Identifying and supporting people at risk on hospital admission

Multidisciplinary teams in Croydon are leading work to identify people at risk of hospital admission and support them in the community, potentially saving substantial sums. Mark Hunter reports

Much of the uptake of health and social services is directed towards a relatively small number of people whose needs are complex and costly to manage.

Numerous attempts have been made to identify these individuals and intervene ­earlier, thereby raising their quality of life and reducing their subsequent need for admission to hospital or residential care.

However, the group is not static. Those who are at risk of needing emergency care this year may be completely different from those at risk in future.

“This is why it is so important to identify who will be at risk next year, rather than who is at risk now,” says Dr Geraint Lewis, a senior fellow at health research body the Nuffield Trust. “If you are a practitioner working with people who are currently at risk you’ll think you are doing a wonderful job because their admission rate is going down. But it would have done that anyway.”

Advance identification of people with increasing needs is known as predictive case finding and has been used in the health service for several years. It is now being adapted for use in social care (see overleaf).

One of two models available free of charge to NHS organisations in England is called the combined predictive model, and was developed by the King’s Fund.

Information, covering patients’ medical conditions, their treatment, response to treatment and admissions history, is fed into a computerised algorithm which allocates each patient a risk score of between one and 100. The score reflects the patient’s risk of needing an unplanned hospital admission in the coming 12 months.

“Those identified as most at risk tend to have a range of complex issues involving at least one long-term condition, such as diabetes, coronary artery disease or COPD (chronic obstructive pulmonary disease), but also other issues such as mental health problems, drug or alcohol abuse, sensory impairment or complex social care needs,” says Dr Lewis.

Having identified these at-risk individuals, the next question is how they should be managed.

Four years ago Dr Lewis was a specialist registrar in public health at Croydon Primary Care Trust and helped integrate the combined predictive model into an innovative programme of community care known as the “virtual ward”.

Patients identified as high risk by the algorithm are referred by their GPs to a multidisciplinary team led by a community matron. The community matron visits the patient at home, listens to their perspective on their situation and assesses their health and social care needs. The team then develops a care management plan.

“A virtual ward is a cadre for providing support in the community to people with the most complex medical and social needs,” says community matron Geraldine Rodgers, who leads one of the 10 virtual ward teams now operating in Croydon.

“It mimics the hospital ward with staff and systems, but the patient lives in their own home,” she says.

Each ward team covers 100 patients and consists of nurses, a pharmacist, a social worker, a physiotherapist, an occupational therapist and a ward administrator who co-ordinates the team’s activity. The team uses a common set of notes, communicates daily by e-mail and meets at least once a week.

“This is where the magic of the multidisciplinary team kicks in,” says Dr Lewis. “Just having all those people around the same table is extremely useful and something that, in my opinion, is often missing from community care.”

Rodgers claims the virtual ward “creates a spirit of partnership” with patients whose circumstances often leave them marginalised by the health and social care system.

“This sense of belonging improves self-esteem and allows patients to be involved in and understand their care, building on their knowledge and quality of life,” she says.

Whether or not Croydon’s predictive case finding has succeeded in reducing emergency admissions is difficult to judge. In its first year of operation, the PCT reported £1m savings in emergency admissions, although other changes to the A&E system will also have contributed to this figure.

A formal independent evaluation of the scheme and similar systems in Devon and Wandsworth has just been commissioned by the National Institute for Health Research and is about to be conducted by the Nuffield Trust.

A HIGH-RISK CASE

Mr S is 57 and lives in social housing. He has epilepsy, drinks 106 units of alcohol a week, has hearing problems and can only read large print. In the two years before being admitted to Croydon’s virtual ward he attended casualty 82 times for fits or alcohol-related incidents. He was admitted to hospital 30 times.

Although follow-up arrangements were made to continue his care, Mr S regularly missed the appointments and he was discharged without assessment or intervention. At the time of admission to the “virtual ward” he was receiving no continuing services from either health or social services.

Identified as very high risk by the combined predictive model, Mr S was assessed by the community matron and admitted to the virtual ward. The community matron’s care plan included three visits a week, referral to social services for a review of his care needs, a joint assessment with the dual diagnosis team, measures to improve his compliance with medication, a series of blood tests and tracking down some MRI scan results that had gone missing.

Since admission to the virtual ward Mr S is now accepting services to meet his health and social care needs. His hearing has been assessed and he has been supplied with a hearing aid. Support from the dual diagnosis team has helped control his drinking behaviour and he no longer indulges in binge drinking. He has not attended A&E for any further alcohol-related incidents.

Mr S is now complying with his anti-epileptic medication and is currently fit-free. He has been assessed by social services and is receiving help for his visual and hearing impairments and advice from a benefits adviser.

Mr S has had no further unplanned hospital admissions or attendances since his admission to the virtual ward.

Predicting the future level of social care needs

Martin Bardsley, head of research at the Nuffield Trust, describes work on producing a model to spot who will need ­intensive social care in future

While progress has been made in identifying future needs in the NHS, there has been less to cheer about in social care – until now.

Commissioned by the Department of Health, the Nuffield Trust has been attempting to create similar models to identify those most at risk of needing intensive social care so councils and primary care trusts can take earlier action to help people stay in their own homes.

The prize is not only greater independence for older people, but significant potential cost savings for health and social care budgets. This is critical at a time of fiscal restraint and with rising numbers of older people with multiple and chronic health problems.

Current methods of assessing whether someone is likely to end up in a care home rely largely on face-to-face assessments. The models we are looking at can look at much larger populations by exploiting existing computerised information.

We worked with five councils and their local primary care trusts to identify information – down to the level of individual people – about their use of hospital services, including in-patient stays, outpatient appointments and visits to A&E, and social care.

Before transfer to the Nuffield Trust, sensitive information was excluded and an anonymised identifying code created. We were able to analyse data from almost two million people aged over 75. Only between 2% and 3% made the transition to intensive social care in any one year – something that made the statistical analysis challenging. Various models were created and the most important predictors in the risk score were a combination of information, including age, previous use of social care and health variables related to hospital use and some chronic health problems.

We will publish our findings this summer and the signs are encouraging, although there are a range of practical hurdles that must be overcome if predictive tools are to be successfully implemented in social care.

A fundamental issue is how to share information in ways that protect confidentiality. There are then a range of questions on what should or could be done to change the care package – according to the calculated level of risk. This work has also raised some interest in integrating information across health and social care.

Still, these early models hold out the promise of enabling councils to more accurately identify individuals most at risk in their communities so they can target effective interventions that maintain independence.

Read more on cutbacks

This article is published in the 3 June 2010 edition of Community Care under the headline “Bring the ward to the patient”