BACKGROUND: Looked-after children in local authority care are among the most disadvantaged, and measures of their well-being, including educational outcomes, are poorer than other children’s.
METHODS: The study sample consisted of all children in England born in academic years 1993 to 1994 through 1997 to 1998 who were in local authority care at any point during the academic years 2005 to 2006 through 2012 to 2013 and for whom results of national tests in literacy and numeracy were available at ages 7, 11, and 16 (N = 47 500).
RESULTS: Group trajectory analysis of children’s educational progress identified 5 trajectory groups: low achievement, late improvement, late decline, predominant, and high achievement. Being looked after earlier was associated with a higher probability of following a high achievement trajectory and a lower probability of following a late decline trajectory. For children first looked after between ages 7 and 16, having a longer total time looked after by age 16 was associated with a higher probability of following a high achievement trajectory. For children with poor outcomes at ages 7 and 11, being looked after by age 16 was associated with an increased chance of educational improvement by age 16.
CONCLUSIONS: This study provides evidence that early entry into care can reduce the risk of poor educational outcomes. It also establishes group trajectory analysis as an effective method for analyzing the educational progress of looked-after children, with the particular strength that it allows factors associated with a late decline or improvement in educational progress to be identified.
- BIC —
- Bayesian information criterion
- GCSE —
- General Certificate of Secondary Education
- GTA —
- group trajectory analysis
- KS1 —
- key stage 1 (school years 1–2)
- KS2 —
- key stage 2 (school years 3–6)
- KS3 —
- key stage 3 (school years 7–9)
- KS4 —
- key stage 4 (school years 10–11)
- LAC —
- looked-after children
What’s Known on This Subject:
Research shows that looked-after children tend to have lower educational attainment than other children; this can largely be attributed to their high levels of need. There is evidence that being looked after reduces children’s level of educational disadvantage.
What This Study Adds:
Applying group trajectory analysis to the educational progress of looked-after children facilitates the identification of factors associated with low and high achievement educational paths, including factors associated with a late decline or a late improvement in children’s educational progress.
Looked-after children (LAC) are children for whom the state is responsible as a proxy parent. Although the majority of LAC are fostered, some are placed in residential care, adopted, or supported to live independently, depending on the children’s age and level of need. They are some of the most vulnerable and disadvantaged children, and their health outcomes are poor worldwide.1–5 Educational outcomes are inexorably linked to health outcomes and tend to be poorer for LAC than for other children.6–8 However, it has been argued that the educational disadvantage of LAC can mostly be attributed to the difficulties that led to these children being looked after rather than to the effects of government care.7 Indeed, being looked after may protect children from some of the educational disadvantage that they would otherwise experience.6,9
LAC strategy in England is determined by 300 local authorities that compulsorily report annual data to the government.10–12 In this national study of children who were in the care of the state at any point during the academic year of 2005 to 2006 through the academic year of 2012 to 2013, their educational achievement in literacy and numeracy was compared with that of all children in England. Their educational progress was investigated by using group trajectory analysis (GTA), which is a versatile method for identifying groups of individuals showing similar paths for an outcome over time. Originally developed for use in criminology,13 GTA is now used in many other fields,14 including studying educational progress.15
GTA was applied to the educational progress of LAC in literacy and numeracy from age 7 (key stage 1 [KS1]) to age 16 (key stage 4 [KS4]), when pupils take the General Certificate of Secondary Education (GCSE) examinations. Without previous assumptions, the method identifies the trajectory groups that emerge from the data; the demographic and care-related factors associated with membership of each trajectory group are then determined.
To analyze educational progress, we considered the level that children achieved in literacy and numeracy in the tests taken at the end of KS1, key stage 2 (KS2), and KS4.16 Key stage 3 (KS3) tests were dropped in 2010 and were omitted from the analysis. The data were derived from local authority returns on LAC for the academic year of 2005 to 2006 through the academic year of 2012 to 2013. The total number of children in local authority care at any point during these years was 171 097. Educational results were extracted from the National Pupil Database and linked to the annual care returns by using children’s unique pupil number.17
Given the years of data collection, only children born in the academic year of 1991 to 1992 through the academic year of 1997 to 1998 could have results for KS1, KS2, and KS4 tests (N = 80 476); 61 405 did so (76.3%). There was a step change in the rate of entry for LAC for the GCSE examinations taken at the end of KS4 between those born in the academic year of 1991 to 1992 through the academic year of 1992 to 1993 and those born in the academic year of 1993 to 1994 through the academic year of 1997 to 1998. We analyzed the latter group (N = 47 543). Unaccompanied asylum seekers were excluded from the sample (n = 43). The final sample consisted of 47 500 children.
