National Trends in Suicide Attempts Among Adults in the United States

Key Points

Questions  Has a national increase in suicide attempts occurred in the United States in the decade since wave 2 of the National Epidemiologic Survey on Alcohol and Related Conditions?

Finding  In this national epidemiologic survey of 69 341 US adults, the percentage making a recent suicide attempt increased from 0.62% in 2004 through 2005 to 0.79% in 2012 through 2013. The adjusted risk differences for suicide attempts were significantly larger among adults aged 21 to 34 years than among adults aged 65 years or older; adults with no more than a high school education than among college graduates; and adults with antisocial personality disorder, a history of violent behavior, anxiety disorders, or depressive disorders than among adults without these conditions.

Meaning  A recent overall increase in suicide attempts among US adults has disproportionately affected younger adults with less formal education and those with antisocial personality disorder, anxiety disorders, depressive disorders, and a history of violence.

Importance  A recent increase in suicide in the United States has raised public and clinical interest in determining whether a coincident national increase in suicide attempts has occurred and in characterizing trends in suicide attempts among sociodemographic and clinical groups.

Objective  To describe trends in recent suicide attempts in the United States.

Design, Setting, and Participants  Data came from the 2004-2005 wave 2 National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) and the 2012-2013 NESARC-III. These nationally representative surveys asked identical questions to 69 341 adults, 21 years and older, concerning the occurrence and timing of suicide attempts. Risk differences adjusted for age, sex, and race/ethnicity (ARDs) assessed trends from the 2004-2005 to 2012-2013 surveys in suicide attempts across sociodemographic and psychiatric disorder strata. Additive interactions tests compared the magnitude of trends in prevalence of suicide attempts across levels of sociodemographic and psychiatric disorder groups. The analyses were performed from February 8, 2017, through May 31, 2017.

Main Outcomes and Measures  Self-reported attempted suicide in the 3 years before the interview.

Results  With use of data from the 69 341 participants (42.8% men and 57.2% women; mean [SD] age, 48.1 [17.2] years), the weighted percentage of US adults making a recent suicide attempt increased from 0.62% in 2004-2005 (221 of 34 629) to 0.79% in 2012-2013 (305 of 34 712; ARD, 0.17%; 95% CI, 0.01%-0.33%; P = .04). In both surveys, most adults with recent suicide attempts were female (2004-2005, 60.17%; 2012-2013, 60.94%) and younger than 50 years (2004-2005, 84.75%; 2012-2013, 80.38%). The ARD for suicide attempts was significantly larger among adults aged 21 to 34 years (0.48%; 95% CI, 0.09% to 0.87%) than among adults 65 years and older (0.06%; 95% CI, −0.02% to 0.14%; interaction P = .04). The ARD for suicide attempts was also significantly larger among adults with no more than a high school education (0.49%; 95% CI, 0.18% to 0.80%) than among college graduates (0.03%; 95% CI, −0.17% to 0.23%; interaction P = .003); the ARD was also significantly larger among adults with antisocial personality disorder (2.16% [95% CI, 0.61% to 3.71%] vs 0.07% [95% CI, −0.09% to 0.23%]; interaction P = .01), a history of violent behavior (1.04% [95% CI, 0.35% to 1.73%] vs 0.00% [95% CI, −0.12% to 0.12%]; interaction P = .003), or a history of anxiety (1.43% [95% CI, 0.47% to 2.39%] vs 0.18% [95% CI, 0.04% to 0.32%]; interaction P = .01) or depressive (0.99% [95% CI, −0.09% to 2.07%] vs −0.08% [95% CI, −0.20% to 0.04%]; interaction P = .05) disorders than among adults without these conditions.

Conclusions and Relevance  A recent overall increase in suicide attempts among adults in the United States has disproportionately affected younger adults with less formal education and those with antisocial personality disorder, anxiety disorders, depressive disorders, and a history of violence.

