The concept of predictive risk, or using linked person-level data to identify the patients most likely to have future unplanned hospital admissions, is now firmly embedded in the NHS. Here at the Nuffield Trust we recently held our fifth annual conference on the topic.
One of the joys of holding an annual event on the same subject is that it provides an opportunity to see how things change from year to year, and it is certainly true that the world of risk stratification has come a long way over the past few years.
The last year, however, has not quite gone according to plan with many commissioners finding the successful implementation of risk stratification a challenge. We did hear, however, that some areas are making real progress.
It was gratifying to hear that despite the recent changes to legislation and accompanying uncertainties, that this need not necessarily be a complete barrier to progress
In the past the focus of predictive risk has mainly been on how to create risk models, specifically the data and algorithms used, and how the models are applied. This interest specifically reflected the growing awareness of how risk modelling can be used in the management of long term conditions.
At the conference we heard a great deal about how organisations have gone about the implementation of risk stratification and profiling at the front line, with perspectives and practical examples provided by Tower Hamlets Clinical Commissioning Group (one of the integrated care pioneer sites), the London Borough of Enfield and NHS Central Southern Commissioning Support Unit.
It was gratifying to hear that despite the recent changes to legislation and accompanying uncertainties, that this need not necessarily be a complete barrier to progress.
One area that seems to have changed is the number of models: whereas once there were a handful of risk models, mainly PARR, the Combined Model, and a few commercial systems, there is now a much wider spread of models and applications.
In particular, models that use only GP data are becoming more popular – partly as a result of the challenges of linking hospital data to primary care data.
The sense is that the idea of predictive risk has truly taken hold, but it is just the beginning of what can be achieved by linking data from different parts of the care system. Here at the Nuffield Trust we have been hearing about lots of interesting work going on, both in the UK and abroad, involving linked data.
Many of these include projects that are perhaps closer to research or audit, and work with data sets that would not be routinely linked together. For example, a paramedic who has to routinely make decisions about whether to take patients to A&E may not know what happens to the patient after they transfer, thus making it difficult to improve future decision making.
Yet one organisation we are collaborating with has shown that it is possible to link ambulance data to specific A&E departments and perform something akin to a clinical audit. Another project we are involved in links data from cancer registries to hospital admissions data to examine patterns in emergency diagnoses of colorectal cancer, with a view to potentially identifying points for intervention.
It is also interesting to note the use of the Ministry of Justice Data Lab by Tracey Gyateng of New Philanthropy Capital, which is part of an initiative to make better evaluative methods more accessible to small organisations – such as community and voluntary providers of care for offenders.
These employ many similar methods to those we have used at the Nuffield Trust in examining the efficacy of schemes in reducing hospital admissions, but applied here to examining re-offending rates.
All this is part of a larger picture of general advances in using information. Perhaps the most compelling vision of the future of “big data” and health informatics is provided by the recently created Farr Institute.
This innovative collaboration between academia, industry and NHS organisations brings together information from electronic health records and other data sources (and potentially genomic data) to provide leading medical research – and may prove the key to unleashing a new era of personalised medicine.
After all, to quote Farr himself, “Diseases are more easily prevented than cured and the first step to their prevention is the discovery of their exciting causes”.