Summary of the Evidence

Bradley N. Gaynes, M.D., M.P.H^a; Suzanne L. West, Ph.D.^b; Carol A. Ford, M.D.^c; Paul Frame, M.D.^d; Jonathan Klein, M.D., M.P.H.^e; Kathleen N. Lohr, Ph.D.^f

The authors of this article are responsible for its contents, including any clinical or treatment recommendations. No statement in this article should be construed as an official position of the Agency for Healthcare Research and Quality or the U.S. Department of Health and Human Services.

Address correspondence to: Bradley N. Gaynes, M.D., M.P.H.; Department of Psychiatry; CB 7160; University of North Carolina; Chapel Hill, NC 27599; E-mail bgaynes@med.unc.edu

Select for copyright information. The USPSTF recommendations based on this review are online.

The summaries of the evidence briefly present evidence of effectiveness for preventive health services used in primary care clinical settings, including screening tests, counseling, and chemoprevention. They summarize the more detailed Systematic Evidence Reviews, which are used by the U.S. Preventive Services Task Force (USPSTF) to make recommendations.

Background: Suicide is the 11th leading cause of death and the seventh leading cause of years of potential life lost in the United States. Although suicide is of great public health significance, its clinical management is complicated.

Purpose: We systematically reviewed the literature to determine whether screening for suicide risk in primary care settings decreases morbidity, mortality, or both.

Data Sources: MEDLINE® (1966 to October 17, 2002), PsycINFO, Cochrane databases, hand-searched bibliographies, and experts.

Study Selection: For screening, we examined only English-language studies performed in primary care settings. For treatment, we included randomized controlled trials (RCTs), and cohort studies were included if they were performed in any setting where suicide completions, suicide attempts, or suicidal ideation were reported.

Data Extraction: A primary reviewer abstracted data on key variables of study population, design, and outcomes; a second reviewer checked information accuracy against the original articles.

Data Synthesis: No study directly addressed whether screening for suicide in primary care reduces morbidity and mortality; the remainder of the review focused on the questions of reliable screening tests for suicide risk and the effectiveness of interventions to decrease depression, suicidal ideation, and suicide attempts or completion. One screening study provided limited evidence for the accuracy of suicide screening in a primary care setting. Intervention studies provide fair and mixed evidence that treating those at risk for suicide reduces the number of suicide attempts or completions. The evidence suggests mild to moderate improvement for interventions addressing intermediate outcomes such as suicidal ideation, decreased depressive severity, decreased hopelessness, or improved level of functioning.

Conclusion: Because of the complexity of studying the risk for suicide and the paucity of well-designed research studies, only limited evidence guides the primary care clinician's assessment and management of suicide risk.

Suicide is a major public health problem in the United States. In 2001, suicide was the 11th leading cause of death in the United States, accounting for approximately 30,000 deaths with an age-adjusted rate of 10.7 per 100,000 persons.¹ Suicide accounts for 1.3 percent of total deaths, more than double the number due to HIV infection and AIDS.² It is the seventh leading cause of years of potential life lost, surpassing diabetes, liver disease, and HIV infection.¹ Annually, approximately 500,000 individuals require emergency department treatment in U.S. medical centers following attempted suicide.³ One of every 6 young adults describes having suicidal ideation at some point in their lives, and 5.5 percent report ever having made a suicide attempt.⁴ The public health significance of this problem was underscored by The Surgeon General's Call to Action to Prevent Suicide,³ which proposed completion of a National Strategy for Suicide Prevention.⁵

Relevant demographic risk factors have been identified. Individuals aged 65 and older are at the highest risk for completed suicide; white men aged 85 and older have an especially high rate (59/100,000).⁵ Suicide also affects adolescent and young adults; it is the third leading cause of death among persons aged 15 to 24 (10.3/100,000), following unintentional injuries and homicide.¹ Rates of suicide attempts and completions differ by sex; men have a higher rate of suicide completion, whereas women have a higher rate of attempts.⁶ Finally, suicide behaviors vary widely by race and ethnicity. Nearly 75 percent of all completed suicides are by white males,² who have a 2-fold higher risk for suicide than black men (11.7/100,000 vs. 5.5/100,000).¹

Clinical risk factors have also been identified. Suicide is closely related to psychiatric illness. More than 90 percent of those who complete suicide have a diagnosable psychiatric illness at the time of death, usually depression, alcohol abuse, or both.⁷ Hopelessness, often present in those with severe depressive illness and a history of previous suicide attempts are particularly strong and independent prospective risk factors for a suicide.⁸ Although prior suicide attempts are a key risk factor, two-thirds of suicides occur on the first attempt,⁹ and suicide attempts remain substantially more common than completed suicides by a factor between 10 and 20.⁷ Other identified risk factors for completed suicide include being widowed or divorced, living alone, having a recent harmful event (such as job loss or death of loved one), having severe anxiety, having a chronic medical illness (especially a central nervous system disorder), and having a family history of suicide attempts or completions (Hirschfield and Russell⁷ and Mann⁹ provide more detailed reviews).

