for age, sex, and prior suicide attempts) or
Although no harms were identified in any of the adult trials, 4 of the 12 trials reporting suicide
attempts in adolescents reported statistically nonsignificant increases in suicide attempts of 22% to
113% (73, 76, 77, 82). The possibility of harm
cannot be ruled out in treatment of currently or recently suicidal adolescents.
The trial of lithium treatment reported that 13% of the lithium recipients withdrew from the study
because of adverse effects, compared with 2% of placebo recipients. However, the statistical significance
of this difference was not reported (57). Overall withdrawal
rates for any reason were similar between groups. Specific adverse effects were not reported.
Suicide risk can be difficult to accurately assess because some persons may attempt to conceal suicidal
thoughts (creating false-negative results on screening) and some may express suicidal thoughts without
serious intention to kill themselves (creating false-positive results on screening) (113). Even in high-risk populations, suicide is comparatively rare. Furthermore, the
known risk factors associated with suicide are relatively common and individually not very strong predictors
of suicide, even in persons at high risk. This combination of factors makes accurately predicting who will
die by suicide on the basis of known risk factors very difficult. Nonetheless, suicide prevention is of high
national importance, so it is critical to know whether primary care–based screening is likely to help
reduce suicide in the United States by identifying patients in need of treatment and referring them to
appropriate care. The Table summarizes the evidence from this
review for all key questions.
Table. Summary of Evidence
Although screening instruments have been developed for quick risk assessment, few studies have
reported diagnostic accuracy characteristics of sensitivity, specificity, or related statistics
relative to an interview with a clinician or other trained questioner. Minimal evidence (2 studies)
suggested that screening tools can identify adults and older adults in primary care who are at
increased risk for suicide, although these tools produce many false-positive results. Data on the
accuracy of screening were even more limited in adolescents. Neither instrument performed well in
adolescents, and the screening populations in which they were tested had relatively poor applicability
to general primary care patients.
An important limitation of these data is the unknown accuracy of a full clinical interview in
predicting suicide-related events, which are relatively rare and inherently difficult to predict (17). Instrument accuracy aside, we identified minimal data that
examined whether suicide risk screening increased or decreased the likelihood of suicidality or other
distress. Our results are consistent with those of an earlier review of suicide screening in
adolescents, which concluded that data were very limited and future research was essential to
determine whether and how screening can reduce suicide in young persons (114).
Although the included studies involved too few deaths to determine whether a particular treatment
reduced the risk for suicide deaths, they provided useful evidence about attempts. Combined assessment
of all psychotherapy studies found that psychotherapy targeting suicide prevention reduced the risk
for attempts by an estimated average of 32%. Psychotherapy also showed small beneficial effects on
depression, although other beneficial outcomes were sparsely reported or showed no consistent group
differences. Interventions that primarily focused on enhancing usual care had little effect on suicide
deaths, suicide attempts, or related outcomes.
The participants in the included adult treatment trials who reported suicide attempts were at high
risk for suicide, usually based on a history of multiple attempts, which resulted in a very high
incidence of suicide attempts (for example, 11% to 68% in the control groups of the psychotherapy
trial). This finding contrasts with the screening accuracy studies that were done in general primary
care patients, where attempt rates are substantially lower (probably <1% over 1 year) (115). Thus, the indirect evidence linking screening accuracy with
benefits of treatment is not good.
There are several possible explanations for why psychotherapy seemed to be effective in adults,
whereas practice-based approaches or other enhanced usual care interventions were not. First, the care
provided in enhanced usual care trials was generally less time-intensive than that provided in the
psychotherapy trials, which may be associated with smaller effects. In addition, the enhanced usual
care trials may have included slightly lower-risk samples than the psychotherapy trials, as evidenced
by a smaller proportion of control group participants who attempted suicide at follow-up (0.5% to 28%
of control participants).
Alternatively, usual care may have been more effective in these studies, which would also attenuate
the effect (but which we could not examine with available evidence). Assuming that these interventions
are less likely, on average, to be effective, some of the enhanced usual care interventions may
nevertheless be useful components of a larger system-wide approach that includes psychotherapy.
