IRR indicates incidence rate ratio, adjusted for sex of offspring, calendar time at parental death, age at time of parental death, and country.
aData from Denmark, Sweden, and Finland.
bData from Denmark and Sweden.
cData from Denmark.
Owing to data restrictions, the time points have been collapsed so that the underlying number of suicides exceeds 5 in each of the plotted time points.
Guldin M, Li J, Pedersen HS, Obel C, Agerbo E, Gissler M, Cnattingius S, Olsen J, Vestergaard M. Incidence of Suicide Among Persons Who Had a Parent Who Died During Their ChildhoodA Population-Based Cohort Study. JAMA Psychiatry. 2015;72(12):1227-1234. doi:10.1001/jamapsychiatry.2015.2094
Parental death from suicide is associated with increased risk of suicide in the bereaved child, but little is known about the long-term risks of suicide after parental death from other causes. A better understanding of this association may improve suicide prevention efforts.
To examine the long-term risks of suicide after parental death and how the risk trajectories differed by cause of parental death while accounting for major potential confounding variables.
Design, Setting, and Participants
A population-based matched cohort study was performed using information from nationwide registers (data from 1968 to 2008) in 3 Scandinavian countries (for a total of 7 302 033 persons). We identified 189 094 children (2.6%) who had a parent who died before the child reached 18 years of age (ie, the bereaved cohort). Each bereaved child was matched by sex and age to 10 children who did not have a parent who died before they reached 18 years of age (for a total of 1 890 940 children) (ie, the reference cohort). Both cohorts were followed for up to 40 years. Poisson regression was used to calculate the incidence rate ratio (IRR), while accounting for age at parental death, sex, time since bereavement, maternal/paternal death, birth order, family history of psychiatric illness, and socioeconomic status. Data analyses were finalized June 24, 2015.
The main exposure was death of a parent within the first 18 years of life.
Main Outcomes and Measures
Incidence of suicide among persons who had a parent who died during their childhood.
During follow-up, 265 bereaved persons (0.14%) and 1342 nonbereaved persons (0.07%) died of suicide (IRR = 2.02 [95% CI, 1.75-2.34]); IRR = 3.44 (95% CI, 2.61-4.52) for children who had a parent who died of suicide, and IRR = 1.76 (95% CI, 1.49-2.09) for children who had a parent who died of other causes. The IRR tended to be higher for children who had a parent who died before they reached 6 years of age, and the IRR remained high for at least 25 years. During 25 years of follow-up, the absolute risk of suicide was 4 in 1000 persons for boys who experienced parental death and 2 in 1000 persons for girls who experienced parental death.
Conclusions and Relevance
Parental death in childhood is, irrespective of cause, associated with an increased long-term risk of suicide. The consequences of parental death in childhood are far-reaching, and suicide risk trajectories may be influenced by early-life conditions. Future public health efforts should consider helping highly distressed children to cope with bereavement.
The death of a parent is experienced by 3% to 4% of children in Western societies and is one of the most stressful and potentially harmful life events in childhood.1- 4 Most children and adolescents adapt to the loss, but some develop preventable social and psychological problems.5,6 These bereaved children have been shown to have an increased long-term risk of developing mental health problems and committing suicide,6- 10 but little is known about the long-term risk of suicide among children who had a parent who died from other causes.1,8,9,11,12 Few studies have had sufficient size and follow-up time to investigate the long-term risk of suicide among children with additional vulnerabilities, yet studies point to the following factors impacting suicide risk: low socioeconomic status, family history of psychiatric illness,9,12- 15 young age at the time of parental death,9,10,16,17 and maternal suicide.10,18,19
We hypothesized that accidental and suicidal parental death and shared genes increase the suicide risk among the offspring.10,20 In a large population-based cohort study, we aimed to investigate how parental death may influence the long-term risk of suicide among persons who had a parent who died during their childhood and how the risk trajectories may differ by cause of parental death, age of child at time of parental death, sex, time since bereavement, birth order, socioeconomic status, and parental psychiatric history.
