The suicidal process; prospective comparison between early and later stages

Journal of Affective Disorders 82 (2004) 43 – 52 www.elsevier.com/locate/jad

Research report

The suicidal process; prospective comparison between early and later stages Jan Neeleman a,*, Ron de Graaf b, Wilma Vollebergh b a Julius Center for Health Sciences and Primary Care and Division of Psychiatry, University Medical Centre Utrecht Internal Mail Address STRG. 131, PO Box 85500, 3584 GA Utrecht, Netherlands b Trimbos Institute, P.O. Box 725 3500 AS Utrecht, Netherlands

Received 3 June 2003; received in revised form 18 September 2003; accepted 22 September 2003

Abstract Background: Mechanisms contributing to suicidal behaviour may differ according to how far individuals have progressed through the suicidal process. Methods: Lifetime and subsequent 12-month cumulative incidences were obtained of death ideation, death wishes, suicide contemplation and deliberate self-harm in a survey (n = 5618) of Dutch adults. Mokken’s scale analysis was used to examine whether these were compatible with underlying lifetime and 12-month severity dimensions of suicidality. Sociodemographic details and personality traits, 12-month occurrence of negative life events, hopelessness and CIDI-generated DSM-III-R diagnoses were obtained. Ordered logistic regression was used to examine whether the effects of these on the 12-month incidence of suicidality differed by individuals’ prior lifetime history of suicidality. Results: Mental illness was more strongly associated with 12-months’ suicidality in the presence (ordered logistic regression coefficient 0.80, 95% confidence interval 0.48 – 1.12) than the absence of previous suicidality (0.49 [0.25 – 0.74]). The reverse obtained for negative life events (0.18 [0.08 – 0.29] vs. 0.40 [0.33 – 0.48], respectively) and a number of sociodemographic risk factors. Female gender was a risk factor for 12-months’ suicidality (0.47 [0.28 – 0.66]) only when it was first-onset. Limitations: Completed suicides were not recorded. Self-report of lifetime suicidal behaviour may be biased. Conclusion: Environmental influences on suicidal behaviour are most pronounced early in the suicidal process which, after it has progressed, becomes more autonomous and intricately linked with mental illness and depression in particular. Men progress through the suicidal process faster than women. Management of suicidal behaviour depends on the stage of the process the person is at. D 2003 Elsevier B.V. All rights reserved. Keywords: Deliberate self harm; Hopelessness; Suicidal process

1. Introduction Suicidal acts like deliberate self-harm (DSH) are often preceded by milder expressions of suicidality like death ideation or suicide contemplation. This * Corresponding author. Tel.: +31-30-2539341; fax: +31-302539028. E-mail address: [email protected] (J. Neeleman). 0165-0327/$ - see front matter D 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.jad.2003.09.005

progression, in individuals, from mild to more severe forms of suicidality, has been referred to as the suicidal process (Runeson et al., 1996; Van Heeringen et al., 2000). Evidence suggests that, once a person has entered it, i.e. experienced suicidality, however mild, he or she may become more vulnerable to future suicidal behaviour as a result. This is an example of vulnerability accumulation or ‘‘scarring’’ (Post, 1992; Segal et al., 1996) and it would imply that effects of

44

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

factors like life stress, socio-economic circumstances or mental illness on the risk of suicidal behaviour differ according to how far individuals have progressed through the suicidal process before. Indeed, obvious precipitating recent life events appear to be reported more frequently by first time than repeat selfharmers (Arensman and Kerkhof, 1996). However, the scarring hypothesis, implying that later stages in a suicidal career are more autonomous and hence, perhaps, less amenable to prevention than earlier ones, has not been examined directly yet and prospective evidence in particular is lacking. We have therefore undertaken the present analysis in a large Dutch population survey. The focus is on the general idea that the risk factor profile differs between earlier and later stages of the suicidal process, as implied by the ‘‘scarring’’ hypothesis (Post, 1992; Segal et al., 1996). Consistent with the notion of suicidality being a process rather than a once-off phenomenon, the outcome is specified not as a dichotomy but rather as a severity dimension ranging from death ideation through death wishes and suicide contemplation, to DSH (Williams and Pollock, 2000).

2. Method 2.1. Subjects Data were obtained from the baseline (t1; 1996) and second (t2; 1997) waves of the Netherlands Mental Health Survey and Incidence Study (NEMESIS), designed to chart incidence and prevalence of mental illness in Dutch adults. Data collection procedures have been described fully elsewhere (Bijl et al., 1998). Baseline sampling was stratified according to urban density. Municipalities (90) were selected according to size and location, and households according to municipalities’ population size. In selected households, the person (between 18 and 65 years old) with the most recent birthday, was asked to participate. The response rate was 69.7%, giving 7076 participants at t1. The participants in the survey reflected the Dutch population well in terms of gender, marital status and urbanicity (Bijl et al., 1998). Only the 18 – 24 years age group was underrepresented. Of the baseline sample, 5618 (79.4%) also participated at t2.

