Outcomes in adulthood of adoption after long-term foster care: A sibling study

Abstract

Prior research has reported a positive impact of adoption on developmental outcomes for children with experience of foster care. To inform decisions about permanent care arrangements, we used Swedish national population registers to create a sibling population consisting of 194 children born 1973–1982 who had been in out-of-home care (OHC) at least 5 years before adolescence but were never adopted (50% boys) and their 177 maternal birth siblings who also had been in OHC at least 5 years before their teens but were adopted before adolescence (52.5% boys). We constructed 14 outcome variables spanning social, educational, and health outcomes in adult age with information from Swedish national registers. Based on multilevel logistic random effects and fixed effects regression models (supplemented with a sensitivity analysis assessing the potential impact of unobserved confounding), results showed that adopted siblings tended to have considerably better outcomes in adult age in educational achievement, income, criminality, disability, and suicidality. Outcomes related to mental health and substance abuse were more similar, but differences pointed in the same direction. Implications for child welfare policy and practice are discussed.

Keywords

Adoption foster care longitudinal outcome sibling study

Permanency planning has for decades been a guiding principle in U.S. child welfare, aiming to reduce the number of children growing up in nonrelative foster homes. The idea of permanency planning has also over time increasingly influenced UK legislation and practice (Rushton, 2003). Otherwise, there are major differences between Western countries in legislation, policy, and attitudes among professionals toward adoption from foster care (Warman & Roberts, 2004). Sweden, the context for this study, does not permit adoptions without birth parental consent. Adoption is simply not considered an option in child welfare. Domestic adoptions are rare—in stark in contrast to international adoption, where Sweden for decades was one of the major receiver countries in the world (Hjern, Lindblad, & Vinnerljung, 2002).

From the 1950s to the 1970s, adoptions by Swedish foster parents, usually after long stays in foster care, were far more common than today and encouraged by social workers (Vinnerljung, 1992, 1996a, 1996b). In Bohman’s and Sigvardsson’s seminal adoption studies, children placed in long-term foster care as infants were used as one comparison group when examining development over time among national adoptees born in the 1950s (Bohman & Sigvardsson, 1978, 1980a, 1980b). But at age 7, 70% of these foster children had been adopted by their foster parents (Bohman & Sigvardsson, 1990). However, the practice of adoption from long-term foster care seems to have been abandoned by professionals later.

In the early 1990s, Barth (1992) was surprised to find that Swedish social workers seemed to discourage adoption by foster parents, even when birth parental consent was given. In 2014, the National Board of Health and Welfare found that social workers were reluctant to promote adoption of foster children and did so in only a handful of cases each year (Socialstyrelsen, 2014). The Board recommended that promotion of adoption (with parental consent) should be given more priority, especially for children placed in foster care in early years with weak prospects of reunification with birth parents. A negative attitude among social work professionals to adoption from foster care is also common in the other Scandinavian countries. Danish and Norwegian legislation permits adoption without parental consent, especially for children who have been in foster care for a long time, but this legal opportunity is rarely used in practice (Bryderup, Engen, & Kring, 2017). Generally, adoption from foster care remains a controversial issue in several European countries, including the UK (McSherry, Malet, & Weatherall, 2016; Warman & Roberts, 2004).

In Bohman’s and Sigvardsson’s works, the foster children (of whom the great majority was adopted by age 7) were compared with peers who were adopted in infancy. At all developmental stages (age 11, 15, 18 and as young adults), the foster children had poorer cognitive, behavioral, and social outcomes (Bohman & Sigvardsson, 1980a). Generally, research evidence from other comparisons of adopted children and peers that grow up in OHC is mixed. A crude summary suggests far better stability for adoption (Christoffersen, Hammen, Andersern, & Jeldtoft, 2007), mostly few and relatively weak differences in various outcome areas during the latency period and early adolescence (summaries in Christoffersen et al., 2007; Rushton, 2003, 2004, 2007), but better outcomes for adoptees in young adult age (Christoffersen, 2012; Dumaret & Stewart, 1985; Triseliotis, 2002; Vinnerljung & Hjern, 2011). It seems that “recovery” from troubled adolescent years is common among adoptees (Feigelman, 1997). A general problem in comparative studies of adoptees and foster children is residual confounding from a black box of familial risk factors associated with hereditary factors, fetal exposures to toxic substances and early childhood environment before coming into care (Vinnerljung & Hjern, 2011), and birth parental pathology (Thomas, 2013). One way to reduce such confounding is to use a sibling design. The only sibling study that we know of where adopted and fostered siblings are compared in a longitudinal frame comes from France. Dumaret and Stewart (1985) looked at cognitive competence (IQ) and school performance in teens of 35 children who were adopted into families of high socioeconomic status during infancy, and 21 birth siblings who had grown up in out-of-home care (OHC; foster family and residential care). The adoptees had far higher cognitive competence than the OHC siblings and 80–90% had done well at school. In contrast, all the OHC siblings were considered “school failures”.

Using Swedish longitudinal registry data, the purpose of this study is to test the hypothesis that adoption has a positive impact on developmental outcomes in children with long-term experience of foster care. This is achieved by comparing children who spent at least 5 years in OHC before their 13th birthday with their maternal birth siblings who also spent at least 5 years in OHC before their 13th birthday but were adopted before their 13th birthday.

