This article was downloaded by: [University of Auckland Library] On: 01 December 2014, At: 15:18 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Aging & Mental Health Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/camh20
Risk factors for suicide among older adults with cancer ab
c
d
e
Thomas B. Cole , J. Michael Bowling , Michael J. Patetta & Dan G. Blazer a
Department of Social Medicine, University of North Carolina School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA b
Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA c
Department of Health Behavior and Health Education, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA d
Training and Education Department, SAS Institute, Cary, North Carolina, USA
e
Department of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Duke University, Durham, North Carolina, USA Published online: 07 Mar 2014.
To cite this article: Thomas B. Cole, J. Michael Bowling, Michael J. Patetta & Dan G. Blazer (2014) Risk factors for suicide among older adults with cancer, Aging & Mental Health, 18:7, 854-860, DOI: 10.1080/13607863.2014.892567 To link to this article: http://dx.doi.org/10.1080/13607863.2014.892567
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Aging & Mental Health, 2014 Vol. 18, No. 7, 854–860, http://dx.doi.org/10.1080/13607863.2014.892567
Risk factors for suicide among older adults with cancer Thomas B. Colea,b*, J. Michael Bowlingc, Michael J. Patettad and Dan G. Blazere a
Department of Social Medicine, University of North Carolina School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; bDepartment of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; cDepartment of Health Behavior and Health Education, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; dTraining and Education Department, SAS Institute, Cary, North Carolina, USA; eDepartment of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Duke University, Durham, North Carolina, USA
Downloaded by [University of Auckland Library] at 15:18 01 December 2014
(Received 6 September 2013; accepted 2 February 2014) Objective: To determine whether the increased risk of suicide for individuals with cancer may be explained by functional limitations, lack of social support, or other factors. Method: In this population-based case-control study, interviews of primary informants for suicides in the state of North Carolina were compared to interviews with participants in the Piedmont Health Study of the Elderly to estimate adjusted odds ratios for suicide and self-reported, physician diagnosed cancer, heart attack, stroke, and hip fracture. Results: Adjusting for all other factors, there was a statistically significant association of suicide and cancer (odds ratio [OR] 2.62, 95% confidence interval [CI] CI 1.84–3.73), but not heart attack, hip fracture, or stroke. The risk of suicide was also elevated for men vs. women (OR 17.15, CI 10.88–27.02), whites vs. blacks (OR 9.70, CI 6.07–15.50), and individuals with stressful life events (OR 2.75, CI 1.97–3.86) or limitations of instrumental (OR 2.93, CI 2.03–4.22) but not physical activities of daily living. Suicide cases were not more likely to be short of breath or poor sleep quality. Suicide was statistically significantly less likely for study participants who were married with spouse living vs. other (OR 0.61, CI 0.43–0.88) or who had one or more indicators of social support (OR 0.27, CI 0.19–0.39). Conclusion: After adjustment for other risk factors, suicide was strongly associated with cancer but not with other disabling, potentially fatal conditions. Keywords: suicide; social support; functional status; epidemiology (mental health); physical disorders
Introduction Cancer is an established risk factor for suicide (Fall et al., 2009; Llorente et al., 2005; Misono, Weiss, Fann, Redman, & Yueh, 2008), but it is not clear whether functional limitations (Akechi et al., 2002; Breitbart et al., 2000; Robson, Scrutton, Wilkinson, & MacLeod, 2010), lack of social support (Breitbart et al., 2000; Spoletini et al., 2011), or other factors account for this elevated risk. A study of terminally ill cancer patients found that 17% expressed a desire for hastened death (Breitbart et al., 2000), which was associated with measures of depression, hopelessness, limited physical functioning, and lack of social support. Factors associated with suicide, suicidal thoughts, or a desire for hastened death in cancer patients include functional impairment (Akechi et al., 2002; Breitbart et al., 2000; Robson et al., 2010), shortness of breath (Suarez-Almazor, Newman, Hanson, & Bruera, 2002), lack of social support (Breitbart et al., 2000; Spoletini et al., 2011), depression (Akechi et al., 2002; Spoletini et al., 2011; Suarez-Almazor et al., 2002), hopelessness (Breitbart et al., 2000; Robson et al., 2010), and shame or guilt (LoConte, Else-Quest, Eickhoff, Hyde, & Schiller, 2008). The relative risk of suicide has been reported to be greatest in the first week after receiving a diagnosis of cancer (Fang et al., 2012). Some risk factors for suicide can be addressed with social and behavioral interventions, and therefore it is of interest to learn whether cancer may be a *Corresponding author. Email: [email protected] Ó 2014 Taylor & Francis
proxy for other factors or whether a diagnosis of cancer is an independent risk factor for suicide (Druss & Pincus, 2000). We conducted a case-control study of suicide and cancer, adjusting for other risk factors, in a population of individuals aged 65 years and older, an age group at high risk for cancer (Siegel et al., 2012) and cancer-associated suicide (Fall et al., 2009; Llorente et al., 2005; Misono et al., 2008). To determine whether the risk of suicide is specific to a diagnosis of cancer or is also elevated for other disabling and potentially fatal conditions of old age (Juurlink, Herrmann, Szalai, Kopp, & Redelmeier, 2004; Spoletini et al., 2011), we estimated the risk of suicide associated with self-reported heart attack, stroke, and hip fracture. To determine whether cancer is an independent risk factor for suicide in old age, we controlled for other risk factors for suicide in cancer patients such as functional limitations (Akechi et al., 2002; Breitbart et al., 2000; Robson et al., 2010), living alone (Blazer, 2003), lack of social support (Breitbart et al., 2000; Spoletini et al., 2011), stressful life events (Blazer, 2003), poor sleep quality (Blazer, 2003), and serious illnesses (Juurlink et al., 2004; Spoletini et al., 2011).
Methods In this case-control study, a case was defined as an individual aged 65 years or older who died in the state of
Downloaded by [University of Auckland Library] at 15:18 01 December 2014
Aging & Mental Health North Carolina in 1990 or 1991 and whose death certificate indicated the cause of death was suicide. In North Carolina, all suicides are certified by the North Carolina Medical Examiner System, which is mandated by state statute to investigate all sudden, unexpected deaths, including fatal injuries, fatal overdoses, and suicides. For each death, a primary informant was identified from the death certificate and was contacted to participate in an in-person interview. Six to twelve months after the date of death, a letter explaining the study was mailed to a primary informant for each suicide case. The letter explained that the study was designed to collect information about the circumstances that lead to untimely deaths of older adults, asked for the informant’s help in providing information for the study, and said that a research assistant would call within two weeks to arrange an inhome interview. Approximately two weeks after the letter was sent, a telephone call was made to each primary informant to request participation in the study and request a personal interview. If the primary informant declined to participate, he or she was asked to give the name of another potential informant who personally knew the suicide case. If the primary informant agreed to participate but declined a personal interview, a telephone interview was offered instead. Interviewers were trained by Piedmont Health Study of the Elderly (PHSE) staff using the interview protocol of the PHSE. Nearly all interviews for this study were conducted by one interviewer. Institutional review board (IRB) approval was obtained from the Duke University School of Medicine. The controls were the complete cohort of older adults enrolled in the PHSE, one of the study sites for the Established Populations for Epidemiologic Studies of the Elderly project. The purpose of the PHSE was to obtain data on community-dwelling elders for longitudinal studies of risk factors for diseases, disabilities, and mortality (Cornoni-Huntley et al., 1993). The PHSE cohort was selected using a four-stage sampling design from a fivecounty region of North Carolina; one county was primarily urban and the other four were primarily rural. Housing units were stratified by race, with blacks over-sampled to yield approximately 55% of the sample. The response rate at baseline was 80%, producing a sample of 4162 men and women, of whom 2260 were (non-Hispanic) blacks and 1876 (non-Hispanic) whites. Baseline data were collected during in-home interviews in 1986. To obtain information about cases, each primary informant was asked a subset of questions from the baseline survey of the PHSE. Information about controls was obtained from the same subset of survey questions. To obtain information about clinical diagnoses, case informants and controls were asked if a doctor had diagnosed ‘cancer, a malignancy, or a malignant tumor of any type’; ‘a heart attack, coronary, myocardial infarction, coronary thrombosis, or coronary occlusion’; ‘a stroke or a brain hemorrhage’; or ‘a broken or fractured hip.’ Possible covariates of the association of serious illness and suicide included age group, race, sex, marital status, functional limitation, social support, stressful life events, sleep quality, and shortness of breath. Age group was
855
dichotomized as 71 years or older vs. 65–70 years of age, race as white or black, sex as man or woman, and marital status as married, spouse living vs. other. We developed indices of functional limitation using seven items from the Activities of Daily Living Scale and seven items from the Instrumental Activities of Daily Living Scale (Branch & Meyers, 1987). (For these and all other survey items in this study, the same questions were asked of the informants about suicide cases as were asked of the older adults in the PHSE about themselves, although the wording of the questions was modified slightly to make it clear that the questions referred to a deceased case or a living control.) The seven items from the Activities of Daily Living Scale asked if the case or control needed assistance with walking across a small room, bathing, personal grooming, getting dressed, eating, getting from a bed to a chair, or using the toilet. The seven items from the Instrumental Activities of Daily Living Scale asked if the case or control needed assistance to use the telephone, drive a car or travel alone on buses or taxis, shop for groceries or clothes, prepare meals, do housework, take medicine, or handle money. For each functional index, cases and controls were dichotomized as needing assistance with one or more items in the index vs. none of the items. We also developed an index of social support. Cases and controls were classified as having some social support if they had a child, other relative, or close friend that they saw at least once a month or could talk to about private matters or call on for help, or if they belonged to any clubs or organizations, such as church-related or other groups. Cases and controls who did not have a child, other relative, or close friend that they saw at least once a month or could talk to about private matters or call on for help, and did not belong to any clubs or organizations, were classified as having no social support. Cases and controls were classified as having a stressful life event if during the past year they had been hospitalized; missed a week or more of work or usual activities because of an illness or injury; experienced the death of a spouse, close family member, or friend; or had trouble with the law; or if a close family member or friend experienced a serious illness or injury or had trouble with the law. Cases and controls were classified as having poor sleep quality if they reported sometimes or most of the time having trouble falling asleep, waking up during the night, waking up too early and not being able to go back to sleep again, or feeling the need to sleep during the day. Cases and controls were classified as having shortness of breath if it was severe enough to require them to stop and rest. Data analysis Age group, race, and sex were compared for suicide cases with and without a primary informant who was willing to be interviewed for this study. Then cases were compared to controls for each of the study variables. All study variables were included in a logistic regression model using SAS version 9.2, and odds ratios with 95% Wald
856
T.B. Cole et al.
Table 1. Comparison of eligible cases with and without interviews. Eligible cases
Downloaded by [University of Auckland Library] at 15:18 01 December 2014
Age group 65–70 years 71þ years Race White Black Other Sex Male Female
An informant was interviewed (%) (N ¼ 217)
An informant was not interviewed (%) (N ¼ 117)
68 (31.5) 148 (68.5)
47 (40.9) 68 (59.1)
191 (88.0) 24 (11.1) 2 (0.9)
107 (91.5) 10 (8.5) 0 (0.0)
184 (84.8) 33 (15.2)
96 (82.1) 21 (17.9)
p-value 0.09 0.44
0.52
Some eligible cases were missing information on age group.