We found minimal data on medication's effectiveness in preventing suicidal behavior. These data were
limited to a single, short-term, fair- to poor-quality lithium trial that had high attrition. Lithium
is commonly used for treating bipolar disorder and has been shown to reduce the risk for suicide in
observational studies (116, 117) and in controlled
trials of patients with unipolar and bipolar depression who were not necessarily suicidal, compared
with placebo or other agents (pooled Peto odds ratio of randomized trials, 0.26 [CI, 0.09 to 0.77])
(118). We found no studies on lithium use in patients
identified through screening for suicidality. Lithium is associated with important adverse effects
that were not described in the 1 trial included in this review (119–121).
Our findings were generally consistent with other recent reviews of treatment to prevent suicide or
self-harm (122–125). Each of
these recent reviews generally included similar bodies of research but grouped the trials differently.
Nonetheless, they all found insufficient evidence for an effect on suicide deaths because of the small
number of events. They also generally found small to moderate (usually nonsignificant) reductions in
suicide attempts or self-harm, and all were limited by the included trials' sparse reporting of other
outcomes. The most comprehensive of these reviews, published by the National Institute for Health and
Clinical Excellence, concluded that psychological and psychosocial interventions may be effective
compared with usual care. However, variations in populations, treatment methods, and comparison groups
Psychotherapy did not reduce the risk for suicide attempts in adolescents in contrast to adults. Data
did not allow us to rule out the possibility that the risk for suicide attempts was paradoxically
increased. Psychotherapy showed small beneficial effects on depression for adolescents, as it did for
adults. Other outcomes either showed no consistent beneficial effects or were sparsely reported.
The research on iatrogenic suicidality related to antidepressants suggests that adolescents react
differently from adults to pharmacologic treatment (126).
Research also suggests that risk factors and methods of committing suicide differ between younger
versus older teenagers (127). Thus, different age groups
seem to have different treatment needs and risks. The evidence base in adolescents is still small, and
few approaches have been replicated. Replication is important, as shown by the trials of developmental
behavior therapy in this review; beneficial results in a first trial (80) were not replicated in 2 subsequent good-quality trials (75, 77). Overall, our findings were consistent with the National Institute
for Health and Clinical Excellence review, in which only 1 trial showed a beneficial effect in
Psychotherapy trials primarily involved high-risk youth, most with a recent suicide attempt or acute
suicidal ideation. These samples are consistent with those in the screening studies but may have low
applicability to youth identified through primary care screening. Suicidal youth need treatment, but
caution, close monitoring, and care coordination are also warranted (128). These trials suggest that active parental involvement in treatment may be important.
Further research is urgently needed.
One important limitation is that most of the treatment literature was in high-risk populations, so the
generalizability of these results to screening-detected populations is unknown. In addition, there was
very little evidence on the effectiveness of treatment in older adults and racial or ethnic minorities.
Differences in suicide rates among ethnic groups suggest that cultures vary, both in motivation for and
meaning of suicide; thus, culturally tailored risk-based screening and interventions may be important
The lack of power and reports of suicide death is another important limitation. Suicide attempts and
self-harm are not good surrogates for suicide death. As a result, we cannot assume that the reduced
attempts seen with psychotherapy interventions will decrease the number of deaths (130). Because suicide death is relatively rare and predicting such deaths is
difficult, very large collaborative trials are probably required for sufficient power to see an effect on
suicide deaths (131). For example, if all participants in all
psychotherapy trials that reported suicide deaths were treated as a single study that found a 57%
reduction (0.62% in the intervention group vs. 1.44% in the control group), 4 times the number of actual
participants would be needed to achieve statistical significance.
Power would probably be even more dramatically limited in studies of screening-detected patients.
Assuming an annual suicide rate of 100 per 100 000 persons (twice as high as that of older white
men, who have the highest rates of any age, sex, or racial subgroup) and the ability of a treatment to
reduce suicide by 40%, more than 83 000 persons per group would be required to generate a
statistically significant result. Thus, building a coherent chain of evidence from broad population-based
screening through treatment to reduce suicide deaths will be difficult, because treatment studies will
necessarily be limited to very high-risk groups in order to have sufficient power to detect a treatment
Suicide prevention is a topic of high national importance in which primary care providers may have a
role. Although evidence was limited, primary care–feasible screening tools could probably identify
adult patients at increased risk for suicide who may need treatment. A larger body of evidence showed
that psychotherapy can reduce the risk for suicide attempts in high-risk populations.
