Age Differences in Behaviors Leading to Completed Suicide

Abstract

The authors described retrospectively the premorbid self-destructive behaviors of suicide victims to determine whether these behaviors differ with age. One hundred forty-one suicide victims, age 21–92 years, were studied by psychological autopsy. Older age was significantly associated with more determined and planful self-destructive acts, less violent methods, and fewer warnings of suicidal intent. Age differences in the behaviors leading to suicide indicate that intervention in the midst of a suicidal crisis may be less effective in elderly persons than in younger people. Primary prevention should be the focus of efforts to decrease suicide rates in late life. The authors described retrospectively the premorbid self-destructive behaviors of suicide victims to determine whether these behaviors differ with age. One hundred forty-one suicide victims, age 21–92 years, were studied by psychological autopsy. Older age was significantly associated with more determined and planful self-destructive acts, less violent methods, and fewer warnings of suicidal intent. Age differences in the behaviors leading to suicide indicate that intervention in the midst of a suicidal crisis may be less effective in elderly persons than in younger people. Primary prevention should be the focus of efforts to decrease suicide rates in late life. In young adulthood, suicide attempts occur up to 200 times as often as completed suicide.1Curran DK Adolescent Suicidal Behavior. Hemisphere, New York1987Google Scholar In contrast, the ratio of attempted to completed suicide among elderly persons is approximately 4 to 1.2McIntosh JL Santos JF Hubbard RW et al.Elder Suicide: Research, Theory, and Treatment. American Psychological Association, Washington, DC1994Crossref Google Scholar This difference has been ascribed to a variety of factors. Because older people who commit suicide are more socially isolated,3Conwell Y Suicide in elderly patients.in: Schneider LS Reynolds III, CF Lebowitz BD Diagnosis and Treatment of Depression in Late Life. American Psychiatric Press, Washington, DC1994: 397-418Google Scholar they are less likely to be rescued after a self-destructive act. The greater physical frailty of older people makes any injury more potentially lethal, as well. Other findings suggest that the large number of completed suicides in late life, relative to the number of attempts, results from a greater intent to die among suicidal older people. Merrill and Owens4Merrill J Owens J Age and attempted suicide.Acta Psychiatr Scand. 1990; 82: 385-388Crossref PubMed Scopus (85) Google Scholar found higher scores on a measure of suicidal intent among older persons admitted after an overdose than among younger age groups. Frierson5Frierson RL Suicide attempts by the old and the very old.Arch Intern Med. 1991; 151: 141-144Crossref PubMed Scopus (118) Google Scholar also noted greater premeditation of the act among older than younger attempters. Examining persons who did take their own lives, one study has shown that older victims chose more deadly methods,6McIntosh JL Santos JF Methods of suicide by age: sex and race differences among the young and old.Int J Aging Hum Dev. 1985–86; 22: 123-139Crossref PubMed Google Scholar and another, that they gave fewer warnings to others of their suicidal intentions, than did younger samples.7Carney SS Rich CL Burke PA et al.Suicide over 60: The San Diego Study.J Am Geriatr Soc. 1994; 42: 174-180Crossref PubMed Scopus (143) Google Scholar Differences in the behaviors leading to suicide have important implications for prevention. Different approaches may be indicated at different ages. For example, if suicidal elderly persons are more deliberate and planful in their self-destruction than younger people, they will be more difficult to identify as being at imminent risk and more difficult to rescue after initiation of the act. Measures for this high-risk population may be better aimed at preventing development of the suicidal state rather than at its recognition and subsequent intervention. As part of a psychological autopsy (PA) study of suicide victims across the adult age-span, we gathered information on the subjects' recent and remote self-destructive acts. Our objective of the analyses reported here was to describe the behaviors systematically and test