This site is intended for health professionals only

First validated GP risk tool to predict emergency admissions

Researchers have developed the first validated risk tool for GPs to use to predict patients’ likelihood of being admitted as an emergency to hospital, which they suggest practices could use to meet the requirements of the risk profiling DES introduced this year.

Nottingham University researchers have developed the QAdmissions algorithm, which they say is at least as accurate as other available scores for admissions risk profiling and means GPs can use data they already collect routinely, without needing to link to hospital records.

Developed by the same team that created QRISK2 and QDiabetes using the QResearch database and published in BMJ Open, the new tool comes as practices start taking the new risk profiling DES forward this month, with CCGs tasked with providing guidance to practices on what risk prediction score to use and which patients to actively case manage.

Doubts have been raised about the evidence base for the DES after the approach failed to reduce emergency admissions, and suggestions it could even lead to an increase in patients being admitted to hospital. The service has also been criticised by the GPC because it lacks nationally applied specifications, including any directive on what risk profiling algorithm to use.

According to the developers of the new QAdmissions risk tool, scores that are currently available are beset with problems, not least that they do not run off data routinely collected in GP computer systems.

The researchers say the two Department of Health-recommended scores, the PARR (Patients at Risk of Re-hospitalisation) and the CPM (Combined Predictive Model) developed by the King’s Fund, are based on old data and need updating. Other tools, including several commercial scores, either do not come from primary care populations or have not been published or validated, they say.

Professor Julia Hippisley-Cox, professor of clinical epidemiology and general practice at University of Nottingham and a sessional GP in the city, and Professor Carol Coupland, reader in medical statistics at University of Nottingham, have now developed a new score, using data from the QResearch database linked to hospital episode statistics (HES) data, which they say overcomes these drawbacks.

To come up with the score, the researchers analysed a total of 2,849,381 patients from 405 practices contributing to QResearch, of whom 265,573 had one or more emergency admissions to hospital over a two-year follow-up lasting from the beginning of January 2010 to the end of December 2011.

From this, they developed an algorithm incorporating 30 variables that are associated with an increased risk of emergency admissions. These include sociodemographic variables, including Townsend score, lifestyle factors, such as smoking and alcohol intake, and morbidity variables, such as obesity, as well as prior admissions, medications and laboratory test results.

The positive predictive value for the top 10% of patients with the highest risk was 42% with GP-HES linked data and 40% with the GP data only, while sensitivity was a respective 39% and 37%.

Professor Hippisley-Cox told Pulse: ‘The data required to generate a risk score are at present in the electronic health record or are automatically generated by the algorithm behind the tool, as is the case for the Townsend score, for example. Increasingly, GPs are recording hospital admissions in the clinical record from discharge letters.’

Professor Hippisley-Cox said the score is available now and could easily be integrated into practice software.

She said: ‘QAdmissions is available now (, and software development kits) and can be used to illustrate an individual’s risk of an emergency admission to hospital.  

‘EMIS is developing an integrated solution which is information governance compliant, time saving and cost effective for practices and commissioners.’

Commenting on the report, Dr James Kingsland, national clinical lead for the NHS Clinical Commissioning Community and a GP in Wallasey, Merseyside, said the risk tool appeared to ‘tick all the boxes’, but said NHS England should advise on which tool to use.

He said: ‘The tools are variable and the BMJ Open one is as good as I’ve seen - certainly for its methods, number of participants and conclusions it’s drawn it’s as good as any if not better.  But to say how we implement the DES in terms of what tool to use, that has got to be a discussion between the commissioner and provider, which is practice and area team.

‘NHS England should be analysing what’s available, looking at the cost effectiveness and then advising on which tool or range of tools to use.’

BMJ Open 2013, online 19 Aug