a) Age

As shown in Table 10, suicide rates vary with age, gender, and race or ethnicity. Annual rates in the general U.S. population rise sharply in adolescence and young adulthood, plateau through midlife, then rise again in individuals over age 65. The increased rates of suicide in youths are even more dramatic in some ethnic and racial subgroups of the population. For example, the suicide rate among American Indian males between ages 15 and 34 years averaged about 36 per 100,000 during the period from 1979 to 1993, whereas Alaska Native males between ages 14 and 19 years had an even more dramatic rate, at 120 per 100,000 (661). Black male youths, who were historically at low risk for suicide, now have a suicide rate comparable to their white peers. Although the suicide rate in adolescents, like the overall U.S. suicide rate, has dropped in the past decade, the relative suicide risk of youths remains high, and this has been attributed to increases in alcohol and substance abuse (662), breakdown in extended family and intergenerational support, and increased availability of firearms, especially for young African American males (663).

Individuals over age 65 are disproportionately represented among those who die by suicide. Compared with suicide rates in men ages 55 to 64 years, suicide rates in men over age 85 are two- to threefold higher for all races except African Americans. For elderly women, suicide rates are relatively unchanged with increasing age, with the exception of Asian women over age 85, whose suicide rate increases threefold from middle age.

Overall suicide rates among those over age 65 have decreased substantially over the course of the last century, with a further decrease over the past decade. Although the reasons for the decline are unknown, a variety of mechanisms have been postulated, including improved access to social and health care resources by older adults with the implementation of Social Security and Medicare legislation and the more widespread use of safe and effective antidepressant medications (664). The incidence of suicide among elderly persons may increase again, however, as the large, post–World War II baby boom generation continues to age. Relative to age groups born in earlier or later periods, baby boomers have been distinguished by suicide rates that have been comparatively higher at all ages (665). Of additional concern is the fact that elders are the fastest growing segment of the U.S. population. Thus, as large numbers of this high-risk cohort enter the phase of life associated with greatest risk, the absolute number of suicides among older adults may increase dramatically (666).

Suicidal ideation and suicide attempts are more frequent in younger age groups than in later life (14). Kuo et al. (29), using prospectively gathered data from the Epidemiologic Catchment Area (ECA) survey, found a progressive decrease in the annual incidences of suicidal ideation and suicide attempts with increasing age. Compared with the rate in individuals over age 65, the rate of suicide attempts was 10-fold greater in those ages 18 to 29 years, at approximately 310 per 100,000 person-years. The rate of suicidal ideation in individuals ages 18 to 29 was approximately 630 per 100,000 person-years, a rate that was sixfold greater than that in those over age 65. Duberstein et al. (13), in a study of adults age 50 years and older, also found that people are less likely to report suicidal ideation as they age.

In other studies, estimates of the prevalence of suicidal ideation in older adults have varied with the population sampled and the site, time frame, and study methods. Lish and colleagues (667) found that 7.3% of an older sample in Department of Veterans Affairs (VA) primary care practices had thoughts of suicide, and elders with a history of mental health treatment were at far greater risk. Callahan and colleagues (668) used a far more stringent definition of suicidal ideation, limited ascertainment to within the past week, and required the ideation to include a specific suicide plan. They found that 0.7%–1.2% of elders in primary care had suicidal ideation, all of whom had a simultaneous mood disorder. Skoog et al. (669), in a survey of nondemented Swedes age 85 years and older, inquired about the presence of both active and passive suicidal ideation in the month preceding the interview. They found that 16% of the subjects had thoughts of suicide. Again, the rate was higher in subjects with mental disorder, in those taking anxiolytic and neuroleptic agents, and in those with significant physical illness. Among community-dwelling Floridians 60 years of age and older, less than 6% reported ever having had suicidal thoughts in a study by Schwab et al. (670), while in the Berlin Aging Study (671) 21% of subjects over age 70 reported having had suicidal ideation. Again, psychiatric illness was present in virtually all subjects, suggesting a need for careful screening for psychiatric disorder in elders with suicidal ideation.

b) Gender

In the United States, epidemiologic data show that suicide is more frequent in men than in women. For example, data from the National Center for Health Statistics for the year 2000 showed an age-adjusted suicide rate for males that was approximately 4.5-fold that for females (18.08 per 100,000 and 4.03 per 100,000, respectively) (11). This differential is comparable to the male-to-female ratio for suicide found in the National Longitudinal Mortality Study for the years 1979 to 1989 (672). Within the U.S. population, males are disproportionately represented among deaths by suicide in all racial and ethnic groups, with rates that range from more than 5.5-fold greater than that for females among African Americans and Hispanics to threefold greater than that for females among Asian/Pacific Islanders. This is not the case in other parts of the world, however. For example, in China the suicide rate for women is 25% higher than that for men (18).

The male-to-female predominance in suicide in the United States persists across the lifespan. Adolescent and young adult males are about 5.5 times more likely to die from suicide than females, whereas in midlife the male-to-female ratio is approximately 3.5 to 1. After about age 65, however, there is a steadily widening male-to-female ratio of suicide rates in all groups except Asians, with differences of more than 10-fold after age 80.

Differences in suicide risk with gender may be explained in part by factors that contribute to risk in general but that are present to differing degrees in men and in women. For example, men are less likely than women to seek help, admit the severity of their symptoms, or accept treatment, increasing their likelihood of suicide. In contrast, women tend to be less impulsive, have more social support, and have lower rates of comorbid alcohol and substance use disorders, all of which may have a protective effect (21). Among African American women, the potential protective factors of religion and extended kin networks have been suggested as possible explanations for this group's very low rate of suicide (22).

Despite their lower rate of suicide, women have higher rates of depressive illness than men (23, 673). Furthermore, in a 10-year follow-up study using data from the National Longitudinal Mortality Study, unemployment was associated with a greater and longer-lasting effect on the suicide rate of women compared to men (24). Compared to men, women also have an increased likelihood of having been physically or sexually abused, which may also increase the risk for suicide (36). The relative lethality of the suicide methods chosen by women remains less than those chosen by men; however, the recent, more frequent use of firearms among women suggests that this distinction may be diminishing (11, 674).

Suicide rates have also been examined in pregnant women and during the postpartum period. Dannenberg et al. (675) reviewed New York City medical examiner records of 293 pregnant or recently pregnant women ages 15 to 44 years who died of injury during a 4-year period. Of these, 15 died by suicide, a rate that was not significantly different from the expected age- and race-specific rates in the general population. However, Marzuk et al. (676) analyzed autopsy data from female residents of New York City who were of childbearing age and found the standardized mortality rate for suicide during pregnancy to be one-third the expected rate. Appleby (677), using retrospective population data for England and Wales from 1973 to 1984, also noted decreased rates of suicide among pregnant women and among women during the first year after childbirth, with SMRs of 0.05 and 0.17, respectively. In contrast to decreased suicide rates for women in general during pregnancy and the puerperium, Appleby et al. (27) subsequently found an extremely high suicide rate among women who had been psychiatrically hospitalized during the postpartum period. In this study of 1,567 women admitted to Danish psychiatric hospitals within the first year after childbirth, the SMR for suicide within 1 year was more than 70 times the expected rate. Although risk was greatest within the first month postpartum, it persisted throughout the initial year after childbirth. In addition, women who died by suicide after childbirth often used violent methods. Thus, although evidence is limited, women with severe postpartum psychiatric disturbances appear to be at significantly increased risk during the initial year after childbirth. Other groups with a particularly increased postpartum risk include teenagers and women of lower socioeconomic status (27, 28). For women as a group, however, a protective effect seems to be present during pregnancy and the postpartum period (25).

In terms of suicide attempts, women in the United States are reported to attempt suicide three times as often as men. This female predominance of suicide attempters varies with age, however, and in older adults the ratio of female-to-male suicide attempters approaches 1:1 (11). Similar trends are observed in the incidence of suicidal ideation. For example, Kuo et al. (29), using data from 3,481 prospectively followed individuals from the Baltimore ECA study, found that females ages 18 to 29 years had a higher incidence of suicidal ideation and suicide attempts than their male peers. However, this female-to-male predominance in suicidal ideation and suicide attempts was not observed for older age groups or for the sample as a whole.

As noted earlier, women are more likely to have experienced domestic violence or physical or sexual abuse, all of which have been associated with higher rates of suicidal ideation and suicide attempts (32–34). In a study of psychosocial outcomes in 1,991 same-sex twin pairs, Nelson et al. (35) found that childhood sexual abuse was three times more common in women and was associated with an increased risk of attempting suicide. Borderline personality disorder is also present more often in women (515) and is itself associated with increased rates of suicidal ideation, suicide, suicide attempts, and other self-injurious behaviors. In addition, borderline personality disorder is particularly common in women who have experienced childhood sexual abuse, physical abuse, or both (31). As a result, physical and sexual abuse and domestic violence should be given particular consideration in the assessment and treatment of women with suicidal ideation, suicide attempts, and other self-injurious behaviors.

c) Race, ethnicity, and culture

Race, ethnicity, and culture are all associated with variations in rates of suicide. In the United States for the year 2000, the overall age-adjusted rates of suicide were highest in Native Americans and non-Hispanic whites, at 13.6 and 12.1 per 100,000, respectively (11). In contrast, the age-adjusted rate of suicide in Hispanics was substantially less, at 6.13 per 100,000, and was similar to the rates for non-Hispanic African Americans and Asian/Pacific Islanders, at 5.8 and 6.0 per 100,000, respectively.

For immigrant groups, suicide rates in general tend to mirror rates in the countries of origin, with trends converging toward the host country over time (40, 41). In a large epidemiological study, Singh and Siahpush (39) found that between 1979 and 1989, foreign-born men in the United States were 52% less likely to die by suicide than native-born men, but the difference narrowed in the older age cohorts. Data for immigrant women were not statistically significant because of the small number of deaths.

In the United States, racial and ethnic differences are also seen in the rates of suicide across the lifespan (Table 10). Among European-American non-Hispanic whites, Hispanics, and Asian/Pacific Islanders, the highest suicide rates occur during the senior years, in those over age 65. In contrast, among Native Americans and African Americans, the highest suicide rates occur during adolescence and young adulthood. For example, in Native American and African American males ages 15 to 24, suicide rates in the year 2000 were 36.81 and 14.66 per 100,000, respectively. Young African American men have been described as being caught in a cycle of drug abuse, criminal activity, and self-devaluation and may view an early death as inevitable or as an alternative to the wearying struggle that life has become (678). Additional risk factors for suicide in young African American males include substance abuse (662, 679), presence of a firearm (663), and in particular the combination of cocaine abuse and the presence of a firearm (679). Suicidal ideation and suicide attempts are also common in urban African American young adults, with 6-month prevalences of 1.9% and 0.4%, respectively (680).

In contrast to young African American males, African American women have a very low rate of suicide. Gibbs (22) attributes this low rate to the protective factors of religion, including the role of religion in the civil rights movement, women's central involvement in the church, and strong values for endurance in the face of adversity. Women-dominated kinship networks are also believed to be protective, providing flexible roles, resource sharing, and social support (681).

Although black women are less likely to die from suicide than white women, they attempt suicide and express negative emotional states such as hopelessness and depression just as frequently. In addition, both black men and black women are less likely than their white counterparts to pursue professional counseling in the face of depression or other mental illness. Instead, African Americans are more likely to view depression as a "personal weakness" that can be successfully treated with prayer and faith alone some or almost all of the time (682). When depression is discussed, it may be described in different terms such as having "the blues" or "the aching misery" or "being down" (678). Consequently, sensitivity to language and beliefs about illness are important in recognizing depression and other risk factors for suicide among African Americans.

Among Native Americans (American Indians and Alaska Natives), suicide also is predominately an epidemic of the young and is the second leading cause of death for Native Americans between ages 15 and 24 years. As with other racial and ethnic groups, Native American and Alaska Natives are a very heterogeneous population, with different tribal identities, varying degrees of urbanization, different levels of tribal organization, and diverse approaches to historical and cultural integration. For example, in a study of three groups of Native Americans in New Mexico, the Apache had the highest suicide rate (43.3 per 100,000) and the highest degree of acculturation but also had the lowest degree of social integration and generally viewed religion as unimportant (683). In contrast, the Navajo had the lowest suicide rate (12.0 per 100,000) and the lowest level of acculturation but had moderate social integration and were organized into bands with a strong matrilineal clan influence. In the third group, the Pueblo, the subgroup with the most acculturation, had a higher suicide rate than the most traditional subgroup, again suggesting an effect of acculturation on suicide risk. Acculturation has also been proposed as a contributor to the extremely high suicide rate in Alaska Native youths, which in one study approached 120 per 100,000 (661). Theories to explain these high rates tend to rely on family disintegration, social disruption, and alcohol use (684), as well as rapid social and cultural changes associated with intensive energy development projects in the Arctic and the resulting stress of acculturation. In contrast, in Hawaiian youths, the relationship between acculturation and suicidal behavior is less clear, with increased numbers of suicide attempts in those with stronger Hawaiian cultural affiliation (685).

Research on suicide among Hispanics in the United States is limited and rarely differentiates among different Hispanic groups. In addition, many individuals of Hispanic origin are undocumented workers who are not represented in census data or epidemiological studies. Large-scale grouping of diverse ethnic groups also obscures intracultural variations in important social and economic categories. For example, Cuban American women and Mexican Americans and Puerto Ricans of both genders were reported to have lower than expected suicide rates, relative to 1-year prevalence of major depression, than were whites, blacks, and Cuban American males (23). In terms of suicidal ideation, higher levels have been reported in Central American immigrants experiencing heightened levels of acculturative stress (43). In addition, lifetime age- and gender-adjusted rates of suicidal ideation were significantly lower for Mexican Americans born in Mexico (4.5%) than for Mexican Americans born in the United States (13%) or for non-Latino whites (19.2%) (686). Similarly, rates of suicide attempt were lower among Mexican Americans born in Mexico (1.6%) and higher among both Mexican Americans born in the United States (4.8%) and non-Latino whites (4.4%). The rate of suicide attempt is also elevated among Hispanic youths, who had higher numbers of reported suicide attempts compared to non-Hispanic youths in a nationwide survey of high school students (687).

The suicide rate for Asians overall is the lowest of all of the major American ethnic groups, but Asian Americans themselves have diverse ethnic backgrounds, languages, and cultures. Some groups, such as the Japanese, have been in the United States for generations. Others, such as the Chinese, include both recent immigrants and descendants of 19th-century immigrants, whereas the Vietnamese have arrived in large numbers only since the 1960s. These individuals bring with them attitudes toward coping and suicide from their home countries, which can influence the circumstances of suicidal behavior (688). In Japan, for example, suicide is permissible or even appropriate in particular contexts, and ritual suicide has been an honorable solution to certain social dilemmas. For example, the disgrace of bankruptcy in Japan can shame the family for generations, making suicide a preferable way to resolve debt. When it is culturally important for a man to be physically healthy and able to support his family, suicide may be viewed as an option if a serious physical illness impairs his ability to function. For example, in Hawaii, 20.5% of suicides by Japanese American men occurred in individuals with health problems, in contrast to only 11.8% of suicides by Caucasian men and 3.0% by Hawaiian men (42). In addition, for individuals who come from a culture in which mental illness is highly stigmatized, receipt of a psychiatric diagnosis may increase the risk for suicide. Although Chinese societies have not generally codified suicide as socially acceptable, more recent suicide rates in China are quite high, particularly in women and in rural settings, where use of agricultural poisons is a common suicide method (18).

In the United States, acculturation and acculturative stress may be a contributor to suicide risk among Asian Americans. For most Asian Americans, the family unit is central to identity. Children are socialized into awareness that their individual actions reflect upon the entire family, including extended family members (689). While this feature may impede a family's willingness to seek treatment for a troubled relative, the strong sense of family as a support and source of obligation protects against suicide as well. At the same time, family conflict as a reason for suicide is more common in Eastern societies (42). For example, if a young woman from a traditional society experiences conflicts with her in-laws that have no apparent solution, the woman may be more likely to view suicide as an option than would someone from a different family system in which close family relationships are not as imperative. Transition to the individualistic, communication-oriented U.S. society is a major and stressful change for many families (44). The group most at risk appears to be traditionalists who live in tight-knit groups resistant to acculturative processes. They appear to function relatively well until their elderly years, when the culture clash between the values of the larger society and the Confucian tradition of strong family identity results in alienation of elders and contributes to suicide in the style of the old country (44). For example, a major factor in the high suicide rate of elderly Asian/Pacific Islander women was reported to be the failure of younger family members to provide support for their elderly parents, especially widowed mothers (690). Such deaths occurred predominantly by hanging, which was traditionally seen as an act of revenge, since someone who died by hanging was believed to return to haunt the living as a ghost (690).

In summary, race, ethnicity, and culture may all influence population-based rates of suicide and suicide attempts. Of equal importance to the clinician, however, each of these factors may modify suicide risk within the individual. Views of death and cultural beliefs regarding suicide can vary widely, even among members of apparently homogeneous racial, ethnic, or cultural groups. Thus, as part of the assessment and treatment planning process, it can be helpful for the psychiatrist to explore the patient's beliefs about death and suicide and the role of cultural and family dynamics in these beliefs.

d) Marital status

Marital status has been correlated with variations in suicide mortality in a number of studies. Smith et al. (691) used data from the U.S. National Center for Health Statistics for the years 1979 to 1981 to calculate age-adjusted suicide rates for each marital status. Regardless of age or racial group, the suicide rate was consistently lowest in married individuals. An intermediate rate was seen in those who had never been married, with a relative risk that was about twice that in married individuals. The highest suicide rate was found for divorced or widowed individuals, with a relative risk that was about threefold greater than that in married individuals. Whereas divorced women had a higher age-adjusted suicide rate than widowed women, the opposite was true among men, with a particularly striking rate of suicide in young widowed men.

Kposowa (672) applied Cox proportional hazards regression models to data from the 1979–1989 follow-up of the National Longitudinal Mortality Study and made adjustments for age, sex, race, education, family income, and region of residence to estimate the effect of marital status on suicide risk. Although in this sample being single or widowed had no significant effect on suicide risk, divorced and separated persons had suicide rates that were more than twice that of married persons. Stratification of the sample by sex showed that the effect of marital status on suicide rates occurred only among men.

Luoma and Pearson (46) also examined whether marital status is associated with variations in suicide rates. Suicide rates broken down by race, 5-year age groups, sex, and marital status were calculated by using data compiled from the U.S. National Center for Health Statistics Multiple-Cause-of-Death Files for the years 1991 to 1996. Widowed white and African American men under age 50 were found to have substantial elevations in suicide rates, with 17-fold and ninefold higher rates, respectively, compared with married men under age 50. At younger ages, for women as well as for men, being widowed was associated with a higher suicide rate, compared with being married.

Using data from the National Suicide Prevention Project in Finland, Heikkinen et al. (402) investigated age-related variations in marital status as well as other social factors in a sample of 1,067 individuals who died by suicide during a 1-year period and for whom relevant data were available. Compared with the general population, individuals who died by suicide were more commonly divorced, widowed, or never married. Among individuals under age 50 who died by suicide, more males than females had never been married. Among those over age 50, more women than men were widowed.

Other data from Finland obtained through the Finnish Population Register and cause-of-death files also suggest that the rate of suicide is elevated among widowed individuals (45). Among 95,647 persons who were widowed during 1972–1976 and followed up to the end of 1976, 7,635 deaths were observed, of which 144 were due to suicide. During the initial month of bereavement, men had a much greater increase in suicide mortality than women (17.2-fold versus 4.5-fold), but this disproportionate ratio primarily resulted from occurrences of homicide-suicide. In the remaining first year of bereavement, men had a 3.1-fold increase in suicide mortality and women a 2.2-fold increase, and rates remained higher than expected throughout the follow-up period.

Overall, these studies suggest that married individuals have a significantly lower rate of suicide than unmarried individuals. In addition, elevations in the suicide rate are especially striking for widowed men in general and young widowed men in particular. What remains unclear is whether this protective effect of marriage on the suicide rate relates to specific benefits of marriage, such as a greater likelihood of social integration. In contrast, the decrease in social integration and the psychological experience of loss with widowhood and with divorce may increase the tendency for suicide. The suicide rate among divorced individuals could also be higher because individuals who stay married have a greater likelihood of stable mental health at baseline. Other confounding factors, such as differences in substance use or socioeconomic status with marital status, could play additional roles that should be considered in the assessment process.

e) Sexual orientation

It remains unclear whether suicide rates in gay, lesbian, and bisexual individuals differ from the suicide rate among heterosexual individuals. One psychological autopsy study compared gay males to all other similarly aged males in the sample and did not find any characteristics that distinguished the two groups (692). However, research on suicide among gay, lesbian, and bisexual individuals is particularly complex because of many factors, including small sample sizes, difficulties in achieving random sampling, problems in obtaining baseline prevalences, and problems in reliability of postmortem reports of sexual orientation. In addition, individuals may choose not to disclose their sexual orientation to researchers or may engage in same-sex behavior but not identify themselves as gay or lesbian.

The risks for suicide attempts and suicidal ideation in gay, lesbian, and bisexual individuals have been assessed by using several approaches. Fergusson et al. (51), analyzing longitudinal data gathered on a New Zealand birth cohort, found that those who identified themselves as gay, lesbian, and bisexual or reported having a same-sex partner since the age of 16 had elevated rates of suicidal ideation (odds ratio=5.4) and suicide attempts (odds ratio=6.2). A study by Cochran and Mays (50) examined lifetime prevalences of suicide-related symptoms among men with same-gender partners and found that approximately one-half (53.2%) of the men reported experiencing at least one suicide-related symptom in their lifetime, with a suicide attempt reported by 19.3%. In contrast, in men with female partners only, 33.2% had at least one suicide-related symptom and 3.6% reported a suicide attempt. Corresponding figures for those with no sexual partners were 28.1% and 0.5%, respectively. Using the population-based Vietnam Era Twin Registry, Herrell et al. (52) identified a subsample of 103 middle-aged male twin pairs in which one of the twins from each pair reported having a male sexual partner after age 18 while the other did not. Suicide attempts were more common in the men with same-gender sexual orientation, with 15% reporting a suicide attempt, compared with only 4% of their twin brothers. In the Twin Registry sample as a whole, which included 16 twin pairs concordant for having a male sexual partner after age 18 and 6,434 twin pairs concordant for having no adult same-gender partners, the men with same-gender sexual orientation had more than a fourfold increase in suicidal ideation and more than a 6.5-fold increase in suicide attempts.

Gay, lesbian, and bisexual youths may be at particular risk for suicidal behaviors. Paul et al. (53), in a study of a large urban population–based telephone probability sample of gay men, found that 21% had made a suicide plan and 12% had attempted suicide. Of the latter, almost one-half had made multiple attempts, and most had made their first attempt before age 25. The importance of sexual orientation to suicidal behaviors in youths is also highlighted by the findings of a statewide population-based study of public high school students by Remafedi et al. (48). In this study, suicide attempts were reported by 28.1% of bisexual/homosexual males, 20.5% of bisexual/homosexual females, 14.5% of heterosexual females, and 4.2% of heterosexual males. For males, but not for females, a bisexual/homosexual orientation was associated with suicidal intent (odds ratio=3.61) and with suicide attempts (odds ratio=7.10).

