Suicides in older adults: A case–control psychological autopsy study in Australia

Journal of Psychiatric Research 47 (2013) 980e988

Contents lists available at SciVerse ScienceDirect

Journal of Psychiatric Research journal homepage: www.elsevier.com/locate/psychires

Suicides in older adults: A caseecontrol psychological autopsy study in Australia Diego De Leo a, *, Brian M. Draper b, c, John Snowdon d, Kairi Kõlves a a

Australian Institute for Suicide Research and Prevention, National Centre of Excellence in Suicide Prevention World Health Organization Collaborating Centre for Research and Training in Suicide Prevention, Griffith University, Australia b School of Psychiatry, University of New South Wales, Sydney, Australia c Academic Department for Old Age Psychiatry, Prince of Wales Hospital, Randwick, Australia d Discipline of Psychiatry, Sydney Medical School, University of Sydney, Australia

a r t i c l e i n f o

a b s t r a c t

Article history: Received 9 July 2012 Received in revised form 17 February 2013 Accepted 18 February 2013

Aim: The present study aims to analyse predicting factors of suicide among older adults compared to sudden death controls and middle-aged suicides. Methods: During the period 2006e2008, at two Australian sites, the psychological autopsy method was utilised to investigate suicides of individuals over the age of 35 by interviewing next-of-kin and healthcare professionals. A caseecontrol study design was applied using sudden death cases as controls. Initial information was gathered from coroner’s offices. Potential informants were approached and interviews were conducted using a semi-structured format. Results: In total, 261 suicides (73 aged 60þ) and 182 sudden deaths (79 aged 60þ) were involved. Older adult suicides showed a significantly lower prevalence of psychiatric diagnoses (62%) when compared to middle-aged suicide cases (80%). In both age groups, subjects who died by suicide were significantly more likely to present a psychiatric diagnosis, compared to controls; however, diagnosis did not remain in the final prediction model for older adults. Hopelessness and past suicide attempts remained in the final model for both age groups. In addition, living alone was an important predictor of suicide in older adults. Conclusion: Although mood disorders represent an important target for suicide prevention in old age, there should be increased attention for other risk factors including psychosocial, environmental, and general health aspects of late life. Ó 2013 Elsevier Ltd. All rights reserved.

Keywords: Suicide Older adults Predictive model Psychiatric disorders

1. Introduction Internationally, suicide rates increase with age. Consequently, global suicide rates are highest among older people, especially in males older than 75 years of age (Bertolote and Fleischmann, 2009). However, there are remarkable differences between countries (De Leo et al., 2011). Despite recent rate declines in Anglo-Saxon countries, including Australia, suicide in older people remains an important public health issue (De Leo et al., 2011). However, when compared to younger age groups, older suicides receive relatively less attention from both the research and prevention perspectives. Within suicide research, psychiatric disorders are widely studied, with clear evidence indicating that psychopathological problems increase the risk of suicide across all age groups. According to * Corresponding author. Australian Institute for Suicide Research and Prevention, Griffith University, Mt Gravatt Campus, QLD 4122, Australia. Tel.: þ61 7 3735 3377; fax: þ61 7 3735 3450. E-mail address: D.DeLeo@griffith.edu.au (D. De Leo). 0022-3956/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jpsychires.2013.02.009

previous psychological autopsy (PA) studies, psychiatric disorders are found in up to 90% of suicide cases (Arsenault-Lapierre et al., 2004; Cavanagh et al., 2003). However, the number of studies that specifically analyse the prevalence of psychiatric disorders among older people is rather limited and discordant in terms of findings. For example, psychiatric diagnoses were reported in 42.1% of elderly suicide cases in Canada (Preville et al., 2005), but in 96.5% of cases in Sweden (Waern et al., 2002). Considering that interventions focussing on mental health problems seem to have had a modest effect in reducing suicidal behaviours in older adults, attention has shifted to other factors (Lapierre et al., 2011). A recent systematic literature review of the social factors in suicidal behaviour in older adulthood showed limited evidence for the contribution of the lack of social connectedness to suicidal behaviours (Fässberg et al., 2012). Furthermore, a recent large-scale Australian study indicated suicidal ideation in older people was predicted by psychiatric factors, such as depression, anxiety, current use of antidepressants, and past suicide attempt, and somatic factors, like pain and poor