A population-based cohort was established by linking data from national registers from Denmark, Finland, and Sweden. The Nordic Perinatal Bereavement Cohort21 consists of all persons born in Denmark during the period from 1968 to 2008 (n = 2 809 393), all persons born in Sweden during the period from 1973 to 2006 (n = 3 400 212), and, owing to data protection rules, a random sample of 89.3% of persons born in Finland during the period from 1987 to 2007 (n = 1 136 409). The unique personal identification number used in the Nordic countries permits linkage of individual-level data between different registers.21,22
We excluded children who died within the first 6 months of life because mortality in this age group is mainly due to perinatal and congenital conditions.23 Offspring who died within 30 days of the parental death were also excluded because the majority of these offspring deaths were due to the same external causes, such as accidents or natural catastrophes.
A matched cohort was established. Each child in the cohort of 189 094 children who had a parent who died before the child reached 18 years of age (ie, the bereaved cohort) was matched to 10 children who did not have a parent who died before they reached 18 years of age (for a total of 1 890 940 children) (ie, the reference cohort) based on the offspring’s age at the time of parental death (±28 days), sex, and country of residence. The matching algorithm was applied without replacements (ie, each person served as reference only once). If a nonbereaved became bereaved, this person changed from reference to bereaved participant. Follow-up time varied according to the study entry time, which was defined as the date of death of the parent of the child in the bereaved cohort and the matched cases in the reference cohort. Participants were observed from entry time until death, emigration, or end of study period (December 31, 2009, in Denmark; December 31, 2008, in Sweden; and December 31, 2010, in Finland), whichever came first.
The main outcome was suicide among children who had a parent who died during their childhood. We studied whether the incidence rate ratios (IRRs) varied by specific characteristics shown to be risk factors for suicide: sex, child’s age at time of parental death, time since parental death, maternal or paternal death, parity, family history of psychiatric illness, socioeconomic status, and parental education level.6,7,13,16,18,24,25
Data on cause of death classified according to the International Classification of Diseases (ICD) were obtained from national registers of causes of death.26 In Denmark, the ICD-8 was used during the period from 1978 to 1993, and the ICD-10 was used during the period from 1994 to 2007. In Sweden, the ICD-8 was used during the period from 1973 to 1986, the ICD-9 was used during the period from 1987 to 1996, and the ICD-10 was used during the period from 1997 to 2008. In Finland, the ICD-9 was used during the period from 1987 to 1995, and the ICD-10 was used during the period from 1996 to 2010.21 Suicides among offspring were categorized by ICD codes (ICD-8 and ICD-9 codes E950-E959 and ICD-10 codes X60-X84). Parental mortality was classified into death from suicide (same ICD codes as suicide in offspring), accidents (ICD-8 and ICD-9 codes E80-E99 and ICD-10 codes V01-Y98), and other causes.
Potential confounders or effect modifiers included in the analyses were sex of offspring (male or female), sex of deceased parent (male or female), time since bereavement (0-4, 5-9, 10-14, 15-19, 20-24, or ≥25 years), calendar time at parental death (before April 4, 1995 or April 4, 1995, or later), offspring age at time of parental death (0-5, 6-11, or 12-18 years), birth order (1, 2, or ≥3), age of surviving parent at bereavement (<35, 35-44, or >45 years), country, parental socioeconomic status (not in labor market, unskilled worker, skilled worker, or white collar worker), parental education level (≤9, 10-11, or ≥12 years), and psychiatric illnesses of parents. Psychiatric illnesses of the children were considered a confounder but were not included in the analysis because too few cases were available.
Data on birth order were retrieved from the national medical birth registers,21 which contain information on all births in the study countries.27 Data on parental education and socioeconomic status were obtained from the Swedish Register of Education28 and the Danish Integrated Database for Labor Market Research.21 Data on education and socioeconomic status were available annually for Denmark for the period from 1980 to 2008 and for Sweden for 1980, 1985, and 1990, whereas these data were not available for Finland. The data used for the analysis were baseline characteristics of the children at birth year. Data on psychiatric diagnoses for parents and children were obtained from the national patient registers in Denmark and Sweden, which record psychiatric diagnoses for patients seen by a psychiatrist.29 Records on psychiatric diagnoses were not applied to the cohort from Finland. Psychiatric diagnoses were identified by the following ICD codes: affective disorders (ICD-8 codes 296.09-296.99, 298.09, 300.49, and 300.1-300.19; ICD-9 codes 296B, 296X, 300E, and 33; and ICD-10 codes F30-33, F34.1, F38.8, and F39.0) and schizophrenia or other psychoses (ICD-8 code 295, ICD-9 codes 291-292 and 295-298, and ICD-10 code F20).
