Introduction
Several studies have reported inverse associations between cognitive ability and suicide (Allebeck et al. Reference Allebeck, Allgulander and Fisher1988; O'Toole & Cantor, Reference O'Toole and Cantor1995; Gunnell et al. Reference Gunnell, Magnusson and Rasmussen2005; Hemmingsson et al. Reference Hemmingsson, Melin, Allebeck and Lundberg2006; Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008; Osler et al. Reference Osler, Nybo Andersen and Nordentoft2008; Gravseth et al. Reference Gravseth, Mehlum, Bjerkedal and Kristensen2009) and between cognitive ability and attempted suicide (Jiang et al. Reference Jiang, Rasmussen and Wasserman1999; Fergusson et al. Reference Fergusson, Horwood and Ridder2005; Osler et al. Reference Osler, Nybo Andersen and Nordentoft2008; Alati et al. Reference Alati, Gunnell, Najman, Williams and Lawlor2009; Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010). However, the mechanisms underlying these associations are not fully understood. Various pathways have been suggested as contributing to the relationship. Childhood socio-economic position does not appear to confound the association between cognitive ability and suicide (Allebeck et al. Reference Allebeck, Allgulander and Fisher1988; Hemmingsson et al. Reference Hemmingsson, Melin, Allebeck and Lundberg2006), but socio-economic position in adulthood appears more important (Hemmingsson et al. Reference Hemmingsson, Melin, Allebeck and Lundberg2006). Mental illness, an established risk factor for suicide and attempted suicide (Hagnell et al. Reference Hagnell, Lanke and Rorsman1981; Allebeck & Allgulander, Reference Allebeck and Allgulander1990b; Fergusson et al. Reference Fergusson, Woodward and Horwood2000; Tidemalm et al. Reference Tidemalm, Elofsson, Stefansson, Waern and Runeson2005; Nock et al. Reference Nock, Borges, Bromet, Cha, Kessler and Lee2008), has been suggested as a confounder as well as a mediator (Batty et al. Reference Batty, Deary and Gottfredson2007). Controlling for mental illness has been found to attenuate the associations between intelligence quotient (IQ) and suicidal behaviour (Osler et al. Reference Osler, Nybo Andersen and Nordentoft2008). However, among people with psychotic illness, no increased risk with lower IQ has been found (Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008; Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010; Webb et al. Reference Webb, Langstrom, Runeson, Lichtenstein and Fazel2011). Further, the impacts of different measures of maladjustment and deviant behaviour on the association between cognitive ability and suicide attempt have been investigated among young adults, yielding inconsistent results (Fergusson et al. Reference Fergusson, Horwood and Ridder2005; Alati et al. Reference Alati, Gunnell, Najman, Williams and Lawlor2009).
Brezo et al. (Reference Brezo, Paris and Turecki2006) found in a systematic review that some aspects of personality are associated with suicidal behaviour, although few of the reviewed studies were prospective and conducted with representative population-based samples. There is also some evidence that cognitive ability is associated with certain aspects of personality, although findings are inconsistent and mechanisms unclear (Ackerman & Heggestad, Reference Ackerman and Heggestad1997; Chamorro-Premuzic & Furnham, Reference Chamorro-Premuzic and Furnham2004). Moreover, in a small US community sample, an association between education, which is highly correlated with intelligence, and self-reported suicide attempts was partly explained by the personality traits self-directedness and harm avoidance (Grucza et al. Reference Grucza, Przybeck and Cloninger2005).
Low cognitive ability is also associated with some indicators of psychological distress as well as later suicidal behaviour. Self-reported symptoms, such as depressive feelings, irritability and somatic symptoms were found to be associated with childhood cognitive ability in the 1958 National Child Development Survey and the 1970 British Cohort Study (Gale et al. Reference Gale, Hatch, Batty and Deary2009). In the Isle of Wight Study, interview-based reports of worry and irritability in adolescence were associated with self-reported suicide plans and suicide attempts in middle age (Pickles et al. Reference Pickles, Aglan, Collishaw, Messer, Rutter and Maughan2010).