Preventing suicide is a leading public health¹ and research² priority. However, despite policy and clinical initiatives aimed at reducing suicide, the rate of suicide in the United States increased by approximately 2% per year from 2006 to 2014.³ One recognized approach to preventing suicide involves improving the identification and treatment of individuals at high risk, including those who plan or attempt suicide.⁴

Suicide attempts are the most powerful known risk factor for completed suicide.^5-7 During the first year after a suicide attempt, the risk for completed suicide varies from 0.8% to 3.0% for men and from 0.3% to 1.9% for women.^6,8-10 In a Swedish study,⁸ the rate of suicide among individuals in the year after a suicide attempt was nearly 100-fold higher than the corresponding suicide rate among age- and sex-matched community control individuals. By 10 years, 5% to 10% of adults making serious suicide attempts have completed suicide.^9-12 Because 15% to 25% of adults who die by suicide have received treatment for a suicide attempt within the past year,^11,13,14 a substantial proportion of suicide deaths are potentially subject to prior intervention that could be identified with a suicide attempt.

Suicide attempts are important clinical events. They are a major source of distress, morbidity, and economic burden. Most adults who make suicide attempts have anxiety or mood disorders and many have substance use disorders.¹⁵ In addition, approximately 18% of individuals who attempt suicide make a second attempt during the following year.¹⁶ In 2013, the total annual estimated economic burden of suicide attempts in the United States exceeded $8 billion.¹⁷

Population-based surveillance of suicide attempts could help to assess progress in efforts to reduce suicidal behavior. Several prior reports^18-21 have characterized trends in deliberate self-harm events among individuals presenting for emergency medical or mental health care. A limitation of these reports is that they provide no information about self-harm events that do not result in use of health care services. As a result, relatively little is known about the underlying epidemiology of suicide attempts and how it may have changed in recent years. According to the National Longitudinal Alcohol Epidemiologic Survey and National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), the lifetime prevalence of suicide attempts among US adults 18 years and older remained unchanged from 1991 through 1992 to 2001 through 2002 at 2.4%.¹⁸ However, an analysis of the National Comorbidity Surveys¹⁵ revealed that the proportion of adults in the United States who made a suicide attempt in the past year was 0.4% in 1990 through 1992 and 0.6% in 2001 through 2003. More recently, the rate of past year suicide attempts was reported as 0.5% among adults according to the 2008-2009 National Survey on Drug Use and Health.²²

Despite an increase from 2004 to 2014 in the US annual suicide rate from 11.0 to 13.0 per 100 000 population,^3,23 whether a corresponding coincident increase in suicide attempts has occurred remains unknown. If an increase has occurred, a characterization of which groups are at high and increasing risk would help focus prevention and early intervention initiatives. Therefore, the present report examines trends in recent suicide attempts among nationally representative general population samples collected from 2004 to 2005 and from 2012 to 2013. Because mental disorders²⁴ and socioeconomic disadvantage²⁵ have been hypothesized to contribute to the risk for suicide attempts, we sought to identify whether recent trends in suicide attempt risk have differentially affected subgroups with common mental disorders that are often a focus of clinical efforts to reduce suicide risk. Because an economic downturn occurred during the period under study, we also assessed whether adults with markers of socioeconomic disadvantage, including lower levels of educational attainment and lower family income, experienced a disproportionate increase in suicide attempt risk during this period.

The wave 2 NESARC (2004-2005) and NESARC-III (2012-2013) were separate nationally representative face-to-face interview surveys of 34 653 and 36 309 adults, respectively, residing in households and group quarters (eg, boarding and group homes) that were conducted by the National Institute on Alcoholism and Alcohol Abuse.^26,27 Multistage probability sampling was used to randomly select respondents. First, primary sampling units, which consisted of individual counties or groups of contiguous counties, were selected. Next, secondary sampling units, which were groups of census-defined blocks, were selected. In the third stage, households in the sampled secondary sampling units were selected. This sample involved random selection of eligible adults in sampled households. The analytic sample was restricted to all persons 21 years and older. The 69 341 adult study participants included 34 629 in the 2004-2005 cohort and 34 712 in the 2012-2013 cohort. The institutional review boards of the National Institutes of Health and Westat approved the study protocols. All participants provided electronic informed consent.

