Predicting and Reducing Re-admission to Hospital

In 2005, Essex Strategic Health Authority commissioned The King’s Fund to develop a case-finding algorithm on behalf of the Department of Health, the NHS Modernisation Agency and England’s strategic health authorities. The Department of Health took over management of this contract in May 2007.

The project team developed the Patients at Risk of Re-hospitalisation (PARR) tool by using five years of Hospital Episode Statistics (1999/2000 to 2003/4).

The team examined admissions in 2003/4 to identify a 'triggering' admission for each patient. Then they looked back at data relating to them from the previous three years to see what factors might predict the triggering admission and any further admissions in 2003/4.

The PARR tool used a broad range of variables to establish this risk. This included the patient’s previous use of hospital services, demographic data including location to pick up different practice styles and the hospital of current admission to pick up on different practice styles and admission thresholds.

A final version, PARR++ was issued in November 2007.

Subsequently, The King’s Fund, Health Dialog and New York University developed another model, the Combined Predictive Model, which links inpatient, outpatient, accident and emergency, GP and other routinely collected primary care data.

Because it draws on more data sources than PARR, the Combined Predictive Model can identify patients at risk of emergency admission to hospital from across the spectrum of care. This means that the model is able to identify patients who would otherwise be missed by PARR; ie patients whose condition is deteriorating but who have not yet entered a sprial of admissions. This offers the PCT the opportunity to intervene early and to prevent deterioration.