Family formation in adulthood might also contribute to the association of IQ with suicidal behaviour. In Scottish cohorts, men with higher cognitive ability in childhood tended to be more likely to get married (Taylor et al. Reference Taylor, Hart, Smith, Whalley, Hole, Wilson and Deary2005). Being married (Johansson et al. Reference Johansson, Sundquist, Johansson, Qvist and Bergman1997; Andres et al. Reference Andres, Collings and Qin2010) and having children in the household (Andres et al. Reference Andres, Collings and Qin2010) are in turn both associated with a reduced risk of suicide.
In the present study we investigated the relationship between cognitive ability among young men, measured at compulsory military conscription in 1969–1970, and suicide and suicide attempt over a 36-year follow-up period. Such associations have previously been shown in a larger cohort of Swedish conscripts (Gunnell et al. Reference Gunnell, Magnusson and Rasmussen2005; Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010), adjusting for a limited number of relevant factors. The aim in the present study was to investigate the extent to which the associations between cognitive ability in early adulthood and later completed and attempted suicide can be explained by the following factors: socio-economic position in childhood and adulthood; indicators of maladjustment, mental illness and personality in young adulthood; and later family formation. We used a cohort of 49 321 Swedish men, born in 1949–1951, who were conscripted in 1969–1970. In this database we had access to data on psychological, psychiatric and social conditions from questionnaires, psychological interviews and medical examinations. Record linkages with national censuses provided information on socio-economic conditions and family formation.
Method
Study population
The study was based on detailed health, psychometric and socio-economic data collected on 49 321 young Swedish male conscripts aged 18–20 years, who were conscripted for compulsory military service in 1969 and 1970. The background of the Swedish conscription surveys and the variables included have been presented in detail elsewhere (Allebeck & Allgulander, Reference Allebeck and Allgulander1990a; Larsson et al. Reference Larsson, Hemmingsson, Allebeck and Lundberg2002; Karlsson et al. Reference Karlsson, Ahlborg, Dalman and Hemmingsson2010). Only 2–3% of all Swedish men were exempted from conscription at that time, in most cases due to severe handicaps or congenital disorders. Of all men conscripted in 1969 and 1970, 98% were born in 1949–1951; the remaining 2% were born before 1949. Ethical approval was granted by the Research Ethics Committee of the Karolinska Institutet, Stockholm.
Exposure
Psychometric assessments were conducted using four subtests, measuring verbal, logic-inductive and visuospatial ability and technical comprehension. The results were converted to normally distributed standard-nine (stanine) scales for each subtest, with scores 1–9. These were combined and transformed into a new stanine scale as a global measure of general ability, corresponding to approximate IQ bands of <74, 74–81, 82–89, 90–95, 96–104, 105–110, 111–118, 119–126 and >126. Of the men, 49 262 (99.9%) had a score on cognitive ability (Hemmingsson et al. Reference Hemmingsson, Melin, Allebeck and Lundberg2006).
Outcomes
Information on suicide mortality 1971–2006 was obtained from the National Cause of Death Register and information on attempted suicide from the Swedish Hospital Discharge Register, for which data were available for 1973–2006. We classified suicide and suicide attempt through the following International Classification of Diseases (ICD) codes: suicide, E950–9 (ICD-8/9) and X60–X84 (ICD-10); deaths of undetermined intent, E980–9 (ICD-8/9) and Y10–Y34 (ICD-10). Deaths of undetermined intent were included, as previous studies have shown that most of these are, in fact, likely to be suicides (Allebeck et al. Reference Allebeck, Allgulander, Henningsohn and Jakobsson1991; Ohberg & Lonnqvist, Reference Ohberg and Lonnqvist1998). Analyses excluding undetermined cases yielded similar results. The same codes were used to identify hospital admissions for suicide attempts. All men admitted to hospital and staying at least overnight were included. Reasons for hospitalization in these cases were mainly observation and psychiatric evaluation.