Thus, although evidence is limited, there is clearly an elevated risk for suicide attempts among cohorts of gay, lesbian, and bisexual individuals that is particularly striking among youths. In addition to addressing risk factors such as psychiatric and substance use disorders in the assessment and treatment planning processes, it is also important for the clinician to address stresses that are unique to being gay, lesbian, or bisexual (e.g., disclosure of sexual orientation to friends and family, homophobia, harassment, and gender nonconformity). Since suicide attempts themselves increase the risk for later suicide, it is presumed that suicide rates may also be increased in gay, lesbian, and bisexual individuals. However, this hypothesis remains to be tested empirically.

f) Occupation

Occupational groups differ in a number of factors contributing to suicide risk. These factors include demographics (e.g., race, gender, socioeconomic class, and marital status), occupational stress, psychiatric morbidity, and occupationally associated opportunities for suicide. Although many studies have reported increased rates of suicide in specific occupational groups, most have not controlled for other suicide risk factors. In one study, however, that controlled for basic demographic correlates of suicide across 32 occupations (54), risk was found to be highest among dentists and physicians, compared with the rest of the working-age population, with multivariate logistic regression odds ratios of 5.43 and 2.31, respectively. The odds of suicide were also significantly higher in nurses (1.58 times the risk), social workers (1.52 times the risk), mathematicians and scientists (1.47 times the risk), and artists (1.30 times the risk). Rates of suicide among physicians have also been found to be elevated, compared with rates for other white male professionals, with white male physicians having a 70% greater proportionate mortality ratio for suicide (58). In well-designed epidemiological studies, police officers have generally not been found to be at higher risk for suicide than age- and sex-matched comparison subjects (54, 57).

Factors that may play a role in the increased suicide rates in specific professions may include occupational stresses, as is seen in helping professionals (54), or social isolation, as is seen in sheepherders, who had the highest suicide rate of 22 occupational groups studied in Washington State (56). Although data are inconsistent, additional work stress may occur with infrequent role sets such as female laborers or pilots (55, 57) or in nontraditional occupations (693). In some occupations, suicide rates may be influenced by greater access to lethal methods such as medications or chemicals, as in health care professionals, scientists, and agricultural workers (57).

Differential rates of psychiatric illness may be present in some occupations and may predate employment. Artists, for example, have higher rates of psychiatric morbidity and suicide than the general population. Highly educated people with depressive disorders also have a higher suicide rate. Among physicians, such individuals may tend to specialize in psychiatry (56).

In general, specific occupations do seem to be associated with an increased risk for suicide, but more research is needed to distinguish occupational from nonoccupational stressors (56) and to determine whether it is the occupation itself or associated factors such as psychiatric morbidity that affect suicide risk.

a) Mood disorders

Major depressive disorder and other depressive syndromes are the most commonly and most consistently identified axis I diagnoses in individuals who die by suicide (694, 695). For example, Robins et al. (60) found that among 134 persons who died by suicide, 98% were psychiatrically ill and most had depression or chronic alcoholism. Barraclough et al. (65), in a similar study, found that of 100 individuals who died by suicide, 93% were mentally ill and 85% had either depression or alcoholism. Henriksson et al. (59), using psychological autopsy methods to investigate current mental disorders among a random sample of 229 persons who died by suicide during a 1-year period in Finland, found that 93% of those persons had received at least one axis I diagnosis and that 59% had a depressive disorder.

In patients with bipolar disorder who die by suicide, the majority are experiencing either a depressive or mixed episode of illness (69, 72, 315). For example, Isometsa et al. (68) noted that among 31 patients with bipolar disorder identified in a group of 1,397 persons who died by suicide in Finland in a 12-month period, 79% died while in a major depressive episode and 11% while in a mixed state. In a study of more than 300 patients who were discontinued from lithium treatment, Baldessarini et al. (696) found that the majority of suicidal acts occurred either during a major depressive episode (73%) or during a dysphoric-mixed episode (16%).

In addition to being highly prevalent in individuals who die from suicide, mood disorders have long been associated with an increased risk for suicide. For example, in 1970, Guze and Robins (697) reviewed 17 studies that assessed the risk of suicide in individuals with primary affective disorders and calculated the frequency of suicide as a percentage of all deaths. High suicide rates were found, with the ultimate risk of suicide estimated to be about 15%, or approximately 30 times that seen in the general population. For major depression, review of the literature suggests that overall rates of suicide mortality range from 5% to 26% and are about twice as high for men as for women (694). However, these studies generally assessed severely ill patient populations and individuals early in the course of their illness, when suicide rates are known to be highest.

Several investigators have subsequently reexamined these estimates of lifetime suicide risk in individuals with mood disorders. For example, Inskip et al. (94), using cohort-based curve-fitting techniques and data from previous studies, estimated the lifetime risk for suicide in mood disorders to be 6%. In addition, Bostwick and Pankratz (77) used data from prior studies to calculate case fatality prevalences (the ratio of suicides to the total number of subjects) to determine suicide risks for three groups of patients with affective disorders—outpatients, inpatients, and suicidal inpatients. With this method, which provides a less biased estimate of risk, they found a gradation in suicide risk that varied with treatment setting as well as with hospitalization for suicidality. For example, in patients with mood disorders who were previously hospitalized for suicidality, the estimated lifetime prevalence of suicide was 8.6%, compared to a lifetime risk of 4% for those with a psychiatric hospitalization for any reason. For mixed inpatient/outpatient populations, the prevalence of suicide was 2.2%, whereas for the populations without affective illness, it was less than 0.5%. For individuals with major depressive disorder, Blair-West et al. (205) used age- and gender-stratified calculations to arrive at comparable estimates for lifetime suicide risk of 3.4%, with a lifetime risk for males more than six times than for females (6.8% versus 1.1%).

Harris and Barraclough (64), in their meta-analysis of suicide as an outcome in psychiatric illness, assessed relative suicide risk in mood disorder by calculating SMRs. Their analysis used data from published English-language studies that had mean or median follow-up periods of at least 2 years and that provided sufficient data to calculate ratios of observed to expected numbers of suicides. For patients with major depressive disorder, 23 studies that included a total of 351 suicides among more than 8,000 patients yielded an SMR of 20.35, or a 20-fold increase in risk. A key finding was that risk in patients with major depressive disorder was highest immediately after hospital discharge (698, 699). For patients with bipolar disorder, data from 15 studies including a total of 93 suicides among 3,700 subjects yielded an SMR for suicide of 15.05. Although patients with dysthymia also had an elevated SMR for suicide, of 12.12, the nine studies that contributed to this estimate were extremely heterogeneous in their findings and most had extremely small samples, which raises some question about the validity of this approximation.

Several studies have examined rates of suicide in longitudinal follow-up in individuals hospitalized for mood disorder. Hoyer et al. (75) used data from the Danish Psychiatric Case Register to determine SMRs for suicide among 54,103 patients (19,638 male and 34,465 female patients) who had an initial admission to a Danish psychiatric hospital between 1973 and 1993 and who received a mood disorder diagnosis. During the study period, 29% of the patients died, and of those, suicide occurred in 20%. Standardized mortalities for suicide were comparable for patients with ICD-8 diagnoses of unipolar major depression, psychotic reactive depression, and bipolar disorder, with SMRs of 19.33, 18.67, and 18.09, respectively. In contrast, the SMR for suicide in patients with neurotic depression was significantly less, at 10.51. In all diagnostic subgroups and regardless of age and gender, the risk of suicide was greatest during the first year after the initial admission, decreased over the subsequent 5 years, and then stabilized. Overall, the risk for suicide was comparable in men and women, except in patients with bipolar disorder, for whom the SMR for suicide was somewhat greater in women than in men (20.31 versus 18.09).

In a similarly designed study using data from a Swedish inpatient register, Osby et al. (73) obtained the date and cause of death for patients hospitalized between 1973 and 1995 with a diagnosis of bipolar disorder (N=15,386) or unipolar depressive disorder (N= 39,182). SMRs for suicide were found to be significantly increased in women and in patients with a unipolar depressive disorder diagnosis (15.0 for male bipolar disorder patients, 20.9 for male unipolar depressive disorder patients, 22.4 for female bipolar disorder patients, and 27.0 for female unipolar depressive disorder patients). Suicide mortality was more pronounced in younger individuals and with shorter intervals from the index hospitalization. Although SMRs decreased in all age groups with increasing time of follow-up, some suicide risk persisted even at long follow-up intervals.

Baxter and Appleby (188) used the Salford (U.K.) Psychiatric Case Register to identify 7,921 individuals who had received psychiatric or mental health care and determined their mortality rates (estimated as rate ratios) over a follow-up period of up to 18 years. Among individuals with affective disorders, there was a 12.2-fold elevation in observed suicide mortality in men, compared to expected mortality based on population rates. For women, the relative increase in suicide mortality was even greater, with a 16.3-fold elevation.

Angst et al. (74) followed a sample of 406 hospitalized patients with mood disorders (220 with bipolar disorder and 186 with unipolar depressive disorder) on a prospective basis for 22 years or more and found an overall standardized mortality rate for suicide of 18.04, comparable to the SMRs found in the Swedish and Danish longitudinal follow-up studies. Sixty-one percent of the sample had manifested psychotic symptoms at least once over their lifetime, suggesting that this was a particularly ill group of patients. The suicide rate was greatest near the age of illness onset; however, from ages 30 to 70 years, the rate was remarkably constant, suggesting a persistence of risk throughout the illness course. The suicide mortality in women was greater than that in men (SMR of 21.87 for women, compared to 13.49 for men), in part reflecting the greater rate of suicide for men in the general population. Patients with unipolar depressive disorder had a significantly higher rate of suicide than patients with bipolar I disorder or bipolar II disorder, with an SMR for suicide of 26.7, compared with 12.3 for bipolar disorder patients. The SMR for suicide did not differ significantly between bipolar I disorder patients and bipolar II disorder patients.

Some evidence suggests that in individuals with mood disorders, the rate of suicide may be increasing over time. For example, Harris and Barraclough (64) noted that the suicide risk for patients with major depression in cohorts treated before 1970 was increased by 17-fold in contrast to a 36-fold increase in risk for cohorts treated after 1970. In the study described earlier, Hoyer et al. (75) noted an increase in both the absolute and relative risks for suicide over the 20-year study time period, and they suggested that the increase may have been related to changes in the health care delivery system and the availability of psychiatric inpatient services. In addition, Baldessarini et al. (563) observed that the annualized rates of suicide and suicide attempts in patients with major affective disorders appear to have risen across the decades since 1970. This trend was sustained and statistically significant for both suicides and suicide attempts, as well as for treated and untreated samples considered separately. Although this apparent secular trend could reflect increased recruitment of more severely ill patients to more recent studies or increased reporting of suicidal behaviors, the percentage reduction of suicide risk with lithium treatment did not decline across the years, suggesting that the patient populations are in fact comparable and that the prevention of suicide in major affective disorders is becoming increasingly challenging (558). Furthermore, suicide attempts that do occur in individuals with major mood disorders may be more lethal than suicide attempts by individuals in the general population. The reported ratio of suicide attempts to deaths from suicide averages between 3:1 and 5:1 among persons with mood disorders, whereas in the general population the suicide attempt rate has been estimated to be about 10–20 times (average, 18 times) greater than the suicide rate, or about 0.3% per year (700).

For individuals with mood disorders, it is also important to note factors that are particularly associated with increased risk. Fawcett et al. (79, 313) determined time-related predictors of suicide in a sample of 954 psychiatric inpatients in the NIMH Collaborative Program on the Psychobiology of Depression, about one-third of whom had bipolar disorder and the rest of whom had other mood disorders. During the initial 10 years of follow-up, 34 patients died by suicide, an overall rate that was extremely low, at 0.36% per year. The first year of follow-up was the time of highest risk, with 38% of suicides occurring during that period. Within 1 year of admission, six factors were associated with suicide: panic attacks, severe psychic anxiety, diminished concentration, global insomnia, moderate alcohol abuse, and anhedonia. The three factors associated with suicide that occurred after 1 year were severe hopelessness, suicidal ideation, and history of previous suicide attempts. By 14 years, among individuals for whom follow-up information was available, 36 had died by suicide, 120 had attempted suicide, and 373 had no recorded suicide attempt (247). Analysis at that time point showed that patients who died by suicide and patients with suicide attempts shared core characteristics, including a history of previous suicide attempts, alcohol and substance abuse, impulsivity, and psychic turmoil within a cycling/mixed bipolar disorder. In contrast to suicide within 12 months of intake, which was predicted by clinical variables, suicide beyond 12 months was prospectively predicted by temperament attributes, such as higher levels of impulsivity and assertiveness. Stressful life events (701), executive dysfunction (702), and higher levels of depression (10, 78, 221, 222, 703) may also be associated with greater risk, as may an awareness of the discrepancies between a previously envisioned "normal" future and the patient's likely degree of future chronic disability (273).

In summary, mood disorders are consistently identified as conferring a significant increase in the risk for suicide as well as for suicide attempts. However, among individuals with mood disorders, a variety of factors commonly modify that risk and should be taken into consideration during the assessment and treatment planning processes. These factors include the specific mood disorder diagnosis and duration of illness, the type and severity of the mood episode, the prior history of treatment, the presence of comorbid diagnoses or specific psychiatric symptoms such as severe anxiety or agitation, and the occurrence of significant psychosocial stressors. It is important to note, however, that this increased risk of suicidal behaviors among individuals with mood disorders has been consistently shown to be modifiable with treatment (see Section VI.D, "Somatic Therapies").

b) Schizophrenia

Schizophrenia has also been associated with an increase risk of suicide in multiple studies. Harris and Barraclough (172), for example, analyzed data from 38 studies that had follow-up periods of up to 60 years. Acknowledging that some heterogeneity in the diagnosis of schizophrenia across studies was likely as a result of changes in diagnostic criteria, the authors noted 1,176 suicides among more than 30,000 patients with schizophrenia, yielding an SMR for suicide in schizophrenia of 8.45. Baxter and Appleby (188), in a case registry study of long-term suicide risk in the United Kingdom, found an even higher 14-fold increase in rate ratios for suicide among individuals with schizophrenia. In contrast, using cohort-based curve-fitting techniques and data from 29 studies of mortality in schizophrenia, Inskip et al. (94) estimated the lifetime risk for suicide as 4%.

In addition to assessing suicide rates among patients with schizophrenia, longitudinal follow-up studies have also examined factors associated with increased risk of suicide. Black et al. (98) found that suicide occurred in 14 of 688 schizophrenia patients (2%) who were admitted to an Iowa psychiatric hospital over a 10-year period, with the majority of deaths occurring within 2 years of hospital discharge. Although women were found to be at relatively greater risk, the numbers of suicides significantly exceeded expected rates for both male and female patients. Nyman and Jonsson (101) found that suicide occurred in 10 of 110 (9%) young patients with schizophrenia who were hospitalized between 1964 and 1967 and followed for up to 17 years. In this group, suicide was associated with a more chronic course as well as with social and financial dependency. Dingman and McGlashan (103) longitudinally followed 163 Chestnut Lodge patients with a diagnosis of schizophrenia and noted that the 13 patients who died by suicide were predominantly male and had a later onset of illness, less chronic illness, better premorbid functioning, and a greater ability for abstract and conceptual thinking. At a later follow-up (mean=19 years), 6.4% of the Chestnut Lodge sample had died by suicide, and this group had exhibited fewer negative symptoms but more severe delusions and suspiciousness at index admission than those who did not die by suicide (93). A group of young psychotic patients who had not exhibited a chronic course was followed after discharge from an index hospitalization by Westermeyer et al. (83), who found that 36 patients died by suicide and 550 did not. Suicide occurred in about 9% of individuals with schizophrenia and was more likely during the early years of their illness, particularly within 6 years of initial hospitalization. At greater risk for suicide were unmarried white male patients with chronic symptoms, relatively high IQs, and a gradual onset of illness.

De Hert et al. (89) studied outcomes for 870 patients (536 men and 334 women) with schizophrenia (87%) or schizoaffective disorder (13%) after a mean duration of follow-up of 11.4 years. Sixty-three individuals died by suicide, yielding a suicide rate of 635 per 100,000 per year and an SMR for suicide of 39.7. The frequency of suicide in men was twice that in women, although the SMR and the age at the time of suicide did not differ significantly between the sexes. Of the suicides, 33 (52.4%) occurred while the patient was hospitalized (although only nine actually took place in the hospital) and 12 (19.1%) occurred during the first 6 months after discharge. When the patients who died by suicide were compared with an age- and sex-matched group of 63 patients from the remaining sample, a number of differences between the groups were observed. Those who died by suicide were more likely to have a family history of suicide, had had more and shorter hospitalizations and more past suicide attempts, and were more likely to have used a highly lethal method in prior suicide attempts. They also had higher total WAIS IQ scores and were more likely to have been psychotic or depressed or to have suffered a major loss in the 6 months before death or follow-up. Compared with control subjects, the patients who died by suicide were also less likely to have received community-based care and were less likely to have had a useful daily activity, remission of symptoms, or an early onset of prominent negative symptoms.

Among individuals with schizophrenia who die by suicide, a number of demographic factors seem to be present more often than in living control subjects. In a cohort of 9,156 patients with schizophrenia, Rossau and Mortensen (95) individually matched 10 control subjects to each of 508 individuals who were admitted to Danish hospitals between 1970 and 1987 and who later died by suicide. They found suicide risk to be particularly high during the first 5 days after discharge, with some excess suicides during temporary hospital leaves. Increases in risk were also associated with multiple psychiatric admissions during the previous year, previous suicide attempts, previous diagnosis of depression, male sex, and previous admissions to general hospitals for physical disorders. Breier and Astrachan (102) compared 20 schizophrenia patients who died by suicide with a randomly selected sex-matched group of nonsuicidal schizophrenia patients and a group of persons without schizophrenia who died by suicide. Patients with schizophrenia who died by suicide were more likely to be men and tended to be young, white, and never married. In contrast to the persons without schizophrenia who died by suicide, the schizophrenia patients who died by suicide tended not to show a temporal relationship of suicide with suicide attempts or stressful life events.

Among individuals who died by suicide, comparisons have also been made between those with schizophrenia and those with other diagnoses. Heila et al. (100) used psychological autopsy data for 1,397 individuals who died by suicide over a 1-year period in Finland and compared the 92 individuals with schizophrenia (7%) to the remainder of the sample. They found that suicide occurred at any point during the course of schizophrenia and over a large age range. In addition, among the individuals with schizophrenia, 71% had a history of suicide attempts, and, particularly in women, active illness and depressive symptoms were often observed immediately before the suicide. Significant life events, however, were seen less often before suicide in individuals with schizophrenia than in those with other diagnoses (46% and 83%, respectively).

Other studies have found suicidal ideation and suicide attempts to be common among individuals with schizophrenia. For example, in the Chestnut Lodge sample, over an average of 19 years of follow-up, 40% of the patients with schizophrenia spectrum disorders reported suicidal ideation since their initial hospitalization, and 23% reported at least one suicide attempt (93). Radomsky et al. (80) evaluated lifetime rates of suicidal behavior among 1,048 consecutively admitted psychiatric inpatients with DSM-III-R psychotic disorders. Of the 454 individuals with a diagnosis of schizophrenia, 27.3% reported at least one lifetime suicide attempt, with an additional 26.4% reporting suicidal ideation only. For the 159 patients with schizoaffective disorder, 42.8% and 27% reported suicide attempts and suicidal ideation, respectively. Roy et al. (117) found that 55% of a series of 127 consecutively admitted patients with chronic schizophrenia had previously made a suicide attempt. Harkavy-Friedman et al. (108), in a sample of 104 individuals with schizophrenia or schizoaffective disorder, found that 33% had made a suicide attempt, with 60% of those reporting multiple attempts. Attempts were often medically serious, requiring medical inpatient care in 57% of cases and emergency medical evaluation in an additional 11%, and were associated with strong suicidal intent (in the 76% of patients for whom this information was available). As with suicide in schizophrenia, initial suicide attempts tended to occur early during the course of the illness.

A number of specific factors appear to increase the likelihood of a suicide attempt among individuals with schizophrenia. For example, in the study by Roy et al. (117), those who had attempted suicide had significantly more psychiatric admissions and were more likely to have experienced a major depressive episode or received antidepressant treatment, compared with those who had not attempted suicide. Young et al. (704), in a longitudinal study of 96 individuals with recent-onset schizophrenia who were followed for a 1-year period, noted that depression was moderately correlated with concurrent suicidality but was not independently associated with future suicidality, whereas the presence of suicidal ideation even at low levels increased the risk for significant suicidal ideation or a suicide attempt during the subsequent 3 months. In their sample, Harkavy-Friedman et al. (108) found that suicide attempts were reported to be precipitated by depression (27%), loss of a significant other or other stressful life event (24%), being bothered by psychotic symptoms (11%), and responding to command hallucinations (4%). In a subsequent study of 100 individuals with schizophrenia, Harkavy-Friedman et al. (106) found that 8% of suicide attempts were associated with command auditory hallucinations for suicide and that individuals with previous suicide attempts were at particularly increased risk. In a prospective study of 333 patients with chronic schizophrenia (705), multivariate analysis suggested that current and lifetime suicide attempts and suicidal ideation were associated with hopelessness and possibly with greater levels of insight or higher cognitive functioning. Increased insight, specifically awareness of delusions and negative symptoms, has also been noted in individuals with schizophrenia who experience recurrent suicidal thoughts and behaviors (706).

In summary, an increase in the risks of suicide and suicide attempts is seen in individuals with schizophrenia and should be taken into consideration in the assessment and treatment planning process. Additional factors that modify risk include the duration of illness, the patient's insight into the illness's implications, the patient's history of treatment, and the presence of comorbid diagnoses or specific psychiatric symptoms, such as depression, hopelessness, or negative symptoms. As with mood disorders, however, increasing evidence also suggests that the risk of suicidal behaviors among individuals with schizophrenia can be modifiable with treatment (see Section VI.D, "Somatic Therapies").

c) Anxiety disorders

Data on lifetime rates of suicide among patients with anxiety disorders are limited but suggest that these diagnoses are associated with an increase in suicide risk. At the same time, it is not clear whether anxiety disorders represent an independent risk factor for suicide or whether this increased risk is attributable to the presence of depressive disorders or substance use disorders, which commonly co-occur with anxiety disorders.

