Using the psychological autopsy method, the authors sought to determine whether widowed people who commit suicide more than 4 years after their spouse's death (n=21) can be clinically distinguished from those who commit suicide after a shorter period of widowhood (n=14). The latter had a higher rate of psychiatric treatment (P=0.018), early loss/separation (P=0.03), and a nonsignificantly higher rate of lifetime substance abuse (P=0.07). Spousal bereavement increases the likelihood of physician visits, so the recently widowed represent a population for whom interventions may be readily implemented. Clinicians should monitor suicide risk in their recently widowed patients, especially those with psychiatric, substance abuse, and/or early loss/separation histories.
Spousal bereavement is a common event in late life. Cohort studies have established that the death of a spouse increases the risk of mortality in the survivor.
1
, 2
, 3
Suicide may be the most preventable cause of death among bereaved persons. Given that healthcare utilization increases after spousal loss4
, 5
and older suicide victims frequently see a physician in the month before their death,6
, 7
it is likely that many recently widowed individuals who commit suicide see healthcare professionals in the interval between their spouse's death and their own demise. Data on their clinical characteristics may thus aid efforts to prevent late-life suicide.At least six studies bearing on late-life suicide,
7
, , 9
, 10
, 11
, 12
some with psychological autopsy data13
, 14
have been published. None, however, has described the functional limitations, early loss/separation histories, and psychiatric diagnoses, symptoms, and treatment histories in a sample of widowed suicide patients. Given MacMahon and Pugh's finding that suicide risk is greater in the first 4 years of widowhood than thereafter, we describe bereaved suicide victims, attempting to distinguish those who died in the first 4 years of widowhood (“≤4;” recently widowed) and those suicide patients who were widowed for a longer period of time (“>4;” remotely widowed). On the basis of previous suicide and bereavement research,16
, 17
, 18
, 19
, 20
, 21
, 22
, 23
, 24
our primary hypothesis was that ≤4s would be more likely to have a history of separation or loss early in life. We also hypothesized that they would be more likely to 1) have a lifetime history of substance abuse; 2) have at least one pre-widowhood Axis I diagnosis; 3) have a history of psychiatric treatment; 4) have made at least one suicide attempt prior to widowhood. We did not expect any differences in functional status and depressive symptoms in the weeks before death.METHODS
Subjects
A complete description of our methods is presented elsewhere;
25
the present analyses include cases for whom psychological autopsy (PA) data were collected both before and after the sampling frame used in that publication. The population for study included all deaths among residents of Monroe County, NY, age 50 years and over, that occurred in the years 1988–1994 (inclusive), and were ruled suicides by the Office of the Medical Examiner (ME). Recognizing that any age cutoff is somewhat arbitrary, we focused on “older” widowed suicides primarily because there are substantial differences among older and younger widowed people,5
and we wanted to limit the variability that may be age-related. Of the 171 late-life suicides in Monroe County during this period, 52 were widowed at the time of death. We secured PA data on 35 widowed suicides (67%) with whom informants had contact during the last week of life; 10 cases were excluded because potential informants refused to participate; 6 were not pursued in 1988 because of inadequate staffing; and 1 was excluded because we were unable to locate the next-of-kin.Of the 35 victims, 18 (51%) were women, 34 (97%) were white, 21 (60%) lived alone, and 14 (40%) committed suicide within 4 years of spousal loss. Of the 17 exclusions, there were 12 women; 16 were white, and 10 people lived alone. The included and excluded victims were compared by use of available demographic information. No statistically significant differences in gender, race, or living situation were observed. Excluded cases were, on average, 7 years younger than those included (mean age=79.54±7.87 vs. 72.25±7.13 years; t[51]=3.17, P=0.003). For the included cases, information on length of widowhood was gathered in the PA interview. By using the Medical Examiner's report, we attempted to determine whether the excluded victims were >4s or ≤4s. Because of inadequate information, only 6 of the 17 excluded cases could be classified, of whom 4 were >4s. The possibility of sample bias as a function of widowhood duration cannot be ruled out, but the available information suggests that, if indeed there was a bias, its magnitude was relatively small.