Primary care physicians have a key role in identifying and managing suicidal tendency. Patients endorsing suicidal ideation, a key factor in the assessment of suicide risk, are not uncommon in primary care; between 2 percent and 3 percent of primary care patients report suicidal ideation in the pervious month.^10,11 Furthermore, most patients completing suicide have recently visited their primary care physician. Approximately one-half to two-thirds of individuals who commit suicide visit physicians within 1 month of taking their lives; 10 percent to 40 percent visit in the week before.^12-15

The clinical management of suicide risk is complicated. Suicide is a rare event. It has a low prevalence in the general population (0.01 percent)¹⁶ and, despite a 10-fold increase in adults with depression, most depressed patients (99.9 percent) do not commit suicide.¹⁷ As a result, many clinical trials on the management of suicide risk have focused on high-risk patients, such as those with a history of deliberate self-harm.

Deliberate self-harm, understood as an intentionally initiated act of self-harm with nonfatal outcome (including self-poisoning and self-injury), encompasses terms such as attempted suicide and parasuicide.¹⁸ Deliberate self-harm is not synonymous with attempted suicide. Attempted suicide, understood as a self-initiated act with the intent of ending one's life, is only a single example of deliberate self-harm. Still, deliberate self-harm is a recurrent behavior with important long-term risks. Between 15 percent and 23 percent of patients who are seen for deliberate self-harm will be seen for treatment of a subsequent episode within 1 year¹⁹and are at high risk for repeat deliberate self-harm in the weeks following an episode.²⁰ Of those with an episode of deliberate self-harm, 3 percent to 5 percent die by suicide within 5 to 10 years.²¹ Identification of deliberate self-harm is relevant to primary care practice, since two-thirds of patients who deliberately harm themselves visit their general practitioner within 12 weeks of the episode.²² Patients with borderline personality disorder are at increased risk for deliberate self-harm, with groups from psychiatric and primary care settings having similar self-harm profiles.²³

Given the Surgeon General's call to action, clarification of the available evidence base guiding the clinical management of suicide risk is especially pertinent. As part of the USPSTF update of its 1996 recommendation,²⁴ we examined the evidence addressing whether primary care identification and treatment of suicide risk improves outcomes in patients whose risk had previously been unidentified. Our full systematic evidence review set out to answer 8 key questions (Appendix Figure 1). In this article, we report on the 3 key questions for which we found data meeting our selection criteria:

1. Can a screening test reliably detect suicide risk in primary care populations?
2. For those at risk, does treatment result in decreased suicide attempts or completions?
3. For those at risk, does treatment result in improved intermediate outcomes (e.g., decreased suicidal ideation or depressive severity)?

Using USPSTF methods,²⁵ we developed an analytic framework and 8 key questions to guide our literature searches. (Appendix Table 1, Appendix Table 2, Appendix Table 3, and Appendix Figure 1.) Our population of interest was primary care patients with previously unidentified suicide risk.

To identify relevant articles, we searched the MEDLINE® database from 1966 to October 17, 2002, beginning with the terms suicide or suicide, attempted. We supplemented these sources by using the same search terms in PsycINFO; searching the Cochrane Collaboration Library; and hand searching the bibliographies of systematic reviews, relevant original articles, and the 1996 edition of the Guide to Clinical Preventive Services.²⁴ We additionally reran searches using deliberate self-harm as a search term and identified no further articles. In this paper, we present our findings for studies involving adults.

We found 1 well-conducted, recent systematic review by Hawton et al. concerning treatment of deliberate self-harm, which was relevant to reducing suicide attempts or completions, our primary outcome.¹⁸ We found another recent, well-done systematic review relevant to intermediate outcomes.²⁶ We checked our study results against the studies in these reviews, and we examined in detail only those studies that had not been included in the systematic reviews.