We have no data that can explain the obvious question: Why were some siblings adopted while others grew up in state care? Are there selection effects? The sample was delimited in the national Child Welfare Register, which contains full national cohort data from 1971 on all placements in OHC. But data are mainly restricted to date of entry and exit, and form of care (foster family or residential care). The register contains no information about reason for placement, no data on the care environment (except foster family/residential care), and do not enable reliable estimates of placement stability since only moves between residential and foster family care are recorded, not transfers from one foster family to another. The only thing we know with certainty is that the average adopted sibling entered OHC at a younger age (mean 1.2 years of age) than their sister or brother who remained in long-term OHC (mean 2.6 years of age). We are left with speculations about possible selection effects. Probably, children with fair disposition and temperament were more likely to evoke healthy attachment patterns from their substitute parents, which logically would increase the chance that their foster parents wanted to adopt after a few years of foster care (Ge et al., 1996; Rutter, 2006). It may also be that birth parents were more likely to agree to adoption for some children and not others. But there could also be a selection effect related to the foster parents. In the 1970s and 1980s, it was commonly assumed among Swedish child welfare workers that most foster parents with long-term placements wanted to adopt “their” children (like in Bohman’s and Sigvardsson studies of children born in the 1950s). But this assumption was probably wrong. Vinnerljung (1992) found in a local case file study of 108 children in long-term foster care that even when the birth parents gave consent to adoption, a majority of the foster parents that were given the opportunity to adopt “their” children declined. Swedish legislation today stipulates that after 2–3 years of placement in the same family, foster parents should be offered legal guardianship. However, a large sample study by the National Board of Health and Welfare found that a wide majority of foster carers did not want to assume legal guardianship, mainly for reasons related to economy and access to supportive services (Socialstyrelsen, 2006). The foster parents in our study who adopted may have been a qualitatively selected group within the foster family population.

With this study, we hope to contribute to our understanding of what is potentially supportive of positive long-term development for children who become “wards” of child welfare authorities. The study may also provide some answers to questions frequently being asked by child welfare policymakers and practitioners about the relative benefits and shortcomings of these two forms of substitute parenting.

Methods

This study was based on record linkages between national registers with data on the entire national population, held by the Swedish National Board of Health and Welfare, Statistics Sweden, the National School Board, and the Crime Prevention Council. The overall quality of the registers is regarded as high (Cnattingius, Ericson, Gunnarskog, & Källén, 1990; Ludvigsson et al., 2011). The registers were linked by use of the individually unique 10 digit personal id number (PIN) number that follows all Swedish residents from birth (or date of immigration) until death. The PIN enables researchers to trace individuals over the life course in national databases, with low attrition (Ludvigsson et al., 2011). In research data sets, the PIN is replaced with an anonymized “control number” by the national authorities responsible for the administration of the national registers. The study was approved by the regional ethics committee in the Stockholm region.

Study population

The population was defined as (a) all persons born in Sweden 1973–1982, according to the Medical Birth Register (held by National Board of Health and Welfare), (b) who were alive and residing in Sweden at their 15th birthday, according to the Total Population Register (held by Statistics Sweden) and (c) who had been placed in OHC before their 13th birthday, according to the Child Welfare Register (held by the National Board of Health and Welfare.

From this population, we created our study group: all persons who had been in OHC at least 5 years before their 13th birthday but were never adopted (n = 194), and who had a maternal sibling within the cohort, who also had been in OHC at least 5 years before age 13 but had been adopted before the 13th birthday. These adopted siblings (n = 177) were added to the study population (total n = 371). Siblings were identified in the Multi-Generation Register (held by Statistics Sweden). The selection criteria were inspired by an earlier Swedish sibling study, where 107 adults (from one Swedish city) with experiences of long-term foster care before adolescence were compared to their 128 maternal siblings who grew up in the care of their birth mothers (Vinnerljung, 1996a).

Long-term OHC of small children in these cohorts almost always refers to foster family care, albeit in some cases preceded by a short stay in residential care before placement with foster parents (Vinnerljung, 1996a, 1996b). All adoptions from foster care in Sweden that have been identified in previous studies were done by the child’s foster parents (Vinnerljung, 1992, 1996a, 1996b).

The sibling configurations in our sample were rare in comparison with the number of all cohort members who had been in long-term OHC (n = 4,116). Of these, 60% (n = 2,514) had at least one sibling. Henceforth, the 194 persons in our sample who (a) had grown up in foster care, and (b) had a sibling that was adopted after long-term care before age 13, constitute but a small proportion (7.7%) of all long-term foster children with at least one sibling.

Outcome variables

We used several national registers to create altogether 14 outcome variables. While the first seven outcomes are educational and social indicators, the remaining outcomes refer to general health, mental health, and substance abuse. Due to the age dependency of the indicators and the time-dependent availability of the information in the registers, the delimitation of the study population varies some between variables. The maximum number was 371 (for suicidality from 15 years to 2012) and the lowest number 353, the population that was alive and residing in Sweden in 2012.

University/college degree in 2012

Data retrieved from “the Longitudinal Integration Database for Health Insurance and Labor Market Studies” (LISA), held by Statistics Sweden (yes/no).

Only compulsory education in 2012

Data retrieved from LISA (yes/no).

Working in 2012

Any income from employment or self-employment in November 2012, according to LISA (yes/no). This variable is measured nationally by Statistics Sweden each year in November.

Low income in 2012

Annual income in the lowest national quintile income bracket (income from employment or self-employment), according to LISA (yes/no).