Instruments
The Structured Clinical Interview for DSM-III-R (SCID).26
This widely used structured research instrument served to establish the victims' psychiatric diagnoses
27
and obtain information concerning age at onset of diagnosis and the presence of previous suicide attempts. In order to establish its validity with informants, we independently interviewed psychiatric inpatients and family informants (n=26 pairs). Kappa coefficients for the diagnoses of any substance use disorder, affective disorder, and their comorbidity ranged from 0.61 to 0.75. For the purpose of the analyses presented here, we combined subjects with abuse or dependence diagnoses because they may be difficult to discriminate by psychological autopsy, and the diagnoses made by DSM-III-R criteria reflect differences in severity rather than qualitatively distinct disorders.28
The Hamilton Rating Scale for Depression (Ham-D).29
An informant-report version of this 24-item scale was used with family members and other informants familiar with the victims' behavior and symptoms in their last week of life. All victims automatically receive a score of 4 on the suicide item.
Instrumental Activities of Daily Living (IADLS)30 and Physical Self-Maintenance (PSMS)30 Scales.
The IADLS assesses the degree to which people are able to manage the activities of daily living—laundry, housekeeping, cooking, etc. The PSMS assesses the extent to which people have difficulty managing basic self-maintenance activities, such as toileting, feeding, ambulating, and grooming.
Early loss or separation from one or both parents.
This variable was operationalized as separation from parents because of parental death, divorce, or separation before the subject's 18th birthday.
Procedures
Next-of-kin were contacted by letter 2 weeks after the subject's death and invited to participate in this project. Soon thereafter an in-person interview was conducted by two members of the research team who are mental health professionals. The interview lasted 2½ to 3 hours. Telephone interviews were conducted with physicians, other healthcare providers, and, when feasible and appropriate, with the victims' friends, acquaintances, coworkers, and relatives. Police, prison, medical, and educational records were obtained. The average number (±standard deviation [SD]) of informant sources was 4.3±1.9 for the ≤4s and 4.1±2.5 for the >4s. Consensus Axis I diagnoses were made on the basis of an integration of all data sources according to DSM-III-R criteria
27
in a meeting attended by members of the research team, the ME, deputy MEs, and the County's forensic toxicologist.Statistical analysis.
Fisher's exact test (FET) and univariate χ2 analyses were conducted to determine whether the ≤4s and >4s could be distinguished in terms of their gender, living situation, and rates of Axis I diagnoses, suicide attempts, psychiatric treatment, and history of early separation. Univariate t-tests were conducted to determine whether there were group differences in age, as well as in Ham-D, PSMS, and IADLS scores. A multiple logistic regression was conducted to determine whether the ≤4s and >4s could be distinguished on the basis of their loss/separation histories, after adjusting for age, gender, and substance abuse history—a clinical variable often associated with early loss and separation.
RESULTS
Basic demographic data are shown in Table 1. Given that the groups were defined on the basis of widowhood length, it is not surprising that the >4s were somewhat older. There was a greater proportion of women than men among the more recently bereaved victims, but no difference in the proportion living alone.
TABLE 1Demographic characteristics
| ≤4 Years (n = 14) | >4 Years (n = 21) | Statistic | |
|---|---|---|---|
| Aae, years, mean ± SD | 76.6 ± 7.6 | 81.5 +7.4 | t[33] = 1.81; P = 0.038 |
| Gender, female, n (%) | 11 (79) | 7 (33) | χ2[1] = 6.88; P = 0.009 |
| Living situation, alone, n (%) | 10 (71) | 11 (52) | χ2[1] = 1.27; NS |
a one-tailed.
Data on the five variables for which we had directional hypotheses are shown in the top rows of Table 2. Whereas significantly higher rates of psychiatric treatment were observed among the ≤4s (FET=0.018; one-tailed), there was no difference in the observed rate of prewidowhood Axis I diagnoses, suicide attempts (FET=0.098; one-tailed), or lifetime history of substance abuse (FET=0.072; one-tailed). However, the latter two findings may be interpreted as trends in the predicted direction. Finally, the ≤4s had a higher prevalence of early loss or separation. This was seen both in the univariate analysis (FET=0.030; one-tailed) as well as in the logistic regression (χ2[1]=2.58; P=0.054; one-tailed). As shown in the middle and bottom rows of Table 2, differences in the rates of nonsubstance diagnoses, or level of depressive symptoms, and functional impairment in the week before death did not approach significance.