Among broadly defined groups of individuals with anxiety disorders, increased rates of suicide have been seen in several studies. Khan et al. (118) used the U.S. Food and Drug Administration (FDA) database to assess the risk of suicide among patients who were participating in recent clinical trials of antianxiety medications and had diagnoses of panic disorder, social phobia, generalized anxiety disorder, posttraumatic stress disorder, or obsessive-compulsive disorder. Among the 20,076 patients, 12 died by suicide, yielding a suicide risk among patients with anxiety disorders of 193 per 100,000 patients, or at least 10-fold higher than that in the general population. This finding is particularly striking since the patients were receiving treatment and since current suicidality is generally an exclusion criterion for clinical trials. Allgulander (119) also noted an increased risk of suicide in individuals with anxiety disorders. Data on 9,912 patients with anxiety neurosis in the Swedish National Psychiatric Case Register between 1973 and 1983 yielded SMRs for suicide before age 45 of 6.7 and 4.9 for men and women, respectively. Suicide risk was highest within 3 months of discharge and was two- to threefold less than the risk in individuals with depressive neurosis.

Several studies have examined characteristics of patients with panic attacks or panic disorder who have died by suicide. Henriksson et al. (707) used data on suicides in Finland in a 1-year period to examine the relationship between panic disorder and suicide. All of the 17 persons with a current diagnosis of panic disorder who died by suicide—1.22% of the 1,397 suicides in Finland in the 1-year period—also had another axis I disorder, most often major depression. A substance use disorder was found in one-half of these individuals, with almost one-half of those persons also receiving an axis II diagnosis. These results are in accord with those of a study by Barraclough et al. (65), which found a principal diagnosis of either alcoholism or depression in virtually all persons who died by suicide and who had had a panic attack in the week before death.

Two smaller follow-up studies of patients with panic disorder yielded similar conclusions. Noyes et al. (122) found that 4% (three of 74) of patients with panic disorder followed up after 7 years had died by suicide, with an additional 7% (five of 74) having made a serious suicide attempt. Comorbid diagnoses, particularly major depression and axis II disorders, were more likely to be present in those who died by suicide and in serious suicide attempters. Coryell et al. (123) found that 35 years after an index admission, approximately 20% of 113 patients with panic disorder had died by suicide and that alcoholism and secondary depression may have had a role in those deaths.

Rates of suicidal ideation and suicide attempts are also increased in individuals with anxiety disorders, but again, comorbid diagnoses may play a role in mediating this effect. In a random sample of 18,011 adults from five U.S. communities derived from the ECA study, Weissman et al. (125) found that the presence of suicidal ideation and suicide attempts varied. Levels were highest among subjects with a lifetime diagnosis of panic disorder, followed by those who had panic attacks but not panic disorder and those with other DSM-III disorders but not panic attacks or panic disorder; lower levels were found in individuals with no prior panic attacks or DSM-III diagnoses. Weissman et al. also found that 20% of the subjects with panic disorder and 12% of those with panic attacks had made suicide attempts. Furthermore, this increase in risk was not solely attributable to comorbid diagnoses, since the lifetime rate of suicide attempts for persons with uncomplicated panic disorder was consistently higher than that for persons with no psychiatric disorder (7% and 1%, respectively) (708). This conclusion contrasted with the findings of Hornig and McNally (137), who reanalyzed the ECA data with the effects of comorbid disorders and sociodemographic variables controlled in the aggregate rather than singly. Using both stepwise and backward logistic regression analyses, they did not find panic disorder to be associated with a significant increase in risk for suicide attempts beyond that predicted by the presence of other disorders.

Other investigators have assessed other populations to determine whether panic attacks or panic disorder is associated with increases in suicidal behaviors. Pilowsky et al. (130), in a study of 1,580 adolescents in an urban public school system, found that suicidal ideation was three times more likely and a history of suicide attempts twice as likely in individuals with panic attacks, even after the effects of demographic factors, major depression, and substance use were controlled. Fleet et al. (136) assessed 441 consecutive patients who presented to an emergency department with chest pain and who underwent a structured psychiatric interview. Of the total sample, 108 (25%) met the DSM-III-R criteria for panic disorder. The investigators found that more of those with panic disorder had experienced suicidal ideation during the preceding week than of those without panic disorder (25% and 5%, respectively), even after controlling for the effect of coexisting major depression. In addition, of the 44 patients (10% of the sample) who had experienced suicidal ideation during the preceding week, 60% met the DSM-III-R criteria for panic disorder (709). Thus, in both of these populations, panic attacks or panic disorder was a significant risk factor for suicidal ideation or suicide attempts, independent of comorbid disorders.

Other studies have assessed psychiatric outpatients with panic disorder and have demonstrated substantial variability in its effect on suicidality. Cox et al. (124), for example, used the suicide questions from the ECA study to assess 106 patients with panic disorder and found that 31% of the patients reported suicidal ideation and 18% reported a history of suicide attempts. Very few individuals with suicidal ideation reported actual suicide attempts within the preceding year. However, when suicide attempts did occur, they were predominantly in the context of depressed mood. In a sample of 100 outpatients with panic disorder, Lepine et al. (129) found that 42% had a prior suicide attempt. Suicide attempters were more likely to be female or unmarried, and 88% of the patients met the DSM-III-R criteria for at least one additional diagnosis, predominantly major depressive disorder (52%) or substance use disorder (31%). Warshaw et al. (127) followed 498 patients with panic disorder and found a 6% risk of suicidal behaviors over a 5-year period. Being married or having children were protective factors, whereas mood disorders, substance use, eating disorders, personality disorders, female sex, and a prior history of suicide attempts were associated with increased risk. In the absence of other risk factors, the risk of a suicide attempt in persons with panic disorder was minimal. King et al. (126) studied 346 depressed outpatients and found a significant difference in the frequency of suicide attempts in those with a history of panic attacks compared with those without such a history (26.9% and 16.8%, respectively). Paradoxically, however, depressed patients with a history of infrequent panic attacks had a higher incidence of suicide attempts than those with panic disorder (32.3% and 21.5%, respectively). Friedman et al. (710) assessed 293 patients with panic disorder, of whom 59 had comorbid borderline personality disorder. A past history of suicide attempts was reported by 25% of the patients with comorbid borderline personality disorder and by 2% of those without that comorbidity. In contrast, Beck et al. (128) found that none of the 73 patients with primary panic disorder in a study of 900 consecutive psychiatric outpatients reported having made a prior suicide attempt.

Other anxiety disorders, although less well studied, may also influence suicide attempts or suicidal ideation. For example, in recent clinical trials of new antianxiety medications that included patients with a broad range of diagnoses, the risk of suicide attempts was increased relative to the general population, with attempts occurring in 28 of 20,076 patients, for an annualized risk of 1,350 per 100,000 patients (118). Cox et al. (124) found that of 41 outpatients with a diagnosis of social phobia, 14 (34%) had experienced suicidal ideation and two (5%) had made a suicide attempt within the prior year, although five (12%) had at least one lifetime suicide attempt. Oquendo et al. (131) assessed 156 inpatients with a diagnosis of major depressive episode and found that those with comorbid posttraumatic stress disorder were more likely to have attempted suicide, a finding that was more prominent in women than in men and that was independent of the presence of borderline personality disorder. Schaffer et al. (132) retrospectively reviewed the assessments of 533 patients with major depression and found that suicidal ideation was present in 57.8%. Suicidal ideation was more likely to be present in the 43.2% of the sample that had a lifetime anxiety disorder, and this association was independent of either age or severity of depressive symptoms. In contrast, in a study of 272 inpatients with at least one major depressive episode, Placidi et al. (220) found that rates of panic disorder did not differ between the 143 patients who had attempted suicide and the 129 patients who had not. In fact, agitation, psychic anxiety, and hypochondriasis were more severe in the nonattempter group, and these effects were independent of severity of aggression and impulsivity. However, rates of comorbid borderline personality disorder were much greater in those who attempted suicide, which may have contributed to these findings.

Even subsyndromal anxiety symptoms may contribute to an increase in risk. For example, Marshall et al. (135) found that rates of suicidal ideation increased linearly and significantly with an increasing number of subthreshold symptoms of posttraumatic stress disorder. They reported that for the 2,608 of 9,358 individuals who were screened in 1997 as part of National Anxiety Disorders Screening Day and who reported at least one symptom of posttraumatic stress disorder of at least 1 month's duration, the risk of suicidal ideation was increased, even after controlling for the effect of comorbid major depressive disorder.

Anxiety disorders may be overrepresented among individuals with suicidal ideation or suicide attempts. Pirkis et al. (383) analyzed data from 10,641 respondents in the Australian National Survey of Mental Health and Wellbeing and found that the relative risk of anxiety disorder was increased 3.5-fold in individuals with suicidal ideation in the prior year and increased sevenfold in those with a suicide attempt in the prior year.

Thus, available evidence suggests that anxiety disorders, particularly panic disorder, may be associated with increased rates of suicidal ideation, suicide attempts, and suicide. It remains unclear whether panic attacks and panic disorder represent independent risk factors for suicide or whether elevations in suicidality associated with these disorders are simply a reflection of comorbidity with other disorders such as depression, substance use disorders, or personality disorders. Nonetheless, individuals with anxiety disorders warrant explicit evaluation and follow-up for comorbid diagnoses and for suicide risk. Psychiatrists should also be alert for masked anxiety symptoms and for anxiety disorders that are misdiagnosed as physical illnesses.

d) Eating disorders

Studies point to eating disorders in general as a risk factor for death and as a likely risk factor for suicide. For example, Harris and Barraclough (64) calculated SMRs for suicide using data from 15 studies and found a 23.1-fold increase in risk in patients with eating disorders. Herzog et al. (138), in an 11-year longitudinal study of 246 women with eating disorders, noted a crude mortality rate of 5.1% and an SMR for death by any cause of 9.6. Three of the women died by suicide, yielding a significantly elevated SMR for suicide of 58.1.

The risk associated with specific eating disorders is less clear. Eckert et al. (139), in a similar longitudinal study, examined the clinical course and outcome of anorexia nervosa in 76 severely ill females. Although none of the deaths were attributed to suicide, by the time of 10-year follow-up, five subjects (6.6%) had died, yielding an almost 13-fold increase in mortality.

Coren and Hewitt (140) extracted data from all death certificates in the United States registered with the National Center for Health Statistics from 1986 through 1990. Of 5.5 million females who died in that period, 571 had anorexia nervosa listed as an underlying cause or accompanying condition of death. Of these, 1.4% died by suicide, compared to 4.1% of a matched control sample, suggesting that the risk of suicide in persons with anorexia nervosa is, if anything, lower than that in control subjects. However, substantial underreporting bias may be present, since personnel recording information on death certificates may not recognize anorexia as a contributory comorbid condition.

Eating disorders, particularly bulimia nervosa, have also been associated with an increased rate of suicide attempts, and, conversely, suicide attempters may have an increased rate of abnormal eating behaviors. Kent et al. (141) compared 48 women who were referred for psychiatric assessment after an act of deliberate self-poisoning with 50 control subjects who were evaluated in a hospital emergency department after a minor accidental injury. Even after controlling for the effect of differences in rates of depression, the investigators found that disordered eating behaviors were significantly more prevalent in the self-poisoning group. Compared to the general community, for whom surveys suggest rates of bulimia nervosa of 1%–2%, four subjects (8%) in the self-poisoning group met the diagnostic criteria for bulimia nervosa. Thus, awareness of eating disorders may be important in evaluating patients after a suicide attempt.

By the same token, suicide attempts may be more likely in women with eating disorders. Using anonymous survey data gathered from 3,630 girls in grades 6 through 12 in the upper Midwest, Thompson et al. (142) found that eating disturbances and aggressive behavior were significantly associated with substance use and with attempted suicide. In addition, adolescents reporting disturbed eating behaviors were three times more likely to report suicidal behaviors than were other respondents.

In summary, individuals with eating disorders may be at increased risk for suicidal behaviors. Anorexia nervosa seems more likely to be a potential risk factor for suicide, whereas bingeing, purging, and bulimia may be more likely to be associated with suicide attempts. The role of comorbid diagnoses in increasing the risk of suicidal behaviors remains unclear. Also unclear is whether the self-imposed morbidity and mortality associated with severe caloric restriction or bingeing and purging should be viewed as a self-injurious or suicidal behavior. Regardless, clinicians conducting suicide risk assessment should be attentive to the presence of eating disorders and especially the co-occurrence of eating disorders with other psychiatric disorders or symptoms such as depression or deliberate self-harm.

e) ADHD

The relationship between ADHD and suicidal behavior is unclear, with some but not all studies indicating an association between the diagnosis of ADHD and suicide attempts or suicide. To identify psychiatric risk factors for adolescent suicide, Brent et al. (145) used psychological autopsy data to match 67 adolescents who died by suicide to community control subjects. At the time of death, 89.6% of those who died by suicide had a psychiatric disorder, with major depression, bipolar disorder–mixed episode, substance use disorder, and conduct disorder seen at increased rates relative to the rates for the community control subjects. In contrast, the rate of ADHD in those who died by suicide was 13.4% and did not differ from the rate in the control subjects (145). Similarly, in a case-control study of adolescent suicide attempters, ADHD was actually less likely in those who attempted suicide than in the control subjects (146).

In a group of subjects between ages 17 and 28 years, Murphy et al. (144) compared 60 subjects with ADHD, combined type, to 36 subjects with ADHD, predominantly inattentive type, and to 64 community control subjects. A higher proportion of the group with ADHD, combined type (15%), reported attempting suicide, compared with the group with ADHD, predominantly inattentive type (2.8%), and the control group (0%). Compared to the control group, both ADHD groups had greater amounts of psychological distress, received more prescriptions for psychiatric medication and more types of psychiatric services, and had a higher prevalence of alcohol/cannabis use disorders and learning disorders. The groups did not differ in comorbidity of conduct disorder, major depressive disorder, or anxiety disorders. Patients with the combined type of ADHD are clinically more likely to present with distractible and impulsive behavior, whereas patients with the predominantly inattentive type of ADHD are more likely to present with problems of staring, daydreaming, confusion, passivity, withdrawal, and sluggishness or hypoactivity. These differences in clinical features may account for the differences in the numbers of suicide attempts in the two subgroups.

Nasser and Overholser (143) examined the lethality of suicidal behavior in 60 hospitalized adolescent inpatients who had recently attempted suicide. The subjects were divided into three equal groups on the basis of the qualities of their suicidal acts (nonlethal, low-lethal, and high-lethal). The groups did not differ significantly in terms of hopelessness, depression, substance abuse, and self-esteem or in diagnoses of major depression, adjustment disorder, substance abuse, and bipolar disorder. However, the group of high-lethal attempters included four individuals with a diagnosis of major depressive episode and comorbid ADHD. Thus, it may be the comorbidity of ADHD with other disorders that increases the relative lethality of suicide attempts.

In summary, evidence for an independent association between ADHD and risk for suicide or attempted suicide appears weak. Individuals with ADHD, combined type, seem to be at greater risk than those with ADHD, predominantly inattentive type, perhaps because of an increased level of impulsivity. In addition, there may be a relationship between ADHD and suicide risk that relates to comorbidity with conduct disorder, substance abuse, and/or depressive disorder. Given the frequent occurrence of ADHD in patients with other psychiatric disorders, it is important for psychiatrists to be aware that comorbid ADHD may augment the risk of suicidal behaviors.

f) Alcohol use disorders

The presence of an alcohol use disorder increases suicide risk. Estimates based on computerized curve fitting of data from 27 studies have suggested a 7% lifetime risk of suicide in individuals with alcohol dependence (94). Other approximations of lifetime suicide risk have ranged from 3.4% to as high as 15% (148, 157) but vary by country and depend on the definition of alcoholism used. In fact, the vast majority of studies have not used the DSM-IV criteria for alcohol use disorders, making comparisons across studies difficult. As a result, descriptions of studies in this document will use the diagnostic terms employed by the study authors.

Harris and Barraclough (64) used data from 32 publications, including findings for more than 45,000 individuals with follow-up periods for up to 30 years, to calculate an SMR for suicide of 5.86 among persons with alcohol abuse or dependence. The overall suicide rate for women with alcohol abuse or dependence was about 20 times the expected rate, whereas the rate for men was only about four times the expected rate. Beck et al. (227) also found a risk of suicide in alcoholics that was about fivefold greater than in nonalcoholics in a sample of 413 patients hospitalized for a suicide attempt and prospectively followed for 5–10 years. They also noted that the timing of suicides was spread throughout the follow-up, with no particular period of increased risk.

The association between alcohol use disorders and suicide is also demonstrated by psychological autopsy studies, which show alcohol use disorders to be common among individuals who die by suicide. For example, Henriksson et al. (59), in a random sample of 229 Finnish suicide deaths during a 1-year period, found that alcohol dependence was present in 43% of cases. In the United States, Conner et al. (150) found that 39% of 141 individuals who died by suicide over a 2.5-year period had had a history of alcohol use disorder.

Significant rates of alcohol use were also seen in a sample of youth suicides that included older adolescents. Brent et al. (151) examined death certificates and coroners' reports for all suicides, undetermined causes of death, and questionable accidents for 10- to 19-year-old residents of Allegheny County, Pennsylvania, from 1960 to 1983. Altogether, 159 definite suicides and 38 likely suicides were noted, but the suicide rate increased markedly over the study period, particularly among white males ages 15–19 years. During the study period there was also a 3.6-fold increase in the percentage of suicides with detectable blood alcohol levels (12.9% in 1968–1972, compared to 46.0% in 1978–1983). In addition, the rate of suicide by firearms increased much faster than that by other methods (2.5-fold and 1.7-fold, respectively), and persons who died by suicide with firearms were almost five times more likely to have been drinking than individuals who used other suicide methods.

A number of factors have been specifically observed with suicide in individuals with alcohol use disorders (153). Murphy et al. (152), in a study of 50 alcoholics who died by suicide, found that 26% had experienced interpersonal loss within 6 weeks of their death. These findings were comparable to those in a prior group of 31 alcoholics who died by suicides, one-third of whom had experienced the loss of a close interpersonal relationship within 6 weeks of the suicide. An earlier study by Murphy and Robins (156) also found a high proportion of recent interpersonal disruptions, as did a study of suicides in San Diego by Rich et al. (67). To identify other factors associated with increased risk for suicide among alcoholics, a subsequent study by Murphy et al. (149) pooled these two similar groups of alcoholics who died by suicide and compared them to two control samples of white male alcoholics, one from a psychiatric patient population and one from the ECA community-based population. Clinical features that were significantly more frequent among those who died by suicide than among the control subjects included current alcohol use, poor social support, serious physical illness, unemployment, living alone, and having made a suicidal communication. Eighty-three percent of the alcoholics who died by suicide had four or more of the seven risk factors.

Pirkola et al. (154) also examined factors associated with increased likelihood of suicide among alcohol misusers. They found that alcohol misusers who died by suicide (N=349) were more likely to be young, male, and divorced or separated, compared with individuals who did not misuse alcohol in the several months preceding their suicide (N=648). Alcohol misusers were also more likely to be intoxicated with alcohol at the time of death or to have died from an overdose of medications. Those with alcohol misuse had also experienced more adverse life events close to the time of their suicide despite having better psychosocial adjustment earlier in their lifetime. For example, alcoholics who died by suicide had more often worked but were also more likely to be recently unemployed.

A number of studies have identified comorbid disorders as being common among individuals with alcohol use disorders who die by suicide. In a series of 1,312 alcoholics admitted to a Swedish psychiatric hospital between 1949 and 1969 and followed through 1980, Berglund (157) found that alcoholics who died by suicide had a higher rate of depressive and dysphoric symptoms than alcoholics who died of other causes or who were alive at the end of the follow-up period. Murphy et al. (152) also found that concurrent depression was present in most but not all of their sample of alcoholics who died by suicide, suggesting that depression was neither a necessary nor a sufficient precondition for suicide. In a later study, Murphy et al. (149) found that major depressive episodes were significantly more frequent among alcoholics who died by suicide than among alcoholic control subjects and also found that 58% of the alcoholics who died by suicide had comorbid major depression.

Shaffer et al. (159) compared 120 individuals under age 20 who died by suicide to 147 age-, sex-, and ethnicity-matched community control subjects and found that 59% of the subjects who died by suicide and 23% of the control subjects met the DSM-III criteria for a psychiatric diagnosis based on information obtained from the subject's parents. When information from multiple informants was obtained, 91% of the subjects who died by suicide met the criteria for a DSM-III psychiatric diagnosis. In addition, with increasing age, there was an increased prevalence of a psychiatric diagnosis in general and of a substance and/or alcohol use disorder in particular. Previous suicide attempts and mood disorders were risk factors for suicide in both male and female subjects, whereas substance and/or alcohol abuse occurred exclusively in males and was present in 62% of 18- to 19-year-old subjects who died by suicide.

Even in individuals whose alcohol use disorder has remitted, suicide risk may still be increased but is likely to be influenced by comorbid disorders. Conner et al. (150) analyzed data from a community sample of 141 individuals who died by suicide and found that 39% (N=55) had a history of alcohol misuse. Compared with those who were actively using alcohol, those with remitted alcohol use disorders were predominantly younger individuals with psychotic disorders or older individuals with major depression.

In addition to being associated with an increased risk of suicide, alcohol use disorder is associated with a greater likelihood of suicide attempts. For example, Petronis et al. (163) analyzed data from 13,673 participants in the ECA survey and found that active alcoholism was associated with an 18-fold increase in the relative odds of making a suicide attempt. Gomberg (162) compared 301 women admitted to 21 alcohol treatment facilities to an equal number of age-matched nonalcoholic women from the community. Alcoholic women were far more likely to have attempted suicide (40%, compared with 8.8% of nonalcoholic women), and suicide attempts were particularly likely among alcoholic women under age 40. Alcoholic women who had attempted suicide were more likely to have used other drugs, and they reported significantly more tension, explosiveness, indecisiveness, fearfulness/anxiety, and difficulty concentrating and getting up in the morning.

Among alcoholics, differences also have been noted between those who attempt suicide and those who do not. Roy et al. (165), for example, performed a case-control study to determine the differences between alcoholic suicide attempters and alcoholic nonattempters. Of the 298 alcoholic patients studied, 19% had attempted suicide. Compared with the nonattempters, the attempters were significantly more likely to be female, to be young, and to have a lower economic status. They also were more likely to have first- or second-degree relatives who abused alcohol, to consume a greater amount of alcohol when drinking, and to have begun heavy drinking and experienced the onset of alcohol-related problems at an earlier age.

In addition, comorbid diagnoses are frequently identified among alcoholics who attempt suicide. Roy et al. (165), for example, found the most common comorbid psychiatric diagnoses among alcoholic suicide attempters to be major depression, antisocial personality disorder, substance abuse, panic disorder, and generalized anxiety disorder. Hesselbrock et al. (166), in a sample of 321 inpatients (231 men, 90 women) in alcoholism treatment centers, found that suicide attempters typically had multiple psychiatric diagnoses (e.g., depression, antisocial personality disorder, and substance abuse) and more severe psychiatric symptoms than nonattempters. Two-thirds of alcoholics who attempted suicide had a lifetime diagnosis of major depressive disorder, and most reported symptoms of depression within 2 weeks of the interview. Alcoholic suicide attempters tended to have a parental history of alcoholism, to have begun abusing alcohol at an early age, and to have abused other substances in addition to alcohol.

