Introduction

Cardiotocographs are used to monitor fetal well-being during labor and were introduced approximately 50 years ago [1]. The cardiotocograph is a part of routine intrapartum care monitoring in middle-income [2] and high-income countries [1]. Cardiotocograph monitoring and interpretation aims to prevent intrauterine fetal death by confirming fetal well-being. Unfortunately, cardiotocography has faced significant scrutiny over the years due to inconsistencies in interpretation. The interpretation of tracings varies among professionals, and the same practitioner may interpret the same tracing differently at different times [3]. This may lead to inappropriate intrapartum interventions [4]. Adding to the ongoing discourse on cardiotocograph monitoring, research has demonstrated its association with an increased incidence of cesarean sections and instrumental deliveries [5].

Intrapartum cardiotocographs are interpreted according to clinical practice guidelines. Healthcare professionals in different parts of the world use specific official guidelines to interpret cardiotocographs. There are approximately 13–17 cardiotocograph interpretation guidelines available worldwide [6–8]. Among these guidelines, the International Federation of Obstetrics and Gynecology (FIGO) guidelines [9], the American College of Obstetricians and Gynecologists (ACOG) guidelines [10], and the National Institute for Health and Care Excellence (NICE) guidelines [11] are the most commonly used [12]. These guidelines form part of the baseline knowledge that healthcare professionals should possess when interpreting cardiotocographs.

A normal cardiotocograph comprises a stable baseline fetal heart rate (FHR) of 110–160 bpm without significant decelerations. Normal variability of 5–25 beats per minute (bpm) and alternate periods of increased variability with or without acceleration is called cycling behavior and is a vital behavioral state of normal-term or near-term fetuses [13]. Anything outside these parameters is considered abnormal and warrants further consultation and interventions. Cardiotocograph surveillance enables healthcare professionals to study different FHR patterns and uterine contractions during the intrapartum period. These patterns indicate the fetal state and adaptation to the stress of labor. Unfortunately, fetal heart rate patterns can be very complex, which may contribute to a lack of standard interpretation among healthcare professionals [4]. Previous studies have identified the following factors contributing to interpretational variation among healthcare professionals: years of experience [14], paper speed settings [15], and the use of four-tier ST segment analysis (STAN) guidelines, with the availability of electrocardiographic ST waveform data [16].

Visual interpretation of cardiotocographs and fetal heart rate monitoring through a handheld Doppler device and the Pinard fetal stethoscope remain the primary methods of monitoring fetal well-being during labor. Additional tests used in low- and middle-income countries include fetal acoustic stimulation, fetal scalp stimulation tests, and fetal scalp oximetry [2]. High-income countries have attempted to improve cardiotocograph interpretation by introducing fetal blood sampling and STAN of the fetal electrocardiogram [17,18]. Fetal blood sampling has been discontinued in countries with a high burden of the Human Immunodeficiency Virus (HIV), such as South Africa, as part of the prevention of mother-to-child transmission strategy [19]. Additionally, fetal electrocardiograms cannot be performed in these countries because they require a fetal scalp electrode, which is an invasive procedure and contributes to the vertical transmission of HIV.

Recently, a systematic review summarized and assessed the existing inter- and intraobserver reliability research on intrapartum FHR interpretation [12]. This review included studies where fetal monitoring was performed via cardiotocography and intermittent auscultation. The findings of this review revealed substantially more intraobserver agreement than interobserver agreement. Furthermore, clinicians have demonstrated greater reliability and agreement for basic fetal heart rate features than for overall classification [6]. Another systematic review compared visual and computerized cardiotocograph interpretations to determine whether automated cardiotocograph interpretation resulted in improved perinatal outcomes [20]. These reviews, however, did not identify the factors contributing to inter- or intraobserver variability during FHR interpretation. Therefore, this literature review assessed the factors contributing to variations in visual intrapartum cardiotocograph interpretation among healthcare professionals.

Review question

What are the factors contributing to intrapartum cardiotocograph interpretation variation among healthcare professionals?

Methods

Data extraction

The first author and a librarian conducted the literature search. A total of 1912 results were identified during the database search. EBSCO Host automatically removed 99 duplicates during extraction. A total of 1813 titles were imported into EndNote 21 reference management software. Thereafter, the results were exported into Rayyan Software for Systematic Reviews [22]. A total of 332 duplicate studies were detected and removed. The total numbers of remaining titles and abstracts was therefore 1481. The remaining titles and abstracts were then dually screened in accordance with the inclusion and exclusion criteria. A total of 60 full-text articles were then screened for eligibility. One further result was identified from the citation search during the full-text screening of the included studies. Only nine studies were included in the review. This process is presented in the PRISMA flow diagram in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram of the literature review search process.

1. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. International journal of surgery. 2021;88:105906.

https://doi.org/10.1371/journal.pone.0315761.g001

Quality assessment of studies included in the review

The quality of the included studies was assessed using the Quality Appraisal Tool for Studies of Diagnostic Reliability (QAREL) [24]. This is an eleven-item scale used to assess studies reporting diagnostic reliability. The agreement of two or more observations using the same medium is called diagnostic reliability [24]. Table 1 below is an overview of the QAREL quality appraisal of the studies included in the review.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. QAREL quality appraisal.

https://doi.org/10.1371/journal.pone.0315761.t001

Results

Description of included studies

The researchers selected 60 articles for full-text review. Of those, nine met the inclusion criteria for the integrative review. The studies included in this review used quantitative designs and were conducted in Norway [25], Ireland [3], Finland [26], Sweden [27], Portugal, Boston and London [28], Thailand [29], Italy [30,31], and China [32]. The studies used methods such as asking observers to interpret the same traces using two different FHR classification systems [29] and asking observers to interpret various types of cardiotocograph traces using a single FHR interpretation classification system [25,26,28,32]. Four of the studies reported intra- and interobserver agreement [3,27,29,30], while the other five reported interobserver agreement [25,26,28,31,32].

Statistical analysis measures for intrapartum cardiotocograph interpretational variation.

The studies included in this review used different statistical methods to assess reliability and agreement. Reliability and agreement are two separate concepts that describe how a particular instrument detects real variability between assessors, and whether repeated measures of the instrument yield consistent results [33]. Agreement shows the exact magnitude of how close test scores for repeated measures are [33]. Reliability is defined as the proportion of variation between scores of the same subjects by different observers, which may even be different from the total variability of all scores in the sample [34]. The kappa (κ) statistic provides a numerical rating of the difference between inter- and intraobserver agreement, which is actually present, and that which is expected to be present by chance alone. The kappa scale ranges from -1 to 1, where 0 represented exact agreement by chance, 1 represented perfect agreement, and negative values represented agreement less than chance [35]. The weighted kappa (κw) reflects the degree of disagreement, so that greater emphasis is placed on large differences between ratings and not small ones [35].

The proportions of positive agreement and specific agreements depict the overall agreement between observers for each category under investigation [33]. The result is often expressed as a percentage; for example, 0.85 indicates 85% agreement between observers [36]. The main outcome measurements of each study are shown in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Overview of studies included in the review.

https://doi.org/10.1371/journal.pone.0315761.t002

The advantage of the kappa statistic is that it measures agreement beyond that of chance, which proportion of agreement does not [27].

Key themes

This review investigated the factors contributing to variation in the visual interpretation of intrapartum cardiotocographs among healthcare professionals. Thematic analysis revealed two themes, namely, interpretational variations among health professionals for cardiotocograph characteristics and factors leading to increased interpretational variation among healthcare professionals. Seven subthemes were also identified. The themes and subthemes are presented in Table 3 below.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Table of themes and subthemes.

https://doi.org/10.1371/journal.pone.0315761.t003

Theme 1: Interpretational variations in cardiotocograph characteristics among health professionals.

The current review identified key areas of interpretational variation among healthcare professionals related to cardiotocograph features.

Cardiotocograph features. Agreement of visual interpretations of intrapartum cardiotocographs ranged from fair to good for baseline FHR, baseline variability, accelerations, and frequency of contractions. Interobserver agreement was excellent for the classification of decelerations [3]. Palomaki et al. [26] reported greater proportional agreement for normal features on cardiotocographs than for abnormal features. The interpretation of individual cardiotocograph characteristics using the National Institute for Child Health and Human Development three-tier and 5-tier systems revealed an interobserver agreement of κ = 0.75 for the FHR baseline, κ = 0.01 for baseline variability, κ = 0.39 for accelerations, κ = 0.59 for variable decelerations, κ = 0.37 for late decelerations and κ = 0.66 for prolonged decelerations [29]. For example, a study conducted in China revealed that proportional agreement between participants was good for baseline, variability, acceleration, variability, and late and prolonged decelerations. The feature with the least interobserver agreement was early decelerations (Pa = 0.39) [32]. One study evaluated whether extended cardiotocograph interpretation training improved cardiotocograph interpretation and reported that extended training greatly improves intraobserver agreement for the FHR baseline, variability, accelerations, and contractions. Interobserver agreement was better for the baseline, variability and accelerations [27].