TABLE 2Hypothesized vulnerabilities and other psychiatric diagnoses and symptoms
| ≤4 Years (n = 14) | >4 Years (n = 21) | P | |
|---|---|---|---|
| Hypothesized Indices of Vulnerability, n/N (%) | |||
| Substance use disorder | 5/14 (36) | 2/21 (10) | FET = 0.072 |
| Previous suicide attempt | 6/13 (46) | 4/21 (19) | FET = 0.098 |
| History of psychiatric treatment | 5/12 (42) | 1/20 (5) | FET = 0.018 |
| Prewidowhood Axis I diagnosis | 6/14 (43) | 6/21 (29) | FET = 0.304 |
| Early separation/loss | 6/14(43) | 2/21 (10) | FET = 0.030 |
| Other Psychiatric Diagnoses, n/N (%) | |||
| Major depression | 9/14 (64) | 14/21 (67) | FET = 1.00 |
| Somatoform disorder | 2/14 (14) | 0/21 (0) | FET = 0.15 |
| Dysthymia | 2/14 (14) | 1/21 (5) | FET = 0.55 |
| Other | 4/14 (29) | 4/21 (19) | FET = 0.68 |
| None | 1/14 (7) | 5/21 (24) | FET = 0.37 |
| More than one Axis I diagnosis | 6/14 (43) | 5/21 (24) | FET = 0.28 |
| More than one active Axis I diagnosis | 5/14 (36) | 3/21 (15) | FET = 0.22 |
| Symptoms, mean ± SD | |||
| Ham-D | 25.2± 10.3 | 24.0±7.9 | t[25] = 0.27; NS |
| IADLS | 8.9±6.8 | 8.2±6.5 | t[33] = 0.30; NS |
| PSMS | 3.1 ±2.6 | 2.6±3-l | t[33] = 0.51; NS |
Note: FET = Fisher's exact test; SD = standard deviation; Ham-D = Hamilton Rating Scale for Depression; IADLS = Instrumental Activities of Daily Living Scale; PSMS = Physical Self-Maintenance Scale.
a This diagnosis preceded bereavement in only two cases, both of whom were ≤4 s.
b Ham-D data were collected on 14 ≤4 s and 13 >4 s.
DISCUSSION
Psychological autopsy studies are uniquely positioned to generate new clinical knowledge and treatment and prevention strategies.
13
This study, which required 7 years of data collection, is the first to report on the characteristics of older widowed people who commit suicide. Its message must be carefully weighed against both its limitations and uncertainties, some of which are unavoidable. For example, questions have been raised about the possibility that some suicides are misclassified as natural or accidental. These questions cannot be definitively resolved. Underreporting of suicides is probably less likely to occur in jurisdictions with forensically trained medical examiners and investigators, such as Monroe County, NY, than in regions with appointed coroners.25
Two potential sampling problems may bear on the interpretation of the findings. First, with respect to potential sample bias, participants and nonparticipants in the current study did not differ in race, gender composition, or in the proportion who lived alone. However, victims for whom we had PA data were somewhat older than nonparticipants. Age was not the primary variable of interest in this investigation, but it is, of course, related to widowhood duration, suggesting that perhaps a greater proportion of ≤4 suicides than >4 suicides were excluded. The available data on widowhood duration are inconclusive. Similarly, it is unclear why women were overrepresented among the ≤4 suicides. Artifactual explanations must be eliminated before substantive hypotheses can be entertained. Unfortunately, given the paucity of information on length of widowhood among our excluded subjects, sample bias cannot be eliminated as one potential explanation. Replication in a contemporary, consecutive sample is warranted. Second, despite the lengthy period of data collection, the relatively small sample size may still have prevented the detection of significant differences or led to spurious findings. Rather than focus on any one result, therefore, we wish to emphasize the overall pattern of findings and its consistency with previous research and our hypotheses, most of which were supported.
As predicted, subjects who had been widowed for 4 years or less had higher rates of psychiatric treatment and early loss or separation. Findings on attempted suicide and lifetime history of substance abuse approached statistical significance in the predicted direction, but differences in the rates of prewidowhood Axis I diagnoses did not. Consistent with our predictions, there were no differences in rates of affective, somatoform, or anxiety disorders or scores on the Ham-D, IADLS, or PSMS, nor were there clinically important differences in the length of the first major depressive episode. Of the 23 victims who met criteria for active major depression at the time of death, 12 had been syndromic for less than 4 months (5/9 ≤4s vs. 7/14 >4s). Although there were no differences in the proportion of >4s and ≤4s with active Axis I diagnoses, their patterns of comorbidity may differ somewhat: recurrent major depression was diagnosed in all three of the >4s with psychiatric comorbidity and only one of the five ≤4s; substance disorders were present in one of the comorbid >4s and four of the comorbid ≤4s.