2.2. Data Data collection procedures were similar on both occasions. Trained interviewers assessed participants in their homes using the computerised version of the Composite International Diagnostic Interview (CIDI) (Smeets and Dingemans, 1993) generating DSM-III-R diagnostic categories (American Psychiatric Association, 1987). The CIDI interview contains four items on suicidality covering, respectively, death ideation (Have you had a period of 2 weeks or more during which you were preoccupied with your own death, others’ death or dying in general?), death wishes (Have you had a period of 2 weeks or more during which you wanted to be dead?), suicide contemplation (Have you been so down that you thought of committing suicide?) and actual DSH (Have you attempted suicide?). Participants were not given a narrower definition of attempted suicide. At t1 these were asked on a lifetime basis and at t2 they referred to the year elapsed since the previous assessment. At t 2 participants also completed Eysenck’s personality inventory (14 items; (Eysenck, 1959)). Socio-economic position (educational attainment, income, employment status), living arrangement, country of birth and religious affiliation (membership of a religious congregation) and salience (stated importance of respondents’ own religious beliefs) were assessed at t1 and family history of completed suicide at t2. The occurrence of nine types of negative life events (relating to respondents’ personal health and that of important others; negative changes in respondents’ life or that of others like divorce, death or sacking; negative changes in respondents’ living arrangement or social network; anticipation of negative events like impending redundancy; failure to attain important goals and other negative events in respondents’ own lives or that of important others) during the preceding 12 months was also established at t2. 2.3. Suicidality Mokken’s scale analysis for dichotomous items (Mokken, 1971) was applied to the 4 CIDI items covering suicidality scored at the subsequent interviews. This procedure, implemented in the program MSP5 (Molenaar and Sijtsma, 2000), examines whether scores on series of dichotomous (no – yes) items are compatible with a single latent, underlying dimension.

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

MSP5 gives an overall scalability coefficient H (ranging from 0 to 1) which approaches 1 if most persons replying affirmatively to the most severe item (in this case; DSH) also report positively on each of the less severe items. Responses to series of dichotomous items are considered compatible with an underlying dimension if the overall scale H is 0.5 or higher and the H coefficients for the individual items 0.3 or higher (Mokken, 1971) in which case a single score can be constructed from them to obtain a representation of respondents’ position (person parameter) on the dimension, in this case suicidality. The suicidality score at t2 referring to the previous year was the outcome of interest, and the suicidality score at t1, referring to the suicidality during the lifetime already elapsed until then, analysed as one of its principal predictors. 2.4. Predictors Next to prior lifetime suicidality, we focused on well-known risk factors for suicidality including gender and age (Bille Brahe, 1993), low socio-economic status (Gunnell et al., 1995), unemployment (Platt, 1984), neuroticism (Neeleman et al., 1998), negative life events (Arensman and Kerkhof, 1996), mental illness (Beautrais et al., 1996) and hopelessness during the preceding year (Beck et al., 1985). Some less well-established risk factors were included as well like immigration status (Neeleman et al., 1996) and religious affiliation and salience (felt importance of one’s own religious belief) (Neeleman et al., 1997). Hopelessness is one of the six facultative symptoms (next to appetite/weight problems, insomnia, fatigue, low self-esteem and poor self-confidence, undecisiveness and concentration problems) required for the DSMIII-R diagnosis of dysthymia (American Psychiatric Association, 1987). To allow separation of the links of mental illness (including dysthymia) and hopelessness with suicidality, the diagnosis of dysthymia was reformulated excluding hopelessness as a contributing symptom, by requesting the sumtotal of the remaining 5 facultative symptoms to be 1 or more instead of 2 or more. This resulted in minimal changes in the number of diagnoses of dysthymia between t1 and t2 (80/5618 instead of 73/5618). A similar approach was used to separate the outcome of interest, suicidality, from the diagnosis of major depression for which a positive reply to any of the four suicide items counts as a

45

positive score on one of the nine symptom groups (depression/anhedonia, loss of interest, weight/appetite problems, sleep trouble, agitation or retardation, fatigue, feelings of worthlessness or guilt, impaired concentration, thinking or deciding, suicidality) for which a total of five or more is required by DSM-III-R (American Psychiatric Association, 1987). The diagnosis major depression was reformulated by requesting the sumtotal of these symptom groups, excluding suicidality, to be 4 instead of 5 for a diagnosis of major depression. The number of major depressions between t1 and t2 rose from 310 to 415 as a result. 2.5. Analysis Ordered logistic regression was used to examine associations between the selected predictors and the 12-month suicidality score between t1 and t2. This procedure is based on a threshold-liability model and assumes that a latent continuous trait (dimension) underlies an ordinal outcome (Greenland, 1985). The model estimates the thresholds and generates regression coefficients with 95% confidence intervals (CI). The regression coefficients depend on persons’ position (the person parameter) on the underlying dimension. If smaller than zero they indicate negative, and if larger than zero, positive associations. The crude, unadjusted associations between the selected risk factors and the 12-month suicidality score were then examined for modification or confounding by prior lifetime suicidality scores. Likelihood ratio (LRI) tests were used to assess the significance of interaction terms which were retained in the model if p < 0.050. Finally, multivariate ordered logistic models were specified for the best and most parsimonious prediction of 12-month suicidality scores according to levels of prior lifetime suicidality, covariates being retained in the model only if their p-values, estimated by LR tests, were 0.050 or smaller.