Cardiotocograph category classification. The cardiotocograph category classification varied among different healthcare professionals, while intraobserver agreement also varied in some instances. Devane and Lalor [3] revealed that intraobserver agreement for trace classification was altered between time 1 and time 2 for normal traces 18% of the time. Intraobserver variation was also noted for 29% of the time for suspicious traces and 11% of the time for pathological traces. The same study revealed that interobserver agreement was greatest for suspicious traces (κ = 0.77) and lowest for normal traces (κ = 0.54). In contrast, Blix et al [25] reported that interobserver agreement was highest for normal traces (Pa >0.85). A similar result was observed by Palomaki et al. [26], who reported that interobserver agreement was greater for traces that did not require any intervention (Pa = 0.63) than for traces where clinical intervention was recommended (Pa = 0.55). Similarly, Amadori et al. [30] reported moderate interobserver agreement in classifying cardiotocograph traces as normal but reported no consensus on suspicious and pathological traces. Another study that compared the National Institute of Child Health and Human Development (NICHD) three-tier and five-tier classification systems revealed that both classification systems showed strong agreement for normal cardiotocograph traces [29].

Theme 2: Factors leading to increased interpretational variation among healthcare professionals.

This review revealed several factors linked to variations in the interpretation of intrapartum cardiotocographs, including work experience, professional category, clinical and anamnestic information, and cardiotocography guidelines.

Work experience. One study [26] compared interobserver agreement between senior and junior obstetricians and revealed that senior obstetricians demonstrated better agreement when assessing cardiotocograph abnormalities (Pa = 0.59) than did junior obstetricians (Pa = 0.51). In that study, senior doctors were defined as doctors with more than 4 years of work experience, while those with less than 4 years of work experience were categorized as junior doctors. The same study also reported that both senior and junior obstetricians recommended interventions equally. However, junior doctors were 30% more likely to recommend that patients undergo cesarean section. In contrast, another study on interobserver reliability among junior and senior obstetricians reported that interobserver agreement and reliability were not strongly affected by years of experience [32]. That study categorized years of experience into three categories, namely, senior level: having more than 10 years of work experience; intermediate level: having 5–10 years of obstetric experience; and junior level: having less than 3 years of obstetric experience.

Professional category. The current review revealed that midwives tend to have more interpretational agreement than other professionals within the multidisciplinary team [30]. That study assessed inter- and intraobserver reliability among midwives and obstetricians. The results of that study found greater intraobserver reliability for midwives (mean agreement % = 77.05%) than for obstetricians (mean agreement % = 65.75%). Midwives were also more likely to have interobserver agreement (κ = 0.58) than obstetricians (κ = 0.39).

Clinical and anamnestic information. A study performed in Italy [30] established that a patient’s clinical and anamnestic information reduced obstetricians’ cardiotocograph classification from pathological to suspicious and even to normal. Senior midwives, however, remained relatively consistent in their analysis of cardiotocograph traces with or without clinical and anamnestic data. The same study also revealed that clinical and anamnestic data played a role in the classification of a trace as suspicious or pathological but did not affect the classification of normal traces.

Cardiotocograph guidelines. Three doctors with varying levels of expertise compared the NICHD three-tier cardiotocograph classification system and the 5-tier cardiotocograph classification system in Thailand [29]. The same study also compared trace interpretations using FIGO guidelines and an automated computerized system. Overall, the interobserver agreement was fair (κ = 0.52) for the three-tier classification system and moderate (κ = 0.30) for the five-tier classification system. The intraobserver agreement was greater for the NICHD 3-tier system (κ = 0.75) than for the NIHCD 5-tier system (κ = 0.62). Another study performed in Italy investigated the level of interobserver agreement between the Royal College of Gynecologists and the Piquard classification systems [31]. The Piquard classification system is specific to the second stage of labor and describes six different cardiotocograph patterns in relation to fetal pH at birth [37]. The results of that study revealed moderate interobserver agreement for both the RCOG and the Piquard classification systems. A multicountry study investigated interobserver agreement between the FIGO, ACOG, and NICE cardiotocograph classification systems [28]. The results of that multicountry study established that the reliability of the FIGO classification system (κ = 0.37, CI = 0.31–0.43) and the NICE classification system (κ = 0.33, CI = 0.28–0.39) was significantly greater than that of the ACOG classification system (κ = 0.15, CI = 0.10–0.21).