Unfortunately, whether similar benefits would be found in screening-detected patients is unknown. There
was little evidence on the accuracy of screening in adolescents, and what little data are available
showed that treatment did not demonstrate a positive effect. Results in adolescents also did not rule out
the possibility of harm (that is, increased suicide attempts) with some forms of psychotherapy. More
research on how to effectively identify and treat adolescents at increased risk for suicide is urgently
needed. There is also a need for research on the effect of psychotherapy to prevent suicide attempts in
primary care patients who screen positive for suicide risk, as well as whether treatments actually lead
to lower suicide death rates, even in high-risk populations.
The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.
Richard G. Malish, Robert L. Czech, David W. Cole
USA MEDDAC Fort Stewart, GA and the 3rd Infantry Division
June 5, 2013
Conflict of Interest:
The opinions and assertions contained herein are the private views of the authors and are not to be construed as official or reflecting the views of the U.S. Army Medical Activity Fort Stewart, Georgia, the 3rd Infantry Division, the United States Army, or the United States Department of Defense. 1. O’Connor E, Gaynes BN, Burda BU, Soh C, Whitlock EP. Screening for and Treatment of Suicide Risk Relevant to Primary Care: A Systematic Review for the U.S. Preventive Task Force. Ann Intern Med. 2013; 158: 741-754.2. Screening for Suicide Risk: Recommendation and Rationale. Ann Intern Med. 2004: 140: 820-821.3. Zoroya G. Army, Navy suicides at record high. 18 November 2012. USA Today.
Evidence to Support Screening for Suicide Remains Insufficient
We would like to congratulate O’Conner et al. for their comprehensive review of the literature regarding the evidence in support of screening to prevent suicide. However, we do not believe the evidence presented in the manuscript supports the conclusion that, “primary care-feasible screening tools could probably identify adult patients at increased risk for suicide.”
1 We suspect that this conclusion may be flavored by unsupported optimism. We instead feel the 2004 United States Preventive Task Force (USPTF) statement is more technically sound. Specifically, “the evidence is insufficient to recommend for or against routine screening by primary care clinicians to detect suicide risk in the general population.”
2 In their article, O’Conner et al. recognize the growing epidemic of suicides in the general population. Because of its characteristics as a contained and easily studied population, the United States Army has recognized this phenomenon for several years. It would not be unfair to characterize our efforts to prevent suicide as a collective “obsession.” To this end, the Army has adopted many initiatives of which aggressive screening is one of two prominent pillars. Because of these efforts, U.S. military service members are likely (anecdotally) the population most screened for suicide in the world. Not only are Soldiers screened at every primary care visit, they are also screened annually as part of a periodic health assessment. Furthermore, non-medical military leaders complete periodic questionnaire-based interviews with Soldiers to identify their risk for suicide, substance abuse, sexual abuse, and other personal and societal dangers. These interviews vary by unit but occur no less than yearly. Finally, deploying Soldiers are screened for depression and suicidal ideation at 4-5 deployment-related touch-points.In spite of these screening efforts (which have escalated rapidly over the past five years), Army suicide rates have continued to increase.