Covariates
Childhood socio-economic conditions
Information on socio-economic position and crowded housing during the conscripts' childhood, at age 9–11 years, was obtained from the National Population and Housing Census of 1960. Socio-economic position was based on the occupation of the head of household, most often the father. Occupation was classified into six socio-economic groups: (1) unskilled workers; (2) skilled workers; (3) assistant non-manual employees; (4) non-manual employees at intermediate or higher level; (5) farmers; (6) those not classified in any other socio-economic group. Crowded housing was classified as more than two people per room (kitchen not included). Body height was measured at conscription and is used as an indicator of poor circumstances in childhood. Short stature was classified as height lower than 171 cm when investigating its distribution in the IQ categories, in accordance with previous studies (Hemmingsson et al. Reference Hemmingsson, Melin, Allebeck and Lundberg2009).
Conscription data
A medical examination and a psychological interview were carried out at conscription. In cases where psychiatric illness was suspected or reported, the conscript was referred to a psychiatrist for diagnosis. Any mental disorder was recorded in accordance with the ICD-8. Nearly 5800 of the conscripts (11.7%) received a psychiatric diagnosis at their conscription assessment; 32 of these men received a diagnosis of psychosis (0.06% of the conscripts). Whilst a psychiatric diagnosis at conscription has previously been associated with lower IQ score (David et al. Reference David, Zammit, Lewis, Dalman and Allebeck2008) and also increased suicide risk (Allebeck & Allgulander, Reference Allebeck and Allgulander1990a), amongst people with a subsequent diagnosis of psychosis, higher IQ has not been shown associated with a lower risk of suicidal behaviour (Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008; Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010; Webb et al. Reference Webb, Langstrom, Runeson, Lichtenstein and Fazel2011).
Personality factors were assessed by a trained psychologist in a structured interview. The psychologist asked about, for example, anxiety, adjustment problems and conflicts, as well as about successes, taking responsibility and emotional coping in critical situations. Personality factors were rated on a scale from one to five. A low rating of ‘emotional control’ was given to conscripts who seemed to lack an ability to control nervousness and anxiousness, channel aggression, commit emotionally and/or had documented functional reduction due to psychosomatic symptoms. A high rating was given to conscripts who seemed able to act calmly in most kinds of situations. This variable has previously been associated with suicide in this cohort (Allebeck et al. Reference Allebeck, Allgulander and Fisher1988; Larsson et al. Reference Larsson, Hemmingsson, Allebeck and Lundberg2002). A low rating of ‘social maturity’ was given to conscripts who seemed dependent on others, irresponsible and/or showed signs of social maladjustment. A high rating was given to those who seemed willing to take responsibility and who showed signs of independence, dominance and extroversion. This variable has also been related to suicide risk in early adulthood (Allebeck et al. Reference Allebeck, Allgulander and Fisher1988).
Psychologists' ratings were regularly checked for inter-rater reliability (Lilieblad & Stahlberg, Reference Lilieblad and Stahlberg1977). When the personality variables were dichotomized for the purpose of examining their relationships with the exposure variable, IQ, and the outcomes, suicide and suicide attempt, ratings of 1 and 2 were treated as ‘low’, with ratings of 3–5 as reference. In other analyses they are used in their original categorical form of values 1–5.
All men completed questionnaires concerning information on social background, upbringing, behaviour and adjustment, health, and psychological and psychosomatic symptoms. The items regarding contact with police or child-care authorities and truancy were used as indicators of maladjustment and deviant behaviour, as in previous studies (Allebeck & Allgulander, Reference Allebeck and Allgulander1990b; Hemmingsson & Kriebel, Reference Hemmingsson and Kriebel2003). Self-reported symptoms at conscription have previously been associated with higher mortality in later life (Larsson et al. Reference Larsson, Hemmingsson, Allebeck and Lundberg2002). Self-reported medication for nervous problems was also included as an indicator of psychological distress. The questions and response alternatives are presented in Appendix 1.
Conditions in later adulthood
Information was obtained from the National Population and Housing Census of 1985 regarding socio-economic position, marital status and children in the household at 34–36 years of age. The classification into eight socio-economic groups was based on occupation: (1) unskilled workers; (2) skilled workers; (3) assistant non-manual employees; (4) non-manual employees at intermediate level; (5) non-manual employees at higher level; (6) farmers; (7) self-employed (mostly skilled workers or drivers); (8) those for whom no occupation was reported (e.g. unemployed, early retired or disabled). Marital status was dichotomized as married/cohabiting or unmarried/living alone. Whether the man was living in a household with at least one child under the age of 15 years, regardless of biological relationship, was also assessed as a potential covariate.