Any welfare in 2012

Any income from public means-tested welfare during 2012 (yes/no), according to LISA.

Any crime

At least one conviction for a crime from the 18th birthday to 2012 (with exception of traffic violations), according to the Swedish Crime Register, held by the Crime Prevention Council (yes/no). All crime data are retrieved from this register.

Severe criminality

At least one sentence to prison, probation, or forensic psychiatric care from age 18 to 2012 (yes/no). All these form of legal sanctions indicate either repeated offending or a conviction for serious crime. This variable has been used in several previous studies and has shown good discriminatory traits (Berlin, Vinnerljung, & Hjern, 2011).

Disability pension in 2012

Income from disability pension during 2012, according to LISA (yes/no). Disability pension is awarded to individuals who are deemed unfit to work because of a chronic disorder or disability. We do not have access to data on reason for disability pension but know from other studies that most disability pensions in this age-group are related to impaired mental health (Vinnerljung, Brännström, & Hjern, 2015).

Suicide attempts/suicide after age 15

All deaths after the 15th birthday to 2012 with the cause of death registered as suicide or “uncertain” in the Cause of Death Register (as recommended by Sainsbury, 1986), or at least one hospitalization after the 15th birthday to 2012 with a diagnosis of suicide attempt/self-inflicted injury, according to the National Patient Register held by the National Board of Health and Welfare (yes/no).

Psychiatric care

Any record in the National Hospital Register of hospital care with a psychiatric diagnosis from age 18 to 2012 (yes/no). After 2002, these data include outpatient clinical care, before that only episodes of inpatient care were registered.

Alcohol-related hospital care

At least one instance of hospital care from age 18 to 2012 with an alcohol-related diagnosis (main or contributory diagnosis) according to the National Patient Register (yes/no).

Indication of illicit drug problems

At least one conviction for possession or distribution of illegal drugs in the National Crime Register, dating from age 18 to 2012 (yes/no) or at least one instance of hospital care from age 18 to 2012 with an illicit drug-related diagnosis (main or contributory diagnosis) according to the National Patient Register (yes/no).

Antianxiety medication

Retrieval of at least one prescription of anxiolytics from July 2005 to the end of 2012, according to the Pharmaceutical Register, held by Statistics Sweden (yes/no). This register became operative with full coverage from July 2005, and contains records of all retrieved prescribed pharmaceuticals from pharmacies in Sweden, but does not include prescribed medicines administered in hospitals. For this, and the next outcome variable (antidepressive medication), the study population was restricted to persons alive and residing in Sweden in 2005 (N = 355). Subsequently, the study population was 23–32 years of age at the start of the follow-up and 30–39 at the end.

Antidepressive medication

Retrieval of at least one prescription of antidepressants July 2005–2012 according to the Pharmaceutical Register (yes/no).

Confounders

Data on sex and year of birth were collected from the Total Population Register, age at first placement and total time in OHC before age 13 from the Child Welfare Register. With the exception of sex, these variables were entered in the statistical analyses as continuous variables.

Statistical analyses

Siblings with experiences of long-term foster care before adolescence were compared to their maternal siblings who also had experiences of long-term foster care but were adopted before adolescence. Simple descriptive analysis of the sample and variables, including bivariate comparisons of the foster care group and the adopted group with results from two-sample test of proportions/means, is presented in Tables 1 and 2.

Table 1.

Sample descriptives: Observed background variables by type of substitute parenting.

Variable	Foster care		Adoption		No. of siblings nested within no. of biological mothers
	n_i	%	n_i	%	n_i /n_j	Difference^a
Type of substitute parenting	194	52.3	177	47.7	371/161	4.6 n.s.
Sex (male)	194	50.0	177	52.5	371/161	2.5 n.s.
Mean year of birth	194	1976.6	177	1977.5	371/161	−0.9**
Mean age placed in OHC	194	2.6	177	1.2	371/161	1.4**
Mean years in OHC before age 13	194	9.6	177	7.7	371/161	1.9**

Note. n_i = Number of siblings; n_i /n_j = Number of siblings nested within number of mothers. OHC = out-of-home care; n.s. = Not statistically significant.

^a Difference between foster group and adopted group. Results from two-sample test of proportions/means, accounting for clustering within mothers. **/*Statistical significance at the 1% and 5% level, respectively.

Table 2.

Sample descriptives: Outcome variables by type of substitute parenting.

Variable	Foster care		Adoption		No. of siblings nested within no. of biological mothers	Difference^a
	n_i	%	n_i	%	n_i /n_j
Educational and social outcomes
Only compulsory education 2012	185	33.5	168	22.0	353/160	11.5*
University degree 2012	185	11.9	168	20.2	353/160	8.3*
Working 2012	185	58.9	168	73.8	353/160	14.9**
Low income 2012	185	50.3	168	33.3	353/160	17.0**
Any welfare 2012	185	19.5	168	9.5	353/160	10.0**
Any crime	193	54.9	169	32.0	362/161	23.0**
Serious criminality	193	14.5	169	4.3	362/161	10.2**
General health, mental health and substance abuse outcomes
Disability pension 2012	185	17.8	168	8.3	353/160	9.5**
Psychiatric care	193	40.9	169	26.6	362/161	14.3**
Antidepressive medication	187	39.0	168	31.5	355/160	7.5 n.s.
Antianxiety medication	187	36.4	168	29.2	355/160	7.2 n.s.
Suicide attempts/suicide after age 15	194	14.9	177	6.8	371/161	8.1*
Alcohol-related hospital care	193	11.9	169	5.3	362/161	6.6*
Indication of illicit drug problems	193	15.5	169	13.0	362/161	2.5 n.s.

