Mortality rate ratios are stratified by all deaths, deaths from diseases, and deaths from external causes. Error bars indicate 95% CIs.
eTable 1. Cause-Specific Mortality Rate Ratio (MRR) After Sibling Death in Childhood
eTable 2. Mortality Rate Ratio (MMR) After Sibling Death in Childhood by Type of Death, With Additional Adjustment for Comorbid Diseases Among Index Children
eTable 3. Mortality Rate Ratio (MRR) After Sibling Death in Childhood by Type of Death With a Fine Categorization of Age and Calendar Year in 1-Year Interval
eTable 4. Mortality Rate Ratio (MRR) After Sibling Death in childhood including All Children After Birth
eTable 5. Mortality Rate Ratio After Sibling Death in Childhood by Sex Difference Between Sibling Pairs in Singletons
Customize your JAMA Network experience by selecting one or more topics from the list below.
Yu Y, Liew Z, Cnattingius S, et al. Association of Mortality With the Death of a Sibling in Childhood. JAMA Pediatr. 2017;171(6):538–545. doi:10.1001/jamapediatrics.2017.0197
Does sibling death in childhood lead to increased mortality in bereaved individuals?
In this population-based cohort study, a sibling death in childhood was associated with a 71% increased all-cause mortality risk among bereaved persons. The higher mortality risks were found in the first year and among same-sex sibling pairs or sibling pairs with a small age difference.
The minimization of potential adverse effects of sibling death in childhood is important, especially in the first year and for sibling pairs of close age and same sex.
The death of a close relative is associated with an increased mortality risk among the bereaved, but much less is known about the potential association of the death of a sibling in childhood with mortality in this population.
To examine the association between sibling death in childhood and subsequent mortality risk.
Design, Setting, and Participants
This population-based cohort study of 5 005 029 participants evaluated linked national registers in Denmark (January 1, 1973, through December 31, 2009) and Sweden (January 1, 1973, through December 31, 2008). A total of 2 060 354 Danish and 2 944 675 Swedish children who survived the first 6 months of their life were included. We excluded 14 children who died of the same external cause as their siblings within 30 days. Data were analyzed from November 2, 2015, through October 14, 2016.
Participants were classified as exposed if a sibling died in childhood (age <18 years).
Main Outcomes and Measures
Poisson regression was used to estimate mortality rate ratio (MRR) with the exposure as a time-varying variable.
Among the 55 818 participants who experienced sibling death in childhood (51.5% male and 48.5% female; median age at loss, 7.0 [interquartile range, 3.3-12.1] years), all-cause mortality risk was increased by 71% (MRR, 1.71; 95% CI, 1.57-1.87) during the follow-up of 37 years. The excess mortality risk was observed for groups with specific causes of death, and the higher MRRs were found when the sibling pairs died of the same cause (death due to disease [MRR, 2.16; 95% CI, 1.87-2.49]; death due to external cause [MRR, 1.91; 95% CI, 1.54-2.37]). The increased mortality risk after sibling death was seen across the follow-up period, regardless of the age at bereavement and the type of death among bereaved siblings, but the magnitude of association was stronger during the first year after sibling death (MRR, 2.51; 95% CI, 1.79-3.54). Higher MRRs were found among sibling pairs with the same sex (MRR, 1.92; 95% CI, 1.70-2.18) and close age (MRR, 1.94; 95% CI, 1.58-2.37).
Conclusions and Relevance
Bereavement in childhood because of the death of a sibling was associated with an increased risk for mortality in the short and long term. Health care professionals should be aware of individuals’ vulnerability due to sibling death, especially for sibling pairs of close age or the same sex. Social and health care support may help to minimize the potential adverse effects on the bereaved sibling.
Bereavement caused by the death of a close relative is ranked as one of the most severe life events1 and is likely to cause grief regardless of coping mechanisms.2 An increased risk for mortality and adverse health outcomes has been observed among bereaved spouses, parents, and children.2-5
Estimates suggest that nearly 8% of individuals in the United States experienced a sibling death in childhood,6 but much less is known about the health consequences to the survivors. Sibling relationship tends to be the longest and the most intimate in families; thus, the death of a sibling can be a devastating life event.7 Previous Swedish studies8-11 found that sibling death in adulthood might affect the bereaved sibling’s mortality. However, these studies were restricted to adult siblings, and the experience of psychological stress due to a sibling death in childhood may vary compared with that due to a sibling death in adulthood. Previous studies have suggested that the experience of a sibling death in childhood is associated with an increased risk for developing psychosocial distress and mental disorders later in life.12-17 However, to our knowledge, no study has investigated the association of sibling death in childhood with subsequent mortality in bereaved siblings with a long follow-up.