The overall survey response rate for the wave 2 NESARC was 70.2%.²⁷ For NESARC-III, the household screener response rate was 72.0% with a person-level response rate of 84.0% to yield an overall response of 60.1%, comparable to rates for other current US surveys.^28,29 The samples were weighted to adjust for nonresponse at the household and person levels, selection of 1 person per household, and oversampling of young adults and Hispanic and African American individuals. After weighting, the data were adjusted to be representative of the US population for variables that included region, age, sex, and race/ethnicity based on the Decennial Census and American Community Survey.³⁰

Sociodemographic measures included age, sex, race/ethnicity, marital status, educational attainment, family income, and current employment by self-report. The Alcohol Use Disorder and Associated Disabilities Interview Schedule–DSM-IV version (AUDADIS-IV) was used in NESARC,³¹ and the AUDADIS DSM-5 version (AUDADIS-5) was used in NESARC-III.³² Past-year substance use disorders (alcohol use disorders and drug use disorders, excluding nicotine dependence), past-year anxiety disorders (panic disorder, generalized anxiety disorder, and social phobia), past-year depressive disorders (major depressive disorders and dysthymic disorder), and lifetime personality disorders (borderline, antisocial, and schizotypal disorders) were assessed by structured diagnostic interviews. Test-retest reliability of AUDADIS-IV is good to excellent for substance use disorders (κ = 0.51-0.74) and fair to good for other psychiatric disorders (κ = 0.40-0.67),^33-36 whereas reliability of the AUDADIS-5 is good to excellent for substance use disorders (κ = 0.50-0.85) and fair to good for other psychiatric disorders (κ = 0.35-0.54).³⁷

A series of questions were also asked of respondents to evaluate whether they had ever engaged in violence, including starting a lot of fights, forcing a person to have sex against their will, swapping blows with a partner, using a weapon in a fight, hitting a person so hard that they required medical care, physically hurting another person on purpose, or robbing or mugging an individual.³⁸ To evaluate suicide attempts, respondents were first asked if they had ever attempted suicide (“In your entire life, did you ever attempt suicide?”). Those who responded affirmatively were asked their age at the first and most recent times that they attempted suicide. Individuals who indicated that their most recent attempt was within 3 years of their current age were defined to have made a recent suicide attempt. A history of suicide attempts was defined as reporting that the first suicide attempt occurred more than 3 years before their current age.

Proportions of 2012-2013 respondents with recent suicide attempts were computed overall and stratified by demographic and clinical subgroups. Because suicide attempt risk varies by age, sex, and race/ethnicity,¹⁵ multivariable analyses were controlled for these respondent characteristics. Because sociodemographic characteristics (educational attainment, marital status, employment, and family income) were conceptualized as being potentially in the causal pathway of trends in suicide attempts, the multivariable analyses were not controlled for these variables.

We used χ² tests to evaluate group differences in demographic and clinical characteristics of the 2004-2005 and 2012-2013 respondents with recent suicide attempts. Proportions of individuals with a suicide attempt within the past 3 years were then compared between the 2 surveys. Risk differences adjusted for age, sex, and race/ethnicity (ARDs) assessed associations between the survey periods (2004-2005 vs 2012-2013) and the risk for a recent suicide attempt. Adjusted risk differences were obtained from SUDAAN (version 11.0; RTI International) software using the predicted marginal approach that back transforms the estimates from the logistic regression to the probability scale.³⁹ The independent variable of interest was the survey period effect, with the 2004-2005 survey as reference. Separate adjusted regression models using the average marginal prediction approach³⁹ tested whether the ARDs significantly varied across different levels of each stratification variable (additive interactions).⁴⁰ All statistical analyses were performed with SAS (SAS Institute; version 9.4) or SUDAAN (version 11.0; RTI International) software to accommodate the complex sample design and weighting of observations.

