Article Text
Abstract
Background Suicide is a leading cause of death in young adults. Several risk factors are well known, especially those related to adult mental health. However, some risk factors may have their origin in the very beginning of life. This study examines suicide in the general Norwegian population in a life course perspective, with a main focus on early life factors.
Methods In this study, several national registers were linked, supplying personal data on biological and social variables from childhood to young adult age. Participants were all Norwegians live born during the period 1967–1976, followed up through 2004. Persons who died or emigrated before the year of their 19th birthday, at which age follow-up started, were excluded. Thus, the study population comprised 610 359 persons, and the study outcome was completed suicide.
Results 1406 suicides (0.23%) were recorded, the risk being four times higher in men than in women. Suicide risk factors included not being firstborn (adjusted HR in men and women (95% CIs): 1.19 (1.05 to 1.36) and 1.42 (1.08 to 1.88)), instability of maternal marital status during childhood, parental suicide (mainly in women), low body mass index (only investigated in men), low education and indications of severe mental illness.
Conclusions Suicide in young adults may be rooted in early childhood, and the effect is likely to act through several mechanisms, some of which may be linked to the composition and stability of the parental home. A life course perspective may add to our understanding of suicide.
- Conscription data
- education
- life course epidemiology
- Norway
- suicide
- education FQ
- suicide SI
Statistics from Altmetric.com
Among the top three causes of death in young adults in most countries, suicide is estimated to represent 1.8% of the total global burden of disease according to figures from the World Health Organization,1 with suicide being up to four times more prevalent in young men than in young women.
Among the wide range of risk factors identified for suicide, mental disorders seem to be the most prevalent risk factor in adults2 as well as in young adults.3 That disorders such as schizophrenia,4 affective disorders,5 substance abuse disorders6 and personality disorders7 8 imply a strongly increased suicide risk seems well established. There is also evidence that factors indicating instability of the parental home, such as parental divorce,9 parental suicide9 10 and parental unemployment or low income,9 may contribute to increasing risk. An inverse effect of parental educational level has been reported,9 but this effect may be complex since interactions with the study persons' own educational level11 or skills12 may exist. The effect of study subjects' own educational level has been found to be inverse,9 but not necessarily significant.13 The literature on parity effect on suicide risk is sparse, but one recent study has found higher maternal parity to be associated with higher suicide risk of offspring in their young adulthood.14 Increasing number of stressful life events has been found to be associated with increased probability of suicide ideation/attempt, especially in genetically disposed persons.15 Other risk factors that have been identified are low birth weight,16 low intelligence test scores12 13 and low body mass index (BMI).13 17 18
This wide and heterogeneous range of risk factors acts at different stages of life, but few epidemiological studies have been able to comprehensively study suicide risk factors in the general population. In the present study, we use prospectively collected data from a large birth cohort of all Norwegians born between 1967 and 1976 to explore risk factors for suicide in a life course perspective, taking exposures acting at different life periods into account. Our main focus will be on parental and other early life background factors, and our data allow consideration for other well-known risk factors, especially those related to mental health status.
Methods
Participants and linking procedures
The study population comprised a total of 610 359 index persons (312 358 men and 298 001 women) born in Norway in 1967–1976, as registered by the Medical Birth Registry of Norway, who were alive and had not emigrated by 1 January in the year of their 19th birthday. The original birth cohort comprised 626 928 persons; 4544 persons were excluded because of emigration before the age of 19 years, and 12 025 persons were excluded because of the occurrence of death before this age. Two hundred and six of these deaths were attributed to suicide (156 men and 50 women). The Medical Birth Registry contains the 11-digit national identification number of index persons and their parents, allowing linkage with the Central Population Register, the Education Register and the Cause of Death Register of Statistics Norway, the benefit registries in the Norwegian Labour and Welfare Organisation and with the Norwegian Armed Forces Personnel Data Base (men only). This linkage provided longitudinal data for index persons and their parents, with annual updates through 2004 (Cause of Death Register) or 2003 (other registers).
Outcome variable
The outcome variable was completed suicide, as registered in the Norwegian Cause of Death Register in the period 1986–2004. Suicides in the period 1986–1995 were coded according to International Classification of Diseases ninth revision (ICD9) and, starting from 1996, according to ICD10. The following codes were defined as suicides: 950–959 (ICD9) and X60–X84 (ICD10). Method of suicide was also recorded.