Literacy was identified as reading level at KS1, English level at KS2, and English GCSE result at KS4. Numeracy was identified as mathematics level at KS1 and KS2 and mathematics GCSE result at KS4. To place children’s KS4 English and mathematics GCSE results on a commensurate scale to the KS1 and KS2 test results, approximate equivalences were used: “GCSE passed at grade A* to C” = level 8, “GCSE passed at grade D to G” = level 5, “GCSE entered but failed” = level 2, and “Not entered for GCSE” = level 0.18
Results were obtained for literacy, numeracy, and overall educational attainment, which was defined as the mean of the levels achieved in literacy and numeracy. Because the results were similar for all 3 outcomes, we report results for overall educational attainment only.
The following covariates were included in the analysis: child’s sex, ethnic group, first language, age in school year, academic year of birth, special educational needs, category of need, placement type (foster care, adoption, independent living, residential), whether the child had had more than 1 placement type by the end of KS4, and whether the child had spent ≥1 period in care by the end of KS4.
When children are placed in local authority care, they are recorded as having 1 of 8 categories of need. A small proportion of children (5.7%) had more than 1 category of need recorded during their time being looked after; for these children the earliest recorded category of need was used. The category of need, "low income" (n = 135), was merged with "family in acute stress." The child’s placement type was the last placement recorded before the end of KS4. Placement type “independent living” includes placements with the child’s own parents or guardian, placement in an independent accommodation (including staying with friends or in “bed and breakfast” accommodation), and residential employment. The placement type “other” (n = 278) was merged with "independent living."
The total number of looked-after periods and the total time looked after were calculated from the annual returns. A period in care may consist of ≥1 placement; the number of individual placements was not recorded in the annual return. The age and educational stage that a child was first looked after were found by subtracting the length of the first recorded period in care from the date of the corresponding record. Only periods in care that extended into the reporting years of 2005 to 2006 through 2012 to 2013 were recorded in the data.
Results for LAC at KS1, KS2, and GCSE were compared with those for all children in England.19–21 Educational progress was analyzed by using the GTA method developed by Nagin13 and as implemented in the Stata “traj” plugin.22,23 In the GTA modeling procedure, the trajectory groups are derived from the outcome variable data. The covariates associated with membership of the trajectory groups are then found through multivariate logistic regression analysis. In the first model (model 1), the principal covariate of interest was the educational stage when children were first looked after: during preschool, or KS1, KS2, KS3, or KS4. Children who were first looked after later than the end of KS4 were used as a reference group.
Model selection was conducted by maximizing the Bayesian information criterion (BIC)13,24: (1) the number of trajectory groups was increased from 2 until the BIC was maximized; (2) all trajectories were initially assumed to be quadratic, and we tested whether modeling each group trajectory as linear improved the BIC; (3) finally, the potential covariates were added successively and retained if adding the covariate increased the BIC.
In model 2, a refinement of the first model, children first looked after at each educational stage were divided into 3 groups by using tertiles of the total length of time they had spent in care by the end of KS4.
Demographic and care history variables are summarized in Table 1. There was no missing covariate data.
GTA identified 5 trajectory groups; (1) low achievement (15.9%), (2) late improvement (7.0%), (3) late decline (20.5%), (4) predominant (43.3%), and (5) high achievement (13.3%); see Fig 1. All the trajectories were found to be quadratic except the low achievement group, which was linear. A breakdown of children by educational stage at which they were first looked after and trajectory group membership is given in Table 4. The results of model 1 are given in Table 5. A breakdown of children by educational stage at which they were first looked after and tertiles of total time looked after by the end of KS4 is given in Table 6. The results of model 2 are given in Table 7.