confidence intervals were calculated for the association of suicide and each of the study variables. Finally, a sensitivity analysis was performed to compare the controls, which were selected from a five-county region of the state, to the subgroup of cases from the same five-county region, using logistic regression. Results During the two-year period 1990–1991, 334 deaths of individuals aged 65 years or older were certified as suicides in the state of North Carolina. A primary informant agreed to provide information in an interview for 217 of these 334 (65.0%) eligible suicides. Of the primary informants who provided interviews, 79 were sons or daughters, 70 were spouses, 19 were brothers or sisters, 12 were nieces or nephews, 27 were other family members, and 10 were non-family members who knew the suicide cases well. Of the 217 proxy interviews, 150 were obtained in person and 67 by telephone. Eligible suicides with primary informant interviews were more likely than eligible suicides without interviews to be older, of black race, and of male sex (Table 1). Compared with controls, suicide cases were more likely to be older, white, male, married with spouse living; to have a diagnosis of cancer, heart attack, broken hip, or stroke; to have one or more limitations in physical and instrumental activities of daily living; to be short of breath; to have one or more indicators of social support; to have had one or more stressful events in the past year; and to have poor sleep quality (Table 2). When adjusted for all other factors, there was a statistically significant association of suicide and cancer (odds ratio [OR] 2.62, confidence interval [CI] 1.84–3.73), but not of suicide and heart attack, hip fracture, or stroke (Table 3). The risk of suicide was also estimated to be greater for men vs. women (OR 17.15, 95% CI 10.88– 27.02) and for whites vs. blacks (OR 9.70, CI 6.07– 15.50). Suicide cases were statistically more likely than controls to have had a stressful life event in the past year (OR 2.75, CI 1.97–3.86) or to have one or more
limitations of instrumental (OR 2.93, CI 2.03–4.22) but not physical activities of daily living. Suicide cases were not statistically more likely to be short of breath or have poor sleep quality. Suicide was statistically significantly less likely for study participants who were married with spouse living vs. other (OR 0.61, CI 0.43–0.88) or who had one or more indicators of social support (OR 0.27, CI 0.19–0.39). In this study, all 100 counties of the state of North Carolina were eligible to contribute cases in 1990 and 1991, but controls were selected from a five-county region of the state, under the assumption that the study findings would not be influenced by unknown or unmeasured risk factors associated with residence in a particular county of the state. To test this assumption, the subset of suicide cases from the five-county region was compared to controls. Although CIs for the odds ratios were wider, due to increases in standard errors resulting from smaller sample sizes, associations of suicide with cancer, male sex, a stressful life event in the past year and limitations of instrumental activities of daily living that were observed in the full-state analysis remained elevated and statistically significant (Table 3), and the lower risk of suicide and social support also remained statistically significant. However, associations of suicide with race and marital status that were observed in the full-state analysis were not statistically significant in the analysis restricted to cases from the five-county region. Discussion The results of this population-based case-control study confirm the previously reported association of suicide with a diagnosis of cancer (Bj€orkenstam, Edberg, Ayoubi, & Rosen, 2005; Breitbart, 1987; Hem, Loge, Haldorsen, & Ekeberg, 2004; Llorente et al., 2005; Spoletini et al., 2011) but extend prior research by demonstrating that the increased risk is not accounted for by other known risk factors for suicide, including age group, sex, race, marital status, functional limitations, social support, stressful life events, shortness of breath, and poor sleep quality. After adjustment for these factors, suicide cases were still twice as likely to have had cancer as controls. We did not find an increased risk of suicide for study participants reporting a diagnosis of heart attack, hip fracture, or stroke, which supports our finding that a diagnosis of cancer is an independent risk factor for suicide that is not explained by factors associated with potentially disabling or fatal illness in general. Previous studies have reported that patients with cancer were more likely to consider, attempt, or commit suicide if they had limitations of activities of daily living (Spoletini et al., 2011); were separated, divorced, or widowed or lacked other sources of social support (Breitbart et al., 2000; Spoletini et al., 2011); or experienced certain symptoms associated with cancer, including depression (Breitbart et al., 2000; Derogatis et al., 1983; Spoletini et al., 2011). Depression is a strong risk factor for suicide in old age (Harwood, Hawton, Hope, & Jacoby, 2001; Sun, Xu, Chan, Lam, & Schooling, 2012); almost two-
Aging & Mental Health
857
Downloaded by [University of Auckland Library] at 15:18 01 December 2014
Table 2. Characteristics of cases and controls (listing each variable in the study). Suicide cases (%) (N ¼ 217)
Controls (%) (N ¼ 4162)
Age group 65–70 years 71þ years
68 (31.5) 148 (68.5)
1659 (39.9) 2500 (60.1)
0.0137
Race White Black
191 (88.0) 24 (11.1)
1876 (45.4) 2260 (54.6)
Aging & Mental Health Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/camh20
Risk factors for suicide among older adults with cancer ab
c
d
e
Thomas B. Cole , J. Michael Bowling , Michael J. Patetta & Dan G. Blazer a
Department of Social Medicine, University of North Carolina School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA b
Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA c
Department of Health Behavior and Health Education, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA d
Training and Education Department, SAS Institute, Cary, North Carolina, USA
e
Department of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Duke University, Durham, North Carolina, USA Published online: 07 Mar 2014.