whether they differed as a function of age at death. One hundred forty-one victims, age 21 years and over, whose deaths were ruled by the Office of the Medical Examiner (ME) of Monroe County to have been suicides, between August 1, 1989, and February 28, 1992, were studied by the PA method. Information sources included next-of-kin and others (friends, coworkers, health care providers) with knowledge of the victims' recent and remote histories, including their symptoms and behaviors in their last week of life. As described elsewhere,8Conwell Y Duberstein PR Cox C et al.Relationships of age and Axis I diagnosis in victims of completed suicide: a psychological autopsy study.Am J Psychiatry. 1996; 153: 1001-1008Crossref PubMed Scopus (643) Google Scholar initial contact with the next-of-kin occurred within 2 weeks of the subject's death. The average length of time (±standard deviation [SD]) between the death and the first interview was 9.2±6.4 weeks. In addition to in-person interviews with the victims' families, investigators conducted telephone interviews with physicians and other health care providers (e.g., substance abuse counselors, psychotherapists). Interviews with friends, acquaintances, coworkers, and relatives were also conducted when it was feasible and appropriate. Police, prison, medical, and educational (primary and secondary school) records were also obtained. Information was gathered from a mean of 4.6±2.2 separate sources for each case. There was no statistically significant correlation between age at death and the number of informant sources. Interviews were conducted under the auspices of the ME; informants' voluntary participation was determined by the County of Monroe and the University of Rochester Research Subjects Review Board to constitute informed consent. Data on the subjects' suicidal behaviors were obtained by means of a semi-structured interview. In addition to a history of previous suicide attempts and direct and indirect comments to informants about suicidal intent, ratings were derived from interview for the first eight items of the Suicide Intent Scale (SIS).9Beck AT Schuyler D Herman I Development of suicidal intent scales.in: Beck AT Resnick HCP Lettieri D The Prediction of Suicide. Charles Press, Bowie, MD1974: 45-56Google Scholar The SIS is a widely used measure of the degree to which a self-destructive act is intended to result in death. Only those items were included that reflect observable behaviors: physical isolation from others, timing of the act to avoid intervention, precautions against discovery, efforts to get help after the act, preparations for death (e.g., preparation of a will, gifts), preparation for the act (e.g., obtaining the means), suicide notes, and communication of intent to others. Each item is scored on a 0–2 scale, with a maximum score of 16. Higher values suggest greater intent to die. Information was available to derive a total SIS score on 136 of 141 subjects, and in 137 cases, a confident determination of the presence or absence of previous self-destructive acts could be made. For the continuous variable of total SIS score, our statistical approach was multiple analysis of covariance. For analyses in which outcome measures were dichotomous (use of violent method of suicide, a past history of suicide attempt[s], and warnings to others of suicidal intent in the last week and month of life), multiple logistic regression was used. Age at death was the principal variable of interest. Gender, an age × gender interaction term, and other factors likely to influence SIS score were incorporated as covariates. These included psychiatric diagnosis (substance use disorder, affective disorder, or schizophrenia), living situation (alone vs. with others), and education (fewer than 12 years vs. 12 or more years). To aid visual inspection of the data, Table 1 and Table 2 divide the sample into four age-groups—representing “young adult” (21–34 years), “middle aged” (35–54 years), “young-old” (55–74 years), and “old-old” (75–92 years) subjects. In addition to demographic characteristics, Table 1 lists the methods used to commit suicide and the proportion of each age group that used a violent (firearm, hanging, cutting/stabbing, jumping, or burning oneself to death) rather than nonviolent (ingestion, carbon monoxide poisoning, or drowning) means. Younger age (χ2[1] = 3.49; P = 