Note. n_i = Number of siblings; n_i /n_j = Number of siblings nested within number of mothers. n.s. = Not statistically significant.

Since all our outcomes were binary, we estimated logistic random effects (RE) regression models (Hox, Moerbeek, & van de Schoot, 2017). To control for unobserved common fixed childhood background endowments shared by the siblings, we also estimated fixed effects (FE) models, that is, conditional logistic models (Allison, 2009). From the perspective of precision, effect estimates from FE regression models are often less precise than corresponding RE estimates. This particularly holds when observations are thrown away due to no between-sibling variation in outcomes. Yet, since both approaches provide different perspectives on the same data, the combination of RE and FE models may advance our understandings of the addressed phenomena (Bell, Fairbrother, & Jones, 2018).

All analyses were performed using Stata 14/SE-version (StataCorp LP, College Station, TX). The bivariate analyses were based on the prtesti/ttesti commands and account for the hierarchical data structure (siblings nested within biological mothers). The regression analyses were based on the xtlogit command using cluster-robust standard errors to account for the dependence on biological mothers (Cameron & Trivedi, 2010).

We assessed to what extent our results were sensitive to unobserved confounding by calculating E-values for both the observed exposure−outcome association estimates (after adjustments for observed confounding) and for the limit of the confidence interval (CI limit) closest to the null. A large E-value suggests considerable unobserved confounding would be needed to explain away an effect estimate. A small E-value thus implies little unobserved confounding would be needed to explain away an effect estimate. The size of the CI limit value is interpreted in a similar way: a large/small CI limit implies that a large/small unobserved confounding would be needed to make the effect estimate statistically insignificant. (VanderWeele & Ding, 2017). E-values and CI limits were estimated using the web-based calculator provided by Mathur, Ding, Riddell, and VanderWeele (2018).

Results

The study population is described in Table 1. Adopted siblings tended to be slightly younger (mean birth year 1977.5 vs. 1976.6 for the fostered siblings). Also, more adopted siblings were placed in infancy (mean age of placement in OHC 1.2 years compared to 2.6 years for siblings who remained in foster care). The figures indicate that adoptions by foster parents usually took place several years after entry into foster care. Almost half of adopted siblings were adopted between age 8 and 13 (mean time in OHC 7.7 years before age 13).

Looking at frequencies of outcomes in Table 2, adopted siblings tended to fare better in all outcome variables, although statistical power was too low in some mental health and substance abuse outcomes to reach statistical significance (p < .05). One in seven foster children (15%) had records of suicidal behavior after age 15, compared to one in fifteen adopted siblings (6.8%). Crime conviction levels were also higher for the group that remained in foster care. In 2012, when our study population were 30–39 years old, more than one in six (17.8%) in the foster care group was living on disability pension, compared to one of twelve adopted siblings (8.3%).

To describe the exposure−outcome associations for educational and social outcomes, Table 3 reports odds ratios (ORs) from adjusted RE and FE logistic regression models. Due to no variation in outcomes between siblings nested within the same biological mother, it should be noted that FE estimates are based on fewer observations than corresponding RE estimates. Yet, after adjustments for sex, year of birth but also for age at entry into foster care and time in foster care before age 13, the patterns of the estimated ORs from the two regression models roughly point in the same direction for all outcome variables. Educational and social outcomes were considerably better in the adopted group, with differences being particularly pertinent for crime (although some of the estimates have poor precision).

Table 3.

Educational and social outcomes for foster group versus adopted group.

Outcome	RE estimates		FE estimates
	n_i/n_j	OR (95% CI)	n_i/n_j	OR (95% CI)
Only compulsory education 2012	353/160	2.4 (1.3–4.5)	148/61	1.9 (0.9–3.9)
University degree 2012	353/160	0.4 (0.2–0.9)	92/40	0.4 (0.2–0.9)
Working, Nov 2012	353/160	0.4 (0.2–0.7)	203/83	0.4 (0.2–0.7)
Low income 2012	353/160	2.3 (1.3–4.1)	196/84	2.5 (1.4–4.7)
Any welfare 2012	353/160	2.1 (0.9–4.8)	88/36	1.6 (0.6–3.7)
Any crime	362/161	4.5 (2.3–9.0)	214/87	7.9 (3.1–19.9)
Severe criminality	362/161	5.8 (1.7–19.2)	61/32	8.1 (1.7–39.4)

Note. Results from adjusted RE and FE logistic regression analyses. All regression models are adjusted for sex, year of birth, age at first placement in OHC, and time in OHC before age 13. Intercepts and control covariates suppressed. n_i/n_j = Number of siblings nested within number of mothers. RE = Random effects; FE = Fixed effects; OR = Odds ratio; CI = Confidence interval.

Outcomes of general health (disability pension), mental health, and substance abuse are presented in Table 4. The pattern here is more similar between the study groups, although there is a threefold difference for disability pension and statistically significant differences also for psychiatric care and alcohol-related hospital care, differences that are confirmed also in the FE analyses. It should also be noted that no outcome in Tables 3 and 4 favored the foster care group in either analysis.

