METHODS

This study was based on data from national registers held by the Swedish National Board of Health and Welfare, and Statistics Sweden. The key to these registers is the unique personal identification number, which follows each Swedish resident from birth to death. This number was used to link data from the registers to each participant.

Study population

The study population consisted of the entire Swedish population born 1973–82 who had settled in Sweden before 7 years of age and were recorded to be living in family households in the November 1990 census, 989 871 people. Four study groups were created within this population in a stepwise hierarchical fashion so that every person could only be included in one study group.

Firstly, 12 240 intercountry adoptees were identified by having a record of being born outside of Europe, no record of a biological parent in the Swedish Parent Register, immigration to Sweden before 7 years of age and parent(s) who were recorded to have been born in Sweden without any record of emigration or immigration after 1968. Secondly, 6437 long term foster children were defined by records of foster care for at least two years before their 13th birthday in the Swedish Register of Children and Young Persons subjected to Child Welfare Measures. The mean time in care in this group was 7.7 years and the median time 7.3 years. As long term placements in residential care are very rare for children who enter Swedish care before their teens,⁶ care experience in this group is dominated by foster family care.

Thirdly, 15 868 other children in families where at least one child had been subjected to child welfare measures before their 13th birthday were identified. The measures ranged from temporary placements in foster or residential care for less than two years to regular support by an adult outside of the family. For the purpose of simplification, this group will be referred to as having received “other child welfare interventions” in the text. The remainder of the population; 955 326 children, were labelled the general population.

Sociodemographic variables

Mother was defined as the biological mother in the Swedish Parent Register with the exception of the intercountry adoptees where the female caregiver in the household in the census of 1990 was considered to be the mother. The Swedish Population and Housing Census 1990 was the source of the following information about the mother: year of birth, sex, socioeconomic status (SES) of the household, housing situation, family status (lone parent household), and geographical location of the home (residency). SES was defined according to a classification used by Statistics Sweden, which is based on occupation but also takes educational level of occupation, type of production, and position at work of the head of the household into account.¹⁰ The Total Enumeration Income Survey for 1990 was the source for social welfare benefits received (yes/no) in 1990 by the mother.

The ethnicity of the study subjects was classified according to the country of birth of the adults in the household of the mother in the census of 1990. When the adults in the household had diverse ethnicity the child was classified according to the oldest woman, and if there were no female adults in the household, the oldest man in the household. If at least one, but not all, adults in the household were born in Sweden, the child was classified in an intermediate group.

Parental morbidity

Risk factors related to psychiatric illness and addiction in the biological parents (thus excluding intercountry adoptees) were obtained through individual record linkage to the Swedish Hospital Discharge Register for the years 1987–1990. The variables were defined according to the 9th revision of the WHO International Classification of Diagnosis (ICD-9)

• Any psychiatric disorder: a main diagnosis of 290–319 (ICD-9)

• Drug misuse: a main or contributory diagnosis of 292, 304, 965.0, 968.5, 969.6, 969.7.

• Alcohol misuse: a main or contributory diagnosis of 291, 303, 305.0, 357.5, 425.5, 535.3, 571.0-571.3, E860, E980+980

Outcome variables

The concept of avoidable mortality has been developed within research on the quality of medical and psychiatric care.^11,12 This concept is intended to increase the value of mortality rates as quality indicators of care by focusing on deaths that could possibly have been avoided by prevention or adequate medical care. In this study the concept included deaths from natural causes related to alcohol and substance misuse, deaths from natural causes that could possibly have been avoided by proper medical care, and intentional and unintentional injuries (table 1).

Statistical methods

The study population was followed up prospectively from the time of the Swedish Population and Housing Census in November 1990 or from the 13th birthday, whichever came last, to December 2000 in the Swedish National Cause of Death Register. Multivariate analyses were conducted with Cox regressions of proportional hazards of time to event with the variables defined in table 1 as the outcome variables. Time in the study was calculated with date of birth from the census of 1990, date of death from the National Cause of Death Register and date of emigration from the Total Enumeration Income Survey. Birth year was entered as a continuous variable in the regression models, as death rates increased in a linear fashion with this variable. Other sociodemographic variables were entered as categorical variables (using dummy variables) into the models. Missing values in socioeconomic variables were included in the models as a separate category.

