£10 note planeDaloni Carlisle 

An algorithm that will help PCTs predict which patients are set to become so-called “frequent flyers” even before their first emergency admission is set to be released this autumn by the Department of Health.

It is already being trialled in at least two PCTs to identify at-risk patients and intervene to avoid emergency admission.

Croydon PCT has set up “virtual wards” in which the highest risk patients identified by the tool are given intensive support in their own homes.

South Norfolk PCT has identified patients at medium risk and offered them nurse-led telephone coaching in a bid to improve their self-management.

The tool – known as the combined predictive model – uses data from in-patient episodes, out-patients, A&E and general practice to predict which people in a population are most likely to need emergency admission to hospital the following year.

All PCTs are signed up to a Public Service Level Agreement to reduce emergency bed days by 5 per cent by 2008. The stumbling block so far has been how to identify patients before they become unwell. This is the gap the algorithm aims to fill.

NHS commissioned

The combined model was developed by the Kings Fund, New York University and Health Dialog, a data analytics and care management company. It was commissioned by the NHS and will be freely available to PCTs.

Matt Siegel, director of analytics at Health Dialog, said it was similar to algorithms used in the US to manage care. But, he added: “We think this is the first time anyone has been able to predict risk right across a population in a given geographical area”

It builds on a previous model known as PARR – Patients At Risk of Readmission case finding tool – which is based on in-patient data alone.

A study of PARR published in the BMJ this summer showed it was able to predict around eight in ten patients who would have a readmission in the coming twelve months, with a 15 per cent false positive rate. Because of its limited data source (in-patient data and census data), PARR is able to identify only those patients who have already had a hospital admission.

The combined model, with its richer data sources, predicts patients at emerging risk of admission, before they first come into hospital.

An interim report by the Kings Fund published this summer indicated that although there is some overlap between the at-risk groups identified by PARR and the combined model, the new model identifies many new patients.

Mr Siegel said: “PARR is an excellent tool but it analyses data on only the 5 per cent of people who have been admitted to hospital as an emergency in the previous year. If you rely on PARR alone you will not reach the 5 per cent target reduction in emergency bed days by 2008 because it does not cover enough of the population.”

Financial modelling indicates that the combined tool may save money. Figures based on the two PCTs whose data were used to develop the model indicates that if a PCT were to intervene with the 250 highest risk patients identified by the combined model at a cost of £500 per patient, it would save £95,920 in reduced hospital admissions alone.

Virtual wards

In Croydon PCT, public health registrar Dr Geraint Lewis has used the combined model to underpin his “virtual wards” project in which patients at highest risk of hospital admission receive intensive case management at home from a community team that is structured to mirror a hospital ward.

He said: "When we run the model what comes out on our desk is a list of encrypted NHS numbers for everybody in the borough, ranked in order of how likely they are to be admitted to hospital. We take the top 100 of these and, with their GP’s consent, offer them a place on the virtual ward.”

A community matron leads the day to day running of the ward, liaising closely with the patients’ GPs. The ward has a team of nurses, a pharmacist, social worker, physiotherapist, occupational therapist and a ward clerk – all of whom share a single set of electronic notes.

The algorithm was invaluable, he added. “This sort of modelling is the only evidence-based way to predict frequent hospital attenders and offer them preventative care.”

Croydon now has two 100-bed virtual wards and plans for a further eight. A growing number of PCTs across the country are planning similar initiatives. Dr Lewis is applying for funding to evaluate the whole project in a randomised controlled trial.

Mr Siegel said using the combined algorithm was not easy. “It is hard work but then anything worth doing is.” The challenges included working closely with GPs to turn the anonymised data produced by the algorithm into named patients while protecting their confidentiality.

Dr Geraint Lewis has set up a virtual wards discussion group. He can be contacted at geraint.lewis@nhs.net.

The Kings Fund has several pages of information about PARR and the Combined Model here.