3. Results 3.1. Lifetime and 12-month suicidality The life-time cumulative incidence of DSH in the full sample (n = 7076) was 2.9% and in the subsample (n = 5618) that was assessed on both occasions, 2.8%

46

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

(n = 155). The remainder of these results refers to this subsample. The 12-month cumulative incidence of DSH (between t1 and t2) was 0.9%. Milder expressions of suicidality occurred at far higher rates, up to 35% for lifetime death ideation (Table 1). Mokken scale analysis for dichotomous items yielded scalability coefficients larger than 0.5 supporting the existence of a latent dimension underlying the responses to the four suicide items (Table 1). However, there was no clear difference in cumulative incidence between the death wish and suicide contemplation items. Therefore, for both lifetime and 12-month suicidality scores, subjects were scaled from 4 (DSH) to 3 (death wish or suicide contemplation but no DSH), to 2 (death ideation but no more than that) to 1 (the remainder). 3.2. Twelve-month suicidality scores; crude predictors The association between the lifetime and the 12month annual suicidality scores was given by a linear coefficient of 0.94 (95% CI 0.86, 1.02); p < 0.001, corresponding with a predicted 12-month cumulative incidence of DSH in persons without any prior suicidality of 0.37%, rising to 5.9% in case of individuals who previously self-harmed. There was no evidence to reject a linear association between the 12 months and prior lifetime suicidality scores (LR test against linearity (LRL) v2 = 3.0, df = 2; p = 0.228). Higher 12-month suicidality scores were, overall, associated with female gender, lower income and educational attainment, lack of paid employment, single status (especially if combined with parenthood), lack of religious affiliation and immigrant

status. Of the non-sociodemographic variables (assessed at t2), recent experience of negative life events, of mental illness and of hopelessness were associated with increased 12 month suicidality scores as were higher levels of neuroticism and a family history of completed suicide (Tables 2 and 3, col. 2). In the year preceding t2, 922 subjects in the sample qualified for one or more psychiatric diagnoses. There were 545 diagnoses of mood disorder (bipolar (n = 50), major depression [redefined] (n = 415), dysthymia [redefined] (n = 80)), 5 of schizophrenia, 431 anxiety, 11 eating and 264 substance/alcohol use (dependence and/or misuse) diagnoses. Each separate diagnostic group was associated with increased suicidality; the effect being weakest for substance (including alcohol) misuse/dependence (0.91 (0.64, 1.18); p < 0.001) and strongest for eating disorders (2.14 (0.95, 3.33); p < 0.001). 3.3. Twelve-month suicidality scores; effect modifiers and confounders LR-tests for interaction indicated that the associations of the 12-month suicidality score with gender (LRI v2 = 4.1; df = 1; p = 0.042), income (LRI v2 = 5.2; df = 1; p = 0.023), employment status (LRI v2 = 8.0; df = 1; p = 0.005), religious salience (LRI v2 = 4.3; df = 1; p = 0.038) negative life events (LRI v2 = 8.9; df = 1; p = 0.042) and recent mental illness (v2 = 12.0; df = 1; p < 0.001) differed according to subjects’ prior lifetime suicidality scores (Table 2). Female gender was a risk factor only in those without prior suicidal behaviour and the effects of lack of paid employment and low income status were strongest in case of

Table 1 Life-time and 12-month cumulative incidence of suicidality Lifetime suicidality at t0 Death ideation Death wish Suicide contemplation Suicide attempt Scalability coefficient, H Scale, Z Reliability, q a

5615 valid responses.

12-month suicidality at t1

Cumulative incidence (n = 5618)a

Item H

Cumulative incidence (n = 5618)a

Item H

1991 (35.0%) 597 (10.6%) 656 (11.7%) 155 (2.8%) 0.66 58.5 0.68

0.64 0.68 0.63 0.74 – – –

890 (15.9%) 163 (2.9%) 187 (3.3%) 51 (0.9%) 0.61 46.8 0.60

0.66 0.65 0.56 0.53

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

47

Table 2 Risk factors for the 12-month cumulative incidence of suicidality in 5618 adults; stratified by prior lifetime suicidality; ordered logistic regression coefficients, 95% CIs and p-values Prior lifetime history of suicidality Overall (5618)

No prior suicidality (3380)

Death ideation only (1376)

Wish/plan with or without ideation (705)