Our study was approved by the Danish Data Protection Agency, the Research Ethics Committee of the Central Denmark Region, the Research Ethics Committee at the Karolinska Institute in Sweden, and Statistics Finland and the National Institute for Health and Welfare in Finland. Informed consent from the participants was not necessary because the data were deidentified and not recognizable at an individual level.
Descriptive statistics were applied to explore baseline characteristics and incidence of suicide. To assess the absolute risk, we calculated the unadjusted cumulative incidence proportion for boys and girls by using the competing risks method while accounting for offspring death by other causes.30 The IRR for suicide in the offspring was calculated using a Poisson regression model with the logarithm of the person-years as an offset variable.31,32 To account for heterogeneity between strata, we applied a cluster-robust variance estimation. Regression analyses were adjusted for sex of offspring, calendar time at parental death, offspring age at time of parental death, and country of residence. The association between suicide and parental death was examined across subgroups of children based on sex, calendar time at parental death, maternal/paternal death, cause of parental death, offspring’s age at time of parental death, psychiatric diagnoses at baseline of surviving parent, psychiatric diagnoses of deceased parent, country of birth, age of deceased parent, age of surviving parent, birth order, parental education level, and parental socioeconomic status. A hypothesis of no interaction was assessed using a Wald test. A forest plot was used to illustrate associations (Figure 1). Furthermore, we examined the association between incidence of suicide and time since bereavement. All data handling and statistical analyses were performed with SAS statistical software, version 9.2 (SAS Institute Inc), and Stata, version 11 (StataCorp).
Within the total population of 7 302 033 persons, we identified 189 094 offspring (2.6%) who had a parent who died before the child reached 18 years of age (ie, the bereaved cohort) and 1 890 940 children who did not have a parent who died before they reached 18 years of age and were matched by age, sex, and country of residence (ie, the reference cohort). During 28.8 million person-years of follow-up (range, 1 day to 40 years; median, 13 years), 1607 of 2 080 034 persons (0.08%) died from suicide (ie, both cohorts), 265 (0.14%) in the bereaved cohort and 1342 (0.07%) in the reference cohort. Baseline characteristics show that the parents in the bereaved cohort were more likely to be older and to have 9 years of education or less, a low socioeconomic status, and several psychiatric disorders (Table). The crude risk of suicide was higher in the bereaved cohort than in the reference cohort for all groups presented.
During the first 25 years of follow-up, the cumulative incidence proportion of suicide was higher among boys (4 suicides in the bereaved cohort and 2 suicides in the reference cohort, per 1000 boys) than among girls (2 suicides in the bereaved cohort and 1 suicide in the reference cohort, per 1000 girls) (Figure 2). Compared with nonbereaved children, the overall IRR of suicide was 2.02 (95% CI, 1.67-2.44) for bereaved children. The risk of suicide was particularly high for children who had a parent who died of suicide (IRR = 3.44 [95% CI, 2.61-4.52]) but was also high for children who had a parent who died of other causes (IRR = 1.76 [95% CI, 1.49-2.09]) (Figure 1). Children bereaved by parental death by suicide had an 82% higher risk of suicide than children bereaved by parental death by accident after adjusting for age, country, and sex (IRR = 1.82 [95% CI, 0.98-3.38]). The IRR of suicide tended to be particularly high for boys who had a mother who died (IRR = 2.52 [95% CI, 1.93-3.27]), children who experienced parental death before reaching 6 years of age (IRR = 2.83 [95% CI, 2.12-3.78]), and first-born children (IRR = 2.22 [95% CI, 1.75-2.82]). We found no substantial differences in IRR for suicide in subgroups of children based on country of residence, study period, parental psychiatric history, or parental socioeconomic status.
Overall, the IRRs of suicide remained high for at least 25 years after parental death (Figure 3). The IRRs for risk of suicide based on time since bereavement did not vary much by age of the child at the time of bereavement (results not presented).