Hazard ratios of relative risk (RR) were analysed in three regression models. The first model was adjusted only to the year of birth and sex of the study population. This model was used for comparisons between the study groups as we did not have information about to what extent intercountry adoptees had been exposed to the other confounders. The second and third models were created to allow for a comparison with children who had not received child welfare intervention but had been raised in homes with similar parental (model 2) and social characteristics (model 3) as the biological homes of the child welfare recipients. Model 2 included indicators of addiction and psychiatric illness in the biological parents and model 3 added socioeconomic indicators of the household of the biological mother to the variables in model 2. The intercountry adoptees were excluded from the second and third models because of the lack of information about their biological parents. The SPSS software package, version 11.0, was used in all statistical analyses.

RESULTS

The sociodemographic characteristics of the intercountry adoptees stand out among the study groups with a strong female preponderance (59%), the highest SES of the household, the highest maternal age at birth of the child (mean: 32.2 years) and the highest proportion living in houses (86%). The original homes of the long term foster children, on the other extreme, had the youngest mothers at the birth of the child (mean: 25.4 years), the lowest SES, the lowest level of maternal education, and most often received social welfare in 1990 (51%). Forty one per cent of the long term foster children had one or two parents who had been admitted to hospital because of alcohol misuse, substance misuse, or psychiatric disorder, or all three. Children who had received child welfare interventions but had mainly grown up in their original family had characteristics somewhere in between the general population and the original families of the long term foster children for psychosocial risk factors as well as socioeconomic variables (table 2).

There were 2170 deaths (26/10 000 person years) in the general population, 52 (89/10 000 person years) among the long term foster children, 53 (50/10 000 person years) among the intercountry adoptees, and 90 deaths (66/10 000 person years) in children who had received other kinds of child welfare interventions. Deaths because of severe congenital disorders, such as Down’s syndrome, tended to be most common among the long term foster children and other recipients of child welfare; 5/10 000 person years compared with 1/10 000 person years in the general population (table 3).

Long term foster children and intercountry adoptees had increased age and sex adjusted hazard ratios for suicide, relative risks (RRs) 4.3 and 3.5 (table 4) compared with the general population, while those that had received other kinds of child welfare interventions had a hazard ratio between these groups and the general population (RR: 2.7). When the analysis was adjusted to the morbidity and socioeconomic situation of the biological parents, to create a comparison with adolescents raised in similar psychosocial circumstances as the original home, the RR decreased considerably but were still significantly higher than those of the general population for long term foster children (RR 2.2) as well as for children with other forms of child welfare interventions (RR 1.7).

Children in long term foster care as well as children that had received other kind of child welfare interventions had higher age and sex adjusted hazard ratios for “other avoidable deaths” (avoidable natural causes, addiction as well as unintentional injuries and homicide) in comparison with the general population (RRs 2.5 and 2.8 respectively) as well as the intercountry adoptees (table 4). When the hazard ratios for other avoidable deaths were adjusted for the morbidity and socioeconomic status of their biological home the risks decreased considerably for children in long term foster care, and remained significantly higher than those in the general population for children with other forms of child welfare interventions only (RR 1.8).

The hazard ratios of all avoidable deaths were analysed in two age specific groups; 13–17 years and 18–27 years. In the younger age group the RRs after adjustment for year of birth and sex were 2.3 (95%CI 1.3 to 3.8) for intercountry adoptees, 2.5 (95%CI 1.2 to 5.0) for long term foster children and 2.8 (95%CI 1.9 to 4.6) for other child welfare recipients compared to the general population in the age group 13–17 years, while the RRs in the age group 18–27 were 1.8 (95%CI 1.2 to 2.7) for intercountry adoptees, 3.3 (95%CI 2.4 to 4.7) for long term foster children, and 2.6 (95%CI 2.0 to 3.3) for other child welfare recipients compared with the general population.

