To find out why there is such a wide range of referral rates between practices, David Keene began to collect data that has made some GPs question their own clinical practice The NHS plan requires all primary care groups and trusts to monitor referral rates. I was appointed to the post of GP referrals adviser for City and Hackney primary care trust a year ago, having spent 20 years as a GP. I now work three days a week as a GP referrals adviser.

City and Hackney has 53 general practices and a wide range of rates of referral to secondary care. It became clear that, in general, those practices with high referral rates in one specialty had high referral rates in all specialties - and similarly for lowreferring practices. City and Hackney is typical of many inner city areas with significant social deprivation. It has particular problems relating to ethnic diversity, and large numbers of residents whose first language is not English, including asylum seekers.

I decided that an important part of my job would be to visit all practices and discuss referrals with them. The aims of the visits were to raise awareness of the issue of referrals, present each practice with its referral rates, and explore with each one alternative models of referring.

I felt it was important that these visits should be seen as advisory and educational by the practice. I obtained approval for my visits to be accredited for the post-graduate educational allowance. I also sought to be non-judgmental in my approach to practices and to divorce my visits from the commissioning process. I avoided terms such as 'appropriate' and 'inappropriate' referrals.

I prepared for each visit by obtaining referral rates for each practice from East London and the City health authority. I then made the data anonymous and prepared bar charts showing an individual practice's position for each specialty in relation to all other practices within the PCT.

I use first attendance at outpatients as a proxy for referral rate, though this does not include those patients who cancel or miss their first appointment.

This information is usually sufficient to form the basis for discussion. Although most practices do not collect or audit their referral rates, they are fascinated by the data I present to them.They can also appreciate the inaccuracies of their actual referral rates.

This data raises many questions for GPs about their clinical practice and practice policy (or lack of it) towards referrals. It also often prompts them to suggest that they should audit some of their referrals. I then introduce a discussion of referral models.Most GPs still view referrals in terms of a very traditional model of 'GP refers patient to consultant, for patient to attend outpatient department'.

Despite enormous changes in general practice organisation over the past decade and the development of sophisticated practice teams, together with increasing internal referral, the traditional model of referral is still viewed by most GPs as the only way to access secondary care routinely.

We discuss the use of nurse specialists, enhanced physiotherapy, GP specialists, priority scoring and direct booking.We also discuss how to obtain specialist consultant advice without the need to send the patient to outpatients.

Finally, I ask the practice to complete a questionnaire on the referral issues we have discussed.

After each visit, I write a report that I send to the practice. I give them the opportunity to comment on anything I have included in the report and correct any errors. I also enter the data from the questionnaire on to a spreadsheet, together with the referral rates and basic demographic and prescribing data.

I hope in future that this information can be collated to give qualitative data, which may start to explain why there is such a wide range of referral rates between practices.This will enable me to return to practices with suggestions about how to improve their referral decision-making.

I am already able to identify several factors that appear to influence the decision to refer.

I hope these will be confirmed by visiting more practices and that research evidence will support my impressions.

All these factors have been put forward by practices, and I have attempted to categorise them.

They include: rapid changes to practice list size; waiting-list sizes; patient anxiety; doctors' attitudes; and increased stress or workload.

I would like to return to practices with further data on referral rates to enable me to demonstrate trends in clinical practice. In addition, I hope that I will be able to bring together groups of GPs, present them with de-anonymised data and explore differences between their referral rates.

Much more work will need to be done in order to establish what constitutes an appropriate referral.

Visiting practices has started to stimulate debate among clinicians in City and Hackney.The establishment of GP referrals advisers across the country should extend this debate nationally. l Dr David Keene is GP referrals adviser at City and Hackney primary care trust.