On March 14, 1932, George Eastman, the fabulously wealthy industrialist and philanthropist who founded the Eastman Kodak Company, took his own life with a gunshot to the left chest (1). He was 77 years old. A suicide note left on his bedside table said simply, “Friends. My work is done. Why wait?” These last words seemed to reflect the autonomy and self-determination that had made Mr. Eastman so successful in life. The reality, however, was far different. For several years Eastman had been racked with pain from a spinal disorder. Becoming progressively more disabled, he was required to cede control of his company. Isolated from friends and struggling to find meaning in life, Eastman became despondent and ended his own life.
Other than for his riches, Eastman is typical in many respects of older adults who take their own lives. With that backdrop, the following sections provide a brief review of the epidemiology of suicide among older adults in the United States, current knowledge regarding risk and protective factors, and evidence for the most promising approaches to reducing suicide-related morbidity and mortality in later life.
As depicted in Figure 1, suicide rates vary greatly as a function of age, sex, and race (2). Women of all ages and race/ethnicities tend to have lower rates of suicide than men, and whites have higher rates than nonwhites. For both African American and American Indian men, the suicide rate peaks in young adulthood followed by steady declines thereafter. White men show a markedly different pattern in which rates rise to a peak at midlife, diminish somewhat then escalate dramatically to a rate in the oldest-old (50.8/100,000) that is over four times higher than that of the general population (12.1/100,000). In 2010, almost 6,000 people over the age of 65 years died by their own hand. In contrast to the U.S., most countries for which World Health Organization statistics are available report that rates for suicide rise steadily throughout the life course for both men and women (3)
Figure 1.Suicide Rates in the United States by Age, Sex, and Race, 2010
Source: Centers for Disease Control and Prevention, Web-based Injury Statistics Query and Reporting System [WISQARS])
Unlike completed suicide, period prevalence rates of both suicidal ideation (4) and attempted suicide (2) decrease in frequency with greater age. As a consequence, the ratio of completed to attempted suicides is far higher among older adults than in younger and middle-aged populations. Studies estimate one suicide death of an older adult for every two to four who are hospitalized with a nonfatal self-inflicted injury (5). Among the general population, that ratio is estimated to be 1:30, and among female adolescents as much as 1:200 (6). Possible explanations for this pattern include the greater physical frailty and lesser resilience of older people, making death more likely as a result of any injury; that older adults are more likely than younger and middle-aged people to live alone, and thus are less likely to be rescued in the event of self-injury; and because older people in suicidal crises are more planful and determined to die (7). Whereas just over half of all suicides in the United States are with a firearm, almost three quarters of older adults who take their own lives do so with a gun (2). Because older adults are less likely to endorse suicidal ideation or have prior histories of suicide attempts than younger people, the detection of those at imminent risk is that much more difficult.
The fact that older people at elevated risk for suicide are both more likely to escape notice and more likely to die as a result of any initiated self-destructive act has two important implications. First, concern that an older person might be suicidal requires aggressive clinical intervention to maintain their safety, assess their risk status, and intervene as indicated. Second, special emphasis should be placed on approaches that prevent development of suicidal states in later life, because once an older adult enters a suicidal crisis, death is a far more likely outcome than for a younger person in that condition.
The design of preventive interventions hinges on adequate understanding of those factors that predispose or protect from suicide. Figure 2 depicts one useful way to organize current knowledge about risk and protective factors – their categorization into five domains analogous to the five axes of psychiatry’s Diagnostic and Statistical Manual of Mental Disorders (8).
Figure 2.Domains of Risk for Suicide in Older Adults
Axis I: major psychiatric illness
Table 1 lists the results of five case-controlled psychological autopsy studies of suicide in the second half of life (9–13). Results were consistent with previous uncontrolled psychological autopsy studies in demonstrating that a high proportion (80%–100%) of suicides die with a diagnosable axis I disorder (14). Mood disorders consistently showed the highest associations with suicide case status across all studies. Both major depressive disorder and other affective syndromes were associated with increased risk in this age group. In contrast, only two of five studies found a significant association between substance use disorders and completed suicide in these older adult samples, with similar inconsistent findings for anxiety and schizophrenic spectrum disorders. Only one of four studies that examined the role of dementia or delirium found a significant association—an apparent protective effect. This unintuitive finding may represent an artifact of the retrospective psychological autopsy method. Individuals with dementia may be at greatest risk for suicide early in the course of the illness when affective symptoms are most common, but before formal diagnosis is likely to be made and when family members and other informants are unaware of its presence. At later stages of dementia when diagnosis is more easily established, higher levels of supervision and difficulty planning and carrying out a suicidal act may explain lower relative risk. Neuropathology studies of Alzheimer’s-type changes in postmortem brains of suicides and controls have yielded mixed results (15, 16).