DSH (155)

Gendera

0.62 (0.48, 0.76); p < 0.001

0.55 (0.31, 0.78); p < 0.001

0.47 (0.20, 0.73); p < 0.001

0.29 ( 0.02, 0.61); p = 0.070

Incomeb

0.08 ( 0.14, 0.16); p = 0.913

Religious salience (5-point Likert scale) Negative life events (0 – 8)

0.24 ( 0.33, 0.14); p < 0.001 0.73 (0.5 0.91); p < 0.001 0.01 ( 0.03, 0.05); p = 0.647 0.50 (0.44, 0.55); p < 0.001

0.29 ( 0.50, 0.09); p = 0.005 0.55 (0.21, 0.90); p = 0.002 0.08 ( 0.18, 0.01); p = 0.079

0.57 (0.47, 0.67); p < 0.001

0.08 ( 0.24, 0.09); p = 0.355 0.55 (0.19, 0.91); p = 0.003 0.01 ( 0.09, 0.08); p = 0.908 0.37 (0.28, 0.47); p < 0.001

0.33 (0.21, 0.44); p < 0.001

Mental illness, n = 922

1.48 (1.32, 1.64); p < 0.001

1.11 (0.82, 1.39); p < 0.001

0.92 (0.62, 1.22); p < 0.001

1.33 (1.01, 1.64); p < 0.001

0.21 ( 0.84, 0.41); p = 0.505 0.18 ( 0.61, 0.24); p = 0.392 0.72 (0.12, 1.32); p = 0.019 0.05 ( 0.22, 0.12); p = 0.588 0.10 ( to 0.11, 0.30); p = 0.348 1.50 (0.88, 2.12); p < 0.001

Employmentc

0.04 ( 0.33, 0.41); p = 0.846 0.03 ( 0.04, 0.11); p = 0.382

a

U (3007) vs. h (2611). < 2601 Euro (1565), 2600-4400 Euro (2163), >4400 (1648). c 731 unemployed vs. 4879 employed. b

higher cumulative lifetime suicidality. The association between recent negative life events and 12-month suicidality was most pronounced in those without prior suicidality whilst, conversely, recent mental illness was associated with suicidality most strongly in persons with relatively high levels of prior lifetime suicidality. There was a non-significant indication that in persons with higher levels of lifetime suicidality, higher religious salience conferred a protective effect. Effects of the other variables (age, education, living situation, family history of suicide, religious affiliation, coutnry of birth, neuroticism and hopelessness) on 12 month suicidality did not differ according to whether or not there had been previous lifetime suicidality (all LRI tests giving p-values>0.050) and these were therefore adjusted for it (Table 3). This resulted, on the whole, in attenuation of regression coefficients but most of them remained significant. Higher academic achievement, single parenthood, a family history of suicide, birth in foreign country, neuroticism and hopelessness were associated with increased 12-month suicidality irrespective of the previous history (Tables 3 and 4).

3.4. Best model for 12 month suicidality according to prior levels of suicidality As the association of 12 month suicidality scores with important risk factors like mental illness differed according to prior lifetime suicidality, models were specified separately for those that had experienced no more than mere death ideation during their lifetime and for those who, prior to t1, had progressed to death wishes, suicide contemplation or actual DSH. The two baseline models contained all variables listed in Tables 2 and 3. Variables were removed from the models if their contribution was non-significant ( p>0.050), as assessed by LR tests. Removal was stepwise with the weakest contributors being deleted first. The final most parsimonious models for 12month suicidality scores were highly different according to the suicidality levels that had been reached before. The link of mental illness with suicidality was strongest when prior suicidality had been high in which case hopelessness and the presence of a family history of completed suicide also made independent contributions. Conversely, the effect of negative life

48

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

Table 3 Risk factors for the 12-month cumulative incidence of suicidality in 5618 adults; crude and adjusted for prior lifetime suicidality; ordered logistic regression coefficients, 95% CIs and p-values

Age

Education

Living with

Suicide in family Religious affiliation Country of birth Neuroticism Hopelessness

18 – 34 (1669) 35 – 49 (2232) 50 – 65 (1717) Lower (1442) Intermediate (2056) Vocational (1579) Academic (471) Partner (1716) Partner + children (2243) Single parent (242) Alone (1061) Other (356) No (5497) Yes (97) No 2568 Yes 3049 Netherlands (5267) Elsewhere (351) Standardized No (5516) Yes (99)

Unadjusted for prior suicidality

Adjusted for prior suicidality

0.02 (

0.04 (

0.09 (

0.07, 0.11); p = 0.684

0.17,

0.02); p = 0.015

(1) 0.12 ( 0.29, 0.05); p = 0.177 0.82 (0.53, 1.12); p < 0.001 0.24 (0.04, 0.43); p = 0.017 0.02 ( 0.29, 0.32); p = 0.918 (1) 0.84 (0.42, 1.26,); p < 0.001 (1) 0.20 ( 0.34, 0.07); p = 0.004 (1) 0.35 (0.09, 0.61); 0.009 0.77 (0.71, 0.83); p < 0.001 (1) 2.62 (2.24, 3.01); p < 0.001

events was strongest in the absence of prior suicidality. The effects of neuroticism were comparable in both groups. Annual suicidality scores were higher in