In this cohort study, parental death during childhood was associated with an increased risk of suicide for the offspring, which lasted for at least 25 years after the parent’s death. In the bereaved cohort, the highest risk was found among children bereaved by parental suicide (particularly boys who had a mother who died of suicide), first-born children, and children who had a parent who died before the child reached 6 years of age, but the risk was also markedly increased for children bereaved by parental death from other causes. The absolute risk of suicide during the first 25 years of follow-up was 4 in 1000 boys and 2 in 1000 girls who had a parent who died during their childhood.
Our finding of a more than 3-fold higher risk of suicide among offspring who had a parent who died of suicide during their childhood is consistent with several other studies.6,7,9,18,24 However, to our knowledge, only 1 study9 has examined whether this risk also applies to the death of a parent due to other causes, which is a more common exposure. A cohort study9 conducted in Sweden in a subset of our study population observed 503 229 offspring who had a parent who died before the child reached 25 years of age and found an IRR of suicide of 1.9 after parental suicide, but no statistically significantly increased risk among offspring who had a parent who died of other causes. A Danish register-based case-control study6 revealed no association between parental death and subsequent suicide attempts. In comparison, the cohort in our study is based on data from 3 Nordic countries known for high statistical precision, and we found that parental death from other causes was associated with an increased risk of suicide.
Only a few studies have had sufficient size and follow-up time to evaluate the risk of suicide among high-risk children and to consider timing. We are aware of only 1 previous study9 that assessed the suicide risk according to age of child at time of parental suicide. Exposure to parental suicide before the child has reached 17 years of age was associated with a 3-fold increased risk of suicide compared with exposure in early adulthood (≥17 years of age).9 One other study18 found that children who had a mother who died of suicide had a higher risk of suicide attempts compared with children who had a father who died of suicide. In the present study, the data allowed us to establish an increased risk after maternal death; this risk was much more pronounced among boys than girls. Furthermore, first-born children had a slightly higher risk of suicide after the death of a parent compared with second-born or later-born children. This is in contrast to the finding of a recent British birth cohort study,19 which reports an increased risk of suicide among later-born children. The underlying mechanism for an association between birth order and risk of suicide after parental death is unknown. We speculate that this might be explained by a number of first-born children being thrust into a parental role and children classified as “first-born” who are actually the only child in the family, which might increase their vulnerability to familial loss. In addition, an association between suicide and a family history of psychiatric disorders has been established in earlier Danish register-based studies.12,13 Generally, inconsistencies in the results might be explained by differences in exposure (parental suicide rather than parental death from other causes) and in outcome (suicide rather than suicide attempts).
The sample size of our study is unparalleled compared with other studies on the risk of suicide and provides estimates with high statistical precision, even in most of the subgroup analyses. We established a cohort of all children in Denmark and Sweden and a large random sample of children in Finland, and we observed the study participants for up to 40 years, virtually without loss to follow-up. Bias due to selection of study participants, therefore, cannot explain our results. Although, to our knowledge, the present study is the largest of its kind, the investigation of the interaction between suicide and psychiatric disorders was challenged owing to the rarity of the events, and psychiatric disorders in offspring who died by suicide were so rare that this factor could not be fully investigated.
In the Nordic registration system, the overall validity and the overall completeness of the records of death are close to 100%, which ensures that the classification of overall death is accurate. However, the quality of registrations of cause of death is known to vary on death certificates. If physicians are more likely to classify a death as a suicide for a person who had a history of parental death, we may have overestimated the association between parental death and suicide. On the other hand, psychiatric disorders related to parental suicide may have been underreported.17 Because psychiatric disorders were recorded for persons whose conditions were diagnosed by a psychiatrist only, our study holds no information on the large group of persons who are treated for mental health issues in primary health care. However, we believe that the effect of such potential misclassification is likely to be small. Furthermore, data on parental education level were available only for Denmark, and data on socioeconomic status and psychiatric disorders were available only for Denmark and Sweden.
