| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
Correspondence: Address correspondence to Carl I. Cohen, MD, Division of Geriatric Psychiatry, SUNY Downstate Medical Center, Box 1203, 450 Clarkson Avenue, Brooklyn, NY 11203. E-mail: carl.cohen{at}downstate.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Suicide • African Americans • Elderly • Race • African Caribbeans
There have been few population-based studies examining risk factors for suicide in older African American populations. An exploration of such factors is critical for guiding prevention and service strategies (Cook et al., 2002). In examining risk for suicide, the literature has included factors affecting active ideation (i.e., wanting to die) and passive ideation (i.e., thoughts pertaining to death). The combination of these two factors has been termed suicidality (Cook et al., 2002). Gallo, Cooper-Patrick, and Lesikar (1998) conducted the only representative population-based study of racial differences in suicidality using Epidemiologic Catchment Area (ECA) data. They found that Blacks had more passive ideation but less active ideation than Whites. Using multivariate analyses, Cook and coauthors found that suicidality among older African Americans living in public housing developments in Baltimore was associated with life satisfaction, religiosity, and depression. There have been several studies of mixed racial samples of older adults in clinical settings that have identified a variety of social and clinical factors associated with suicidality (e.g., female gender, social support, religiosity, life satisfaction, depression, anxiety, alcoholism, physical illness, cognitive deficits, and help-seeking behavior; viz., Bartels et al., 2002; Callahan, Hendrie, Nienaber, & Tierney, 1996; Lish et al., 1996). Callahan and colleagues found no racial differences in suicidality, whereas the other two studies found suicidality to be lower among Blacks than Whites.
Each of the earlier studies has had some limitations. Although the ECA study provided representative population-based data, it was conducted more than two decades ago, and since then, the composition of the Black population within the United States has been altered by the influx of more than 2 million African Caribbean immigrants (Pierre-Pierre, 1993). Moreover, the ECA study did not provide multivariate data with respect to risk factors for suicidality. The other large-scale community study (Cook et al., 2002) was restricted to a subgroup of elderly adults in urban housing projects living mostly alone and with few financial resources. Consequently, the findings are probably unrepresentative of the older African American population in the general community. Other studies of suicidality in older adults that have included large numbers of African Americans have been limited to nursing homes or psychiatric or primary care settings. Among the latter, the largest study and potentially the most generalizable (Bartels et al., 2002) excluded individuals already receiving outpatient specialty mental health treatment and who were presumably among those most at risk for suicide. Moreover, none of the studies specifically examined risk factors within each racial group that might account for these differences.
The aim of this article is to use epidemiological data of older African Americans and Caucasians living in a large urban community to compare those factors associated with active or passive suicidal ideation (hereafter, suicidality) in each racial group. We adapted George's (1994) social antecedent model of depression in older persons as the theoretical framework for the rational inclusion of the predictor variables. She postulated six stages of risk factors, with later stages hypothesized to be increasingly proximate precursors of depression. The first three stages consist of demographic factors (e.g., race, gender, ethnicity), early events and achievements (e.g., education, early traumatic events), and later events and achievements (e.g., financial status, later traumatic events). The fourth and fifth stages consist of social integration and support and various vulnerability factors (e.g., chronic stressors such as physical illness, alcoholism, mental distress). The sixth stage consists of provoking agents (e.g., acute life events) and coping strategies (e.g., beliefs and behavior). Thus, the model has the potential to test the concurrent effects of social and clinical variables.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Our study design called for interviewing a minimum of 200 individuals in each of four ethnic groups: Caucasians, U.S.-born African Americans, African Caribbeans from English-speaking islands, and African Caribbeans from French-speaking islands. Data collection took place from 1996 to 1999. We made an effort to interview all eligible individuals in a selected block group by knocking on doors. To enhance response rates, we also recruited participants from the selected block group at senior centers, churches, and through personal references; 30% of the sample came from these sources. The overall response rate was 77% of those contacted. The number of individuals contacted represented about 4% of the potentially eligible population according to census data.
