Interobserver agreement in CTG interpretation using the 2015 FIGO guidelines for intrapartum fetal monitoring

Visual analysis of cardiotocographic (CTG) tracings has been shown to be prone to poor intra- and interobserver agreement when several interpretation guidelines are used, and this may have an important impact on the technology's performance.

The aim of this study was to evaluate agreement in CTG interpretation using the new 2015 FIGO guidelines on intrapartum fetal monitoring.

A pre-existing database of intrapartum CTG tracings was used to sequentially select 151 cases acquired with a fetal electrode, with duration exceeding 60minutes, and signal loss less than 15%. These tracings were presented to six clinicians, three with more than 5 years' experience in the labor ward, and three with 5 or less years' experience. Observers were asked to evaluate tracings independently, to assess basic CTG features: baseline, variability, accelerations, decelerations, sinusoidal pattern, tachysystole, and to classify each tracing as normal, suspicious or pathologic, according to the 2015 FIGO guidelines on intrapartum fetal monitoring. Agreement between observers was evaluated using the proportions of agreement (Pa), with 95% confidence intervals (95%CI).

A good interobserver agreement was found in the evaluation of most CTG features, but not bradycardia, reduced variability, saltatory pattern, absence of accelerations and absence of decelerations. For baseline classification Pa was 0.85 [0.82-0.90], for variability 0.82 [0.78-0.85], for accelerations 0.72 [0.68-0.75], for tachysystole 0.77 [0.74-0.81], for decelerations 0.92 [0.90-0.95], for variable decelerations 0.62 [0.58-0.65], for late decelerations 0.63 [0.59-0.66], for repetitive decelerations 0.73 [0.69-0.78], and for prolonged decelerations 0.81 [0.77-0.85]. For overall CTG classification, Pa were 0.60 [0.56-0.64], for classification as normal 0.67 [0.61-0.72], for suspicious 0.54 [0.48-0.60] and for pathologic 0.59 [0.51-0.66]. No differences in agreement according to the level of expertise were observed, except in the identification of accelerations, where it was better in the more experienced group.

A good interobserver agreement was found in evaluation of most CTG features and in overall tracing classification. Results were better than those reported in previous studies evaluating agreement in overall tracing classification. Observer experience did not appear to play a role in agreement.