Returning children home from care: What can be learned from local authority data?

International Human Rights and child rights conventions as well as U.K. wide legislation and guidance require that children in care should be returned home to one or both parents wherever possible. Reunification with parents is the most common route out of care, but rates of re-entry are often higher than for other exit routes. This study used 8 years of administrative data (on 2,208 care entrants), collected by one large English local authority, to examine how many children were returned home and to explore factors associated with stable reunification (not re-entering care for at least 2 years). One-third of children (36%) had been reunified, with adolescent entrants being the most likely age group to return home. Three quarters (75%) of reunified children had a stable reunification. In a fully adjusted regression model, age at entry, being on a care order prior to return home, staying longer in care, being of minority ethnicity, and having fewer placements in care were all significant in predicting chances of stable reunification. The results underline the importance of properly resourcing reunification services. The methods demonstrate the value to local authorities of analysing their own data longitudinally to understand the care pathways for children they look after.

go home and to support families for "as long as it takes" (Action for Children, 2008) after the child's exit from care.

| How many children return to parents and which children are most likely to do so?
There are a range of studies of reunification, from many different countries (see scoping reviews of Biehal, 2006;Carlson et al., 2019).
However, different approaches to child welfare across different jurisdictions make direct comparisons of findings difficult; hence, our literature review will focus on studies of reunification in England. The Department for Education in England publishes annual statistics on children in care drawing on the administrative data records (the SSD903A returns) submitted by local authorities. Data for the most recent year available  show that 31% of children left care to live with a "parent or another relative or person with parental responsibility" (Department for Education, 2018b). The current study focuses on children who entered care in the last 10 years (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) in one local authority. Examination of the national statistics show that over this time period, the percentage of children each year leaving care to be reunified with parents declined steadily from 39% in 2010 (DfE, 2014) to the current rate of 31%. No breakdown of information about the characteristics of reunified children, for example, their age, ethnicity, duration of stay, or number of placements in care, is published by DfE. Neither do they report how many reunified children re-enter care.
Factors affecting the likelihood of a child being reunified include circumstances relating to the child, the family, and to social work practice. Regarding child characteristics, gender does not appear to be a factor (Biehal, 2006). Minority ethnicity has been linked to lower chances of reunification in some U.S. studies, but these findings cannot be extrapolated to those in U.K. (Biehal, 2006;Thoburn, Robinson & Anderson, 2012). Factors that indicate the child's increased vulnerability such as disability (Biehal, Sinclair, & Wade, 2015;Brandon & Thoburn, 2008;Cleaver, 2000), entering care as an infant (Sinclair, Baker, Lee, & Gibbs, 2007), and coming into care because of abuse or neglect (Biehal et al., 2015;Cleaver, 2000;Farmer & Parker, 1991) make reunification less likely.
Reunification is mostly likely to happen within 6 months of care entry and is afterwards much less likely, this often being referred to as the "leaving care curve" (Bullock, Little, & Millham, 1993;Rowe, Hundleby, & Garnett, 1989;Sinclair et al., 2007).
Contact between the child and parent whilst the child is in care has been linked to reunification (Biehal et al., 2015;Bullock et al., 1993). Such contact may not directly cause reunification but may indicate other influential factors such as the quality of the parent/child relationship, parental motivation, and/or good social work support (Biehal, 2006). In terms of the child's connection with their family, retaining their place in the family system matters (Bullock et al., 1993).
At the same time, change in the family may be necessary, particularly in terms of improvement of problems linked to the care admission (Biehal et al., 2015;Bullock et al., 1993;Cleaver, 2000;Sinclair et al., 2007). Active assessment of the home situation and planning towards return by the social worker can facilitate returns home (Farmer, Sturgess, O'Neill & Wijedasa, 2011).