D. De Leo et al. / Journal of Psychiatric Research 47 (2013) 980e988

perceived health. Nevertheless, social factors such as living alone, poor social support, and financial stress contributed independently to the increased risk of suicidal ideation (Almeida et al., 2012). Consequently, there is a need to study a variety of different factors predicting suicide, in order to develop adequate suicide prevention programs for older people. The present study aims to estimate the contribution of different psychosocial and psychiatric factors to suicide among older suicides in Australia when compared to: (i) sudden death controls; and, (ii) middle-aged suicides. 2. Methods 2.1. Design The present study is part of a larger project, which aimed to examine the last clinical contact subjects who died by suicide had with health professionals, in order to determine whether this contact offers the opportunity for suicide prevention. The present study utilised the PA method to investigate older (60þ years of age) and middle-aged (aged 35e59 years) suicide cases from two Australian states: Queensland (Qld) and New South Wales (NSW). Interviews were conducted with next-of-kin (NOK) and healthcare professionals between 2006 and 2008. The methodology was previously tested in a pilot study at the NSW site (April 2003 to April 2004) (Draper et al., 2008). A caseecontrol study design was applied using sudden death victims as controls. The sudden death group included heart attacks, road traffic accidents (RTA), and other accidents; it excluded accidental overdoses, homicides, and single vehicle RTAs (Fig. 1). 2.2. Data collection process Initial information about the deceased was gathered from the Queensland Office of the State Coroner, the Queensland Police Service (Queensland), and the Glebe Coroners Court (Sydney). Subsequently, a letter introducing the study to potential informants, along with the study information sheet and consent form, were sent to participants. On receipt of the consent form, or an NOK’s signed authorisation form (in Queensland given by the Police Service), clinical interviewers followed up with a phone call to invite the NOK to participate and arrange a time and place for the interview. Clinical interviewers then conducted semi-structured interviews with the NOK of the deceased, lasting between one

981

and 3 h. These interviews aimed to establish the presence, or absence, of recognised predictive factors of suicide by following a semi-structured format and using validated scales/questionnaires. The following information was obtained from the informant about the deceased: demographics, psychopathology, personality, physical health status, healthcare contacts within the previous three months, life circumstances, prior suicidal behaviours, social support, aggression, and memory function. The response rate of NOKs was 46.6% for suicide cases and 36.5% for controls. In most cases, people did not provide reasons for their refusal to participate. Difficulty in understanding the benefits of the research (especially in the control group) and time constraints were recorded as main reasons. In total, 277 NOKs were interviewed in relation to 261 suicide cases, and 183 NOKs in relation to 182 sudden death cases. Between suicide cases and controls, there were no significant differences in the relationship to the deceased (e.g., spouse, parent, children, etc.). Most of the NOKs had regular contact (at least once a week) with the deceased, without significant differences between cases and controls (68.9% in suicides and 76.8% in the control group). Sudden death cases were most frequently attributed to diseases of the circulatory system (68.1%), followed by other external causes of death (15.4%). Healthcare professionals (HCPs) were identified through interviews with the NOK and other HCPs, as well as from coroners’ files and other medical files. HCPs (e.g., GPs, counsellors, social workers) were included only if the deceased had contact with them during the six months before death. Information was gathered from 211 healthcare professionals (in relation to 152 suicide cases) and 92 health care professionals (for 81 sudden death cases). However, the present study will analyse information only from the NOKs, as not all cases and controls visited an HCP or the HCP refused to participate in the study, so interviewing all HCPs was not possible. 2.3. Scales/questionnaires The NEO Five-Factor Inventory (NEO-FFI) was used to evaluate personality (Costa and McCrae, 1991). The NEO-FFI is a shortened version of NEO PI-R, a standardised tool used for informant report, and assesses 5 major domains of personality (neuroticism, extraversion, openness, agreeableness, and conscientiousness). Physical health status was assessed using the Cumulative Illness Rating Scale (Linn et al., 1968). A total score, excluding psychiatric items, was calculated for the purposes of the current analysis. Physical selfmaintenance and functioning were measured using the Instrumental Activities of Daily Living Scale and the Physical Self Maintenance Scale by Lawton and Brody (1969). Considering that social support and interaction between people is a complex phenomenon, the present study used the Bille-Brahe Social Support Scale, which brings out different dimensions of interaction (Bille-Brahe and Jensen, 2004). The scale distinguishes between receiving and giving practical and emotional social support from family and friends. 2.4. Psychiatric diagnoses

Fig. 1. Design of the study and numbers per study group.

Psychiatric diagnoses were determined using the Structured Clinical Interview for DSM-IV, Research Version (SCID I) (Spitzer et al., 1994) conducted by qualified clinical interviewers who had a common training day prior to commencement of the study. Using data provided to them by the trained clinical interviewers, two psychiatrists from each site, Qld and NSW, made psychiatric assessments by DSM-IV Axis I criteria; final consensus diagnoses are presented. Psychiatrists judged cases blindly (they did not know if the case was a suicide or a sudden death).

982

D. De Leo et al. / Journal of Psychiatric Research 47 (2013) 980e988

2.5. Ethics

2010). A probability level of 0.05 was employed for all statistical tests. IBM SPSS version 20.0 was used for data analyses.