Preuss et al. (167), using data for 3,190 alcohol-dependent individuals from the Collaborative Study on the Genetics of Alcoholism, found that alcohol-dependent individuals with a history of suicide attempts were more likely to be dependant on other substances and more likely to have other psychiatric disorders. In addition, subjects with suicide attempts had a more severe course of alcohol dependence and more first-degree relatives with suicide attempts. In a subsequent study that followed 1,237 alcohol-dependent subjects over 5 years, Preuss et al. (168) found that the 56 alcohol-dependent subjects with suicide attempts during the follow-up period were more likely to have a diagnosis of a substance-induced psychiatric disorder or be dependent on other drugs. Furthermore, among 371 alcohol-dependent individuals who had made a suicide attempt and also had had an episode of depression, the 145 individuals (39.1%) with alcohol-independent mood disturbance had a greater number of prior suicide attempts and were more likely to have an independent panic disorder but reported a less severe history of alcohol dependence and were less likely to have been drinking during their most severe attempt (169). These findings suggest that in taking a clinical history in suicide attempters it is useful to identify comorbid depression but also to determine whether depressive episodes are alcohol induced or not.

That the presence of prior attempts is predictive of future attempts also highlights a need for taking a thorough history of past suicidal behaviors. Preuss et al. (168) followed 1,237 alcohol-dependent subjects over 5 years and found that the 56 alcohol-dependent subjects with suicide attempts during the follow-up period were more likely to have made prior attempts than subjects with no suicide attempts. Persons with comorbid major depression and alcohol use have higher rates of suicidal symptoms than those with either alone. Cornelius et al. (170) compared 107 patients with both major depression and alcohol dependence to 497 nondepressed alcoholics and 5,625 nonalcoholic patients with major depression assessed at the same psychiatric facility using a semistructured initial evaluation form. Depressed alcoholics had a significantly greater degree of suicidality, as reflected by a global measure that included wishes for death, suicidal ideation, and suicidal behaviors. They also differed significantly from the nonalcoholic depressed patients in having lower self-esteem and greater impulsivity and functional impairment.

In a subsequent study, Cornelius et al. (171) found that among psychiatrically hospitalized alcoholics with major depression, almost 40% had made a suicide attempt in the week before admission, with 70% having made a suicide attempt at some point in their lifetime. There was a significant association between recent suicidal behavior and recent heavy drinking, with most subjects also reporting drinking more heavily than usual on the day of their suicide attempt. In addition, these suicide attempts were usually impulsive. Suicidal ideation, however, was not increased by more recent heavy alcohol use, suggesting that alcohol increases suicidal attempts by increasing the likelihood of acting on suicidal ideation.

In summary, alcohol use disorders are associated with increased risks of suicide and suicide attempts. Conversely, rates of alcohol use disorders are elevated among those who die by suicide as well as among suicide attempters. The common occurrence of comorbid psychiatric symptoms and diagnoses suggests a need for thorough assessment and treatment of such complicating factors in users of alcohol. Also, the frequent presence of psychosocial stressors including unemployment and interpersonal losses should also be taken into consideration in the assessment and treatment planning process.

g) Other substance use disorders

As with disorders of alcohol use, other substance use disorders may be associated with an increased risk of suicide. Harris and Barraclough (172) noted that the SMRs for suicide varied widely across studies and that calculations were often confounded by the subjects' simultaneous use of multiple substances and by the difficulties in distinguishing accidental overdoses from suicide. Nonetheless, their meta-analysis of published literature found that substance use disorders were associated with a substantial increase in suicide risk. The SMRs for suicide were 14.0 for those with opioid abuse or dependence; 20.3 for those with sedative, hypnotic, or anxiolytic abuse or dependence; and 19.2 for individuals with mixed substance abuse or dependence.

Among individuals with substance use disorders, suicide may be more likely in the presence of comorbid diagnoses such as mood disorders. For example, in a study comparing 67 adolescents who died by suicide to 67 demographically matched community control subjects, Brent et al. (145) found that substance abuse conferred more significant risk when it was comorbid with affective illness than when it was present alone (odds ratio of 17.0 and 3.3, respectively). Lesage et al. (174) compared 75 male subjects ages 18–35 years who died by suicide to a group of 75 demographically matched living control subjects and found significantly greater rates of DSM-III-R psychoactive substance dependence among the subjects who died by suicide (22.7% versus 2.7%). They also found that comorbid major depression or borderline personality disorder was common among those with substance dependence who died by suicide.

In psychological autopsy studies, diagnoses of substance use disorders are particularly common among individuals under age 30 who die by suicide. For example, Fowler et al. (173) studied a subset of 128 individuals from the San Diego Suicide Study (67) who were under 30 years old and found that 53% had a diagnosis of substance abuse. Of this group, about one-half had an additional psychiatric diagnosis such as atypical depression, atypical psychosis, or adjustment disorder with depression. Despite the young age of the study sample, substance abuse was typically a chronic condition that had been present for an average of 9 years. Abuse of multiple substances was the norm, with marijuana, alcohol, and cocaine being the most frequently abused substances. Other data from the San Diego Suicide Study sample as a whole (110) showed that most substance users abused alcohol as well as other substances, with relatively small numbers of "pure" alcoholics or "pure" substance users. They also noted that interpersonal conflicts or loss occurred more frequently near the time of death for substance abusers with and without depression than for persons with mood disorders alone.

Although the majority of persons with substance use disorders who die by suicide are male, it is important to recognize that men and women with substance use disorders may differ in their characteristics and their risk for suicide. Pirkola et al. (175) used data from a nationwide psychological autopsy study in Finland to study the characteristics of a sample of 172 men and 57 women who died by suicide and had a DSM-III-R diagnosis of psychoactive substance dependence. They found that women were more likely than men to have abused or been dependent on prescribed medication. In addition, women were more likely than men to have a substance use disorder preceded by a comorbid axis I disorder (45% and 18%, respectively). Borderline personality disorder, previous suicide attempts, and suicidal communications were more common in women age 40 years or younger. In addition, alcohol-dependent women died at a younger age than women with nonalcohol substance dependence and also died at a younger age than men with either alcohol dependence or nonalcohol substance dependence.

Available evidence suggests that suicide attempts are common in substance users and that substance use disorders are associated with an increased risk of suicide attempts. Borges et al. (176) used data from the U.S. National Comorbidity Survey, a nationally representative sample of 8,098 persons age 15–54 years that was carried out in 1990–1992, to examine whether retrospectively reported substance use, abuse, and dependence are predictors of the onset of suicidal behavior. After controlling for the effects of sociodemographic factors and comorbid psychiatric disorders, the investigators found that subsequent suicide attempts were predicted by use of alcohol, heroin, or inhalants. Current substance use, rather than a history of use, increased the likelihood of suicidal behavior, with the number of substances used being more important than the types of substances used. In addition, among those with suicidal ideation, current substance use, abuse, and dependence were significant risk factors for unplanned suicide attempts.

Rossow and Lauritzen (178) assessed the self-reported prevalences of nonfatal overdoses and suicide attempts in 2,051 individuals who were being treated for substance abuse. Almost one-half (45.5%) reported having had one or more life-threatening overdoses, and nearly one-third (32.7%) reported one or more suicide attempts. Suicide attempts were more often reported among those who had overdosed, and the number of life-threatening overdoses and number of suicide attempts were positively and moderately associated. Individuals who had exhibited both life-threatening behaviors also showed higher rates of HIV risk-taking behaviors, poor social functioning, and use of multiple substances. Suicide attempters also had more symptoms of depression and anxiety as measured by the Global Assessment Scale. Thus, there is substantial covariation between suicide attempts and drug overdoses in individuals with substance use disorders that is also associated with other risk-taking behaviors and poor social integration.

Individuals with substance use disorders also have an increased likelihood of making a suicide attempt, compared to control subjects. Beautrais et al. (177) compared 302 individuals who had made medically serious suicide attempts to 1,028 control subjects who were randomly selected from local electoral rolls. Overall, those who had made a serious suicide attempt had high rates of substance use disorders (odds ratio=2.6). Furthermore, of those with a serious suicide attempt, 16.2% met the DSM-III-R criteria for cannabis abuse/dependence at the time of the attempt, compared with 1.9% of the control subjects (181). Mann et al. (31), in a study of 347 consecutive admissions to a university psychiatric hospital, found that the 184 patients who had made a prior suicide attempt had a greater likelihood of past substance use disorder or alcoholism. Johnsson and Fridell (179) assessed 125 substance abusers 5 years after hospitalization for detoxification and short-term rehabilitation. Although seven patients were dead at the time of follow-up, none of the deaths were from suicide. Of 92 interviewed subjects, nearly one-half the group (45%) reported having attempted suicide at some point in their lives, with about 50% of that group having attempted suicide with prescribed psychotropic drugs such as antidepressants or sedatives. Only a few of the suicide attempts were made by using the individual's primary substance of abuse. The most common reasons given for suicide attempts were the loss of a person whom they loved and feelings of loneliness. Compared to those who had never made a suicide attempt, the suicide attempters were more likely to have had childhood psychiatric hospitalizations or experienced loss of significant others in childhood. They also were more likely to experience depressive moods or other psychiatric comorbidity.

The combination of cocaine use plus alcohol use also appears to increase the risk of suicide attempts. Cornelius et al. (711) found that of 41 consecutively admitted depressed alcoholic inpatients, 16 had made a suicide attempt and 10 had used cocaine during the week before their hospitalization. The proportion of patients making a suicide attempt in the week before admission was greater in those who had used cocaine than in those who had not (70% and 32%, respectively). Suicidal ideation was also more prevalent in the depressed alcoholics who also used cocaine. Roy (180) studied the characteristics of cocaine-dependent patients in a substance abuse treatment center and compared the 130 individuals who had never attempted suicide with the 84 individuals who had made prior suicide attempts (a mean of 2.1 prior attempts). Compared with nonattempters, attempters were more likely to be female and to have a lifetime history of alcohol dependence (58.3%, compared with 34.6% for nonattempters) and a family history of suicidal behavior (25%, compared with 5.4% for nonattempters). Attempters also had significantly higher childhood trauma scores for emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect than the nonattempters; however, these scores were not corrected for the differences in the gender ratios in the two groups. Thus, suicide attempts are common among individuals seeking treatment for cocaine dependence, and factors that seem to augment risk are similar to those for other groups of suicide attempters.

In summary, studies indicate that substance use is a significant risk factor for suicide attempts and suicide. This is particularly true in younger individuals, leading some researchers to hypothesize that increasing suicide rates among youths may be related to increasing rates of substance use. Individuals with chronic substance use disorders, those who have experienced life-threatening nonsuicidal overdoses, and those who abuse multiple substances, including alcohol, may be at even greater risk. Moreover, substance use disorders may complicate mood disorders (182), increasing susceptibility to treatment-resistant illness and psychological impairment, and on that basis may contribute to an elevated risk for suicide and for suicide attempts. As a result, it is important to evaluate individuals with suicidality in the context of substance use for the presence of comorbid mood disorders as well as other comorbid psychiatric diagnoses. The evaluation should also be aimed toward identifying patterns of recent substance use and psychosocial factors such as recent interpersonal loss or history of childhood trauma that may also affect the likelihood of suicidal behaviors among substance users.

h) Personality disorders

Although personality disorders are often comorbid with substance use disorders and with other psychiatric diagnoses, they also appear to confer an independent risk for suicide. In addition, among individuals with personality disorders, the rate of suicide may be equivalent to rates in individuals with other major psychiatric syndromes. For example, in a meta-analysis of 14 case-control samples and nine longitudinal samples of patients with personality disorders, Linehan et al. (184) found rates of suicide that were between 4% and 8%. For patients with borderline personality disorder, studies have shown suicide rates ranging from 3% to 9% (183). Harris and Barraclough (64) also found suicide risk to be increased, calculating an SMR for suicide of 7.08 among individuals with personality disorder. However, the majority of patients included in their analysis were male and from a Veterans Administration study, raising questions about the representativeness of the population. Baxter and Appleby (188), in a large case registry study of long-term suicide risk in the United Kingdom, found even higher risks for suicide among individuals with personality disorder diagnoses, with a 12.8-fold elevation of risk in men and a 20.9-fold elevation of risk in women with personality disorders.

In a longitudinal follow-up study of individuals with personality disorders, Stone et al. (185) found that 18 of the 196 patients who were able to be located had died by suicide by 16.5 years. Compared to a suicide rate of 8.5% for the borderline personality group as a whole, those with alcohol problems had a twofold increase in the rate (19%), with a 38% rate of suicide among women who had a combination of alcoholism, major affective disorder, and borderline personality disorder. Other factors that appeared to contribute to suicide in individuals with borderline personality disorder were continuing alcohol abuse, impulsivity, and a history of parental brutality, specifically sexual molestation.

Psychological autopsy studies also show significant rates of personality disorder diagnoses among individuals who die by suicide. Duberstein and Conwell (186) reviewed case-based and cohort studies on suicide in individuals with personality disorders and found that approximately 30%–40% of suicides occur in individuals with personality disorders, with increased risk conferred by the presence of borderline, antisocial, and possibly avoidant and schizoid personality disorder diagnoses.

In a random sample of all persons who died by suicide in Finland within a 1-year period, Isometsa et al. (187) found that 29% of the subjects (N=67) had an axis II disorder. All individuals with a personality disorder also had at least one axis I diagnosis, which in 95% included a depressive syndrome, a substance use disorder, or both. Individuals with cluster B personality disorders were more likely to have substance use disorders and to have had a previous suicide attempt and were less likely to have had a health care contact during their final 3 months of life. In the same group of subjects, Heikkinen et al. (190) examined data on recent life events for 56 subjects with personality disorder who died by suicide and matched those subjects to control subjects who did not have a personality disorder diagnosis. Those with a personality disorder were more likely to have experienced one or more stressful life events in the last 3 months of life as well as in the week preceding the suicide. Specifically, of those with a personality disorder, 70% had a significant event in the week before suicide, with job problems, family discord, unemployment, and financial difficulty reported most commonly. Thus, these findings suggest that individuals with personality disorders who die from suicide have high rates of comorbid depression and substance use as well as high rates of significant life stressors that precede suicide.

The increased risk of suicide with personality disorders seems to be a particular factor that contributes to risk in young adults. Lesage et al. (174) compared 75 young men who died by suicide to a demographically matched group of men in the community and found that the 6-month prevalence of borderline personality disorder was substantially increased among those who died by suicide (28.0% versus 4.0%). In a study of adolescents and young adults who had been admitted to a regional poisoning treatment center because of deliberate self-poisoning or self-injury, Hawton et al. (384) compared 62 individuals who died by suicide or possible suicide to 124 matched control subjects and found that an increased risk of death was associated with the presence of a personality disorder (odds ratio=2.1).

Suicide attempts may also be more likely to occur in individuals with personality disorders than in those with other diagnoses. In a review of the topic, Linehan et al. (184) noted that suicide attempts are estimated to occur in 40%–90% of individuals with personality disorders. Soloff et al. (193) examined data for 84 patients who met the DSM-III-R criteria for borderline personality disorder and found that 61 patients (72.6%) had a lifetime history of suicide attempts, with an average of more than three attempts per patient. Risk factors for suicide attempts in patients with borderline personality disorder included older age, prior suicide attempts, antisocial personality, impulsive actions, and a depressed mood, but not comorbid mood disorder or substance use disorder. Ahrens and Haug (194), in a case-control study of 226 patients with a personality disorder who were admitted to a psychiatric hospital, found that patients with a personality disorder (including, but not limited to, borderline personality disorder) were more likely than other hospitalized patients to have had a suicide attempt immediately before admission, with persistent clinically relevant suicidal behavior within the first 24 hours after admission (39% versus 24%). Furthermore, in patients with a personality disorder, suicidality was not related to the presence of a specific mood disorder, since only 3% of the patients with personality disorder met the criteria for a major affective syndrome. However, the rates of reported suicide attempts in individuals with personality disorder diagnoses varied with the treatment setting. Pirkis et al. (198), in a study of suicide attempts by psychiatric patients under active treatment, observed a rate of suicide attempts in acute inpatients that was 10-fold greater than those for individuals in community-based and for individuals in long-stay inpatient care (22.7 attempts per 1,000 episode-days, compared with 2.3 and 2.1 attempts per 1,000 episode-days, respectively).

Conversely, among individuals who attempt suicide, personality disorders are commonly observed. Mann et al. (31), in a study of 347 consecutive patients who were admitted to a university psychiatric hospital, found that comorbid borderline personality disorder was more common among the 184 patients who had attempted suicide than among those with no prior suicide attempts. Beautrais et al. (177) compared 302 consecutive individuals who made serious suicide attempts with 1,028 randomly selected comparison subjects. Multiple logistic regression showed that those who made suicide attempts had a high rate of conduct disorder or antisocial personality disorder (odds ratio=3.7, 95% confidence interval=2.1–6.5). Thus, both borderline personality disorder and antisocial personality disorder appear to occur more frequently among suicide attempters. In a study of consecutive patients who had attempted suicide, Suominen et al. (191) compared 65 patients who did not have a personality disorder diagnosis to 46 patients who received a diagnosis of personality disorder. Of those with a diagnosis, 74% had a cluster B personality disorder and 46% had a diagnosis of borderline personality disorder. Individuals with a personality disorder were more likely to have attempted suicide in the past (78%, compared to 57% of those without a personality disorder diagnosis) and were more likely to have had psychiatric treatment in their lifetime (85% versus 57%); however, those with and without personality disorders did not differ in their degree of intent, hopelessness, somatic severity, or impulsiveness. Personality disorders were associated with a high degree of comorbidity, with comorbid alcohol dependence being particularly common and associated with greater difficulty in pursuing follow-up.

A number of additional factors may act as contributors to risk for suicide attempts among individuals with personality disorders. Brodsky et al. (196) analyzed data for 214 inpatients with a diagnosis of borderline personality disorder according to a structured clinical interview and examined the relationship between the specific DSM-IV criteria for borderline personality disorder and measures of suicidal behavior. After excluding self-destructive behavior and controlling for the effects of lifetime diagnoses of depressive disorder and substance abuse, they found that impulsivity was the only characteristic of borderline personality disorder that was associated with a higher number of suicide attempts. In addition, the number of previous suicide attempts was associated with having a history of substance abuse.

Comorbid mood disorders are also common among suicide attempters with personality disorder diagnoses. For example, Van Gastel et al. (192), in a study of 338 depressed psychiatric inpatients, found significantly more suicide attempts and more suicidal ideation among those with a comorbid personality disorder diagnosis than among depressed inpatients without a personality disorder. In addition, Oquendo et al. (131) found that among 156 inpatients with a diagnosis of a major depressive episode, having a history of suicide attempts was independently related to the presence of a cluster B personality disorder and to PTSD.

Corbitt et al. (195) also examined the effects of comorbid borderline personality disorder in 102 individuals with mood disorders and found that the 30 patients with major depressive disorder and comorbid borderline personality disorder were just as likely to have made a highly lethal suicide attempt as the 72 patients with major depressive disorder alone. However, those with comorbid borderline personality disorder were more likely to have a history of multiple serious suicide attempts, and past suicidal behavior was better predicted by the number of personality disorder symptoms than by the number of depressive symptoms. Thus, they suggested that the severity as well as the presence of comorbid cluster B personality disorder symptoms should be ascertained in assessing the risk of suicide attempts in patients with major depressive disorder.

Soloff et al. (197) compared the characteristics of suicide attempts in 77 inpatients with major depressive episodes to suicide attempts in 81 patients with borderline personality disorder, 49 of whom had a concomitant major depressive episode. Compared to patients with borderline personality disorder alone, all of the depressed patients had more severe observer-rated depression and lower levels of functioning. Patients with borderline personality disorder had higher rates of impulsivity, regardless of whether depression was also present. However, the diagnostic groups did not differ in their subjective intent to die, their degree of objective planning for death, the violence of the suicide method, or the degree of physical damage in the attempt as measured by the Beck Suicide Intent Scale.

In summary, individuals with personality disorders, and particularly those with a diagnosis of borderline personality disorder or antisocial personality disorder, have increased risks for suicide and for suicide attempts. These risks appear to be further augmented by the presence of comorbid disorders such as major depression, PTSD, and substance use disorders. The severity of symptoms such as impulsivity may also play a role in increasing risk, suggesting that such factors should be identified and addressed in the assessment and treatment of individuals with personality disorders.

a) Anxiety

Anxiety has been suggested to increase the risk of suicide even when a specific anxiety disorder is not present. In a review of 46 cohort or case-control studies that used standardized or structured assessments of psychological dimensions to assess psychological vulnerability to suicide, Conner et al. (217) noted anxiety to be one of five constructs that is consistently associated with suicide. Busch et al. (218) reviewed the charts of 76 patients who died by suicide while in the hospital or immediately after discharge and found that 79% (N=60) met the criteria for severe or extreme anxiety and/or agitation according to Schedule for Affective Disorders and Schizophrenia ratings. In addition, Fawcett et al. (79), in a study of 954 psychiatric patients with major affective disorders, found that panic attacks and severe psychic anxiety were factors associated with suicide within 1 year of index evaluation. In these studies, the anxiety levels associated with suicide often took the form of anxious ruminations, panic attacks, or agitation and were in the severe range (i.e., severe anxiety most of the time).

With respect to the effect of anxiety on the risk of suicide attempts, the findings are more mixed. For example, Hall et al. (212) studied the characteristics of 100 patients who made a severe suicide attempt and found that severe anxiety and panic attacks were among the factors that were associated with the attempt. However, Placidi et al. (220), in a study of 272 inpatients with at least one major depressive episode, found that agitation and psychic anxiety were more severe in those who had not reported making a past suicide attempt.

Thus, although the relationship between anxiety and suicide attempts is unclear, and specific measures of anxiety have not been found to be predictive of suicide (78), severe anxiety does seem to increase suicide risk at least in some subgroups of patients. In particular, psychic anxiety, which may not be obvious to the clinician, should be specifically assessed, since such symptoms can respond rapidly to aggressive short-term treatment with benzodiazepines, second-generation antipsychotic medications, and possibly anticonvulsant medications (109).

b) Hopelessness

Hopelessness has been consistently identified as a factor associated with an increased risk of suicide, independent of diagnosis (217). Many studies that have assessed hopelessness have used the Beck Hopelessness Scale (712), which is described further in Section VI.B.1, "Rating Scales". For example, Beck et al. (222) followed 207 patients who were hospitalized for suicidal ideation but who had not made a recent suicide attempt to identify predictors of later suicide. After a follow-up period of 5–10 years, 14 individuals (6.9%) had died by suicide. Although a score of 10 or more on the Beck Hopelessness Scale correctly identified 91% of the patients who eventually died by suicide, there was significant overidentification of at-risk patients, with a false positive rate of 88%. The group who died by suicide also had a higher mean score on clinicians' ratings of hopelessness (223).