The severity of this bereavement is expected to depend on the closeness of the siblings.18-20 We hypothesized that experiencing a sibling death in childhood could lead to an excess mortality risk in the surviving children, particularly for sibling pairs with a small age difference and of the same sex. We examined the association between sibling death in childhood and subsequent mortality and evaluated effect modification by the siblings’ sex and age differences.
The unique personal identification number in Scandinavian countries permits accurate linkage of personal data between national registers,21 and we conducted this population-based cohort study based on linked national registers in Denmark and Sweden.22 We linked live births in Denmark (January 1, 1973, through December 31, 2004) and Sweden (January 1, 1973, through December 31, 2006) with their mothers and thereafter identified their siblings. We excluded those who died within the first 6 months of life, because infant mortality in this period is likely attributable to preterm birth, malformations, or neonatal complications.23,24 We included only children who had at least 1 sibling in childhood (age <18 years). We also excluded 14 children who died within 30 days of the same external cause as their siblings (of whom 11 [79%] died of transport accidents, for which both siblings might die due to the same accident). Our study included 5 005 029 children in the final analysis. The study was approved in Denmark by the Data Protection Agency and the Research Ethics Committee of the Central Region, and in Sweden by the Research Ethics Committee at the Karolinska Institute. The study was based on deidentified data, for which the ethics committees do not require informed consent.
All index children were included in the unexposed group at the start of follow-up (6 months after birth). The exposure was a time-varying variable. An index child would be changed to the exposed group when he or she was aged 6 months to younger than 18 years and had a sibling who died at any age. The other children remained in the unexposed group. Each child was followed up until death, emigration, or the end of the study (December 31, 2009, in Denmark, and December 31, 2008, in Sweden), whichever came first. If multiple siblings died at different times, the exposure time started at the time of the death of the first deceased sibling. The causes of death were retrieved from the Danish and the Swedish cause of death registers22 and coded using the International Classification of Diseases, Eighth Revision (ICD-8), International Classification of Diseases, Ninth Revision (ICD-9), and International Statistical Classification of Diseases and Health Related Problems, Tenth Revision (ICD-10). The causes of death were divided into deaths due to diseases (ICD-8 codes 000-799, ICD-9 codes 000-799, and ICD-10 codes A00-R99) and external causes of death (ICD-8 codes E800-E999, ICD-9 codes E800-E999, and ICD-10 codes V01-Y98) according to the European Shortlist Causes of Death.25
The main outcomes were all-cause mortality, cause-specific mortality, and mortality by type of death (death due to diseases or external causes of death as defined above). We also investigated the following causes of death by groups: neoplasms; endocrine, nutritional, and/or metabolic diseases; diseases of the nervous system; diseases of the circulatory system; transportation accidents; suicide and intentional self-harm; and other diseases (ICD-8, ICD-9, and ICD-10 codes are shown in eTable 1 in the Supplement).
