London councils are using data analytics to predict which children are at risk for neglect and abuse, allowing them to act before crisis occurs and prevent more costly, extensive interventions. Developed by technology company Xantura and trialled by councils including Hackney and Thurrock, the model analyses various data sources, including school and health records, to judge families’ risk scores. Around 80% accurate, it identifies and alerts social workers to those who need extra support.
Results & Impact
One local authority has identified nearly 400 additional families to receive support, and more than 80% of risk alerts in Hackney have been deemed valid. Councils are expected to save over $910,000 for early targeted interventions, $160,000 by replacing human-conducted screenings with an automated system, and $193,000 for improving access to multi-agency data.
Xantura, London Councils, London Ventures
Xantura’s “Early Help Profiling System” (EHPS) translates data on families into risk profiles, sending monthly written reports to council workers with the 20 families in most urgent need of support. Various statistics are used from multiple agencies, including information about school attendance and attainment, families’ housing situations, and economic indicators. With three years of time series data, the modellers have used the first two years to build the algorithm and train the model, testing it against the third year of data. This helps to weigh the variables and continuously test the model, which keeps improving as more data is fed into it. Each month, social workers are given a manageable list of 20 families most at risk in the area, with whom they are not currently working. The risk score created for each family comes in 40 bands, and is put through a “natural language generator” to give summaries outlining why each family passed the risk threshold. Social workers then use this information as they see fit. Early intervention usually consists of extra support from a children’s centre or a health assessment. The software's development has been funded by London Ventures, a program run by EY and London Councils designed to improve public services innovation.
London and the South East of England
Cost & Value
Xantura’s software has cost around $1.25 million to develop since the project was launched in January 2015.
Pilot running since 2015
The project requires a significant change in how councils run their children’s protection screening, and as a result, it has taken time for councils to trust the model. Meanwhile, the statistical process is still developing, which means it is not yet completely effective or accurate, though more data and time should improve performance and accuracy. The data itself is also not continuously updated: for example, school information is only refreshed every half term.
Similar child protection projects are being developed in Pittsburgh and Florida. Predictive analysis is also already in use by other public bodies such as the National Health Service and Department for Work and Pensions.
London’s boroughs are using big data to predict which children are at risk of abuse, and intervene before anything happens.
With a statistical model developed by Xantura, a technology company, social workers in participating local authorities are given a monthly report summarising the children most at risk of future violence or neglect.
London Councils, the authority which oversees London’s 32 boroughs, is currently trialling the model. In one London council, the system has already identified nearly 400 additional families to receive support. Early action helps to prevent the situation from escalating to a stage where more extensive, costly interventions are needed. It is therefore suggested that the model could save a council more than $910,000 per year.
“You actually don’t have to prevent that many children from going into care to make quite a significant saving, given that the costs per child per annum of a child coming into care are in the order of 50-odd thousand pounds [$677,000 USD] depending on the nature of the placement that they are in,” said Steve Liddicott, Interim Assistant Director of Children and Young People’s Services at the London Borough of Hackney.
Participating councils, including Hackney, Thurrock, Newham and Tower Hamlets, have been using the software since 2016.
Xantura’s “Early Help Profiling System” (EHPS) uses statistics from multiple agencies, including information about school attendance and attainment, families’ housing situations, and economic indicators, and turns them into risk profiles for individual families.
With three years of time series data, the modellers have used the first two years to build the algorithm and train the model, testing it against the third year of data. With the model being continuously tested, the council could “refine the software that Xantura have developed in terms of the amount of weight that’s given to different indicators,” said Liddicott.
As a result, each month the system provides the council’s team of social workers with a list of 20 families whose risk score indicates they are most in need. Importantly, the model flags up families they would not otherwise be working with at that time.
The onus is on social workers themselves to apply their professional judgement to the information as they see fit; the system is not devised to replace that process, but to inform it. Early intervention usually consists of extra support from a children’s centre or a health assessment.
The alerts come in as reports, as opposed to simply risk scores, and include factors which might raise flags about a family, such as whether it is in debt. This makes the information easier to process for social workers.
Liddicott explains that, during the development process, Xantura “developed what they call a natural language generator, which is taking data and then making that into something sensible than you can read, rather than just being a series of numbers and indicators”.
The other aspect which helps social workers is the manageable number of 20 families they are alerted to each month. By comparison, in the current system, each council currently tends to receive anywhere from 8 to 10,000 referrals every year. The software therefore makes it easier to prioritise early intervention help where it is most needed, which is especially important when resources are scarce.
The accuracy of the alerts has taken a while to improve, but the database is continually updated and the model re-run, which means it is constantly improving. In Hackney, one of the participating local authorities, more than 80% of the risk alerts have been deemed valid. Another challenge is that the data is not continuously updated: school information, for example, is only refreshed every half term.
The eventual aim is that the screening of at-risk children can become fully automated, and that data can be shared across all professional agencies working with children and adults. Hackney Council will run a pilot to share the data with doctors in fall 2017. It is expected that improving access to multi-agency data could save a council $193,000 per year. Meanwhile, full automation of family screenings could save $160,000 because less staff would be needed.
To implement the system, it would require a significant change in the operation of children’s protection screening, which has presented a challenge in encouraging councils to use the model. Its development has been funded by London Ventures, a program designed to improve public services innovation, run in partnership by London Councils and EY. Xantura’s software has cost around $1.25 million to develop since the project was launched in January 2015.
If councils choose to continue using the software, funding will have to come out of their own budgets. The potential for long-term savings, though, may convince more councils to embrace this new big data strategy.
(Picture credit: Flickr/Kat Grigg)