At the beginning of the interview, we administered the Mental Status Questionnaire (Kahn, Goldfarb, Pollack, & Peck, 1960) in order to exclude individuals with moderate to severe cognitive impairment (i.e., scores of 4 or less; Zarit, Miller, & Kahn, 1978). The final community sample consisted of 860 Blacks and 214 Caucasians. The weighted totals based on the 1990 census were 440,616 individuals, of whom 122,988 were Black and 317,628 were Caucasian. The weighted totals by gender were 175,120 men and 265,496 women.
After providing a complete description of the study to the participants, we obtained written informed consent. The State University of New York Downstate Medical Center, Brooklyn, New York, Institutional Review Board approved this study.
Instruments and Variables
As described in the introduction, our review of the literature identified a variety of social and clinical variables that may be associated with suicidality in later life (Baker, 1996; Callahan et al., 1996; Cook et al., 2002; Lish et al., 1996). George's (1994) theoretical model also posits that coping strategies might moderate the effects of some of these factors, although none of the earlier studies had specifically examined these variables. Thus, our model was the scaffolding on which to incorporate the relevant variables described in the literature along with several additional coping variables that we postulated might be important.
We operationalized 1 dependent variable and 19 predictor variables (see Table 2) that we derived from the following instruments: (a) Center for Epidemiologic Studies–Depression scale (CES-D; Radloff, 1977), a 20-item measure of depression; (b) Anxiety Status Inventory (ASI; Zung, 1971), a 20-item scale that has been validated for use with older adults (Sheikh, 1991) with converted scale scores ranging from 25 to 100; (c) a "physical illness" score that represented the sum of 29 physical health items derived from the Multilevel Assessment Inventory Physical Health Status, with possible scores ranging from 29 to 60 (lower scores indicate worse health; Lawton, Moss, Fulcomer, & Kleban, 1982); (d) Instrumental Activities of Daily Living Scale, with possible scores ranging from 17 to 54 (lower scores indicate more impairment; Lawton et al., 1982); (e) Conflict Tactics Scale (Strauss, 1979), which we subdivided into seven subscales of responses to how the participant might handle a dispute with a close relative or friend (prays, discusses things calmly, keeps feelings inside, hits, shoves, hurts, or uses a weapon) using principal components analysis with varimax rotation; (f) Pearlin Coping Scale (Pearlin, Mullan, Semple, & Skaff, 1990), a measure of what the participant would typically do in response to stress or emotional upset, which we subdivided into four scales using principal components analysis with varimax rotation (uses medications and/or seeks professional help, uses distraction, accepts, finds meaning); (g) Trauma and Victimization Scale (Cohen, Ramirez, Teresi, Gallagher, & Sokolovsky, 1997), which examined 18 lifetime traumatic events (e.g., physical abuse, sexual assault, physical assault, witnessing violence), with scores ranging from 18 to 72 (higher scores indicate more trauma); (h) Acute Stressors Scale (Chatters, 1993), which examined 11 negative life events such as health, money, family, or crime that may have occurred in the past month, with possible scores ranging from 11 to 44 (higher scores indicate more stressors); (i) Financial Strain Scale (Pearlin, Lieberman, Menaghan, & Mullan, 1981), with possible scores ranging from 9 to 21 (lower scores denote more strain); (j) a three-item Religiosity Scale, which we created based on the frequency of church attendance, intensity of religious feelings, and whether God was perceived as a source of strength (possible scores ranged from 3–16); and (k) Network Analysis Profile (Sokolovsky & Cohen, 1981), which generated variables concerning network size, intimacy, reliability of contacts, advice giving, and material support. "Spiritualist" consisted of two items based on whether the participant saw a "spiritualist or neighborhood healer" or "purchased items from a botanica or religious shop." Participants received a score of 1 if they responded positively to one of the items and 2 if they responded positively to both items. We determined the presence of possible alcohol problems if the participant responded positively to any of the four CAGE (Ewing, 1984) items, past or present. We based "significant loss" on whether a significant person had died or moved away in the past 5 years. We based the "entitlement" score on the number of personal resources such as various types of health insurance, income support, and various social service supports. We also included a variety of demographic items and an item that queried as to whether the participant had used a physician or psychiatrist (any medical doctor) for "mental health reasons in the past year." We did not include cognitive status in our model because we had eliminated cognitively impaired individuals from the study sample.
