Linking children's social care data to information about their care proceedings to understand the use of care proceedings and their effects on parents, children and local authorities

Funding information Nuffield Foundation, Grant/Award Number: 43316; Economic and Social Research Council, Grant/Award Number: ES/M008541/1 Abstract Care proceedings are the most powerful child protection intervention; local authorities make applications for over 20,000 children each year in England and Wales. The Department for Education holds two administrative databases of children who receive services, for children looked after and children in need. These databases do not record the use of care proceedings. Data linkage is a powerful tool for revealing patterns not visible from a single data set; the Outcomes for children before and after care proceedings reform study linked these administrative records with a research database of care proceedings to examine children's care and service journeys associated with care proceedings. The researchers also devised a simplified method for use by local authorities with their own records and tested this with North Yorkshire County Council. Analysis of the linked data reveals how care proceedings are used and the impact of care proceedings on care demand. It provides the basis for analysis of cohorts of children subject to proceedings and feedback to social workers, children's guardians, and the judges and lawyers on their decisions and children's outcomes.

family with new legal parents, an outcome frequently referred to as "draconian" by the courts. They can involve findings against parents, which make it likely that further children will be removed; approximately 25% of mothers involved in care proceedings during 2008-2011 were involved again within 7 years (Broadhurst, Mason, Bedston, et al., 2017: 5). Not all care proceedings have these extreme effects; approximately 20% of children in care proceedings remain with or return to their parents without any order or with only a supervision order (Harwin et al., 2019;Harwin, Alrouh, Bedston & Broadhurst, 2018). Courts are required to make proportionate orders (European Convention on Human Rights, Article 8(2); Re B 2013), that is to limit any intervention to the least, which will protect the child, and there is an explicit policy of maintaining children within their wider family, wherever possible (Children Act 1989, s.17(1), Department for Education [DfE] 2014: 2.2, Re L 2007.
Bringing care proceedings makes very considerable demands on local authorities, particularly on social workers and managers making decisions (Beckett, McKeigue, & Taylor, 2007;Dickens, 2006) and preparing applications for court (Beckett, Dickens, & Bailey, 2014;DfE, 2014;Masson et al., 2013). Resources are expended on court fees, lawyers, and tasks, which are only required for cases taken to court-preparing evidence statements and attending hearings; they commonly result in orders requiring placements and/or services for  (Cafcass, 2018). Although only a minority of children enter care through court proceedings, children subject to care orders stay longer; consequently, increasing the use of care proceedings adds to the demand for care placements. On March 31, 2018, 80% of the care population were subject to an order made in care proceedings (DfE, 2018a).

| Children's social care data and care proceedings
The DfE maintains two large, longitudinal, administrative databases on children in need (CiN;DfE, 2018b, Emmott, Jay, & Woodman, 2019 and looked after children (CLA;DfE, 2018c, McGrath-Lone, Harron, Dearden, Nasim, & Gilbert, 2016 and publishes summary and local authority level, snapshot analyses of these data, showing the numbers of children helped each year and trends over 5 years (DfE, 2018d). Some longitudinal analyses are undertaken by the department, for example, to examine the extent to which children achieve permanence (DfE, 2013) and, more recently, children's journeys through services over time (DfE, 2018e). The Welsh government also maintains a database of children in care but only collects aggregate data and undertakes a census of services for CiN based on a sample (Welsh Government, 2018aGovernment, , 2018bGovernment, , 2018a. Despite the impact of care proceedings on local authorities and the care system, the use of care proceedings is not recorded in any of these administrative databases. Information is collected on a child's legal status for each episode of care, but the published information only details legal status on entering and leaving care and the numbers of children with each legal status on the census date. Also, in most official statistical publications, (temporary) interim care orders, granted during care proceedings, are not distinguished from full care orders made at the end of proceedings (DfE 2018f).
Local authorities in England make annual returns to the DfE relating to the services provided for every child referred as in need or looked after. Similar processes operate in Wales for Welsh local authorities. Returns are compiled from recording systems for service provision and use the same unique identifier for each child each year.
Guidance on completing the return, particularly on coding, is set out in separate documents for each return (DfE, 2018b(DfE, , 2018c and revised and reissued each year. Rigorous checks are made for the accuracy and internal consistency of these data, and the cleaned data are made available to local authorities for their own use. Data collected for these returns are primary sources for local authorities' own analyses. Specifically, the DfE has created an interactive tool (Local Authority Interactive Tool), which allows local authorities to compare descriptive statistics and examine these alongside other data; the only additional proceedings-related data included are rates of care proceedings applications provided by Cafcass, the Children and Family Court Advisory and Support Service (DfE, 2017).