| How many reunified children come back into care, and what factors link to failed reunification?
Conclusions on the proportion of reunified children who come back into care vary depending on research methodologies, study focus, and follow-up period (Thoburn, Robinson & Anderson, 2012;Biehal, 2006). For example, Farmer, Sturgess, O'Neill and Wijedasa (2011) reported that 2 years following reunification, 47% of children had come back into care. Looking just at neglected children, Lutman and Farmer (2012) found that 65% of those reunified had returned to care 5 years later. Highs rates of return to care are also experienced by maltreated children, 59% of whom had re-entered care after 4 years in the study by Wade, Biehal, Farrelly, and Sinclair (2011). Most recently, McGrath-Lone, Dearden, Harron, Nasim, and Gilbert (2017) used administrative data on a large sample of children who exited care in England in 2008 to explore who came back into care (with a followup period of 5 years). Children who returned to parents were the group most likely amongst all care leavers to return to care, 40.5% returning within 5 years. Other factors linked with higher rates of returning to care included age at exit (age 11-15 years was the highest risk group), white or mixed ethnicity, history of previous care entry, and more placement changes in care. Children who had been in care for longer than 1 year had a lower chance of returning to care.
This study however explored factors affecting return to care across all exit routes, rather than focussing specifically on reunified children. Mixed methods studies have been able to explore a wider range of factors associated with the stability of returns home. Farmer & Wijedasa (2012) found older children, abused children, and those with less stable care histories to be more at risk of returning to care. Their study highlighted several practice-related factors such as the importance of good assessments, planning, setting conditions to be met, and support for families (including from agencies independent of social services). Children who went home on care or supervision orders had more stable returns, possibly because they were younger, and the return was better planned and supported. Findings about the importance of planning and supporting reunification are echoed by Biehal et al. (2015) and Carlson et al. (2019). Some children who return home end up experiencing inadequate parenting or further maltreatment (Wade et al., 2011;Brandon & Thoburn, 2008;Farmer & Wijedasa, 2013), demonstrating the need to consider carefully which children should go home and to effectively support children and parents before and after return home.
Although much has been learned from the existing studies in the English context, there are limitations to our knowledge. Crosssectional studies will underrepresent children who stay short term in care, and therefore over represent children whose situations are more complex (Biehal, 2006). Where studies include only certain groups such as children who have been maltreated, they underrepresent children with less challenging backgrounds or characteristics. Only two studies have focused on all care entrants (Dickens, Howell, Thoburn, & Schofield, 2007;McGrath-Lone et al., 2017), and both of these relate to samples of children who entered care over 10 years or more ago. Further study of reunification rates and stability focusing on the last 10 years and including all care entrants is therefore warranted. We aim to address this gap by analysing the administrative data for a complete sample of all children who entered care between 2009 and 2015 in one large English local authority. We aimed to find out firstly which characteristics of the child and their care history were associated with the likelihood of return home. Second, we aimed to explore what factors were associated with a stable reunification (lasting at least 2 years, and for some, up to 8 years) and for those whose return home failed to examine the characteristics of their second stay in care.

| The dataset and sample population
The data used were from 8 years of the "SSDA903" records The study population consisted of the 2,208 children who started to be looked after over 6 years from April 1, 2009, to March 31, 2015. Children on an agreed series of short-term breaks or who came into care because they had (or were alleged to have) committed a crime were excluded. Children could have multiple periods of care. A subset included only children who returned home during the 6 years (n = 802). Both the total cohort and the "returned home" subset were followed-up until March 31, 2017, a follow-up of between 2 and 8 years. Children were coded as "reunified" when it was recorded that they had "returned home to live with parents or relatives." However, most children who leave care to live with a relative do so under the terms of a legal order (Special Guardianship Order or Residence Order [now Child Arrangements Order]; coded elsewhere); therefore, this code is used almost exclusively for return to a parent.

| Variables included in the analysis
All available variables that related to children's demographic details and their placement pathways were selected for the analyses and coded as follows: Age at entry was recoded into groups: 0-2, 3-6, 7-11, and 12-17. These groups were based on preliminary inspection of children's end trajectories (where they were last time we observed them in the data set, the options being: "returned to a parent," "the making of an Adoption Order," "leaving care to independent living," "leaving care following the making of a Special Guardianship or Residence Order (SGO or RO)," "other" reason for leaving care, "still in care").
Children with similar end trajectories were grouped together .
Ethnicity was grouped as "white" and "non-white ethnic background"; this grouping was used as the study population was mainly white (90%). Main reason for entering care was grouped as "abuse or neglect" and "other." "Abuse and neglect" was chosen as the reference category because it was the first reason that can be entered from a hierarchical list of eight options and is also the most common reason given for care entry. "Other" included child's disability, parental illness or disability, family in acute stress, family dysfunction, socially unacceptable behaviour, low income, and absent parenting.