Statistical analysis
The associations of IQ test performance at conscription with suicide 1971–2006 and suicide attempt 1973–2006 were estimated using Cox proportional hazard models in SAS 9.2 (SAS Institute, Inc., USA). The proportionality assumption was assessed by examining the log–log survival plot. IQ was modelled as a continuous variable, with hazard ratios (HRs) given for a decrease of one (1) on the stanine scale. Dependent variables were suicide mortality and first suicide attempt. Subsequently, the effect of controlling for variables measured in childhood and at conscription was investigated. The analyses were repeated with follow-up from 1986, adding adulthood variables recorded in 1985 to the models. In the analysis of the association with suicide attempt 1986–2006, men with previous suicide attempts were excluded. We also performed analyses excluding men with a diagnosis of psychosis at conscription or in hospital records 1973–2006 (n = 762 in total), as men with psychoses are shown to differ regarding the association of cognitive ability with later suicidal behaviour (Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008; Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010; Webb et al. Reference Webb, Langstrom, Runeson, Lichtenstein and Fazel2011).
Results
Of the 49 321 men conscripted for military service in 1969–1970, there were 44 560 (90.3%) with information available on all the variables included in the final analyses for the full follow-up period. Men in the study sample were very similar to the full cohort with regard to various characteristics in childhood and early adulthood (data available upon request).
Table 1 shows the prevalence of potential risk factors in the nine categories of IQ. Most factors were graded across the stanine scale, with more favourable conditions typically being more common towards the higher end of the scale, and vice versa. Psychiatric diagnoses at conscription were more prevalent at lower levels of IQ, as too were low emotional control and low social maturity.
Table 1. Description of the cohort: men with information on all childhood and conscription variables (n = 44 560)
Data are given as percentage prevalence of risk factors.
IQ, Intelligence quotient; stanine, standard-nine; SEP, socio-economic position.
a Prevalence among men with information on adulthood as well as earlier variables (n = 43 479).
Table 2 shows the associations between the potential risk factors at conscription and suicide and suicide attempt. During the full follow-up periods, there were 553 deaths from suicide (1971–2006) and 1058 of the conscripts were admitted to hospital following a suicide attempt (1973–2006). Almost all the variables were significantly associated with both outcomes, but generally more strongly so with suicide attempts than with completed suicide. Psychiatric illness, personality dimensions and deviant behaviour measured in early adulthood, and family formation in later adulthood, were among the factors with the strongest associations. Among the socio-economic position categories, having no reported occupation was the strongest risk factor.
Table 2. Associations between risk factors and suicide 1971–2006 and suicide attempt 1973–2006
Data are given as crude hazard ratio (95% confidence interval).
SEP, Socio-economic position.
a Excluding men with a previous suicide attempt.
Table 3 shows the associations between IQ at conscription and subsequent suicide and suicide attempt. IQ was inversely associated with both outcomes. The crude HR for completed suicide was 1.20 [95% confidence interval (CI) 1.15–1.24] per one-unit decrease in IQ score, while the corresponding risk for attempted suicide was greater, with a crude HR of 1.28 (95% CI 1.25–1.32). The increased risks were graded across the stanine scale for both outcomes (Fig. 1). Controlling for all variables collected in childhood and at conscription, in a multivariable model, attenuated the associations by about 40% for both outcomes. Among the possible confounding variables, the personality variable ‘social maturity’ had the greatest single impact on the associations. The personality variable ‘emotional control’ also had a fairly substantial impact on the associations, as did psychiatric diagnosis and self-reported contact with police or child-care authorities. The effects of controlling for other factors were generally marginal.
Fig. 1. Risk for suicidal behaviour across the standard-nine (stanine) scale. Hazard ratios (HRs) are plotted against intelligence quotient (IQ) scores for suicide 1971–2006 and suicide attempt 1973–2006. The highest IQ score (9) is the reference unit (HR = 1). The full model adjusts for socio-economic position and crowded housing in childhood; and height, psychiatric diagnosis, personality evaluation (emotional control and social maturity) and self-reported behavioural factors and psychological symptoms at conscription.
