By Connie Reguli

This 2018 study from Finland should be required reading for EVERY judge with the power to remove children from their homes.

If the judicial branch will not educate themselves on this then we must.

Children in adverse home situations have better outcomes when raised in their homes than when separated from their families. End of story.

Except where their is substantial risk of harm the government should not interfere with our Fourteenth Amendment right to family integrity.

Read this study and you will see that the research has been done.

America is shameless in not conducting its own research because the child welfare money is too alluring.

#banasfa #childwelfareform

Articles

Out-of-home placement in early childhood and psychiatric diagnoses and criminal convictions in young adulthood:

a population-based propensity score-matched study

Sylvana M Côté, Massimiliano Orri, Mikko Marttila, Tiina Ristikari

Summary

Background To ensure their protection and healthy development, children exposed to adverse family circumstances might be placed in foster homes, institutions, or kinship care (out-of-home placement). We aimed to compare the rates of psychiatric diagnoses and criminal convictions in young adulthood (ages 18–25 years) among children who were first placed at ages 2–6 years with those of children who were not placed and who had similar sociodemographic and family characteristics.

Method We did a population-wide cohort study using the 1987 Finnish Birth Cohort, which collects longitudinal data linking nationwide child welfare, medical, and criminal registers for all 59476 livebirths in Finland in 1987. The exposure was the first out-of-home placement at ages 2–6 years. Outcomes were rates of psychiatric diagnoses, criminal convictions, and prescriptions for psychotropic medication filled at ages 18–25 years. We matched cases to non-placed controls using propensity score matching on parental characteristics (eg, age, psychiatric diagnoses, education, family structure) and child characteristics (eg, neurodevelopmental problems, prematurity). Differences in adult outcomes between children placed and matched controls were assessed by use of logistic regression on the matched cohort.

Findings Of 54 814 individuals with complete data, 388 (1%) were first placed at ages 2–6 years; matched controls were identified for 386 of these children. At ages 18–25 years, those who had been placed as children had greater odds than never-placed controls of substance-related disorders (odds ratio 2·10, 95% CI 1·27–3·48), psychotic or bipolar disorders (3·98, 1·80–8·80), depression or anxiety (2·15, 1·46–3·18), neurodevelopmental disorders (3·59, 1·17–11·02), or other disorders (2·06, 1·25–3·39). Participants who were placed had more psychotropic medication prescriptions (1·96, 1·38–2·80) and higher rates of criminal convictions (violent offences, 2·43, 1·61–3·68; property offences, 1·86, 1·17–2·97).

Interpretation Preschool children who are placed out-of-home are at risk of adverse outcomes as adults, even accounting for their initial circumstances. It is important to explore which conditions lead to more or less favourable outcomes in child protection.

Lancet Child Adolesc Health

2018; 2: 647–53

Published Online

July 26, 2018 http://dx.doi.org/10.1016/ S2352-4642(18)30207-4

See Comment page 623

Department of Social and Preventive Medicine, University of Montreal, Montreal, QC, Canada

(S M Côté PhD); Research Centre, Sainte-Justine Hospital, Montreal, QC, Canada

(S M Côté); Bordeaux Population Health Research Centre, INSERM U1219, University of Bordeaux, Bordeaux, France (S M Côté,

M Orri PhD); McGill Group for Suicide Studies, Douglas Mental Health University Institute, Department of Psychiatry, McGill University, Montreal, QC, Canada (M Orri); and Welfare Department, National Institute for Health and Welfare, Helsinki, Finland (M Marttila MSc,

T Ristikari DsocSc)

Correspondence to:

Dr Sylvana M Côté, Research Centre, Sainte-Justine Hospital, Montreal, QC H3T 1C5, Canada sylvana.cote.1@umontreal.ca

Funding Academy of Finland.

Copyright © 2018 Elsevier Ltd. All rights reserved.

Introduction

In an attempt to help children in unfavourable circumstances when their parents are deemed unsuitable to their wellbeing, child protection agencies can remove children from their parents and place them in foster homes, institutions, or kinship care.1 In Europe, more than a million children live in foster care or in residential care institutions.2 In Finland, rates of placement doubled in one decade from 3·1% of children born in 1987 to 6% of children born in 1999. Although reasons for placement might vary across countries, the Finnish child protection system has a family support orientation, and placement is used as a last-resort measure. The most common reasons for placement are parents’ inability to care for the child because of their physical or mental illness or because of the child’s need for special care and education. Parental abuse and maltreatment are less common motives for placement.3,4 By age 18 years, most placed children have experienced

several different placements,5 and decisions for such changes are made by social workers who take into account the best interests of the child. Therefore, in early childhood, more children are placed in family care than in institutions, a type of placement that is more common in adolescence. Additionally, Finnish child protection law encourages a return to the biological family as soon as possible, and adoption is almost never used as a child protection measure.6

Studies have shown that children placed out-of- home have poorer long-term outcomes in terms of education and income, substance use, criminal convictions, and mental health, as compared with non-placed children in the general population.7–15 Children placed out-of-home usually come from low-income families, with few resources or social support,7,8 high rates of physical violence, psychological and social problems,8,9 and abuse or neglect.16,17 As these same characteristics are also risk factors for poor mental health and criminal behaviour, it is difficult

http://www.thelancet.com/child-adolescent Vol 2 September 2018

647

Articles

Research in context

Evidence before this study

We systematically searched Medline and PubMed without data and language limitations up to March 2, 2018, using the keywords “foster”, “foster care”, “placement”, “out of home placement”, “psychiatric”, “mental health”, “psych*”, “anxiety”, “anx*”, “depress*”, “substance”, “bipolar”, “criminal*”, “medication”, “psychotropic”, “population”, “population based”, “longitudinal”, “cohort”, and “register”. We selected only studies with population-based samples assessing mental health, criminal behaviour, or psychotropic medication use. We found 15 relevant studies, including a systematic review. Studies generally investigated placement across the whole of childhood and early adulthood and considered the criminal convictions, behavioural problems, and substance use as outcomes. The results of these studies suggested that placed children had worse outcomes compared with children from the general population. However, two studies that accounted for selection bias using a propensity score matching (matched for background and family variables [child age, race, income, marital status], child vocabulary, and childhood behaviours) found little evidence of an increased risk of behavioural problems among placed children after propensity

scores were applied. This finding suggests that pre-existing risk factors might account for the worse outcomes in placed children.

Added value of this study

To our knowledge, this is the first study to assess the association between placement and psychiatric diagnoses using propensity score matching to limit selection bias, and specifically assess placement in early childhood, thus limiting the heterogeneity of children exposed to placement. We showed that individuals who had out-of-home placement in early childhood (ages 2–6 years) had greater odds of psychiatric diagnoses and criminal convictions than never-placed controls, even after accounting for family and child characteristics before placement.

Implications of all the available evidence

Evidence, including psychiatric diagnoses from official records, suggests that children placed out of home are at risk of adverse outcomes as adults. The effects on psychiatric diagnoses, criminal convictions, and psychotropic medication use that we identified persisted after accounting for preplacement risk factors. It is important to explore which conditions lead to more or less favourable outcomes in child protection.

648

http://www.thelancet.com/child-adolescent Vol 2 September 2018

See Online for appendix

to establish which observed negative outcomes for placed children are attributable to placement or to those back- ground characteristics.18–20 Propensity score matching is a robust methodological design enabling comparison between placed children and children who are not placed but have similar family and sociodemographic character- istics.21 To our knowledge, no study has used this method to test the association between out-of-home placement before school entry (age 7 years in Finland) and the risk of adult psychiatric diagnoses. Children’s dependence and attach- ment to the family is fundamental in early childhood, and disrupting attachment via out-of-home placement could have a negative effect on future outcomes. Alternatively, reducing exposure to a potentially pathogenic family environment via placement might have a positive impact.

We aimed to quantify the risk of psychiatric disorders, use of psychotropic medications, and criminal convictions in adulthood (ages 18–25 years) among people placed out- of-home in early childhood (ages 2–6 years) compared with, first, people from the general population, and second, propensity score-matched individuals with similar background risks to the placed individuals. Because reasons for placement vary with age,9 and because we wanted to investigate the role of early childhood placement, we focused on placement during the preschool period.

Methods

Participants and data collection

We did a population-wide cohort study using the 1987 Finnish Birth Cohort (FBC-87), which collects longi- tudinal data linking nationwide registers for all children born in Finland in 1987 (see appendix for a full list of data

sources).22 Data were available up to the age of 25 years (ie, 2012). Data on out-of-home placement in early childhood were extracted from the Child Welfare Register of the Finnish National Institute for Health and Welfare. Children were classified as being placed if they were placed outside the home at least once, with the first time being at ages 2–6 years. In the context of this study, out-of-home placement includes voluntary care agree- ments and placements in residential, foster, and community or kinship care.

Use of the FBC-87 data was approved by the ethics committee of the Finnish National Institute for Health and Welfare (Ethical committee 28/2009) and all people with access to the data obtained the necessary permissions from each administrative registry. Data were handled in accordance with Finnish data privacy laws.

Outcomes

Psychiatric data were obtained from the Finnish Health Information System, which includes all information on inpatient and outpatient visits at public hospitals, as submitted to the Finnish Hospital Discharge Register. Coverage of registered diagnoses in hospitals can be considered to be near-complete.22 Psychiatric diagnoses were made during inpatient or outpatient visits to specialised hospital units, in accordance with the International Statistical Classification of Diseases and Related Health Problems (ICD), 10th Revision (ICD-10) codes F00–F99 and ICD 9th revision (ICD-9) codes 290–319. The categories of disorders studied (table 1) were substance-related disorders, psychotic and bipolar disorders, depression and anxiety disorders,

neurodevelopmental disorders, and other disorders. We also included a variable reflecting the number of diagnoses from each distinct class (ie, comorbidity). Prescriptions were assessed using the Anatomical Therapeutic Chemical classification codes N05–N06. Filled prescriptions were obtained from purchases as per the Social Insurance Institution of Finland’s register on reimbursed prescribed medicine. Information on criminal convictions (violent offences, property offences, and any offence) was extracted from the registry of the Finnish Legal Register Centre.

Confounders

Potential confounders that could be used for propensity score matching based on individual and family char- acteristics from before the child was aged 2 years were obtained from registers.22,23 Family variables were parental age at the birth of the index child (mother or father younger than 20 years), education (either parent did not complete high school), family structure (divorced or not divorced), death of a parent, use of social assistance (either parent received social assistance benefits at any time), and psychiatric and neurodevelopmental disorders of either parent. Child variables were child’s order of birth (ie, first-born or later), preterm birth (gestational age <37 weeks), nicotine exposure during pregnancy, and diagnosed intellectual disability or neurodevelopmental disorder at age 2 years or younger.

Statistical analysis

Propensity score variables were entered into a multiple logistic regression model estimating the likelihood of exposure to out-of-home placement (propensity score).

We then matched each case (child placed out-of-home) to a control (child not placed out-of-home) using the propensity score as a matching criterion. Propensity score matching was done using the R package matchit. We did nearest neighbour matching, without replacement, with a calliper of 0·1 times the SD of the logit-transformed propensity scores, and exact matching on sex. The chosen calliper represents the maximum permitted difference between matched subjects. Matching variables included parental use of social assistance benefits and education, young parental age, diagnosis of parental mental health problems, maternal smoking during pregnancy, divorce, and preterm birth. If no suitable matched control could be found, the case was discarded from further analysis. Covariate balance was assessed with the standardised bias before and after matching.

Differences in adult outcomes between individuals placed as children and matched controls were assessed with logistic regression on the matched cohort. Results were reported as the odds ratio (OR) and 95% CI. Complementary analyses were done using logistic regression on the entire cohort, adjusting for baseline differences between placed and not-placed individuals. The adjustment variables in regression analyses were the same as those used in propensity score matching.

The main unmeasured confounding variable in our study was child maltreatment or neglect. To test the sensitivity of the results to this unmeasured confounder, we did a sensitivity analysis based on the approach suggested by Rosenbaum.24 We relaxed the assumption of equal probab- ility of treatment assignment within the matched pairs, instead allowing the treated and control individuals to vary

Articles

Child characteristics

Parent and family characteristics

Before matching

Not placed (n=54 426)

Placed Standardised (n=388) bias

After matching

Not placed Placed Standardised (n=386) (n=386) bias

Sex

Male

27 829 (51%)

197 (51%)

–0·007

197 (51%)

197 (51%)

0·000

Female

26 597 (49%)

191 (49%)

0·007

189 (49%)

189 (49%)

0·000

First-born

32 800 (60%)

240 (62%)

0·033

249 (65%)

238 (62%)

–0·059

Preterm birth

2611 (5%)

33 (9%)

0·133

42 (11%)

32 (8%)

–0·093

Nicotine exposure during gestation

8093 (15%)

224 (58%)

0·867

208 (54%)

222 (58%)

0·073

Intellectual disability before age 2 years*

5 (0%)

0

··

0

0

··

Neurodevelopmental disorder†

74 (0%)

2 (1%)

0·053

3 (1%)

2 (1%)

–0·036

One or both parents did not finish high school

5176 (10%)

165 (43%)

0·667

172 (45%)

165 (43%)

0·022

Received social assistance benefits at any time

7114 (13%)

314 (81%)

1·725

310 (80%)

312 (81%)

0·013

Parent aged <20 years at birth of index child

1809 (3%)

56 (14%)

0·316

49 (13%)

56 (15%)

0·052

Psychiatric or neurodevelopmental disorder‡

735 (1%)

56 (14%)

0·372

51 (13%)

54 (14%)

0·030

Parent (or parents) died

135 (0%)

4 (1%)

0·077

6 (2%)

3 (1%)

–0·077

Parents divorced

546 (1%)

18 (5%)

0·173

16 (4%)

18 (5%)

0·025

All characteristics are presented as n (%). ICD-10= International Statistical Classification of Diseases and Related Health Problems, 10th Revision.*Defined as ICD-10 codes F70–F79 and ICD-9 codes 317–319. †Defined as ICD-10 codes F00–F69 and F80–F99 and ICD-9 codes 290–319. ‡Defined as ICD-10 codes F10–F99.

Table 1: Baseline characteristics

http://www.thelancet.com/child-adolescent Vol 2 September 2018

649

Articles

in their odds of treatment up to a factor of Γ. For a given value of Γ, we could then determine the possible range of observed p values obtained via a McNemar’s test, given the presence of a confounder associated strongly with both the treatment assignment and outcome.

Analyses were performed with R, version 3.4.0 (R Foundation for Statistical Computing).

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. MM had access to all the raw data and the corresponding author had final responsibility for the decision to submit for publication.

Results

Data were available for 59476 individuals up to age 25 years, of which we analysed the 54814 who had complete data for the study variables. Of these individuals, 388 (1%) were placed in out-of-home care for the first time between the ages of 2 years and 6 years (figure 1).

Before matching, individuals who were placed as children were more likely than non-placed controls to have come from families that used social assistance benefits, and have parents who were very young, less educated, smoked during pregnancy, divorced, or diagnosed with a mental health issue (table 1). Placed individuals were more likely to have been born preterm, but otherwise were not significantly different from the non-placed group in terms of sex ratio, birth order, intellectual disability, or neuro- developmental disorders before the age of 2 years.

With propensity score matching, we identified 386 pairs of placed individuals and matched control, excluding two people for whom no suitable control could be found. Of these 386 individuals, 249 (65%) were placed in foster care, while the remaining 137 (35%) were sent to institutions or some other form of placement. The length of placement varied between a few days to 16 years (mean 7·3 years, SD 5·8). After matching, differences between placed and not placed individuals were substantially reduced in all selected covariates; the standardised bias before and after matching suggested that the procedure was successful in

59 476 children born in Finland in 1987

321 excluded

199 died before age 2 years 122 placed out of home

before age 2 years

59 155 alive and never placed out-of-home at age 2 years

430 placed out-of-home at age 2–6 years

58 725 not placed out-of-home at age 2–6 years

42 had missing data

4299 had missing data

388 placed as children

54 426 not placed as children

Propensity score matching

386 placed as children

386 matched controls

Figure 1: Study population

Before placement

After placement

Any mental disorder†

Substance-related disorders‡

Psychotic and bipolar disorders§

Depression and anxiety disorders¶

Neurodevelopmental disorders||

Other disorders** Psychotropic medication use Any criminal conviction

Conviction for property offences Conviction for violent offences

Not placed (n=54 426)

6168 (11%) 1457 (3%) 1115 (2%) 4134 (8%)

434 (1%) 2713 (5%) 5713 (10%)

5700 (10%) 1701 (3%) 1673 (3%)

Placed (n=388)

129 (33%) 50 (13%) 30 (8%) 88 (23%)

14 (4%)

50 (13%) 104 (27%) 139 (36%)

81 (21%) 54 (14%)

Unadjusted OR (95% CI)

3·90 (3·15–4·82) 5·38 (3·98–7·27) 4·01 (2·75–5·84) 3·57 (2·81–4·54)

4·66 (2·71–8·01) 2·82 (2·09–3·80) 3·12 (2·49–3·92) 4·77 (3·87–5·88) 5·10 (3·81–6·82) 8·18 (6·37–10·50)

Adjusted OR* (95% CI)

2·24 (1·79–2·79) 2·19 (1·60–3·01) 2·47 (1·66–3·67) 2·09 (1·63–2·68)

3·02 (1·70–5·36) 1·74 (1·28–2·38) 1·84 (1·45–2·32) 2·05 (1·64–2·56) 1·78 (1·31–2·42) 2·58 (1·97–3·37)

Not placed (n=386)

67 (17%) 25 (6%) 8 (2%)

46 (12%)

4 (1%) 26 (7%) 61 (16%) 82 (21%) 38 (10%) 31 (8%)

Placed (n=386)

128 (33%) 49 (13%) 30 (8%) 87 (23%)

14 (4%)

50 (13%) 104 (27%) 139 (36%)

81 (21%) 54 (14%)

Matched OR (95% CI)

2·36 (1·68–3·31) 2·10 (1·27–3·48) 3·98 (1·80–8·80) 2·15 (1·46–3·18)

3·59 (1·17–11·02) 2·06 (1·25–3·39) 1·96 (1·38–2·80) 2·09 (1·51–3·87) 1·86 (1·17–2·97) 2·43 (1·61–3·68)

Data are n (%) unless specified otherwise. Odds ratios for placed vs not placed individuals were estimated with logistic regression for unmatched samples and conditional logistic regression for matched samples. OR=odds ratio. ICD-10= International Statistical Classification of Diseases and Related Health Problems, 10th Revision. *Adjusted for parental use of social assistance benefits, young parental age, low parental education, maternal smoking during pregnancy, divorce, parental diagnosis of mental health problems, and preterm birth. †Defined as ICD-10 codes F10–F99. ‡Defined as ICD-10 codes F10 and F11–19. §Defined as ICD-10 codes F20, F25, F21–F24, F28, F29, F30,

and F31. ¶Defined as ICD-10 codes F32–F34, F38, F39, F42, F40, F41 (excluding F41.2), F93, and F94; ||Defined as ICD-10 codes F70–F79, F84, F80–F83, F90, and F95. **Defined as ICD-10 codes F90.1, F91, F92, F50, F41.2, F43–F45, F48, F51 (excluding F51.3, F51.4), F52–F55, F59, F60, F62, F63, F65, F66, F68, F69, and F9.

Table 2: Associations between placement at ages 2–6 years and mental health outcomes at ages 18–25 years

650

http://www.thelancet.com/child-adolescent Vol 2 September 2018

Articles

creating a comparable control group (table 1; appendix).

Of the 386 individuals placed as children, 128 (33%) were treated in specialised inpatient or outpatient hospital units for psychiatric or neurodevelopmental disorders, com- pared with 67 (17%) of the 386 matched controls (OR 2·36, 95% CI 1·68–3·31; table 2). The risk for placed individuals was higher than for non-placed matched controls in all diagnostic categories: neurodevelopmental disorders (3·59, 1·17–11·02), psychotic and bipolar disorders (3·98, 1·80–8·80), depression and anxiety (2·15, 1·46–3·18), and substance-related disorders (2·10, 1·27–3·48). In terms of comorbidities, 61 (16%) placed individuals had one psychiatric diagnosis compared with 39 (10%) matched controls (1·93, 1·25–2·98); 38 (10%) placed individuals had two comorbid diagnoses compared with 15 (4%) matched controls (3·13, 1·69–5·82); and 29 (8%) placed individuals had three or more comorbid diagnoses compared with 13 (3%) matched controls (2·79, 1·40–5·41; figure 2). Prescriptions for psychotropic medications were filled by 104 (27%) placed individuals and 61 (16%) matched

controls (1·96, 1·38–2·80; table 2).

Among individuals placed as children, 139 (36%) of

386 had a criminal conviction between ages 18–25 years, compared with 82 (21%) of 386 matched controls (2·09, 1·51–3·87; table 2).

For most outcomes, analyses with multivariate logistic regression yielded similar results to the matched analyses (table 2). For psychotic and bipolar disorders, however, the difference between the two estimates was substantial (OR 2·47, 95% CI 1·66–3·67 for regression analyses vs 3·98, 1·80–8·80, for propensity score matching), although the CIs overlapped.

Our sensitivity analysis indicated that an unmeasured confounder that is strongly associated with psychiatric disorders (eg, maltreatment) would have to increase the odds of being placed by about 80% (ie, would unbalance the groups by a factor Γ=1·814) for the association between placement and any psychiatric disorder to be non- significant (ie, p>0·05). Additional sensitivity analyses investigating patterns of associations between length of placement and adult outcomes suggested a U-shaped association, with higher rates of problems among individuals placed for shorter durations (ie, <2 years) and those placed for longer durations (ie, between 11 years and 16 years; appendix). These descriptive statistics must be interpreted with caution because of the small or absent data points.

Discussion

In this population-wide birth cohort study using data from linked administrative databases, we found that children removed from their parents and placed in institutions or foster care at ages 2–6 years had odds of having a psychiatric diagnosis, filling prescriptions for psychotropic medications, or having a criminal record in young adulthood (18–25 years) that were twice as high compared with their non-placed counterparts. We

matched placed and non-placed individuals on the basis of propensity scores relating to family and individual characteristics before the age of 2 years, and our findings suggest that placed children have an excess risk of poor outcomes in young adulthood compared with non-placed children, over and above what might be explained by measured risk factors reflecting sociodemographic characteristics and family background.

A third of individuals placed as young children were diagnosed with a mental health disorder in adulthood, compared with about one-tenth in the entire cohort. The odds of having two or three comorbid diagnoses was three times higher in placed individuals than in matched controls. By the age of 25 years, a fifth of those who had been placed were convicted of a property offence.

Sensitivity analyses indicated that possible unmeasured confounding (such as maltreatment) would have to increase the odds of being placed by about 80% to sufficiently unbalance the groups and render the results not significant. Although it is difficult to determine whether child abuse or maltreatment could increase the odds of being placed by 80% in our cohort, this analysis suggests at least a moderate level of robustness to our

100

75

50

25

0

Not placed

No diagnosis One diagnosis

Placed

n=319

n=258

n=61

n=39

n=38

n=15

n=13

n=29

Placement

Two diagnoses

Three or more diagnoses

http://www.thelancet.com/child-adolescent Vol 2 September 2018

651

Figure 2: Number of psychiatric diagnoses at 18–25 years in placed individuals versus matched controls

Cumulative percentage of individuals with zero, one, two, or three or more psychiatric diagnoses in adulthood in placed individuals (n=386) and the matched controls (n=386).

Study population (%)

652

http://www.thelancet.com/child-adolescent Vol 2 September 2018

Articles

results. It is also important to consider that, in the Finnish context, the main reason for placement is not abuse, but some level of neglect or inability to care for the child as a result of parental poor mental health, financial difficulties, and the accumulation of problems.25 Psychiatric and financial problems were used as matching variables.

Notably, the increased risk of psychiatric diagnoses varied by disorder. The odds ratios were small to moderate (OR 2–3) for depressive, anxiety, and substance use disorders and moderate to large (OR 3–5) for neuro- developmental and psychotic and bipolar disorders. The odds ratios for criminal convictions were also small to moderate, with a greater odds ratio for violent offences than for common property offences.

Our findings are consistent with those of previous studies showing a 2–3 fold increased risk of mental health symptoms,8,10,12 substance use,9,10 criminal convic- tions,7,9 and all medication use, including use of psycho- tropic medication26 in individuals placed in out-of-home care as children, but they add to the literature in several ways. To our knowledge, this is the first study to use data from medical records for psychiatric problems and legal records for criminal offences. Given that self-reported mental health assessments are subject to bias, especially for severe psychiatric illnesses such as psychotic or bipolar disorder, this point constitutes a meaningful contribution to the field of child and adolescent health, and to mental health in particular.

We have provided estimates of the risks of placement in early childhood, as compared with previous studies that covered the entirety of childhood. Reasons for placement vary widely by age. Children placed during adolescence are more likely to have shown behavioural and mental health problems and criminal convictions before placement.10 Exploring the very young age range, together with the notion of first placement, makes it easier to disentangle the association with pre-existing risk factors from that of placement itself and also provides additional information on very young children, with which to inform decision making in child protection.

No previous study has, to our knowledge, used propensity score matching to address the effect of placement in early childhood on adult psychiatric diagnoses and criminal conviction outcomes. We have shown that almost half of the associations between placement and adult mental health could be explained by family and child characteristics before placement (41% for psychotropic medications use, 57% for criminal convictions, and 40% for any psychiatric disorders). Estimation of the putative role of placement requires an informative control group. Two studies found that the negative effect of placement on non-psychiatric outcomes disappeared when propensity matching was used.27,28 This difference from our findings might be attributable to factors such as the relatively high statistical power of our study compared with most previous studies, our reliance on conservative information from official records as outcomes (as opposed to self-reports),

differences between countries, and differences in the ability of studies to control for confounding.