Potential confounders were selected a priori based on previous literature4: preterm birth (<37 or ≥37 weeks), birth weight (<2500, 2500-3249, 3250-3999, or ≥4000 g), singleton birth (yes or no), parity (1, 2, or ≥3), Apgar score at 5 minutes (0-6, 7-9, or 10), maternal age (≤26, 27-30, or ≥31 years), maternal education (0-9, 10-14, or ≥15 years), maternal social status (not in the labor market, unskilled worker, skilled or white collar worker, or top-level status), maternal smoking during pregnancy (yes or no), country of birth (Denmark or Sweden), sex of child (male or female), age of the child (<1, 1-4, 5-9, 10-14, 15-19, 20-25, 25-30, or >30 years), and calendar period (1973-1975 or in a 5-year interval during 1976-2010). The age of bereaved siblings and calendar year period were treated as time-varying variables. A missing indicator was created for a missing value of each variable. Information on maternal and birth characteristics were retrieved from the medical birth registers in Denmark and Sweden,26,27 whereas socioeconomic factors were retrieved from the Danish Integrated Database for Longitudinal Labour Market Research, the Danish Civil Registry System, the Swedish Multigeneration Registry, and the Swedish Registry of Total Population and Population Changes.22
Data were analyzed from November 2, 2015, through October 14, 2016. We used a log-linear Poisson regression model to estimate the mortality rate ratios (MRRs) with 95% CIs to investigate the association of sibling death in childhood with subsequent mortality of the surviving sibling. A child who emigrated during the follow-up was censored at the time of emigration. Exposed participants who died were divided into the following 2 subgroups: those who died of the same cause as their sibling and those who died of a different cause, with additional adjustment for the index child’s Charlson comorbidity index28 to account for comorbidity. When cause-specific mortality and mortality by type of death were estimated, death from other causes was considered as a competing event and treated as a censored case. The association between sibling death and mortality was examined across subgroups according to the age when the child experienced the sibling’s death (<6, 6-11, or 12-17 years) and time since the sibling’s death (<1, 1-4, 5-9, 10-14, or ≥15 years). Because siblings with the same sex or a small age difference in general are socially closer,20 they may experience greater effects after bereavement. We therefore also investigated the association by age difference (<2, 2-5, 6-11, or ≥12 years) and sex difference between the deceased and bereaved siblings.
In addition, based on data availability of covariates and ICD code changes, subanalyses were performed that were restricted to children born after 1978, 1980, 1982, 1987, 1991, 1994, and 1997 and in children with information for all covariates (complete case analysis). Additional analyses were performed with children stratified by country of birth. We also performed analyses through inclusion of all children after birth. The analyses were performed using STATA/SE (version 12; StataCorp) and SAS (version 9.2; SAS Institute, Inc) software.
A total of 55 818 of 5 005 029 participants (1.1%; 51.5% male and 48.5% female) experienced the death of a sibling in childhood (from 6 months after birth to the day before 18 years of age). The median age of bereaved persons at the time of loss was 7.0 years (interquartile range, 3.3-12.1 years). Mothers of bereaved persons were more likely to be young and to have low educational levels and high parity and smoked more often during pregnancy. Compared with unexposed individuals, exposed individuals more often were of multiple births, were born preterm, and had a low birth weight and low 5-minute Apgar scores (Table 1).
During the follow-up to 37 years, 534 participants died in the bereaved group and 25 591 died in the nonbereaved group. Compared with those who did not experience sibling loss, the bereaved group had an increased all-cause mortality risk (MRR, 1.71; 95% CI, 1.57-1.87). The exposed group had increased all-cause mortality when their siblings died of disease (MRR, 1.72; 95% CI, 1.55-1.90) or an external cause (MRR, 1.65; 95% CI, 1.40-1.94) (Table 2). Bereaved participants also had excessive deaths from disease (MRR, 1.94; 95% CI, 1.70-1.21) and external cause (MRR, 1.55; 95% CI, 1.36-1.76). A stronger association was observed if bereaved participants had the same cause of death (disease or external cause) as the deceased sibling (Table 2). Increased mortality risks were found among bereaved persons after additional adjustment for their comorbid diseases irrespective of the type of deaths among sibling pairs (eTable 2 in the Supplement).
Next, we investigated cause-specific mortality. As expected, individuals who experienced the death of a sibling also had a particularly high risk for death from the same cause as their deceased sibling (eTable 1 in the Supplement). For example, the relative risks were the highest for endocrine, nutritional, and metabolic diseases (MMR, 158.08; 95% CI, 97.19-257.13); diseases of the nervous system (MMR, 24.90; 95% CI, 14.68-42.24); and diseases of the circulatory system (MMR, 44.90; 95% CI, 18.57-108.53). We also found a much higher mortality risk for suicide among individuals whose deceased siblings died of suicide (MMR, 8.01; 95% CI, 5.34-12.00).
With respect to time of follow-up, we observed increased MRRs of sibling death associated with subsequent mortality in the short and long term regardless of the age at bereavement and type of death among the bereaved group (Figure), but the strongest association was seen in the first year after sibling death (MRR, 2.51; 95% CI, 1.79-3.54). For external causes of death, the magnitude of association was notably stronger during the first year of follow-up (MRR, 8.58; 95% CI, 3.99-18.48). The proportions of deaths from external causes tended to decrease with the increasing length of follow-up. For death from disease, the association also appeared to be stronger in the first year after bereavement (MRR, 2.06; 95% CI, 1.39-3.04) and the estimated effect seemed to persist into older ages.
