Person based resource allocation can predict future costs of general practices

A new formula that can predict future health costs more accurately than previous models could help guide commissioning budgets for general practices under the government's new Health Bill, finds a study published on bmj.com today.

The model (known as the 'person based resource allocation' or PBRA) was developed to help allocate resources for commissioning hospital care to all general practices in England, based on the health needs of each individual registered in each practice.

A team of researchers led by The Nuffield Trust set out to develop a formula that was accurate enough to predict future costs of a general practice's population and thus could serve as a basis for setting budgets for the next year's expenditure on hospital care.

For everyone registered with a general practice in England, the team exploited the rich amount of information routinely collected by the NHS on their health needs, for example their illnesses as recorded in hospital. Importantly this was done in a way that protected patient confidentiality. Using this information they devised a series of mathematical models to predict future costs.

In this study, models used data from 2005-6 and 2006-7 to predict costs in 2007-8. Predicted costs were then compared with each practice's actual costs for 2007-8 to assess how the model performed.

The best model performed well by international standards, predicting 77% of hospital costs per practice. Subsequent analysis has increased this figure to 85%. The researchers say that this research could be very useful when the NHS Commissioning Board looks to set budgets for Clinical Commissioning Groups.

The best formulae tested, however, could predict only about 12% of next year's hospital costs per individual. This suggests that the models should be used only to guide allocations to practices, and are not recommended for use at an individual level, say the authors.

"With current information, it is not possible to know what the "true" level of health needs is in each practice," they conclude. "However, compared with previous methods, the PBRA makes far greater use of information from individuals, relies less on area level information attributed to individuals, and is more predictive of future costs. As such, it is an improvement on current methods but is only one step further in assessing health needs of practice populations. Further steps include using information on health needs collected in general practices, which will be possible if data from GP clinical systems are made available for such analysis."