Table 1.Odds Ratios for Suicide by Axis I Diagnosis in Case-Controlled Psychological Autopsy Studies of Older Adults
| Add to My POL
|Study||Number of Cases||Age||Gender (M/F)||Odds Ratios|
|Suicides||Controls||Suicides||Controls||Any Axis I Diagnosis||Any Mood Disorder||Major Depressive Episode||Substance Use Disorder||Anxiety Disorder||Schizo-phrenic Spectrum||Dementia/Delirium|
|Harwood et al., 2001 (9)||54||54||≥ 60||n/a||n/a||--||4.0||--||n.s.||--||n.s.||0.2|
|Beautrais, 2002 (10)||53a||269||≥ 55||27/26||n/a||43.9||184.6||--||4.4||--||--||--|
|Waern et al., 2002 (11)||85||153||≥ 65||46/39||84/69||113.1||63.1||28.6||43.1||3.6||10.7||n.s.|
|Chiu et al., 2004 (12)||70||100||≥ 60||32/38||43/57||50.0||59.2||36.3||n.s.||n.s.||>1||n.s.|
|Conwell et al., 2009 (13)||86||86||≥ 50||63/23||63/23||44.6||47.7||12.2||n.s.||5.9||n.s.||n.s.|
While other axis I psychiatric illnesses likely play a role in late life suicide, affective disorders are the most prominent factor, associated with far higher odds ratios than any other putative risk factor.
Axis II: personality and coping
Based on the Five-Factor Model of personality, traits of high neuroticism (the tendency to experience negative affect) and low openness to experience (preferring the familiar to the novel, blunted affective and hedonic responses) were associated in one retrospective case-controlled study of suicide in later life (17). A separate study found that anankastic (obsessional) and anxious traits were also associated with late life suicide (9).
Axis III: physical health
A variety of physical illnesses have also been shown in both retrospective psychological autopsy and record linkage studies to be associated with suicide (18–20). Specific illnesses most frequently identified as risk factors include malignancies and central nervous system disorders (e.g., epilepsy, spinal cord injury, Huntington’s disease), chronic obstructive pulmonary disease, congestive heart failure, and chronic pain. The impact of physical illness may be cumulative. In a retrospective case-control study of late-life suicide, Juurlink and colleagues showed that the relative risk of suicide increased with the number of comorbid physical disorders (19). Compared with patients with no identified illness, for example, patients with three illnesses had over three times higher relative risk of suicide (odds ratio=3.5, 95% CI=2.9–4.2); patients with five illnesses were at almost six times greater risk (odds ratio=5.7, 95% CI=4.4–7.4).
Studies comparing older adults who took their own lives with matched controls show that social factors determine suicide risk independent of psychiatric illness. In addition to losses common in older adulthood (e.g., bereavement, retirement, and disability), stressors that lead to social disconnectedness are particularly salient. Beautrais reported that serious relationship problems distinguished older adults with near fatal suicide attempts from controls in New Zealand (10), and in both Sweden (21) and the U.S (22, 23), family discord was significantly more common in the lives of older adult suicides than in matched, living comparison samples. Social connectedness appears also to serve as a protective factor. Individuals who report a strong family connection are less likely to report suicide ideation (24). In other retrospective studies older adult suicides were significantly less likely to have a confidante than controls (25), more likely to live alone than their peers in the community (26), and less likely to participate in community activities (23), be active in organizations, or have a hobby (21).