Deprivation of the home address
Placement at entry/when last observed in care were grouped as "foster care" and "other." "Foster care," which included placement with kinship foster carers, was chosen as the reference category because it was the most common placement. "Other" contained a wide variety of options including placement for adoption, placement with parents, and children's homes. These options individually were not common enough to analyse their association with type of reunification.
Placement changes was defined as the number of moves between in-care placement settings (i.e., one move equals two placements).
Remaining at the home address whilst "in care" or returning to the home address was not included as a change in placement.
Placement changes per year was defined as the number of placement changes divided by the time spent in care measured in years. If the time spent in care was less than a year, the total number of placement changes was used instead to avoid upward bias.
Stable reunification was defined as not re-entering care within 2 years of leaving.
Unstable reunification was defined as re-entering care within 2 years of return home.

| Statistical analyses
Contingency tables were created to summarize the profiles of three groups of children (no reunification, an unstable reunification, or a stable reunification). Using Kaplan-Meier estimators, time in care was summarized for these three groups. For children with an unstable reunification, Kaplan-Meier estimators were also obtained to summarize the time spent out of care and the time of the re-entered period in care.
Logistic regression models were fitted to estimate the associations between the outcome of returning to parental care and the independent variables of children's characteristics at entry (gender, age group, ethnicity, reason for care entry, legal status, placement, and financial year). For the subset of reunified children, logistic regression models were fitted to estimate the associations between stable reunification and the independent variables of demographic characteristics at entry and care characteristics at departure (gender, age group at entry, ethnicity, reason for care entry, financial year at entry, legal status and placement at last observation in care, number of placement changes per year, and time in care). In the regression analyses, deprivation of the home address was excluded as it would otherwise substantially reduce the sample size and bias the results towards children with a longer period of care.
First, regression models were fitted for each independent variable separately to obtain unadjusted odds ratios, which ignore the effects of the other independent variables associated with the outcome. Next, regression models were fitted that included all independent variables together to obtain adjusted odds ratios, which take into account the effects of the other independent variables associated with the outcome. The regression model assumptions were checked and assessed on overall performance (McFadden's R 2 ) and discrimination (specificity, sensitivity, and overall accuracy).

| How many children returned to parental care and what factors were associated with being reunified?
Of the 2,208 children who entered care between 2009 and 2015, 802 (36%) were returned home in that same period (Table 1). Reunification happened between 1 day and nearly 6 years of being in care, the average being 53 days (1.7 months). Children who were not reunified were a diverse group and spent on average 2 years in care, ranging from 1 day to remaining in care at the end of the study period of 8 years.
Children aged 12-17 at entry were significantly more likely to be reunified compared with the younger children (42% versus 32-35%).
Those in care for reasons other than abuse or neglect were significantly more likely to return home (40%) than abused/neglected children (33%). There was no clear trend in the incidence of reunification by deprivation of the home address, although the incidence was significantly higher in children with unknown deprivation (i.e., missing postcode) compared with children with known deprivation (54% versus 22-24%, respectively). Children who at follow-up were accommodated under Section 20 were significantly more likely to return home (56%) compared with those on care orders (15%). Children who had their last placement with foster carers were significantly more likely to be reunified (47%) than children who had a different placement (15%). Finally, in later cohorts, significantly fewer children were reunified, with the incidence decreasing from 45% in 2009/11 entry to 28% in 2013/15 entry.
In the fully adjusted model, the outcome of reunification was significantly associated with financial year at entry, age at entry, and placement type at entry (Table 2). These independent variables contributed 48%, 19%, and 18%, respectively, to the model's explained variance in reunification. Compared with children who entered care in 2009/11, children who entered in 2011/13 or 2013/15 were 0.8 or 0.5 times less likely to be reunified. Compared with children aged 0-2 at entry, children aged 12-17 were 1.6 more likely to be reunified whereas there was no significant difference for children entering aged 3-6 or 7-11. Finally, compared with children who had their first placement with foster carers, children who had a different first placement were almost half as likely to be reunified.
3.2 | What factors were associated with a stable reunification and, for children whose return home did not last, what were the characteristics of their second stay in care?
There was a trend in the incidence of stable reunifications by deprivation, stable reunifications being observed for 48% of the children from the three most deprived quintiles, 60% of the children from the two least deprived quintiles, and 88% of the children with a missing deprivation quintile. Children entering or exiting care on a care order had significantly more stable reunifications (83% and 90%, respectively) compared with children entering or exiting care being accommodated under S20 (both 72%). The incidence of stable reunifications significantly decreased with increasing number of in care placement changes per year, from 91% when there was less than one change per year to 67% when there were two or more changes per year. Finally, children with a stable reunification were looked after on average for slightly longer than children with an unstable reunification, with the median time to reunification being 1.9 and 1.3 months, respectively. This difference was not significant.
In the fully adjusted model, the outcome of stable reunification was significantly associated with age at entry, legal status at exit, number of placement changes per year, time in care, and ethnicity (Table 3). These independent variables contributed 28%, 27%, 17%, 14%, and 8%, respectively, to the model's explained variance in stable reunification. Compared with children aged 0-2 at entry, children aged 12-17 at entry were almost half as likely to have a stable reunification; there was no significant difference for children aged 3-6 or 7-11 at entry. Compared with children accommodated under S20 before they left care, children on a care order were three times more likely to have a stable reunification. Compared with children with less than one placement change per year, children with 1-2 or ≥2 placement changes per year were 0.6 or 0.4 times less likely to have a stable reunification. There was a positive relationship (with diminishing returns) between time in care and the likelihood of a stable reunification; the longer a child stayed in care, the higher the likelihood of a stable reunification, with the highest likelihood at approximately 2.5 years in care before return home. Children who were reunified at 1 month, 2 months, 1 year, or 2 years, were 1.1, 1.2, 2.0, or 2.8 times more likely to have a stable reunification. Finally, children from a non-white background were 1.8 times more likely to have a stable reunification than white children.
Children with an unstable reunification re-entered care between 1 day and 2 years later (median = 4 months). Their second stay in care was on average significantly longer than their previous period, with the median time in care being almost 10 months longer. Apart from a few exceptions (n < 10) children re-entered care for the same reason as their previous period. A care order was more common the second time a child entered care; 26% of children in the second period in care had a care order compared with 20% in the first period. Foster care placement was less common the second time a child entered care; 69% of children in the second period in care had a foster placement compared with 75% in the first period in care. There were fewer placement changes per year in the second period in care, where 24% had <1 per year, 51% had between one and two per year, and 25% had two or more per year (for the first period in care these percentages were 3%, 70%, and 27%, respectively).
Most children who had an unstable reunification had left care again for the second time during the study period. Just over one third (35%) went home again (although it is not known whether this was to T A B L E 1 Characteristics of children by reunification (n = 2,208)