Independent variables
The different registers provided data on several covariates, occurring throughout the life span. The Medical Birth Registry provided data on birth weight, birth order, birth year, mother's marital status at birth and whether the father's identity was known. Birth weight was standardised for sex and birth order into z scores, where birth weight at the mean was assigned a zero score and the SD was 1. From the Norwegian Labour and Welfare Organisation, we had data on index persons' benefits due to chronic childhood disease19 and data on disability pension for the parents (occurring at index persons' age 0–17 years) and the index persons. The Central Population Register provided date of death or emigration for the index persons and category of industrial composition and urbanity in the municipality of residence of the index persons at the age of 16; the latter variable was divided into seven categories, according to the standard of Statistics Norway. The Central Population Register also provided data on maternal marital status in the year of the index persons' 18th birthday. This information was combined with maternal marital status at birth into the variable “maternal marital status according to index persons' age”, with six categories such as married both at birth and at age 18, unmarried both at birth and at age 18, not married at birth/married at age 18, divorced at age 18, widowed at age 18 and dead at age 18 (the last three categories irrespective of status at birth). There was no data on ethnicity. However, according to the Central Population Register, 99.4% of the study population had a Norwegian citizenship; because of this uniformity, this variable was not adjusted for. The Cause of Death Register provided data also on parental suicides, according to the same ICD codes as for index persons' suicides. From the Education Register of Statistics Norway, we had information on educational level for the index persons as well as for both parents, based on the Norwegian standard classification of education NUS2000.20 In the main analyses, the original eight levels were dichotomised into whether upper secondary education was completed at age 19. Parental education refers to the parent with the highest educational level, in the year of the index persons' 16th birthday.
Finally, the Norwegian Armed Forces Personnel Data Base provided results from conscription tests (men only). National military service is compulsory to all Norwegian men, taking place at age 18 or 19 years. At conscription, men are tested for general intellectual performance, and height and weight are measured. Furthermore, as part of the medical examination, an evaluation of mental health is conducted to sort out whether the conscripts are suited for military service. For the purpose of the present study, men with any mental health impairment were categorised in one group and compared with mentally healthy men. The test of intellectual performance that is used in Norway is highly correlated with the Wechsler Adult Intelligence Scale, with a correlation of 0.73 in a small sample.21 Intellectual performance is recorded in a stanine scale as a single digit ranging from 1 (lowest) to 9, the scores being normally distributed in the general population with mean=5 and SD=2.21 From the height and weight, we computed BMI and divided it into four groups: <18.50 (underweight), 18.50–24.99 (normal weight), 25.00–29.99 (overweight) and ≥30.00 (obesity).
Important factors known to be associated with suicide, such as family situation (marital status and having own children) and employment status, were not included because of the design and age profile in the study. Follow-up started at age 19 years, which is before most people establish family and working life.
Statistical analysis
Stata/SE V.9.2 software was used in all analyses. Cox' proportional hazards models were used, with start of follow-up on 1 January in the year of the study participants' 19th birthday. It was assumed that age at this time was, on average, 18½ years. Individuals who died (from other causes than suicide) or emigrated during the observation period, which lasted until 31 December 2004, were not included in the analysis. It was assumed that suicides and censoring on average took place in the middle of the year. Suicides thus took place in the age span 19–37 years, and the follow-up time varied from 0.5 to 19 years. The proportional hazard assumption was assessed comparing log–log survival curves. We computed crude and adjusted hazard ratios (HRs) and the corresponding 95% confidence interval (CIs) for the study variables. The adjusted HRs were mutually adjusted for all other variables. Generally, the value expected to be associated with the most favourable outcome was selected as the reference category. The reference value for intellectual performance was a score at the mean (ie, score=5).
Suicide occurrence in groups was also computed as risks (cumulative incidence) and rates (suicides per 100 000 person years). Throughout, missing covariate values were included in the models as separate categories, and all analyses were made separately for male and female subjects.
Results
During the observation period, a total of 1406 suicides (0.23%) (1143 men and 263 women) were recorded. Three thousand and five hundred thirty-six men and 1243 women died from other causes than suicide, and 7905 men and 8950 women emigrated; thus, altogether, 21 634 persons (3.5%) were censored because of emigration or non-suicide death during the observation period. Total observation time was 8 816 603 person years, with a sex distribution of 51% vs 49% for men and women, respectively. The suicide risk for men was about four times the risk for women: 0.37% vs 0.09%. Figure 1 shows suicide rates by age and sex.
