"Some ideas are flash in the pan, lasting until the next thing comes along. Others stick around, slowly developing and spreading. Predictive risk modelling is one of the stickers"
Health Service Journal, Daloni Carlisle
Public sector predictive models they tend to be based on ‘personal data’ from individual service silos, and once developed tend to be used by these service silos.
For example, the Patients at Risk of Readmission (PARR) tool, developed by the King’s Fund, is used by the majority of Primary Care Trusts to support strategic planning and commissioning of A&E services.
Although most widespread in the health sector, these techniques and resulting models are clearly applicable in other business domains.
Despite the fact that these techniques have a key role to play in delivering smarter, more efficient public services their development and adoption is limited. There are a number of reasons for this; not least the fact that modelling processes require highly specialised statistical skills.
In addition, in the public sector, development of these models tends to require the involvement of academic institutions – in order to bring appropriate understanding of underlying risk factors and to provide reassurance to service teams.
Finally, public sector organisations do not necessarily have the tools and processes in place to make use of these models – whether to support strategic planning processes or the more effective targeting of services.
In this context, Xantura brings a unique combination of technology and statistical modelling skills, specifically;
- the Xantura Technology Platform (XTP) that delivers a cross agency view of citizens
- specialist statistical resources with a proven delivery track record in the public sector
- relationships with a number of academic organisations
- a set of end user applications that can help organisations deploy these models into strategic planning and operational service delivery processes
- practical experience of integrating these models into existing systems and processes to deliver business benefits
What we offer
Our Predictive Analytics service has been designed to help our clients understand how both existing and new predictive models could be applied to support improved outcome and efficiency objectives.
Our engagement methodology involves four key phases, described in summary below:
- Feasibility study
- Identification of key strategic cross agency theme(s) to be targeted
- Development of data sharing Memorandum of Understanding across agencies
- Evaluation of existing Data Sharing Protocols and assessment of data sharing attitudes
- Mapping of available data to the XTP data assets matrix
- Engagement of an academic partner
- Creation of modelling specification
- Development of target benefits case
- Outline specification of In-field analytical pilot
- Analytical pilot
- Creation and analysis of modelling data sets
- Creation and testing of predictive model
- Business case validation and refinement
- Development of detailed Data Sharing Protocols
- In- field analytical pilot
- Set-up secure data transfer and data sharing protocols in the XTP
- Integration of predictive model into the XTP
- Calibration / configuration of the model in the
- Set-up end user XTP applications and integration
- Test configured XTP environment
- Conduct in-field pilot
- Model management / refinement
- Cyclical review of model performance
- Integration and testing of additional risk factors
- Tuning model to match local client characteristics
In summary, delivery of a smarter more efficient public sector will require a fundamental rethink about how we can securely and safely use cross agency data.
Our predictive analytics service has been designed to; improve the quality of models that can be developed, improve the testing and deployment of these models and finally to make sure that these services are delivered in a secure way that doesn’t raise civil liberty issues.