This controlled intervention study has shown that GPs’ behaviours were changed and practice improved by a structured educational programme, albeit not in a way that was reflected in all measures. In particular, the fall in RR (the primary outcome measure) was statistically insignificant; on the other hand, GPs made more disease-specific diagnoses while requesting far fewer investigations, and they became much more willing to initiate treatment. It should be noted that, despite the difference in their sizes, the two groups of patients on whom these comparisons were based were demographically similar. Although the age difference (43.2 versus 46.8 years) was significant statistically, it was not so clinically.
To our knowledge this is the first study to demonstrate empirically that GPs’ practice in the field of headache can be favourably influenced by education. The Dutch study of Smelt et al. , aimed specifically at migraine management and recruiting patients already on triptan therapy, failed to show a beneficial effect. This controlled trial employed clinical outcomes, and perhaps demonstrated the difficulties associated with them. The Norwegian study of Kristoffersen et al. , targeting medication-overuse headache only, improved outcomes in an intervention group but the essential element of the intervention was to equip GPs with a “simple and effective instrument” as a management aid; the educational element was secondary to this. Both these studies had much narrower focus than ours; we assessed the provision of headache care to unselected patients, which was a major strength in both purpose and study design.
That there was only a small and statistically insignificant reduction in the primary outcome measure – the overall RR – was disappointing, especially since baseline RR was much higher (39 %) than anticipated (25 %). However, for patients with clearly headache-related GP-made diagnoses, RR fell by one fifth (from 40 to 32 %; p = 0.08), suggesting some gain in confidence in managing patients with primary headache without referral to a neurologist.
Only for patients diagnosed with (pericranial) myalgia was RR reduced significantly, but this outcome was substantially influenced by a considerable reduction in use of this diagnosis (M79.1). We did not analyse diagnostic changes case-by-case or perform quality analysis of diagnoses, but can reasonably speculate that, with better knowledge applied to recognizing and diagnosing tension-type headache, inappropriate use of M79.1 gave way to correct usage of G44.2 for this disorder. In keeping with this supposition, employment of the latter diagnosis significantly increased post-intervention. Also there was a significant reduction in diagnoses that did not specify a primary headache type. In other words, patients left their GPs’ offices less frequently with diagnoses of headache as a symptom and more frequently with diagnoses naming the disease causing this symptom. GPs’ practice shifted towards themselves diagnosing the most common primary headaches.
Key to this is that primary headaches are diagnosed clinically, rarely with need for investigations. Our results show that, post-intervention, GPs substantially reduced their demands for diagnostic investigations, eliminating them altogether in the case of migraine. Not only was this further evidence of learning and of confidence gained as a result, but also it produced immediate cost savings. Our study was not designed to measure cost-effectiveness, but this finding may be one of particular importance because it indicates that health-care resources can be conserved.
In line with an increased rate of specific diagnosis, the study found a greatly enhanced rate of GP-initiated treatment. At baseline, 42 % of patients appeared to receive no treatment recommendations from their GPs; post-intervention, this fell to 19 % – that is to say, 81 % of patients left their GPs’ offices with clear treatment options recommended or prescribed. It was not possible in the context of the study to make any judgement of the appropriateness of treatment initiated.
Interestingly, all these changes in GPs’ practice post-intervention had no discernible influence on patient-reported outcomes – satisfaction with care, satisfaction with health and quality of life, lost productive time or quality-of-life measures. It should be noted here that the instruments used, while not directly validated in the study population, had all been employed in multiple countries, cultures and languages. Where patient satisfaction is concerned, there are multiple determinants that might explain failure to indicate benefit. For example, we sought to reduce the use of investigations because these do not contribute usefully to the diagnosis of headache disorders in primary care. Similarly we aimed to reduce RR because the proportion of patients who should be referred – for diagnostic or management difficulties or for secondary headache – is much smaller than was the baseline RR . Patients, however, might not agree that these reductions were in their interest. Also we have to recognise that the study was underpowered at outset for these secondary outcome measures, added to which the proportions responding to the patient-directed enquiries were about 30 %. Realistically we should not attempt to make anything of these, because, with such low response rates, bias was also likely. Unfortunately, such response rates are not at all unusual  and, as was the case here, are one of the limitations of this type of study.
Other limitations here related to the scope of the study. Its aims did not include quality review of diagnoses, investigations or treatments; to do any of these would have required an entirely different approach. Also this study cannot comment on the duration of found effects, which was outside the scope of the protocol. In other words we are not able to answer the question about need for repeated interventions over time. Further studies, which are planned, are required for these purposes.
Finally it must be said that the gains achieved, at least in the principal outcome measure, were less than had been hoped for, but the fact that gains were made suggests more would be achieved with more educational input. The cost-effectiveness of this must be investigated, but is very likely to be favourable given the enormously high socioeconomic burden of headache [1, 2].