Table 1 shows suicides by sex and methods. Hanging was the most prevalent method for both sexes but, for men, only slightly more frequent than the use of firearms, which was rare for women. Overdose was the method of suicide in a considerably higher proportion of suicides in women than in men.
Tables 2 and 3 show risks and HRs for suicide according to the different study variables for men and women, respectively. For both sexes, there was a parity effect. Not being firstborn was associated with increased suicide risk. With parity as a dichotomous variable, adjusted suicide HR for not first-born persons (95% CI) was 1.19 for men (1.05 to 1.36) and 1.42 for women (1.08 to 1.88). Mother being divorced or widowed when the index person was 18 years was also associated with increased suicide risk, whereas mother being dead increased the risk for men only. If, however, mother had been unmarried both at the time of birth of the index person and when the index person was 18 years or if mother had married during the index person's childhood, no increased risk of suicide was observed. Parental disability and father's identity being unknown were only weak and borderline significant risk factors, the tendency being stronger in women than in men. Parental suicide strongly increased the risk for women but had only borderline significant impact for men.
There was a substantially increased suicide risk for men with impaired mental health at conscription and a clear inverse effect of BMI (table 2). Intellectual performance measured at the time of conscription had a modest effect that vanished in the adjusted model. However, subjects who scored 1 or 2 on the intellectual performance test (corresponding to an IQ ∼70–80) still had a 40% increased risk of suicide, compared with essentially all other scores. Having been granted a disability pension increased the suicide risk considerably for both sexes (tables 2 and 3); however, for men, this was only the case for disability pension with a diagnosis of schizophrenia.
For both sexes, there was a strong and inverse association between educational level and the risk of suicide, that is higher suicide risk for persons with low education (tables 2 and 3). Concerning parental education, there was a sex difference; there was essentially no effect of parental education in men, whereas there was a significantly positive association in women. There also was an interaction between parental and own educational level, with particularly high suicide mortality among study persons with low education and highly educated parents.
There was no clear effect of birth weight or childhood disease benefit; this was the case for both sexes. Municipal category of residence at age 16 had no effect on suicide risk.
Discussion
The study has revealed several risk factors for suicide, including not being firstborn, instability of maternal marital status during childhood, parental suicide, low BMI, low education and indications of mental illness.
The main strengths of this study are the size of the study population and the >1400 suicides, the high quality of the register data, the prospective design and the almost complete follow-up. It is important to note that the use of register-based data, however rich they may be, will imply a lack of several interesting variables. Our information on mental health, for instance, only includes severe conditions, leading to disability or to being disqualified from military services, whereas information on minor psychiatric disorders, being much more prevalent in the population, is lacking. We have no data on specific personal characteristics, and the important conscription data are missing for women. Our data on adverse life events are only those related to conditions affecting parental stability, whereas we have no information on the index persons' exposure to violence or abuse and other stressful life events. Although our data on the stability of maternal marital status are richer than in previously published studies, they are still limited to the index persons' birth and the year of their 18th birthday. We have no data on what possible changes mothers may have made with respect to marital status in the time span between these points in time.
Suicide risk seems to have many origins, of which some may be operative from the very beginning of life. The effect of birth order found in our cohort is probably related to such early life environmental factors. Only in the recent Scottish study by Riordan et al14 has this so far been reported. They speculated that the effect could be mediated through a possibly higher prevalence of mental illness in subjects who were not firstborn, since their mothers could have been under higher stress during the pregnancy and first years of their child's life. In contrast to the data set in the study of Riordan et al, our data allowed us to study this possible explanation more closely. As we have shown, the birth order effect remained significant even after adjusting for impaired mental health. We also did the adjusted model, excluding the mental health variables, and these analyses showed that the birth order effect was completely independent of the mental health data. To reveal whether the birth order impact in reality was an effect of family size, we also made analyses separating only children from other firstborns. This did not show any significant differences between these two groups. Since neither major mental impairment nor adjustment for any other available data, such as social background, could explain the association between birth order and suicide risk, we should rather look for mediating mechanisms among personal characteristics such as coping and resilience. It is possible that children who have enjoyed their mother's undivided attention during their first year(s) of life could somewhat more easily develop resilience against suicidal reactions to stressful conditions later in life than other children. Previous studies have found high scores on measures for personal characteristics such as self-esteem,22 23 social competence24 and coping capacity25 to be protective of suicidal ideation and behaviour. According to a meta-analysis of published research by Sulloway,26 there is evidence of a birth order effect on important aspects of personality development along the lines we have indicated.