3 Without evidence to the contrary, we must acknowledge the possibility that screening is as likely to be neutral or harmful as it is to be beneficial based on this observation. Repetitive screening may conceivably introduce suicide as a potential course of action for relief of psychic pain and emotional distress through suggestion. Anecdotal observations could as easily support a negative-leaning statement against screening as they do a slightly positive supporting one. In supporting screening for suicide, we believe O’Conner et al. may have fallen prey to the common misconception that detecting suicidal ideation in populations equates to preventing completed suicides. In fact, to our knowledge, no study designed to detect suicidal ideation has resulted in a reduction in suicide. This makes ideation endpoints invalid as predictors of suicide and inappropriate surrogates. While the authors correctly identify this pitfall in their text, the studies used to support the faintly positive recommendation in favor of screening, in fact, rely upon screening to detect suicidal behavior. None of the supporting studies was designed to demonstrate an effect on completed suicide. O’Conner et al. are conscientious in their efforts to qualify their assertion in favor of screening. But even the suggestion that “limited evidence” exists is misleading.It is important that scientific verbiage is technically precise because military and political leaders rely upon it to support multi-million dollar programs. Even the suggestion by experts that unproven methodologies “might help” may result in their adoption by government officials desperate to curb this alarming trend. This may divert resources from pivotal research or other pillars of suicide prevention (such as building resiliency or creating policy).
We should instead be quite frank and deliberate. Recognizing that many of the authors are likely the same for both works, nothing new is presented in the 2013 O’Conner et al. manuscript that repudiates the 2004 statement, “the USPTF found no evidence that screening for suicide reduces suicide attempts or mortality.”2
Elizabeth A. O'Connor, PhD, Evelyn P. Whitlock, MD, MPH, Bradley N. Gaynes, MD
July 2, 2013
We thank Dr. Malish and his colleagues for their comments. First, we would like to clarify that our review is a statement of the evidence related to suicide screening that the U.S. Preventive Services Task Force (USPSTF) used to develop their recommendation for suicide screening in primary care in the United States (1). Our task was to accurately characterize the body of evidence through application of rigorous systematic review methods. The USPSTF then used this review to interpret the strength and clinical significance of the evidence as it is related to primary care. The USPSTF’s draft recommendation has been published separately, and is consistent with the 2004 USPSTF recommendation that the evidence is insufficient to recommend for or against screening for suicide in primary care (2).
We agree that there was very little direct evidence that screening for suicide risk is either beneficial or harmful (3). Therefore, we examined the evidence for (a) whether there were screening instruments that adequately identify patients at increased risk of suicide, and (b) whether treatment was effective in patient populations similar to those who would be identified through screening. When we said that “primary care-feasible screening tools could probably identify adult patients at increased risk for suicide” we were referring only to the ability of the screening instruments to identify people who are suicidal, not to whether screening would reduce the rate of suicide.
Dr. Malish and colleagues point out that despite very high rates of screening in military populations, suicide in active military and veteran populations is still very high and has even increased, suggesting that screening has not been effective in preventing suicide and might even be counterproductive in this population. As outlined in our paper and review, we did not include such ecological data in our evidence review because this type of data cannot establish a causal connection between screening and suicide rate, especially given war-related events and potential social/cultural shifts that may have occurred in the same time period.
In summary, we agree that evidence on the effectiveness of screening programs is lacking and cannot assume that increased identification of suicidal persons will reduce the suicide rate. The experience of the military provides valuable insights that could have important implications for other high-risk populations and we appreciate the extensive efforts undertaken on behalf of service members and veterans to prevent suicide. We do not yet know how to prevent suicide at a population level, and we hope researchers or institutions, including the armed services, will help to identify measures that are effective in preventing suicide in the near future.
Elizabeth A. O’Connor, PhD
Evelyn P. Whitlock, MD MPH
Bradley N. Gaynes, MD
(1) U.S.Preventive Services Task Force. U.S. Preventive Services Task Force Procedure Manual. Rockville, MD: Agency for Healthcare Quality and Research; 2008. [PMID: None]
(2) U.S.Preventive Services Task Force. Screening for suicide risk in adolescents, adults and older adults: U.S. Preventive Services Task Force Recommendation Statement [DRAFT]. USPSTF Website [cited 2013 July 1]; Available from: URL:http://www.uspreventiveservicestaskforce.org/uspstf13/suicide/suicidedraftrec.htm [PMID: None]
(3) O'Connor E, Gaynes BN, Burda BU, Soh C, Whitlock EP. Screening for and treatment of suicide risk relevant to primary care: a systematic review for the U.S. Preventive Services Task Force. Ann Intern Med. 2013;158:741-754. [PMID: 23609101]
Results provided by:
Copyright © 2016 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use
This PDF is available to Subscribers Only