Following the Government's decision not to upgrade existing predictive risk models, this guide aims to help commissioners now tasked with choosing risk tools from an open market.
Predictive risk models are an important part of the strategy for managing patients with chronic illness. They are used to identify patients most at risk of future unplanned hospital admissions, and who may benefit from preventive measures such as that provided by community matrons, or through admission to a community-based ‘virtual ward’.
Such cutting-edge statistical methods have been found to be more accurate at predicting who is most likely to be admitted than other ‘case finding’ approaches including threshold models and clinical opinion.
Despite this, in August 2011 the Department of Health announced that it would not be commissioning a national upgrade of existing predictive modelling tools used by the NHS, such as the Patients at Risk of Re-hospitalisation tool (PARR++) or the Combined Predictive Model (CPM). Instead, the current policy promotes an ‘open market’ in terms of suppliers of risk tools, making it unclear whether in future predictive models in England should best be procured or built at a local, regional or national level.
The financial situation in the NHS makes it even more important that tools such as risk prediction are used to target investment to reduce the need for hospital admissionNuffield Trust Director of Research Dr Martin Bardsley
This user-friendly guide is intended to help commissioners now tasked with selecting such tools from an open market. The guide advises commissioners to consider a range of factors when choosing whether to ‘make or buy’ a predictive model, including the outcome to be predicted, the accuracy of the predictions made, the cost of the model and its software, and the availability of the data on which the model is run.
Although there are opportunities for improving the health status of patients with complex needs while making net savings for the NHS, the evidence for hospital-avoidance interventions is patchy and therefore robust evaluations should be built into any proposed local strategies.
This short guide aims to explain some of the key principles involved in procuring a predictive model and to provide a guide for people who might be new to this field, or who are engaged in selecting a predictive model for their organisation. We believe it will be particularly useful to clinical commissioners, public health specialists and others involved in the redesign of services for patients with long-term conditions. We will endeavour to keep this resource updated as new developments emerge.
The Nuffield Trust has considerable experience in the application of predictive risk modelling techniques in the UK. This includes work on the development of case finding tools; a feasibility study of models that predict future use of social care; work on person-based resource allocation; and a range of national evaluations of interventions to reduce hospital use, including the Whole System Demonstrator Project, integrated care pilots and selected Partnership for Older People Projects (POPP) and Virtual Wards. Further information on this work is available from our dedicated area of work page on predictive risk.
Lewis G, Curry N and Bardsley M (2011) Choosing a predictive risk model: a guide for commissioners in England. Briefing. Nuffield Trust.