Our finding that early separation from parents was more prevalent in the ≤4s than >4s has implications for research on early loss and suicide.
31
Vulnerability to stressors in late life may be tied in part to earlier life events. Some people, because of their histories or propensities, are more affected by specific life events than others.32
The relationship between early loss and completed suicide is probably indirect, moderated by long-standing personality and attachment vulnerabilities, and precipitated by intervening events, such as spousal loss. With respect to the substance abuse findings, Murphy and Robins16
reported that disruptions of interpersonal relationships in alcoholic patients clustered in the 6 weeks before suicide. This pattern was not observed in a comparison sample of depressed suicide completers, but it has been confirmed in three independent studies17
, 18
, 19
and in a fourth that examined stressors 3 months before suicide.20
Substance abusers are apparently more likely to commit suicide after loss than those who do not abuse substances. The trend (P=0.07) for ≤4s to have a higher rate than >4s of lifetime substance abuse diagnoses may be interpreted to mean that, given an adequate sample size, Murphy and Robins's groundbreaking work16
may be extended to a different stressor (spousal loss) and time frame (4 years).The present data reveal significant yet predictable heterogeneity among late-life suicide victims, and illustrate the value of attempting to identify potential risk factors in specific demographic subgroups. Widowhood per se may not be a particularly compelling (or clinically useful) risk factor for suicide
11
or other cause of morbidity and mortality.24
Cohort studies identifying significant mortality after bereavement1
, 2
, 3
have not examined the moderating roles of psychiatric and developmental vulnerability, perhaps because the data have been unavailable. The current findings, however, suggest that risk is conferred by psychiatric symptoms that develop after bereavement in an already vulnerable population.24
, 33
, 34
This hypothesis may be worth pursuing by use of the psychological autopsy method, given that it has obvious implications for risk identification and intervention.Suicide prevention and risk recognition require the identification of particular risk factors in specific clinically or demographically defined groups. Recently widowed older patients with a previous psychiatric or developmental vulnerability may merit special attention. The recently widowed tend to see healthcare providers
4
, 5
and thus represent a population for whom interventions may be readily tailored and implemented. Clinicians should vigilantly monitor suicide risk in their recently widowed patients, especially those with histories of substance abuse and those with early loss histories.References
- Mortality after bereavement: a prospective study of 95,647 widowed persons.Am J Public Health. 1987; 77: 283-287
- Mortality after the death of a spouse: rates and causes of death in a large Finnish cohort.Am J Public Health. 1996; 86: 1087-1093
- Mortality following conjugal bereavement and the effects of a shared environment.Am J Epidemiol. 1995; 141: 1141-1152
- Secondary morbidity among the recently bereaved.Am J Psychiatry. 1986; 143: 158-163
- Bereavement and Health: The Psychological and Physical Consequences of Partner Loss. Oxford University Press, New York1987
- Suicide in elderly patients.in: Schneider LS Reynolds CF Lebowitz BD Diagnosis and Treatment of Depression in Late Life: Results of the NIH Consensus Development Conference. American Psychiatric Press, Washington, DC1994: 397-418
- Suicide in the elderly: recent developments in psychogeriatrics.Br J Psychiatry supp. 1971; 6: 87-97
- The Final Months. Oxford University Press, New York1981
- Suicide in later life: psychological autopsy findings.Int Psychogeriatr. 1991; 3: 59-66
- Secular trends in psychiatric diagnoses of suicide victims.J Affect Disord. 1991; 21: 127-132
- Suicide over 60: The San Diego Study.J Am Geriatr Soc. 1994; 42: 174-180
- Mental disorders in elderly suicide.Int Psychogeriatr. 1995; 7: 275-287