|
We derived the dependent variable from six items: "currently (past 2 weeks) wanted to die," "ever wanted to die," "currently thought about suicide," "ever thought about suicide," "currently attempted suicide," and "ever attempted suicide." The first two categories composed passive suicidality, and the latter four categories composed active suicidality. We combined the six items forming the passive and active categories to form the dependent variable (suicidality). We opted to use lifetime prevalence (current or ever) of suicidality as the dependent variable because we felt that the low prevalence of current suicidality (4.8%) might distort the statistical analyses; lifetime prevalence was 13.5%. Including current and past suicidality seemed to be a reasonable option because many studies have found strong associations between past, current, and future suicidality (Stong, 2006). In order to compare our results with the ECA data (Gallo et al., 1998), we also examined whether the participant currently or had ever "thought a lot about death."
For Haitian participants who were not fluent in English, we had the questionnaire translated into Creole, and then back-translated according to the methods described by Flaherty and colleagues (1988). Interviewers were trained using audio- and videotapes, and they were generally matched to participants from similar ethnic backgrounds. Interrater reliability using the intraclass correlation ranged from.86 to 1.0 for the scales in this sample. We periodically monitored interviewers by using audiotapes.
Data Analysis
In order to control for intrablock clustering effects and for without-replacement sampling, we performed data analysis using Survey Data Analysis (SUDAAN) 7.5.4A (Research Triangle Institute, 2000). The unit of analysis was the individual. We conducted all analyses using sampling weights. With respect to the selection of independent variables, for measures that generated more than one variable, we conducted preliminary bivariate analyses and selected one or more variables from that measure that had substantial association (p <.10) with any of the dependent variables (preliminary analyses not shown). For the final analyses, we included three demographic variables: gender, age, and education. Based on these preliminary analyses, we selected two coping subscales—copes by keeping calm (possible scores ranging from 3–21) and copes with stress or emotional upset by using medications and/or seeks professional help (possible scores range from 2–4)—from the Conflict Tactics Scale and Pearlin Coping Scale, respectively. In order to test for interactions with race, we created four interactive variables between race and several key variables (viz., CES-D scores, ASI scores, religiosity, and copes by using medications) using centroid transformations. We selected these variables for interaction with race because they were significant in a preliminary logistic regression examining predictors of suicidality for the entire sample. Because of the high correlations between the CES-D and ASI scores (r =.76), we could not enter these variables simultaneously in the logistic regression analyses, and so we examined them in separate analyses. Among the remaining selected variables entered into the logistic regression analyses, there was no evidence of collinearity.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
Using logistic regression, we examined the suicidal impact of the 19 independent variables in concert within each racial group (see Tables 3 and 4). (Note that we entered CES-D and ASI scales in separate analyses.) Three variables were significant within both races: higher CES-D scores, copes by using medications and/or seeks professional help, and lower religiosity. Two variables were associated with suicidality only within the White group: higher use of spiritualists and copes by keeping calm. One variable, greater use of doctors for mental health problems, was significant only within the Black group. When we substituted ASI scores for CES-D scores, higher ASI scores were significantly associated with suicidality in both racial groups (Whites: OR = 1.07, 95% CI = 1.02–1.12, p =.008; Blacks: OR = 1.07, 95% CI = 1.03–1.12, p =.001).