0.05, 0.13); p = 0.352

0.08 (

0.16,

0.01); = 0.042

(1) 0.09 ( 0.27, 0.09); p = 0.311 0.39 (0.08, 0.71); p = 0.015 0.01 ( 0.19, 0.22); p = 0.897 0.04 ( 0.27, 0.35); p = 0.802 (1) 0.70 (0.26, 1.14); p = 0.002 (1) 0.11 ( 0.26, 0.03); p = 0.125 (1) 0.37 (0.09, 0.64); 0.009 0.59 (0.52, 0.65); p < 0.001 (1) 1.80 (1.41, 2.19); p < 0.001

women than men but only as regards new onset suicidality. Apart from negative life events, sociodemographic variables did not contribute to the risk

Table 4 Best independent predictors of the 12 month cumulative incidence of suicidality in 5618 adults; multiple ordered logistic regression, stratified by prior lifetime suicidality; ordered logistic regression coefficients, 95% CIs and p-values

Gender Income Educational attainment Living arrangement With Partner (1716) Partner and children (2243) Single parent (242) Alone (1061) Other (356) Negative life events Mental illness Neuroticism Family history of suicidal behaviour Hopelessness

No prior suicidality or ideation only, n = 4756

Death wishes, suicide ideation/plan and/or attempts, n = 860

0.47 (0.28, 0.66); < 0.001 0.18 (0.04, 0.32); 0.014 0.11 ( 0.22, 0.01); 0.044

– – –

(1) 0.09 ( 0.31, 0.12); 0.404 0.65 (0.23, 1.08); 0.003 0.15 ( 0.46, 0.15); 0.316 0.31 ( 0.10, 0.72); 0.140 0.40 (0.33, 0.48); < 0.001 0.49 (0.25, 0.74); < 0.001 0.41 (0.30, 0.51); < 0.001 –

–

– LR test v2 = 7.3; df = 7; p = 0.404

0.18 0.80 0.41 0.83

(0.08, (0.48, (0.30, (0.06,

0.29); 1.12); 0.52); 1.60);

0.001 < 0.001 < 0.001 0.035

0.73 (0.21, 1.25); 0.006 LR test v2 = 9.9; df = 12; p = 0.628

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

of further suicidality in those with prior experience of it even though, in the univariate analyses (Tables 2 and 3), the effects of unemployment and low income status were stronger on repeat than new suicidality. New onset suicidality was associated with educational attainment, income levels and living arrangement, lone parents being at increased risk compared to individuals living in partnerships or even single individuals. All else being equal, low educational attainment was an independent risk factor for new onset suicidality as where higher income levels. 3.5. Attrition Of the 7076 subjects that had participated at baseline 1458 (21%) had dropped out by t2. Attrition was unrelated to a lifetime history of any mental illness (odds ratio 1.06; 95% CI 0.95, 1.19; p = 0.302). A lifetime history of suicide attempts was a stronger predictor of attrition but this effect also failed to reach significance (odds ratio 1.25; 95% CI 0.90, 1.73; p = 0.176).

4. Discussion First-onset suicidal behaviour has more and stronger links with social and economic risk factors and negative life events than repeat suicidality. Conversely, later on in the suicidal process, mental illness, hopelessness and a positive family history of completed suicide are the main associations of suicidality. This suggests, compatible with the study’s main hypothesis, that repeat suicidality has a more autonomous character and is less dependent on external influences, than first-onset suicidality. It concurs with the general idea that there are differences between early and later phases of the suicidal process (Van Heeringen et al., 2000). For instance, first-onset, less serious suicidal behaviour has been described as a means of protest, and more serious suicidality, often taking place in the context of a prior history, as an expression of defeat and hopelessness (Williams and Pollock, 2000). Incidence figures for the separate items are in line with those reported from other sources. Internationally, the cumulative lifetime incidence of DSH ranges from 0.7% to 5.9% and of suicide contem-

