|Home | About | Journals | Submit | Contact Us | Français|
Assessing predictors of suicide and means of completion in patients with dementia may inform the development of interventions to reduce suicide risk among the growing population of individuals with dementia.
This national, retrospective, cohort study used data from the Department of Veterans Affairs (fiscal years 2001–2005). The sample included patients ≥ 60 years old with the diagnosis of dementia (N=294,952), of which 241 died by suicide. Potential predictors of suicide were identified using logistic regression. Suicide methods are also reported.
Increased suicide risk was associated with white race (OR 2.4, 95% CI: 1.2, 4.8), depression (OR 2.0, 95% CI: 1.5, 2.9), a history of inpatient psychiatric hospitalizations (OR 2.3, 95% CI: 1.5, 3.5), and prescription fills of antidepressants (OR 2.1, 95% CI: 1.6, 2.8) or anxiolytics (OR 2.0, 95% CI: 1.5, 2.7). Nursing home admission was associated with lower suicide risk (OR 0.3, 95% CI: 0.1, 0.8). Severity of medical comorbidity did not impact suicide risk. Sensitivity analysis indicated that the majority of suicides occurred in those with new dementia diagnoses. Firearms were the most common suicide method (73%).
Given the higher rate of suicide in those receiving treatment for psychiatric symptoms and the high proportion that died using firearms, closer monitoring and assessment of gun access may be an important part of initial treatment planning for older male patients with dementia, particularly those with symptoms of depression or anxiety.
In many countries, including the United States, older adult men are at greater risk for suicide than other segments of the population. Already a significant public health concern, the number of suicides in later-life is expected to increase as the population ages. Developing effective interventions to reduce suicide risk among elders, requires the identification of risk and protective factors.(1) To date, several risk factors for later-life suicide have been identified, the most significant of which is the presence of a major psychiatric illness.(2) Major depression is the psychiatric disorder most strongly associated with suicide in the elderly, although psychotic disorders, anxiety disorders and substance abuse also increase risk, (3) as does poor physical health status and impaired functional capacity.(4)
Dementia is a prevalent disorder of later life and is associated with neuropsychiatric symptoms such as depression, psychosis and anxiety that have been themselves identified as suicide risk factors. Although the risk of dementia increases with age (5), the role of dementia as an independent risk factor for suicide is not well understood. While some investigators have reported increased risk for suicide among older adults with dementia diagnoses, especially early in the course of dementia (6, 7), others have found no association between cognitive impairment and subsequent death by suicide in later life.(8) A recent review concluded that the risk of suicide in dementia appears to be the same as that of the age-matched general population, but noted that many studies had significant methodological limitations.(9)
Although studies to date are mixed regarding the role of dementia itself as a risk factor for suicide, investigation of predictors of suicide among patients with dementia is warranted given the prevalence of dementia among elders, particularly the “oldest old”. Assessing risk factors within the population of patients with dementia such as demographics, the presence of associated neuropsychiatric symptoms, and use of health services may help researchers to understand which patients are particularly at risk and who might need closer monitoring and assessment as part of dementia care. Assessing suicide risks by stage of dementia may also be helpful in targeting interventions. Traditionally, suicide risk has been thought to increase at the time of diagnosis and during the early stages of dementia, when patients have fears of future physical and mental decline and concerns about burdening their families emotionally and financially. In contrast, suicide risk may decrease when dementia becomes advanced and cognitive and functional impairments diminish both insight and the ability to prepare and execute a plan.(10)
Further investigation regarding the relationship between dementia and suicide is warranted given the aging population and the importance of both dementia and suicide as public health issues. The VA provides a useful context in which to examine these critical issues for a number of reasons, including: 1) a large aging patient population; 2) VA high quality pharmacy and health services data that can be linked to other critical large-scale data resources; and 3) the ability of the VA to translate research findings into clinical practice enhancements. The current study aims to examine potential predictors of suicide in older Veteran patients diagnosed with dementia.
In this study, we compared VA patients with dementia who committed suicide during the study period vs. those who did not by demographic characteristics, medical comorbidity, health care utilization and medication use variables. In addition, we examined the relationship between dementia severity and suicide and the methods used by those who killed themselves.
Patients aged 60 years or over who had received a dementia diagnosis in VA healthcare settings between fiscal year 2001 through 2005 (FY01 to FY05) were identified in the Department of Veterans Affairs (VA) National Care Patient Database (NCPD). Vital status and cause of death for these patients were obtained from the National Death Index (NDI). The NDI compiles death record data for all US residents from state vital statistics offices. The NDI has the greatest sensitivity in determining vital status among population-level sources of mortality data and is considered the “gold standard” for mortality assessment.(11) Established procedures were used to identify “true” matches in cases where the NDI search resulted in multiple records as potential matches.(12)
The VA Ann Arbor Healthcare System Institutional Review Board approved the study. All identifiers were removed to assure confidentiality.