Two of the authors independently reviewed all titles and abstracts. If either reviewer determined that the study met inclusion criteria, we retrieved the full paper for further evaluation. Two of the authors subsequently reviewed the studies to determine final inclusion, adjudicating disagreements by consensus discussion.

A primary reviewer abstracted relevant information into evidence tables. As part of this abstraction, the primary reviewer rated the internal and external validity for each article using criteria developed by the USPSTF Methods Work Group.²⁵ A second reviewer checked the accuracy of the abstracted information against the original articles, while the first author reviewed all quality ratings to ensure consistency.

We required that screening studies be performed within a primary care setting, but treatment studies could be performed in either primary or specialty care settings. This strategy reflected our idea that screening must be performed in primary care but should suicide risk be identified, a primary care physician could refer patients for subsequent treatment.

For screening studies, inclusion required comparison with a gold standard. For treatment studies, inclusion required that trials report suicide completions, suicide attempts, or suicidal ideation as primary outcomes. We excluded the following: clinical trials targeting patients with chronic psychotic illnesses because these patients would already be identified as having increased suicide risk; randomized controlled trials (RCTs) that did not supply sufficient detail to allow direct comparison of outcomes between intervention and control groups; and cohort studies that did not have either a similar clinical presentation for intervention and control groups or an independent control group.

This research was funded through a contract to the RTI International-University of North Carolina Evidence-based Practice Center from the Agency for Healthcare Research and Quality (AHRQ). Agency staff and Task Force members participated in the study design and reviewed draft and final manuscripts A more comprehensive report was distributed for review to content experts and revised accordingly before preparation of this manuscript.²⁷

Can a Screening Test Reliably Detect Suicide Risk in Primary Care Populations?

Our evaluation identified 1 relevant article assessing an instrument's operating characteristics for identifying suicide risk in primary care.¹⁰ The Symptom-Driven Diagnostic System for Primary Care (SDDS-PC),²⁸ a 62-item self-report instrument designed to help identify psychiatric illness in primary care settings, contains 3 items assessing suicide risk (all within the past month).¹⁰ Data on suicidal thoughts, plans, and past attempts were systematically collected using a nurse-administered, face-to-face structured interview conducted immediately before the medical visit. The individual operating characteristics of the 3 items were compared with a structured interview for identifying a plan to commit suicide (the gold standard).

"Thoughts of death" had 100 percent sensitivity, 81 percent specificity, and 5.9 percent positive predictive value for detecting patients with a plan to commit suicide. Endorsing "wishing you were dead" had 92 percent sensitivity, 93 percent specificity, and 14 percent positive predictive value; "feeling suicidal" had 83 percent sensitivity, 98 percent specificity, and 30 percent positive predictive value. Of those "feeling suicidal," 85 percent had a psychiatric disorder as determined by structured clinical interviews.²⁹ Only major depression (odds ratio [OR], 33.1; 95 percent confidence interval [CI], 10.9-99.6) and drug abuse or dependence (OR, 16.7; 95% CI, 3.9-71.4) were independently associated with suicidal ideation. Of note, only 46 percent of those eligible for this study agreed to participate.

Given the rarity of suicide attempts in the primary care population, finding an accurate screening strategy for suicide risk in a primary care setting is a daunting challenge. This is illustrated by the following hypothetical situation. Consider a screening instrument (e.g., endorsing "feeling suicidal") that identifies patients at high risk and has reasonable test characteristics (e.g., sensitivity of 80 percent and specificity of 70 percent, figures similar to screening tools for depression).^30,31 Apply this tool to a population of 10,000 in which 10 patients will attempt suicide (10-fold more than the 10 in 100,000 persons who will complete suicide), it will produce 8 true-positive results, 2 false-negative results, and 2,997 false-positive results, a positive predictive value of 0.3 percent. This high proportion of false-positive results could generate a substantial time and cost burden. Using the higher specificity result (98 percent) from the prior study produces substantially fewer false-positive results (n = 200), but only a slightly improved positive predictive value (3.8 percent).

Does Treatment Reduce Suicide Attempts or Completions?

We report findings first from RCTs and then from cohort studies. Within each study design section, we provide the evidence stratified by age whenever possible.