49

plation from 2.1% to 11.7% (Weissman et al., 1999). Death wishes and suicide contemplation are usually referred to as suicide ideation in the literature. Its lifetime cumulative incidence has been reported to be around 15% (Kessler et al., 1999; Pirkis et al., 2000) and its 12-month cumulative incidence 3.4% (Pirkis et al., 2000). Only few studies have considered death ideation as part of the suicidal spectrum (Deykin and Buka, 1994) even though it figures among the facultative DSM criteria for major depressive episode next to DSH, suicide plans and suicide ideation (American Psychiatric Association, 1987). Suicidality, in this study, was conceived of as a continuum ranging from fleeting death ideation at its mildest, to actual DSH at its severest end. Scale analysis of the four suicide items lent support to this idea of the suicidal continuum, which has also been described by others (Lewinsohn et al., 1996). No specification was given to participants about what they should consider ‘‘attempted suicide’’. This is line with current thinking in which neither medical seriousness nor suicidal intent are part of the definition of DSH (Williams, 1997). It has to be noted that the wording of the suicide contemplation item in the CIDI version used in the present study is such that it cannot be equated with concrete suicide planning which is usually considered an important component of the suicidal continuum (Kessler et al., 1999). This lack of information on whether or not concrete suicide plans ever existed and the medical seriousness of prior acts, limits the study’s potential in that the suicidality dimension can be expected to only refer to the depressive and not the impulsivity component of suicidal behaviour (Caspi et al., 1996; Neeleman et al., 1998). Completed suicides are, of course, also part of the spectrum of suicidality (Neeleman et al., 1998) but were, unfortunately, not registered. However, given the notion of the continuum, completed suicide, in all likelihood, shares many of its risk factors with milder forms of suicidal behaviour (Neeleman et al., 1998). It is therefore unlikely that attrition from the study by completed suicides would threaten the study’s overall findings since their exclusion would have lead to mitigation, not inflation, of the associations in question. This idea is also supported by the somewhat raised levels of attrition among persons with prior suicidal behaviour.

50

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

Female gender was associated with an increased risk of first-onset suicidality but not of repeated suicidal behaviour. This is in line with reports that men tend to progress through the suicidal process faster than women. It has, for instance, been reported that the time from the first suicidal communications to completed suicide in men is around 12 months compared with 42 months in women and that more completed suicides in men than women are not preceded by prior, milder suicidal behaviour (Runeson et al., 1996). Clear age effects on suicide and DSH have been described (Bille Brahe, 1993) but were not found in the present study even when the individual items were analysed. For the lifetime suicidality figures this absence of an age effect can be readily explained by noting that the longer lifetime of the elderly is likely to obscure any associations of suicidality with particular age groups. Twelve-month suicidality scores may have been too low to allow clear age effects to emerge but age effects may also be genuinely less pronounced in the Netherlands compared to elsewhere. It has been reported that DSH in the Netherlands tends to be associated with other age groups than, for instance, the UK (Grootenhuis et al., 1994). The pattern of socio-economic risk factors that were, in this survey, associated with an increased overall 12-month incidence of suicidality, is in line with existing knowledge. Being without paid employment is a known overall association of DSH (Platt, 1984) and completed suicide (Mortensen et al., 2000). Indices of low socio-economic status like low levels of income and educational attainment have also been reported in connection with DSH and completed suicide (Gunnell et al., 1995; Mortensen et al., 2000). The strong association between lone parenthood and suicidality has not been reported as such before but comes as no surprise, given the relatively poor socio-economic circumstances of many single parents. On crude analysis, indices of low socio-economic status appeared more strongly associated with repeated than first onset suicidal behaviour but once mental illness and hopelessness had been controlled for, the opposite was true. Thus, this study qualifies the existing knowledge regarding the socio-economic associations of suicidal behaviour by restricting them to first onset suicidality only.

The existence of an overall negative link between religious affiliation and suicidality concurs with many previous ecological findings (Neeleman et al., 1997) but relatively few individual-level findings have been reported on this matter (Foster et al., 1999). Protective effects of religious affiliation were, in the present study, most pronounced on first onset suicidality. However, they were no longer apparent once other socio-economic variables had been taken into account suggesting that religious affiliation has no independent effects on suicidality. Reports on the associations of suicidality with immigration status and ethnicity are divided and the direction of influence probably depends on the particular ethnic group and host country concerned (Neeleman et al., 1996). However, increased levels of suicidal behaviour among younger members of the main ethnic minority groups in the Netherlands, Surinamese, Turkish and Moroccans, have been reported before (Beker and Merens, 1994). However, as for religious affiliation, immigration status has no effects on suicidality that are independent of the effects of other socio-economic variables. Among the different types of mental illness, the impact of the mood disorders is the largest not only because of their strong association with suicidality at the level of individuals but also due to their relatively high prevalence. Mental illness is associated much more strongly with repeat than first onset suicidality and this effect outweighs that of socio-economic risk factors. This concurs with reports that serious suicide attempts occur in the context of a major depression more often in older persons who have had the opportunity to accumulate more suicidality during their lifetime than younger individuals (Beautrais et al., 1996). The importance of hopelessness as a factor in the suicidal process is confirmed by the fact that it retained its effects even after the presence of mental illness had been adjusted for. The finding that its effect was limited to repeat suicidality confirms the idea that later stages of the suicidal process reflect defeat rather than protest (Williams and Pollock, 2000) and that hopelessness is one of the best predictors of eventual suicide among persons with prior suicidality (Beck et al., 1985). Repeat but not firstonset suicidality was strongly associated with a positive family history of completed suicide. However, it must be noted that it is not possible, in this study, to