| Care proceedings data
Local authority legal departments keep records of all care proceedings cases, usually by reference to their court application number so that the lawyers responsible can monitor progress with other lawyers and the court. Legal department records do not usually include the children's social care identifier; most children are not in care at the start of care proceedings  but, whether they have a child protection plan or not, all are in need of protection-that is the purpose of the court application. It is extremely rare for any child unknown to the local authority to be the subject of immediate care proceedings (Masson et al., 2008). National data relating to care proceedings are held in three separate systems: Cafcass has a database (e-cms) with information on all family proceedings in England, with which it has any involvement; a similar system is operated in Wales by Cafcass Cymru. The courts service operates a case management system for England and Wales (FamilyMan) through which courts manage the different court processes for family cases and is the source for court data on care proceedings (MoJ, 2018(MoJ, , 2018a. Alrouh and Broadhurst (2015) explained the structure of the Cafcass database, and how it was used to establish the numbers of mothers repeatedly involved in care proceedings. Both Cafcass and FamilyMan data are used in the MoJ data visualization tool, Public Law Applications to Orders (PLATO; MoJ, 2018b), which consequently limits some analyses to England.

| DATA LINKAGE
Data linkage is a powerful tool for revealing patterns not visible in single data sets (Harron, 2016). Sebba and colleagues linked children's social care data with their education records in the National Pupil Database to examine the educational progress of CLA and the impact of care experiences on educational achievement. In contrast with common beliefs about the negative impact of care on education, the researchers showed that children who were looked after at age 10 years (KS2) made above average progress to age 16 years (KS4) and better progress than CiN who were not looked after. However, those who entered at age 15 years had much worse educational outcomes, especially if they had moved placements and/or schools while in care (Sebba et al., 2015). A study for the Children's Commissioner similarly linked care and education data (Children's Commissioner, 2017) to look at the number of moves in care and of school experienced by CLA, combining this with a survey on changes of social worker to create a stability index. This indicated that 6% of children experienced high levels of instability in all three measures with multiple placements, a midyear change of school and multiple changes of social worker (Children's Commissioner, 2017a, 5). The DfE has also linked its CLA and CiN databases to gain new insights into children's experiences in the social care system by building a picture of all children's journeys over time from referral to child protection plan and/or becoming looked after and/or case closure and so forth. Nearly 200 separate pathways were identified for children over 3 years (DfE, 2018e). Court data have also been linked with police data to examine offending patterns for children who have been subject to family proceedings; children in public law proceedings (mostly care proceedings) were found to have a higher rate of juvenile offending than the general population (Forty & Sturrock, 2017), a finding that conflates children's behaviour and police and criminal justice responses to it. A study of a cohort of children who became looked after in 2012-2013 aged five and under, linked two Scottish administrative data sets comparable with the CLA and the Cafcass databases. A previous referral to the Children's Hearing System had been made for 42% of the linked children, with a hearing arranged as a result of this previous referral for 12% of these children and leading to a statutory order for the child to be looked after in just over half those cases. The children's pathways revealed differences in the source of referral between children who became looked after at home or away from home. All children were more likely to have been referred from social work than from the police at the index referral, which led to the child becoming looked after; however, this was only the case at first referral in the group who became looked after away from home (Hooper, Cusworth, & Whincup, 2019).
The Scoping study for the Family Justice Observatory recommended that there should be "routine linkage" between data held by the MoJ (which includes Cafcass), the DfE's databases (including the CiN and the CLA), and local authority data (Jay, Woodman, Broadhurst, & Gilbert, 2017). Work has been done to achieve this with the children in family justice data share and the creation of the PLATO and WATCh tools (Kaspiew, 2018;MoJ, 2018b).