- Assessment in absentia: the value of the psychological autopsy method for studying antecedents of suicide and predicting future suicides.in: Maris RW Berman AL Maltsberger JT Assessment and Prediction of Suicide. Guilford, New York1992: 144-182
- Some clinical considerations in the prevention of suicide based on a study of 134 successful suicides.Am J Public Health. 1959; 49: 888-899
- Suicide in the widowed.Am J Epidemiol. 1965; 81: 23-31
- Social factors in suicide.JAMA. 1967; 199: 303-308
- Suicide and alcoholism: interpersonal loss confirmed as a predictor.Arch Gen Psychiatry. 1979; 36: 65-69
- San Diego Suicide Study, III: relationships between diagnoses and stressors.Arch Gen Psychiatry. 1988; 45: 589-594
- Interpersonal stressors, substance abuse, and suicide.J Nerv Ment Dis. 1993; 181: 80-85
- Differences in recent life events between alcoholic and depressive nonalcoholic suicides.Alcohol Clin Exp Res. 1994; 28: 1143-1149
- A study of suicide in the Seattle area.Compr Psychiatry. 1960; 1: 349-359
- Broken homes and attempted and completed suicide.Arch Gen Psychiatry. 1965; 12: 213-216
- Recent bereavement in relation to suicide.J Psychosom Res. 1972; 16: 361-366
- Traumatic grief as a risk factor for mental and physical morbidity.Am J Psychiatry. 1997; 154: 616-623
- Relationships of age and Axis I diagnoses in victims of completed suicide: a psychological autopsy study.Am J Psychiatry. 1996; 153: 1001-1008
- Structured clinical interview for DSM-III-R–Patient Version. Biometrics Research Department, NY State Psychiatric Institute, 1986
- Diagnostic and Statistical Manual of Mental Disorders. 3rd Edition. American Psychiatric Association, Washington, DC1987 (Revised (DSM-III-R))
- Alcohol abuse and dependence.in: Robins LN Regier DA Psychiatric Disorders in America: The Epidemiologic Catchment Area Study. The Free Press, New York1991
- A rating scale for depression.J Neurol Neurosurg Psychiatry. 1960; 23: 56-62
- Assessment of older people: self-maintaining and instrumental activities of daily living.Gerontologist. 1969; 9: 179-186
- Environmental, psychosocial, and psychoanalytic aspects of suicidal behavior.in: Blumenthal SJ Kupfer DJ Suicide Over The Life Cycle: Risk Factors, Assessment, and Treatment of Suicidal Patients. American Psychiatric Press, Washington, DC1990: 39-96
- Personality continuity and change across the life course.in: Pervin LA Handbook of Personality: Theory and Research. Guilford, New York1990: 549-575
- Past personal history of dysphoria, social support, and psychological distress following conjugal bereavement.J Am Geriatr Soc. 1994; 42: 712-718
- Depression through the first year after the death of a spouse.Am J Psychiatry. 1990; 148: 1346-1352
Article info
Publication history
Accepted:
March 17,
1998
Received in revised form:
February 9,
1998
Received:
September 17,
1997
Footnotes
Previous versions of this paper were presented at the Annual Meeting of the American Association of Suicidology, St. Louis, MO, April 26, 1996; the 6th European Symposium on Suicidal Behavior, Lund, Sweden, June 14, 1996; and the 49th Annual Scientific Meeting of the Gerontological Society of America, Washington, DC, November 20, 1996.
Josephine Lauri assisted in manuscript preparation. The authors also acknowledge the assistance of John H. Herrmann, Diane Gill, Andrea DiGiorgio, Jill Eichele, other LSS associates, and the personnel of the Office of the Medical Examiner of Monroe County for their help in conducting this study.
This study was supported, in part, by Public Health Service grants P30-MH40381, T32-MH18911, K07-MH00748, and K07-MH01135.
Identification
Copyright
© 1998 American Association for Geriatric Psychiatry. Published by Elsevier Inc.
User license
Creative Commons Attribution – NonCommercial – NoDerivs (CC BY-NC-ND 4.0) | How you can reuse 
Elsevier's open access license policy
Creative Commons Attribution – NonCommercial – NoDerivs (CC BY-NC-ND 4.0)
Permitted
For non-commercial purposes:
- Read, print & download
- Redistribute or republish the final article
- Text & data mine
- Translate the article (private use only, not for distribution)
- Reuse portions or extracts from the article in other works
Not Permitted
- Sell or re-use for commercial purposes
- Distribute translations or adaptations of the article
Elsevier's open access license policy
