Category: Child protection
What follows is a transcription of the audio recording. Due to differences between spoken and written English, the transcript may contain quirks of grammar and syntax.
JM - Janice McGhee
JM What I'll be partly doing is drawing on a recently edited book that Lorraine and I have been involved in and are the editors of that is, has this title New Directions in Safeguarding Children and I'll also be drawing on some of my work as the Social Work lead for the Administrative Data Research Centre. It's a Centre that is funded by the ESRC. There is one in each of the four parts of the UK and it's really intended to support and promote the use of data that is already routinely collected by Government, by Public Agencies, by the third sector and use that for research purposes. So, I'll be bringing, in a sense, both hats together to explore the contribution of particular forms of research evidence that might expand our knowledge in child protection.
Just to say a little bit about the book it draws on both humanities and social science contributors. We have contributors from around the globe, New Zealand, Australia, Finland, the UK and Italy and is intended to be slightly anti-conventional to, to kind of listen to authors both who are researchers but who come from different perspectives to really think about and re-examining the social problem of child abuse and neglect. So, it's a kind of attempt to go back to first principles and we know that violence against children is a global problem. I mean the UN have documented it very well. We know in a way that violence is part of everyday life as of course is the capacity to reject violence and that violence comes in many forms and not just physical violence of course but violations of personal dignity and human worth. What two very well-known anthropologists Nancy Scheper-Hughes and Philippe Beuchwar say, you know, "violence has a human face". So, I think when we are looking at research and child protection we really need to keep sight of that question, that very core question "Why are children harmed?" And so, we need to keep in mind the violence that children experience in their everyday lives across the world and across very different setting and connecting research to practice. And also, you know, children and particularly in conditions of armed conflict and political violence where they are often not simply bystanders to that violence but are often the targets of that violence.
I think also research in child protection again needs to recognise particularly the position of women and I think again the UN has been really at the forefront of looking at how these global problems of conflict, of instability, of inequality, of violent crime have very significant consequences for women and I think we need to bear in mind both in our practice but in our research of the position of women because in a sense the majority of social work clients and children and families are poor women. It's well documented in a range of research studies. Studies often talk about parents but they rarely desegregate women and men. So I think it's really important that as we take research forward that we provide an opportunity to hear the voices of women who are involved in child protection as mothers, as child welfare involved mothers and work is being done, for example, in Torsletbo in Norway has brought together birth parents, foster carers and social workers, that's birth parents who have lost children, to child protection systems and brought them together and had very productive discussions about ways forward, about how organisational change can be made and I think some of the work that was done in New York City by David Tobas in terms of developing parent advocacy groups which were primarily, were men and women but, you know, training parents to be advocates for other parents and how supportive that was both in terms of developing skills for the women, about improving the care of their children but actually having a real direct input into the organisation and delivery of services and actually what they saw was a significant reduction in the number of children coming into public care. So, I think that's in a sense our back to basics and thinking about what we mean when we're talking about child protection.
So, I'll go a little bit back to a more traditional perspective. So, I kind of was thinking about what I would talk about today. Lorraine and I were both thinking about this and we kind of thought, "Let's think about, what are the sort of traditional forms of research that are being undertaken in child protection social work primarily in the UK?" And obviously, we have the very traditional case studies. You know I would include serious case reviews as a single case and there's been meta-analysis of those by Marion Brandon at East Anglia. Very much drawn out some of the lessons for us for the future. One could see the public enquiries into those very serious cases of child death or very serious injury in a sense as another form of single case and obviously, that includes the historic abuse enquiries. A lot of work is kind of case serious, analysis of cases probably drawn from one or more local authorities within one jurisdiction in the UK. Often without a kind of comparator group so work like Elaine Farmer's done on patterns of reunion and admission to public care in cases of neglect. There are not many cohort studies. Some of the work that Jim Wade at York has done around again patterns of children coming in and exiting public care and the Avon Longitudinal Study is one of the few that has actually been able to look at very vulnerable families and outcomes for that and it's a bit unfortunate in some ways considering the UK is actually a very leader in birth cohort studies globally. And then obviously, we have very small ethnographies thinking of the work of Harry Ferguson or Sue White or interview based studies. Thinking of the work that Donald Forrester undertakes. So, that's kind of been the range of the work and what they've done in Australia, which is my colleague, Fiona Arney who's written a chapter in the book. What they actually did was in Australia they had a very national stocktaking of research. They wanted to see well what research is there, what international research can we draw upon to illuminate how we might approach the protection of children in Australia and they also used that as a way to set a national direction but to also think, "Well what are some of the gaps in the research base?" And I guess what they identified, and I think this is true for the UK as well, is that we know we have a lot of descriptive research on the social problem of child abuse and neglect. We have a lot of factors that are associated with child abuse and neglect, but perhaps we're, still less is know about how best to intervene with vulnerable children and their families. You know apart from some evaluations of very specific programmes. There's very limited research on the long term impact of social work intervention, This work by Morag Henderson and Jonathan Scourfield at Cardiff has been trying