Among the total sample of 69 341 study participants, 42.8% were men, 57.2% were women, and the mean [SD] age was 48.1 [17.2] years. In the 2012-2013 survey, women (0.92%) were more likely than men (0.64%) to have made a recent suicide attempt (Table 1). In adjusted analyses, recent suicide attempts were also significantly correlated with younger adults (adjusted odds ratio [AOR], 12.65; 95% CI, 6.91-23.18); being widowed, separated, or divorced rather than married or cohabiting (AOR, 4.09; 95% CI, 2.68-6.24); lower educational attainment (AOR, 4.05; 95% CI, 2.45-6.70); current unemployment (AOR, 3.37; 95% CI, 2.50-4.55); and a lower level of family income (AOR, 5.71; 95% CI, 3.43-9.50). Each of the mental disorders, especially borderline (AOR, 13.55; 95% CI, 10.29-17.85), schizotypal (AOR, 7.12; 95% CI, 5.44-9.33), and antisocial personality disorders (AOR, 6.67; 4.45-10.02), and a prior suicide attempt (AOR, 23.54; 95% CI, 16.46-33.67) were strongly associated with the risk for a recent suicide attempt.

Adults with recent suicide attempts in both surveys were predominantly female (60.17% and 60.94%), white (67.89% and 68.92%), and not currently employed (58.30% and 59.37%). Mental disorders were common. Approximately one-half of adults with recent suicide attempts reported having made a prior suicide attempt. In both surveys, nearly two-thirds of those with recent suicide attempts had borderline personality disorder. Compared with adults from the 2004-2005 survey who had recently attempted suicide, those in the 2012-2013 survey were younger (21 to 34 years of age, 49.98% vs 41.51%) and more likely have a depressive disorder (53.93% vs 25.52%), antisocial personality disorder (22.90% vs 13.13%), and a history of violent behavior (55.05% vs 43.52%). In a post hoc analysis, the proportion of respondents aged 35 to 49 years with suicide attempts was significantly larger in the 2004-2005 survey (43.24%) than in 2012-2013 survey (30.40%; P = .02). In relation to their 2004-2005 counterparts, the adults with suicide attempts in the 2012-2013 survey were also significantly less likely to have an anxiety (45.36% vs 60.45%) or substance use disorder (49.15% vs 61.28%) (Table 2).

During the study period, the percentage of US adults who reported making a recent suicide attempt increased from 0.62% in 2004-2005 to 0.79% in 2012-2013 (ARD, 0.17%; 95% CI, 0.01%-0.33%; P = .04) (Table 3). In adjusted trends analyses, significant risk differences in recent suicide attempts were observed among adults aged 21 to 34 years (ARD, 0.48%; 95% CI, 0.09%-0.87%; P = .02), non-Hispanic white (ARD, 0.24%; 95% CI, 0.04%-0.44%; P = .02) and black (ARD, 0.28%; 95% CI, 0.01%-0.55%; P = .04) individuals, and adults with no more than a high school education (ARD, 0.49%; 95% CI, 0.18%-0.80%; P < .002).

We also tested whether the change across surveys in the percentages of adults who reported recent suicide attempts differed across strata (adjusted additive interaction P values). As an example, we considered whether the ARD for men (0.13%; 95% CI, −0.07% to 0.33%) was significantly different from that for women (0.21%; 95% CI, 95% CI, −0.02% to 0.44%) (Table 3). In these analyses, the ARD in suicide attempts was significantly larger for adults aged 21 to 34 years (0.48%; 95% CI, 0.09% to 0.87%) than for 65 years or older (0.06%; 95% CI, −0.02% to 0.14%). The increase in risk was also significantly larger for adults with no more than a high school education (0.49%; 95% CI, 0.18% to 0.80%) than for those who had graduated from college (0.03%; 95% CI, −0.17% to 0.23%).

After controlling for potentially confounding demographic characteristics, we found significant increases in recent suicide attempts among adults with a history of violent behavior, antisocial personality disorder, substance use disorders, depression disorders, and anxiety disorders and among adults without anxiety disorders or substance use disorders (Table 4). In adjusted models, the increase in suicide attempt risk was significantly greater among adults with anxiety disorders (1.43% [95% CI, 0.47% to 2.39%] vs 0.18% [95% CI, 0.04% to 0.32%]; interaction P = .01), depressive disorders (0.99% [95% CI, −0.09% to 2.07%] vs −0.08% [95% CI, −0.20% to 0.04%]; interaction P = .05), antisocial personality disorder (2.16% [95% CI, 0.61% to 3.71%] vs 0.07% [95% CI, −0.09% to 0.23%]; interaction P = .01), and a history of violent behavior (1.04% [95% CI, 0.35% to 1.73%] vs 0.00% [95% CI, −0.12% to 0.12%]; interaction P = .003) than among adults without these conditions (Table 4).