A later longitudinal study that included 1,958 consecutive psychiatric outpatients examined whether the level of hopelessness at intake could predict eventual suicide (221). In this patient population, those with a suicide death scored significantly higher on both the Beck Hopelessness Scale and the Beck Depression Inventory. Although a Beck Hopelessness Scale score of 10 or more was associated with an 11-fold increase in the likelihood of suicide, the specificity was again low. Since a high level of hopelessness is common in psychiatric patients, applying this Beck Hopelessness Scale cutoff to a larger population would identify 100 patients as being at risk for every one or two eventual suicides (713).

In a subsequent study that included an expanded sample of 6,891 psychiatric outpatients seen between 1975 and 1995 and followed for up to 20 years (with a median length of follow-up of 10 years), Brown et al. (78) used survival analysis to identify factors associated with increased risk for suicide. Along with higher levels of suicidal ideation and depression, hopelessness was identified as a risk factor for suicide, with patients who scored above 8 on the Beck Hopelessness Scale being at four times greater risk for suicide in a given year than those with lower scores.

The effect of hopelessness on suicide risk may vary by diagnosis, however. Fawcett et al. (79), in a longitudinal follow-up study of 954 patients with major affective disorder, found that severe hopelessness was one of several factors associated with an increased risk of suicide more than 1 year after the index assessment. However, among subjects who met the criteria for alcohol or substance abuse at any time, those who were not pervasively hopeless had the highest suicide risk at 5-year follow-up (226). This pattern is consistent with the findings of Beck et al. (227), who followed 161 alcohol-abusing patients for 7–12 years after they were hospitalized for a suicide attempt. Comparison of the 18 individuals who died by suicide to the remainder of the group failed to show a relationship between suicide and either hopelessness or depression.

Hopelessness at the termination of treatment may also reflect an increased risk for suicide. Dahlsgaard et al. (236) compared 17 cognitive therapy outpatients with mood disorder who died by suicide with a matching group of 17 outpatients who did not. Although the sample was small, those who died by suicide had higher levels of hopelessness at the end of treatment and were more likely to have ended treatment prematurely.

In addition to being a risk factor for suicide, hopelessness is more prominent in individuals who have reported previous suicide attempts, compared to individuals without such a history. Cohen et al. (229), for example, found greater levels of hopelessness in the 43 suicide attempters among 184 individuals with a first admission for psychosis. Hall et al. (212), in a study of 100 patients who had made severe suicide attempts, also noted feelings of hopelessness to be associated with suicidal behavior. Among 84 inpatients with DSM-III-R major depression, Malone et al. (230) found that the 45 individuals who had made a suicide attempt had higher subjective ratings of hopelessness and depression severity and that these ratings were inversely correlated with "reasons for living." In a study comparing 148 low-income African American women who had made a suicide attempt to 137 demographically similar women who presented for general medical care, Kaslow et al. (230) found hopelessness to be associated with a nearly eightfold increase in the risk of a suicide attempt in a univariate analysis. In a multivariate logistic regression analysis, hopelessness was independently associated with an increased risk of suicidal behaviors. Van Gastel et al. (192) also found that hopelessness was associated with suicidal ideation among 338 depressed inpatients and that the presence of a comorbid personality disorder was associated with additional increases in suicidal ideation and suicide attempts.

Across diagnostic groups, hopelessness appears to relate to the seriousness of suicidal ideation and intent. Soloff et al. (197) assessed the relationship of hopelessness to suicide attempts in inpatients with major depressive disorder (N=77) as well as in inpatients with borderline personality disorder alone (N=32) or in combination with major depressive disorder (N=49). Across groups, increased hopelessness was associated with an increased number of suicide attempts as well as an increase in the lethal intent associated with attempts. In addition, in patients with both disorders, higher levels of hopelessness were associated with objective planning of suicide attempts, which would further enhance risk. In a sample of 384 individuals who had attempted suicide, Weissman et al. (233) found that hopelessness contributed to the severity of suicidal intent in those with substance use disorders (N=86) as well as in those without substance use disorders (N=298).

Other evidence suggests that the level of hopelessness cannot be considered independently of other factors. For example, Uncapher et al. (228) analyzed data for 60 institutionalized elderly men and found that the relationship between hopelessness and suicidal ideation varied with the level of depressive symptoms and was most pronounced at moderate or higher levels of depression. Mendonca and Holden (232) assessed 97 outpatients and found the strongest predictors of the seriousness of current suicidal inclinations (as measured by the Beck Scale of Suicidal Ideation) to be hopelessness (as measured by the Beck Hopelessness Scale) and "unusual thinking" (defined as a state of cognitive distress with confused, disorganized thinking, including "trouble concentrating" and "mind going blank").

Furthermore, cross-sectional assessments of hopelessness may not necessarily be as relevant to risk as the level of hopelessness at baseline when the individual is not depressed. Young et al. (234), in a longitudinal study of 316 individuals, found that the baseline level of hopelessness was a better predictor of suicide attempts than either the level of hopelessness when depressed or the relative change in hopelessness from baseline levels during depression.

In summary, hopelessness is well established as a psychological dimension that is associated with an increased risk for suicide and suicide attempts and an increased level of suicidal intent. This relationship between hopelessness and suicidality holds true across diagnostic groups, with the possible exception of individuals with alcohol use disorder. These findings suggest the importance of inquiring about current levels of hopelessness as well as inquiring about usual levels of optimism about life and plans for the future. They also suggest the use of interventions to reduce hopelessness as a part of treatment.

c) Command hallucinations

Although command hallucinations have been regarded clinically as being associated with increased suicide risk, there is limited evidence that addresses this question. In addition, those studies that are available have included relatively small numbers of patients, making it difficult to detect differences in rates of suicide or suicide attempts between patient groups. Furthermore, in psychological autopsy studies, it is impossible to determine whether command hallucinations were present immediately before death or may have contributed to suicide.

Two small studies have noted whether command hallucinations had been present on index assessment in individuals who later died by suicide. Breier and Astrachan (102) described 20 schizophrenia patients who died from suicide and found that none had previously reported hallucinated suicidal commands. In contrast, Zisook et al. (239) found that command hallucinations, which were often violent in content, had been reported by 46 of 106 outpatients with schizophrenia, including the two patients who died by suicide during the study.

Other studies have tried to determine the rates at which patients follow command hallucinations and the factors that contribute to following or resisting such commands. For example, Junginger (241) used a semistructured psychiatric interview and hospital chart review to obtain information on 51 psychiatric inpatients and outpatients, all of whom had experienced recent command hallucinations. Of these subjects, 39.2% reported that they had followed the commands, 47.1% reported that they did not follow the commands, and 13.7% were unable to recall their response. Patients with hallucination-related delusions and identifiable hallucinatory voices were more likely to follow the commands than patients who were unable to identify the voices that they heard. In a subsequent study of 93 psychiatric inpatients who had a history of at least one command hallucination, the most recent command hallucination reported by the subject was rated for level of dangerousness and level of compliance with the command (242). Of the 93 subjects, 52 (56%) reported at least partial compliance with their most recent command hallucination, and 40 (43%) reported full compliance. Individuals who experienced less dangerous commands or who could identify the hallucinated voice reported higher levels of compliance, although reported compliance with more dangerous commands was not uncommon. Commands experienced in the hospital were less dangerous than those experienced elsewhere, tended to be specific to the hospital environment, and were less likely to be followed. Based on these self-reports, the authors concluded that psychiatric patients who experience command hallucinations are at risk for dangerous behavior and that a patient's ability to identify the hallucinated voice is a fairly reliable predictor of subsequent compliance. Also, the level of dangerousness that results from compliance with command hallucinations may be a function of the patient's environment.

Erkwoh et al. (245) used a 24-item questionnaire to assess the psychopathological characteristics of command hallucinations in 31 patients with schizophrenia. Like Junginger, they found that following the commands was predicted by recognizing the voice. In addition, patients were more likely to comply with commands from hallucinations that they viewed as "real" and that produced an emotional response during the hallucination.

Kasper et al. (243) compared 27 psychotic patients with command hallucinations to 27 patients with other hallucinations and 30 patients with other psychotic symptoms. Although the groups did not differ in aggressive or violent behavior or in most nonhallucinatory symptoms, 84% of the patients with command hallucinations had recently obeyed them, even during their hospital stays. Among those with command hallucinations, almost one-half had heard and attempted to obey messages of self-harm during the previous month. Rogers et al. (244), in a study of 65 forensic inpatients with psychotic disorders, also found that a significant number of individuals (44%) often responded to command hallucinations with unquestioning obedience.

These findings, that significant numbers of individuals comply with at least some hallucinated commands, are in contrast with the findings of Hellerstein et al. (240). Among 789 consecutive inpatients admitted over a 2-year period, they found that 19.1% had auditory hallucinations within 2 weeks of hospital admission, and, of these, 38.4% heard commands to behave violently or self-destructively. It is not surprising that hallucinations were more common in the 159 patients with schizophrenia, with 50.3% experiencing auditory hallucinations and 18.2% experiencing command hallucinations. Among 167 patients with affective disorder, rates of auditory and command hallucinations were 13.2% and 4.2%, respectively. The presence of auditory hallucinations was significantly associated with use of maximal observation and seclusion. However, patients with command hallucinations were not significantly different from patients without command hallucinations on demographic and behavioral variables, including suicidal ideation or behavior and assaultiveness. This finding suggests that command hallucinations alone may not imply a greater risk for acute, life-threatening behavior. In addition, consistent with the findings of Goodwin et al. (714), these findings imply that many patients are able to ignore or resist command hallucinations.

The most specific assessment of the role of command hallucinations in suicidal behaviors is that of Harkavy-Friedman et al. (106). They interviewed 100 individuals with schizophrenia or schizoaffective disorder who were hospitalized on an inpatient research unit about their experiences with command auditory hallucinations as well as about suicide attempts. Suicide attempts were reported by 33% of the sample, and the relative frequency of individuals with command hallucinations did not statistically differ between those who had and those who had not reported a suicide attempt (30% and 18%, respectively). Command hallucinations were present in 22% of the sample as a whole, and, of these, 45% had made at least one suicide attempt. Among individuals with command hallucinations who had made a suicide attempt, however, 80% had at least one attempt in response to the hallucinations. Thus, these findings suggest that for some individuals, particularly those with prior suicide attempts, suicidal behavior may occur in the context of auditory command hallucinations.

In summary, study findings are inconsistent about whether patients with command hallucinations are likely to obey them. Patients who recognize the hallucinated voices or view them as real or benevolent may be more likely to follow their directives. In addition, patients with prior suicide attempts may be more likely to follow suicidal commands. In terms of suicide risk, per se, patients with command hallucinations may not be at greater risk than other severely psychotic patients. However, existing studies include too few subjects to draw strong conclusions. In addition, since some patients do seem to act in response to auditory command hallucinations, it is important to identify such hallucinations, assess them in the context of other clinical features, and address them as part of the treatment planning process.

d) Impulsiveness and aggression

Factors such as impulsivity, hostility, and aggression may act individually or together to increase suicide risk. For example, many studies have provided moderately strong evidence for the roles of impulsivity and hostility-related affects and behavior in suicide (217, 246). In particular, impulsivity and aggression have been shown to be associated with suicide in patients with schizophrenia as well as in those with mood disorder. For example, De Hert et al. (89) compared 63 patients who died by suicide and 63 control subjects from a consecutive admission series of patients with a diagnosis of schizophrenia, all of whom were under age 30 on admission. In this sample, impulsive acting-out behavior was associated with an increased likelihood of suicide (odds ratio=6.4). Among 529 patients with affective illness who were followed naturalistically for up to 14 years and who either attempted suicide or died by suicide, Maser et al. (247) also found that impulsivity was a core characteristic of patients with suicidal behaviors. In fact, beyond 12 months, higher levels of impulsivity and assertiveness were the best prospective predictors of suicide.

Angst and Clayton (248) found a significant effect of premorbid aggression on the risk of suicide attempts or suicide. To assess the effect of personality traits on suicidal behaviors, they administered the Freiberg Personality Inventory to 6,315 Swiss army conscripts. Twelve years later, 185 of these individuals were identified as receiving psychiatric treatment during that time period, and a record review was conducted to establish a blind diagnosis and assess measures of suicidality and mortality in a subgroup of 87 of those individuals. Those who made suicide attempts or died by suicide were found to have scored higher on aggression than control subjects. In contrast, subjects with suicidal ideation alone scored lower on aggression, suggesting a role for premorbid aggression in suicidal behaviors.

Multiple other studies have demonstrated increased levels of impulsivity and aggression in individuals with a history of attempted suicide. For example, Mann et al. (31), in a study of 347 consecutive patients admitted to a university psychiatric hospital, found that rates of lifetime aggression and impulsivity were greater in the 184 patients who had attempted suicide than in those without a history of suicide attempts. Hall et al. (212) found that the recent onset of impulsive behavior was an excellent predictor of suicidal behavior in 100 patients who had made a severe suicide attempt. Kotler et al. (249) compared 46 patients with PTSD to 42 non-PTSD anxiety disorder patients and 50 healthy control subjects and found that impulsivity was positively correlated with the risk of suicidal behavior in the PTSD group.

Impulsivity and aggression have also been associated with suicide attempts among patients with mood disorders. Brodsky et al. (250), in a study of 136 depressed adult inpatients, found that individuals with at least one prior suicide attempt had significantly higher scores on measures of impulsivity and aggression than individuals without reported suicide attempts. Placidi et al. (220) analyzed data for 272 inpatients with at least one major depressive episode and found significant increases in measures of aggression and impulsivity in those with a history of suicide attempts, compared to those without suicide attempts. Finally, in a study of 44 individuals with a DSM-III-R diagnosis of bipolar disorder, Oquendo et al. (252) found that suicide attempters were more likely to have more lifetime aggression than nonattempters, although lifetime rates of impulsivity were not increased among those with a prior suicide attempt.

Suicide attempters with borderline personality disorder similarly have been reported to exhibit increased levels of aggression and impulsivity. Soloff et al. (197) compared 32 inpatients with borderline personality disorder alone and 77 inpatients with major depressive episode alone to 49 patients with both diagnoses and found that a greater number of suicide attempts was associated with a diagnosis of borderline personality disorder or with increases in either hopelessness or impulsive aggression. Soloff et al. (193) studied the characteristics of 84 patients with borderline personality disorder, of whom 61 had a lifetime history of suicide attempts (72.6%), with an average of 3.39 (SD=2.87) attempts per patient. Those with a history of suicide attempts were found to have had more impulsive actions than patients who had never attempted suicide.

Many individuals with borderline personality disorder and other cluster B personality disorders have a history of suicide attempts, but they may also have a history of self-mutilatory behaviors. However, it is important to recognize that these three characteristics define overlapping but not identical groups of individuals. Stanley et al. (251) compared 30 suicide attempters with a cluster B personality disorder and a history of self-mutilation to a matched group of 23 suicide attempters with a cluster B personality disorder but no history of self-mutilation. Individuals with a history of self-mutilation had higher levels of impulsivity and aggression than those without such a history. Herpertz et al. (253) examined characteristics of self-mutilatory behaviors and found that an ongoing tendency for behavioral dyscontrol was present only in patients exhibiting impulsive self-mutilatory behaviors and not in those with premeditated self-mutilatory behaviors. Thus, although self-mutilatory behaviors and impulsivity share many associated features and antecedents and are common among individuals with borderline personality disorder or histories of physical or sexual abuse, self-mutilatory behavior cannot be regarded as synonymous with impulsivity. In a similar fashion, the presence of other risk-taking behaviors such as reckless driving or unsafe sexual practices is not necessarily a reflection of increased impulsivity per se. Nonetheless, the presence of impulsivity, violence, risk-taking, or self-mutilatory behaviors requires a careful assessment and plan of treatment to address these clinical characteristics and minimize their effect on the risk of suicide and suicide attempts.

a) Alcohol intoxication

Intoxication with alcohol and/or with other substances is often found in individuals who have died by suicide, independent of whether they meet the diagnostic criteria for a substance use disorder. Hayward et al. (255) reviewed coroners' records for 515 consecutive suicides in Western Australia and found that 35.8% of the persons who died by suicide had a nonzero blood alcohol level, with 24.5% being moderately to significantly impaired by alcohol at the time of death. Alcohol consumption before suicide was more prevalent in younger individuals, with 44.8% of teenagers and 35.1% of those age 20–24 years having used alcohol, in contrast to 25.9% of individuals over age 45 years. In addition, those with nonzero blood alcohol levels were more likely to have experienced a breakup in a relationship but less likely to have a history of psychiatric illness or treatment.

As part of the National Suicide Prevention Project in Finland (1987–1988), Ohberg et al. (121) conducted toxicological screening in 1,348 consecutive suicides in a 1-year period and found alcohol use before suicide in 35.9% of the sample. Alcohol was present in men twice as often as in women, whereas prescribed medications, which were found on toxicological screening in 41.6% of suicides overall, were more commonly noted in women.

Brent et al. (151), in a study of suicides in 10- to 19-year-old residents of Allegheny County, Pennsylvania, from 1960 to 1983, found that the proportion of persons who died by suicide with detectable blood alcohol levels rose from 12.9% in 1968–1972 to 46.0% in 1978–1983. In addition, individuals who used a firearm for suicide were 4.9 times more likely to have been drinking than individuals who used other suicide methods.

That intoxication increases the likelihood of suicide is also suggested by the role of employment in modulating suicide risk among alcohol users. Specifically, Pirkola et al. (260) found that alcohol misusers who were employed were more likely to have died by suicide on a weekend than those who were unemployed, suggesting that alcohol use per se contributes to risk, perhaps by increasing impulsivity.

Alcohol intoxication is also a common concomitant of suicide attempts. Borges et al. (715) assessed measures of alcohol intoxication in 40 emergency department patients who had attempted suicide and compared them to 372 patients who presented to the emergency department because of animal bites or workplace or recreational accidents. Patients with suicide attempts were significantly more likely to be under the influence of alcohol, as measured by breath alcohol testing or by self-report of alcohol consumption in the preceding 6 hours. In a study of 325 individuals with deliberate self-poisoning who presented to a Brisbane hospital over a 12-month period, McGrath (453) found that almost one-third had consumed alcohol before their suicide attempt. Varadaraj and Mendonca (454) found similar rates of intoxication in a study of 158 emergency department patients who had attempted suicide by overdose, with 41% consuming alcohol prior to the attempt and 29% having serum alcohol levels above 80 mg/dl.

Individuals who have made a suicide attempt while intoxicated are also at increased risk of later suicide. A study by Suokas and Lonnqvist (258) included data for 1,018 individuals who made a total of 1,207 suicide attempts and were evaluated in the emergency department of a Helsinki, Finland, hospital in a single year. Of these patients, 62% had recently consumed alcohol. Suicide attempts that occurred while intoxicated were more likely to be impulsive. After 5.5 years of follow-up, suicide had occurred in 3.3% of those who had used alcohol with their index suicide attempt. The majority of deaths occurred within the initial year of follow-up, yielding a 51-fold increase in risk of suicide, compared to the general population in the initial year, and a 17-fold increase in risk for the follow-up period as a whole. These findings suggest a need for careful follow-up of intoxicated individuals who present with a suicide attempt. In addition, they suggest a need to determine whether prior suicide attempts occurred in the context of intoxication.

b) Past suicide attempts

A substantial percentage of individuals will die on their initial attempt at suicide. For example, Isometsa and Lonnqvist (266) found that 56% of the 1,397 individuals in the Finnish psychological autopsy study had died with their first suicide attempt and that this pattern was particularly evident in males (62%, compared with 38% of females). In addition, however, individuals with nonfatal suicide attempts have an increased likelihood of later suicide. From a public health standpoint, this finding is particularly important, given the high occurrence of attempted suicide, which in recent decades has had annual rates ranging from 2.6 to 1,100 per 100,000, with lifetime prevalence rates ranging from 720 to 5,930 per 100,000 (2, 518).

Multiple studies have indicated that suicide attempts increase the risk of subsequent suicide. In fact, depending on the length of the follow-up period, from 6% to 10% of those who attempt suicide will ultimately die by suicide. For example, in follow-up studies of patients seen in psychiatric emergency settings after a suicide attempt, 4%–12% die by suicide within 5 years (268, 269, 275). In a 14-year follow-up of 1,018 deliberate self-poisoning patients, Suokas et al. (271) found a 6.7% rate of suicide overall, with the rate in men approximately twice that in women. Among 1,573 individuals who had been hospitalized after attempted suicide and followed up 4–11 years later, Nordstrom et al. (275) found an overall mortality of 11%, with a suicide risk of 6%. Tejedor et al. (272), in a 10-year follow-up of 150 patients admitted to a psychiatric department after a suicide attempt, found an even higher mortality rate from suicide (12%) as well as from natural causes (10%). Furthermore, in a meta-analysis of literature on psychiatric disorders and suicide, Harris and Barraclough (64) found that attempted suicide had a relative risk of later suicide that was greater than that of any psychiatric disorder. Compared to the general population, patients who attempted suicide were at 38 times greater risk of suicide, with the majority of evidence suggesting that this increase in suicide risk is related to the recency of the suicide attempt. This effect has also been noted by Nordstrom et al. (275), who found the greatest risk for suicide during the first year after an attempt. In addition, among the 1,397 individuals in the Finnish psychological autopsy study, Isometsa et al. (266) found that a nonfatal suicide attempt had occurred in 19% of the males and 39% of the females in the year preceding their suicide. In contrast, Fawcett et al. (79) found that among patients with a major affective disorder, a history of previous suicide attempts was associated with suicide that occurred more than 1 year after index hospital admission.

Other factors may also modulate suicide risk following a suicide attempt. Harris and Barraclough (64) found that risk for suicide after an attempt varied with measures of social cohesion and was increased by the presence of long-standing physical illness or a history of multiple previous attempts or prior psychiatric treatment. Risk of suicide following a suicide attempt may also vary with gender, since Nordstrom et al. (275) found a twofold increase in suicide risk in males, compared with females, with the risk for younger male attempters being four times that for younger females.

Other investigators have examined factors associated with subsequent suicide attempts following an index suicide attempt. Hjelmeland (164), for example, studied 1,220 patients who had attempted suicide and compared those who had a repeated attempt within 12 months to those who did not. Although there were no gender differences between repeaters and nonrepeaters, repeaters were more likely to be unmarried, to be unemployed, to abuse alcohol, and to report their own psychiatric problems as their main concern. Repeaters were also more likely to have had a history of sexual abuse, a criminal record, a recent address change, or a relative or friend who had attempted suicide. Others have confirmed an increased risk of repeated suicide attempts in individuals with multiple prior attempts (272, 278). Aborted suicide attempts are also common among those who attempt suicide (276).