|
|
2 = 1.88, p =.39; or lifetime suicidality (13% U.S. born, 10% English Caribbean, 8% French Caribbean), 2 = 1.12, p =.57. |
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Our findings also differ from those of several studies conducted in outpatient medicine settings in which African Americans had significantly lower rates of suicidality than Whites (Bartels et al., 2002; Lish et al., 1996). However, our results are similar to those of Callahan and colleagues (1996) for older primary care patients in Indianapolis, where no racial difference was noted. The differences between our study and the primary care studies may be due to variations in look-back criteria for suicidality (up to 12 months in some studies), geographic differences, and whether it was a clinical or community sample.
Another important finding is the lack of significant intragroup differences in suicidality among Blacks. The fact that individuals of French Caribbean background had the lowest suicidality levels, albeit nonsignificant, may reflect the greater number of Roman Catholics in this group versus the other Black ethnic groups (72% vs 17%). Catholicism has especially strong sanctions against suicide.
Unlike earlier studies, we were able to separately examine factors associated with suicidality in both races. Of the eight variables that were significantly associated with suicidality, four were significant for both Blacks and Whites. Three of these variables, lower levels of religiosity and greater depressive and anxiety symptoms, have been consistently identified as variables associated with suicidality in younger and older groups (Bartels et al., 2002; Callahan et al., 1996; Cook et al., 2002; Lish et al., 1996, Nisbet, Duberstein, Conwell, & Seidlitz, 2000), although Bartels and colleagues did not find religion to be significant. With respect to depression and anxiety symptoms, because they had a shared variance of 58%, we entered each one into separate logistic regression analyses along with the other predictor variables. For both races, the very slight differences in the ORs of these two variables suggest that they were relatively equivalent predictors of suicidality. The fourth variable found among both Blacks and Whites was the greater use of medications and/or seeking professional help when stressed or upset emotionally. Although researchers have not previously examined coping strategies with respect to suicidality, findings with depressed elders have pointed to the value of including coping strategies in analyses (Cohen, Magai, Yaffee, & Walcott-Brown, 2005; George, 2004). Although we cannot determine causal direction because the data were cross-sectional, it seems more reasonable to assume that this coping strategy represents a response to distress rather than a cause of increased suicidality. The impacts of these four variables were additive rather than multiplicative (i.e., none of these variables' interactions with race were significant). Thus, these variables affected suicidality in a linear fashion (i.e., greater levels were associated with more suicidality).
There were two variables associated with increased suicidality among Whites but not among Blacks: greater use of spiritualist or their products, and current or past alcohol abuse. Only one variable, greater use of physicians for mental health problems, was significant among Blacks but not among Whites. Of these three variables, greater use of physicians/psychiatrist and alcohol abuse have been noted previously in the literature (Cook et al., 2002; Lish et al., 1996), although it is not clear why the former did not obtain among Whites with suicidality and the latter did not obtain among Blacks. One possibility is that the low prevalence of these variables may have made the estimates unstable. Indeed, there was a trend (p =.08) among Whites for suicidality to associate with greater use of physicians for mental health problems. Although it is possible that seeing a physician/psychiatrist for mental health reasons could trigger more recall of suicidality, we believe that examining between-group racial differences can more plausibly explain the within-group racial differences in addressing mental health issues. For the sample as a whole, we found a nonsignificant trend for Whites to more frequently use physicians/psychiatrists for mental heath problems than Blacks (2.2% vs 5.6%), 2 = 3.69, p =.06. This finding is consistent with reports in the literature (Baker, 1996), in which Blacks have been found to be less likely to discuss mental health issues or use their primary care providers or psychiatrists for mental health concerns. However, in contrasting Whites and Blacks, we found that the latter had a higher ratio of having used a physician/psychiatrist for mental health problems among those with a history of suicidality versus those without histories. This is despite the fact that a smaller percentage of Blacks than Whites with suicidality had discussed mental health issues. However, of concern was that among these at-risk groups of elderly Blacks and Whites with histories of suicidality, only 14% and 21%, respectively, had discussed mental health issues with a physician/psychiatrist in the past year.