| DISCUSSION
This analysis of local authority data found that just over a third of care entrants (36%) had returned home within 6 years of care entry, with most returning home soon after entering care (an average of

months). This observation of generally rapid returns concurs with
other studies in the United Kingdom (Biehal, 2006), Australia (Delfabbro, Fernandez, McCormick, & Ketter, 2015), and the United States (Wulczyn, 2004). The "average" time to reunion masks differences between children relating to age at entry and other circumstances, for example, older children who go home after years because of a placement breakdown versus young children returning quickly after initial assessments have been completed (Farmer & Wijedasa, 2013;Esposito et al., 2014).
Children entering care earlier in the study period were more likely to be reunified, this downward trend mirroring patterns in England as a whole. This study cannot answer questions as to why fewer children went home. However, as with the rise of children entering care, overlapping factors such as a lack of family support services and rising levels of deprivation may be part of the picture (Thomas, 2018).
Children who, on entry to care, were placed in a setting other than a foster home were less likely to go home. Some of these children may have been in specialist settings, possibly because of needs such as emotional and behavioural problems or disabilities. Children entering care aged 12-17 were more likely to go home, as found in previous studies (Esposito et al., 2014;Sinclair et al., 2007). In this sample, adoption and special guardianship were routes out of care only for younger children, and long-term foster care was used mainly for those entering age 7-11 ; thus, for teenagers, reunification was effectively their main chance of permanency.
The reported 25% rate for re-entry to care after reunification is generally lower than found in other studies (Farmer & Wijedasa, 2013;Wade et al., 2011;McGrath-Lone et al., 2017). This may be because we have studied all children entering care rather than subgroups and/or because of differences in length of follow up, changing thresholds for care entry over time, or local authority variation (Farmer, 2018;McGrath-Lone et al., 2017). Also, children re-entering care in other local authorities would have been missed in our sample.
Stable reunifications were more likely for children who entered Farmer's conclusions (Farmer, 2018). As she notes, age at entry and T A B L E 2 Unadjusted and adjusted odds of reunification at home (n = 2,208) placement instability may both be associated with greater child problems such as emotional difficulties that could impact on the chances of a successful return home, and instability in care may in itself cause children to be distressed and disturbed.
Children who went home on a care order were less likely to reenter care, as also found by Sinclair et al. (2007) and Farmer & Wijedesa (2013). As Farmer notes, legal status may link to age as well as to differences in pre-return assessment and planning, and post return monitoring and services. Our analysis controlled for age and reason for entry, but we had no data about assessment, planning, or support, factors that may be more important than legal status per se (Farmer, 2018). Our finding that reunions were more likely to be stable for children who spent longer in care is congruent with other studies and may be a warning against local authorities (and courts) aiming for children to remain in care for as short a time as possible. As other studies have shown, it is vital that problems, which led to the need for care, are adequately resolved before reunification takes place.