Given this increased likelihood of suicide or additional suicide attempts, particularly in the first few years after a suicide attempt, assessment and treatment of suicide attempters should be an integral part of risk reduction. All too often, however, suicide attempters do not receive a psychiatric assessment or follow-up care (2, 279, 280). Thus, in addition to a thorough psychiatric assessment, determining a patient's history of suicide attempts (including aborted suicide attempts) yields information that is important in estimating the level of suicide risk of an individual patient. Additional factors such as psychiatric diagnosis, comorbid alcohol abuse, physical illness, or psychosocial stressors may augment risk following a suicide attempt. Furthermore, the significant mortality observed in suicide attempters underscores the need for careful aftercare planning for suicide attempters.

c) History of childhood physical and/or sexual abuse

Multiple studies have examined the association between childhood abuse and suicidal behaviors, although few have examined the effect of childhood abuse on risk for suicide per se. Plunkett et al. (36), however, assessed 183 young people who had experienced childhood sexual abuse and individuals from a nonabused comparison group 9 years after study intake. Those who had experienced childhood sexual abuse had a suicide rate that was 10.7–13.0 times the national rate, whereas no suicides occurred in the control group.

The bulk of studies have assessed the effects of childhood abuse on suicidal ideation and suicide attempts, both of which are common among individuals reporting childhood abuse. For example, in the study by Plunkett et al. (36), 43% of the 183 young people who had experienced childhood sexual abuse had thought of suicide, whereas 32% had made a suicide attempt.

Other studies have examined the effect of childhood abuse on the risk of suicidal ideation and suicide attempts. Fergusson et al. (286), in an 18-year longitudinal study of a birth cohort of 1,265 New Zealand children, found that those reporting childhood sexual abuse had higher rates of suicidal behaviors than those not reporting such abuse. In addition, the extent of childhood sexual abuse was consistently correlated with risk, with the highest risk of suicidal behaviors in those whose childhood sexual abuse involved intercourse. Even after controlling for the effects of confounding variables, the investigators found that those who reported harsh or abusive childhood experiences were also at increased risk for suicide attempts (282).

Brown et al. (294) followed a cohort of 776 randomly selected children over a 17-year period to adulthood and found that adolescents and young adults with a history of childhood maltreatment were three times more likely to become suicidal than individuals without such a history. Again, the effects of childhood sexual abuse on suicidal behavior were greater than the effects of other forms of abuse, with the risk of repeated suicide attempts being eight times greater for youths with a history of sexual abuse.

Several Australian investigators have used data from twin pairs to assess the effect of childhood abuse on the risk of suicidal ideation or suicide attempts. In structured telephone interviews with 5,995 Australian twins, Dinwiddie et al. (292) found that the 5.9% of women and 2.5% of men who reported a history of childhood sexual abuse were more likely to report suicidal ideation or prior suicide attempts. Nelson et al. (35), using data from 1,991 Australian twin pairs, found even greater rates of childhood sexual abuse (16.7% of women and 5.4% of men) but confirmed that a history of childhood sexual abuse significantly increased risk for suicide attempts, with the greatest risk associated with sexual abuse that involved intercourse. Even in twin pairs who were discordant for childhood sexual abuse, both twins had increased rates for many adverse outcomes, probably as a result of shared family background risk factors. Nonetheless, the twin who reported experiencing childhood sexual abuse had an even greater risk of a subsequent suicide attempt than the co-twin, which suggests an independent contribution of childhood sexual abuse to the risk for suicidal behaviors.

In contrast to the authors of the longitudinal studies described earlier, Romans et al. (291) selected a random community sample of New Zealand women and compared those who reported having been sexually abused as children to those who did not report such abuse. The presence of self-harming behaviors was associated with sexual abuse in childhood and was most marked in individuals who were subjected to more intrusive and frequent abuse.

Cross-sectional assessments of nonpsychiatric populations in the United States have also found associations between suicide attempts and childhood abuse, particularly childhood sexual abuse. Among the 2,918 respondents in the Duke University ECA study, Davidson et al. (293) found that subjects reporting a history of sexual assault also reported higher lifetime rates of suicide attempts than individuals without such a history. In women, a history of sexual trauma before age 16 was a particularly strong correlate of suicide attempts. Among U.S. women physicians (N=4,501), data from a nationally distributed questionnaire showed that the 4.7% of respondents with a history of childhood sexual abuse were more likely to report a history of suicide attempts (287). Kaslow et al. (230) compared 148 low-income African American women who presented to the hospital following a suicide attempt to a similar group of 137 women who presented for general medical care and found a threefold greater risk of childhood maltreatment among suicide attempters. Molnar et al. (33) analyzed data for 5,877 individuals from the National Comorbidity Survey and found that individuals with a history of sexual abuse were more likely to attempt suicide than those without such a history. This risk differed by sex, with a two- to fourfold increase in risk among women and a four- to 11-fold increase in risk among men. Dube et al. (281), in a sample of 17,337 adults (mean age=56 years), also found that the risk of suicide attempts was increased in those who had experienced childhood abuse. They observed this risk to be augmented by multiple other factors, including parental separation or divorce, witnessing of domestic violence, and living with substance abusing, mentally ill, or criminal household members.

In addition to the augmentation of suicide risk associated with sexual abuse, risk appears to be further increased among individuals who have experienced multiple forms of abuse. Anderson et al. (289) examined the association between childhood abuse and adult suicidal behavior in a sample of low-income African American women. Compared to the women who did not report experiencing any emotional, physical, or sexual childhood abuse, those who experienced one, two, or three forms of abuse were, respectively, 1.83, 2.29, or 7.75 times more likely to attempt suicide. In addition, women who reported all three types of abuse were more likely to attempt suicide than women who reported one or two types of abuse.

Childhood abuse is particularly frequent among individuals with psychiatric diagnoses and appears to increase the likelihood of suicide attempts even after the effects of psychiatric comorbidity are controlled. For example, in a study of 251 psychiatric outpatients (68 men and 183 women), Kaplan et al. (284) found that 51% of the subjects had reported experiencing childhood abuse, with 15% reporting sexual abuse alone, 17% reporting physical abuse alone, and 18% reporting a combination of physical and sexual abuse during childhood. Abusive experiences in adulthood were reported by 38% of the subjects, with physical abuse alone in 21%, sexual abuse alone in 8%, and both physical and sexual abuse in 9%. Compared to control subjects without a history of abuse, subjects with a history of abuse were more likely to have been suicidal at a younger age and to have made multiple suicide attempts. Among patients with a history of abuse, suicide attempters could be distinguished from nonattempters on the basis of higher levels of dissociation, depression, and somatization. In analyzing data from the National Comorbidity Survey, Molnar et al. (33) also found that those with a comorbid psychiatric disorder were younger at the time of their first suicide attempt than those without concomitant psychiatric illness. In addition, a history of childhood sexual abuse remained a risk factor for attempting suicide even after adjustment for the effect of a lifetime psychiatric diagnosis.

The presence of childhood physical and/or sexual abuse has also been associated with an increased likelihood of suicidality in studies of patients with specific psychiatric diagnoses. For example, Brodsky et al. (196), in a study of 214 inpatients with a diagnosis of borderline personality disorder, found that the number of lifetime suicide attempts was correlated with a history of childhood abuse. Van der Kolk et al. (285) assessed 74 individuals with personality disorders or bipolar II disorder and found that histories of childhood sexual and physical abuse were highly significant predictors of self-cutting and suicide attempts. During a follow-up period that averaged 4 years, the patients who continued being self-destructive were those with the most severe histories of separation and neglect and those with past sexual abuse.

For individuals with major depressive disorder, evidence in the literature is more complex. Brodsky et al. (250) found that adults with major depressive disorder who had a history of childhood physical or sexual abuse were more likely to have made a suicide attempt than those who did not report an abuse history, even after adjustment for the effects of impulsivity, aggression history, and presence of borderline personality disorder. Zlotnick et al. (716), in a study of 235 outpatients with major depression, found substantial rates of diagnostic comorbidity, primarily with borderline personality disorder and PTSD. After controlling for the effects of the presence of these diagnoses, however, they did not find an independent contribution of childhood sexual abuse to the likelihood of suicide attempts.

Childhood abuse is also prevalent among individuals with substance use disorders and, again, is associated with increased rates of suicide attempts. In a group of 481 male and 321 female alcoholic inpatients (age 19–57 years), Windle et al. (288) found a high prevalence of reported childhood abuse. For women, the rates of physical abuse only, sexual abuse only, and dual abuse were 10%, 26%, and 23%, respectively, whereas for men the corresponding rates were 19%, 7%, and 5%, respectively. For both sexes, a reported history of childhood abuse was associated with a higher rate of suicide attempts, with an even larger effect associated with a history of both physical and sexual abuse. Roy (290) examined abuse histories in a consecutive series of 100 male cocaine-dependent patients and found that the 34 patients who had attempted suicide reported significantly higher scores for childhood emotional abuse, physical abuse, sexual abuse, and emotional and physical neglect than the 66 patients who had never made a suicide attempt.

In addition to increasing risk for suicide attempts within community samples and across subgroups of psychiatric patients, the presence of a childhood abuse history in individuals who have made a suicide attempt should alert the psychiatrist to a further increase in the risk of repeated attempts (284). Elliott et al. (717) compared 65 patients hospitalized for a medically serious suicide attempt to 32 patients seen in the emergency room for a suicide attempt but who were not medically hospitalized. Those with attempts that were not medically serious had higher rates of previous sexual and physical abuse as well as higher rates of traumatic life events and borderline personality disorder. Hjelmeland (164) also found that patients in a Norwegian county who required medical treatment after an initial suicide attempt were more likely to have a repeated suicide attempt during 6 years of follow-up if they had a history of being sexually abused.

In summary, there is consistent evidence, in multiple samples studied with multiple study designs, that a history of abuse augments the risk for later suicidal ideation and suicide attempts. The effect of abuse on suicide per se has been less well studied, but the few findings that are available suggest that abuse increases suicide risk. Childhood abuse and particularly childhood sexual abuse appear to be associated with greater increases in risk than childhood physical abuse or abuse during adulthood. Individuals who have experienced multiple forms of abuse are at particularly increased risk of suicidal ideation and behaviors. Although a history of abuse is common in individuals with suicide attempts and in individuals with a psychiatric diagnosis, the contribution of childhood abuse to the risk of suicidal behaviors seems to be independent of the effects of psychiatric diagnoses. Consequently, in patients who have attempted suicide as well as in those presenting for any type of psychiatric treatment, it is important to inquire about childhood and adult experiences of physical, sexual, or emotional abuse and to incorporate this information into the risk assessment and treatment planning process.

d) History of domestic partner violence

Although studies have not directly assessed the effects of domestic partner violence on risk for suicide, domestic partner violence has been associated with increased rates of suicide attempts and suicidal ideation. For example, after adjustment for the effects of sociodemographic characteristics and alcohol use in a nationally representative sample of 5,238 U.S. adults, Simon et al. (298) found that being physically assaulted was associated with suicidal ideation or behavior (odds ratio=2.7) and that this pattern was particularly true for individuals who sustained injury (odds ratio=3.4) or were assaulted by a relative or intimate partner (odds ratio=7.7). McCauley et al. (300) surveyed 1,952 respondents in a primary medical care practice and found that 5.5% had experienced domestic violence in the year before presentation. Compared with women who had not recently experienced domestic violence, those with recent experiences of such violence were four times more likely to have attempted suicide. Among women physicians (N=4,501 respondents) who responded to the Women Physicians' Health Study questionnaire, suicide attempts were significantly more prevalent among the 3.7% of respondents with a history of domestic partner violence (34).

Domestic partner violence is particularly a risk factor for suicide attempts among women in low-income urban environments. In a group of 648 women, most of whom were young and unemployed and had an annual household income of less than $10,000, Abbott et al. (299) found that among the 418 women with a current male partner, 11.7% reported being recently assaulted, threatened, or intimidated by their partner. For the entire sample, the cumulative lifetime prevalence of exposure to domestic violence was 54.2%. Women with any exposure to domestic partner violence had an increased rate of suicide attempts, compared to women without such exposure (26% and 8%, respectively). Kaslow et al. (230) compared 148 low-income African American women who presented to the hospital after a suicide attempt to a similar group of 137 women who presented for general medical care. Women who presented with a suicide attempt had a greater likelihood of having experienced either physical or nonphysical partner abuse (odds ratios=2.5 and 2.8, respectively). Thompson et al. (301), in a sample of low-income, inner-city women, found that suicide attempters (N=119) were approximately three times more likely to experience significant physical partner abuse, nonphysical abuse, and PTSD than nonattempters (N=85). In addition, increased suicidality in individuals who were experiencing physical partner abuse appeared to depend on the presence of PTSD rather than the independent contribution to risk of the abuse.

Increased risk for suicide attempts is also seen in battered women presenting to emergency department settings or to women's shelters. Muelleman et al. (296) surveyed 4,501 women between age 19 and 65 years who presented to 10 hospital-based emergency departments in two cities serving inner city, urban, and suburban populations. Of these, 266 (5.9%) had definite or probable battering injuries and an additional 266 (5.9%) were currently in a physically abusive relationship but did not present with evidence of a battering injury. Compared to the 3,969 women (88.2%) who were not currently in a physically abusive relationship, women in physically abusive relationships were more likely to present to the emergency department after an attempted suicide. In a cross-sectional study of 203 women seeking refuge in battered women's shelters, Wingood et al. (297) found, after controlling for the effects of sociodemographic characteristics, that women experiencing both sexual and physical abuse were more likely to have attempted suicide than women experiencing physical abuse alone.

Although much more commonly experienced by women, domestic partner violence also affects men. Ernst et al. (302) surveyed 233 men and 283 women who presented to an inner-city emergency department for past and current histories of domestic partner violence and found that such experiences were associated with increased rates of suicidal ideation in both sexes. It is important to note that men with a history of domestic violence toward their partners may also be at increased risk for suicide. Conner et al. (303), for example, noted that one-half of the 42 male alcoholics who died by suicide and were originally described by Murphy et al. (149) had a history of domestic violence.

Domestic violence in the home may also affect the risk for suicide attempts among those who witness that violence. Dube et al. (281) examined the relationship to suicide attempts of eight adverse childhood experiences, including witnessing domestic violence, in 17,337 adults (mean age=56 years) and found that childhood exposure to parental domestic violence increased the risk of later suicide attempts.

Thus, although data on suicide risk per se are not available, there is clear evidence that domestic partner violence is associated with an increased risk of suicide attempts. In addition, although evidence is more limited, individuals who become violent with their partners or who observe domestic partner violence may also be at increased risk for suicidal behaviors. Since a past or current history of domestic partner violence is often overlooked, even in settings such as emergency departments where it is quite prevalent, it is important to specifically ask about domestic partner violence as a part of the suicide assessment.

e) Treatment history

A past history of treatment of mental illness, including a past history of hospitalization, should be viewed as a marker that alerts the clinician to an increase in suicide risk (64, 198). Furthermore, greater treatment intensity is associated with a higher rate of eventual suicide. For example, Bostwick and Pankratz (77) used meta-analytic techniques to calculate suicide risks for outpatients, inpatients, or suicidal inpatients and found a hierarchy in suicide risk among patients with affective disorders. The estimated lifetime prevalence of suicide in those ever hospitalized for suicidality was 8.6%, compared to a lifetime suicide rate of 4.0% for all hospitalized patients. For mixed inpatient/outpatient populations, the lifetime suicide prevalence was even lower, at 2.2%, whereas for the population without affective illness, it was less than 0.5%. A similar phenomenon was noted by Simon and VonKorff (718) among patients treated for depression in a large health plan in western Washington State. Computerized discharge diagnoses, outpatient visit diagnoses, and outpatient prescription records were used to identify all enrollees who received treatment for depression during a 3-year period. During the study period, 35,546 individuals received some treatment for depression and accounted for 62,159 person-years of follow-up. Thirty-six individuals (4.2% of all deaths) were classified as having definitely or possibly died by suicide, yielding an overall suicide mortality rate of 59 per 100,000 person-years, with the rate for men more than threefold higher than the rate for women. Patients who received any inpatient psychiatric treatment had a risk for suicide of 224 per 100,000 person-years, with suicide rates among those who received outpatient specialty mental health treatment and those treated with antidepressant medications in primary care of 64 and 43 per 100,000 person-years, respectively. No patient with a diagnosis of depression who was treated only in primary care and who did not receive antidepressant medication died by suicide.

In terms of suicide attempts, Pirkis et al. (198) analyzed data for 12,229 patients in 13,632 episodes of care and found that the risk of suicide attempts was 10-fold higher in acute inpatient settings, compared with longer-stay inpatient or community-based settings (5.4 attempts per 1,000 episode-days, compared with 0.6 and 0.5 attempts per 1,000 episode-days, respectively). Thus, the rate of suicidality is increased in individuals with prior inpatient treatment, although it is not clear whether the rate is higher because the patients have more severe illnesses (and are deemed to be at increased risk for suicide) or because hospitalization increases suicide risk by increasing emotional or psychosocial distress.

Temporally, the risk for suicide appears to be greatest after changes in treatment setting or intensity (304). Recently admitted and recently discharged inpatients show particularly increased risks (64, 72), and this pattern is seen across diagnostic categories (91, 95, 305–308). Rates decline with time since discharge but may remain high for as long as several years (91, 306, 309). For example, Pirkis and Burgess (309) systematically reviewed the literature on suicide and health care contacts and found that up to 41% of those in the general population who die by suicide may have had psychiatric inpatient care in the year before death, with up to 9% dying by suicide within 1 day of discharge. Appleby et al. (304) compared individuals who died by suicide within 5 years of discharge from psychiatric inpatient care to surviving demographically matched patients and found that those who died by suicide were more likely to have had their care reduced at the final appointment in the community before death (odds ratio=3.7).

Black et al. (91) assessed 5,412 patients admitted to the University of Iowa Psychiatric Hospital and found that 331 died over a 9-year follow-up period. Ninety-nine percent of all premature deaths occurred during the initial 2 years after discharge, with the risk for premature death being greatest among women and the young. Over the initial 2-year period, 29% of deaths were by suicide and suicide occurred at a rate that was more than 50 times the expected rate for the group as a whole.

Roy (96) compared 90 psychiatric patients who had attempted suicide (53 male patients and 37 female patients) to a group of 90 matched control subjects who had not attempted suicide. Of the 75 patients who had died by suicide as outpatients, 58% had seen a psychiatrist within the previous week, 81% had been admitted in their last episode of contact, and 44% of those who had been inpatients attempted suicide within 1 month of discharge.

Goldacre et al. (307) determined the risk of suicide within a year of psychiatric discharge in a population-based study in Oxford, U.K., and found that SMRs for suicide in the first 28 days after discharge from inpatient care were 213 and 134 for male and female patients, respectively. The rate of suicide in the first 28 days after discharge was 7.1 times higher for male patients and 3.0 times higher for female patients than the rate during the remaining 48 weeks of the first year after discharge.

More recently, Appleby et al. (306) analyzed data for 10,040 individuals in the United Kingdom who died by suicide over a 2-year period and found that 2,370 (24%) had been in contact with mental health services in the 12 months before death. Of these, 358 (16%) were psychiatric inpatients at the time of death, and one-fifth of those patients were being monitored with special observation procedures. An additional 519 suicides (24%) occurred within 3 months of hospital discharge, with the highest number occurring in the first week. Rossau and Mortensen (95) found that 508 suicides occurred among 9,156 patients who were admitted to psychiatric hospitals in Denmark between 1970 and 1987 and who received a diagnosis of schizophrenia for the first time. Suicide risk was particularly high during the first 5 days after discharge, and risk was also increased in individuals with multiple admissions during the prior year.

Similar findings have been reported for suicide attempts, which are also more frequent in the period following hospitalization. Oquendo et al. (267) followed 136 patients after hospitalization for major depressive disorder and found that 15% of the subjects made a suicide attempt within 2 years, with more than 50% of attempts occurring within the first 5 months of follow-up.

Given that the intensity of past treatment is associated with risk for suicide and suicide attempts, the treatment history is an important part of the assessment process. In addition, these observations suggest specific points in the course of treatment (e.g., hospital discharge or other changes in treatment setting) at which risk of suicidal behaviors may be particularly increased. Awareness of these factors will allow the psychiatrist to take them into consideration in developing a plan of treatment with the patient.

f) Illness course and severity

In some psychiatric disorders, suicide risk is greater at certain points in the illness or episode course. Multiple studies have shown that suicidality tends to occur early in the course of affective disorder, often before diagnosis or before treatment has begun (310, 719). These observations emphasize the importance of early identification of these disorders and early implementation of effective interventions.

Appleby et al. (304) compared individuals who died by suicide within 5 years of discharge from psychiatric inpatient care to surviving demographically matched patients and found that suicide was more likely in those whose index hospitalization was at the beginning of their illness (odds ratio=2.0). Bradvik and Berglund (317) followed 1,206 inpatients who had received a discharge diagnosis of severe depression/melancholia between 1956 and 1969. At the time of the initial follow-up in 1984, 22% had died by suicide, whereas by the second follow-up in 1998, an additional 4% had died by suicide. Although mortality due to suicide declined with time, the standardized mortality was still increased late in the course of depressive illness (SMR=1.3). Osby et al. (73) identified all patients in Sweden with a hospital diagnosis of bipolar disorder (N=15,386) or unipolar depressive disorder (N=39,182) between 1973 and 1995 and determined the date and cause of death using national registries. They found that the SMR for suicide was especially high for younger patients during the first years after initial diagnosis, although an increasing SMR was found for female patients with major depressive disorder over the course of the study. Fawcett et al. (313) found that for the 954 patients with major affective disorder in the NIMH Collaborative Program on the Psychobiology of Depression, 32% of the 25 suicides occurred within 6 months and 52% occurred within 1 year of entry into the study.

Suicide has been noted to be more likely early in the illness course in individuals with schizophrenia. Westermeyer et al. (83), for example, compared 36 patients with schizophrenia who died by suicide to a similar group of patients who did not die by suicide and found that individuals with schizophrenia and other psychotic disorders were especially vulnerable to suicide within the first 6 years of their initial hospitalization. Suicides were present throughout the course of schizophrenia in the National Suicide Prevention Project in Finland (100).

Although patients' risks for suicide and suicide attempts later in the illness course are less than those earlier on, their risks remain greater than those in the general population (74, 100, 316, 317). Angst et al. (74) followed 406 hospitalized patients with affective disorder for 22 years or more and found that the suicide rate was most elevated at the age of onset but that, from age 30 to 70 years, the suicide rate was remarkably constant despite the different courses of illness. Ahrens et al. (316) examined the illness course of 310 patients with mood disorder, 98 of whom had made a suicide attempt, and found no significant correlation between age and suicide attempts, suggesting that the rate of suicide attempts was not declining as the patient aged and the illness progressed. Malone et al. (312) analyzed data for 100 inpatients during a major depressive episode and noted that the first 3 months after the onset of a major depressive episode and the first 5 years after the lifetime onset of major depressive disorder represented the highest-risk period for attempted suicide, independent of the severity or duration of depression.

Risk for suicide may also vary with the severity of symptoms. For example, Brown et al. (78) prospectively followed 6,891 psychiatric outpatients and found that in the 49 (1%) who died by suicide the severity of depression, hopelessness, and suicide ideation were significant risk factors. In contrast, in individuals with schizophrenia, low levels of negative symptoms have been associated with increased suicide risk (93).

