[Pre-partum cardiotocography: comparison between the traditional and the computed systems]

The objective of our study was to determine the usefulness of a computerized antepartum fetal heart rate (FHR) analysis instead of the common visual evaluation of cardiotocographic parameters. From a total of 4,000 antepartum FHR tracings performed in our institute in the years 1994-1996, and analyzed on-line by the software HP 2CTG, 100 were visually assessed by five physicians in three different moments. To each reader a qualitative and quantitative interpretation was required. With use of k coefficient, we have analyzed the differences in observer's qualitative interpretation, then, with use of the intraclass correlation coefficient (RINTRA) and the contingency table X2 we have compared the quantitative analysis with that of the computer. Our results show that variability is the most difficult parameter to interpret, so that the interobserver agreement isn't in anyway sufficient (k 0.43). The agreement between observers on the clinical judgment of the FHR traces proves to be low (k 0.54). The agreement between each of the observers and the computer readings is good for the estimation of the FHR (RINTRA between 0.96 e 0.98) and low for the mean term variability (RINTRA between 0.81 and 0.58). There is a valuable agreement with the computer for tracings with 0 or 1 large acceleration (94.7% and 97.5% respectively), but it is significantly poorer (p < 0.001) for tracings with 0 or 1 small acceleration (57.4% and 50.8% respectively). Decelerations are the most difficult pattern of interpretation. Usually observers do not identify 36% of the FHR traces having at least one deceleration according to computer analysis. Our study confirms the wide variability in fetal monitoring interpretation, a difficulty which can be overcome thanks to a computerized analysis.