Axis V: functional impairment
Because physical illness and functional limitations are the norm in older people, assessment of functional capacity and any resulting disablement is a necessary component of comprehensive geriatric assessment. Evidence now shows that functional limitations and disablement make substantial independent contributions to suicide risk in older people, and therefore represent potential targets for preventive interventions. In their case-controlled study of suicide in later life, Waern and colleagues reported a significant association between suicide and need for help with activities of daily living in those over age 75 years (27). Tsoh and colleagues found that older adults who had attempted or completed suicide had greater functional impairment than nonsuicidal older adult controls (28), and our group has reported that deficits in instrumental activities of daily living significantly differentiated suicides from controls, even after accounting for presence of psychiatric disorders (13). Hospitalization for medical or surgical reasons as well as use of visiting nurse or home health aide services increased risk as well. Findings of Dombrovski and colleagues highlight more specifically the role of neurocognitive deficits in late life suicidal behavior as well (29). They have reported impaired reward/punishment learning in older adult suicide attempters, but not ideators, positing that older adults who attempt suicide over-emphasize present reward/punishment contingencies to the exclusion of past experiences. More research is clearly needed that links studies of brain structure and functioning, using refined measures of discrete cognitive processes and carefully characterized samples of older adults with and without suicidal behavior.
Given that such a high proportion of older adults who die by suicide used a firearm, it is important to know whether access to guns is itself a risk factor. We compared gun ownership and storage among matched samples of older adults who killed themselves and living controls (30). Suicides were significantly more likely to have a handgun in the home; easy access to long guns did not distinguish the groups.
Access to and familiarity with firearms has been postulated to explain the increased risk for suicide observed among veterans of the armed forces at all ages (31). The elevated risk associated with veteran status is particularly pertinent to suicide prevention in later life because two thirds of men over age 65 have served in the military (32).
One final point warrants emphasis for clinical practice. Research that specifically examines the impact on suicide risk of interaction between factors is scarce. Nonetheless, clinicians should be increasingly concerned about their older patients, not only as the number and severity of risk factors for suicide within any domain rises, but as the number of domains represented in the individual’s risk assessment increases as well. Figure 2 illustrates common scenarios among older adults at the areas of interface between domains of risk. Where a larger number of domains overlap, risk is increased. Where all five domains are represented, referred to here as the area of highest convergent risk, the likelihood of suicide is greatest.
In order to design effective preventive interventions, one must know not only characteristics that place older adults at risk for suicide that are amenable to change, but also where older adults with these risk characteristics can be most efficiently identified and engaged in prevention activities. Older people at risk for suicide seek help from mental healthcare providers far less often than younger and middle aged cohorts. On the other hand, one-quarter to a third of older adults who took their own lives were seen in a primary care practitioners office within the last week of life, and a half to three-quarters within the last month (12, 33, 34). Primary care, therefore, represents one important setting in which to detect at-risk elders and intervene. Another is home health and community-based long-term care supports and services, clients of which have been shown also to have a high prevalence of mood disorders and suicidal ideation as well as physical illness burden, functional impairment, and other social stressors (35–38). Given the large number of older adult men who are veterans, a group at even greater risk for suicide, Veterans Service Organizations and Veterans Health Administration facilities are likely to be important venues for prevention programming as well.
The Institute of Medicine classifies preventive interventions into three types (39). The first, and most familiar to clinicians, is “indicated” prevention, which targets individuals at high risk with detectable symptoms of major psychiatric illness and/or other proximal risk factors for suicide. The second is “selective” preventive interventions, which target asymptomatic or presymptomatic individuals or groups with distal risk factors for suicide, or who have a higher than average risk of developing mental disorders due to presence of more distal factors. Finally, there are “universal” preventive interventions that address risk in an entire population irrespective of the risk of any individual or subgroup. “Multilevel” preventive interventions refer to those approaches that combine components from more than one level (for example, a combination of indicated and selective interventions.)
Table 2 lists published studies in which suicidal ideation or behavior in older adults was the targeted outcome. Of eight studies listed, five are best characterized as indicated interventions (40–44), one as a selective approach (45), one universal (46), and one multilevel (47). Because suicidal ideation and behavior are uncommonly expressed in later life, their study is challenging and, as a result, the evidence base for preventive interventions is limited. Further complicating interpretation of the available evidence is that relationships between suicidal ideation and behavior in later life have yet to be fully defined. For example, do wishes for an early death and thoughts of taking one’s own life carry the same risk of future suicide or suicide attempts? Who among those older persons with histories of prior suicidal behavior is most likely to take his own life? It is premature, therefore, to assume that interventions effective in addressing suicidal ideation will have the same effect on attempted or completed suicide in later life.