All the potential model covariates were included in the models except for the child’s first language. Model coefficients give the probability of trajectory group membership as odds ratios relative to a reference group. The predominant group is used as the reference group for the low achievement, high achievement, and late decline groups. The late improvement group followed a similar trajectory to the low achievement group during KS1 and KS2, then showed a dramatic improvement in results by KS4 when GCSEs are taken. The question of interest here is, “What factors are related to whether children with poor achievement at KS1 and KS2 improve by the end of KS4?” To answer this question, the low achievement group was used as the reference group for the late improvement group.
Model 1: Educational Stage First Looked After
See Table 5 and Fig 2. The probability of belonging to the high achievement group was similar for children first looked after between preschool and the end of KS2, then declined for children first looked after during KS3 and KS4. Children first looked after at any educational stage up to the end of KS4 were more likely to follow a high achievement trajectory than children first looked after later than the end of KS4.
The probability of a late decline trajectory followed a complementary pattern: it was similar for children first looked after between preschool and the end of KS2, then rose for children first looked after during KS3 and KS4. Children first looked after at any stage up to the end of KS4 were less likely to follow a late decline trajectory than children first looked after later than the end of KS4.
The probability of a late improvement trajectory for children who had had a poor educational start was higher for all children first looked after during KS1 to KS4 than for children first looked after later than the end of KS4.
The probability of following a low achievement trajectory was lower for all children first looked after before the end of KS4 than for children first looked after later than the end of KS4. Children first looked after during KS1 had the lowest risk of following this trajectory.
Model 2: Total Time in Care
See Table 7 and Fig 3. For children first looked after during KS2 to KS4, those in the highest tertile of time looked after by the end of KS4 were significantly more likely to follow a high achievement trajectory than those in the lowest tertile. That is, a longer time in care was significantly associated with having the most positive outcome.
For children first looked after during KS3, a longer time in care was significantly associated with a lower probability of following a late decline educational trajectory.
The length of time in care by the end of KS4 did not significantly affect the probability of a late improvement within each educational stage first looked after.
For children first looked after during KS2, a longer time in care by the end of KS4 was significantly associated with a lower probability of following a low achievement trajectory. For children first looked after during KS3, this pattern was reversed, with children with a longer time in care being more likely to follow a low achievement trajectory.
This study has significant strengths: (1) the large sample size; (2) the application of trajectory analysis to the educational progress of LAC, which we believe to be unprecedented; and (3) the high quality of the local authority data returns, meaning that no children had to be removed from the sample because of missing covariate data.
Some limitations should also be noted. Firstly, the sample includes only those children born in academic year of 1993 to 1994 through the academic year of 1997 to 1998 who were looked after at some point during the years 2005 to 2006 through 2012 to 2013. There will be a significant number of LAC who were born in these academic years whose care history did not extend into the years 2005 to 2006 through 2012 to 2013, with the result that they do not appear in the sample. Many of these children who left local authority care permanently at a fairly early age will have gone on to have good educational outcomes. The absence of these children from the sample means this study may underestimate the beneficial effects on children’s educational progress of being looked after at an early stage.
Secondly, periods in care were recorded only if they extended into the years 2005 to 2006 through 2012 to 2013. Thus, some children in the sample may have had earlier periods of being looked after that are not recorded in the data. Thus, some children recorded as being first looked after at a given educational stage may have been first in care somewhat earlier and also had a greater total time looked after than the data indicate. This measurement error is likely to weaken the observed effects on trajectory group membership of educational stage first looked after and total time of being looked after.
Of the children who could have had data available from the KS1, KS2, and KS4 tests, 76.3% did so. If the probability that a child’s educational data were present is determined by the observed model covariates, then these data are missing at random and do not lead to biased results. We suggest that the wide range of demographic-, need-, and care-related variables included in the models means that it is likely that the assumption that the data are missing at random is met.
The final limitation to be noted is that correlation cannot be assumed to imply causation. Indeed, in a study of LAC and their educational progress, this point needs special emphasis because the problems that cause children to be taken into local authority care, the care environment itself, and educational progress have a complex relationship of mutual influence on each other over time.