Discussion

This current review aimed to identify the factors contributing to variations in the interpretation of intrapartum cardiotocographs among healthcare professionals. Evidence suggests that there is marked intra- and interobserver variability in the visual interpretation of intrapartum cardiotocographs. This variability may be attributed to the subjective nature of visual interpretation. The current review also revealed that the best agreement was observed for normal intrapartum cardiotocograph traces. There are conflicting data about whether work experience influences the level of agreement between professionals during interpretation. Anamnestic and clinical data may also alter interpretation, and the cardiotocograph classification guidelines may influence interpretation.

Most of the included studies reported high interobserver agreement for normal traces [25,26,29]. These findings align with those of a systematic review by Englehart et al. [6], who reported greater reliability and agreement between observers for individual cardiotocograph features than for overall cardiotocograph classification. Furthermore, their review reported high interobserver agreement for the classification of normal traces. This is concerning because it is suspicious and pathological traces which warrant interventions during labor.

Studies in this current review revealed more inter- and intraobserver agreement among midwives [30]. Amadori et al. [30] observed that midwives might be more inclined to interpret cardiotocographs accurately because they are present throughout labor, which allows them to watch labor unfolding both for physiological and pathological traces. Determining whether work experience affects cardiotocograph interpretation is difficult due to conflicting results. Thellesen et al. [38] revealed that participants with more than 15–20 years of work experience scored higher on cardiotocograph interpretation than participants with less than 15 years of work experience. The findings of that same study also revealed that the size of the maternity unit affects cardiotocograph interpretation expertise. On the topic of interventions following cardiotocograph interpretation, this current review revealed that junior doctors were most likely to recommend a cesarean section as an intervention following cardiotocograph interpretation.

With regard to the influence of clinical and anamnestic data on cardiotocograph interpretation variation, the NICE [11], FIGO [9], and ACOG [10] guidelines all agree that clinical and anamnestic data should be considered during cardiotocograph interpretation. The three guidelines recommend that cardiotocograph interpretation be performed simultaneously with an ongoing risk assessment of antenatal or intrapartum risks that may affect the fetus and the progress of labor.

Implications for practice

• Labor wards with junior staff should create opportunities for mentoring relationships between junior and senior staff in the area of cardiotocograph analysis to allow for knowledge sharing and transfer.
• The practice of sending cardiotocographs to obstetricians without any anamnestic or clinical patient data for analysis and clinical decision-making should be critically reviewed.
• Continuous training on cardiotocograph interpretation focusing mainly on suspicious and pathological traces should be offered to all staff who are involved in intrapartum care.
• Low- and middle-income countries need to find innovative solutions to aid midwives and obstetricians who do not have access to adjunct technologies found in high-income countries to aid in the confirmation of fetal distress during labor. Solutions should be cost-effective, guarantee maternal and fetal safety and ensure practical implementation. This will help to achieve accurate cardiotocograph interpretation, which will ultimately reduce the cesarian section rate.

Limitations and strengths of the review

This review was limited to studies investigating noninvasive cardiotocograph interpretation without adjunctive technologies such as ST segment analysis, fetal blood sampling, and umbilical cord pH analysis. Another limitation is that only studies written in the English language were included. The reviewers conducted a comprehensive search of the included databases and a citation search of all the included full-text articles. However, it is possible that some relevant studies could have been missed. However, the review included all the studies that met the inclusion criteria. Furthermore, a quality appraisal of all included studies was performed using the QAREL checklist.

Conclusion

This integrative review investigated factors contributing to variability in the visual interpretation of intrapartum cardiotocographs among healthcare professionals. The results suggest that there is agreement in the interpretation of cardiotocograph features; however, discrepancies are noted when professionals need to classify traces. The work experience of clinicians may affect the agreement of cardiotocograph interpretation. Patient clinical and anamnestic data also guide healthcare professionals in placing cardiotocographs into appropriate categories. Guidelines for classifying cardiotocographs also affected the interpretational agreement between healthcare professionals. Although our review revealed conflicting evidence, more efforts need to be made to develop methods for verifying cardiotocograph results that can also be applied in resource-limited settings with a high incidence of HIV. The training of midwives and obstetricians on visual interpretation should also focus on the interpretation of suspicious and pathological traces.