Table 3. Associations between IQ at conscription 1969–1970 and suicide 1971–2006 and suicide attempt 1973–2006
Data are given as crude or adjusted hazard ratio (95% confidence interval) per one-point decrease in IQ score on a stanine scale.
IQ, Intelligence quotient; SEP, socio-economic position; stanine, standard-nine.
Excluding men with a diagnosis of psychosis resulted in no difference with regard to the association of IQ with attempted suicide (HR 1.29, 95% CI 1.25–1.34), and a slightly weaker association with completed suicide (HR 1.18, 95% CI 1.13–1.23), compared with the original models.
The associations between IQ score at conscription and suicide and suicide attempt with follow-up starting in 1986 are shown in Table 4. The associations were similar in magnitude to those for the full follow-up period, although slightly weaker for suicide attempt. Controlling for all the childhood and conscription factors also had an impact on these associations, and controlling for all the adulthood variables had an impact of similar magnitude. In models controlling for all the childhood, conscription and adulthood factors, the associations were attenuated by about 45%.
Table 4. Associations between IQ at conscription 1969–1970 and suicide and suicide attempt 1986–2006
Data are given as crude or adjusted hazard ratio (95% confidence interval) per one-point decrease in IQ score on a stanine scale.
IQ, Intelligence quotient; SEP, socio-economic position; stanine, standard-nine.
a Excluding men with a previous suicide attempt.
Discussion
In this register-based, longitudinal study, we found cognitive ability in early adulthood to be associated with suicide and suicide attempt during 36 years of follow-up. The associations were partly attenuated when controlling for covariates in childhood and in early and later adulthood.
Previous studies of the associations of cognitive ability with subsequent suicide (Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008; Osler et al. Reference Osler, Nybo Andersen and Nordentoft2008) and suicide attempt (Fergusson et al. Reference Fergusson, Horwood and Ridder2005; Osler et al. Reference Osler, Nybo Andersen and Nordentoft2008; Alati et al. Reference Alati, Gunnell, Najman, Williams and Lawlor2009) have found crude associations similar to ours among male and mixed populations. Gradients across the full IQ scale for suicide (Gunnell et al. Reference Gunnell, Magnusson and Rasmussen2005; Hemmingsson et al. Reference Hemmingsson, Melin, Allebeck and Lundberg2006; Gravseth et al. Reference Gravseth, Mehlum, Bjerkedal and Kristensen2009) and for suicide attempt (Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010), corresponding to those indicated by our findings, have been presented in previous studies of Scandinavian conscripts.
The effect of personality on the associations between cognitive ability and suicidal behaviour has to our knowledge not been previously investigated. Much like intelligence, personality traits have been found to be stable over time (Deary et al. Reference Deary, Weiss and Batty2010), and findings from twin studies indicate that intelligence and personality aspects share common genetic bases (Luciano et al. Reference Luciano, Wainwright, Wright and Martin2006; Bratko et al. Reference Bratko, Butkovic, Vukasovic, Chamorro-Premuzic and von Stumm2012). Personality-related behaviours observed at the age of 3 years have been shown to be predictive of suicide attempt and other psychiatric outcomes in early adulthood (Caspi et al. Reference Caspi, Moffitt, Newman and Silva1996). Intelligence and personality have previously been related to mortality, and found to be about as important as traditional risk factors including smoking, high blood pressure (Batty et al. Reference Batty, Shipley, Dundas, Macintyre, Der, Mortensen and Deary2009) and socio-economic position (Roberts et al. Reference Roberts, Kuncel, Shiner, Caspi and Goldberg2007). The effect of controlling for personality on the association between IQ and total mortality has been studied previously. Among men in the Vietnam Experience Study cohort, an interaction effect was detected between intelligence and neuroticism (a trait comparable with low emotional control), such that men low in intelligence and high in neuroticism were at particularly high risk of early death (Weiss et al. Reference Weiss, Gale, Batty and Deary2009). Further, among Scottish men and women, the association between cognitive ability in childhood and later mortality was attenuated but remained significant after controlling for dependability, a variable composed of evaluations of perseverance, stability of moods and conscientiousness (Deary et al. Reference Deary, Batty, Pattie and Gale2008). However, as mentioned, the mechanisms through which cognitive ability and personality aspects are related are unclear and assumedly complex (Ackerman & Heggestad, Reference Ackerman and Heggestad1997; Chamorro-Premuzic & Furnham, Reference Chamorro-Premuzic and Furnham2004).