0.031) and male gender (χ2[1] = 10.66; P = 0.0006) were significantly associated with having used a violent method, whereas substance use disorder predicted use of less violent means (χ2[1] = 4.08; P = 0.022).TABLE 1Characteristics of completed suicidesAge, yearsaAge is analyzed as a continuous variable, but is displayed by group to aid visualization.21–3435–5455–7475–92Totaln (%)46 (32.6)45(31.9)36 (25.5)14 (9.9)141 (100)Men (% of age-group)36 (78.3)40 (88.9)28 (77.8)9 (64.3)113(80.1)Education (% ≥ 12 years)58.777.852.842.961.7Living alone, n (%)7(15.2)13 (28.9)14 (38.9)6 (42.9)40 (28.4)Using violent method, n (%)bViolent = firearm, hanging, cutting/stabbing, jumping, burning.36 (78.3)30 (66.7)24 (66.7)8(57.1)98 (69.5)Method Firearm18(39.1)13 (28.9)17 (47.2)5 (35.7)53 (37.6) Hanging13 (28.3)11 (24.4)4(11.1)2 (14.3)30(21.3) Ingestion5 (10.9)9 (20.0)8 (22.2)2 (14.3)24 (17.0) Asphyxiation/CO poisoning1 (2.2)6(13.3)3 (8.3)3(21.4)13(9.2) Jumping4 (8.6)5(11.1)009 (6.4) Drowning3 (6.5)01 (2.8)1(7.1)5 (3.5) Other2 (4.3)1 (2.2)1 (2.8)04 (2.8) Cutting/Stabbing002 (5.6)1 (7.1)3(2.1)a Age is analyzed as a continuous variable, but is displayed by group to aid visualization.b Violent = firearm, hanging, cutting/stabbing, jumping, burning. Open table in a new tab TABLE 2Behaviors of suicide victimsAge, years21–3435–5455–7475–92TotalSIS, mean ±SDaInformation was available to derive a total SIS score on 136 subjects and establish presence or absence of previous suicide attempts in 137 victims. Data were available on 141 subjects for other variables.7.4 ± 2.78.3 ± 3.79.0 ± 3.110.2±3.58.4 ± 3.3Previous suicide attempt, n (%)23 (53.5)18 (40.9)14 (38.9)3(21.4)58 (42.3) Men21 (61.8)16(41.0)8 (28.6)045 (40.9) Women22 (22.2)2 (40.0)6 (75.0)3 (60.0)13(48.1)Warning in previous week, n (%)20 (46.5)14(31.8)12(33.3)3(23.1)49 (36.0)Warning in last month, n (%)25 (59.5)19 (44.2)16(47.1)5 (35.7)65 (48.9)Note: SIS = Suicide Intent Scale; SD = standard deviation.a Information was available to derive a total SIS score on 136 subjects and establish presence or absence of previous suicide attempts in 137 victims. Data were available on 141 subjects for other variables. Open table in a new tab Note: SIS = Suicide Intent Scale; SD = standard deviation. Table 2 lists the distribution of other suicidal behaviors in each age-group. Although there was no association between age and communication of intent to commit suicide within the last month of life, older victims were significantly less likely than younger victims to have given warning to others in the last week (χ2[1] = 5.43; P = 0.010). Whereas older men were less likely to have made previous attempts than their younger counterparts, older women were more likely than younger women to have attempted suicide in the past. The age × gender interaction was highly significant (χ2[1] = 9.11; P = 0.0013). Also, the presence of an affective disorder (χ2[1] = 4.08; P = 0.022) or schizophrenia diagnosis (χ2[1] = 10.90; P = 0.0005) significantly predicted previous suicide attempts. Older age at death was a significant predictor of higher SIS total scores (F[1,125] = 3.80; P = 0.026), indicating an association between age and lethality of intent in completed suicides. Of the covariates, only living alone had a significant independent association with the SIS score (F[1,125] = 9.06; P = 0.0016). The results of this study must be interpreted in light of its limitations. The psychological autopsy (PA) is a retrospective process, subject to potential recall and reporting biases of informants. The work of Brent and colleagues10Brent DA Perper JA Kolko DJ et al.The psychological autopsy: methodological considerations for the study of adolescent suicide.J Am Acad Child Adolesc Psychiatry. 1988; 27: 362-366Abstract Full Text PDF PubMed Scopus (122) Google Scholar supports the validity of the PA method in studies of completed suicide, but additional research is needed. To minimize bias, we took a number of steps that have been shown to improve the accuracy of informant-based diagnostic assessments. Multiple informants were interviewed as soon as possible after the death by use of standardized measures designed to assess the presence and severity of observable signs and behaviors.11Brent DA The psychological autopsy: methodological considerations for the study of adolescent suicide.Suicide Life Threat Behav. 