In contrast to religiosity, which was associated with lower suicidality, the use of spiritualists or their products was associated with greater suicidality only among Whites. Perhaps the different directional associations of these variables with suicidality reflect temporal differences. That is, religiosity may serve to diminish suicidality, whereas at the time spiritualists or their products are used, the person already feels suicidal. However, one must view these explanations as provisional given the cross-sectional nature of the data. The fact that this variable was significant in Whites and not Blacks may also reflect more common use of spiritualists or their products among Blacks in general (see Table 2 for prevalence data); the difference between Blacks and Whites was significant, t(172) = 2.39, p =.02. Finally, the finding of a second coping variable—coping with interpersonal conflicts by keeping calm—that in this case was associated with suicidality only among Whites reinforces the point made previously, that coping strategies may play a role in affecting suicidality. This also deserves further exploration.
Finally, several variables that we thought might be risk factors for suicidality (e.g., social supports, physical illness, acute stressors, lifetime trauma, and impaired functioning) were significant for one or both racial groups in bivariate analysis, but they no longer attained significance after we entered them concomitantly with other variables in the logistic regression analysis. This finding underscores the importance of using a multivariate model to assess the impact of various risk factors on suicidality.
This study has several limitations, so readers need to interpret findings cautiously. Data were cross-sectional, and the results derived from one geographic area. We did not have precise data on individuals excluded because of cognitive deficits, although we believe such cases were rare (well under 5%) given our cutoff criteria and recruitment methods. We also did not have data on whether there might have been any ethnic or gender biases in the recruitment methods within each randomly selected block group (i.e., knocking on doors vs using community sites). Also, the prevalence rates of some of the variables such as alcoholism and use of physicians for mental health reasons were low, and their ORs may have been unreliable. Despite the high prevalence of depression among individuals with suicidality, our use of lifetime suicidality as the dependent variable may have affected the association of some of the independent variables with suicidality, because the latter depends on recall of suicidal thoughts over a lifetime. Future studies must include longitudinal data, larger samples, and more diverse geographic settings. Of importance is the fact that the study's findings are strengthened by our use of random sampling, the use of a statistical program that controlled for design effects of a complex sample, the generation of a large multiracial sample, and the utilization of a theoretical model that enhances the generalizability of the results.
In summary, we found no interracial or intraracial (among Blacks) differences in the prevalence of active or passive suicidality. Rowe, Conwell, Schulberg, and Bruce (2006) pointed out that extrapolating from findings about suicidal ideation to suicide prevention strategies is complicated because the associations between the two are uncertain. Because suicidal ideation is a risk factor for suicide, studies of suicidal ideation can help inform prevention strategies. At the very least, reducing suicidal ideation can enhance quality of life. It is noteworthy that virtually all of the factors associated with suicidality in our study are potentially ameliorable. Among both racial groups, suicidality is likely to be impacted by addressing depressive and anxiety symptoms and, when appropriate, by encouraging various coping strategies, especially religiosity. About one third of older adults with a lifetime history of suicidality report current symptoms of suicidality; however, fewer than 1 in 5 individuals with a history of suicidality report having discussed their mental health with any type of physician in the past year. Targeting this potentially at-risk population should be an important public health goal.
|
Footnotes |
|---|
1 Division of Geriatric Psychiatry, State University of New York Downstate Medical Center, Brooklyn.
2 Department of Psychiatry, Montefiore Medical Center, Bronx, NY.
3 Shirley M. Ehrenkranz School of Social Work, New York University, New York.
Decision Editor: Linda S. Noelker, PhD
Received for publication October 13, 2006. Accepted for publication April 16, 2007.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
