I was interested to read the paper presented by Bushnell on behalf of the MaGPIE group.1 However, I have a number of concerns about the validity of the conclusions, as published.
My chief concern is that there is insufficient information about their methods, in both this and the accompanying paper.2 The basis of this study is a measure of continuity, or ‘consultation frequency’ as the group refers to it. However, detail on how they determine this is missing. Measurement of continuity is fraught with methodological problems,3 yet the authors do not seem to have adopted any of the existing instruments. The importance of this issue, and its potential influence on the findings, is not discussed at any point.
If the influence of continuity on symptom recognition was a prior research question, why was the study conducted with a sample size underpowered to detect differences? Although in their introduction Bushnell et al acknowledge the importance of other factors, such as severity of symptoms, in influencing recognition, no account appears to have been made for these in the final analyses. In addition, the researchers do not appear to have excluded any patients with known psychiatric disorders or in receipt of prescriptions for psychotropic medication. I assume the GPs in this study were not ‘blind’ to the medical records, which may have influenced reported recognition rates, even in ‘unknown’ patients.
We are told that the level of psychological problems recognised by GPs was collected from two questionnaires, referring to the index encounter and the previous year respectively, but only the 12 month findings appear to be reported. Does this reflect any bias that favours the presentation of positive findings?
Finally, this study relies on cross-sectional data, yet no consideration is given to the issue of causality. Although Bushnell et al suggest that frequency of attendance leads to improved GP recognition, the relationship may in fact run the other way.