When estimating the effect of sex difference between sibling pairs, we observed a higher effect size in same-sex sibling pairs (MRR, 1.92; 95% CI, 1.70-2.18) than in opposite-sex pairs (MRR, 1.53; 95% CI, 1.33-1.77) (Table 3). The analyses stratified by sex of the bereaved persons revealed that bereaved male participants had a higher mortality risk than bereaved female participants regardless of whether the deceased sibling was a male or female. For bereaved female participants who died of an external cause, we did not observe an increased mortality risk after sibling death. The strength of association was stronger in close age group (<2 years) (Table 4), particularly for bereaved participants who died of external cause (MRR, 2.04; 95% CI, 1.50-2.77).
In sensitivity analyses, a fine categorization of age and calendar year in 1-year interval (eTable 3 in the Supplement) and inclusion of all children after birth (eTable 4 in the Supplement) did not essentially change the result. The overall MRRs from the separate analyses restricted to singletons, individuals with a gestational age of at least 37 weeks, a birth weight of at least 2500 g, and a 5-minute Apgar score of 10; to those with complete case analysis; and to individuals born after 1978, 1980, 1982, 1987, 1991, 1994, and 1997 were consistent with the results from the main analyses. The results of the analyses stratified by countries were largely similar in the Danish (MRR, 1.78; 95% CI, 1.57-2.02) and Swedish (MRR, 1.62; 95% CI, 1.44-1.83) populations. The analyses on sex difference between sibling pairs restricted to singletons yielded similar results (eTable 5 in the Supplement).
In this large population-based cohort study, we found that loss of a sibling in childhood was associated with an increased mortality risk later in life. The increased mortality persisted across the follow-up period, irrespective of age at bereavement and type of death, but the magnitude of association was strong during the first year after sibling death. The higher MRRs were found when the siblings had the same cause of death or the siblings had the same sex or a small age difference.7,19,29
Sibling relationships generally last longer than other familial relationships and may have a significant influence on one’s social and emotional lifetime development.7,19,29 Bereaved individuals might be directly affected by their siblings’ deaths and also indirectly through parents’ and other family members’ reaction to the deceased siblings.30 Studies examining potential adverse effects of sibling death in childhood are sparse, compared with research on parental death in childhood or sibling death in adulthood, probably owing to fewer available data. The magnitude of grief due to sibling death in childhood can be comparable to that due to parental death in childhood or sibling death in adulthood.4,5 In fact, we found that the overall mortality risk after sibling death in childhood was slightly higher than overall mortality risk after parental death in childhood.4 A previous study also indicated that the death of a sibling in childhood was associated with a higher risk for serious emotional and/or behavioral problems than parental death.16
We observed a stronger association when both the deceased sibling and the bereaved person died of disease or an external cause. Similar patterns have also been reported for parental death in childhood4,5 and sibling death in adulthood.10,11 With regard to cause-specific mortality from diseases of the endocrine, nutritional, and metabolic systems; the nervous system; and the circulatory system among bereaved participants, we found notably larger MRRs when their deceased sibling died of the same cause. This finding may be attributed to shared family environmental exposures and genetic susceptibility. The association for sibling pairs who both died of an external cause might be more complex and could reflect the interplay of genetics and social or environmental factors. On the other hand, we also observed an increased mortality risk among bereaved persons if their sibling died of a discordant cause, although of smaller magnitude. Adjustment for comorbidity did not essentially change these risks. An increased risk for discordant causes of death is less susceptible to genetic confounding, and the finding may reflect an association with bereavement. The psychological stress after bereavement has been proposed to play a significant role.5,22
A stronger association was seen when the sibling pairs had the same sex or a smaller difference in age. Siblings of the same sex and close age tend to have more frequent interactions and experience greater feelings of warmth and closeness.18-20 At the same time, perceived conflict is expected to be great in siblings of the same sex or a close age,18,31 which might lead to the guilt of the surviving siblings about the death of their sibling. The resulting guilt may contribute to further distress and increase the vulnerability of the surviving siblings.32 Our finding that bereaved male participants had a higher mortality risk than bereaved female participants was consistent with previous research4 in which boys who experienced the death of a parent in childhood had a higher mortality risk than did girls. However, a study using Swedish data11 found a higher mortality risk among female participants who experienced the death of a sibling in adulthood than among male participants. Additional research is needed to explore the underlying reasons for these differences.
