There remains, however, the possibility of self report bias; patients who attended the clinics may have reported preventive behaviour without practising it.
Our failure to show any reduction in smoking, however, provides persuasive evidence that significant self report bias was unlikely.
Our findings on behaviour were also consistent with a previous randomised trial in primary care in Belfast that reported improved diet and physical activity but no effect on smoking from health promotion to patients with angina.
To avoid the problems associated with cluster randomisation (especially if practice performances at baseline are variable), we randomized by individual rather than practice.
There may, therefore, have been some effect on control group patients by the presence of the intervention within their general practice: either dilution of effect by contamination, or exaggeration of effect if usual treatment was deliberately withheld from non-intervention group patients.
Previous study of primary coronary prevention clinics suggests that the effect, if present, was more likely to be contamination, but that it was unlikely to have been substantial.
Practices were recruited from a stratified random sample of all general practices in northeast Scotland and the recruitment rate was 68%.Of patients with identifiable coronary heart disease in general practice, 71% agreed to take part. Characteristics of respondents and non-respondents to the initial study invitation have been reported previously. Nonrespondents were slightly less likely to have had aspirin or â blockers prescribed or their blood pressure or cholesterol levels checked in the past three years, but these differences were
modest.Of those who agreed to take part in the study, there were few withdrawals.
We believe, therefore, that the study practices and patients were reasonably representative of northeast Scotland.
In other areas, local factors could affect the performance of replicated clinics.
We found that the most influential confounding variable was baseline performance: if this was already high then patients had less to gain.
Baseline United Kingdom data on secondary prevention, however, were reported in the ASPIRE study and presented a similar suboptimal picture to the baseline position in northeast Scotland.
In an attempt to assess the importance of baseline performance at the general practice level, we divided study practices into quartiles.
Patients were found to benefit in all groups of practices, although most improvement probably occurred in those with the lowest starting point.
It seems likely, therefore, that the results of this study will be relevant to other areas.