Table 4.

General health, mental health and substance abuse outcomes for foster group vs. adopted group.

Outcome	RE estimates		FE estimates
	n_i/n_j	OR (95% CI)	n_i/n_j	OR (95% CI)
Disability pension 2012	353/160	3.0 (1.3–7.2)	93/38	2.6 (1.0–6.3)
Psychiatric care	362/161	1.9 (1.1–3.2)	198/83	1.8 (1.1–3.2)
Antianxiety medication	355/160	1.4 (0.8–2.3)	201/83	1.5 (0.8–2.6)
Antidepressive medication	355/160	1.6 (0.9–2.7)	198/84	1.7 (1.0–2.9)
Suicide attempt/suicide	371/161	1.7 (0.8–3.7)	93/39	2.4 (0.9–6.4)
Alcohol-related hospital care	362/161	2.9 (1.1–7.5)	72/30	3.2 (1.1–9.7)
Illicit drug problem	362/161	1.6 (0.8–3.2)	105/44	1.1 (0.5–2.4)

Sensitivity analyses

To see how unobserved confounding might change our RE and FE estimates, Table 5 reports the E-values and CI limits for each exposure−outcome association. Looking at the overall pattern from the two regression models, a crude summary would be that most OR estimates are fairly robust to unmeasured confounding: E-values range from 1.4 to 15.7. As expected, we can see that larger ORs are associated with larger E-values and are thus less sensitive for unobserved confounding than smaller ORs. Regarding the outcome “disability pension” with an observed RE odds ratio of OR = 3.0, for example, an unobserved confounder that was associated with both the outcome and the exposure by a risk ratio of 5.5-fold each, above and beyond the observed confounders, could explain away the estimate, but weaker confounding could not. The E-value for the corresponding FE estimate (OR = 2.6) was 4.6.

Table 5.

Sensitivity analyses for RE and FE logistic regression OR estimates.

Outcome	RE estimates		FE estimates
	E-value	CI limit	E-value	CI limit
Educational and social outcomes
Only compulsory education 2012	2.5	1.5	2.1	N/A
University degree 2012	1.8	1.2	1.8	N/A
Working 2012	1.8	1.4	1.8	N/A
Low income 2012	2.4	1.5	2.5	1.6
Any welfare 2012	2.3	N/A	1.8	N/A
Any crime	3.7	2.4	5.1	2.9
Serious criminality	11.1	2.8	15.7	2.8
General health, mental health, and substance abuse outcomes
Disability pension 2012	5.5	1.9	4.6	N/A
Psychiatric care	2.1	1.4	2.0	1.3
Antianxiety medication	1.6	N/A	1.7	N/A
Antidepressive medication	1.8	N/A	1.9	N/A
Suicide attempt/suicide	2.8	N/A	4.2	N/A
Alcohol-related hospital care	5.2	1.4	5.9	1.4
Illicit drug problem/any indication	2.6	N/A	1.4	N/A

Note. OR = Odds ratio; CI = Confidence interval. E-value = The minimum strength of association (on the risk ratio scale) that an unmeasured confounder need to have with both the exposure and outcome, conditional on the observed confounders, to fully explain away the exposure−outcome association. CI limit = Same as above but to drive the exposure−outcome association point estimate to be not statistically significant (i.e., lower/upper level 95% CI covers OR = 1.0). N/A = Not applicable since 95% CI overlaps OR = 1.00.

Since a large number of the OR estimates have poor precision (i.e., wide 95% CI), it is not surprising that the estimates of the CI limits were smaller and thus more sensitive to unobserved confounding. The CI limits vary from 1.2 to 2.9. To move the 95% CI for the outcome “psychiatric care” (in which the RE lower level 95% CI was 1.1) to include the null, for example, an unobserved confounder that was associated with the outcome and the exposure by a risk ratio of 1.4-fold each could do so, but weaker confounding could not.

Discussion

This study confirms the hypothesis that adoption (before adolescence) has a positive impact on developmental outcomes for children with long-term experience of foster care. Individuals who were adopted after at least 5 years in OHC before their teens had better educational and social outcomes than birth siblings with a similar OHC history, but who were never adopted. In addition, outcomes associated with health tended to point in the same direction.

Outcomes in both study groups were considerably poorer than in the general population, if we compare our results with previous national cohort studies of the same cohort (Vinnerljung, Berlin, & Hjern, 2010). Results for the fostered nonadopted group are in line with basically all previous international and Swedish research. Generally, scholars have used a host of different comparison groups with a background of childhood adversity, including maternal birth siblings raised by their mothers (Vinnerljung, 1996a) in studies of foster care outcomes. Regardless of methodology and outcome measures, basically none has found better long-term outcomes for foster children (e.g., Berzin, 2008; Buehler, Orme, Post, & Patterson, 2000; Doyle Jr, 2007, 2008). Swedish national population studies have shown that long-term outcomes for young adults raised in foster care are worse than for children growing up in the 2–3% poorest families in the country (households on long-term welfare), nearly irrespective of how outcome is measured (Vinnerljung et al., 2010).