Between the 2004-2005 and 2012-2013 surveys, recent suicide attempts became increasingly prevalent in the United States. The increase was particularly evident among young adults and those with no more than a high school education. The increase was also larger among individuals with antisocial personality disorder, a history of violent behavior, anxiety disorders, and depressive disorders than among those without these conditions. In the 2012-2013 survey, the highest-risk group consisted of adults with prior suicide attempts. Other high-risk groups included persons with borderline, schizotypal, or antisocial personality disorders and those with anxiety and depressive disorders. These findings highlight an increasing prevalence of suicide attempts and underscore the prominent role of mental disorders, including personality disorders, in risks for suicide attempts at the population level.

The upward trend in suicide attempts coincided with a national increase in suicide, although the 2 trends varied across demographic groups. For example, the risk difference in suicide attempts was greatest among adults aged 21 to 34 years, whereas the risk differences in suicide during this period were largest among adults aged 45 to 64 years.³ Although demographic differences in the risk profiles for suicide attempts and completion exist, including age and sex, several clinical risk factors are similar, including depression, anxiety, and substance use disorders^41,42; genetic risk factors may also be similar.⁴³ Population-based suicide attempt data complement traditional suicide mortality as a measure of the national population burden of self-injurious behavior.

The risk for suicide attempts was elevated among adults with high levels of economic insecurity, including those who were unemployed and had low family income and low educational attainment.^44,45 During the period when the NESARC-III survey data were collected and the 3 prior years, the monthly US unemployment rate (7.7%-10.1%) was considerably higher than the unemployment rate period during and preceding collection of the wave 2 NESARC survey (4.9%-6.3%).⁴⁶ Young adults and those with less formal education, 2 groups who experienced disproportionately large increases in suicide attempt risk during this period, may have been particularly vulnerable to economic stress and psychological distress associated with deterioration in the US economy.

Prior studies examining associations between economic factors, most commonly unemployment, and suicide^47,48 and suicidal behavior^49,50 have yielded mixed results across countries and periods. Contextual factors, such as the generosity of safety net programs and personal savings rates, likely account for much of this variation. In the present study, adults with no more than a high school educational level experienced a significantly larger increase in suicide attempt risk than did adults who had graduated from college. This pattern suggests that these socioeconomically disadvantaged individuals have borne a disproportionate share of risk associated with the recent increase in suicide attempts. However, because trends in suicide attempt risk did not significantly vary across family income level or current employment status, the recent recession did not seem to influence suicide attempt risk in a predictable manner.

Consistent with prior research on emergency department–treated deliberate self-harm events,^18-21 recent suicide attempts in this nationally representative sample of community-dwelling adults were more common among women than men and decreased with age. During the study period, the risk for attempting suicide increased for young adults but did not significantly increase for middle-aged or older adults. Together with a recently reported national increase in the prevalence of major depressive episodes among young adults,⁵¹ the increase in risk for suicide attempts among young adults signals the importance of focusing on early detection of mental health risk factors of suicidal behavior and treatment initiatives in this age group.

In both surveys, nearly two-thirds of adults with recent suicide attempts had borderline personality disorder. This high proportion may in part reflect the broad spectrum of self-harm behaviors captured by the suicide attempt survey item. Adults with borderline personality disorder have been previously found to be at increased risk for completed suicide.⁵² Because borderline personality disorder is characterized by “recurrent suicidal behavior, gestures, or threats,”^53(p663) the strong correlation with suicidal behavior is not surprising. A high prevalence of suicide attempts combined with a tendency of some front-line clinicians to hold negative views of borderline personality disorder⁵⁴ underscores the importance of developing clinician training programs to help improve the management of deliberate self-harm among patients with this condition.⁵⁵ One encouraging finding is that although most adults in the 2012-2013 survey who had recent suicide attempts had borderline personality disorder, the risk of attempted suicide among adults with borderline personality disorder significantly decreased during the study period. This trend may reflect increasing access to more effective interventions for impulsivity in borderline personality disorder. A survey of US psychiatric residency programs revealed that 40.8% of programs currently offer training in dialectical behavior therapy for borderline personality disorder.⁵⁶

