We thank Drs Bernardes and Costa-Pereira for their continued interest in our article and would like to make the following response to the issues raised in their letter.
- 1They expressed some concern over the methodology we used. All possible combinations of all baseline differences between the experts (n = 3664, after excluding missing data) were calcu lated and compared with that between the computer and experts. The minimum and maximum differences between experts was −60 to +60 bpm, with the mean difference being −0.26 bpm and a standard deviation of 6.22 bpm. The 95% confidence interval of these differences was not computed from the standard deviation. It was instead estimated by deter mining the 2.5th and 97.5th percentile values using the standard frequency distribution function available in SPSS.
- 2We concur that our algorithm is not foolproof and not always able to determine a baseline value. We anticipate that even clinicians would have difficulty defining a baseline value in heart rate pattern D or a second stage trace without any preceding trace in pattern A or B antenatally and some pre- second stage tracings respectively. The computer has the advantage of being consistent in its interpretation and it is not expected that any interpretation, human or machine, should be able to determine a baseline value under all circumstances.
- 3Our approach differs from others working in this area, in that we have deliberately processed only the heart rate information to obtain its variations, before integrating this with physiologi cal and clinical information to produce an overall pathophysiological diagnosis. Only by doing this do we feel that it will be possible to distinguish between how the heart rate varies and what these variations mean.