| Strengths and limitations of the research
This is one of only a small number of English studies to analyse data on a full and large cohort of entrants to care, and the only such study analysing children entering care in the last 10 years. The regression analyses showed the importance of adjusting for children's characteristics when estimating the likelihood of (stable) reunification. The adjusted regression models were able to distinguish between children who returned to parental care and those who did not (total accuracy of 65%), and between children who had a stable reunification and children who came back into care (total accuracy of 75%).
A key limitation is that we analysed data from only one local authority, and practices across local authorities vary widely (Dickens et al., 2007;Harwin, Alrouh, Bedston, & Broadhurst, 2018;Sinclair et al., 2007). The main variable used to identify children reunified with parents may also have included small numbers of children returned to relatives. The analysis was limited to the variables included in the administrative dataset, and other factors (for example, the role of ongoing contact and good social work practice) could not be examined. Higher accuracies of prediction could potentially be achieved with more information on children's backgrounds and especially if we had fuller data on deprivation, which is likely to have had an impact on entry, reunification, and re-entry rates (Bywaters et al., 2015).

| Implications for practice
The study has a number of implications for local authority managers, social workers, and their family justice partners. The 2015 Guidance on Permanence (DfE, 2015, pp. 4-5) draws attention to return to parental care as a permanence option that must be carefully planned and supported. However, permanency through reunification is only given cursory attention in policy and practice (Farmer, 2018) even though the task of social workers supporting reunified children may be more complex and sensitive than for other routes out of care. For some parents and children, anxiety stemming from the distress of their earlier separation is likely to come through as a reluctance to accept assistance or seek it if stresses build up.
With a quarter of reunified children re-entering care, the importance of carefully timed and properly planned and resourced specialist support for children's return to parents is underlined. The financial cost of this is likely to be much lower than cost of children re-entering care (Holmes, 2014). Formal and specialist approaches to supporting family reunification are underdeveloped and patchy (Hyde-Dryden et al., 2015). Thoburn et al. (2012) identified 11 research reports (mainly from the USA) published since 2005 of "promising" interventions aimed at improving success rates when children return from care. A detailed analysis of helpful approaches has been set out by Farmer (2018). Planning towards reunification should begin at an early stage with assessment focusing on precipitating problems and the nature of the parent/child relationship. Use of assessment tools should be considered, such as the risk assessment framework developed as part of the NSPCC framework for return home (Farmer & Patsios, 2016;Wilkins & Farmer, 2015). This framework also provides guidance on using parental agreements, goal setting, maintaining family links, and planning the support needed prior to and after return home.
In our study, reunifications failed on average 4 months after care exit, a reminder of the importance not to close cases quickly but to provide ongoing support extending beyond an initial "honeymoon" settling in period (Wilkins & Farmer, 2015). There may be important roles that foster carers, residential workers (Farmer & Wijedasa, 2012;Fernandez, 2012), and parent mentors (Berrick, Cohen, & Anthony, 2011) can play in helping parents engage with social workers and in providing additional support. Within a specialist family reunification service, it may be important to recognize and meet the needs of subgroups of children that are more at risk such as teenagers, young people who have experienced instability in the care system, and those who have been placed in specialist settings. Whilst addressing child behaviour problems may be important, it is also essential not to address families' socio-economic needs (Akin, Brook, Lloyd, & McDonald, 2017).
Although we found that returning home on a care order was associated with more successful reunions, a preoccupation with legal status may be unhelpful (Farmer, 2018). Care orders should not be used simply as a means of guaranteeing support for families, and the needs of children going home from voluntary care must not be underestimated (Farmer, 2018). Where families primarily need support rather than monitoring, a supervision order may align more closely with their needs than a care order (Fargas, Mc-Sherry, Pinkerton, & Kelly, 2017). However, recent research questions the benefit of supervision orders because of their time-limited nature (Harwin et al., 2019).
Finally, our analysis suggests local authorities can make better use of their routinely collected annual returns of statistical data by linking data across different years to track the success of reunifications. This could then be used to explore other data held locally (such as case file data) to explore why reunifications succeed or fail disproportionately for subsamples of children identified from the administrative data.