Although we adjusted for several potential confounding factors, residual confounding by unmeasured factors cannot be ruled out. Our register-based study had no information on important risk factors such as genetic factors, the parent-child attachment pattern, the social network, lifestyle factors in the family, and social and psychological distress in the immediate aftermath of the parent’s death. Bereavement is just 1 step on the causal pathway between parental death and suicide risk among offspring and, therefore, should not be adjusted for.33
The underlying causal mechanism for the association between parental death and subsequent suicide risk among offspring remains unknown. Because many potential mechanisms may lead to suicide after parental death, several explanatory models have been considered. Some studies suggest that suicide after parental death may reflect a shared genetic and epigenetic component triggering mental health problems.12,20,25,34 Melhem et al35 have shown that variations among offspring reactions to parental death may be explained by factors that antedate the death of the parent. Other studies have indicated that suicide is a consequence of psychological and biological stress responses to the psychological trauma caused by the loss of an attachment figure, which has been shown to lead to hypothalamic-pituitary-adrenal dysfunction, complicated grief, impaired emotion regulation, subsequent mental health problems, and susceptibility to suicidal behavior.4,5,36- 39 Furthermore, a social-oriented explanation points to social and economic disadvantages in the bereaved family and the potential effects on parental resources, the social network, and environmental settings, although the psychological health and resilience of the surviving parent will still play a vital role.1,2,20 These potential mechanisms may interact and thus increase their effect on suicidal behavior in vulnerable groups. Hence, the pathways from exposure to outcome are likely to be individualized, multifactorial, highly complex, and hard to map. Nevertheless, our study suggests that a trajectory leading to suicide can have its roots in early-life conditions, such as the death of a parent.
Because the present study is based on Nordic health registers, the generalizability of the findings may be limited to primarily Nordic and Western societies. The Nordic health care systems are publicly funded and easily accessible, which may affect both health care patterns and risk factors. Yet, the associations revealed in our study add to the literature on the far-reaching effects of parental death during childhood.
Suicidal behavior is preventable, and early mitigation of risk may have beneficial effects. Yet, the development of clinical and public health efforts is challenged by the rarity of suicide, the highly complex pathways from death of a parent to suicide, and the long-term risk profile of offspring who had a parent who died of suicide. Hence, the challenge is to identify the very small subset of people for whom the risk of suicide is high enough to make treatment ethically and financially worthwhile. Future public health strategies could focus initiatives on families with a history of suicide and psychiatric disorders and provide social welfare programs to reduce exposure to social and financial stressors. One targeted preventive strategy would be to monitor distress in bereaved children and provide support to help highly distressed children cope with bereavement.
In conclusion, parental death during childhood was associated with a long-lasting increased risk of suicide among offspring who had a parent who died of suicide or other causes. The underlying causal mechanism may be attributed to shared genetic dispositions, environmental factors, social changes, and psychological stress originating from early-life conditions. Preventive efforts are challenging because of the highly complex interactions among family members and the long-term risk profiles observed among offspring. Our study points to the early mitigation of distress to reduce the risk of suicidal behavior among children who had a parent who died during their childhood.
Corresponding Author: Mai-Britt Guldin, PhD, Research Unit for General Practice, Department of Public Health, Aarhus University, Bartholins Alle 2, 8000 Aarhus C, Denmark (email@example.com).
Published Online: November 11, 2015. doi:10.1001/jamapsychiatry.2015.2094.
Author Contributions: Drs Pedersen and Vestergaard had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Guldin, Li, Pedersen, Obel, Agerbo, Gissler, Vestergaard.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Guldin, Pedersen, Agerbo.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Guldin, Pedersen, Agerbo, Gissler.
Obtained funding: Li, Cnattingius, Vestergaard.
Administrative, technical, or material support: Cnattingius, Olsen.
Study supervision: Li, Agerbo, Gissler, Cnattingius, Olsen, Vestergaard.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by the Lundbeck Foundation (unrestricted grant R155-2012-11280) and the European Commission’s Seventh Framework Programme (grant ERC-2010-StG-260242-PROGEURO). Data recruitment was supported by the Danish Council for Independent Research/Medical Sciences (grant 09-072986), the Swedish Council for Working Life and Social Research (grant 2010-0092), and the Nordic Cancer Union (grant 2013-78760).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.