to use the birth cohort studies that do exist in the UK that at least have some reference to whether there has been contact with social work to look to see what differences there might be between those children who've grown to adults who have had contact with social work and those who have not, and so as I say birth cohort studies are where, you know, there's a kind of random national selection usually of all children born in a particular week or two weeks in the UK and they're followed up over periods of time. So, there's a birth cohort from 1946, there's one from 1958, there's one from 1970 and then our millennium cohorts. So, it's similar to the growing up in Scotland studies. So, they've been trying to see what can be used within those cohort studies that aren't set up particularly to look at child protection or to look at families and children who have contact with social work but to see, you know, to see what is usable and I've put all the references at the end of the slide so if you have a particular interest in a study you can take that forward. And I think we also have a lack of research in what you might call service as usual. The kind of routine service that most families where there are concerns about a care of the child might receive. So, you know obviously, involvement from social work, or Health Visitor or some other professionals. That kind of multiagency intervention that's very strongly embedded in Scottish policy both in terms of GIRFEC but more longstandingly in terms of child protection. And we don't really routinely, we don't really have a programme research that's kind of said, "Well how effective is this in the longer term for children and their families?"
So, it's left me with a, a kind of, some questions really to think about. As I say is there a need for some kind of stocktaking of research that has been undertaken here in Scotland and in the UK? There was an audit of child protection research that was undertaken by what was then the Scottish Childcare and Protection Network and is currently WithScotland. I think that was back in two thousand and, can you remind me, 2004? 2005?
JM 2005, yes where we tried to gather together all the research that had been carried out in Scotland related to child protection. Not just in social work but also in health and education and I think it was quite a useful review and database that was available but with all these things of course you need to constantly update.
What I'm going to talk about now really is to think about, "Can we expand the scale of social work research in child protection in terms of the kind of advancing technologies?" I know the Scottish Government has a very strong policy on better use of the data that's already collected and so can we use the data that we have already that is easily machine readable to answer some of our questions? So, I'm going to really talk a little bit now and I'm moving into my, my other hat which is the Social Work Lead for Administrative Data Research to think about what might be the added value that linking data that's already routinely collected, has extensively been done and held for many, many years, could add to this field of research. And I'll talk a little bit, just to give you some examples of how this data has been used in different types of research in different parts of the world. It's particularly been taken forward in the States, in California in particular, and also in Australia, particularly in Western Australia and those of you who are kind of familiar with this research will know that in places like Finland where everyone's documented from birth and every single record is linkable, that they have been doing this type of research for many years.
So, what do we mean when we talk about routinely collected administrated data? Well we're talking about the information that is already collected by Government but also, you know, by third sectors to some extent but I'll primarily talk about Government today on really our large population. So, data that you collect, for example, on educational performance of children. Every child when they go to school they're allocated a unique number and their performance, attendance is tracked over time, or capable of being tracked over time. Housing, health. Every time you go into hospital you'll, there'll be a record of that attendance in hospital attached to your unique Community Health index number. We all have this CHI number and that's been collected since, I think it's, is it 1983. So, every single... You know so this data has been collected on all of us all the time and it's obviously for the purposes of Governments, of distribution and allocation of resources, but what we're saying in a sense, it's also now capable or can be used for research. DWP obviously benefits that's the kind of area that the Administrative Data Research Centre in Scotland is working on to try and access the benefits data. So, what we've seen particularly in the last 20 years is this huge increase in data that is easily read. It's machine readable so it doesn't have to be, it's already entered. It doesn't have to be kind of re-entered and it's essentially available for analysis and it's possible to link individual records and records of an individual across different administrative sectors and it is becoming, I think, recognised as a source of research and particularly important for policy research. So, you can begin to potentially see the impact of changes in policy. What we mean by, as I say, data linkage or record linkage is bringing the information from two records or more that relate to the same entity so it doesn't necessarily, it could be a business for example, but primarily we're talking about individuals. So, a very straightforward example that's already been done here in Scotland is the linking of the looked after children and education data sets. So, each, as I say, each child has a unique number when they enter the education system called their Scottish Candidate number. Every child who's looked after, there's an annual return to the Scottish Government on episodes of care so each individual has, each local authority when they return that data to Scottish Government adds the Scottish Candidate number and that would, that's a form of deterministic linkage, you know, where you've got this unique identifier so you're bringing those two data sets together and it enables you to say, "Well how well are children who are looked after doing in comparison to the general population?" But what's particularly being looked at is to see, "Well how well are children who are looked after at home, doing educationally compared to children who are looked after away from home?" And what they found is that children who are looked after away from home appear to be doing better educationally. That's been one very straightforward example and where this data linkage has been used broadly in the field of child welfare. So, another kind of unique identifier of individuals in Scotland of course is what's called our Community Health Index number. We all have this unique number. So, it's possible to link all our health data together. This, what I'm really talking about is for research purposes. It's, it's the linkage would be done in what's called a "safe haven" by a third party which is the National Records in Scotland. It's done in a way to ensure that the researchers don't ever get those unique identifiable numbers. Everything is de-identified when it finally reaches the researcher and I can go into detail on the technicalities of that but it's really a privacy protecting method to ensure that those working with the data aren't able to identify any individuals and it doesn't arrive like on your computer, your office in the University to analyse, you have to go to what's called a "safe haven" which is a special lab which, you know is, the computer has no ports for downloads or anything like that and you have to run and do your analysis there and then and then before you can take anything out that analysis is checked to make sure there's nothing disclosive in it. So, you know, for example, in census data, you know, very small cell sizes are not permitted because it might identify individuals. So, there are a whole range of safeguards to ensure people's privacy is protected, that people can't be identified to make that access for researchers and probabilistic is more demographic data might be age, gender, dates of birth. And this is just a kind of quick summary of the sort of added value that can come from using administrative date for research purposes. So, you can have access to very large populations. It's well known in public health research and epidemiology. They've been doing it for many years in those fields, but we, we're moving it forward now to see if we can actually use it in terms of more social questions as opposed to health outcomes. You can have, you can look backwards and forwards. So, you could take every child who's been looked after in 2007/8 which is when the individual records first began, and you could look forward and if you could link their data to their health outcomes, for example, you could see how well they were doing. What might potentially have been the impact of that experience of care, or you could take all your current, perhaps, say care leavers and look backwards. As I say the strength is the capacity to link across sectors so that you can see pathways from childhood through to adolescence. You can see the temporal ordering of risks. So, you know, what comes first? For example, if a parent has a mental health problem, you know, does that impact on a child's educational outcomes or their attendance, or what is the pattern of that. As I say data on the children who are unreported to Child Protection Services that may have similar risk profiles and we do know from self-report studies both from children and from adults that there are many children who experience child abuse and neglect who never come to the attention of authorities, capturing rare events ...So and I guess for this audience today particularly the way that it can inform policy but also perhaps give some sense of how policy is working on the ground and also the potential for both international and cross UK comparative research.
So, let's just give one or two examples because I think that makes it come alive, makes it a bit more interesting. As I say health has been doing record linkage for many, many years and this is just a very straightforward example just solely using health data and it was across UK analysis and I wanted to try and tease out, well let's look at some differences in this time period and in a sense to slightly do a retrospective analysis and think, "Well what might have been the policies that might have contributed to these particular outcomes?" So I looked at all unplanned injury admissions between all the hospitals in Scotland and all the hospitals in England and particularly looking at the trends for maltreatment and violence related injuries and we adjust obviously for seasonal effects because you get a lot more certain types of injuries in the summer holidays for example, you'd like get a bit more broken arms and things like that or in the icy weather and also to make sure that any differences weren't just reflected in all types of injuries at particular periods of time. So, we broke it down to looking at infants, 1 to 10 year olds and 11 to 18 year olds and both jurisdictions saw an increase in rates of admissions for these for infants. So, you can see this was the kind of pattern an increase in England, decline in Scotland and then a decline for adolescents really that was much steeper in Scotland. So obviously, the kind of questions you have to ask yourself as a researcher is, "Is this a real change in incidence?" You know, "What might be the alternative explanations for these differences? Was there some kind of change in the way this is recorded in either of the jurisdictions?" We didn't think that's what it was so we began to think, "Well is it some impact of policy and policy changes that have contributed to this difference between the, from what are effectively two very similar jurisdictions?" So, you know, for example has the work of the Violence Reduction Unit in Glasgow actually been a contributor to fewer of these types of injuries presenting at our hospitals? So, it can give us some hints about policy. Some examples from other parts of the world, Melissa O'Donnell and her colleagues in Western Australia, they have had a very detailed, they have been collecting data, individual level on their child protection incidents, investigations and outcomes for 25 years. I think we're a little bit behind the times in terms of Scotland so we've only had since about 2010 the individual level return on child protection and at the moment it's quite hard to link it to anything because of the limited amount of data that is collected. But they've been able to link all the live births over that period and linked it to their child protection data sets and they're particularly looking around neonatal withdrawal symptoms and they found that there was a kind of peak, it peaked in the early 2000's and the rate started to drop and they think that's to do with some of the policies they had implemented around that time. But you can see, you know being able to have that 25 years' worth of date would actually open up quite a lot of rich research questions. It can also sometimes confound some of our assumptions so this is work that Emily Putnam-Hornstein did in the States, in California and they linked all the maternal birth records and their child protection data sets for this period of time because they wanted to really examine this question about racial and ethnic disparaties. So there is well recorded overrepresentation of black children, minority ethnic children in child protection systems and what they actually found was, when you controlled for socioeconomic and health status things like disability actually black children in lower socioeconomic groups were less likely to be referred to child protection services and had a lower risk of substantiation, that means when an investigation then substantiates that abuse or neglect has happened, and