Paul Myers is a GP and senior lecturer at the department of general practice and primary care, St Bartholomew's and the Royal London School of Medicine.

There is an increasing need for ways to measure objectively the performance of those working in primary care, and a wide range of methods may be used to assess GPs' performance.1

Much research has already been undertaken in the field of GP training and medical audit, and there is debate about performance assessment for the purpose of reaccreditation and in the area of clinical governance.

There is also a drive by the profession and outside agencies to evaluate GP performance and identify those who are 'poorly performing'.2

Although performance reviews are frequently used to produce an overall picture of the performance of GPs as a group, there is a tendency to look at individuals in an attempt to identify poorly performing doctors.

A knowledge of the range of methods used by health authorities and others for assessing the performance of practising family doctors would be helpful to embryonic primary care groups faced with the task of developing tools for promoting clinical governance.

Evaluation of these indicators to examine whether they are valid as measures of poor performance would be of equal importance.

I attempted to identify the range and type of performance indicators in use by HAs and other agencies to screen for under-achieving GPs. Each indicator was then examined for its validity as a measure of GP performance.

A literature search was conducted via the Medline database. This was complemented by independent requests direct to specialist libraries.

The initial searches yielded a total of 408 articles related to performance and clinical competence, but most were based on work in the US, and many concerned undergraduate teaching.

The results of the literature search and informal discussions with HAs, local medical committees and GP educationalists produced a list of 61 performance indicators that had been used over the past eight years to monitor GPs' performance in the UK.

Of these, 19 related to prescribing, seven to disease prevention and chronic disease management, seven to the interface between primary and secondary care and 13 to practice administration, leaving 15 miscellaneous indicators. The subjective evaluation of these gave a number of equivocal results, reflecting the difficulty in making assessments of competence.

Few of the indicators were of definite direct clinical importance and few were thought to be reliable and discriminative. Therefore, very few were considered to fulfil the 'gold standard' for markers which would be of clinical importance, discriminative and understandable, relate to the GP's knowledge and skill, be underpinned by accepted research evidence, be acceptable to the assessors as well as those assessed, be seen to be reliable and not set for purely financial reasons.

Performance indicators which most closely approached this standard were mainly in the area of prescribing - for example, the use of generic prescribing, and the avoidance of the prescribing of barbiturates and appetite suppressants. Prescribing indicators are already widely used and have been reviewed elsewhere.3

The use of performance indicators in primary care is a controversial subject, which needs to be approached with great care if indicators are to be used as proxy markers to identify poorly performing doctors. The latter term requires definition, and such doctors should be differentiated from GPs who for historical reasons have been poorly supported, for example, in terms of education or finance.

There is a danger in looking at a marker and then drawing abstract conclusions about the doctor or the practice. Although there has been considerable research on individual markers, there is no consensus on which are valid measures of the quality of an individual GP's performance. There is a risk that a performance indicator is chosen because it is easily measured and relates to current processes in which data is recorded, rather than because it has validity.

The principles for clinical governance include the requirement that poor clinical performance will be recognised at an early stage and dealt with to prevent harm to patients.4

But it is not clear how GP performance will be monitored in future, and to what extent this will be at HA, PCG or practice level. Whatever indicators are chosen should conform to defined standards.

Suitable indicators for performance management have a number of key attributes.1 For example, an indicator should cover an outcome that is relevant and important to policy-makers, health professionals and patients. It should be robust, the measurement of the indicator should be reliable and coverage of the outcome measured should be high and responsive. The data generating the indicator should be usable and should be readily available within a reasonable timescale.

There is a clear need to differentiate between a performance indicator and a quantitative measure of process within primary care. Evidence is needed that the marker under consideration is a true measure of GP performance.

At the present time there is no single marker or group of markers that can reliably be used for the specific purpose of identifying poorly performing doctors. Each marker looks at a single aspect of care. But one should be cautious about drawing conclusions about a GP's overall performance.

The problem with the use of individual indicators is that although each looks at a very specific area, inevitably a range of confounding factors reflects variations, such as patient population characteristics and infrastructure support, which preclude the validity of pronouncing a GP poorly performing.5

It will only be possible to depend on these proxy indicators for performance, either individually or in combination as a quantitative scoring system, if it can be shown that a poorly performing GP would have an indicator profile inferior to other GPs. More research is therefore required.

The model of the GP (registrar) training practice assessment visit represents the most widespread method of objectively assessing GP practices. This method, which many predict will be used when reaccreditation is introduced, may be considered to be the only validated method of examining a GP's performance, as although it does look at some of the markers discussed here, it does so not in isolation but in relation to a detailed examination of the behaviour of the whole practice, not just an individual practitioner.

In the minds of many patients and doctors, a poorly performing GP would be one who does not demonstrate certain values considered to be characteristic of a 'good doctor'.6

These include a high level of clinical knowledge and skills, good communication abilities, high moral and ethical values, and good organisational and decision-making abilities. Many of these are not measured by the performance indicators which appear to be in use at present.

Perhaps the most important attribute of a doctor considered to be performing satisfactorily is being 'safe'. For a doctor to be practising 'safely' - for example, spotting important clinical conditions, referring and using drugs appropriately - there would have to be a minimum standard of clinical knowledge and skills, set against a suitably organised primary care infrastructure. Again, whether a poorly performing doctor, viewed in these terms, could be identified by the performance indicators currently in use is debatable.

A strategy for primary care groups

Despite the difficulties, the need to improve standards in the clinical care provided by GPs remains a priority. If the present indicators are too crude a measure of performance, are there any other strategies for monitoring and improving performance in primary care?

The move away from performance indicators towards continuous quality improvement should be encouraged. The search for the poorly performing doctor currently being attempted is flawed and should be replaced by new systems.

One possibility would be for primary care group boards to provide members with regular updates on their performance, presented in a manner which allows comparison with pooled data from other local practitioners.

In encouraging a culture of continuous quality improvement, such a system could run in parallel with an incentive system, in which practitioners who could provide evidence of quality improvement could also be eligible for support. Thus it would be up to the PCG board to set up innovative systems to support developing practices that need it, at the same time as rewarding more developed practices for improving existing services.