We found that the increased mortality persisted across the follow-up period. The strength of grief caused by the death of a sibling is generally expected to weaken over time, but the death of a sibling could be a long-lasting stressor leading to allostatic load or long-term grief.33,34 Bereaved persons were found to have a prominently increased mortality rate in the first year after their siblings’ deaths, particularly if bereaved persons died of external causes. The posttraumatic stress of the bereaved parents may be directly related to the quality of care and support that the bereaved individuals subsequently received during the very early time after their siblings’ deaths. This process could lead to an immediate adverse health consequence.35-37 Our results also showed that most deaths in the first year after bereavement were attributable to external causes. In contrast, a Swedish study11 showed that the association between sibling death in adulthood and mortality was stronger among bereaved individuals in the long term than in the short term (>1 and ≤1 year since sibling death, respectively). The discrepant results may stem from whether the sibling death occurred in childhood or adulthood. In previous research on mortality after parental death in childhood,4 MRRs for diseases tended to increase during follow-up, whereas the highest MRR for external causes was seen at the beginning of follow-up. The potential mechanisms underlying the differences of how an individual reacts to parental and sibling deaths remain unknown.
The main strength of our study is the use of large-scale register-based data from Denmark and Sweden. The population-based cohort study, which includes all individuals from both countries and provides accurate information on mortality, exposure, and other covariates, eliminates the possibility of recall bias or selection bias. Also, we adjusted for a broad range of potential confounders. The sufficient sample size with a long follow-up period provides a unique opportunity to investigate cause-specific mortality and evaluate long-term consequences of sibling death by the age at bereavement and type of death.
Limitations must also be noted. We do not have data on shared social environment and family characteristics between sibling pairs, which might help to further elucidate the underlying mechanisms linking sibling death and subsequent mortality among the bereaved individuals. Although associations were found in increased mortality risks for death from different causes among sibling pairs, we cannot rule out the possibility of residual confounding by unmeasured confounders. Furthermore, an interplay between bereavement and other factors may exist, including stress and family or social factors (such as individuals’ socioeconomic and/or psychosocial characteristics, parental marital status, and quality of the marital relationship). However, some of these variables are not well defined, and few are available in our data set. Additional research is required to understand the mediating or moderating effects of these factors, which will broaden the perspective on the mechanism and inform the design of intervention. In addition, although we have a large cohort from 2 countries, numbers of events are small in some cause-specific mortality groups, especially in the analyses examining the same or discordant causes of death within sibling pairs.
Experiencing a sibling death in childhood was associated with short- and long-term increased mortality risks. Health care professionals should be aware of children’s vulnerability after experiencing sibling death, especially for same-sex sibling pairs and sibling pairs with close age. Social support may help to reduce the level of grief and minimize potential adverse health effects on the bereaved individuals.
Corresponding Author: Yongfu Yu, PhD, Department of Clinical Epidemiology, Aarhus University Hospital, Olof Palmes Alle 43-45, 8200 Aarhus N, Denmark (firstname.lastname@example.org).
Accepted for Publication: January 25, 2017.
Published Online: April 24, 2017. doi:10.1001/jamapediatrics.2017.0197
Author Contributions: Drs Yu and Li had full access to all 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: Yu, Olsen, Vestergaard, Li.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Yu, Olsen, Fu.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Yu, Liew, Fu, Parner, Qin, Zhao.
Obtained funding: Vestergaard, Li.
Administrative, technical, or material support: Cnattingius, Olsen.
Study supervision: Liew, Cnattingius, Olsen, Vestergaard, Fu, Qin, Zhao, Li.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grant ERC-2010-StG-260242-PROGEURO from the European Commission’s Seventh Framework Programme/European Research Council (Dr Li), projects 09-072986 and DFF-6110-00019 from the Danish Council for Independent Research (Dr Li), grant 2010-0092 from the Swedish Council for Working Life and Social Research (Dr Cnattingius), grants 176673 and 186200 from the Nordic Cancer Union (Dr Li), the 2016 Karen Elise Jensens Fond (Dr Li), and unrestricted grant R155-2012-11280 from the Lundbeck Foundation (Dr Vestergaard).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: