Samples

We combined data from three surveys of looked after British children and one survey of British children in private households; all these surveys were nationally representative and used the same instruments to assess psychopathology and some of its correlates (Meltzer et al, Reference Meltzer, Gatward and Goodman2000, Reference Meltzer, Gatward and Corbin2003, Reference Meltzer, Lader and Corbin2004a ,Reference Meltzer, Lader and Corbin b ).

Children looked after by local authorities

Random samples of children (aged 5–17 years) looked after by local authorities were selected from the relevant databases in England, Scotland and Wales (Fig. 1). A contact person in each administrative area was sent child summary forms for each child selected from that area. After obtaining whatever consent the local authority deemed necessary, the contact person was responsible for ensuring that the child's social worker completed the information on the child summary forms and for returning the completed forms to the Office for National Statistics.

Fig. 1 Diagram illustrating the sample selection and response rate from the surveys of children looked after by local authorities in England, Wales and Scotland, listing the five main reasons for ineligibility in each country.

Similar proportions of children from each country were deemed ineligible, and response rates among eligible children were very high. As previous analysis demonstrated that the prevalence rates of psychiatric disorders were not significantly different between the three countries (Meltzer et al, Reference Meltzer, Lader and Corbin2004a ,Reference Meltzer, Lader and Corbin b ), and we were not trying to establish an accurate measurement of prevalence but were primarily interested in the association of disorder with correlates, we have combined data from all three surveys without weights. Analyses comparing the children looked after by local authorities with children living in private households excluded looked after children aged 16–17 years (n=290) to fit with the age range in the private household survey.

Measures

We used the Development and Well-Being Assessment (DAWBA; Reference Goodman, Ford and RichardsGoodman et al, 2000) to assess psychiatric disorder in all four surveys. This structured interview was administered by lay interviewers to the parents or carers of all children, and also to the children themselves if they were at least 11 years old, and the interviewers also recorded detailed verbatim descriptions of any problem areas. An abbreviated version was sent to the child's teacher. A small team of experienced clinicians used the information provided by all the informants, combining information as they would in the clinic, to make diagnoses according to ICD–10 criteria (World Health Organization, 1993). In the validation study, the DAWBA provided excellent discrimination between community and clinical samples (Reference Goodman, Ford and RichardsGoodman et al, 2000). Within the community sample, children with DAWBA diagnoses differed markedly from those without such a diagnosis in external characteristics and prognosis, whereas there were high levels of agreement between the DAWBA and case notes among the clinical sample (Kendall's tau b=0.47–0.70).

Parents, teachers and children aged 11 years or over also completed the Strengths and Difficulties Questionnaire (SDQ; Reference GoodmanGoodman, 2001) a well-validated measure of common childhood psychopathology comprising the following scales: total difficulties, emotional symptoms, conduct problems, hyperactivity, peer problems, prosocial behaviour and impact. Parents or carers were asked whether their child suffered from a list of conditions, and the derived variable ‘neurodevelopmental disorder’ included children reported to have cerebral palsy, epilepsy, muscle disease or weakness, or coordination problems. Teachers provided data on attainment, an estimate of mental age and reported whether the child had a Statement of Special Educational Needs. Social workers and carers of the children looked after by local authorities provided data on the child's care history.

Analysis

The analysis was conducted using Stata 8 and the Statistical Package for the Social Sciences (SPSS) version 12.01 for Windows. We tested differences in socio-demographic characteristics, diagnosis and placement history according to type of placement using chi-squared tests for categorical variables and one-way analysis of variance for continuous variables, in the whole population of children looked after by local authorities (n=1543). We classified children living in private households as disadvantaged if their parents had either never worked or worked in unskilled occupations. We tested differences between the three groups using logistic regression to adjust for age and gender, because the children looked after by local authorities were significantly older and more likely to be boys (see Table 2) and the prevalence of childhood psychiatric disorder varied by both these characteristics (Meltzer et al, Reference Meltzer, Gatward and Goodman2000, Reference Meltzer, Gatward and Corbin2003). As psychological adjustment is more than the presence or absence of a psychiatric disorder, we used the bandings for the carer-completed SDQ to identify particularly well-adjusted children on the basis of scores within the normal range for all of the sub-scales. Finally, we explored the relationship between potential correlates of psychiatric disorder among children looked after by local authorities and the private household survey using logistic regression and used general linear modelling to examine the fit of the data to multiplicative and additive models.

Table 1 Socio-demographic characteristics, care history and psychiatric disorder in British children (age 5–17 years) looked after by local authorities, according to type of placement

Characteristic (number of children with data)	Percentage of children in placement					Percentage of all children looked after by local authorities (n=1543)	P 1
	Foster care (n=839)	Living with natural parents (n=206)	Kinship care (n=168)	Residential care (n=279)	Living independently (n=51)
Gender (n=1543)
Male	57.4	56.3	45.8	65.9	51.0	57.4	0.001
Female	42.6	43.7	54.2	34.1	49.0	42.6
Country (n=1543)
England	74.0	52.9	54.8	63.8	76.5	67.3	<0.001
Scotland	15.9	38.8	29.8	29.4	19.6	23.0
Wales	10.1	8.3	15.5	6.8	3.9	9.7
Ethnic group2 (n=1529)
White	89.5	96.6	92.9	90.7	80.4	90.7	0.008
Black	5.4	1.5	3.6	4.3	5.9	4.5
Other	5.1	1.9	3.6	5.0	13.7	4.8
Age when first looked after (n=1260)
<5	31.2	32.7	38.8	16.5	7.9	29.0	<0.001
5-10	49.6	41.5	47.5	43.6	13.2	46.2
11-15	18.6	25.8	13.7	39.0	68.4	24.0
16-17	0.7	0	0	0.9	10.5	0.9
Years in current placement (n=1411)
<1	31.7	37.3	25.6	50.9	100	35.3	<0.001
1-1.9	22.6	22.2	15.5	24.0	0	22.0
2-3.9	22.6	23.8	25.6	17.9	0	22.1
≥4	23.2	16.7	33.3	7.2	0	20.6
Placement changes last year (n=1417)
0	28.1	25.0	30.9	16.3	12.5	25.5	<0.001
1	47.4	50.0	57.9	39.3	32.5	47.0
2	17.3	18.3	7.2	25.0	25.0	17.9
≥3	7.2	6.7	3.9	19.4	30.0	9.6
Psychiatric disorder3 (n=1543)
Yes	38.6	48.1	31.5	71.0	49.0	45.3	<0.001
No	61.4	51.9	68.5	29.0	51.0	54.7
Emotional disorder (n=1543)
Yes	9.7	18.0	7.7	18.6	17.6	12.4	<0.001
No	90.3	82.0	92.3	81.4	82.4	87.6
Conduct disorder (n=1543)
Yes	32.3	35.9	26.8	61.3	39.2	37.7	<0.001
No	67.7	64.1	73.2	38.7	60.8	62.3
Hyperkinetic disorder (n=1543)
Yes	8.5	9.2	6.0	10.0	2.0	8.4	0.26
No	91.5	90.8	94.0	90.0	98.0	91.6

Table 2 Comparison of children looked after by local authorities with children living in private households

		Prevalence (%)
	OR (95% CI)1	Children looked after by local authorities (n=1253)	Disadvantaged private household sample (n=761)	Remaining children from the private household sample (n=9677)
Socio-demographic variables
Male		57.1***2,3	45.9*3	50.3
Older (11-15 v. 5-10)		59.0***2,3	41.9	43.5
White (v. Black or minority ethnic)		91.6	89.8	91.5
Neurodevelopmental disorder		12.8***2,3	4.5	3.3
Statement of special educational needs		23.0***2,3	4.5**3	2.9
Carer report of learning difficulties		36.9***2,3	12.2***3	8.3
Literacy or numeracy problems		34.3***2,3	20.4***3	10.4
Mental age 60% or less of chronological age		10.7***2,3	1.5	1.3
ICD—10 psychiatric diagnosis
At least one diagnosis	4.92 (4.13-5.85)	46.4***2,3	14.6***3	8.5
Any anxiety disorder	2.09 (1.58-2.76)	11.1***2,3	5.5**3	3.6
Separation anxiety	1.92 (1.05-3.51)	2.0***3	1.7**3	0.7
Specific phobia		1.1	1.1	1.0
Social phobia		0.7	0.5	0.3
Panic disorder		0	0.1	0.1
Post-traumatic stress disorder	11.76 (4.98-27.76)	1.9*2,***3	0.5*3	0.1
Obsessive—compulsive disorder		0.2	0.5	0.2
Generalised anxiety disorder	1.61 (0.78-3.32)	1.9***3	0.9	0.6
Other anxiety disorder	2.86 (1.86-4.39)	4.4**2,***3	1.4	1.2
Depression	2.28 (1.34-3.88)	3.4*2,***3	1.2	0.9
Hyperkinesis	3.90 (2.80-5.42)	8.7***2,3	1.3	1.1
Any behavioural disorder	7.53 (6.21-9.14)	38.9***2,3	9.7***3	4.3
Oppositional defiant	3.60 (1.50-2.66)	12.2***2,3	4.5***3	2.4
Conduct disorder4	9.34 (7.26-12.03)	26.7***2,3	5.3***3	1.9
Autistic-spectrum disorder	1.38 (0.68-2.81)	2.6**2,***3	0.1	0.3
Other disorders5	1.22 (0.33-4.33)	0.6*2	0.8**3	0.2

Children looked after by local authorities

Children looked after by local authorities from different types of placement differed significantly in relation to socio-demographic characteristics (Table 1), with a higher proportion of boys in residential care, and of girls and children from Black and minority ethnic groups living independently. Black and minority ethnic children and children from England and Wales were the least likely to be living with their natural parents. Younger children tended to live with kinship carers (mean age 10.5 years 95% CI 9.9–11.0), foster carers (11.8 years, 95% CI 11.5–12.0) or their natural parents (11.2 years, 95% CI 10.7–11.7), whereas adolescents were more likely to be living in residential settings (13.6 years, 95% CI 13.3–13.9) or independently (16.6 years, 95% CI 16.4–16.8); F=55.4, P<0.001. Placement type also varied by care history, with children who entered the care system in adolescence and more recently most likely to be living in residential care or independently, in contrast to children who were taken into care at a very early age or had been in care for several years, who were more likely to be living in a family placement.

There were also significant differences in the prevalence of psychiatric disorder according to the type of placement, with high rates of emotional and conduct disorder among children in residential placements or living independently. Hyperkinetic disorder showed a similar trend and the failure to detect a significant association may relate to lack of power and/or the older age of children living in this setting, who would be less likely to have these symptoms. Children with psychiatric disorder entered the care system later (mean age 7.7 years, 95% CI 6.7–7.3, v. 7.0 years, 95% CI 7.3–8.0; F=8.6, P=0.003), reported more changes in placement within the past 12 months (1.4 changes, 95% CI 1.3–1.5, in those with a psychiatric disorder v. 1.0 changes, 95% CI 0.9–1.1, in those without, F=39.4, P<0.001), and had lived for less time in their current placement (2.4 years, 95% CI 2.3–2.6 v. 3.0 years, 95% CI 2.8–3.1, F=30.4, P<0.003). However, missing data about all three care-related variables varied systematically with the type of placement, in that children living with their natural parents or independently were particularly likely to have missing data (age first looked after χ²=10.4, d.f.=4, P=0.03; changes of placement within the past year χ²=879, d.f.=4, P<0.001; duration of current placement χ²=27.0, d.f.=4, P<0.001). Children with psychiatric disorder were also overrepresented among those with missing data (χ²=8.9, d.f.=1, P=0.003).

Comparison of the two groups

Children looked after by local authorities had a higher prevalence of educational and neurodevelopmental difficulties than the disadvantaged and non-disadvantaged children living in private households (Table 2). After adjusting for age and gender, the prevalence of most psychiatric disorders was also significantly higher, whether the comparison group was all children from private households or just the most disadvantaged children from private households. For most disorders, the highest prevalence was for the children looked after by local authorities and the lowest prevalence was for the non-disadvantaged children from the private households; disadvantaged children from private households generally had intermediate rates.

The proportion of children without a psychiatric disorder who were particularly well adjusted in terms of scoring in the normal range on all six SDQ sub-scales was 9% (58 of 670) among the looked-after sample as opposed to 41% (265 of 649) among the disadvantaged children and 53% (4619 of 8733) among the rest of the private household sample (χ²=506, d.f.=2, P<0.001).

Complete data on correlates were available for multivariate analysis on 72% of children, but children with a psychiatric disorder (χ²=14.6, d.f.=1, P<0.001) and children who were looked after (χ²=32.7, d.f.=1, P<0.001) were more likely to have missing data on correlates. The odds ratios in Table 2 indicate that looked after status was an independent correlate of all the psychiatric disorders that were more common in this group with the exception of autistic-spectrum disorders and generalised anxiety disorder, even after adjusting for other potential correlates. Looked after status had the strongest association with disorders in which environmental factors are believed to have a leading role, such as post-traumatic stress disorder and conduct disorder. Literacy and numeracy problems were correlates of all disorders except depression among children who were looked after, whereas global learning disability was related only to pervasive developmental disorders and generalised anxiety disorder (more details of this analysis are available from the authors). The relationship of age and gender varied with the different types of psychiatric disorder in the looked after children, as one would predict from epidemiological findings in children living in private households. Thus, older children were more likely to have generalised and other anxiety disorders, post-traumatic stress disorder, depression and conduct disorder, whereas younger children were more likely to have oppositional defiant disorder, hyperkinetic disorder and separation anxiety disorder. Girls were more likely to have post-traumatic stress disorder; boys were more likely to be diagnosed with hyperkinetic disorder, and conduct or oppositional defiant disorder.

Table 3 shows how the influence of looked after status interacted with other correlates in relation to the presence of psychiatric disorder. Looked after status interacted with learning difficulties in an additive way, with learning difficulties resulting in a similar percentage increase in the prevalence of psychiatric disorder in children who were and were not looked after. In contrast, looked after status interacted with gender and age in a multiplicative way, with similar odds ratios in children who were and were not looked after. The findings for neurodevelopmental difficulties were intermediate, with an interaction that was more than additive but less than multiplicative.

Table 3. Comparison of the impact of four risk factors on rates of psychiatric disorder in children looked after by local authorities and children in private households.

	Prevalence of any ICD—10 psychiatric diagnosis		Interaction
	Looked after	Private household	Additive model	Multiplicative model
Gender
Male, %	53.7	10.7
Female, %	36.6	7.1
Difference (male—female), %	17.1	3.6	P<0.001
Odds ratio (male:female) (95% CI)	2.01 (1.60-2.52)	1.56 (1.36-1.79)		P=0.07
Age
11-15 years, %	50.1	10.5
5-10 years, %	41.1	7.6
Difference (older—younger), %	9.0	2.9	P=0.0001
Odds ratio (older:younger) (95%CI)	1.43 (1.15-1.81)	1.42 (1.24-1.63)		P=0.9
Neurodevelopmental difficulties
Yes, %	58.4	31.8
No, %	41.6	8.1
Difference (yes—no), %	16.8	23.7	P=0.002
Odds ratio (yes:no) (95% CI)	1.74 (1.25-2.44)	5.29 (4.18-6.70)		P<0.001
Carer-reported learning difficulties
Yes, %	61.7	29.2
No, %	37.4	7.1
Difference (yes—no), %	24.3	22.1	P=0.3
Odds ratio (yes:no) (95% CI)	2.69 (2.12-3.41)	5.41 (4.58-6.39)		P<0.001

Strengths and limitations

The strengths of this study include data drawn from large population-based samples that used the same methodology and measures administered by the same team of researchers to nationally representative groups of children who were and were not looked after. The sample of children who were looked after included children living in all kinds of placement types, rather than just focusing on foster care, as found in other studies (Reference Stein, Evans and MazumberStein et al, 1996; Reference PhillipsPhillips, 1997; Reference Quinton, Rushton and DanceQuinton et al, 1998; Reference Leslie, Landsverk and LoftstromLeslie et al, 2000; Reference Farmer, Burns and ChapmanFarmer et al, 2001), and did not include children in contact with social services for other reasons (Reference Garland, Hough and McCabeGarland et al, 2001; Reference Burns, Phillips and WagnerBurns et al, 2004).

Unfortunately there were few shared potential correlates between the children looked after by local authorities and those living with private households, limiting our comparison. Such lack of access to historical information is one of the common difficulties of working with and studying children looked after by local authorities, and also makes it difficult to draw conclusions about how far looked after status contributes directly to poor mental health. Children looked after by local authorities had higher levels of educational disadvantage than deprived children living in private households, so the relationship of looked after status to psychiatric disorder might be due to confounding by other aspects of social adversity that we were not able to control for. The fact that children living in residential care were more likely both to have a psychiatric disorder and to have had multiple placements within the past year indicates the problem of studying these factors in cross-sectional studies. Are children with psychiatric disorders more likely than other children to suffer multiple breakdowns in placement and end up in residential placements, or do multiple placements and/or communal living precipitate psychiatric disorder? Although psychological difficulties may be the result of placement instability, it is not unusual for children who are looked after to be referred to mental health services with an undetected psychiatric disorder (Reference Rubin, Alessandrini and FeudtnerRubin et al, 2004). We suspect that both processes are at work, but prospective longitudinal studies are needed to assess their individual impact. In addition, both the care-related variables and educational disadvantage may be markers of abuse, trauma or attachment difficulties that might explain both the increased prevalence of psychiatric disorder and the poor educational attainment and care history in these children.

Between a third and a half of the original random samples were deemed ineligible, meaning that despite a high response rate respondents might not be representative of children looked after by local authorities. Children undergoing adoption or returned to their parents might be expected to have a lower prevalence of psychiatric disorder, whereas local authority or carer refusal to grant access may be an indicator of poor mental health, making it difficult to estimate how our findings might be influenced by our difficulty in accessing the children looked after by local authorities. Missing data on care-related variables for the looked after children varied systematically by the type of placement, with more missing data on young people living independently or with their natural parents, making the results less reliable in these groups. The disproportionate loss of data on possible correlates among looked after children with psychiatric disorders reduced our power to detect positive associations, but we can be confident of the associations that we have detected. Despite these limitations, our study is one of the largest and most systematic studies of children looked after by local authorities carried out to date.

Clinical and policy implications

Concerns about the unmet needs and poor outcomes of children in the care system in America led to the development of ‘treatment foster care’ and the increased use of kinship care (Reference Rosenfeld, Pilowsky and FineRosenfeld et al, 1997; Reference ChamberlainChamberlain, 2003). Similar initiatives in Iceland, Norway, Slovenia and the UK aim to minimise the number of children in institutional care, but alternatives to institutional care are underutilised in much of the rest of Europe (Reference Browne, Hamilton-Giachritsis and JohnsonBrowne et al, 2006). In Britain, children looked after by local authorities are recognised by the children's National Service Framework (Department of Health, 2004) and Every Child Matters (Chief Secretary to the Treasury, 2003) as a group who are particularly vulnerable to psychological difficulties and are often denied access to services, leading to the development of dedicated mental health teams.

Our findings underline the need for services to ensure that the emotional and behavioural difficulties of children looked after by local authorities are understood by professionals working with these children. In some cases, the diagnosis of a psychiatric disorder may provide access to evidence-based treatments and reduce the chance of a placement breaking down. In other cases it might be more appropriate to focus on changing the care or educational environment rather than labelling affected children as psychiatrically disordered. Specialist mental health services need to support other professionals working in this area to minimise the impact of being looked after and to allow a greater proportion of these children to fulfil their potential as adults. This study shows that residential social workers are dealing with many children with serious psychiatric disorders, and yet many have little training or support for the identification and management of these difficulties. Evaluations of treatment foster care suggest that foster carers and social workers could also benefit from this kind of input (Reference ChamberlainChamberlain, 2003). Given the high levels of educational disadvantage among children looked after by local authorities, and given that carer-reported learning difficulties were frequently an independent predictor of psychiatric disorder, professionals working with this population should try to ensure that these children are provided with suitable school placements and adequate additional support where necessary.

Future research

Our findings suggest that fewer than one in ten of the children looked after by local authorities had positively good mental health and that their substantially increased prevalence of psychiatric disorder was at least partially explained because they had also experienced particularly high levels of psychosocial and educational adversity. However, there was also a strong association between psychiatric disorder and care-related variables. Longitudinal studies of the mental health of children looked after by local authorities are required to tease apart the causal relationship of care-related variables, early physical and psychosocial adversity, and constitutional factors in the child.