Sussex HIS has developed a risk prediction tool - the Sussex Combined Predictive Model (CPM). This tool predicts future risk of admission within a given population.
Sussex CPM is similar to national tools, but is configurable to work with local data available – for example hospital, community and/or primary care datasets.
The system is used by a number of Trusts in Sussex and is well liked by users. In light of this we now offer a national hosting service of Sussex CPM, as a flexible NHS alternative to ‘off-the-shelf’ versions of the Combined Model.
Risk prediction is about identifying patients mostly likely to be admitted into hospital in the future.
For every person in a given population we calculate a risk score between 0 and 100% - this is their risk of being admitted for an emergency chronic admission over the coming 12 months.
Clinicians can then target high-risk patients for proactive treatment at home, reducing the likelihood of hospital admission.
Risk prediction is about looking at patterns in patient care. We can identify key ‘risk factors’ for a person by looking at previous healthcare given and then use this information to predict those patients most likely to be admitted in the coming year.
Our risk tool looks at hundreds of healthcare factors that affect their risk of admission. Each factor has a different weighting which, when combined, give us a single risk score for each person.
Because the system is jointly owned and developed, we will continue to develop the tool to meet your needs, under the direction of the Sussex CPM user group.
Once authorised, any clinician can access and interrogate risk scores for their practice(s) using a simple, secure web-based interface:
The risk scores are currently updated monthly, although you can access the system as often as you like. You can also view a history of risk scores over time.
An independent evaluation of the Sussex CPM tool by East Sussex Public Health department showed that with hospital data alone the tool provides a level of predictive accuracy equivalent to national tools such as Combined Model.
Any missing, incomplete or wrongly-coded activity will impact on the accuracy of the risk score produced, so full, accurate data recording is, as always, vital for accurate decision-making. Also, NHS number needs to be correctly recorded on all systems to ensure accurate data linkage.
Yes, we provide access to pseudonymised risk score data for the population via the Sussex Database (a Sussex wide Data Warehouse detailing healthcare activity for the population), however analysts can only access patient names if they have Caldicott Guardian approval for this access.
