We’ve never had the ability to place plan in SEND, now we can build new schools based on growth and needs forecasts, Ralph Holloway, Head of SEND Strategy and Innovation, ECC
What started as a conversation two years ago about the lack of a forecast for Special Educational Needs and Disabilities (SEND) turned into a great opportunity to analyse a new dataset, implement new analytical approaches within our team, provide SEND commissioners with insight they had never had before, and share our learning regionally and nationally.
Mainstream schools across Essex have had a trusted and robust forecast methodology in place for many years, but nothing existed for special schools and children with special educational needs and disabilities. Not surprisingly our conversation with SEND commissioners quickly turned into an aspirational list of the intelligence they would need to plan school places for children with SEND. And so, our work began…
To grant their wish we needed to be able to predict the level of need by these different characteristics. This would require something different to our usual forecasting approach. We are normally asked for a single high-level volume of demand, which means we can apply time series or causal forecast methods. The problem with these is that you can’t start drilling-down into the outputs by different characteristics – the only output is the total volume. So, for SEND forecasts we used a simulation method called Markov Chain Modelling. This simulates probable transitions at a child level, which means outputs can be permuted by any input characteristic (giving us not only total SEND demand, but also demand by locality, SEND need, school setting, age, etc).
The data came from the main local authority database (Capita), we tested the initial outputs, and set about building the model. Crucially, the first output was played back to commissioners in draft form, who were able to use their service knowledge to highlight areas and numbers that didn’t feel right. We investigated the reasons behind this, and together identified the latest correct information needed to feed into the model.
There were a couple of iterations in the first year of producing the forecast, and key to its success was the way we worked together with SEND commissioners to review and interpret the outputs and improve the quality of the information, and model.
I think there’s only so much data you can look at and analyse (especially when working remotely!) before you need to share with the subject matter experts who can add greater meaning and understanding to the patterns. It is a fine balance though, as you need to have gained enough insight to have something interesting to talk about, without being overwhelmed by too much data.
This work is pretty unique, we’ve had lots of interest from other Local Authorities and from Department of Education, Ralph Holloway
In the second year of the forecast, we were able to test for accuracy of the model by checking the forecasted numbers against what had happened and were really pleased with the high levels of accuracy. We shared the model with Eastern Regional SEND Commissioners, the DfE, and some other interested local authorities.
This forecasting work was a springboard for two other pieces of work we are currently involved in. It showed just how much growth had been happening within the Autistic Spectrum Disorder (ASD) need type, and how much this was expected to grow in future. So, commissioners are keen to gain a deeper understanding of the range of needs that sit beneath this broad ASD diagnosis to better plan for different types of provision needed.
Colleagues from my team experienced in natural language processing are analysing the content of case notes of children with Autism. This is beginning to give us a more in-depth understanding of the different types of need and support currently being provided to children with Autism in Essex. This work is also iterative with data outputs being shared with SEND Commissioners as the work evolves, adding value to the analysis outputs, and determining further areas of interest.
The third piece of work focuses on children with SEND in mainstream Essex schools. Not all children and young people with SEND need to be in a special school - about 60% of children with an Education and Health Care Plan, are in mainstream Essex schools. Realising the expected growth over the next five years, Commissioners wanted to understand the extent that mainstream schools’ SEND populations reflect the need in their local area, and how their SEND populations compare with neighbouring schools in the area. This can help identify schools, and communities of schools that might work together in future to support SEND children in mainstream schools.
This work was presented as maps to show the different SEND proportions across the schools, and the different areas of Essex. The geographical nature of this analysis lends itself to engaging with the SEND experts at an area level across Essex, and these sessions are being held within the four main Essex quadrants.
The work was done iteratively and in partnership and that was really refreshing, Ralph Holloway
We continue to have an ongoing conversation between SEND Commissioners and Analysts, as the data and intelligence on both sides evolves and new questions arise, moving us both forward in our understanding of SEND need in Essex and the provision and resources required to meet the need. And just as we’ve joined subject matter expertise with data analysts, we’ve also drawn upon the key data science skills within the team to deliver these different pieces of work, so it’s really been a collaborative, team effort!