Table 2. Interventions Associated With Suicide Risk Reduction in Later Life
| Add to My POL
|Study||Study Design||Prevention Approacha||Intervention||Participants||Age||Outcome Assessed||Effectb|
|Unützer et al., 2006 (U.S.A.) (40)||Randomized controlled trial||Indicated||IMPACT: Primary care-based depression care management; tx algorithms; patient, family, provider education||1801 with major depression/ dysthymia: 996 intervention, 895 controls||≥ 60||Suicidal ideation||Resolution of suicidal ideation: OR=0.7 (95% CI=0.4–0.8)|
|Alexopoulos et al., 2009; Bruce et al., 2004 (U.S.A.) (41)||Randomized controlled trial||Indicated||PROSPECT: Primary care-based depression care management; treatment algorithms; patient, family, provider education||599 with mood disorders: 320 intervention, 279 controls||≥ 60||Suicidal ideation||For patients with major depression, resolution of suicidal ideation at 24 months: OR=3.2 (95% CI=1.1–9.5)|
|Heisel et al., 2009 (Canada) (42)||Case series||Indicated||IPT to improve social functioning + existing treatment||11 referrals from clinicians/medical staff||≥ 60||Suicidal ideation||Pre/post reduction in suicidal ideation score: p=0.01|
|Stone et al., 2009 (U.S.A.) (43)||Meta-analysis||Indicated||Antidepressant medications||372 randomized, placebo-controlled trials, with 99,231 randomized subjects with affective disorders (50%) or other psychiatric conditions (50%)||≥ 18||Suicidal ideation (or behavior [14%])||Decreasing risk of newly emerging suicidal ideation with age: <25 yrs: OR=1.62 (95% CI=0.97–2.71); 25–64: OR=0.79 (95% CI=0.64–0.98); ≥65: OR=0.37 (95% CI=0.18–0.76)|
|Oyama et al., 2008 (Japan) (47)||Meta-analysis||Multilevel||Depression screening, psychoeducation workshops, referral, follow-up, treatment by psychiatry or primary care||Five quasi-experimental studies comparing regions with and without intervention. Men: 20,598 person years; women: 28,437 person years||≥ 65||Suicide||Psychiatrist follow-up: men: IRR=0.3 (95% CI=0.1–0.7), women: IRR=0.3 (95% CI=0.2–0.6); GP follow-up: men: n.s., women: IRR=0.4 [0.2–0.6]|
|De Leo et al., 2002 (Italy) (45)||Ecological study||Selective||24 hr. access to supports as needed; weekly phone contact||Men: 2,983 women: 15,658||≥ 65||Suicide||For women, standardized mortality ratio=16.7% (2.0%–59.9%); for men: n.s.|
|Chan et al., 2011 (Hong Kong) (44)||Cohort study||Indicated||Primary care-based gatekeeper training, referral to geropsychiatry, care management, active aftercare for suicide attempters.||351 suicide attempters received intervention (66 preintervention), all diagnoses||≥ 65||Suicide and suicide attempt||2-year suicide rate: p=0.028; reattempt rates: p=n.s.|
|Ludwig & Cook, 2000 (U.S.A.) (46)||Ecological study||Universal||Relative change in handgun suicides in states that implemented gun control legislation versus those with no new policy implementation.||All 50 U.S. states, vital statistics data reports of suicides from 1985 through 1997||All ages||Handgun suicides||Rate reduction per 100,000 population: −0.92 (95% CI=−1.43 to −0.42) for those ≥ 55 years. No difference for homicide rates or overall suicide rates.|
Studies of interventions that target suicidal ideation
The PROSPECT and IMPACT studies were rigorously conducted randomized controlled trials designed to test whether primary care-based collaborative depression care management for older adults was more effective than enhanced care as usual in reducing suicidal ideation among older adults with major depression and dysthymia (48, 49). Both studies found significantly greater improvement in depressive symptoms and suicidal ideation in those who received the care management intervention (40, 41). In neither study were there sufficient suicide attempts to examine the effectiveness of depression care management on suicidal behavior. Given the importance of primary care as a venue for suicide risk management in later life, and because integrated approaches to the management of comorbid mental illness and chronic physical disorders have been shown so effective (50, 51), the wider dissemination of primary care-based collaborative depression care management is a promising approach to addressing late-life suicide. Whether suicide deaths can actually be reduced remains to be determined.
