The problem with usual care in general practice is its variability. UK GPs often provide care to rival the best in the world, but not always. For example, the report of the pilot study for the Child Death Review cites examples of general practice care of which we should be proud, but it also reveals poor care which contributes to avoidable death.1 Between these extremes, there is a gradient of care of variable quality. The systematic review of usual care for back pain reported by Somerville et al on pages 790–797 provides a good example, showing substantial differences in the clinical approach to a common condition even in selected practices involved in clinical trials.2 As the authors say, usual care is variable in content and effectiveness.
Reducing the variability in usual care is certainly a good thing in clinical trials. Proving that a particular method of managing back pain is better than usual care doesn't help much if what was provided as usual care is unclear. The authors' recommendation that papers reporting clinical trials should more fully describe the usual care delivered is uncontroversial but insufficient.2 GPs involved in such trials need to reach agreement about the content and delivery of best usual care before they start the research. Even if the usual care provided is fully described in a research paper, interpreting the results and applying them to your own clinical practice is very difficult if what is reported is a wide spectrum of care of variable content and quality.
In everyday clinical practice, variability in usual care matters most at the tail-end of the distribution where poor care can lead to adverse outcomes including avoidable death. Although effective regulatory mechanisms to deal with poor care are essential, epidemiological studies suggest they are likely to be less effective in dealing with a tail than trying to shift the whole distribution by driving up average performance.3 The UK Quality Outcomes Framework (QOF), a protocol-driven pay-by-performance incentive scheme for GPs, already appears to have contributed to this aim for two chronic conditions. The mean practice quality score for diabetes and asthma increased from 70% to over 80% in 2003–2005.4 The variability of usual care was also reduced, with progressive decreases in the upper limit of exception reporting5 and in inequalities in the delivery of care.6
Although there will now be pressure from NHS policy makers to extend the QOF, constraining variability in usual care in this way does have potential disadvantages. Firstly, listening carefully to the patients' concerns and reaching a shared decision consistent with their wishes does not sit easily with a pre-specified and financially incentivised management protocol. Secondly, many of the decisions made by GPs, particularly diagnostic decisions, require complex reasoning that is not easily reduced to a simple protocol or target.7 Although diagnostic decision rules can be very helpful in avoiding missed diagnoses of the type reported in the child death review,8 inflexible protocol-defined triage has the potential to simply shift the complex decision-making further up the line, with consequent loss of effective gate-keeping and increase in service costs. Thirdly, the best treatment decisions are not based directly on the estimate of relative effectiveness derived from a clinical trial but on a personalised estimate of absolute benefit and harm for the individual patient. This requires the clinician to exert clinical judgement by applying the research evidence from clinical trials to their own, preferably evidence-informed, assessment of what will happen to the patient without treatment.9 Lastly, allowing some variability in care within a health system is essential because improvement in care depends on innovation and comparative research.
These disadvantages of constrained protocol-driven care may not be so apparent to a politician or health service manager who takes an over-simplistic view of the application of evidence. It is therefore increasingly important that we put effort now into collecting the evidence in general practice to justify the exercise of clinical judgement in usual care. For example, an issue which will become increasingly debated is the optimal level of exception reporting. High exception reporting levels can reflect either game-playing and poor care or very good personalised care. Clinicians delivering the latter need the support of better observational evidence to defend their judgements. The current level of QOF exception reporting may already be too low for optimal care quality, as protocol driven polypharmacy for older people is an emerging feature of primary care and the total number of drugs prescribed is the strongest predictor of potentially inappropriate prescribing.10 Older people seldom suffer from a single illness and the need for further observational research to delineate the benefits and harms of applying disease-specific QOF targets to people with significant comorbidity will increase as the QOF is extended.
Some of the evidence to address these issues can be derived from routine medical records without great expense. Unfortunately the use of such records for observational research has become problematic.11 Research ethics and governance committees now demand that the patient must give individual consent unless their medical record can be fully anonymised, which is often impossible. In the absence of an opt-out approach, gaining individual consent is usually prohibitively costly and produces hopeless bias by excluding disadvantaged populations. While it is essential that we do nothing to jeopardise the trust of patients, it is possible that ethics committees have been over-cautious, providing the public with a level of protection that patients may not want. It is not obvious that a well informed public would agree that the risk of loss of confidentiality outweighs the benefit of using the aggregated data to inform and improve usual care.
However, much of the evidence we need to inform usual care and justify its variability does require primary research. Only a quarter of QOF key indicators for asthma, diabetes, and angina are described as ‘strongly evidence based’.12 The good news is that GPs in the UK seem increasingly prepared to participate in the necessary research as long as they believe it is directly relevant to clinical care. Dormandy et al (pages 759–766) report little difficulty in engaging general practices in a research study comparing different ways of offering antenatal screening for haemaglobinopathies because it answered an important clinical question about the care they were providing — as one practitioner said, ‘we thought it was useful for the patients … we're very aware of the burden of sickle cell disease in the community’.13
This willingness to participate in research to underpin usual care in general practice must in part reflect the growing influence of evidence-based medicine teaching in our universities and the reimbursement by the NHS of the service costs of research participation. It may also reflect increasing understanding that without evidence we run the risk of becoming ‘docs-in-a box’,14 simply implementing protocol-defined usual care devised by others. A senior colleague with whom I discussed the first draft of this editorial commented: ‘It's as a patient that I never want to meet such a person as a “doc-in-a-box”. Indeed it is my biggest fear about the long-term consequences of QOF: that it produces doctors who don't think, and in the end who can't think.’
- © British Journal of General Practice, 2008.