The final sample included 294,952 patients.
Data on gender, marital status, race, ethnicity, and age of the study population were obtained at cohort entry from the NPCD. Age categories included 60–69, 70–79, 80–89 and 90+. Ethnicity was defined as Hispanic or not, and race defined as white, African-American, or other. “Other” race included patients of Asian, Native Hawaiian, Pacific Islander, or Native American race/ethnicity and patients who were multiracial.
We identified patients with dementia diagnoses using codes from the International Classification of Diseases, tenth revision, selected for identifying dementia cases by a collaborative effort of the VA Offices of Geriatric Research & Evaluation and Geriatrics and Extended Care. These codes have also been used in previous studies published by our group. (13, 14) As one of the aims of this study was to examine stage of dementia as a risk factor, an inclusive coding scheme was used in order to capture the earliest stages of dementia where less-specific diagnoses such as “memory loss” may be assigned. For a sensitivity analysis, we also grouped patients as having specific (e.g. Alzheimers Disease) vs. less-specific dementia diagnoses (e.g., memory impairment). Specific dementia diagnoses were represented by the ICD 9 diagnoses 0461, 0463, 290.0, 290.1x, 290.2x, 290.3, 290.4x, 291.2, 294.10, 294.11, 331.0, 331.1, and 331.82. Less specific dementia diagnoses included ICD 9 diagnoses 2948, 2949, 3100, 3101, 3312, 3313, 3314, 3317, 33189, 3319, and 78093.
Suicide deaths during the study period were identified and categorized using the National Death Index (15) and included ICD-10 codes X60–X84, Y87.0, and U03.
Comorbid psychiatric disorders were identified at cohort entry and included PTSD, any other anxiety disorder, depression, and any personality disorder. As in prior studies by our group (16), the diagnosis of depression was inclusive (ICD-9 codes 296.2x, 296.3x, 296.90, 296.99, 298.0, 300.4, 311, 293.83, 301.12, 309.0, or 309.1) given the fact that the majority of later-life depression is treated in primary care settings, and less specific depression diagnoses are often used in primary care settings. The presence of any substance use disorder (except nicotine dependence) and indicators of schizophrenia and bipolar disorder were also obtained. We used a modified version of the Charlson Comorbidity Index (17) to measure medical comorbidity with the original Charlson weights based on 18 medical comorbidities (excluding dementia). In the final model, we used indicators for comorbidity and included only those showing statistically significant associations or meaningfully large coefficients.
Health care use was determined from the NPCD. We examined use of inpatient (IP), outpatient (OP) and nursing home (NH) services during the period of 6 months prior to patients’ entry into the cohort. Health care use was divided into psychiatric (psych) and medical/surgical (nonpsych) care.
Medication use was determined from outpatient VA pharmacy records during the period of 6 months prior to patients’ entry into the cohort. Medications of interest included anti-anxiety agents, cholinesterase inhibitors, antidepressants and antipsychotics.
Given that dementia gradually worsens over time and that changes can generally be seen on a yearly basis, we created a proxy for dementia severity (a variable unavailable in the administrative record). Using a subset of veterans who were found to be consistent users of VA healthcare in the three years prior to the study period, we produced a “time since dementia diagnosis” variable. The latter was defined as the time between the initial appearance of a dementia diagnosis in our administrative data and the patient’s index diagnosis during the study period (FY01-05).
Categorical differences between the two study groups (patients with dementia who died from suicide vs. the remainder of the total patient sample with dementia) in demographic variables, comorbidity characteristics, healthcare utilization, and psychiatric medication use were assessed by means of chi-square tests of independence. Subsequently, multivariate logistic regression models were used to identify potential predictors of suicide. Independent variables included in the multivariate logistic model were sex, age, race, marital status, psychiatric comorbidities, medical comorbidities, healthcare utilization, and medication use. We used SAS 9.20 to perform all analyses.
A total of 294,952 patients aged 60 and over who received the diagnosis of dementia within the VA during the study period were eligible for study inclusion. Given the VA population, the vast majority of the sample was male (97.2%). Almost half of the overall sample was white (47.2%), and a majority were married (62.9%).
Among these patients, 241 (.09%) died by suicide during the study period.
The demographic characteristics of the overall patient sample and the subsample of patients who died from suicide are presented in Table 1, along with bivariate statistics. In bivariate analyses, younger age and white race were associated with higher rates of suicide.
Table 2 presents the comorbidities of both the overall cohort as well as for the suicide vs. non-suicide subsamples. With regard to comorbid psychiatric illness, patients who died by suicide were more likely to have histories of substance abuse, bipolar disorder, anxiety disorders (excluding PTSD) and depression.