Our study design did not allow us to test why placement led to unfavourable outcomes. Placement could lead to psychiatric illnesses because of the hazards of foster or institutional care, including poor quality and disrupted care (ie, change in the mode of care).29 The timing of placement—ie, during early childhood—might have disrupted attachment with parents during a sensitive developmental period. We could not test the hypothesis that disruption or timing was related to better or worse outcomes because of the small number of placed children in the different categories. However, the matching procedure and sensitivity analyses suggest that the home life characteristics that led to out-of-home placement cannot explain the pattern of results. An underlying genetic susceptibility to mental health problems could have partly explained the results. However, parental mental health diagnosis was a matching variable, which reduces the possibility that common genetic predisposition to mental health problems explains the results. Finally, the length of placement might have affected outcomes, but the small number of participants in the different placement duration categories did not allow us to conduct multivariate analyses. Bivariate descriptive analyses suggest a U-shaped association, with higher rates of problems among individuals placed for shorter durations and for longer durations.

The strengths of this study include the use of a total population-based cohort followed from birth to 25 years of age, the linkage of administrative databases of mental health diagnostic information and criminal records, and a study design that uses propensity score matching. In terms of generalisability, Finland’s child protection services, like those in most high-income countries (including the USA and Canada) rely on placement when first-line responses including family treatment and home-based service programmes are considered infeasible.13 The psychiatric diagnostic information was obtained from administrative databases and, as such, reflects service use among patients with mental health problems severe enough to warrant a diagnosis. Thus, the associations might be considered conservative because the rates of diagnosed mental health problems are lower than the true rates. Indeed, according to nationally representative Finnish surveys, the rates of mental health problems among young adults are 20–25%,30 whereas the rate of diagnoses in the present cohort is 11%. Furthermore, although propensity score matching allows us to quantify the effect of an exposure by creating comparable case and control groups, the possibility that unmeasured con- founders might explain part of the results cannot be excluded. Importantly, placed children might have had greater exposure to family situations such as abuse or neglect (which were unmeasured in this study) than their matched counterparts. As shown in sensitivity analyses, accounting for abuse or neglect is unlikely to entirely

explain the associations observed. Data on parental criminal convictions were not available in this study and could potentially explain part of the association between placement and criminal conviction outcomes. Finally, the data were gathered from administrative databases, where missingness caused by incomplete registration of events is not distinguishable from missingness due to an absence of events so it is therefore not possible to estimate the effect of missingness on the results.

In conclusion, young children exposed to out-of-home placement between the ages of 2 years and 6 years were more likely to have a psychiatric diagnosis or a criminal conviction in young adulthood than were non-placed controls with similar family and individual characteristics before placement. Given that placement is done to protect children from adverse family environments, it is important to identify conditions leading to more or less favourable outcomes. Future population-based studies could aim to distinguish between types of placement (eg, institutional care vs kinship care vs foster care), reasons for placement (behavioural problems vs family adversity),9,10,13 and durations of placement, while using robust methods. However, very large samples are needed and we were unable, using this total birth cohort, to examine variations in placement because of small sample sizes in the different categories.

Contributors

MM prepared and coded the data, and conducted the primary statistical analysis. MO contributed to the primary analysis and conducted the sensitivity analyses. All authors contributed to the study design,

data interpretation, and writing of the manuscript.

Declaration of interests

We declare no competing interests.

Acknowledgments

Data were obtained from official Finnish government records.

The Academy of Finland (grant number 308556 PSYCOHORTS and 288960 Time Trends in Child and Youth Mental Health, Service Use and Wellbeing Cohorts) provided support for the salary of MM and TR and had no role in the study. We are grateful to Danielle Buch (medical writer and editor) for critical revision and substantive editing of the manuscript.

References

1 Fernandez E, Barth RP. How Does Foster Care Work? International Evidence on Outcomes. London: Jessica Kingsley Publishers, 2010.

2 Eurochild. Children in alternative care: National surveys— 2nd edition. Brussels: Eurochild, 2010.

3 Gilbert N. Combatting child abuse: international perspectives and trends. Oxford: Oxford University Press, 1997.

4 Rauhala U. Huostaanotto [Taking into custody]. Helsinki: Suomen virallinen tilasto, Sosiaalisia erikoistutkimksia, 1978.

5 Children’s Bureau. Children’s Bureau: an office of the administration for children and families. https://www.acf.hhs.gov/cb (accessed March 31, 2018).

6 Laine S, Pösö T, Ujula T. Adoptio lastensuojelussa–lukumääristä ja ominaispiirteistä [Adoption within child protection—numbers and characteristics]. Yhteiskuntapolitiikka 2018; 83: 199–207.

