Risk prediction: what every CCG needs to know

Blog post

Published: 17/11/2011

Keen readers of this blog will already know about the importance of risk prediction in health. As my colleague, Dr Geraint Lewishas pointed out: “neither doctors, nurses nor case managers [are] able to predict which patients [are] at highest risk of readmission to hospital.”

So, if the NHS is to target effectively the ever increasing rate of emergency admissions, it is clear that it needs some help from predictive risk tools.

When the Department of Health announced in the summer that it had decided not to commission a central update of its two predictive risk tools (PARR++ and the Combined Predictive Model) but instead leave it to the market, it effectively added another responsibility to the list of responsibilities of Clinical Commissioning Groups (CCGs). 

In-line with the direction of current policy, the development and procurement of predictive tools are to take on a distinctly local flavour.  The theory is thus: multiple providers will appear and develop numerous innovative tools from which savvy commissioners demanding high quality, high accuracy and low prices will choose.

But will this happen in reality? Is this something that CCGs need to grapple with or is it simply another thing to add to the ‘shopping list’ of things to outsource to suppliers of commissioning support?

Experience tells me that this is definitely something that CCGs need to engage with. Having witnessed numerous examples of primary care trusts (PCTs) developing a whizzy model but then struggling to get GPs to use it because it either doesn’t predict what they need it to predict or because it’s just too difficult to use, I think it is essential that the predictive risk model is embedded in a wider strategy of population management.

Whilst the contracting and development of the model may easily be outsourced, it is important that CCGs are clear about what they want the model to predict and how it fits in with their overarching strategies. In other words, for an effective market to develop, commissioners need to know what they want. This requires CCGs to have some knowledge of the risk prediction world.

To assist CCGs in getting to grips with risk prediction, the Nuffield Trust has published a simple guide which sets out the considerations a commissioner needs to make in choosing a model. It stresses that choosing a model is not just about cost and predictive power.  Indeed, a whole host of factors need to be considered before the market is even approached.

It is essential that the commissioner first establishes the desired outcomes of the population management strategy and defines the population group it wants to target. It is also crucial that there is clarity about how effective the intervention is likely to be.  After all, key to a successful long-term conditions strategy is the efficacy, equity and cost-effectiveness of the intervention. Data availability is another crucial factor that will play a central part in the decision.

One of the first decisions a CCG will need to make is the scale at which a model should be commissioned. In the past, models have typically been run at PCT level but many CCGs will be smaller units. Commissioning a model jointly with other CCGs might offer economies of scale but CCGs will need to weigh that up against the potential loss of local flexibility.

Only once such issues have been addressed will we see the emergence of savvy commissioning of risk prediction models within an innovative market place.

This article has also been published on GPonline.com.

Suggested citation

Curry N (2011) ‘Risk prediction: what every CCG needs to know’. Nuffield Trust comment, 17 November 2011. https://www.nuffieldtrust.org.uk/news-item/risk-prediction-what-every-ccg-needs-to-know