The role of mental illness in the association between cognitive ability and future suicide and suicide attempt has previously been investigated. Among Danish men born in 1953 and followed up to 50 years of age, controlling for mental disorder at conscription substantially attenuated the associations (Osler et al. Reference Osler, Nybo Andersen and Nordentoft2008). In our study, psychiatric diagnosis at conscription had a somewhat lesser attenuating effect on IQ's association with subsequent suicide and, in particular, suicide attempt, perhaps because of classification differences. Only 4.1% of the men were coded as mentally disordered in the Danish cohort, compared with the approximate 12% who had received a psychiatric diagnosis at conscription in our study.
Among the men and women in the 1958 British Birth Cohort, a global measure of psychosomatic symptoms in early adulthood attenuated the association between childhood IQ and mortality up to the age of 46 years (Jokela et al. Reference Jokela, Batty, Deary, Gale and Kivimaki2009). In our study, however, psychological symptoms contributed only marginally to explaining the associations between IQ and suicide and suicide attempt.
Self-reported contact with the police or child-care authorities had some attenuating impact on the associations, but truancy had little effect. As mentioned, these variables were used as indicators of maladjustment. In comparison, among young New Zealand men and women, controlling for parents' and teachers' ratings of conduct problems in childhood had a substantial impact on the association between childhood IQ and history of suicide attempts at age 25 years (Fergusson et al. Reference Fergusson, Horwood and Ridder2005). However, retrospective self-reports of childhood externalizing behaviour had no impact on the association between IQ at age 14 years and history of suicide attempts at age 21 years among men and women in an Australian birth cohort (Alati et al. Reference Alati, Gunnell, Najman, Williams and Lawlor2009).
Some potential mechanisms underlying the association between cognitive ability and suicidal behaviour have previously been suggested. A common cause, such as neurodevelopmental problems or adversities in childhood, might hamper cognitive development and also increase psychological vulnerability (Gunnell et al. Reference Gunnell, Magnusson and Rasmussen2005; Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008; Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010). This vulnerability could be manifested in psychiatric illness, psychological symptoms, behaviour or personality.
Further, it has been suggested that the association might be mediated by attained socio-economic position (Gunnell et al. Reference Gunnell, Magnusson and Rasmussen2005; Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008; Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010), which indeed has been found to explain some of the association between cognitive ability and suicide in this cohort (Hemmingsson et al. Reference Hemmingsson, Melin, Allebeck and Lundberg2006). In keeping with this, we found that controlling for socio-economic position at age 35 years attenuated the association between cognitive ability and suicide attempt. The effect of family formation (marriage/parenthood) on the association between IQ and suicidal behaviour was limited in the present study, and we are unaware of any previous epidemiological study that has investigated this possible pathway.
Nevertheless, although several covariates and risk factors were included in our analyses, substantial associations remained. Although there is a possibility of residual confounding, it does seem as if cognitive ability has an effect on attempted and completed suicide over and above socio-economic, psychosocial, behavioural and personality factors. One possible mechanism would be the effect of cognitive ability in itself. Problem-solving ability and finding coping strategies to handle emotional distress might increase resilience to stressful events and prevent suicidal acts from appearing as the only solution (Gunnell et al. Reference Gunnell, Magnusson and Rasmussen2005; Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008). Case–control studies showing lower executive functioning among depressed patient with high-lethality suicide attempts compared with other depressed patients (Keilp et al. Reference Keilp, Gorlyn, Oquendo, Burke and Mann2008) and poorer decision-making in people who have attempted suicide, compared with people with a similar affective profile but without a history of suicide attempts (Jollant et al. Reference Jollant, Lawrence, Olie, O'Daly, Malafosse, Courtet and Phillips2010), might indicate a specific role of these cognitive functions. However, because of the retrospective study designs it cannot be ruled out that the suicide attempts could have affected neuropsychological functions, i.e. that reversed causality underlies the relationship. It has also been suggested that locus of control might mediate the association (Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010), as people with higher cognitive ability tend to have a more internal locus of control (Gale et al. Reference Gale, Batty and Deary2008) which in turn is related to lower risk for suicidal behaviour (Pearce & Martin, Reference Pearce and Martin1993). Further, cognitive ability might affect mental health and the risk of suicidal behaviour via its connection to health behaviours (Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010).