7 Lindquist MJ, Santavirta T. Does placing children in foster care increase their adult criminality? Labour Econ 2014; 31: 72–83.

8 Zlotnick C, Tam TW, Soman LA. Life course outcomes on mental and physical health: the impact of foster care on adulthood.

Am J Public Health 2012; 102: 534–40.

9 von Borczyskowski A, Vinnerljung B, Hjern A. Alcohol and drug abuse among young adults who grew up in substitute care— findings from a Swedish national cohort study.

Child Youth Serv Rev 2013; 35: 1954–61.

10 Pecora PJ, White CR, Jackson LJ, Wiggins T. Mental health of current and former recipients of foster care: a review of recent studies in the USA. Child Fam Soc Work 2009; 14: 132–46.

11 Vinnerljung B, Hjern A, Lindblad F. Suicide attempts and severe psychiatric morbidity among former child welfare clients—

a national cohort study. J Child Psychol Psychiatry 2006;

47: 723–33.

12 Gypen L, Vanderfaeillie J, De Maeyer S, Belenger L, Van Holen F. Outcomes of children who grew up in foster care: systematic-review. Child Youth Serv Rev 2017; 76: 74–83.

13 Kessler RC, Pecora PJ, Williams J, et al. Effects of enhanced foster care on the long-term physical and mental health of foster care alumni. Arch Gen Psychiatry 2008; 65: 625–33.

14 Randsalu LS, Laurell L. Children in out-of-home care are at high risk of somatic, dental and mental ill health. Acta Paediatr 2018; 107: 301–06.

15 Stewart CJ, Kum HC, Barth RP, Duncan DF. Former foster youth: employment outcomes up to age 30. Child Youth Serv Rev 2014; 36: 220–29.

16 Gilbert R, Widom CS, Browne K, Fergusson D, Webb E, Janson S. Burden and consequences of child maltreatment in high-income countries. Lancet 2009; 373: 68–81.

17 Turney K, Wildeman C. Mental and physical health of children in foster care. Pediatrics 2016; 138: e20161118.

18 Lund C, Silva MD, Plagerson S, et al. Poverty and mental disorders: breaking the cycle in low-income and middle-income countries. Lancet 2011; 378: 1502–14.

19 Sareen J, Afifi TO, McMillan KA, Asmundson GJG. Relationship between household income and mental disorders: findings from a population-based longitudinal study. Arch Gen Psychiatry 2011;

68: 419–27.

20 Whittle S, Vijayakumar N, Simmons JG, et al. Role of positive parenting in the association between neighborhood social disadvantage and brain development across adolescence. JAMA Psychiatry 2017; 74: 824–32.

21 Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivar Behav Res 2011; 46: 399–424.

22 Paananen R, Gissler M. Cohort Profile: The 1987 Finnish Birth Cohort. Int J Epidemiol 2012; 41: 941–5.

23 Gissler J, Haukka J. Finnish health and social welfare registers in epidemiological research. Nor Epidemiol 2009; 14: 113–20.

24 Rosenbaum PR. Observational Studies, 2nd edn. New York: Springer-Verlag, 2002.

25 Heino T, Hyry S, Ikäheimo S, Kuronen M, Rajala R. The reasons for child placement, background, services and costs. Main findings from HuosTa-project. Report: 3/2016. Helsinki: National Institute for Health and Welfare—THL, 2016.

26 Gardner H, Corsi S, Nelson M, Schaneman J, Pendleton P, Simison B. Polypharmacy and pregnancy among foster children. Child Welfare 2016; 95: 101–16.

27 Berger LM, Bruch SK, Johnson EI, James S, Rubin D. Estimating the ‘impact’ of out-of-home placement on child well-being: approaching the problem of selection bias. Child Dev 2009; 80: 1856–76.

28 Berzin SC. Difficulties in the transition to adulthood:

using propensity scoring to understand what makes foster youth vulnerable. Soc Serv Rev 2008; 82: 171–96.

29 Humphreys KL, Gleason MM, Drury SS, et al. Effects of institutional rearing and foster care on psychopathology at age

12 years in Romania: follow-up of an open, randomised controlled trial. Lancet Psychiatry 2015; 2: 625–34.

30 Marttunen M, Kaltiala-Heino R. Nuorisopsykiatria. In: Lönnqvist, J, Henriksson, M, Marttunen, M, et al, eds. Psykiatria [Psychiatry], 9th edn. Helsinki: Kustannus Oy Duodecim, 2011.

http://www.thelancet.com/child-adolescent Vol 2 September 2018

653

Articles