This document discusses a partnership program in Essex County that uses predictive risk modeling and data from multiple organizations to identify communities at risk of poor outcomes. Specifically, it examines a pilot project that aimed to predict which children in the town of Vange would not be school ready. The project used anonymized data from the county council, city council, and police force matched at the household level. The results identified neighborhoods in the city of Basildon most likely to have children unprepared for school. The model correctly identified half of the children not known to services as being at risk, double the rate of random chance. Future work will use these insights to target interventions and develop additional risk models to tackle other challenges through early prevention.
4. What’s up in Vange…
What's different from any other community engagement
exercise?
We’re combining what they tell us with predicative risk
modelling.
5. A bit of back ground …
A partnership programme of work led by ECC
A pilot programme – testing, learning and building whole-
system capability
Using data to shift from I think to I know this is a problem
Using data to improve outcomes and reduce demand and
cost
Essex Data is a programme that seeks to provide the capability and capacity
to safely share and match data across Essex partners so that it can be used
to provide insight and to predict risk to inform where and how to direct
resources.
6. Essex Data is made up of….
A programme team and
project governance
A safe data sharing
platform
A series of predictive
risk prototypes
7. Essex Data is nationally innovative….
Focus on predictive
analytics, not just
operational flows across a
system
Using whole-system data not
just data from one
organisation or one sector
8. School Readiness….The Ask
Can we use insight from ED to
triangulate with other insight to
commission differently to increase the
number of children who are school
ready in Vange?
Can we predict which
children in Vange will not be
school ready on starting
school in reception?
“In 2013 50% of children
in Vange were not ready
for school”
Predicting communities in Basildon who are most at risk of not being school
ready.
9. School Readiness….What we did
• Data from ECC, Basildon BC and Essex Police which is
pseudonymised at source and matched at household level.
• Results are displayed at output area level and will enable targeted
intervention at a community level which was considered the right
level for this work.
• Using predictive risk models to triangulate with other insight.
11. School Readiness….The Results
• In Basildon those who are known to services are 1.5 times
more likely to not be ready for school
• This leaves a sizeable cohort (1,539 over the four year period
of the data) of school starters that are not known to services
and that were ‘not school ready’.
• The risk model would have correctly identified half of these at
‘risk of being not ready’ – which is 2 times better than random.
• The model is 75% accurate enabling resource to be targeted to
those communities that would benefit from it the most.
• Financial stability & presence of anti-social behaviour are the
most significant social and environmental factors.
12. What’s next…
• Using the data to make changes in Vange
• Developing further risk models that help us to tackle other key challenges
through providing insight to enable earlier intervention and prevention
13. 1. Finding the
right problems
Why is this problem
important?
What actions are available and are
they implementable?
Can data be used to solve
this problem?
What data can we use, is
the data available?
Some key lessons learned so far…
14. Some key lessons learned so far…
2. Ethics issues - balance the duty to protect with the duty to
prevent
How does the public benefit?
How intrusive and identifiable
is the data you are working
with?
How automated are the
decisions?
What is the risk that someone
will suffer unintended
consequences as a result of the
project?
Do the public agree with what
you are doing?
15. Some key lessons learned so far…
3. 20% about the data and 80% about the culture, politics and
organisational capacity