To cite this article: Thomas B. Cole, J. Michael Bowling, Michael J. Patetta & Dan G. Blazer (2014) Risk factors for suicide among older adults with cancer, Aging & Mental Health, 18:7, 854-860, DOI: 10.1080/13607863.2014.892567 To link to this article: http://dx.doi.org/10.1080/13607863.2014.892567
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Aging & Mental Health, 2014 Vol. 18, No. 7, 854–860, http://dx.doi.org/10.1080/13607863.2014.892567
Risk factors for suicide among older adults with cancer Thomas B. Colea,b*, J. Michael Bowlingc, Michael J. Patettad and Dan G. Blazere a
Department of Social Medicine, University of North Carolina School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; bDepartment of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; cDepartment of Health Behavior and Health Education, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; dTraining and Education Department, SAS Institute, Cary, North Carolina, USA; eDepartment of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Duke University, Durham, North Carolina, USA
Downloaded by [University of Auckland Library] at 15:18 01 December 2014
(Received 6 September 2013; accepted 2 February 2014) Objective: To determine whether the increased risk of suicide for individuals with cancer may be explained by functional limitations, lack of social support, or other factors. Method: In this population-based case-control study, interviews of primary informants for suicides in the state of North Carolina were compared to interviews with participants in the Piedmont Health Study of the Elderly to estimate adjusted odds ratios for suicide and self-reported, physician diagnosed cancer, heart attack, stroke, and hip fracture. Results: Adjusting for all other factors, there was a statistically significant association of suicide and cancer (odds ratio [OR] 2.62, 95% confidence interval [CI] CI 1.84–3.73), but not heart attack, hip fracture, or stroke. The risk of suicide was also elevated for men vs. women (OR 17.15, CI 10.88–27.02), whites vs. blacks (OR 9.70, CI 6.07–15.50), and individuals with stressful life events (OR 2.75, CI 1.97–3.86) or limitations of instrumental (OR 2.93, CI 2.03–4.22) but not physical activities of daily living. Suicide cases were not more likely to be short of breath or poor sleep quality. Suicide was statistically significantly less likely for study participants who were married with spouse living vs. other (OR 0.61, CI 0.43–0.88) or who had one or more indicators of social support (OR 0.27, CI 0.19–0.39). Conclusion: After adjustment for other risk factors, suicide was strongly associated with cancer but not with other disabling, potentially fatal conditions. Keywords: suicide; social support; functional status; epidemiology (mental health); physical disorders
Introduction Cancer is an established risk factor for suicide (Fall et al., 2009; Llorente et al., 2005; Misono, Weiss, Fann, Redman, & Yueh, 2008), but it is not clear whether functional limitations (Akechi et al., 2002; Breitbart et al., 2000; Robson, Scrutton, Wilkinson, & MacLeod, 2010), lack of social support (Breitbart et al., 2000; Spoletini et al., 2011), or other factors account for this elevated risk. A study of terminally ill cancer patients found that 17% expressed a desire for hastened death (Breitbart et al., 2000), which was associated with measures of depression, hopelessness, limited physical functioning, and lack of social support. Factors associated with suicide, suicidal thoughts, or a desire for hastened death in cancer patients include functional impairment (Akechi et al., 2002; Breitbart et al., 2000; Robson et al., 2010), shortness of breath (Suarez-Almazor, Newman, Hanson, & Bruera, 2002), lack of social support (Breitbart et al., 2000; Spoletini et al., 2011), depression (Akechi et al., 2002; Spoletini et al., 2011; Suarez-Almazor et al., 2002), hopelessness (Breitbart et al., 2000; Robson et al., 2010), and shame or guilt (LoConte, Else-Quest, Eickhoff, Hyde, & Schiller, 2008). The relative risk of suicide has been reported to be greatest in the first week after receiving a diagnosis of cancer (Fang et al., 2012). Some risk factors for suicide can be addressed with social and behavioral interventions, and therefore it is of interest to learn whether cancer may be a *Corresponding author. Email: [email protected] Ó 2014 Taylor & Francis