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

separate, in this association, environmental from genetic components. The former may be due, for instance, to the negative impact of being reared in an environment with suicidal or chronically mentally ill family members. However, genetic effects are also possible either acting directly on suicidality, or indirectly, through effects on mental illness with a high associated risk of suicidal behaviour (Statham et al., 2000). It is not immediately clear why these effects should be stronger on repeat than first-onset suicidality but it is consistent with the fact that depression, the main single risk factor for suicidal behaviour, especially in the elderly (Conwell et al., 1996), has a larger genetic component when it is recurrent (Sullivan et al., 2000). There has been discussion over whether suicide prevention should be the sole responsibility of the health services or whether it is also a matter of social policy (Gunnell and Frankel, 1994). The present study suggests that social policy measures, targeting for instance deprivation or marital dissolution, may help reduce the numbers of persons entering the suicidal process. Around half of all first time attempters do not repeat and for them, social work or crisis intervention may be all that is needed. However, on the other hand, repeat suicidal behaviour, which accounts for a substantial proportion, around 50% (Sakinofsky, 2000), of all suicidality in the population, is unlikely to be greatly affected by minimal interventions. Secondary prevention in this group in all likelihood requires a multimodal approach combining, probably, pharmacotherapy to alleviate depression (Verkes et al., 2000), and more specialized psychotherapeutic approaches targeting hopelessness (Williams, 1997) which, as this study has reconfirmed, is at the core of recurrent and thus serious suicidality.

Acknowledgements This work was done whilst J.N. was supported by the Dutch Organisation for Scientific Research (Pionier Grant No. 900-00-002). The expert help of S. van Dorsselaar at the Trimbos institute is greatly acknowledged. This work was done whilst J.N. was supported by the Dutch Organisation for Scientific Research (Pionier Grant No. 900-00-002).

51

References American Psychiatric Association, Wilma, 1987. Diagnostic and Statistical Manual of Mental Disorders: DSM-III-R APA, Washington. Arensman, E., Kerkhof, A.J.F.M., 1996. Classification of attempted suicide: a review of empirical studies, 1963 – 1993. Suicide Lifethreat Behav. 26, 46 – 64. Beautrais, A.L., Joyce, P.R., Mulder, R.T., Fergusson, D.M., Deavoll, B.J., Nightingale, S.K., 1996. Prevalence and comorbidity of mental disorders in persons making serious suicide attempts: a case-control study. Am. J. Psychiatry 153, 1009 – 1014. Beck, A.T., Steer, R.A., Kovacs, M., Garrison, B., 1985. Hopelessness and eventual suicide; a 10 year prospective study in patients hospitalised with suicidal ideation. Am. J. Psychiatry 145, 559 – 563. Beker, M., Merens, J.G.F., 1994. Rapportage Jeugd Sociaal en Cultureel Planbureau, Rijswijk. Bijl, R.V., van Zessen, G., Ravelli, A., de Rijk, C., Langendoen, Y., 1998. The Netherlands Mental Health Survey and Incidence Study (NEMESIS); objectives and design. Soc. Psychiatry Psychiatr. Epidemiol. 33, 516 – 581. Bille Brahe, U., 1993. The role of age and sex in suicidal behaviour. Acta Psychiatr. Scand., Suppl. 371, 21 – 27. Caspi, A., Moffitt, T.E., Newman, D.L., Silva, P.A., 1996. Behavioral observations at age 3 years predict adult psychiatric disorders. Arch. Gen. Psychiatry 53, 1033 – 1039. Conwell, Y., Duberstein, P.R., Cox, C., Herrmann, J., Forbes, N.T., Caine, E.D., 1996. Relationship of age and axis I diagnosis in victims of completed suicide: a psychological autopsy study. Am. J. Psychiatry 153, 1001 – 1008. Deykin, E.Y., Buka, S.L., 1994. Suicidal ideation and attempts among chemically dependent adolescents. Am. J. Publ. Health 84, 634 – 639. Eysenck, H.J., 1959. Manual of the Maudsley Personality Inventory University of London Press, London. Foster, T., Gillespie, K., McClelland, R., Patterson, C., 1999. Risk factors for suicide independent of DSM-III-R axis 1 disorder. Br. J. Psychiatry 175, 175 – 179. Greenland, S., 1985. An application of logistic models to the analysis of ordinal response. Biom. J. 27, 189 – 197. Grootenhuis, M., Hawton, K., van Rooijen, L., Fagg, J., 1994. Attempted suicide in Oxford and Utrecht. Br. J. Psychiatry 165, 73 – 78. Gunnell, D., Frankel, S., 1994. Prevention of suicide: aspirations and evidence. BMJ 308, 1227 – 1233. Gunnell, D.J., Peters, T.J., Kammerling, R.M., Brooks, J., 1995. Relation between parasuicide, suicide, psychiatric admissions and socioeconomic deprivation. BMJ 311, 226 – 230. Kessler, R.C., Borges, G., Walters, E.E., 1999. Prevalence and risk factors for lifetime suicide attempts in the National Comorbidity Survey. Arch. Gen. Psychiatry 56, 617 – 626. Lewinsohn, P.M., Rohde, P., Seeley, J.R., 1996. Adolescent suicidal ideation and attempts: prevalence, risk factors, and clinical implications. Clin. Psychol.: Sci. Pract. 3, 25 – 46.

