On Monday 10th December over 100 social care practitioners, academics and data experts gathered at Coram’s new Queen Elizabeth II Centre to share promising innovations and ideas about using data visualisation to improve the support for children who need help from local authority children’s services.
Coram and its partners, The Alan Turing Institute, The Rees Centre and Kent County Council, showcased its data visualisation prototype which mapped how children move through different pathways when they enter care. This visualisation will be publicly launched in 2019. Complementary work with North Yorkshire County Council and the University of Bristol linked care data with court proceedings data and used this to visualise children’s journeys pre and post proceedings. The development of the project was also supported by ideas and research from Lancaster University and University College London.
The project was made possible with funding from the Nuffield Foundation.
This blog outlines our data visualisation journey and our key learning from the project. More about the data visualisation event can be found on Coram’s Twitter feed (@Coram).
How the project began
The Coram data visualisation project has been an unpredictable but rewarding journey.
Coram began from the insight that using a data driven approach in its consultancy work with local authorities has been useful in helping children’s services managers to understand the experiences of children and identify ways to improve services. We believed that giving these managers a better handle on their data offered the prospect of a deeper understanding and more improvement.
We were also aware that university-based researchers were providing findings about children’s care journeys that local authorities could not easily replicate locally due to the way that their data was held.
One solution to both of the above seemed to be through better data visualisation, although which data and how to visualise it had yet to be determined.
What we did
We started off by bringing local authorities, data scientists and university-based researchers together and sharing what we knew and how we understood the problems and opportunities. While this led to interesting discussions, the next step was not immediately obvious. A key insight came from The Alan Turing Institute’s discussion with local authorities – that people like and understand data portrayed as maps.
Further development led to the creation of an animated display of individual children around the entry, service and exit points of the child’s journey through the system. This showed flow over time, depicting speed of progress and highlighting those making repeated circuits.
|An overview of the bespoke interactive map of Kent’s front door and assessment data – the power of visualisation|
Our key learning from the visualisation project
In a scientific process, exploration and description of data are the first steps on a journey to form hypotheses about why events occurred, which factors drove them and which events were largely irrelevant.
Once this understanding is in place we can start to make predictions – the accuracy of which can be improved upon and adjusted as new evidence is taken into account. One example of a prediction is that if a child experiences four or more Adverse Childhood Experiences events then there is a 16% probability of them abusing alcohol as an adult (Felitti et al, 2004). An effective predictive model helps us to identify the decisions that have the most chance of achieving the outcomes being sought.
Visualisation makes a contribution to all of these stages. However, a good data visualisation depends on knowing what questions you want the visualisation to answer and designing it accordingly.
Picking the right question depends on who you are. Researchers may pick a research question; service managers may be more likely to pick an organisational performance question that can guide their decisions today and in the near future. For managers of a service, hypothesis development is often informal and based on observations and experience. For them ‘administrative data’ is not a pejorative term; it is the data that is needed for many strategic decisions.
Competing predictions may be that if a child is re-referred then a) subsequent assessments will generally take less time because a lot is already known or b) each will take more time because the return may be an indication that the situation is more complex or intractable than it first appears.
The tool developed in the project enabled managers to see data as an animated flow. It enabled managers to quickly form and abandon such hypotheses, to see where children’s journeys could be improved and identify areas to look at in more detail.
The university researchers also showed that treating data longitudinally shows the true magnitude of issues such as the proportion of children who have been ‘Children in Need’ at some point before their 18th birthday. Just looking at annual totals can be very misleading and lead one to underestimate how many children are affected.
|An analysis of North Yorkshire County Council’s children & young people’s data, applying outcomes to care proceedings for children study method.|
Linking of data, even within an authority, such as proceedings and care data can give a different understanding of children and young people’s trajectories than looking at each data set individually.
However, when we link data we must remember that linkage error is not random and is worse for certain groups. This can lead to an underestimate of adversity within these groups.
What does population-based, longitudinal data tell us about services?
A child life course perspective – the power at a local and national level.
Finally, the limits on what we can visualise underscores the fact that interventions that are provided to children and families are often described in narrative form only and are painstaking to extract. Greater use of categorical data to describe services and outcomes would make nuanced analysis more feasible for local authorities.
Coram would like to express their gratitude to all stakeholders involved in the project as well as those who attended the seminar on the 10th December. Alongside the experts, academics and local authority representatives you see above, we are thankful for the involvement of Alastair Lee (East Sussex County Council and Chair, National Children’s Services Performance and Data Management Group), the work of Prof Harwin and Dr Alrouh of Lancaster University (whose presented work will be published in due course), as well as our colleagues at Brent County Council, Public View and Perfect Ward for demonstrating their innovations in the sector.
If you are interested in what you have read above and would like to contact Coram-i then please use the following page to contact us.
Felitti, V.J. (2003). Ursprünge des Suchtverhaltens – Evidenzen aus einer Studie zu belastenden Kindheitserfahrungen. [The Origins of Addiction: Evidence from the Adverse Childhood Experiences Study]. Praxis der Kinderpsychologie und Kinderpsychiatrie [Practice of Child Psychology and Child Psychiatry], 52:547-559.