A substantial and increasing proportion of adults who attempted suicide met criteria for antisocial personality disorder. Although individuals with antisocial personality disorder are often perceived as having a high risk for violent behavior toward others,⁵⁷ they also commonly have a history of suicide attempts.⁵⁸ In prospective research, adults with antisocial personality disorder and other externalizing psychopathologic features have been reported to be at increased risk for attempting suicide.⁵⁹ Antisocial behaviors may also be associated with increased risk for suicide.⁶⁰ Although suicidal behaviors are typically considered in relation to depression and other internalizing disorders, associations between antisocial personality disorder and attempted suicide challenge this stereotype. Future clinical research is needed to clarify the social context, triggers, and motivation for suicidal behavior in this population.

In interpreting these findings, several limitations apply. First, the NESARC surveys rely on retrospective self-reports. Responses may be affected by inaccuracies in the recall of the timing of events or the intention of self-harm events. However, we have no reason to believe that recall inaccuracies differentially affected the 2 surveys or that memory of suicide attempts is easily perturbed by recall bias. Second, suicide attempts were assessed with a single survey item that likely captured a wide range of behaviors, including interrupted, aborted, and potentially lethal and nonlethal attempts. Third, we have no means of assessing the influence of changes in the effectiveness of life-saving emergency management of suicide attempts. Fourth, the NESARC does not survey homeless or incarcerated adults, who have relatively high rates of suicidal behavior,^61,62 nor does it include an assessment of schizophrenia. Fifth, minor modifications between DSM-IV and DSM-5 criteria may have biased results of trends in suicide attempts among adults meeting criteria for the various mental disorders,⁶³ although these modifications do not influence the overall trends in suicide attempts or associations with sociodemographic characteristics. Sixth, some important characteristics, such as residence in a rural or urban location,⁶⁴ were not available. Seventh, to increase sample size, a 3-year rather than a 1-year period was used to define recent suicide attempts; this would be expected to attenuate associations with past-year mental disorders. Finally, the surveys did not collect data from individuals who died of suicide. This lack may have led to an underestimation of suicide attempts in each survey.⁶⁵

From the 2004-2005 to the 2012-2013 surveys, a national increase in recent suicide attempts occurred. Because attempted suicide is the greatest known risk factor for completed suicide,^6,10 reducing suicide attempts is an important public health and clinical goal. The pattern of suicide attempts supports a clinical and public health focus on younger, socioeconomically disadvantaged adults, especially those with a history of suicide attempts and common personality, mood, and anxiety disorders.

Corresponding Author: Mark Olfson, MD, MPH, The New York State Psychiatric Institute, Columbia University, 1051 Riverside Dr, New York, NY 10032 (mo49@cumc.columbia.edu).

Accepted for Publication: July 4, 2017.

Published Online: September 13, 2017. doi:10.1001/jamapsychiatry.2017.2582

Author Contributions: Ms Liu had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Olfson, Blanco, Wall.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Olfson, Wall, Liu.

Critical revision of the manuscript for important intellectual content: Olfson, Blanco, Wall, Saha, Pickering, Grant.

Statistical analysis: Wall, Liu, Saha, Pickering.

Obtained funding: Grant.

Administrative, technical, or material support: Pickering, Grant.

Study supervision: Olfson, Wall.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by grants DA019606 and MH 107452 from the National Institutes of Health (NIH) and The New York State Psychiatric Institute (Drs Olfson and Wall). The National Epidemiologic Survey on Alcohol and Related Conditions was funded, in part, by the Intramural Program, National Institute on Alcohol Abuse and Alcoholism, NIH (Dr Grant).

Role of the Funder/Sponsor: The sponsors had no additional role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The opinions expressed in this article are the author's own and do not reflect the view of the NIH, the Department of Health and Human Services, or the US government.