proxy for other factors or whether a diagnosis of cancer is an independent risk factor for suicide (Druss & Pincus, 2000). We conducted a case-control study of suicide and cancer, adjusting for other risk factors, in a population of individuals aged 65 years and older, an age group at high risk for cancer (Siegel et al., 2012) and cancer-associated suicide (Fall et al., 2009; Llorente et al., 2005; Misono et al., 2008). To determine whether the risk of suicide is specific to a diagnosis of cancer or is also elevated for other disabling and potentially fatal conditions of old age (Juurlink, Herrmann, Szalai, Kopp, & Redelmeier, 2004; Spoletini et al., 2011), we estimated the risk of suicide associated with self-reported heart attack, stroke, and hip fracture. To determine whether cancer is an independent risk factor for suicide in old age, we controlled for other risk factors for suicide in cancer patients such as functional limitations (Akechi et al., 2002; Breitbart et al., 2000; Robson et al., 2010), living alone (Blazer, 2003), lack of social support (Breitbart et al., 2000; Spoletini et al., 2011), stressful life events (Blazer, 2003), poor sleep quality (Blazer, 2003), and serious illnesses (Juurlink et al., 2004; Spoletini et al., 2011).
Methods In this case-control study, a case was defined as an individual aged 65 years or older who died in the state of
Downloaded by [University of Auckland Library] at 15:18 01 December 2014
Aging & Mental Health North Carolina in 1990 or 1991 and whose death certificate indicated the cause of death was suicide. In North Carolina, all suicides are certified by the North Carolina Medical Examiner System, which is mandated by state statute to investigate all sudden, unexpected deaths, including fatal injuries, fatal overdoses, and suicides. For each death, a primary informant was identified from the death certificate and was contacted to participate in an in-person interview. Six to twelve months after the date of death, a letter explaining the study was mailed to a primary informant for each suicide case. The letter explained that the study was designed to collect information about the circumstances that lead to untimely deaths of older adults, asked for the informant’s help in providing information for the study, and said that a research assistant would call within two weeks to arrange an inhome interview. Approximately two weeks after the letter was sent, a telephone call was made to each primary informant to request participation in the study and request a personal interview. If the primary informant declined to participate, he or she was asked to give the name of another potential informant who personally knew the suicide case. If the primary informant agreed to participate but declined a personal interview, a telephone interview was offered instead. Interviewers were trained by Piedmont Health Study of the Elderly (PHSE) staff using the interview protocol of the PHSE. Nearly all interviews for this study were conducted by one interviewer. Institutional review board (IRB) approval was obtained from the Duke University School of Medicine. The controls were the complete cohort of older adults enrolled in the PHSE, one of the study sites for the Established Populations for Epidemiologic Studies of the Elderly project. The purpose of the PHSE was to obtain data on community-dwelling elders for longitudinal studies of risk factors for diseases, disabilities, and mortality (Cornoni-Huntley et al., 1993). The PHSE cohort was selected using a four-stage sampling design from a fivecounty region of North Carolina; one county was primarily urban and the other four were primarily rural. Housing units were stratified by race, with blacks over-sampled to yield approximately 55% of the sample. The response rate at baseline was 80%, producing a sample of 4162 men and women, of whom 2260 were (non-Hispanic) blacks and 1876 (non-Hispanic) whites. Baseline data were collected during in-home interviews in 1986. To obtain information about cases, each primary informant was asked a subset of questions from the baseline survey of the PHSE. Information about controls was obtained from the same subset of survey questions. To obtain information about clinical diagnoses, case informants and controls were asked if a doctor had diagnosed ‘cancer, a malignancy, or a malignant tumor of any type’; ‘a heart attack, coronary, myocardial infarction, coronary thrombosis, or coronary occlusion’; ‘a stroke or a brain hemorrhage’; or ‘a broken or fractured hip.’ Possible covariates of the association of serious illness and suicide included age group, race, sex, marital status, functional limitation, social support, stressful life events, sleep quality, and shortness of breath. Age group was