Randomized Controlled Trials

All 30 RCTs that met our inclusion criteria²⁷ involved high-risk groups as identified by a deliberate self-harm episode, a diagnosis of borderline personality disorder, or admission to a psychiatric unit. However, only Motto and Bostrom³² and Rudd et al.³³ identified depressive illness as part of their eligibility criteria, and no studies focused primarily on depressed patients with suicidal ideation. Of the 2 studies directly involving primary care, 1 recruited some of its patients from a primary care setting (although the intervention occurred in a psychiatric outpatient setting)³⁴ and the other conducted its intervention in a primary care setting.³⁵

Trials focusing on adolescents or young adults and elderly adults, the 2 populations of greatest clinical concern, were limited. The included studies involved either adults only or adults and older adolescents but did not differentiate further by age in the analyses; we review them together.

We found no published intervention study for the elderly population. However, the Prevention of Suicide in Primary Care Elderly-Collaborative Trial (PROSPECT) is currently being conducted.³⁶ PROSPECT aims to determine whether placement of a depression health specialist in primary care practices has a favorable impact on rates of depression, hopelessness, and suicidal ideation in elderly primary care patients with depressive illness. Initial outcomes for the 4- and 8-month followup periods were expected in 2003.

We organized our review of the 30 RCTs as follows. Because Hawton et al.¹⁸ had systematically reviewed 21 of these trials,^37-57 we first briefly summarize the results of their meta-analysis (Table 1). We then provide greater detail on the 9 additional RCTs of deliberate self-harm that our literature search identified (Table 2).^32-35,58-63 two articles^58,59 were from a single trial and are counted as 1 trial. Given the substantial heterogeneity of the populations enrolled, the interventions tested, the length of followup periods, and the outcomes measured, we concluded that integrating the new studies into the prior summary was not warranted.

Prior Review of RCTs for Deliberate Self-Harm

Of the 21 studies of adults receiving treatments for deliberate self-harm (Table 1),^37-57 12 included older adolescents.^{37-44,46,47,52,54} Although some trends suggested incremental benefit from certain interventions compared with usual care, interventions for which more than 1 study was performed produced no statistically significant effects by meta-analysis. The most promising intervention was problem-solving therapy, a short-term, cognitively-oriented psychotherapy. In 5 studies of this intervention versus standard aftercare, the summary OR showed a trend toward decreasing deliberate self-harm (OR, 0.70; 95% CI, 0.45-1.11).^37-41 Of note, both the form and duration of treatment varied considerably within these 5 studies.

Intensive care plus outreach versus standard aftercare (6 studies)^42-47 produced a summary OR of 0.83 (95% CI, 0.61-1.14). Again, the form and duration of treatment varied substantially among the studies. One large trial comparing provision of both physician contact and crisis intervention assistance showed a trend toward a decreased likelihood of repeating deliberate self-harm in favor of the intervention (OR, 0.43; 95% CI, 0.15-1.27) compared with standard care.⁴⁸

Two interventions, each supported by a single study involving a maximum of 20 patients in each group, reported statistically significant reduced repetition of deliberate self-harm. Dialectical behavior therapy (DBT), a comprehensive treatment program developed to treat severely dysfunctional individuals with borderline personality disorders by improving emotional and behavioral management skills, significantly reduced repetition of deliberate self-harm for patients with borderline personality disorder and recent deliberate self-harm compared with standard care (OR, 0.24; 95% CI, 0.06-0.93).⁴⁹ Administration of the antipsychotic flupenthixol significantly reduced the proportion of repeated deliberate self-harm for those with a history of at least 2 prior suicide attempts compared with placebo (OR, 0.09; 95% CI, 0.02-0.50).⁵³

Additional RCTs of Deliberate Self-Harm

Of the 9 additional studies involving repetition of deliberate self-harm identified in our literature search (Table 2),^32-35,58-63 2 showed benefit. Guthrie et al. found significant benefit from interpersonal psychotherapy, a time-limited method that focuses on resolving current interpersonal problems to improve symptoms, compared with standard care.⁶⁰ Patients who presented to an emergency department with deliberate self-harm, but did not require medical or psychiatric hospitalization, were enrolled. Participants were randomized to either 4, 50-minute sessions of interpersonal psychotherapy delivered by nurse therapists in the patient's home, or usual care. Of those eligible (n = 119), 51 percent participated; those refusing were at a greater suicide risk as indicated by clinical measures. In an intention-to-treat analysis, those in the interpersonal psychotherapy group were less likely to have a repeat episode of deliberate self-harm in the subsequent 6-month period (8.6 percent vs. 27.9 percent, P <0.001).