How could these results be interpreted, in a reasonable way? From a developmental perspective, stability and security are key requisites for recovery from past adversities and well-being among children placed with substitute parents (Harden, 2004). We do not have reliable data on the stability of placements for our population, but several studies have affirmed that for stability, adoption is superior to long-term foster care (review in Christoffersen et al., 2007). Breakdowns of adoptions do happen, but in Sweden, the prevalence is miniscule compared to long-term foster care. A national population study of 16.500 international adoptees found that 3–4% had at least one time before age 18 been placed in OHC (Elmund, Lindblad, Vinnerljung, & Hjern, 2007; cp. Hoksbergen, 1991; Wijedasa & Selwyn, 2017; for a valuable review of adoption breakdown studies, see Palacios, Rolock, Selwyn, & Barbosa-Ducharne, 2018). This an imprecise indication of breakdown since some children placed in OHC probably returned to their adoptive families (50% in a similar Dutch study by Hoksbergen, 1991). In contrast, a Swedish regional study of children in long-term foster care before adolescence found that every fourth placement disrupted (with no returns to the foster family) during early teens, in average after 10 years in the same foster family (Vinnerljung, Sallnäs, & Berlin, 2017). Instability is regrettably inherent to foster care (Oosterman, Schuengel, Slot, Bullens, & Doreleijers, 2007). Rubin, O’Reilly, Luan, and Localio (2007) concluded that foster children experience placement changes/breakdowns unrelated to their baseline problems, and this instability has per se a significant impact on their behavior and well-bring. Similar findings have been reported from other studies (James, Landsverk, Slymen, & Leslie, 2004; Newton, Litrownik, & Landsverk, 2000).

The alderman of UK foster care and adoption research, John Triseliotis, found that adopted adolescents were more secure in their family compared to fostered peers (Triseliotis, 1983; Triseliotis & Hill, 1987, 1990). Almost all adopted children in his small sample studies felt that they had “a family for life” in the home where they lived, while a sizeable proportion of the foster children were painfully aware that their place in the substitute family could be terminated at will by the foster parents or the child welfare authorities. It seems reasonable to assume that the increased stability and security for the adopted siblings has had a beneficial effect on their long-term development, compared to their fostered brothers and sisters.

The trials and tribulations of young people leaving foster care (in Sweden at the 18th birthday) are well-known in research from several countries (Berzin, 2008; Biehal & Wade, 1996; Höjer & Sjöblom, 2010; Stein, 2006, 2008; Stott, 2013). Their pathway to adulthood is accelerated compared to other peers, and most are left without support from their foster families during the transition to independence. The foster homes they grew up in do not fulfil the role of a “family for life” (Triseliotis, 2002). Even though we know of no studies that have examined adult adoptees’ connections to their adoptive families in detail, we assume from plentiful anecdotal evidence that adoptive parents tend to be like most other parents in this regard, that is they provide economic, emotional, and other support over many years after their children reach age of legal majority.

In addition, poor school and educational performance have set many foster children on a pathway to economic and social marginalization (Forsman, Brännström, Vinnerljung, & Hjern, 2016). They tend to perform substantially poorer in school than peers—and adopted children—with similar cognitive competence (Vinnerljung et al., 2010; Vinnerljung & Hjern, 2011). Low grades in school and having only compulsory education has in Swedish population studies statistically explained most of the over-risks for teenage parenthood and disability pension among alumni from long-term foster care, as well as being a powerful predictor of basically all negative long-term outcomes (Brännström, Karlsson, Vinnerljung, & Hjern, 2018; Brännström, Vinnerljung, & Hjern, 2015, 2016; Vinnerljung et al., 2010, 2015). But Swedish national population studies have shown that this is not the case for adoptees (Vinnerljung et al., 2010).

Strengths and limitations

The main strength of this study is the sibling design, as far as we know the first such study in the literature to address this research question, and the use of high quality outcome indicators up to mature age with minimal attrition from the Swedish national registers.

Albeit a sibling design can be expected to limit the importance of residual confounding from familial risk factors, the situation may change in families over time, and thus a concern is the lack of historical background data about the circumstances leading to the OHC. As stated in the introduction, the main limitation concerns the limited scope of information in the national Child Welfare Register. Selection factors related to the children and within-family heterogeneity (Holmlund, 2005)—undetected by us due to restrictions of relevant data—may account for some of the differences we found.

In addition, representativity remains an issue since the sibling configurations in our study population were rare among children that grew up in foster care. Also, due to contextual factors, we do not know if the results can be extrapolated to countries outside of Scandinavia.

Conclusions and implications

The siblings who were adopted had mostly better long-term outcomes than their brothers and sisters who grew up as foster children. Regardless of the complex—and probably interacting—explanatory factors behind these findings, at least one message seems fair to send to policymakers and professionals. When foster parents want to adopt, there is no valid support from research for social workers to act against this wish, if the reasons for a negative attitude are related to concerns about the child’s long-term development. In light of the above, we propose—in line with recommendations from the Swedish National Board of Health and Welfare—a reorientation of child welfare policy and practice in Scandinavia regarding adoption from long-term foster care.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was financed by Bank of Sweden Tercentenary Foundation.

References

Allison

P. D.

(2009). Fixed effects regression models. Thousand Oaks, CA: SAGE.

Barth

R. P.

(1992). Child welfare services in the United States and Sweden: Different assumptions, laws and outcomes. International Journal of Social Welfare, 1, 36–42.

Bell

Fairbrother

Jones

(2018). Fixed and random effects models: Making an informed choice. Quality & Quantity. Advance online publication. doi:10.1007/s11135-018-0802-x

Berlin

Vinnerljung

Hjern

(2011). School performance in primary school and psychosocial problems in young adulthood among care leavers from long term foster care. Children and Youth Services Review, 33, 2489–2497.