855
dichotomized as 71 years or older vs. 65–70 years of age, race as white or black, sex as man or woman, and marital status as married, spouse living vs. other. We developed indices of functional limitation using seven items from the Activities of Daily Living Scale and seven items from the Instrumental Activities of Daily Living Scale (Branch & Meyers, 1987). (For these and all other survey items in this study, the same questions were asked of the informants about suicide cases as were asked of the older adults in the PHSE about themselves, although the wording of the questions was modified slightly to make it clear that the questions referred to a deceased case or a living control.) The seven items from the Activities of Daily Living Scale asked if the case or control needed assistance with walking across a small room, bathing, personal grooming, getting dressed, eating, getting from a bed to a chair, or using the toilet. The seven items from the Instrumental Activities of Daily Living Scale asked if the case or control needed assistance to use the telephone, drive a car or travel alone on buses or taxis, shop for groceries or clothes, prepare meals, do housework, take medicine, or handle money. For each functional index, cases and controls were dichotomized as needing assistance with one or more items in the index vs. none of the items. We also developed an index of social support. Cases and controls were classified as having some social support if they had a child, other relative, or close friend that they saw at least once a month or could talk to about private matters or call on for help, or if they belonged to any clubs or organizations, such as church-related or other groups. Cases and controls who did not have a child, other relative, or close friend that they saw at least once a month or could talk to about private matters or call on for help, and did not belong to any clubs or organizations, were classified as having no social support. Cases and controls were classified as having a stressful life event if during the past year they had been hospitalized; missed a week or more of work or usual activities because of an illness or injury; experienced the death of a spouse, close family member, or friend; or had trouble with the law; or if a close family member or friend experienced a serious illness or injury or had trouble with the law. Cases and controls were classified as having poor sleep quality if they reported sometimes or most of the time having trouble falling asleep, waking up during the night, waking up too early and not being able to go back to sleep again, or feeling the need to sleep during the day. Cases and controls were classified as having shortness of breath if it was severe enough to require them to stop and rest. Data analysis Age group, race, and sex were compared for suicide cases with and without a primary informant who was willing to be interviewed for this study. Then cases were compared to controls for each of the study variables. All study variables were included in a logistic regression model using SAS version 9.2, and odds ratios with 95% Wald
856
T.B. Cole et al.
Table 1. Comparison of eligible cases with and without interviews. Eligible cases
Downloaded by [University of Auckland Library] at 15:18 01 December 2014
Age group 65–70 years 71þ years Race White Black Other Sex Male Female
An informant was interviewed (%) (N ¼ 217)
An informant was not interviewed (%) (N ¼ 117)
68 (31.5) 148 (68.5)
47 (40.9) 68 (59.1)
191 (88.0) 24 (11.1) 2 (0.9)
107 (91.5) 10 (8.5) 0 (0.0)
184 (84.8) 33 (15.2)
96 (82.1) 21 (17.9)
p-value 0.09 0.44
0.52
Some eligible cases were missing information on age group.