Bateman and Fonagy compared psychoanalytically oriented partial hospitalization with standard psychiatric outpatient aftercare for patients with borderline personality disorder.^58,59 Treatment occurred for a maximum of 18 months. Twenty-two patients were initially randomized to each group; analysis was not intention-to-treat. After 18 months of treatment, the percentage of those with suicide attempts within the prior 6 months was significantly lower in the treatment group than in the control group (53 percent for intervention group, no rate given for control group, but a graph suggests approximately 40 percent [P <0.001]). At the 36-month followup, a significantly smaller proportion of the partial hospitalization group made a suicide attempt compared with the usual care group (18.2 percent vs. 63.2 percent; no OR calculated; P <0.004).

The remaining 7 studies identified no benefit from interventions (Table 2). In 6 studies in which treatment was provided in the primary care setting, interventions included providing an emergency information card,⁶² a letter,³² outpatient day hospitalization,³³ antidepressant medication,⁶¹ antipsychotic medication,⁶³ and DBT.³⁴ All interventions were compared with usual care or placebo (for the 2 medication studies). Of interest, a subgroup analysis in the Evans et al. emergency information card study suggested a need to examine the data by whether previous deliberate self-harm had occurred; for those with a history of deliberate self-harm, the intervention increased the likelihood of repeat deliberate self-harm.⁶²

In the only study that tested an intervention for suicide risk in the primary care setting, Bennewith et al. compared a 3-part, 1-time intervention with usual care.³⁵ The intervention provided general practitioners with:

1. A letter informing them of a patient's deliberate self-harm episode.
2. A letter the physicians could forward to the patient inviting him or her to make an appointment.
3. Guidelines on assessing and managing deliberate self-harm in general practice.

In an intention-to-treat analysis at a 12-month followup, the groups did not differ significantly in the proportion of patients who attempted suicide (21.9 percent vs. 19.5 percent). Adherence to this low-intensity intervention was poor; only 58 percent of the intervention-group physicians sent letters to the patients.

Of note, the investigators reported a subgroup analysis with results opposite to those of Evans et al.⁶² For patients with prior deliberate self-harm, this primary care intervention significantly decreased the likelihood of repeat deliberate self-harm, whereas for those with no prior deliberate self-harm, the intervention increased the likelihood of repetition (Table 2). The variability of adherence in the Bennewith et al. study and the differences in the 2 trials' study populations may partially explain the contradictory results.

Cohort Studies

Two cohort studies, each using depression as part of how they selected participants, met our inclusion criteria.^64,65 Neither study produced statistically significant differences involving repeated suicidal behavior (Table 3).

Using a nested case-control design, Coryell et al. evaluated suicide risk in a long-term cohort of patients with major affective disorders.⁶⁴ In this small study, case-patients were compared with controls for use of lithium in the week before the suicide completion (15 case-patients vs. 15 matched controls) or suicide attempt (41 case-patients vs. 41 matched controls); all were receiving some type of treatment at the time of the episode. The investigators found no relationship between lithium use and suicide or suicide attempts.

Raj et al. compared the use of 10 weekly sessions of cognitive-behavioral counseling with routine medical treatment for patients admitted to the intensive care unit (ICU) of a general hospital following their first or second suicide attempt.⁶⁵ Upon ICU admission, the 40 enrolled patients were sequentially assigned to either the counseling intervention or routine medical care with the option to attend therapy sessions. None of the intervention group repeated a suicide attempt at 2 to 3 months of followup; 1 patient in the control group made a repeated suicide attempt.

Does Treatment Result in Improved Intermediate Outcomes?

We identified 1 systematic review²⁶ and 4 additional articles that studied intermediate outcomes in patients at high risk for suicide.^34,60,65,66 Again, study heterogeneity (interventions tested, treatment duration, followup length, and outcomes used) precluded integration of the new studies into the prior review.

