Atrial fibrillation (AF) has come very much to the fore recently but the fundamental issue of securing an accurate diagnosis is not always as easy as it may seem. Quite often an irregular pulse is followed by a less than clear ECG trace, with the interpretative software found on most GP surgery machines stating: "no P wave seen: atrial fibrillation", or words to that effect.

A paper, in the BMJ, has investigated the diagnostic accuracy of health professionals and interpretative software when analysing ECG traces for AF.

The study was set in 49 general practices in Central England and included 2,595 patients aged 65 or older who were screened for AF as part of the Screening for Atrial Fibrillation in the Elderly (SAFE) study. ECGs were recorded with computer software interpretation and then re-read by GPs and practice nurses, independently of each other and unaware of the computer diagnosis.

The results showed that on standard 12-lead ECGs GPs detected 79 out of 99 cases of AF (sensitivity 80%, 95% CI, 71-87%) and misinterpreted 114 out of 1,355 cases of sinus rhythm as AF (specificity 92%, 95% CI, 90-93%). Practice nurses detected a similar proportion of cases of AF (sensitivity 77%, 95% CI, 67-85%), but had a lower specificity (85%, 95% CI, 83-87%). The interpretative software was more accurate (specificity 99%), but missed 36 out of 215 cases of AF (sensitivity 83%).

Combining GPs' interpretation with that of the computer software resulted in a sensitivity of 92% and a specificity of 91%.

This study has shown that although primary care diagnosis of AF is not bad it is far from perfect, and unfortunately interpretative software is not the answer, as it will tend to miss cases even when combined with the human element.

I have always had my doubts about the accuracy of interpretative software. I suspect that the best approach in cases of uncertainty is to fax the offending ECG to a friendly cardiologist.