The non-suicide group had higher rates of overall medical burden than the suicide group (defined as Charlson score of greater than 1; suicide group=32.0% vs. non-suicide group= 41.3%, χ2=11.81, p<0.003). The non-suicide group also had higher rates than the suicide group on three of the specific disease groupings (cerebrovascular disease, χ2=4.10, p<0.05; diabetes with complications, χ2=4.50, p<0.04; and renal disease χ2=5.94, p<0.02; Table 2).
Table 3 shows health care utilization by study group. A significantly greater number of those who died by suicide had previous inpatient (χ2=70.71, p<0.0001) or outpatient (χ2=69.68, p<0.0001) psychiatric care, whereas significantly fewer had nursing home care (χ2=8.47, p<0.004).
Psychotropic use by study group is shown in Table 4; psychotropic use was higher for every class of medications within the suicide group. Most notably, 61.4% of this group had been prescribed an antidepressant as compared to 30.9% of the non-suicide group (χ2=105.15, p<0.0001).
In multivariate analyses (Table 5), white patients remained at greater risk of dying from suicide (OR= 1.49, 95% CI 1.14–1.95). Among psychiatric diagnoses, only a history of depression placed patients at significantly higher risk of dying by suicide (OR 2.04, 95%CI 1.45, 2.85). Almost 25% of the suicide group had a diagnosed depression, as compared with approximately 9% of the non-suicide group. Schizophrenia was associated with a significantly lower risk of suicide. In the multivariate analyses, none of the relationships between suicide and physical comorbidities retained significance. A history of inpatient psychiatric care was significantly associated with increased risk of suicide (OR 2.31, 95% CI 1.54, 3.46) while an inpatient nursing home stay which was associated with a lower risk of suicide (0.33, 95% CI 0.14, 0.75). Two psychiatric medication classes remained significantly associated with suicide in the multivariate model: anti-anxiety medications (OR 1.98, 95% CI 1.48, 2.65) and antidepressants (OR 2.11, CI 1.57, 2.84).
Using NDI data, we were able to examine methods of suicide among patients who took their lives during the study period. The method used in the vast majority of suicides was a firearm (72.6%). The next most common methods were self-poisoning and hanging (9.5% each). Much less common means used in suicide were harm with a sharp object (2.9%), jumping from a high place or moving object (2.4%), drowning (1.2%) or self-immolation (0.4%).
We examined the likelihood of suicide based on a patient’s stage of dementia at time of cohort entry. This analysis was performed using a subgroup of patients (n=136) with consistent VA healthcare utilization in the three years prior to cohort entry. Time since first dementia diagnosis in the three years prior to study entry was categorized as 1) zero years between initial diagnosis and cohort entry; 2) one year between initial diagnosis and cohort entry; 3) two years between initial diagnosis and cohort entry; or 4) three years between initial diagnosis and cohort entry. In this categorization, a patient with a first diagnosis of dementia occurring during the study period could be construed as a new dementia diagnosis, whereas a patient who had a diagnosis of dementia occurring three years prior to study entry could be construed as reflecting a more advanced stage dementia. The results indicated that the majority of suicides (75.0%) occurred in those with new dementia diagnoses (e.g. diagnosis during the study period).
In additional sensitivity analyses, we examined patients with specific dementia diagnoses (e.g. Alzheimer’s, vascular dementia) as compared to those with less-specific dementia diagnoses (e.g. chronic organic brain syndrome, memory loss). About 56% of cohort had a less-specific diagnosis. Having a specific versus a non-specific dementia diagnosis did not impact the risk for suicide (OR= .806, 95% CI: 0.618, 1.051), controlling for other variables in the model. Given this, analyses for individual dementia diagnoses were not performed.
This large national cohort study allowed the identification of several risk factors for suicide in dementia that are consistent with findings of previous research. Risk factors for suicide in dementia previously described include depression and younger age (9). As anticipated, we found comorbid depression to be associated with substantially increased suicide risk. In contrast, comorbid schizophrenia was found to be a potentially protective factor among elder with dementia. Our results further suggest that the majority of suicides were among patients receiving psychiatric treatment. Additionally, patients who committed suicide in our study were younger than those who did not, and the majority of suicides appeared to occur in those with new dementia diagnoses. However, unlike many prior studies, concurrent medical comorbidities were not found to increase risk. While prior work regarding suicide and self-injurious behaviour in nursing homes has been inconclusive, we found a decreased risk of suicide in patients with nursing home admissions. In terms of methods of committing suicide, the vast majority of patients with dementia took their lives using a firearm (73%).