It is also important to recognize that other factors such as age modulate the effects of symptom severity on risk. With older adults, for example, milder symptoms may be associated with greater risk than moderate symptoms in younger adults (66). Waern et al. (207) analyzed data for 85 individuals over 65 years of age who had died by suicide and 153 randomly selected living comparison subjects and found that elevated suicide risk was associated with minor as well as major depressive disorder.

In terms of attempted suicide, Mann et al. (31) followed 347 consecutive patients after admission to a university hospital and found that the objective severity of current depression or psychosis did not distinguish the 184 patients who had attempted suicide from those who had never attempted suicide. However, those who had attempted suicide reported higher levels of subjective depression and suicidal ideation and fewer reasons for living. Corbitt et al. (195), using data from structured interviews of 102 psychiatric inpatients, found that past suicidal behavior was better predicted by the number of criteria for borderline personality disorder and other cluster B personality disorders that were met than by depressive symptoms.

That the risk for suicidality may be associated with symptom or illness severity suggests that it is important to determine the magnitude and not simply the presence of risk factors as part of the assessment process. In addition, when estimating risk and implementing a plan of treatment to address risk, variations in risk with illness course may need to be considered in the context of other patient-specific factors.

Considerable evidence derived from a variety of sources supports a link between physical illness and suicide (Table 7). Methods used to establish this relationship have included record linkage and prospective cohort studies of clinical samples with specific physical illnesses, as well as retrospective examinations of the prevalence of specific physical illnesses in samples of individuals who took their own lives.

Harris and Barraclough (25) conducted a comprehensive, systematic literature review and meta-analysis to determine the suicide risk associated with 63 specific physical disorders that had been posited to influence this risk. They did not include reports of epilepsy, conceding that the evidence base was already strong for its association with suicide. Their review yielded 235 reports that met the specific criteria for consideration, from which data were abstracted to enable calculation of pooled SMRs for each condition. The authors concluded that the disorders demonstrating significantly increased risk for suicide included HIV/AIDS, Huntington's disease, malignant neoplasm, multiple sclerosis, peptic ulcer disease, chronic hemodialysis-treated renal failure, spinal cord injury, and systemic lupus erythematosus. The data were insufficient to conclude whether amputation, valve replacement, intestinal diseases, cirrhosis, Parkinson's disease, or systemic sclerosis conferred an increased risk for suicide. Surprising findings included the association of amyotrophic lateral sclerosis, blindness, stroke, diabetes mellitus, rheumatoid arthritis, and hypertension with only average risk that was no greater than that in the general population. Pregnancy and the puerperium were associated with a statistically reduced risk for suicide. Although the authors were unable to examine the influence of mental disorders or other mediating or moderating effects, many of the disorders associated with increased risk are also associated with mental disorders (e.g., multiple sclerosis with depression and peptic ulcer disease with alcohol abuse). Therefore, when the influence of comorbid psychiatric illness is accounted for, the independent risk associated with physical illness may be less.

Quan and colleagues (337) reported results of a record linkage study conducted in Alberta, Canada, that was designed to establish whether specific illnesses distinguish persons who died by suicide from comparison subjects who died in accidents. In univariate statistical analyses comparing 822 persons age 55 years and older who died by suicide with 944 subjects of similar age who died in motor vehicle accidents, those who died by suicide were more likely to have had malignant neoplasm, arteriosclerotic heart disease, chronic obstructive pulmonary disease, peptic ulcer disease, prostate disorders, depression, and other psychiatric diagnoses. In multivariate analyses in which the effects of demographic and health characteristics were controlled, arteriosclerotic heart disease and peptic ulcer disease did not differentiate the groups. Among the physical illnesses, only malignant neoplasm and prostate disorders (excluding prostate cancer) remained significant predictors, along with chronic obstructive pulmonary disease in married (but not single) men.

Grabbe and colleagues (338) used data from the National Mortality Followback Survey to identify health status variables related to suicide in older persons, compared with natural deaths and deaths from injury. In comparing suicides with injury deaths, they found that malignant neoplasm, but not lung conditions, was associated with increased risk. The presence of a stroke, paradoxically, appeared to lower risk. The study reinforced the powerful influence of cancer.

Other studies have also specifically looked for associations between suicide and central nervous system disorders that are known to increase the rates of depressive syndromes. Stenager et al. (334), for example, found that patients with a diagnosis of multiple sclerosis after age 40 were at no greater risk for suicide than control subjects, but that men and women who received this diagnosis before age 40 were at approximately three times and two times greater risk, respectively. Stenager et al. (720) also cross-referenced data for all patients discharged with a diagnosis of a cerebrovascular accident in selected areas of Denmark between 1973 and 1990 with death records and found 140 suicides among almost 38,000 patients with a history of cerebrovascular accident and increased suicide risks for both men and women with this diagnosis in all age groups. In general, risk for suicide was higher in women than in men and in age groups under age 60 years, compared to older adults. The group at highest risk was women under age 50 with a stroke, who had a risk for suicide almost 14 times greater than that for women of similar age in the general population. These data provide additional support for an association between suicide and cerebrovascular disease, particularly among younger and middle-aged stroke patients.

Evidence for increased suicide risk in people with epilepsy is similarly strong. Stenager and Stenager (323) examined all published reports concerning the link between suicide and neurologic disorders in order to critically evaluate the strength of the evidence. They identified a variety of common methodologic problems in this body of research, including sources of bias in selection of cases, inadequate definition of control samples, imprecise definitions of disease, inadequate sample sizes, absent or imprecise definitions of suicidal behavior, and inadequate follow-up intervals. Nonetheless, they concluded that sufficiently rigorous studies of patients with multiple sclerosis, patients with spinal cord injury, and selected groups of patients with epilepsy did establish increased risk in these conditions. The most rigorous studies examining risk associated with epilepsy were conducted by White and colleagues in 1979 (324). They followed 2,099 patients with epilepsy who had been committed for institutional care and treatment and compared their risk for suicide with that in an age- and sex-standardized control population. They found that individuals with epilepsy were at 5.4 times higher risk for suicide than the control subjects.

Rafnsson et al. (326) analyzed data for 224 individuals who first received a diagnosis of unprovoked seizures in Iceland between 1960 and 1964 and who were followed for up to 25 years. Among men, the relative risk of dying by suicide was almost six times the expected risk in the general population, and the rates of death from accidents, poisoning, and violence were about three times the expected rates.

Nilsson et al. (325) used data from the Swedish National Cause of Death Register to determine causes of death among 6,880 patients with a diagnosis of epilepsy registered in the Stockholm County In-Patient Register. In a comparison of 26 individuals who died by suicide and 23 individuals suspected of having died by suicide with 171 living control subjects, individuals with an onset of epilepsy before age 18 had a higher risk for suicide than those with comorbid psychiatric diagnoses or those treated with antipsychotics. However, unlike other studies, this study did not find a specific association with particular types of epilepsy, including temporal lobe seizures.

Suicide attempts also appear to be increased in frequency among patients with epilepsy, compared to the general population. Hawton et al. (330) analyzed data for patients admitted after deliberate self-poisoning or self-injury over a 2-year period and found that the number of patients with epilepsy was five times higher than general population prevalence rates. Mendez et al. (329) compared 175 outpatients with epilepsy to a group of 70 comparably disabled outpatients and found that prior suicide attempts were reported by 30% of the patients with epilepsy, compared to only 7% of the control subjects. Rates of depression were similarly increased among the patients with epilepsy (55%, compared with 30% of the control subjects). In a subsequent study, Mendez et al. (331) compared 62 patients with epilepsy to 62 patients with schizophrenia and to 62 patients with both diagnoses and found that suicidal behaviors were more common among individuals with epilepsy.

The association between terminal physical illnesses and suicide is complex. Brown and colleagues (721) found that 34 of 44 terminally ill patients receiving palliative care had never wished for an early death. All of the 10 patients who had wished for an early death were found to have clinical depressive illness, but only three reported suicidal ideation. Chochinov and colleagues (342) interviewed 200 patients who had terminal cancer to determine their psychiatric status and whether they had thoughts of death. Almost 45% had wished for an early death, but in only 8.5% were the thoughts serious and persistent. Predictors of desire for death included pain, a low level of family support, and clinically significant depression. Diagnosable depressive illness was found in almost 60% of those with a desire to die and in 8% of those without a desire to die.

Other features of physical illness that may augment the likelihood of suicidal ideation or suicide include functional impairments (338), pain (340, 341), disfigurement, increased dependence on others, and decreases in sight and hearing (321, 333). Waern et al. (333) compared consecutive records of people who had died by suicide (46 men and 39 women) with those of living control participants selected from the tax register (84 men and 69 women) in Gothenburg, Sweden. In addition to neurological disorders and malignant disease, which were associated with three- to fourfold increases in suicide risk, visual impairment and serious physical illness of any type were also associated with increased risk, with odds ratios of 7.0 and 6.4, respectively. Although the number of women in the sample was small, the risk appeared to be greater among men, particularly in those with a high burden of physical illness. Conwell et al. (321) also found physical illness burden and functional limitations to be more common among individuals seen in primary care settings who die by suicide. They compared 196 patients age 60 years and older from a group practice of general internal medicine (N=115) or family medicine (N=81) to 42 individuals age 60 years and older who had visited a primary care provider and who died by suicide within 30 days of their visit. Those who died by suicide were significantly more likely than control subjects to have had a depressive illness, greater functional impairment, or a larger burden of physical illness.

While several studies have shown that people with HIV and AIDS are at high risk for suicide, the data on the extent of that risk vary. In particular, suicide risk among people with HIV/AIDS is likely to relate to other comorbid factors such as substance abuse and other psychiatric diagnoses, stigma, social isolation, and lack of support (722), as well as the direct effects of HIV on the brain (335, 723). Even at the time of HIV serum antibody testing, suicidal ideation is highly prevalent, being noted by about 30% of individuals and diminishing over time after notification of test results (724). Nonetheless, elevations in the suicide rate are present among persons with AIDS and range from seven to 36 times the rates in comparable age- and sex-matched populations (335, 336, 725). For example, Marzuk et al. (335) studied suicide rates in 1985 in New York City and found a rate of 18.75 per 100,000 person-years for men age 20–59 years without a known diagnosis of AIDS, compared to 680.56 per 100,000 person-years for those with a known diagnosis of AIDS, a 36-fold increase in relative risk. Cote et al. (336) used public-access AIDS surveillance data and National Center for Health Statistics multiple-cause mortality data for the period from 1987 through 1989 to identify suicides among persons with AIDS and found that all but one of the persons who died by suicide were male. Compared to demographically similar men in the general population, men with AIDS had a rate of suicide that was 7.4-fold higher, at 165 per 100,000 person-years of observation. Cote et al. also noted that the suicide risk for persons with AIDS decreased significantly from 1987 to 1989, suggesting that the rate of suicide associated with AIDS may be decreasing.

In summary, physical illnesses are associated with increased risk for suicide. The strength of the evidence for malignant neoplasms, central nervous system disorders, peptic ulcer disease, and HIV/AIDS is strong. Although the evidence is less compelling, indications are that a range of other conditions may also be associated with suicide and suicidal behaviors. It is probable that mood and substance use disorders, either as precipitants or sequelae, account in part for the increased SMRs for suicide ascribed to specific physical conditions in the literature. However, further study is needed to determine the role of social and psychological factors as mediators or moderators of the relationship between physical illness and suicide. As a result, in assessing suicide risk among individuals with physical illness, consideration should be given to the presence of comorbid mood symptoms as well as to the functional effects of the illness.

Findings from at least three types of studies suggests that risk for suicide has a familial and probably genetic contribution. These include: 1) strong and consistent findings that risk for suicidal behavior is much higher among first-degree relatives of individuals with suicide attempts or deaths than in the general population, 2) higher concordance for suicidal behavior among identical versus fraternal twins, and 3) greater risk of suicidal behavior among biological versus adoptive relatives of persons adopted early in life who later died by suicide (365–368). These familial associations appear to be accounted for only partly by familial risks for major affective illness or other clinical risk factors for suicide. Recent efforts to specify molecular genetic markers that segregate or associate with suicidal behavior, including those relating to the serotonin (5-hydroxytryptamine [5-HT]) neurotransmission system, have yielded inconsistent findings that are not easily interpreted (17, 366, 368).

a) Family studies

Evidence from family studies of suicide was recently summarized by Turecki (368). In addition to studies showing increased rates of suicidal behaviors among family members of suicidal individuals (31, 82, 360–364), at least 20 reports of controlled comparisons involving more than 11,000 subjects have been published (202, 214, 312, 344–359). The pooled overall relative risk of suicidal behavior in first-degree relatives of suicidal probands compared to control or population risks, weighted by the number of subjects in each study, was 4.48 (95% CI=3.71–5.25), indicating a nearly 4.5-fold excess of risk of suicidal behavior among relatives of suicidal subjects, compared to nonsuicidal subjects (R. Baldessarini, personal communication, 2002).

Across studies, reported estimates of relative risk for suicidal behavior within families vary greatly, depending in part on the types of behavior included (suicide, suicide attempts of varying lethality, or both) and their defining criteria, the prevalence of psychiatric risk factors for suicide among the control subjects, the closeness of kinship (first-degree relatives, including parents and siblings, with or without second-degree relatives), and differences in sample size. Such studies, while demonstrating a powerful association, do not prove genetic risk nor rule out shared environmental factors. Moreover, it remains to be proved that the relationship for suicide is separable from the well-known heritability of leading risk factors for suicide, including major affective illness. Nevertheless, the findings from pooled family studies strongly support the conclusion that overall risk for suicidal behavior is at least four times greater among close relatives of suicidal persons than among unrelated persons.

b) Twin studies

A powerful method of separating risks that result from shared environments from risks that result from genetic factors is to compare the rate of concordance (index condition appearing in both twins) for a condition between identical, or monozygotic (single-egg), twins and fraternal, or dizygotic (two-egg), twins. Risks for dizygotic twins should be similar to those found among other first-degree family members in family studies. Seven such twin studies pertaining to suicide were identified in the research literature (365, 370–375) and reviewed by Roy et al. (365, 369) and Turecki (368). None of the studies involved samples of twins raised separately from early life, and, thus, the confounding effects of shared environments were possible. Moreover, the size and statistical power of these studies varied markedly, from an analysis of a single monozygotic twin-pair (373) to a study of an entire Australian national twin registry involving more than 1,500 monozygotic and nearly 1,200 dizygotic twin-pairs (375). When the data from all seven twin studies were pooled, the overall concordance rate for suicide or suicide attempts, weighted for the numbers of subjects involved, was 23.5% (401 of 1,704) for monozygotic twin-pairs and 0.135% (two of 1,486) for dizygotic twin-pairs, for a highly significant 175-fold increase in pooled relative risk in the monozygotic twin-pairs (R. Baldessarini, personal communication, 2002). Given the low frequency of suicidal behavior found among fraternal co-twins, this relative risk is likely to be a quantitatively unstable estimate. Nevertheless, its magnitude strongly supports a genetic contribution to suicidal behavior. A highly significant fourfold excess of risk in identical twins remained, even after statistical corrections for depressive and other psychiatric morbidity associated with suicide (375). Therefore, twin studies add strong support for the heritability of suicide risk that is separate from the heritability of risk factors such as mood disorders but that is still likely to be influenced by environmental factors.

c) Adoption studies

A less commonly employed technique to separate genetic from shared environmental factors is to study outcomes for persons adopted from their biological families very early in life. For the study of suicide, this approach has been reported only three times, and each study used the same Danish health and vital statistics registers that included data for 5,483 adoptions in greater Copenhagen between 1924 and 1947 (376–378). When data for suicide were pooled across all studies, to include affectively as well as psychotically ill probands (376, 378), there was an approximately fivefold greater risk among biological than among adoptive relatives (20 of 543 subjects [3.68%] versus two of 263 subjects [0.76%]).

Later, the same American and Danish collaborators (378) compared all adoptees identified as having an affective spectrum disorder (N=71) with matched control adoptees without such disorders (N=71). The index disorders included not only DSM-III major depression and bipolar disorder but also milder "neurotic" depressions and a condition ("affect reaction") marked by affective instability that may resemble some forms of personality disorder in current classifications. In relatives of affectively ill adopted probands, there was a significant, approximately sevenfold greater risk for suicide in biological relatives, compared with adopted relatives (15 of 387 subjects [3.88%] versus one of 180 subjects [0.56%]). Further analysis of the suicide rate for biological relatives, compared with control subjects, also yielded a highly significant 13.3-fold difference (15 of 387 subjects [3.88%] versus one of 344 subjects [0.29%]) (378). In striking contrast, however, when suicide attempts were considered separately, there was a 1.16-fold lower but nonsignificant risk in the biological relatives, compared with the adoptive relatives, of affectively ill adopted probands (13 of 387 subjects [3.36%] versus seven of 180 subjects [3.89%]). A similar comparison of the rate of suicide attempts in biological relatives of adopted probands and in matched but not affectively ill control subjects showed a modest 2.89-fold difference that failed to reach significance (13 of 387 subjects [3.36%] versus four of 344 subjects [1.16%]). Among relatives of index adoptees with a diagnosis of schizophrenia, there was a nonsignificant 2.67-fold greater risk for suicide in biological relatives, compared with adoptive relatives (five of 156 subjects [3.20%] versus one of 83 subjects [1.20%]) (376).

Matched comparison of 57 early-adopted individuals who died by suicide with other adoptees lacking evidence of suicide or psychiatric illnesses also showed a great excess risk of suicide in biological over adoptive relatives (12 of 269 subjects [4.46%] versus none of 148 subjects [0.00%]) (378). Risk of suicide was approximately sixfold greater in relatives of suicidal probands compared to relatives of matched, nonsuicidal control subjects (12 of 269 subjects [4.46%] versus two of 269 subjects [0.74%]). However, this study did not consider the possible coincident heritability of clinical risk factors for suicide, such as major affective illnesses and substance use disorders.

Overall, these adoption studies indicate a greater risk of suicide, but not of suicide attempts, among biological relatives of suicidal probands, compared with adoptive relatives. They also show greater risk among biological relatives of probands, compared with control subjects, that is consistent with the hypothesis that suicidality is heritable. Given the broader range of severity and lethality of suicide attempts and the greater likelihood of environmentally determined actions in many instances, the heritability of suicide may well be much greater than that of suicide attempts.

a) Employment

Unemployment has long been associated with increased rates of suicide (379, 380, 726). Furthermore, the link between suicide and unemployment has been confirmed by a recent study that used U.S. National Longitudinal Mortality Study data to assess whether unemployed individuals were at greater risk for suicide than employed persons (24). At 2-year follow-up, unemployed men were two to three times more likely to have died by suicide, compared with employed men. Living alone, being divorced, and having lower socioeconomic status increased the suicide risk. At or beyond 4 years of follow-up, however, there was no statistical association between unemployment and suicide for men. For women, the relationship between suicide and unemployment was even stronger and longer-lasting. Unemployed women had a much higher risk for suicide at each year of follow-up than employed women. Unemployed women continued to show an elevated risk at 9-year follow-up, by which time they were three times as likely to die by suicide as employed women. As with men, younger unemployed women were more at risk than women over age 45 years. While the number of women who died by suicide was small, the results remain significant and powerful. While in the past men were considered most at risk for suicide after becoming unemployed, it is now known that women are at an even greater risk and for a longer period of time. The relationship between unemployment, suicide, and psychiatric disorders remains unclear. Persons with psychiatric disorders may be more likely to quit jobs or to be fired as well as more likely to die by suicide (727).

Areas with socioeconomic deprivation also have larger numbers of unemployed people, and these differences have been used to examine effects of unemployment on rates of suicide and suicide attempts. Hawton et al. (386), for example, analyzed data for different wards, or communities, within Oxford, England, and found that wards with the highest socioeconomic deprivation were associated with the highest rates of suicide attempts. Individuals who attempted suicide, both men and women, were more likely to be unemployed, living alone, and having problems with housing. For men, but not for women, a strong association was also found between the rate of suicide attempts and socioeconomic deprivation. Men living in less deprived areas who had financial problems were even more likely to attempt suicide, suggesting that the dissonance between one's own financial status and that of the neighborhood may affect risk. Unemployment and financial problems can affect suicide in other ways as well. Alcohol consumption and marital conflict, each of which increases with financial difficulties or unemployment, may also contribute to increased risk for suicidal behaviors.

Political context and large-scale economic changes can also influence suicide and may provide clues about the effect of employment status on suicide rates. During times of war, for example, suicide rates decline (728), whereas increased suicide rates are found in political systems associated with violence or social movements. Areas of the former Soviet Union with high levels of sociopolitical oppression (i.e., Baltic States) have had higher suicide rates than other regions with less oppression (729). From an economic standpoint, research on the business cycle and suicide has relied primarily on unemployment rates, but other indicators include growth rates of the gross domestic product, the Ayres index of industrial activity, change in the stock market index, and the rate of new dwelling construction (730). Especially for men, the data suggest that the greater the prosperity, the lower the suicide rate, and, conversely, the greater the trend toward recession, the greater the suicide rate. During periods of high unemployment, such as the Great Depression, the relationship of unemployment to suicide is strengthened (731). However, studies using the Ayres index of industrial activity and monthly suicide trends have suggested that large swings in industrial production, such as those that occurred during the 1930s, are needed to influence the suicide rate (732).

In summary, it is important to ascertain the patient's employment status as part of the assessment process, since unemployment may increase suicide risk, whereas employment may offer some protection against suicide and suicide attempts. However, a patient's job status should also be considered in terms of other psychosocial stressors that may be related to job loss, such as financial or marital difficulties. In addition, there is often a complex interplay between employment status and psychiatric illness, including substance use disorder, that may influence treatment planning.

b) Religious beliefs

Limited evidence points to religion as a protective factor against suicide. Pescosolido and Giorgianna (733) used suicide rates from the National Center for Health Statistics, data on affiliation rates in various Christian denominations from the National Council of Churches, and data on Jewish affiliation from the American Jewish Yearbook to determine whether suicide rates differ according to religious affiliation. They found that religion affected suicide rates, with Catholicism, Evangelical Protestantism, and membership in Church of the Nazarenes being associated with lower rates and Jewish affiliation producing a small but inconsistent protective effect. In contrast, various denominations of mainstream Protestantism tended to be associated with increased suicide rates.

Within specific religious denominations, the strength of religious belief may also play a role. Maris (394) compared suicide rates among Catholics and Protestants in Chicago between 1966 and 1968. Scores on church attendance, perception of religiosity, and influence of religion were negatively associated with suicidal ideation. After controlling for the effects of confounding variables such as sex, marital status, and socioeconomic status, Maris found that the perceived influence of religion was the most significant correlate of suicidal ideation. In immigrants from Central America, infrequent church attendance and low levels of perceived influence of religion were related to high levels of suicidal ideation (43). Thus, religious involvement may serve as a protective factor against suicide, either by helping to buffer acculturative stress (43) or by enhancing social networks and support (733).