| THE OUTCOMES OF CARE PROCEEDINGS FOR CHILDREN STUDY (2015-2018)
This paper presents analysis and findings from the Outcomes of care proceedings for children study Masson, Dickens, Bader, Garside, & Young, 2017), on the relationships between the use of care proceedings and care demand. It discusses how local authorities can replicate this analysis using their own data and reports on a feasibility trial of this method. The original study was designed as a natural experiment to compare court processes and decisions, before and after the reforms to care proceedings, introduced by the Children and Families Act 2014, and the outcomes for the children involved, 1 and 5 years after the proceedings ended . It linked data extracted by the researchers from documents prepared for court for a random sample of care proceedings, brought by six local authorities in England and Wales, with administrative data in the CiN and CLA databases (and their equivalents in Wales). Data for the pre-reform sample (S1) related to proceedings brought in 2009-2010 and were originally collected for a study of the preproceedings process (Masson et al., 2013). The post-reform sample (S2), applications made in 2014-2015, was collected for the study. The combined sample included 616 children, 290 children in S1 and 326 in S2.
Deterministic methods were used to link the children and their care proceedings data to their administrative records, which were obtained following a formal application to the Department's Data Access Management Panel for specific variables for specific children.
The five English local authorities provided the reference number they used to submit each child's annual data; the DfE provided the extracted data to the researchers, subject to the usual, strict conditions about storage, access, processing, and publication. In Wales, comparable data were provided by the study local authority. Match rates of 90% for S1 and 98% for S2 were achieved with the CLA and CiN data combined. The combined data set was prepared with Stata v15 and analysed using SPSS 24.

| Limitations of the study
The study included only six local authorities, all in southern England or Wales; the sample did not include northern local authorities, which are now known to make substantial use of home Variables were derived from the combined data to identify children who were (or were not) in care during the proceedings so that their distinct care/service journeys could be identified. Using the beginning and end of care proceedings as reference points, it was possible to identify in each sample (that is, before and after the reforms), which children were looked after before the proceedings, how long this period of care had lasted and the legal arrangement for care during the proceedings. Overall, just over half the children were looked after before care proceedings were started, slightly more from the before sample (S1), 58%, than the after (S2) sample 47%, but for more than a third this arrangement had lasted less than a week. Just over a fifth of the children had been in the care system for more than 4 months when proceedings were issued. There was little difference between the before and after reform samples in terms of longer periods in care before proceedings except that all but one of the 13 children who had spent more than a year looked after were from S1. This may relate to the greater focus on avoiding drift in recent years.
Where children were looked after during proceedings, most were subject to interim care orders made early in the proceedings.
However, 70 children, 12% of those in local authority care during proceedings, remained subject to s.20 throughout (Masson, 2018).
This arrangement was acceptable to parents, local authorities, and the court. It demonstrated to the court that the parents were cooperating with the local authority and allowed the child to be looked after. Local authority lawyers were aware that orders would not be granted unless they were required-agreement removed the need for an order. The court could avoid hearing an application for an interim order and use the time for other pressing business. A higher proportion of children from S2, 14%, were subject to s.20 throughout the care proceedings than S1, 9%, and there was also a wide variation between areas. Only 5% of the S2 children from local authorities E and F were subject to s.20 during proceedings compared with 30% or more from B and C, which both took their cases to very overstretched family courts.
The analysis of children's care during proceedings also identified a substantial proportion of children in each sample (S1 15%, S2 20%) who were not looked after before or during the proceedings. These children either remained with their parents or stayed with relatives, informally or with an interim child arrangements order. Interim supervision orders (ISOs) were in place for most of these children; local authorities had either accepted an ISO because the court refused to agree the child's removal under an ICO or had withdrawn their ICO application because a parent contested this. If children remained at home or with relatives (not acting as local authority foster carers), they were very unlikely to be made subject to a care order at the end of the proceedings; only just over 3% of the children entered care at the final hearing, see Figure 1.

| Cohort analysis
Linking administrative data with care proceedings data in this way make it possible to plot the care and service journeys for children subject to care proceedings during specific periods. Such analyses have considerable potential for improving understanding of the impact of different types of orders on children's lives by enabling local authorities, Cafcass officers, and courts to learn about the longer term outcomes of their recommendations and decisions. The FJR recommended that courts should receive feedback on their decisions (FJR, 2011, para 2.210). This could be done by tracking, in the administrative data whether, for example, children with placement orders have been placed for adoption or adopted, or children subject to supervision orders have child protection plans (which would be shown in the CiN data) or have been accommodated (from the CLA data). Cohort analysis also makes it possible to monitor the impact of policy or practice changes on court decisions and answer questions about their effects. The Outcomes study was specifically designed to examine the care proceedings reforms, but the same approach could be used to identify the effects of, for example, a change in a local authority's policy about bringing care proceedings or the appointment of a new Designated Family Judge in the local court.