Berzin

S. C.

(2008). Difficulties in the transition to adulthood: Using propensity scoring to understand what makes foster youth vulnerable. Social Service Review, 82, 171–196.

Biehal

Wade

(1996). Looking back, looking forward: Care leavers, families and change. Children and Youth Services Review, 18, 425–445.

Bohman

Sigvardsson

(1978). An 18-year, prospective, longitudinal study of adopted boys. In Anthony

Koupernik

Chiland

(Eds.), The child in his family: The psychologically vulernable child (Vol. 4, pp. 473–486). New York: John Wiley & Sons.

Bohman

Sigvardsson

(1980a). Negative social heritage. Adoption & Fostering, 101, 25–31.

Bohman

Sigvardsson

(1980b). A prospective, longitudinal study of children registered for adoption: A 15-year follow-up. Acta Psychiatrica Scandinavica, 61, 339–355.

10.

Bohman

Sigvardsson

(1990). Outcome in adoption: Lessons from longitudinal studies. In Brodzinsky

Schechter

(Eds.), The psychology of adoption (pp. 93–107). Oxford, UK: Oxford University Press.

11.

Bryderup

Engen

Kring

(2017). Familiepleje i Danmark. Aarhus: Klim.

12.

Brännström

Karlsson

Vinnerljung

Hjern

(2018). Childhood risk factors for disability pension among adult former Swedish child welfare clients: Same or different as for majority population peers? Children and Youth Services Review, 84, 94–102.

13.

Brännström

Vinnerljung

Hjern

(2015). Risk factors for teenage childbirths among child welfare clients: Findings from Sweden. Children and Youth Services Review, 53, 44–51.

14.

Brännström

Vinnerljung

Hjern

(2016). Child welfare clients have higher risks for teenage childbirths: Which are the major confounders? The European Journal of Public Health, 26, 592–597.

15.

Buehler

Orme

J. G.

Post

Patterson

D. A.

(2000). The long-term correlates of family foster care. Children and Youth Services Review, 22, 595–625.

16.

Cameron

A. C.

Trivedi

P. K.

(2010). Microeconometrics using Stata. College Station, TX: Stata press.

17.

Christoffersen

M. N.

(2012). A study of adopted children, their environment, and development: A systematic review. Adoption Quarterly, 15, 220–237.

18.

Christoffersen

M. N.

Hammen

Andersern

K. R.

Jeldtoft

(2007). Adoption som indsats. En systematisk gennemgang af udenlandske erfaringer. København: SFI.

19.

Cnattingius

Ericson

Gunnarskog

Källén

(1990). A quality study of a medical birth registry. Scandinavian Journal of Social Medicine, 18, 143–148.

20.

Doyle

J. J.

Jr (2007). Child protection and child outcomes: Measuring the effects of foster care. American Economic Review, 97, 1583–1610.

21.

Doyle

J. J.

Jr (2008). Child protection and adult crime: Using investigator assignment to estimate causal effects of foster care. Journal of Political Economy, 116, 746–770.

22.

Dumaret

Stewart

(1985). IQ, scholastic performance and behaviour of sibs raised in contrasting environments. Journal of Child Psychology and Psychiatry, 26, 553–580.

23.

Elmund

Lindblad

Vinnerljung

Hjern

(2007). Intercountry adoptees in out-of-home care: A national cohort study. Acta Paediatrica, 96, 437–442.

24.

Fazel

Långström

Hjern

Grann

Lichtenstein

(2009). Schizophrenia, substance abuse, and violent crime. Journal of the American Medical Association, 301, 2016–2023.

25.

Feigelman

(1997). Adopted adults: Comparisons with persons raised in conventional families. Marriage & Family Review, 25, 199–223.

26.

Forsman

Brännström

Vinnerljung

Hjern

(2016). Does poor school performance cause later psychosocial problems among children in foster care? Evidence from national longitudinal registry data. Child Abuse & Neglect, 57, 61–71.

27.

Conger

R. D.

Cadoret

R. J.

Neiderhiser

J. M.

Yates

Troughton

Stewart

M. A.

(1996). The developmental interface between nature and nurture: A mutual influence model of child antisocial behavior and parent behaviors. Developmental Psychology, 32, 574.

28.

Harden

B. J.

(2004). Safety and stability for foster children: A developmental perspective. The Future of Children, 14, 30–47.

29.

Hjern

Lindblad

Vinnerljung

(2002). Suicide, psychiatric illness, and social maladjustment in intercountry adoptees in Sweden: A cohort study. The Lancet, 360, 443–448. doi:10.1016/S0140-6736(02)09674-5

30.

Hoksbergen

(1991). Understanding and preventing “failing adoptions.” In Hibbs

(Ed.), Adoption: International perspectives (pp. 265–278). Madison, CT: International University Press.

31.

Holmlund

(2005). Estimating long-term consequences of teenage childbearing. An examination of the siblings approach. The Journal of Human Resources, 40, 716–743.

32.

Hox

J. J.

Moerbeek

van de Schoot

(2017). Multilevel analysis: Techniques and applications: London: Routledge.

33.

Höjer

Sjöblom

(2010). Young people leaving care in Sweden. Child & Family Social Work, 15, 118–127.

34.