52

J. Neeleman et al. / Journal of Affective Disorders 82 (2004) 43–52

Mokken, R.J., 1971. A Theory and Procedure of Scale Analysis; With Applications in Political Research Mouton, The Hague. Molenaar, I.W., Sijtsma, K., 2000. MSP5 for Windows ProGAMMA, Groningen. Mortensen, P.B., Agerbo, E., Erikson, T., Qin, P., WestergaardNielsen, N., 2000. Psychiatric illness and risk factors for suicide in Denmark. Lancet 355, 9 – 12. Neeleman, J., Jones, P., van Os, J., Murray, R.M., 1996. Parasuicide in Camberwell; ethnic differences. Soc. Psychiatry Psychiatr. Epidemiol. 31, 284 – 287. Neeleman, J., Halpern, D., Leon, D., Lewis, G., 1997. Tolerance of suicide, religion and suicide rates; an ecological and individual-level study in 19 western countries. Psychol. Med. 27, 1165 – 1171. Neeleman, J., Wessely, S., Wadsworth, M., 1998. Predictors of suicide, accidental death and premature natural death in a general-population birth-cohort. Lancet 351, 93 – 97. Pirkis, J., Burgess, P., Dunt, D., 2000. Suicidal ideation and suicide attempts among Australian adults. Crisis 21, 16 – 25. Platt, S., 1984. Unemployment and suicidal behaviour; a review of the literature. Soc. Sci. Med. 19, 93 – 115. Post, R.M., 1992. Transduction of psychosocial stress into the neurobiology of recurrent affective disorder. Am. J. Psychiatry 149, 999 – 1010. Runeson, B.S., Beskow, J., Waern, M., 1996. The suicidal process in suicides among young people. Acta Psychiatr. Scand. 93, 35 – 42. Sakinofsky, I., 2000. Repetition of suicidal behaviour. In: Hawton, K., van Heeringen, K. (Eds.), The International Handbook of Suicide and Suicidal Behaviour. Wiley, Chichester, pp. 385 – 404. Segal, Z.V., Williams, J.M., Teasdale, J.D., Gemar, M., 1996. A cognitive science perspective on kindling and episode sensitization in recurrent affective disorder. Psychol. Med. 26, 371 – 380.

Smeets, R.M.W., Dingemans, P.M.A.J., 1993. Composite International Diagnostic Interview, Version 1.1 WHO, Amsterdam. Statham, D.J., Heath, A.C., Madden, P.A., Bucholz, K.K., Beirut, S.H., Dinwiddie, S.H., Slutske, W.S., Dunne, M.P., Martin, N.G., 2000. Suicidal behaviour; an epidemiological and genetic study. Psychol. Med. 28, 839 – 855. Sullivan, P.F., Neale, M.C., Kendler, K.S., 2000. Genetic epidemiology of depression: review and meta-analysis. Am. J. Psychiatry 157, 1552 – 1562. Van Heeringen, K., Hawton, K., Williams, J.M.G., 2000. Pathways to suicide; an integrative approach. In: Hawton, K., van Heeringen, K. (Eds.), The International Handbook of Suicide and Suicidal Behaviour. Wiley, Chichester, pp. 223 – 234. Verkes, R.J., van der Mast, R.C., Hengeveld, M.W., Zwinderman, J.P., Tuyl, J.P., van Kempen, G.J.M., 2000. Reduction by paroxetine of suicidal behavior in patients with repeated suicide attempts but not major depression. Am. J. Psychiatry 155, 543 – 547. Weissman, M.M., Bland, R.C., Canino, G.J., Greenwald, S.S., Joyce, R.P., Lee, C.K., Karam, E.G., Lellouch, J., Lepine, S.C., Newman, S.C., Rubio-Stipec, M., Wells, J., Wickramaratne, P.J., Wittchen, H.U., Yeh, E.K., 1999. Prevalence of suicide ideation and suicide attempts in nine countries. Psychol. Med. 29, 9 – 17. Williams, J.M.G., 1997. Cry of Pain: Understanding Suicide and Deliberate Self-Harm. Penguin, London, p. 69. Williams, J.M.G., Pollock, L.R., 2000. Psychology of suicidal behaviour. In: Hawton, K., van Heeringen, K. (Eds.), The International Handbook of Suicide and Suicidal Behaviour. Wiley, Chichester, pp. 79 – 83.