were less likely to enter foster care than similarly placed socioeconomically placed white children. It suggests that perhaps the socioeconomic depravation, because more black children live in socioeconomic deprived circumstances, that's why there is the overrepresentation rather than some kind of bias within the system. So, it's research that can offer us, you know, ways to really examine and disentangle some of these more complex issues. Just a brief word, this is a little project that I hope one day to be doing. It has been approved by Education Analytical Services and we're waiting for the data or the first stage of the data, and it's really to be a demonstration project to show what can be done with some of the data we routinely collect already on looked after children and obviously, placement stability is a significant issue for children. So we're doing this, what's called a sequential analysis, but we're going to compare these sequences of placements over the 7 year period because there is now quite a lot of good data there on looked after children and it's possible to link children who remain in the data sets over time or who exit and come back in and we're going to particularly take a sort of make a little cohort of children from this period because this was the time when the permanence orders came into play and look at this trajectory to see if that's potentially had any effect in terms of stability. So we'll do this first stage and then we hope to then be able to link it to attendance and school exclusion rates, you know, to see what comes first does the move, does instability come first and then attendance and exclusion is impacted or is there potentially, you know, a different direction of effect? So, as I say at the moment we're kind of just waiting on the data and we'll get started as soon as that's available from Scottish Government. So, we do have material already in place. This is work that's been done by, been led by Paul Bywaters in Coventry and it's really trying to disentangle and understand these issues of what you might term or what he terms child welfare inequalities. So, we know there are health inequalities. We know there are educational inequalities but there are also child welfare inequalities because there are huge variations, well essentially there has been a lack of systematic study since Bebbington and Myles in '89 and Lorraine Waterhouse and I did a little bit around the Children's Hearing system but what we do know is the majority of children who are looked after have backgrounds of social and economic deprivation so they might have a much higher chance of being looked after. And what Paul did in a first study was to look, to use all local authorities' data on children in need which includes in England children in need who are not looked after and not on child protection but are receiving services, children with child protection plans which would be our children on the Child Protection Register and those looked after by the local authorities and he used 13 local authorities in English West Midlands and what he linked to each child was their index of multiple deprivation score. Now that's a neighbourhood ranking so they obviously used the maternal postcode to measure the level of deprivation. So, it's not an individual measure of the individual child, it's but using the neighbourhood measure it's a kind of proxy essentially and he sorted it by national deprivation ranking. So, you know you have the top 1% of most deprived, least deprived authorities and then the 10% of most deprived areas and it's easier if I just kind of show you the diagram. So obviously essentially there was a very linear relationship between the greater the level of deprivation then the higher the rates of children being looked after. So, this is the 10 most, least deprived areas and that's the most deprived areas. And we really need to come to terms with this. We really need to tackle this. We really need to think about this, you know, because if there is this relationship, and what he also found is what he called the Inverse Intervention Law that a child who is in a very, let's say an area that's deprived and that neighbourhood sits within an overall highly deprived local authority and you compare that to an area, a child whose from a same level of deprivation in terms of neighbourhood but situated within a more wealthy overall less deprived overall local authority, I'm using very broad terms here, then you find that the child whose in the same area of deprivation but in the more wealthy, in the more least deprived authority has a higher risk of actually being looked after. And we need to try and disentangle that and think, "What does that mean?" because we don't really know. I mean what is the right number of children that should be in public care? You know if we reduce it is that a good thing? If we increase it is that a good thing? So, we really need to disentangle this a bit more and currently Paul is leading a project a cross UK study which is replicating this and there are also within that case studies of this inverse intervention law to try and tease out and think, "Well what is happening in these two very different areas and what impact is it having on rates of children being looked after?" So, I've left the link so you're... This is a very US thing and I haven't seen it done in the UK and it's essentially Barbara Friesler and her colleagues. What they're trying to do is they're using sort of special analysis and geography. They're trying to look at community level factors to try and operationalise them in some way to see what influence they may be having on child abuse and neglect and it would be possible to do something very similar here in Scotland. We do have some of this geographical data. So they were particularly looking, they were comparing geographical areas and what they found was that those areas which had much higher levels of off licences essentially rather than bars, seemed to be associated with increased rates of child maltreatment and whether that's because people were drinking more at home or, you know, rather than going out or, but as I say it's only associations but it is an attempt to try and think a more community level and this is quite an interesting study. They were trying to look at the impact of really drug dealing or drug markets and perhaps this kind of broad perspective can help us think, "Well where do we need to be placing resources, in which areas and what sort of time frame?" And what they found was that when there was very high levels of drug dealing or the drug market, there was fewer reports of child abuse and neglect. There was a sort of time lag and when the drug dealing seemed to move on to another area that's when you saw an increase in reporting of child abuse and neglect and so it was a kind of indicator of, if you have that very high levels of dealing then you maybe need to think about should we put additional resources into that particular area because we can maybe expect an increase in reports of child abuse and neglect. So, there was that kind of interesting time lag.