DISCUSSION

This study shows that long term foster children, other child welfare recipients, and intercountry adoptees have threefold to fourfold increased risks for suicide deaths during adolescence and young adulthood compared with the general population. Both groups of welfare recipients—in contrast with the adoptees—also displayed increased risks of other types of avoidable mortality. It is very rare that an undesirable outcome for adoptees in this age group resembles that of long term foster children.^2,13 This seems to indicate that other factors than those related to the socioeconomic context during adolescence are important in explaining the high risk of suicide in children in substitute care (foster children and adoptees). Many foster children as well as intercountry adoptees have been exposed to multiple risk factors during early childhood, factors such as child abuse and neglect, malnutrition, intrauterine exposure to narcotics, alcohol, and psychotropic drugs as well as the hereditary influence of having parents with psychiatric disorders.^14–16 All children in substitute care are also subjected to separations from their biological parents and the task of creating new bonds with the substitute parents. Attachment theory provides a framework for linking these separations to suicide as recent studies^17,18 have demonstrated that insecure attachment patterns are related to suicidal ideation and suicide attempts in adolescents. Juffer et al¹⁹ have shown that insecure attachment patterns in adoptive families are common but can be identified and modified by counselling. Further studies are needed to clarify the role of this mechanism in suicidal behaviour in children in substitute care. It is also possible, however, that the high risks of suicide for long term foster children and intercountry adoptees are caused by parallel causal mechanisms. For instance, racism²⁰ during adolescence and discrimination on the labour market in young adult years²¹ may explain at least a part of the higher risk for suicide among international adoptees, while there is a strong overrepresentation and an aggregation of various risk factors among long term foster children compared with other child welfare recipients.¹³ Several studies have affirmed that children with poorer dispositions, more damaging early adverse experiences, and more behavioural problems tend to stay longer in care, or return after failed reunification attempts.^22,23

The higher risk of death in foster children compared with adoptees after they leave care (over 18 years of age) highlights the difference of being a permanent and transitory member of a substitute family. It may be speculated that the lack of support during late adolescence after leaving care may be important in explaining this finding.²⁴ Both US and European research has for example consistently shown that former foster children are heavily over-represented in homeless populations, even though the links between care experiences and homelessness are unclear. Also, studies that have found increased mortality risks for foster children have noted that these risks are restricted to time after care, not while the children are in care.^25,26

The risk of death for the foster children as well as other child welfare recipients decreased considerably when they were compared with youth from homes with similar psychosocial characteristics as the original homes of the child welfare recipients. This suggests that social adversity and hereditary factors related to substance misuse and psychiatric disorders are important in explaining this higher risk. On the other hand there was no indication that child welfare interventions improved the chance of survival during adolescence for children at risk, since the risk of death was worse rather than better than for youth from homes with similar psychosocial characteristics. However, it is quite possible that the inclusion of more detailed psychosocial background data of the original family would have enabled better adjustment for confounders related to the specific cause of the child welfare intervention, which naturally could have affected the analysis. The lower mortality rate of the adoptees for other types of “avoidable mortality” than suicide may be interpreted as a hopeful indication that a considerable number of deaths can be avoided with an improved content of interventions from the society.

To our knowledge this is the first study that compares mortality rates in adoptees with children who have received child welfare interventions, and also the first study that uses the concept of avoidable mortality to a child welfare population. The results suggest that this concept improves the analysis by limiting confounding through selection of children with congenital disorders. It is possible, however, that some confounding because of selection of children with acquired chronic disorders is still present in the analysis. This comparatively large national sample allowed us to study suicide apart from other avoidable deaths and to use multivariate models to compare death rates of child welfare recipients with children raised in similar psychosocial environments as the original homes of the adolescents in the study. An important limitation of the study was the general nature of the background information of each child before entering substitute care, where factors such as malnutrition, abuse, and neglect, and multiple separations from caregivers may be important to explain differences in mortality rates between the study groups. Another limitation of the study was the indirect criteria for identifying adoptees made necessary by the limitations of the available registers. These criteria made it necessary to exclude certain smaller groups of intercountry adoptees, such as those with parents who adopted while living abroad and adoptive homes where one or two of the parents were foreign born.