Prior Review of RCTs Involving Intermediate Outcomes

Townsend et al. conducted a systematic review of 6 RCTs^37-41,67 involving brief problem-solving therapy in patients with deliberate self-harm, in which the outcomes included depressive severity, hopelessness, and improvement in problems (Table 4).²⁶ Treatment duration and length of followup varied substantially across studies, and the analyses were not stratified according to age. The 4 studies that evaluated depressive outcomes^37-39,41 used 2 different scales for depression, requiring the authors to calculate a standardized mean difference (SMD; the mean difference divided by the pooled sample standard deviation) to evaluate depressive symptoms. The summary SMD indicated a significantly lower depression score of about one-third of a standard deviation for patients offered problem-solving therapy compared with those receiving usual care (-0.36; 95% CI, -0.61 to -0.11). Three trials measured hopelessness, which is strongly correlated with suicidal ideation^39,40,67 using the Beck Hopelessness Scale.⁶⁸ The authors calculated a weighted mean difference with those receiving problem-solving therapy averaging approximately 3 points less on hopelessness scores at followup than did those receiving standard care (-2.97 points; 95% CI, -4.81 to -1.13). Two trials measured whether problems had improved (a dichotomous measure rated by assessors blinded to treatment).^37,38 Improvement in problems was more likely in those receiving problem-solving therapy compared with those receiving usual care (OR, 2.31; 95% CI, 1.29-4.13) (Table 4).

Additional RCTs Involving Intermediate Outcomes

As shown in Table 5, Guthrie et al. measured suicidal ideation in an RCT comparing interpersonal psychotherapy with usual care.⁶⁰ Suicidal ideation, as measured by the Scale for Suicidal Ideation (SSI)⁶⁹ at 6-month followup, was significantly lower for the psychotherapy group (mean difference, -4.9; 95% CI, -8.2 to -1.6; P <0.001). A priori, the authors had identified a difference of 5 points as being clinically significant.

Koons et al. measured suicidal ideation and depressive severity in their 6-month RCT of women veterans with borderline personality disorder.³⁴ For those completing treatment, DBT was superior to usual care in decreasing suicidal ideation as measured by the SSI⁶⁹ (10-point decrease vs. 4-point decrease; P <0.05). As measured by the self-report Beck Depression Inventory,⁷⁰ DBT produced a significantly greater decrease in depressive symptoms than usual care (2-way analysis of variance, P <0.05), which is inconsistent with the authors' findings using the Hamilton Depressive Rating Scale (HAM-D),⁷¹ which showed no significantly greater decrease for DBT versus usual care.

Montgomery et al. performed a 4-week cohort study comparing the antidepressants mianserin, amitriptyline, and maprotiline.⁶⁶ Suicidal ideation, as measured by the Montgomery-Asberg Depression Rating Scale, was decreased by a significantly greater degree by mianserin compared with maprotiline (P <0.01); a trend favoring mianserin over amitriptyline was also observed (P <0.10). The 3 study drugs showed no differences for the analogous "suicidal thoughts" on the HAM-D, and the overall quality of the study was poor.

Raj et al. also measured the effect of a cognitive-behavioral intervention on suicide ideation.⁶⁵ Assessing the difference in SSI scores between baseline and 2 to 3 months post-discharge for the 2 groups, they found that those who received counseling had a substantially greater reduction in suicidal ideation than the usual care group (mean decrease 15 vs. 2.75; P <0.001).

Evidence for or against the value of screening for suicide risk in primary care settings must be considered within a complex practice and epidemiological context. Suicide is a rare outcome, even among high-risk groups; this fact alone creates methodological challenges. RCTs, the gold standard for showing efficacy in evidence reviews, ethically cannot include a true placebo arm; consequently, all interventions are being compared with treatment arms that in fact may or may not be effective. Finally, patterns of suicide behaviors are very complex. Although a prior suicide attempt is a strong risk factor for completed suicide, sociodemographic characteristics and behaviors clearly differ across groups of those who attempt suicide, practice repetitive deliberate self-harm, and successfully complete suicide. To focus exclusively on completed suicide reveals dramatic differences in rates and methods across the life span, between males and females, and between different race and ethnicity groups. Current research, in large part, does not address this complexity.

Within this context, we have reviewed literature published since 1966 with the goal of better defining the clinician's role in screening for suicide risk in primary care settings. Despite the public health import of suicide and the Surgeon General's call to action, evidence to guide the primary care clinician's assessment and management of suicide risk is extremely limited. No studies address the overarching question of whether screening for suicide risk in primary care patients improves outcome. Consequently, we must approach this issue by analyzing studies examining the intervening linkage questions.

Very little is known about use of screening instruments for suicide risk in primary care populations. One prospective study identified reasonable test characteristics for persons reporting that they were "feeling suicidal" compared with responses indicating the presence of a plan. This study has not been replicated, nor has the specific question identified ("feeling suicidal") been tested independently of the longer instrument.

Regarding whether interventions for those at risk reduce suicide attempts or completions, the poor generalizability of the studies makes the overall strength of evidence fair, at best, while the results are mixed.²⁵Although some trends suggest incremental benefit from several interventions, no consistent statistically significant effects have emerged for interventions for which more than 1 study has been done. Of the interventions for which only 1 study has been done, promising interventions included DBT for borderline personality disorder⁴⁹ and interpersonal psychotherapy for deliberate self-harm.⁶⁰ These interventions, however, require further confirmation.

We should emphasize that our review did not include all of the available clinical trial literature involving suicide attempts or completions. Some literature has examined the effectiveness of medications, such as lithium, in the prevention of suicide among psychiatric patients with major mood disorders, as reflected in a recent meta-analysis by Tondo et al.⁷² We excluded these studies because they did not meet our inclusion criteria of controlled trials with adequate comparison groups.

Several studies showed improvement for intermediate outcomes, primarily for persons at high risk for deliberate self-harm. Specifically, meta-analyses of RCTs using problem-solving therapy have shown benefit, as indicated by improved mood, decreased hopelessness, and improvement in problems.²⁶ In addition, 1 RCT involving interpersonal psychotherapy⁶⁰ and 1 RCT involving DBT³⁴ documented decreased suicidal ideation; finally, 1 cohort study of cognitive-behavioral therapy showed decreased suicidal ideation.⁶⁵

Priorities for a Research Agenda

Our review highlights several important issues involving research on assessing and managing suicide risk. First, the challenge of studying interventions for a rare event is underscored by the fact that, even in a population with a relatively high risk for deliberate self-harm, documenting incremental benefit relative to standard care has been difficult. This difficulty is attributable at least in part to the fact that most studies are underpowered to detect significant differences, thereby increasing the risk for falsely concluding that an effective intervention does not produce a statistically significant benefit, whereas studies that have larger sample sizes typically provide the least intense (and, arguably, likely less efficacious) interventions.³⁵ Future research must consider the feasibility of large, multi-site studies that have sufficient power to identify the benefit of interventions for a substantial health problem that is a relatively rare event.

Second, the generalizability of the available evidence to a primary care population with unidentified suicide risk is poor. The great majority of research has been conducted in psychiatric populations with an already identified risk for suicide rather than among unidentified patients in primary care, who as a group are at lower risk. The existing literature includes only 1 screening study conducted in a primary care setting.¹⁰ Only 1 intervention study involved patients recruited from primary care practices,³⁴ and all the intervention studies involved patients identified as being at high risk for harming themselves (and, consequently, are likely to be in treatment with a mental health professional). Only 1 study conducted the intervention in a primary care setting.³⁵ High priorities for future research include examining the test characteristics of instruments used to determine suicide risk in primary care settings, recruiting patients for intervention studies from primary care settings, and testing interventions in primary care settings.

Third, the available studies focused on those with relatively moderate risk for suicide and, for ethical and clinical reasons, excluded patients at the highest risk. Most identified high-risk patients are likely admitted to a psychiatric unit for safety, which may or may not in itself be an effective intervention. Subsequent research should consider how to stratify at-risk primary care patients and target interventions to risk severity.

Fourth, the lack of evidence for incremental benefit from a particular intervention compared with standard care is not equivalent to saying that nothing works. Standard care in many instances may be a successful intervention; it may be "good enough." However, the components of standard care are poorly described in the existing literature and probably vary across studies, making the comparison to the experimental intervention difficult to evaluate. Subsequent research could address this shortcoming by more carefully monitoring and defining standard care.

Fifth, making meaningful conclusions specific to any particular age group is difficult. Available studies were not stratified by age; as a result, drawing conclusions specific to young adults or elderly adults is a challenge. In addition, despite the concern about increased risk for suicide in the elderly, there is a dearth of information to guide evidence-based assessment and management strategies in primary care. Results from the PROSPECT trial will begin to fill this void.³⁶ Subsequent research should involve populations with more clearly defined age groups and analyses stratified by age to allow more meaningful interpretation for specific high-risk age groups.

Sixth, dramatic differences in suicide behaviors among men and women and among different racial and ethnic groups have drawn little attention. A better understanding of these variations may have direct implications for screening and treatment strategies, and they warrant further research.

Seventh, our review is relevant only to those individuals who access clinical care, which means that a large portion of the population may be ignored. Community-based research can presumably address this question.

