Childrens Social Care demanding modelling

Full Application: Not funded at this stage

The number of Children in Care (CiC), both in Suffolk and nationally, has been rising, whilst expenditure has been rising at an even faster rate. Conversations with practitioners, backed up by the data, also reveal changing characteristics in the CiC population, all of this means that current commissioning models need to be reviewed and amended to reflect these changes. This project aims to help us better understand the impact that changes such as budget, case mix, population or timings have on the number and characteristics of Children in Care.

The project is to develop a proof of concept model with associated guidance and thinking, that will allow users to model inputs such as budgets (resource), case mix (age, gender, reason for care), population and destinations (foster care, residential home etc.) and run scenarios to see the impact that this has on the key outputs including overall case mix and hence capacity requirements, and cost.

The proposed model will use up to 10 years of historical data and will be built to reflect the standard phases of the social care service and will be calibrated for Suffolk as a particular test case. Used in other authorities we expect this model, and the associated documentation, to serve as a robust design pattern for similar needs.

The project will be delivered by Suffolk County Council (SCC) and Mastodon C working together to develop a proof of concept model.

Discovery stage – onsite workshop with senior SCC stakeholders to: define what they already consider to be drivers of CiC numbers; refine our shared understanding of the Suffolk system; confirm what kind of information is actionable for the Director of Children’s Services and what kind of accuracy is needed in that information; confirm any key scenarios they’d like to explore with a model.

Data acquisition and initial exploration

  • SCC analysts and Mastodon C to agree pseudonymous data extracts to be provided from Suffolk’s social care databases, once required information governance is in place.
  • Undertake initial exploratory data analysis once all relevant data has been sent to Mastodon C
  • Extended discussions with SCC to create an overall map of the system to be modelled and how the social care data maps onto it, create basic visualisations, identify any unusual or interesting features of the data, establish whether data quality and availability can support modelling.
  • Deliver a first report on data contents back to SCC

Chart A below is a draft schematic of a possible flow of CiC and some of the elements that may impact on the process.

Develop proof of concept model

  • Determine an appropriate simulation/machine learning approach to meet requirements, given the available data
  • Develop and test POC technical model to test the feasibility of producing numbers as desired including a ‘front end’ which facilitates scenario modelling
  • Document the budget, casemix, and population outputs of the model under different scenarios

Present findings and agree any next steps

  • Present outcomes of development project to SCC stakeholders at another workshop
  • Hand over a writeup of the model and data outputs, to analyst team
  • Agree any next steps

Chart A – Draft Schematic of CiC flow –!AvWShWo8Jytfg_oiqFVENRo_QviYYA

Discussions with other local authorities shows that they are struggling with the same issues of increasing numbers of CiC, more complex needs and upwardly spiraling costs. Yet despite these pressures, and other routine benchmarking of rates and process measures, SCC believes that data and intelligence is underdeveloped in this area. Reporting is limited to ‘what happened’ with little attempt to explain why things happened, or what the implications for the future might be. The use of data to understand this area is hamstrung by the vague and limited coding of reasons for being in need, or coming into care, which stand in stark contrast to the copious amount of coding and data generated by a child staying in hospital, for example. SCC believes that by trying to do more with the data, and by publishing our attempts, it will encourage new approaches and a better understanding of what is possible – which may in time perhaps lead to significant changes in recording and coding which would, in turn, help further the analysis. These issues are national and outside the control of any one Local Authority, but by doing work such as this bid, we hope to be able to highlight them and push for change.

In 2016/17 net current expenditure on Children in Care in England was £4,017,584,000. This was over 47% of the total children’s social care expenditure. For Suffolk the figure was just over 44%; 5% of Suffolk County Council’s total expenditure is now spent on Children in Care (approximately 900 children out of a total population of 750,000 people). (Source: MHCLG, RO3 and RSX data 2016/17).

Due to the initial discovery stages that need to be undertaken on this project, it is not possible to firmly quantify the potential savings in either cost, time or resourcing. But if we saw a 1% decrease in costs, this would equate to a saving of over £406,000 in Suffolk and a national saving of £40,175,840.

Whilst completing work on the CiC Sufficiency Plan (capacity plan of the CiC cohort), it became very clear that the initial high-level analysis of trends, using mainly public health data at population level, has not helped greatly with developing actionable insights. Decision-makers would like greater understanding of the drivers of net CiC numbers and costs and how these numbers, case mix, and budgets might change in future, based on more sophisticated simulation/longitudinal and probability-based approaches, similar in some ways to a multi-state Markov model. Issues such as the age and gender of the child coming into care have an impact on how likely they are to remain in care, for how long and in what setting. Between 2014 and 2018, Suffolk has seen a 50% increase in the proportion of CiC that are aged 16-17.  Therefore, by being able to model our current and possible future casemix scenarios we will be able to analyse the likely future impact on capacity and need of each month’s care starters and leavers over the longer term.