In summary, some evidence suggests that religious beliefs and the strength of those beliefs may offer protective effects in relation to suicide risk. At the same time, these protective effects neither are specific to particular religious denominations nor are invariably present. Indeed, for some individuals, religious beliefs or beliefs about death may increase rather than decrease the likelihood of acting on suicidal thoughts. Consequently, the clinician may wish to gain an understanding of the patient's specific religious beliefs and the depth of the patient's religiosity as well as determine the ways in which these beliefs influence the patient's conceptions of death and suicide.

c) Psychosocial support

Although it is often difficult to distinguish perceived from objective measures of social support, available data strongly suggest that the presence of a social network is a powerful and independent predictor of suicide risk. In particular, those who have (or perceive themselves to have) supportive interpersonal relationships are at lower risk for suicide than those without such actual or perceived supports. Rubenowitz et al. (405) used the psychological autopsy method to compare 85 persons age 65 years and older who died by suicide with 153 age- and sex-matched living persons selected from the tax roster in Gothenburg, Sweden. In addition to identifying a powerful influence of psychiatric disorders, they found that family discord was a significant risk factor for those who died by suicide (odds ratio19). Further, being active in a social club was a significant protective factor for both men and women. Another recent psychological autopsy study compared 53 individuals age 55 years and older who had either died by suicide or made a serious suicide attempt with 269 matched control subjects (403). Psychiatric illness was again a powerful predictor of suicide case status, but, in addition, those who died by suicide had significantly fewer social interactions and significantly more relationship problems, compared to the control subjects. Turvey and colleagues (400) used data from the Established Populations for Epidemiologic Studies in the Elderly database to identify 21 elderly persons who died by suicide over a 10-year follow-up period and compared those subjects to 420 control subjects matched for age, sex, and study site. In addition to depressive symptoms, poor perceived health status, and poor sleep quality, the absence of a relative or friend in whom to confide was a significant risk factor for late-life suicide. Finally, Miller (399) compared 30 men age 60 years and older who died by suicide with 30 men, matched on age, race, marital status, and county of residence, who died of natural causes. He reported that the control subjects were significantly more likely to have had a confidante and that the subjects who died by suicide had significantly fewer visits with friends and relatives. Thus, while social support is a complex construct and the data on this factor come primarily from elderly populations, decreases in measures of social support appear to increase suicide risk, and, conversely, increases in social support may serve as a protective factor in relation to suicide.

d) Reasons for living, including children in the home

An additional protective factor against suicide is the ability to cite reasons for living, which often reflect the patient's degree of optimism about life. Malone et al. (231) assessed 84 patients, 45 of whom had attempted suicide, to determine whether "reasons for living" might protect or restrain patients with major depression from making a suicide attempt. Depressed patients who had not attempted suicide were found to have expressed more feelings of responsibility toward their families, more fear of social disapproval, more moral objections to suicide, greater survival and coping skills, and a greater fear of suicide than the depressed patients who had attempted suicide. Although objective severity of depression and quantity of recent life events did not differ between the two groups, scores for hopelessness, subjective depression, and suicidal ideation were significantly higher for the suicide attempters.

Particularly in women, the presence of children in the home is an additional factor that appears to protect against suicide. Hoyer and Lund (26) used data from the Norwegian Central Bureau of Statistics to prospectively follow 989,949 women over a 15-year period. During that time there were 1,190 deaths from suicide, with parous women of all ages having lower relative risks than nonparous women (relative risk=0.4–0.8, depending on age). For both premenopausal and postmenopausal women, a strong linear decrease in relative risk for suicide was found with an increasing number of children.

Consequently, during the assessment and treatment planning process, clinicians should discuss reasons for living with at-risk patients and the need to develop coping skills that may serve as protective factors during periods of high risk for suicide.

e) Individual psychological strengths and vulnerabilities

A number of personality traits and characteristics have been associated with suicide and suicidal risk and behaviors. Conner et al. (217) reviewed the literature on psychological vulnerabilities to suicide, including 46 publications describing 35 distinct case-control or cohort samples, and found no evidence for a link between suicide and guilt or inwardly directed anger. They did find that suicide was consistently associated with five constructs—impulsivity/aggression, depression, hopelessness, anxiety, and self-consciousness/social disengagement. Although other factors often moderate the relationships between these variables and suicide, they are not always interpretable in the literature because of measurement and definitional issues. Nonetheless, psychological vulnerabilities likely influence suicide risk by exacerbating other psychiatric or social risk factors in individual patients.

A number of other concepts have also been explored in relating suicide to individual vulnerabilities. For example, Duberstein (423) used questionnaires to assess personality dimensions in 81 depressed patients over age 50 and found that individuals who reported lower levels of openness to new experiences were less likely to report suicidal ideation. These findings are consistent with other work, suggesting that elderly persons tend to deny suicidality, whereas younger persons tend to exaggerate it. These findings may also provide support for the protective role of expressing suicidal ideation. Thus, when closed-minded people do come into contact with treatment services, their psychiatric symptoms may not be as obvious and their need for treatment may not be appreciated.

Hughes and Neimeyer (422) assessed 79 hospitalized psychiatric patients, 91% of whom had a principal diagnosis that included depression, and examined the utility of several cognitive variables as predictors of suicidal ideation. Level of pessimism, as measured by the Hopelessness Scale, was the best predictor of subsequent suicidal ideation and was reliably related to placement on either one-to-one observation or every-15-minute checks for suicide precautions. In addition, hopelessness, self-negativity, polarization (all-or-nothing thinking), and poor problem-solving performance were associated with suicidal ideation, whereas self-evaluated problem-solving ability was not. A low level of constriction was related to the intensity of subsequent suicidal ideation and to later suicide attempts.

Josepho and Plutchik (409) investigated the relationship between interpersonal problems, coping styles, and suicide attempts in 71 adult psychiatric inpatients. Patients who were hospitalized after a suicide attempt had more interpersonal problems and also had distinct patterns of coping methods, including more use of suppression and substitution and less use of replacement. These coping styles were also associated with higher scores on a rating of suicide risk. After controlling for the effect of interpersonal problems, the authors found that greater suppression, less minimization, and less replacement were significantly related to increased suicide risk scores. The higher the risk score, the greater the likelihood that the patient was admitted to the hospital secondary to a suicide attempt. Depressed patients also had higher suicide risk scale scores.

Stravynski and Boyer (401) collected data from 19,724 persons who returned the Quebec Health Survey and tested whether there was an association between loneliness and suicidal thoughts or behaviors in the general population. A significant correlation was found between experiencing suicidal ideation or attempting suicide and living alone, having no friends, or feeling alone, with psychological distress being the strongest correlate of suicidal ideation. Of individuals who were severely distressed and very lonely, 25% reported serious suicidal ideation or actions. Overall, thoughts of suicide were reported by 3.1% of the population, and 0.9% had attempted suicide.

Maser et al. (247) examined the correlations between suicide and clinical and personality factors in 955 depressed patients who were followed over 14 years as part of the NIMH Collaborative Program on the Psychobiology of Depression. During that time, 3.8% died by suicide and 12.6% attempted suicide. Suicide within 12 months of intake to the study was associated with clinical variables, including emotional turmoil plus depression in the index episode, a history of both alcohol and drug use disorders, and meeting the criteria for antisocial personality disorder. Additional predictors included hopelessness, delusions of grandeur, indecisiveness, definite delusions or hallucinations during the index episode, reduced functional role, dissatisfaction with life, or any prior history of serious suicide attempts as of the intake episode. Beyond 1 year after intake, suicide was associated with temperamental factors, including high levels of impulsivity and shyness and low sanguinity scores. Suicide attempters and those who died by suicide shared core characteristics, including previous attempts, impulsivity, substance abuse, and psychic turmoil within a cycling/mixed bipolar disorder.

Kaslow et al. (413) conducted an empirical study of the psychodynamics of suicide among 52 patients hospitalized for a suicide attempt and 47 psychiatrically hospitalized control subjects with no history of suicidal behaviors. Overall, 49% of the subjects had depression, 25% had substance use disorders, and 63% had a cluster B personality disorder. Individuals who had attempted suicide were significantly more likely to report childhood loss combined with adulthood loss. Furthermore, they had more impairment in their object relations and viewed relationships in a more negative manner, showing lower levels of individuation and separation. Although self-directed anger was associated with homicidal ideation, there was little support for the psychodynamic concepts that depression, self-directed anger, or ego functioning would be associated with having made a suicide attempt.

In a group of 438 undergraduate college students who ranged in age from 16 to 65 years, Boudewyn and Liem (734) compared low and high scorers on a chronic self-destructiveness scale that measured behaviors such as chronic gambling or unsafe sexual behaviors that had a potential for later negative consequences. Overall, those scoring high in self-destructiveness were younger and reported more childhood and adulthood maltreatment, lower self-esteem, greater depression, greater externality, less need for control in interpersonal relationships, and more frequent suicidal and self-injurious thoughts and acts. These findings suggest that other manifestations of self-destructiveness should be assessed in the individual evaluation of the suicidal patient and that childhood and adult maltreatment should be specifically identified and addressed in the treatment planning process.

Together with extensive clinical observations on individual strengths and vulnerabilities as they relate to suicidality (410, 412, 420, 426), research on various psychological dimensions has demonstrated the need to include such features in assessing suicide risk. In particular, personality traits such as aggression, impulsivity, social disengagement and subjective loneliness, hopelessness, anxiety, low self-esteem (and protective narcissism), dependence, ambivalence, and depression may increase risk for suicidal behaviors. Thinking styles such as closed-mindedness or polarized (either-or) thinking may also augment risk. If dilemmas are seen only in black-and-white terms, with fewer perceived options, patients may see no solution to their problems other than suicide. In addition to personality traits and thinking style, an individual's psychological needs, when not met, can cause intense psychological pain, contributing to a suicidal state. Early trauma and loss may thwart the development of healthy coping skills. In addition, individual perceptions of interpersonal supports, particularly subjective perceptions of loneliness, may also contribute to suicide risk. Thus, in weighing the strengths and vulnerabilities of the individual patient and developing and implementing a plan of treatment, it is helpful to assess the patient's past response to stress, vulnerability to life-threatening affects, available external resources, perceived sense of loneliness, fantasies about death, and capacity for reality testing and for tolerating psychological pain.

a) Presence, extent, and persistence of suicidal ideation

Suicidal ideation is common, with an estimated annual incidence of 5.6% (2). Kessler et al. (427) examined the lifetime prevalence of suicidal ideation and suicide attempts in a sample of 5,877 individuals age 15–54 years as part of the National Comorbidity Survey. The estimated lifetime prevalences of suicidal ideation, plans, and attempts were 13.5%, 3.9%, and 4.6%, respectively. The cumulative probability of moving from suicidal ideation to an unplanned attempt was 26%. The corresponding cumulative probability for transitioning from suicidal ideation to suicidal plans was 34%, with a 72% cumulative probability for going from a suicide plan to an attempt. About 90% of unplanned attempts and 60% of planned first attempts occurred within 1 year of the onset of suicidal ideation, suggesting a need for aggressive aftercare and attention to potentially modifiable risk factors in individuals with suicidal ideation.

Longitudinal studies also demonstrate an increased risk of eventual suicide in patients with suicidal ideation. Among 6,891 psychiatric outpatients who were followed for up to 20 years, Brown et al. (78) found that patients' scores on clinician-administered measures of current suicidal ideation and depression were most closely associated with eventual suicide. Fawcett et al. (79), using a case-control method to determine time-related predictors of suicide among 954 patients with major affective disorder, examined suicidal ideation as one possible predictor of actual suicide over a 10-year period. They found that the presence of suicidal ideation was associated with an increased risk for suicide on a long-term basis but not within the first year after study entry.

Others have examined the association between eventual suicide and suicidal ideation at its worst using the Scale for Suicide Ideation–Worst (SSI–W) (428). In a group of 3,701 outpatients in which there were 30 suicides, patients who scored in the high-risk category on the SSI–W had a rate of later suicide that was 14 times greater than that of the patients in the low-risk category. After controlling for the effects of other factors, the investigators found that only the SSI–W score, and not the scores on measures of current suicidal ideation or hopelessness, was associated with future suicide (428). Consistent with the findings of Clark and Fawcett (273), the authors concluded that retrospective report of suicidal ideation at its worst may be a better predictor of suicide than currently reported suicidal ideation.

Intuitively, since suicidal ideas would be expected to precede suicidal intent or suicidal acts, they may serve as a guide for clinicians in identifying and addressing suicide risk. These studies also suggest that past as well as current suicidal ideation is relevant to the assessment process. However, since the vast majority of individuals with suicidal ideation do not die by suicide, additional factors are likely to be modulating suicide risk even in individuals with suicidal ideas.

b) Presence of a suicide plan and availability of a method

As noted earlier, about one-third of individuals with suicidal ideas go on to develop a suicide plan, and about three-quarters of those with a plan eventually make a suicide attempt. Other individuals, however, go on to attempt suicide in an unplanned manner. Thus, the presence of a suicide plan signifies that the risk of a later attempt is increased, but it by no means indicates that an attempt will occur or even the time frame within which an attempt may occur. By the same token, the absence of a suicide plan does not eliminate suicide risk. In general, however, the presence of a specific plan involving an available method is associated with a greater degree of risk for suicide. In addition, availability of methods with relatively high levels of lethality may increase the likelihood that a suicide attempt, either planned or impulsive, will result in suicide.

A number of studies have examined population-based trends in suicide rates as they relate to the availability of specific methods for suicide. Ohberg et al. (735), for example, evaluated trends in suicide rates and availability of methods used for suicide in Finland from 1947 to 1990. For both sexes, the overall suicide rate in Finland rose significantly in that time period, but method-specific rates of suicide varied. For example, the rate of suicide by using the highly lethal pesticide parathion decreased after its availability was restricted, but this decrease was offset by an increased rate of suicide by other methods. Before 1962, most suicides occurred by hanging or drowning, but after 1963, there was a rapid increase in the use of firearms. Coincident with increases in the availability of antidepressants and neuroleptics, the rates of suicide by overdose of these medications increased. There was a high number of overdoses of tricyclic antidepressants, which accounted for most of the deaths attributed to antidepressants. On the other hand, the number of overdose deaths attributed to nontricyclic antidepressants decreased, despite increased availability, and the number of overdose deaths attributed to barbiturates remained stable despite reduction in their availability.

Gunnell et al. (436) investigated method-specific trends in suicide between 1950 and 1975 in England and Wales. In the 1950s and early 1960s, domestic gas poisoning was the most frequently used method of suicide among men and women, accounting for one-half of all suicides. Changes in domestic gas supply and manufacture resulted in a reduction in its carbon monoxide content and thus lethality, and overall suicide rates declined in men and women of all ages. In women and younger men (younger than age 55 years), the effects of these reductions on overall suicide rates were partially offset by a rise in the rates of drug overdose deaths, but there were no immediate increases in the use of other suicide methods. In older men, a reduction in the rate of suicide by gassing was accompanied by only a slight increase in the rate of suicide by overdose as well as reductions in rates of suicide by using all other methods.

Marzuk et al. (437) investigated the relationship between the availability of lethal methods of injury and suicide rates by prospectively classifying lethal methods according to their accessibility in the five counties of New York City over a 4-year period and then comparing the age- and gender-adjusted method-specific suicide rates of the counties. During the study period, there were a total of 2,820 suicides, a rate of 9.81 per 100,000 persons. The study found marked differences in overall crude suicide rates among the five counties, which ranged from 15.27 per 100,000 persons in Manhattan to 5.58 per 100,000 persons in Staten Island. The counties had similar suicide rates involving methods that were equally accessible to all persons in each county (e.g., hanging, laceration, suffocation, and burns) as well as methods that were accessible to a smaller but similar proportion of the population in each county (e.g., firearms and drowning in waterways). Virtually all of the differences in overall suicide risk among the counties were explained by methods that were differentially available, such as fall from height, overdose of prescription drugs, and carbon monoxide poisoning (explained by access to private parking). The availability of a greater variety of alternative lethal methods in some counties did not suppress the rates of use of other methods, and a relative lack of the availability of a specific method did not result in a comparative increase in the rates of use of alternative methods that were available, as the substitution hypothesis would have predicted. Thus, restriction of the availability of a method may reduce its use for suicide, but other methods may tend to be used instead. At the same time, the accessibility to and lethality of particular methods of suicide may have definite effects on the overall suicide rate.

In the United States, firearms constitute the most common method for suicide (736, 737). Fox et al. (738) used mortality data for 1979–1994 from the Wisconsin Center for Health Statistics and the U.S. Census Bureau population estimates for Wisconsin to describe trends for firearm-related suicides in that state. During that period, there were an average of 588 suicides annually, with firearms eclipsing all other methods combined as the most common method of suicide in the 1980s. Between 1981 and 1992, the proportion of firearm suicides increased from 48% to 57%. While the overall suicide rate remained unchanged over the period, the firearm suicide rate increased 17% in all sex, race, and age categories. Among males, the firearm-related suicide rate rose by 13% during the study period, while the rate of suicide by all other methods combined fell 12%. In comparison, among females, the firearm-related suicide rate rose 20%, and the rate of suicide by all other methods fell 26%.

Kaplan and Geling (434) investigated the sociodemographic and geographic patterns of firearm suicides in the United States using mortality data from the National Center for Health Statistics Mortality detail files and death certificate files reported by each state from 1989 to 1993. During this time period, 59.2% of the 139,566 suicides were by firearms. Married persons had the lowest rate of any form of suicide across all race, sex, and age groups. The adjusted odds of using firearms increased with age among men and decreased with age among women. Widowed men and married women had the highest odds of using firearms, and the odds of using a firearm for suicide were also high among those without college education, those who had lived in nonmetropolitan areas, and those who had lived in the East South Central and West South Central geographic divisions. Rates of nonfirearm suicides were higher than firearm suicides everywhere but in the regions of the South. Thus, the likelihood of firearm suicide varied significantly across sociodemographic and geographic subgroups of the U.S. population and paralleled variations in gun ownership, suggesting that regional cultural factors may account for differential rates in suicidal behavior involving firearms.

In addition to population-based data on firearm availability and suicide risk, some data also suggest an effect at an individual level. Brent et al. (438) performed a case-control study to determine the relationship between the presence of guns in the home, the type of gun, the method of storage, and the risk of suicide among adolescents. Forty-seven adolescents from the community who died by suicide were compared with two control groups from a psychiatric hospital: 47 patients who attempted suicide but survived and 47 patients who had never attempted suicide. The study found that guns were twice as likely to be found in the homes of those who died by suicide as in the homes of the suicide attempters or psychiatric control subjects. There was no significant difference in association with suicide between handguns and long guns, and there was no difference in the methods of storage of firearms among the groups. The authors concluded that the availability of guns in the home, independent of the type of firearm or storage method, appears to increase the risk for suicide, at least among adolescents.

In summary, the presence of a suicide plan and the availability of a method for suicide increase risk and are important issues to address as part of the suicide assessment. Since firearm-related suicides account for a significant fraction of suicides in the United States, the presence and availability of firearms are also an important line of inquiry in a suicide risk assessment. A debate remains over whether a reduction in the availability of a particular method of suicide reduces overall risk, although most evidence indicates that restrictions on the availability of particular types of popular methods result in a lower overall suicide rate. At the individual level, reducing access to specific suicide methods may also be indicated. See Section II.C.2, "Elicit the Presence or Absence of a Suicide Plan", for additional discussion of inquiries, removal, and documentation issues related to firearms and the suicidal patient.

c) Lethality and intent of self-destructive behavior

In addition to being increased by the presence of suicidal ideation, a suicide plan, or an available suicide method, suicide risk is also influenced by the patient's subjective expectation and desire to die as a result of a self-inflicted injury. This factor has generally been termed suicidal intent, although the patient's subjective expectation may or may not correspond to the lethality of an attempt made by using a given method. Other facets of a suicide plan or attempt that are often considered when estimating suicidal intent include the severity and potential lethality of the suicide attempt or aborted suicide attempt, the patient's degree of premeditation, whether precautions were taken to avoid intervention or discovery, and whether the patient's intentions were communicated to others (263, 433, 440).

Several studies have longitudinally assessed the influence of suicidal intent on later suicide risk. In a group of 500 patients who had completed a scale measuring suicidal intent after an episode of self-injury, Pierce (441) found that the seven individuals who had died by suicide by the time of a 5-year follow-up tended to have high suicidal intent scores at the time of their initial self-injury. In addition, individuals with increasing levels of suicidal intent with repeated self-injury appeared to be at greater risk for further repetition of self-injury (739). Suokas et al. (271) also conducted a longitudinal assessment of the effect of suicidal intent on suicide risk. They found that 68 (6.7%) of 1,018 deliberate self-poisoning patients had died by suicide by 14-year follow-up. Risk factors for suicide included being male, having previous psychiatric treatment or suicide attempts, having a somatic disease, and having a genuine intent to die at the time of the index self-poisoning.

Thus, for any patient with suicidal ideation, it is important to determine suicidal intent as part of the assessment process. In addition, for any patient who has made a prior suicide attempt, the level of intent at the time of the previous attempt should be determined.

As noted previously, it is impossible to accurately predict suicide. Nevertheless, given the large number of risk factors and protective factors that can affect the likelihood of suicide, a number of statistical models have been developed to attempt to pinpoint which patients may be at greatest risk. In a longitudinal study by Pokorny (160) that followed 4,800 subjects (4,691 males and 109 females) over a 5-year period, stepwise discriminant analysis was used to select a weighted combination of predictive variables from the identified high-risk characteristics, i.e., being a white male; being single; having a diagnosis of affective disorder, schizophrenia, or alcoholism; having made a previous suicide attempt; or having personality disorder-related traits such as manipulativeness and hostility. This method was able to correctly identify 30 of the 67 subjects who died by suicide but also falsely predicted suicide in 773 individuals. Thus, while it may be possible to identify a high-risk group of patients who warrant more detailed clinical screening, it may not be possible to identify the particular individuals at greatest risk.

Goldstein et al. (740) also used a statistical model that incorporated multiple risk factors for suicide and applied it in a group of 1,906 patients with affective disorders who were admitted to a tertiary care hospital and were followed longitudinally. The identified risk factors included the number of prior suicide attempts, the presence of suicidal ideation on admission, gender, outcome at discharge, and a diagnosis of either bipolar affective disorder (manic or mixed type) or, in individuals with a family history of mania, unipolar depressive disorder. The full statistical model, however, was unable to identify any of the patients who died by suicide, highlighting the difficulty of estimating suicide risk with such methods.

In general, statistical models may be valuable in the epidemiological and research arenas by identifying factors that distinguish high-risk populations of patients. They can also suggest clinically important risk factors that, if identified, are potentially amenable to treatment. However, given the low base rates of suicide in the population, accurate prediction of suicide remains impossible, regardless of the complexity of the statistical model used. Consequently, the psychiatric assessment, in combination with clinical judgment, is still the best tool for assessing suicide risk. In addition, intervention must be based not on the simple presence of risk factors as identified by statistical models but on the interaction of those factors with the individual patient's personal and clinical manifestations and the clinician's assessment of the patient's risk at that particular point in time.