Wales to roll out predictive risk tool

Health Solutions Wales is set to roll out a new tool to predict patients at risk of emergency hospital admission on 1 April.

The tool, called PRISM, allows GPs to log into a web-based system where they can view their patient population stratified for risk of emergency admission to hospital in the next 12 months.

They can use the data to generate lists of named patients and work with them to help prevent admission. GPs can also use the data for benchmarking against other practices. The main aim is to support the care and management of people with long term conditions.

The roll out marks the end of a four-year development period since the Welsh Assembly approved an all-Wales risk stratification service in 2006. It was designed by Bupa Health Dialog and Informing Healthcare with HSW.

Leo Lewis, now on secondment from Informing Healthcare to Carmarthenshire health board where she is project director for the chronic conditions management demonstrator, said: “It is not dissimilar to the combined predictive risk model but is based on fewer variables.”

The model uses data from the QOF, Patient Episode Data for Wales (equivalent to HES in England) and deprivation indices from public health as its source data. All data is pseudonymised and then analysed for risk.

The risk-stratified data is then fed back to PRISM where GPs can log in and marry it up with patient demographics to generate lists of named patients who are at risk of emergency hospital admission.

They can drill down into the data to look at different levels of risk and to benchmark their populations and practices against others. Lewis said: “It gives users a tool to identify cohorts of people that they might work with differently.”

Gareth John, head of information products at HSW, said: “We tried to use existing infrastructure as we developed it. For example, we have used our Data Switching Service to enable practices to feed their data us automatically.”

The tool has been successfully piloted in Cardiff, Carmarthenshire and Gwynedd, she added. “Our approach has always been to do a proof of concept with a view to roll out. From the outset it was felt that the model would be able to support us.”

Lewis predicted that uptake by GPs will be variable. Some were concerned that PRISM would be used for performance management, she said, while others were concerned about data sharing.

She said: “We have listened to their concerns and to practice managers and agreed that we needed to build up trust gradually to ensure that the tool is used as a tool. It is not a panacea.”

Lewis and John say the tool will be refined further over the next year to include data from A&E and from social care.

Opinion and analysis: Read more about the history and future of predictive risk modelling in our feature, Predict and survive.