We found no association, nor any trend, with birth weight. A birth weight–suicide association has been found in only one study.16 The unadjusted results, however, suggested a weak and non-significant increased suicide risk for men with birth weight below the mean. Any actual birth weight–suicide association could have been mediated through some of our adjustment factors, like intellectual performance. Since suicide is such a rare outcome, further evidence of an association between birth weight and suicide needs further and even larger studies.
Results of the maternal marital status variable confirmed what other researchers have found earlier with data from a single point of time, that having a single/divorced mother increases suicide risk.9 10 With our longitudinal data, we can add to this picture that what might be essential is not whether the mother is single or married but her stability with respect to marital status. There was no data on cohabitation. This was less common in Norway during the years of interest than today, but increasing throughout the period. Some of the mothers who were classified as unmarried at both points in time could in fact be cohabiting during the whole period and possibly with the same partner. In this case, the interpretation would be that it is a stable family with two parents that is protective against offspring suicide.
There was a striking sex difference in the effect of parental suicide, women being much more vulnerable to this, in relative terms. This sex difference has also been found earlier,10 although not as pronounced as in our study. A proposed explanation to this has been sex differences in reactions against bereavement.27
The strong inverse association between educational level and risk of suicide confirms that there are social inequalities in suicide risk.9 11 Concerning the risk associated with social class of origin, contradictory results have been reported.9 11 We found essentially no effect in men and a positive association in women. This is similar to a recent Norwegian study,11 which found this association in part to work through adult family status.
The inverse association between BMI and suicide in men has been found in several previous studies13 17 18 28 and seems well established. Concerning the relation between intellectual performance and suicide risk, our crude HRs were very similar to the age-adjusted results in a recent Swedish study.12 However, little happened after adjustment in that study, whereas the effect more or less disappeared in our study, except in subjects who scored 1 or 2 on the intellectual performance test. A likely explanation to this is that we, contrary to the mentioned study, also have adjusted for important risk factors as mental health and maternal marital status. Another aspect of this is that one might question whether it is correct to adjust for educational level, which perhaps is more likely a mediator rather than a confounder in the relation between intellectual performance and suicide. We reran the analyses without adjusting for educational level, and the results indicated that about 50% of the effect of intellectual performance is mediated through educational level.
For all the conscript variables, some of the strongest associations were among those with missing data. Men could have missing conscript data by two reasons: they either did not meet at conscript, or, if they met, they did not carry out some of the tests. In both cases, these are men who are not suited for military service. It is a highly selected and deprived group, and it includes men with severe handicaps or congenital disorders. Men with serious mental disorders will also be over-represented.
Our study increases the understanding of suicides in young adults. It emphasises the influence of early life conditions on adult health, and it particularly sheds light on circumstances concerning the composition of the family of origin. The associations found could also have elements of selection into poor adult circumstances. Lundberg29 introduced the concept “unhealthy life career”, which implicated an accumulated effect of risk factors through different stages of early life on adult health. Persons exposed to adverse conditions during childhood are often exposed to adverse conditions as adults. Future research should, if possible, include information on minor mental health problems and incorporate information on more types of adverse life events and on personal characteristics, in order to further explore the mechanisms behind our findings. It is, however, important to remember that suicide is a rare outcome and difficult to study in data sets collected for research purposes because of statistical power concerns. Register data, with their inherent weaknesses, will thus remain an important basis for comprehensive studies of the interplay of many risk factors.
Our findings that early life factors, such as birth order or maternal marital stability, may be associated with suicide risk in young adulthood underline the importance of targeting families with small children for preventive and supportive measures. Prevention of suicide in young adults will, at least in part, probably depend on effective preventive measures instigated at earlier stages in life.
References
Footnotes
Funding The study was supported financially by grants from the Norwegian Research Council (project no. 161321/V50) and from The Directorate of Labour Inspection. The authors' work was independent of the funders.
Competing interests None.
Ethics approval The Regional Committee for Medical Research Ethics (Department Southern Norway, Oslo) has approved the study (ref. no. S-06028).
Provenance and peer review Not commissioned; externally peer reviewed.