Finally, we did not find studies meeting our inclusion criteria that addressed whether more adequate treatment of depressed patients or substance-abusing patients will decrease the risk for suicide. We think such a clinically guided approach is key for the primary care physician to balance effectively the public health import of suicide with the real challenge of improving the outcome of a rare event. Approximately 90 percent of patients who completed suicide have a diagnosable psychiatric illness, with the great preponderance having depression or substance abuse. A more feasible means of decreasing suicide may be to focus on the high-risk groups, such as depressed primary care patients for whom routine screening is already recommended,⁷³ and to focus efforts to decrease risk toward improving the adequate management of depression.⁷⁴ Improving depression management may both improve depressive outcomes and decrease suicide risk. This strategy is reasonable and practical from a clinical perspective and testable from a research perspective. It is also necessary. Assessing suicidal ideation is the standard of care in the evaluation for depression, and routine depression screening will likely identify more patients with suicidal ideation, for which primary care clinicians will need evidence-based management strategies. Retrospective analyses have suggested that educating general practitioners on better identification and treatment of depression may be an effective method of suicide prevention.⁷⁵ Subsequent prospective clinical trials focusing on primary care are needed to develop this evidence base.

This study was performed by the RTI-University of North Carolina at Chapel Hill (RTI-UNC) Evidence-based Practice Center under contract to the AHRQ, Contract No 290-97-0011, Task Order No. 3, Rockville MD.

We acknowledge the continuing support of David Atkins, M.D., M.P.H., who is the Chief Medical Officer for AHRQ's Center for Practice and Technology Assessment; and Jean Slutsky, P.A., M.S.P.H., AHRQ Task Order Officer for the USPSTF project. We appreciate the considerable support and contributions of members of the RTI International staff: Sonya Sutton, B.S.P.H. and Loraine Monroe. In addition, we thank the staff from the University of North Carolina at Chapel Hill and the Cecil G. Sheps Center for Health Services Research: Carol Krasnov for administrative assistance and coordination and Timothy S. Carey, M.D., M.P.H., Director of the Sheps Center and Co-director of the RTI International-University of North Carolina Evidence-based Practice Center.

Author Affiliations

[a] Department of Psychiatry, University of North Carolina at Chapel Hill, Chapel Hill, NC
[b] Department of Obstetrics and Gynecology, School of Medicine, Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC
[c] Departments of Pediatrics and Internal Medicine, University of North Carolina at Chapel Hill, Chapel Hill, NC
[d] Tri-County Family Medicine, Cohocton, NY
[e] Department of Pediatrics, University of Rochester, Rochester, NY
[f] Department of Health Policy and Administration, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC, and RTI International, Research Triangle Park, NC

Copyright and Source Information

This document is in the public domain within the United States.

Requests for linking or to incorporate content in electronic resources should be sent via the USPSTF contact form.

Source: Gaynes BN, West SL, Ford CA, Frame P, Klein J, Lohr KN. Screening for suicide risk in adults. A summary of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med 2004;140:822-35.

The USPSTF recommendation based on this evidence review is online.

NA = not applicable.
*One additional article abstracted; see text.

Note: CI, confidence interval; DSH, deliberate self-harm; GHQ, Generalized Health Questionnaire; NR, not reported; OR, odds ratio.
^a Adapted from Hawton, et al., 2001.¹⁸

Table 2. Additional Randomized Controlled Trials of Interventions to Reduce Deliberate Self-Harm in Adults and Older Adolescents^a

Note: CI, confidence interval; DSH, deliberate self-harm; NA, not available; NR, not reported; OR, odds ratio; SD, standard deviation.
^a Not in Hawton, et al. review^18
b Inclusion criteria required diagnosis of borderline personality disorder; all others required DSH.
^c Primary care treatment setting; all other studies conducted in specialty care settings.

Note: DSH, deliberate self-harm; NA, not available.

Note: CI, confidence interval; NR, not reported; OR, odds ratio.
^a Data are from Townsend, et al., 2001.²⁶ Population and age range already described for each study in Table 1 except for Patsiokas and Clum, for which the study was conducted in "adult" deliberate self-harm patients recruited from a U.S. inpatient psychiatric ward.

Note: BDI, Beck Depression Inventory (self-report); CI, confidence interval; HAM-D, Hamilton Depressive Rating Scale; MADRS, Montgomery-Asberg Depression Rating Scale; NA, not available; NR, not reported; NS, not significant; RCT, randomized controlled trial; SI, suicidal ideation.