I'm just about at the end you'll be glad to know. It's a bit of a kind of rush through but I wanted to also talk about one final way in which administrative data is being used not just for research purposes. In New Zealand they're actually thinking of implementing this and it's a predictive risk model. You know at the moment we tend to use professional judgement or we use actuarial models, you know, of certain factors that we know are associated but are not predictive of child abuse. We, you know, we use them but they are obviously very open to bias of the assessor. This is really an attempt to try and create a system or a model to produce some accurate assessment of the likelihood that an individual child will be the victim of child abuse and neglect. It's an automated tool and it gathers and processes information from a range of databases. So, I think they use about 143 variables for the model to try and predict future outcomes. So, it's a kind of algorithm that uses dynamic risk factors at the start of any new benefit spell that occurred before a child's second birthday and a new spell in that sense is, could be a change of benefit or the arrival or departure of a partner or a change in the child's care giver. So, they've linked their child protection database to effectively what we would call our DWP database. So immediately there's a factor because it wouldn't include children whose parents are not on benefits. Now I'm not an expert on this type of model but to give you a sense of its predictive value, the way its assessed is something called the Receiver Operating Characteristic Curve. That's a kind of common measure for these types of predictive risk models. The area under the curve gives you some sense of the accuracy of the model. So, in a perfect model where you would only identify true positives and true negatives then the area under the curve would be 100%. In a chance model whereby, you know, you're true or false is by chance then 50% would be under the curve. In this model in New Zealand it's 76% so in terms of how you assess these kind of predictive models, these algorithms it would be called fair approaching good. I think it's sometimes said it's something similar to the effectiveness of breast screening. Now clearly that would be interesting research wise but I think it's very contested in terms of potential to use it for operational purpose because what I've been talking about is using the individual level data routinely collected for purposes of research, for purposes of policy assessment. This is really taking it a step further in saying, "This is about if we can identify children potentially at risk at an early stage, can we then offer services and support?" Obviously, there are issues about stigmatisation because there are significant ethical risks potential to marginalise an already marginalised population. We will obviously generate some false positives and this is all about using data without people's direct consent and I guess it revolves around these arguments about whether our Universal Services will be sufficient or whether we should be having some kind of targeting of services because you won't be able to make the Universal Services so intense that they can pick up and address all cases of child abuse and neglect. We know there is a lot of families who need support, but as I say there major, and if you want to look at it further we certainly have a chapter by Tim Dare who is a philosopher in New Zealand who's really looked at this in detail and it is used already in health research for example. There is a Sparra Project which is a project that actually analyses admissions to hospital of elderly people and predicts who's most likely to have a readmission and then alerts GP practices to that individual for them to try and be more proactive but I think it's very, very controversial but I didn't really want to give the talk today without also letting you know about this kind of work that is being done in New Zealand and, as I say there are great debate's there about whether it's legitimate, whether it's ethical and obviously brings us to these issues around targeted or universal services.
So, that was just a kind of run through of some of the advantages, some of the directions of research that might be possible using administrative data and this is just of course a little advertisement about the ESRC's big data network so those of you who are thinking about research or want to test out some of the policy directions or want to commission research then using this data that's already there then the ADRC's are present and correct and very keen to work with you. Thank you very much for listening.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License