The study was approved by the Griffith University Human Research Ethics Committee and South Eastern Sydney Human Research Ethics Committee. 2.6. Statistical analysis To compare background information between cases and controls, Chi-square tests (c2) were applied. To analyse the predicting variables for suicide compared to sudden death controls, separately for the elderly and middle-aged groups, odds ratios (OR) with a 95% confidence interval (CI) were calculated. Comparisons between older and middle-aged suicides are presented using a similar approach. Considering differences in some important background variables, adjusted ORs were calculated. ORs were adjusted for country of birth, living arrangements, and marital status for older and middle-aged groups; ORs were adjusted for employment status of the middle-aged but not the older group, as most were no longer in the workforce. When calculating ORs or c2 in sub-groups with small numbers (fewer than 3), Fisher’s exact test was used. Multinomial logistic regression was applied to estimate the independent contribution of significant predictors from the analyses. Variables with a significant association with serious suicidal ideation were entered into a backward stepwise logistic regression analysis. Final models were assessed using the concordance (c) statistic for discriminative ability (or area under the receiver operating characteristic [ROC] curve), the HosmereLemeshow goodness-of-fit test to evaluate the calibration (a small c2 value and a p greater than 0.05 would show an acceptable adaptation), and the Nagelkerke R2 to present the proportion of explained variance (Steyerberg et al.,

3. Results Table 1 presents the background information about suicide cases and sudden death controls in the two age groups: older (60 years and over, range 60e95 years) and middle-aged (aged 35e59 years). No significant differences between suicides and sudden death controls were found, in either the older or middle-aged group, by gender, living region, place of birth, non-English speaking background (NESB), or ethnicity. In both groups, there were more males than females, and there were more cases from Qld in the analysis than from NSW. Most of the study subjects were Caucasian and born in Australia. There were significant differences between suicides and controls by living arrangement in the older age group and by marital and employment status in the middleaged group. There were significant differences between older suicides and middle-aged suicides by place of birth: 34.2% of older suicides and 20.7% of middle-aged suicides were born overseas. Older people who died by suicide were more frequently than the middle-aged from a non-English speaking background (20.5% versus 11.2%). Significant differences were also shown in living arrangements; the older age group more frequently lived alone. In older suicide cases, 61.6% had at least one psychiatric disorder at the time of death; this was significantly lower than the 80.1% of middle-aged suicides (Table 2). Mood disorders were the most frequently diagnosed psychiatric disorders in both suicide subgroups (46.6% of older adults and 58.1% of middle-aged cases). Major depression had a similar prevalence in the two age groups,

Table 1 Sociodemographic background of the deceased in suicide and sudden death by age group. Age group 60þ

Age group 35e59

Suicide (a)

Sudden death

N

N

%

Gender Male 54 74 53 Female 19 26 26 Region QLD 58 79.5 53 NSW 15 20.5 26 Place of birth Australia 48 65.8 50 Other 25 34.2 29 From a non-English speaking background Yes 15 20.5 14 No 58 79.5 65 Ethnicity Caucasian 71 97.3 76 Other 2 2.8 3 Marital status Married/de facto Not married/de factor Living arrangements Alone With a family member (partner, child, parent) With friends/relatives or other Employment status Employed (FT, PT or Casual) Unemployed Out of workforce

Chi2

p-Value

% 0.86

%

N

%

142 46

75.5 24.5

87 16

84.5 15.5

148 40

78.7 21.3

86 17

84 16

149 39

79.3 20.7

81 22

78.6 21.4

21 167

11.2 88.8

7 95

6.9 93.1

172 16

91.5 8.5

95 8

92.2 7.8

0.751

63.3 36.7 0.2

N

0.086

67.1 32.9 0.1

Sudden death

0.353

67.1 32.9 2.94

Difference between (a) and (b)

Suicide (b)

0.658

17.7 82.3 Fisher’s ex test ¼ 1.000 96.2 3.9

Chi2

p-Value

Chi2

p-Value

3.17

0.075

0.07

0.794

0.96

0.327

0.02

0.897

0.02

0.902

5.18

0.023

1.41

0.236

3.89

0.049

0.05

0.826

Fisher’s ex test ¼ 0.111

30 42

41.7 58.3

37 42

46.8 53.1

0.41

0.523

81 106

43.3 56.7

58 45

56.3 43.7

4.49

0.034

0.06

0.810

38 29

52.1 39.7

24 38

30.4 48

9.41

0.009

56 101

29.9 54.0

25 64

24.3 62.1

1.80

0.407

11.56

0.003

6

8.2

17

21.5

30

16.0

14

13.6

12

16.4

8

10.2

84

44.9

60

58.3

20.54