Offering a holistic and preventative model of care together with a move to integrated care systems are fundamental elements of the NHS Long Term Plan. A key requirement for this is collaborative data-sharing across organisations, helping to produce a comprehensive picture of current and future needs and informing more insightful care planning and commissioning. None of this is straightforward. Whilst in reality health and social care services operate along a continuum, they’re often still viewed distinctly.
In 2016 City of Wolverhampton Council began working with PredictX, Midlands and Lancashire Commissioning Support Unit (CSU) and Wolverhampton Clinical Commissioning Group (CCG). Combining pseudonymised NHS Secondary Uses Service (SUS) data with Adult Social Care data from assessments and service packages, the initial project built a series of dashboards for use across the system. These dashboards displayed key metrics such as:
Delayed Transfers of Care (DTOCs)
care home and home care service provision
This enabled staff to have an informed view on where the next pressure points would be in the system and address these proactively.
Building on this first stage, with funding from an NHS Digital Social Care Pathfinders Programme grant, the next project phase involved applying predictive analytics to the data. The first proof of concept was predicting the incidence of hospital admissions from A&E attendances. The second, involving social care specifically, was predicting the type of care package a patient would need when discharged from hospital. The predictive model for this second case performed less accurately, due to the fact that there was less social care data available for the model to learn from. However, as the Council looks to combine social care data from across the region, this modelling will become increasingly viable.
With work ongoing to increase the size of datasets, the project has moved away from predicting individual care needs for now, to creating profiles of care service users. This helps inform strategic planning by meeting a population health management approach of stratifying residents in a geographical area by clustering against a series of characteristics.
The work has involved analysing the data of 3,000 domiciliary care users, taking into account:
the services they use
touchpoints they have with organisations in the system
diagnoses of ill-health
long-term conditions they live with
and then layering on top of this socio-economic data such as indices of deprivation.
This has generated seven key profiles and unearthed an insight - amongst
several others - that there are residents with long-term health conditions who do not access many services, whilst there are other residents with no recorded conditions who access multiple services. The richness of data makes it possible to drill down further, ask why and re-organise services to address this.
The next phase of the project being developed is working with local Primary Care Networks to investigate whether there are clusters of residents with a certain profile around specific GP practices, and if so, think about how it might be possible to do things differently to cater for them more proactively.
In light of COVID-19, this will need to be done carefully. Profiles are based on data illustrating patterns of demand over the last 2 years. The emergency response to COVID-19 by health and social care systems will have altered these previous demand and support patterns, both in the short and long term, and has implications for how historical data is used in the future.
It’s unusual for a data-driven health and care integration project to be driven by a social care service, but the Wolverhampton experience has shown that this is possible. Key to the project team’s success has been the buy-in of a senior executive sponsor, who can in turn bring on board leaders from across the Council and partner organisations.
The team has reflected that it could further enhance the project’s credibility by bringing some data science expertise in-house. In particular, this would help to bring independent validation to PredictX’s data science work.
Data sharing such as this, which is not for the purpose of direct care, has had its challenges. Where a project is innovatively combining health and social care data or linking any additional health-related dataset, processes can be lengthy and complex. But Wolverhampton’s pioneering work has meant that it can now share its experience with other local authorities.
Sharing for the purpose of scaling also applies to the predictive analytics tool itself. The next step is to roll out the tool across the Black Country and West Birmingham Sustainability and Transformation Partnership (STP), before extending across the West Midlands.