Analysis of CiC records has given us an indication of how long children remain in each stage of the process. This helps with resource planning of the children in the system but also lead to a realisation that monitoring data at the start and end of a period did not take account of the ‘churn’. For example, in 2017/18 we had 760 CiC at 01/04/2017 and 790 at 31/03/2018. At first glance this looks like an increase of 30 children, but in fact during the year we had 315 new children coming into care. Understanding this will help in the planning of resources.

As CiC numbers and costs are rising, we need to explore whether being able to use approaches like predictive modelling can really help us to manage demand and assess the impact of interventions at different points within the system.   A system needs to be able to show what-if around demand, number of children, care packages available, placements, timing of interventions. 

The latest DfE Children Looked After data for 2016-17 states “The number of looked after children continues to increase; it has increased steadily over the last nine years. At 31 March 2017 there were 72,670 looked after children, an increase of 3% on 2016”.

The increases that we have seen in Suffolk is also reflected nationally across England. This project will not only help in the understanding of factors that impact on children coming into care but will also allow local authorities to model scenario’s through changes in budgets, case mix, population and destinations.

Net expenditure and number of CiC

This shows the net current expenditure on children in need between 2012/13 and 2016/17, this increased by 33% over that period, whilst the number of CiC increased by 8% over the same period. Between 2012/13 and 2017/18, the number of CiC increased by 15%. Undertaking a simple unit cost of net expenditure divided by CiC, this shows nearly a 23% rise in cost, per child in care.

 By the end of the project we will have a;

  • Proof of concept model which will allow us to run various scenarios to see the impact of changes in budgets (resource), case mix (age, gender, reason for care), population and destinations (foster care, residential home etc.)
  • Schematics showing the flows of CiC, including determining average and min/max length of times between stages and how these vary depending on the age of the child.
  • An outline benefits case comparing the costs of the modelling/analytical work undertaken with the possible savings in CiC costs, which could come from intervening with particular groups differently; or perhaps from business cases for capital development of facilities, as much of the spiraling in costs is due to supply side limitations rather than demand side pressures.
  • User research report into the work undertaken, key findings and next steps.
  • Code for the model which will be made open source by our partner, Mastodon C, and hence can be used by the wider LA sector at no further cost to them. SCC will be happy to support other Authorities who wish to adopt the model within SCC’s technical capabilities.

This will give Suffolk County Council the ability to make better decisions on how it spends its diminishing resources by understanding the interactions between factors and the positive or negative impact that these change can have.

A beta project would look at developing this work further through;  

  • Testing the model with one or more other local authorities to discover which elements are reusable and which may need refinement before it becomes accessible to all
  • Building more detail into the model in areas that emerge as important (e.g. the referrals process)Broadening the scope of the model to include other stages of the care system e.g. children passing into adulthood
  • Developing the ability to model more complex scenarios
  • Developing more complex cost accounting features


Ultimately the modelling will impact on the children of Suffolk, that either could or do become children in care and it is imperative that whilst using data and modelling that this is never lost. The users of the model will be a mixture of senior social work managers as well as data and financial experts, who will model a range of scenarios and understand the impact that these will have on children in care.

The project plan outlined in question 2, shows that during the discovery stage, onsite workshops will be undertaken with key stakeholders of Suffolk County Council. This will look at getting a better understanding of the drivers of CiC numbers, understand the processes that SCC has in place. These workshops will be in addition to the extensive involvement with staff that SCC already has in place, for example monthly Directorate Management Group meetings, weekly Directorate Management Team meetings, monthly written communications and a wide network of team and service manager meetings. All of these can be used as required to engage users and seek feedback on the model.

There will also need to be explanations on what the model can and cannot do and the levels of accuracy at each stage of modelling. Once the model has identified areas that could have a major impact on CiC, it will be up to social workers to understand and write policies that bring about the changes that the modelling has shown have the largest or most desired outcomes for CiC.

The users would like to be able to assist senior decision-makers on the drivers that impact CiC numbers and costs and how these numbers, case mix, and budgets might change in future. This will allow them to better plan resources requirements and commission the correct services.

The project will not require any additional support from MHCLG. If the findings of the project are useful and widely applicable, we will circulate those through existing DCS networks, but MHCLG could perhaps be of help in brokering conversations with the Department of Education if changes and improvement to coding is a key finding.

This project has never been granted any funding and no funding has been applied for or received during earlier stages of the work.