James

Landsverk

Slymen

D. J.

Leslie

L. K.

(2004). Predictors of outpatient mental health service use—The role of foster care placement change. Mental Health Services Research, 6, 127–141.

35.

Ludvigsson

J. F.

Andersson

Ekbom

Feychting

Kim

J.-L.

Reuterwall

… Olausson

P. O.

(2011). External review and validation of the Swedish national inpatient register. BMC Public Health, 11, 450.

36.

Mathur

Ding

Riddell

VanderWeele

(2018). Website and R package for computing E-values. Epidemiology, 29, e45–e47.

37.

McSherry

Malet

M. F.

Weatherall

(2016). Comparing long-term placements for young children in care: Does placement type really matter? Children and Youth Services Review, 69, 56–66.

38.

Newton

R. R.

Litrownik

A. J.

Landsverk

J. A.

(2000). Children and youth in foster care: Disentangling the relationship between problem behaviors and number of placements. Child Abuse & Neglect, 24, 1363–1374.

39.

Oosterman

Schuengel

Slot

N. W.

Bullens

R. A.

Doreleijers

T. A.

(2007). Disruptions in foster care: A review and meta-analysis. Children and Youth Services Review, 29, 53–76.

40.

Palacios

Rolock

Selwyn

Barbosa-Ducharne

(2018). Adoption breakdown: Concept, Research and implications. Research on Social Work Practice. Advance online publication. doi:10.1177/1049731518783852

41.

Rubin

D. M.

O’Reilly

A. L.

Luan

Localio

A. R.

(2007). The impact of placement stability on behavioral well-being for children in foster care. Pediatrics, 119, 336–344.

42.

Rushton

(2003). The adoption of looked after children: A scoping review of research. London, UK: Social Care Institute for Excellence.

43.

Rushton

(2004). A scoping and scanning review of research on the adoption of children placed from public care. Clinical Child Psychology and Psychiatry, 9, 89–106.

44.

Rushton

(2007). Outcomes of adoption from public care: research and practice issues. Advances in Psychiatric Treatment, 13, 305–311.

45.

Rutter

(2006). Genes and behavior. Nature-nurture interplay explained. Oxford, UK: Blackwell.

46.

Socialstyrelsen. (2006). Vårdnadsöverflyttning för barn placerade i familjehem. Stockholm, Sweden: Socialstyrelsen.

47.

Socialstyrelsen. (2014). Nationella adoptioner av barn i familjehem. Stockholm, Sweden: Socialstyrelsen.

48.

Stein

(2006). Research review: Young people leaving care. Child & Family Social Work, 11, 273–279.

49.

Stein

(2008). Transitions from care to adulthood: Messages from research for policy and practice. In Stein

Munro

(Eds.), Young people’s transitions from care to adulthood: International research and practice (pp. 289–306). London, UK: Jessica Kingsley Publishers.

50.

Stott

(2013). Transitioning youth: Policies and outcomes. Children and Youth Services Review, 35, 218–227.

51.

Thomas

(2013). Adoption for looked after children: Messages from Research. London, UK: BAAF.

52.

Triseliotis

(1983). Identity and security in adoption and long-term fostering. Adoption & Fostering, 7, 22–31.

53.

Triseliotis

(2002). Long-term foster care or adoption? The evidence examined. Child & Family Social Work, 7, 23–33.

54.

Triseliotis

Hill

(1987). Children and adoption allowances. Adoption & Fostering, 11, 35–39.

55.

Triseliotis

Hill

(1990). Contrasting adoption, foster care, and residential rearing. In Brodzinsky

Schecter

(Eds.), The psychology of adoption (pp. 107–120). Oxford, UK: Oxford University Press.

56.

VanderWeele

T. J.

Ding

(2017). Sensitivity analysis in observational research: introducing the E-value. Annals of Internal Medicine, 167, 268–274.

57.

Vinnerljung

(1992). 235 syskon med olika uppväxtöden—en retrospektiv aktstudie. Lund, Sweden: Lunds universitet.

58.

Vinnerljung

(1996a). Fosterbarn som vuxna. Lund, Sweden: Lunds Universitet.

59.

Vinnerljung

(1996b). Svensk forskning om fosterbarnsvård: en översikt. Stockholm, Sweden: Centrum för utvärdering av socialt arbete/Liber Utbildning.

60.

Vinnerljung

Berlin

Hjern

(2010). Skolbetyg, utbildning och risker för ogynnsam utveckling hos barn. In Social rapport 2010 (pp. 227–266). Stockholm, Sweden: Socialstyrelsen.

61.

Vinnerljung

Brännström

Hjern

(2015). Disability pension among adult former child welfare clients: A Swedish national cohort study. Children and Youth Services Review, 56, 169–176.

62.

Vinnerljung

Hjern

(2011). Cognitive, educational and self-support outcomes of long-term foster care versus adoption. A Swedish national cohort study. Children and Youth Services Review, 33, 1902–1910.

63.

Vinnerljung

Sallnäs

Berlin

(2017). Placement breakdowns in long-term foster care—a regional Swedish study. Child & Family Social Work, 22, 15–25.

64.

Warman

Roberts

(2004). Adoption and looked after children: International comparisons. Oxford, UK: University of Oxford.

65.

Wijedasa

Selwyn

(2017). Examining rates and risk factors for post-order adoption disruption in England and Wales through survival analyses. Children and Youth Services Review, 83, 179–189.