| FACILITATING ANALYSIS OF LINKED COURT AND CARE DATA BY LOCAL AUTHORITIES
Both local feedback and monitoring necessitate local authorities being able to examine their administrative data by reference to their use of F I G U R E 2 (a) Leaving care curves for all children (Sinclair, Baker, Lee, & Gibbs, 2007). (b) Leaving care curve for children subject to care proceedings F I G U R E 1 Entry to and exit from care for children in care proceedings (S1 and S2). The group labelled "sometimes" included different patterns of care including: entry during care proceedings; exit before the end of proceedings: late entry and early exit; and entry, exit, and re-entry care proceedings in individual cases. As explained above, variables relating to care proceedings are not collected for the CLA and CiN databases, and CiN data are not included in the MoJ Datashare.
Therefore, Masson devised a simplified method, which could enable local authorities to analyse the effects of their use of care proceedings on their care population. Local authority legal departments all hold records for the care proceedings they bring, which include the date of application and final order and, commonly, the order made. If only these three variables are added to data held by the children's social care department, the local authority should be able to examine the services different cohorts of children received before and after proceedings were brought and also provide feedback to all working in the family justice system on children's outcomes after proceedings. Such data can be presented in aggregate form, but it would also be possible for local authorities to identify children's casefiles and provide more detailed explanations for individual children as was done in the outcomes for children study , but as Dickens et al.

| Analysis
Using these data, the leaving care curve for children subject to care proceedings was produced. This was similar to Figure 2b (above), but a smaller percentage of children left care around the final hearing, reflecting the use of care orders with home placement rather than supervision orders. It was also possible to compare the rate of leaving care for children subject with care proceedings with that for all CLA, reflecting the approach taken by Sinclair et al (2007 and Figure 2a above). This was not possible in the Outcomes for children study because that sample only included children with care proceedings. Comparing these two groups of children, 70% of children subject to care proceedings remained in care 50 weeks after entering, compared with only 50% where those without proceedings are included, see Figure 3.
Sankey diagrams allow the flow of cases to be plotted between two time points with the width of the line reflecting the number of children with each journey type. Sankey diagrams were used to plot the different care and service journeys of children subject to proceedings, see Figure 4. For clarity here, some uncommon journeys, for example, involving family assistance orders and emergency protection orders have been excluded, and journeys with SGOs and SGOs and supervision orders have been combined. Figure 4 shows that most children who had been subject to s.20 for less than 3 months were made subject to an ICO, but a small number remained looked after under s.20 or were subject to ISOs. Further, although most children subject to ICOs ended proceedings with care, placement, or special guardianship orders, a few left proceedings with no order or a supervision order. A closer look might raise questions about the use of proceedings in these cases.

| Future development
New recording methods and systems in the NYCC legal department will allow instant and automatic matching to take place, with any discrepancies highlighted for correction. Elsewhere, the recording of the child ID number used by the social care department on local authority legal files is facilitating data matching. The introduction of a common IT and data management system for both children's social care and legal departments will make it easier to analyse social care data for children who are or have been subject to proceedings, in the way outlined above.
F I G U R E 3 Leaving care curves for children in care with or without care proceedings (test local authority)

| CONCLUSIONS
Linking data on care proceedings with children's social care data made it possible to identify interactions between bringing care proceedings and demand for care placements and raised questions about the use of this form of intervention, particularly in relation to children who never entered care. A substantial and increasing minority of the children in the sample were subject to care proceedings but remained outside the care system during proceedings, and very few of these children entered care as a result of the proceedings. This may indicate the success of proceedings as an intervention to support changes by parents as in Family Drug and Alcohol Courts (Harwin, Alrouh, Ryan, & Tunnard, 2014), alternatively it may suggest over intervention. Further analysis is needed to examine: the nature and duration of social work support before and during care proceedings; how these children's circumstances differed from children who were in care during proceedings; and what happened to them after the proceedings. This could illuminate both whether proceedings are necessary or appropriate if children can remain at home safely; whether some children in care during proceedings could be at home; and whether courts are adequately safeguarding children who only enter care at the end of proceedings.
A simplified approach, which allowed the analysis of social care data by reference to the use of care proceedings and the orders made in them was successfully applied within the test local authority despite the data being held in different systems which used different case IDs. The resulting analysis clearly highlighted the children's different pathways into and during proceedings, and the numbers of children who took these. More detailed data held by children's social care could be analysed to examine interactions with children's social care (visits, contacts, and other services involved) for children on the edge of care proceedings and used to indicate the likely journeys children would take if proceedings were brought, and to support decisions whether, or not, to bring proceedings. Identifying the orders made for cohorts of children subject to care proceedings enables local authorities to provide feedback to courts, Cafcass and other professionals on children's outcomes in terms of placement, reunification, and leaving care as the FJR suggested, with individual examples provided from children's social care files.
The use of the children's services department's child ID by the local authority legal department avoids the time-consuming process of matching records makes preparing the analysis simpler. Two further local authorities are known to have taken this first step to understanding their use of care proceedings.