Ecological studies of medication prescribing rates and their association with suicide mortality have suggested that antidepressant administration is an effective indicated preventive intervention (52–54). Interpretation of the findings remains a subject of debate, including in older adults, however (55, 56). Stone and colleagues reported results of a large meta-analysis of Food and Drug Administration (FDA) data from 372 randomized, placebo-controlled trials of antidepressant medications (43). The data revealed a statistically greater risk that suicidal ideation would emerge in adolescents and young adults during the course of treatment with active medication than placebo. These findings contributed to the institution by the FDA of a “black box” warning for the use of antidepressant medications in this age group. Less widely appreciated was the finding that among those research subjects over the age of 40, risk of suicidal ideation or behavior emerging during the drug trials was significantly reduced.
Early findings indicate the likelihood that psychosocial interventions may be effective in reducing suicidal ideation in older adults as well. Heisel and colleagues, for example, demonstrated in a case series of suicidal older adults that thoughts of killing themselves significantly diminished over the course of treatment with adapted interpersonal psychotherapy (IPT) (42). More definitive trials of IPT as well as cognitive behavioral therapy for high risk elders are ongoing.
Studies of interventions that target suicide and suicide attempts
Because of complex ethical and logistical constraints, no randomized controlled trials have yet been reported in which the outcome was attempted or completed suicide. Four trials listed in Table 2, however, provide some indication of potential effect of selective, universal, and multilevel approaches tested by less rigorous methods. De Leo and colleagues, for example, reported results of the Tele-help/Tele-check intervention in which older adults at risk for adverse physical and mental health outcomes were provided telephone-based access to supportive services (45). Both on-demand and service-initiated contact by social workers with at-risk elders was associated over 11 years of intervention delivery with significantly fewer suicides than would have been expected in a comparable population (standardized mortality ratio of 0.167). The intervention is best characterized as a selective approach because it targeted a group with risk characteristics of functional impairment and social isolation rather than individuals at high risk.
In five separate studies Oyama and colleagues tested multilevel approaches to suicide prevention that combined varying elements of indicated, selective, and universal preventive interventions for older adults in rural Japanese villages. Components included depression screening for older adult residents, referral to either a general practitioner or mental health specialist for those who screened positive, engagement of older adults in group activities, and community-based psychoeducational sessions. Suicide rates in the intervention villages were then compared with demographically similar regions. Merging the five studies using meta-analytic methods (47), the investigators found that when follow up was conducted by a psychiatrist, the suicide incidence rate ratios in intervention areas were significantly reduced for both men and women. When general practitioners provided the depression care, however, the significant effect was found only for older female participants. Interestingly, more detailed analysis of the Tele-help/Tele-check intervention also revealed an effect only for women (45).
Chan and colleagues reported results of an indicated preventive intervention in Hong Kong in which older adults who survived a suicide attempt were referred to a multicomponent prevention program that included psychiatric evaluation and care and ongoing care management (44). They found significantly fewer suicides occurred during two years of program implementation than in a comparable group in the period before the program was begun. However, there was no apparent pre/postintervention difference in reattempts.
Finally, almost no data are available about the effectiveness of a purely universal preventive approach on reducing suicidal behavior in older people. A signal that universal prevention may be helpful was provided, however, by Ludwig and Cook in an analysis of ecological data associated with implementation of the Brady Handgun Violence Prevention Act of 1994 (46). They observed that in the years following implementation of the legislation there was a significantly greater reduction in firearm suicides by people over the age of 55 years in those states that newly implemented background checks and waiting periods for gun purchase than in states in which no additional gun control regulations were required.
In coming decades, the size of the older adult population in the U.S. will increase dramatically. Similar changes will be observed in countries throughout the world due to increasing life expectancy and falling fertility rates. Far more work must be done in a number of areas to limit suicide-related morbidity and mortality in this vulnerable and rapidly growing population of older people. We must better understand factors that place older adults at risk for suicide, in particular through multivariate research designs that define not only which factors and domains of factors are most potent in determining risk, but how they interact to determine risk status. We must define with greater precision the implications for risk assessment of thoughts of death and suicide in later life. And finally, we must apply that knowledge to the design and rigorous testing of preventive interventions that incorporate the most promising approaches to late life suicide prevention at all levels–indicated, selective, and universal.