1989; 19: 43-57Crossref PubMed Scopus (89) Google Scholar, 12Andreasen NC Rice J Endicott J et al.The family history approach to diagnosis.Arch Gen Psychiatry. 1986; 43: 421-429Crossref PubMed Scopus (460) Google Scholar Their information was supplemented by review of records, and diagnoses were made by consensus of investigators by use of all available data.11Brent DA The psychological autopsy: methodological considerations for the study of adolescent suicide.Suicide Life Threat Behav. 1989; 19: 43-57Crossref PubMed Scopus (89) Google Scholar, 13Beskow J Runeson B Åsgård U Psychological autopsies: methods and ethics.Suicide Life Threat Behav. 1990; 20: 307-323PubMed Google Scholar Suicide rates in the United States are higher among whites than nonwhites at all points in the life-course. The small number of suicide victims in this sample who were nonwhite, reflecting both the epidemiology of suicide and the demographic profile of Monroe County, prevented our inclusion of race as a variable in analyses. Although we are not aware of empirical data to suggest this possibility, the relationships between age and suicidal behavior demonstrated here could vary as a function of race. Our findings differ from previous research in demonstrating that older suicide victims were less likely than younger victims to have used violent methods to end their lives. Choice of method is determined in large part by ease of access, patterns of which may vary widely by region. Nonetheless, the less violent methods used by older people in this sample were equally deadly, perhaps because of the greater lethality with which they were used. Older age at death was a significant predictor of total SIS score, a reflection of greater planning and preparation for the act. Also, older victims were less likely to have been recognized as being at risk for suicide. Men were less likely to have had a history of previous attempts, and older suicide victims of both genders either did not communicate their intent to others as often as younger victims in the last week of life or were less likely to be heard. Recent evidence of increasing suicide rates among older people underscores the importance of this public health issue.14Centers for Disease Control Suicide among older persons: United States, 1980–1992.MMWR. 1996; 45: 3-6PubMed Google Scholar The findings suggest that older people at highest risk for completed suicide present special challenges for prevention that may warrant different approaches than used in persons at younger ages. As in younger adults, aggressive interventions should be immediately enacted when suicide risk is identified in older people. These should include referral for psychiatric evaluation, the institution of diagnosis-specific therapy, mobilization of resources in the patient's environment, and the use of any measures needed, including hospitalization, to guarantee safety through resolution of the suicidal crisis. However, such secondary prevention measures are likely to be less effective in decreasing suicide mortality among elderly persons than in younger people. Not only are elderly persons in a suicidal crisis more difficult to identify as being at imminent risk, but because their suicidal behavior is more planful and deliberate, the likelihood of successful intervention after its initiation is diminished. Efforts designed to treat conditions that predispose to development of the suicidal state promise to be more effective strategies over the long term in preventing suicide among elderly persons. Training of primary care providers to recognize and effectively treat affective illness,15Rutz W von Knorring L Wålinder J Frequency of suicide on Gotland after systematic postgraduate education of general practitioners.Acta Psychiatr Scand. 1989; 80: 151-154Crossref PubMed Scopus (263) Google Scholar the use of “gatekeepers” to increase recognition of vulnerable elders,16Raschko R Spokane Community Mental Health Center Elderly Services.in: Light E Lebowitz BD The Elderly with Chronic Mental Illness. Springer, New York1991: 232-244Google Scholar and development of outreach programs to decrease isolation among older people17DeLeo D Carollo G Buono MD Lower suicide rates associated with a Tele-Help/Tele-Check service for the elderly at home.Am J Psychiatry. 1995; 152: 632-634Crossref PubMed Scopus (110) Google Scholar are alternatives worth further study.