Bearing these caveats in mind, it is notable that being looked after at an earlier stage was associated with a higher probability of following a high achievement trajectory and a lower probability of following a late decline trajectory. For children first looked after during KS2 to KS4, having had a longer time looked after by the end of KS4 was also associated with a higher probability of following a high achievement trajectory. Similarly, for children first looked after during KS3, having had a longer time looked after by the end of KS4 was associated with a lower probability of following a late decline trajectory. The general picture is of better outcomes for those who were looked after earlier and had spent longer in care. However, those children first looked during KS3 who had spent the longest time in care by the end of KS4 had a higher probability of belonging to the low achievement group. This may be a consequence of the higher level of need of these children relative to those who had spent a shorter time being looked after rather than an indication that time in care contributed to low educational achievement.
Children who were first looked after later than the end of KS4 were used as a reference group for children first looked after at earlier stages. In many cases, these children took their GCSEs shortly before being deemed to require local authority care; their relatively poor performance may be at least partly attributed to the high levels of need many of them were experiencing during these examinations. This factor means that the conclusion that being looked after increases the probability of a late educational improvement must be tentative. However, it should be noted that the associations found between being first looked after at an earlier educational stage and having a greater probability of following a high achievement path and a lower probability of following a late decline trajectory do not depend on the choice of this particular reference group and may be regarded as fairly robust; these conclusions follow from comparing groups of children who were first looked after before the end of KS4.
The potential of GTA to identify factors associated with a late decline or improvement in educational achievement is of particular interest. The probability of belonging to the late decline group was lower the earlier children were first looked after, and for those first looked after during KS3, it was lower for those who had spent longer being looked after by the end of KS4. Those born in the first half of the academic year were more likely to belong to this group, which leads to the speculation that some children in this group may be those who were easily ahead at primary school but became bored and fell behind. Other risk factors for a late decline include being a boy; having had >1 period in care by the end of KS4; having the special educational need "behavioral, emotional, and social difficulties"; the category of need when first looked after being "socially unacceptable behavior"; and the child’s placement being independent living or residential rather than fostering.
There is some evidence that for children with poor educational results up to age 11, being looked after is associated with a higher probability of following a late improvement trajectory, although, as noted previously, this conclusion is tentative. The probability of a late improvement was higher for girls and higher for those with specific learning difficulties or behavioral, emotional, and social difficulties than for other children. Those who had >1 period in care by the end of KS4 were less likely to follow a late improvement trajectory, as were those whose placement type was independent living or residential rather than fostering.
In accord with earlier research, this study provides evidence that being looked after at an early stage and for a longer time is generally beneficial to children’s educational progress, a finding that may inform future public policy in this area. We argue that GTA is a flexible and effective tool for analyzing the educational progress of LAC. GTA is particularly effective in identifying factors associated with temporal changes such as a late decline or late improvement in children’s educational achievement. More detailed exploration of the reasons for the temporal changes will require further study, possibly by using qualitative methods focused on individuals identified by trajectory analysis. Research has shown that the factors affecting the educational progress of LAC are internationally comparable.7 We therefore believe that our conclusions are relevant beyond the United Kingdom.
We would like to thank the staff at the National Pupil Database and the Department for Education for providing this data.
- Accepted May 30, 2017.
- Address correspondence to Julian Gardiner, PhD, Department of Education, University of Oxford, 15 Norham Gardens, Oxford OX2 6PY, UK. E-mail:
This report presents independent research. The views expressed are those of the authors and not necessarily those of the National Health Service, the National Institute of Health Research, or the Department of Health.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funded by the Medical Research Council grant MR/L020335/1 (grant holder Sutcliffe) and supported by the National Institute for Health Research Biomedical Research Centre at Great Ormond Street Hospital for Children NHS Foundation Trust and University College London.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Sebba J,
- Berridge D,
- Luke N, et al
- O’Higgins A,
- Sebba J,
- Luke N
- Fletcher J,
- Strand S,
- Thomas S
- Department for Education
- Department for Education
- Department of Health
- Department for Education
- Melhuish E,
- Sylva K,
- Sammons P,
- Siraj-Blatchford I,
- Taggart B
- Department for Education
- Department of Education
- Cooper P
- Department of Education
- Department for Education
- Department for Education
- Copyright © 2017 by the American Academy of Pediatrics