In accordance with previous findings (Andersson et al. Reference Andersson, Allebeck, Gustafsson and Gunnell2008; Batty et al. Reference Batty, Shipley, Dundas, Macintyre, Der, Mortensen and Deary2009), excluding men with psychoses from the analyses did not have any substantial impact on the associations.
Strengths and weaknesses
Our study population consists of men examined at mandatory conscription and is highly representative of all Swedish men born around 1950. Through record linkage, using unique personal identification numbers, conscription information could be connected to reliable data on causes of death and hospital records on attempted suicide. Information obtained at conscription included professionally assessed psychiatric diagnoses and personality evaluations, and also self-reports of various psychological symptoms. Cognitive ability was assessed by a standardized test battery in early adulthood.
Two previous studies on the associations between cognitive ability and later attempted and completed suicide have comprised Swedish conscripts 1969–1994 (Gunnell et al. Reference Gunnell, Magnusson and Rasmussen2005; Batty et al. Reference Batty, Whitley, Deary, Gale, Tynelius and Rasmussen2010), including the men in the present study. The unique characteristics of the present study is that the conscription survey from 1969–1970 contains self-reported data on behaviour and psychosocial and psychological factors, as well as evaluation from psychological assessment on emotional control and social maturity, and that we were able to examine the role of social conditions in adulthood.
However, the fact that our findings are based solely on men born around 1950 is a limitation. In a Swedish cohort comprising both women and men, Andersson et al. (Reference Andersson, Allebeck, Gustafsson and Gunnell2008) found a lower risk for suicide with increasing cognitive ability only among men, not women. However, in the New Zealand birth cohort study, tests for gender differences showed no interaction of gender in the association of cognitive ability with attempted suicide (Fergusson et al. Reference Fergusson, Horwood and Ridder2005).
We have no information on suicide attempts before 1973 and the earlier cases are thus missing. Since information on suicide attempt is obtained from hospital records rather than by self-report, the attempts that did not require medical care are probably missed in this study. However, the use of hospital records does give a valid estimate of number of men admitted for suicide attempts.
Test anxiety is a state negatively correlated with intelligence test performance and positively correlated with trait-related stress reactions (Ackerman & Heggestad, Reference Ackerman and Heggestad1997). In our study, emotional control was associated with IQ. If low emotional control is a marker for test anxiety, the association between IQ and emotional control (shown in Table 1) might be aggregated. The attenuating effect of controlling for emotional control in the associations between IQ and suicidal behaviour would thus be partly accounted for by test anxiety. However, the mechanisms behind the association between test anxiety and intelligence test performance are debatable, and it has been suggested that lower cognitive ability might instead determine test anxiety (Reeve & Bonaccio, Reference Reeve and Bonaccio2008).
This present study confirms previous findings of associations between cognitive ability and subsequent suicide and suicide attempt. In this study we also, unlike other studies, had the possibility to control for several psychological, psychiatric and social covariates. It adds to previous research in showing that after controlling for these relevant risk factors, the association between cognitive ability and suicidal behaviour remained.
Conclusions
In this longitudinal study of a male general population cohort, cognitive ability in young adulthood was found to be associated with subsequent completed and attempted suicide, following a dose–response pattern. The associations were attenuated but remained statistically significant in models controlling for a range of possible confounding and mediating factors.
Appendix 1. Self-reported factors from conscription questionnaires
Acknowledgements
This study was supported by the Swedish Council for Working Life and Social Research (project no. 2008-0907).
Declaration of Interest
None.