confidence intervals were calculated for the association of suicide and each of the study variables. Finally, a sensitivity analysis was performed to compare the controls, which were selected from a five-county region of the state, to the subgroup of cases from the same five-county region, using logistic regression. Results During the two-year period 1990–1991, 334 deaths of individuals aged 65 years or older were certified as suicides in the state of North Carolina. A primary informant agreed to provide information in an interview for 217 of these 334 (65.0%) eligible suicides. Of the primary informants who provided interviews, 79 were sons or daughters, 70 were spouses, 19 were brothers or sisters, 12 were nieces or nephews, 27 were other family members, and 10 were non-family members who knew the suicide cases well. Of the 217 proxy interviews, 150 were obtained in person and 67 by telephone. Eligible suicides with primary informant interviews were more likely than eligible suicides without interviews to be older, of black race, and of male sex (Table 1). Compared with controls, suicide cases were more likely to be older, white, male, married with spouse living; to have a diagnosis of cancer, heart attack, broken hip, or stroke; to have one or more limitations in physical and instrumental activities of daily living; to be short of breath; to have one or more indicators of social support; to have had one or more stressful events in the past year; and to have poor sleep quality (Table 2). When adjusted for all other factors, there was a statistically significant association of suicide and cancer (odds ratio [OR] 2.62, confidence interval [CI] 1.84–3.73), but not of suicide and heart attack, hip fracture, or stroke (Table 3). The risk of suicide was also estimated to be greater for men vs. women (OR 17.15, 95% CI 10.88– 27.02) and for whites vs. blacks (OR 9.70, CI 6.07– 15.50). Suicide cases were statistically more likely than controls to have had a stressful life event in the past year (OR 2.75, CI 1.97–3.86) or to have one or more
limitations of instrumental (OR 2.93, CI 2.03–4.22) but not physical activities of daily living. Suicide cases were not statistically more likely to be short of breath or have poor sleep quality. Suicide was statistically significantly less likely for study participants who were married with spouse living vs. other (OR 0.61, CI 0.43–0.88) or who had one or more indicators of social support (OR 0.27, CI 0.19–0.39). In this study, all 100 counties of the state of North Carolina were eligible to contribute cases in 1990 and 1991, but controls were selected from a five-county region of the state, under the assumption that the study findings would not be influenced by unknown or unmeasured risk factors associated with residence in a particular county of the state. To test this assumption, the subset of suicide cases from the five-county region was compared to controls. Although CIs for the odds ratios were wider, due to increases in standard errors resulting from smaller sample sizes, associations of suicide with cancer, male sex, a stressful life event in the past year and limitations of instrumental activities of daily living that were observed in the full-state analysis remained elevated and statistically significant (Table 3), and the lower risk of suicide and social support also remained statistically significant. However, associations of suicide with race and marital status that were observed in the full-state analysis were not statistically significant in the analysis restricted to cases from the five-county region. Discussion The results of this population-based case-control study confirm the previously reported association of suicide with a diagnosis of cancer (Bj€orkenstam, Edberg, Ayoubi, & Rosen, 2005; Breitbart, 1987; Hem, Loge, Haldorsen, & Ekeberg, 2004; Llorente et al., 2005; Spoletini et al., 2011) but extend prior research by demonstrating that the increased risk is not accounted for by other known risk factors for suicide, including age group, sex, race, marital status, functional limitations, social support, stressful life events, shortness of breath, and poor sleep quality. After adjustment for these factors, suicide cases were still twice as likely to have had cancer as controls. We did not find an increased risk of suicide for study participants reporting a diagnosis of heart attack, hip fracture, or stroke, which supports our finding that a diagnosis of cancer is an independent risk factor for suicide that is not explained by factors associated with potentially disabling or fatal illness in general. Previous studies have reported that patients with cancer were more likely to consider, attempt, or commit suicide if they had limitations of activities of daily living (Spoletini et al., 2011); were separated, divorced, or widowed or lacked other sources of social support (Breitbart et al., 2000; Spoletini et al., 2011); or experienced certain symptoms associated with cancer, including depression (Breitbart et al., 2000; Derogatis et al., 1983; Spoletini et al., 2011). Depression is a strong risk factor for suicide in old age (Harwood, Hawton, Hope, & Jacoby, 2001; Sun, Xu, Chan, Lam, & Schooling, 2012); almost two-
Aging & Mental Health
857
Downloaded by [University of Auckland Library] at 15:18 01 December 2014
Table 2. Characteristics of cases and controls (listing each variable in the study). Suicide cases (%) (N ¼ 217)
Controls (%) (N ¼ 4162)
Age group 65–70 years 71þ years
68 (31.5) 148 (68.5)
1659 (39.9) 2500 (60.1)
0.0137
Race White Black
191 (88.0) 24 (11.1)
1876 (45.4) 2260 (54.6)