Patients in our study who died by suicide were younger than those who did not. This is consistent with previous research in the VA population (18) but at odds with epidemiological observation in the general population over age 60 years, in which risk for suicide in men rises continuously through old-old age. In this sample of veterans with dementia, the greater risk with young-old age could reflect several factors. Those with earlier onset Alzheimer’s type dementia may have a more aggressive, neuropsychiatrically complex illness. Younger age of onset carries a higher likelihood of a family history of dementia which may influence the patient’s outlook on the future. There is growing evidence that depression and anxiety are more likely to occur early in the course of dementia (19). Anxiety related to the diagnosis of dementia at an earlier age may drive suicidal behaviour.
The relationship between depression and later-life suicide is well established. (20) Perhaps not surprisingly, the individuals in this study who died by suicide were significantly more likely to have been diagnosed with depression. Prescription of an antidepressant was as strong a predictor as having a diagnosis of depression. Interestingly, however, more patients were prescribed antidepressants than actually carried the diagnosis of depression. The difference between the number of patients with a depression diagnosis (27,079 total in cohort) and those prescribed antidepressants (91,142) may represent either an underreporting of depression diagnoses or may represent the prescription of antidepressants for reasons other than affective disorder, such as pain or sleep
Prescription of an anti-anxiety medication was also a strong predictor of suicide. Recent work by Pfeiffer et al. (21) found co-morbid anxiety disorders and fills of an anti-anxiety medications was associated with suicide risk among a mixed aged cohort of depressed patients. They also found that prescription of an anti-anxiety medication was a stronger predictor of completed suicide than any anxiety diagnosis consistent with our findings. As expected, we found that patients who died from suicide were far more likely to have had psychiatric care (both inpatient and outpatient). Given that a number of patients with dementia and neuropsychiatric symptoms might not receive a separate comorbid psychiatric diagnosis, our findings suggest that those dementia patients with accompanying neuropsychiatric symptoms (indicated either by an accompanying diagnosis of depression or receipt of antidepressants or anxiolytics) are at higher risk for suicide than those with cognitive symptoms only.
In contrast to many prior studies, concurrent medical comorbidities were not found to increase suicide risk. In perhaps the largest study of this issue, Waern et al. (22) found that serious physical illness was independently associated with suicide in elderly people. Conwell et al. (4) recently reported similar findings, noting that physical illness and associated functional impairments were more common in older adults who committed suicide. However, a limitation of this study is that we were unable to assess the impact of disability resulting from medical comorbidities which represents a potential confounder in the medical comorbidity-suicide relationship.
A noteworthy finding in our study was that the majority (75%) of suicides occurred in those with a new dementia diagnosis. This finding would seem to confirm both the conventional wisdom as well as limited research (6, 7) that the most significant concern for suicide in dementia is early in the course of illness. Placed in this context, our finding that a history of nursing home care decreased the likelihood of suicide significantly could be expected. Although data concerning suicidal behaviour in nursing home residents is limited and contradictory (23, 24), the very structured, supervised nature of such facilities and the limited access to firearms may limit suicidal behaviours. It is also possible that this group of patients represents those with more severe cognitive and/or physical limitations who would thus have more difficulty preparing and executing a suicide plan. Similarly, we found that a diagnosis of schizophrenia, an illness also associated with executive and planning deficits, was potentially protective.
Several limitations are inherent in the use of registry data. Given the VA population, the cohort was primarily male, and the study results may not be generalizable to other populations. Because registry dementia diagnoses are clinical, cases not yet diagnosed with dementia (likelier earlier cases where the risk of suicide may be the greatest) cannot be ascertained. However, our definition of “dementia” was purposefully inclusive in order to capture as many early cases as possible. A final limitation of our study is the use of a proxy as an indicator of dementia severity, i.e. time since dementia diagnosis. To the extent that this might not be an adequate proxy, our analyses must be interpreted with caution. However, given that the existing literature suggests that suicide is an uncommon event in patients with dementia, and that recruiting the large numbers of patients with mild cognitive impairment and dementia necessary for prospective, longitudinal studies may be difficult, large registry studies such as ours are important to address potential risk factors.
At present, our findings suggest that patients with accompanying neuropsychiatric symptoms of depression and anxiety in psychiatric treatment who are early in their course of dementia are at greatest risk for suicide. Timely identification and intervention addressing the complex issues of depression and dementia in these patients may help to mitigate their increased risk. Our findings also support the growing evidence that restricting access to firearms is one of the most effective strategies for suicide prevention.(25) Given the high rate of suicide by firearms, closer monitoring and assessment of access to firearms may be an essential part of treatment planning for older male patients with dementia.
This research was supported by grants from the Department of Veterans Affairs, Health Services Research and Development Service, IIR 04-211-1 and by the National Institute of Mental Health, R01-MH078698-01. Resources were also contributed by the Serious Mental Illness Treatment, Research, and Evaluation Center, Ann Arbor, MI. The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.