Fetal heart rate tracings: observers versus computer assessment

ELSEVIER

European Journal of Obstetrics & Gynecology and ReproductiveBiology68 (1996) 83-86

Fetal heart rate tracings: observers versus computer assessment T. Todros*, C.U. Preve, C. Plazzotta, M. Biolcati, P. Lombardo Dipartimento di Discipline Ginecologiche e Ostetriche, Via Ventimiglia 3, 10126 Torino, Italy

Received 25 March 1996;accepted 17 May 1996

Abstract

Cardiotecography (CTG) is widely used despite the fact that its diagnostic accuracy is far from satisfying. This is due, among other reasons, to the great intra- and interobserver variation in reading the fetal heart rate tracings. Computerized analysis might be a means to overcome the latter problem. Objective: the present study was designed to assess the reproducibility of CTG readings among observers and between observers and a computer system. Study design: 63 fetal heart rate tracings were read by 4 clinicians (2 experienced and 2 inexperienced) and by the 2CTG computerized system. The variables considered were: baseline fetal heart rate (FHR), long-term variability (amplitude bandwidth around the baseline), number of large accelerations and number and type of decelerations. Results: the agreement among observers, assessed by means of Kcoefficient, ranges from fair to good. The agreement between each of the observers and the computer readings, ranges from 0.18 to 0.48 for FHR baseline, from 0.16 to 0.74 for variability, from 0.37 to 0.64 for the number of accelerations and from 0.41 to 0.54 for the number of decelerations. The agreement on the type of decelerations is very low (0.01-0.25). Conclusion: it is concluded that interobserver variability between experienced observers, inexperienced observers and 2CTG is considerable and that the use of a computer system should overcome this problem. Keywords: Fetal heart rate; Computerized cardiotocography; Fetal monitoring; Reproducibility

1. Introduction Cardiotocography (CTG) is widely used despite the fact that its diagnostic accuracy is far from satisfying [1,2]. Very different values of sensitivity (from 30% to 100%) and specificity (from 60% to 100%) are reported [1]. The inconsistent results are attributable to differences in study population (low vs. high risk pregnancies) and/or in the end point chosen to calculate the accuracy (perinatal mortality, intrapartum fetal distress, low Apgar score, acidosis). Moreover sensitivity and specificity values are negatively affected by the great intra and interobserver variation in reading the fetal heart rate tracings [3-8]. The computerized fetal heart rate analysis might be a means to overcome the latter problem, at least in part [9,10]. It is also believed that computerized systems are an advantage for inexperienced clinical staff [10]. * Corresponding author, Tel.: 39 11 3134436;fax: 39 11 6647910; e-mail: [email protected].

The present study was designed to assess the reproducibility of CTG readings by expert observers, inexperienced observers and a computer system. 2. Materials and methods

Sixty-three fetal heart rate tracings recorded from high and low risk pregnancies between 30 and 41 weeks of gestational age were randomly chosen. From each, a 25' strip was randomly selected. The recordings were obtained with a Hewlett Packard 8040 external monitor at a paper speed of I cm/min and simultaneously analyzed by the 2CTG computerized system [11] (Fig. 1). Of the computer output variables, the following were considered for the study: the baseline heart rate, the amplitude bandwidth around the baseline (8) - - which is a measure of long-term variability - - expressed in beats per minute (bpm), the number of large accelerations (amplitude > 15 bpm above the baseline lasting > 15 rain), the number of decelerations (amplitude > 20 bpm below the baseline lasting > 30 min or amplitude

0301-2115/96/$15.00 © 1996 ElsevierScienceIreland Ltd. All rights reserved P l l S030 I-21 i 5(96)02487-6

84

T. Todros et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 68 (1996) 83-86 Ist.

Oineooioola

•

Ostetrioia

E e a m e di:

cart.

AUALISI

dl

G.O.

-

Oniversit&

Gestazione:

38 Sett.

41

Torino

e 2 gg.

P ~ I C A C T @

10:30:51 55 minuti 128 19 0 15,4 bpm

Studio Iniziato alle: D u r a t a Studio: Fhr Base: Accelerazioni : Decelerazioni: Delta:

180

-

del 2S-Nov-1992

Accelerazloni Piccole: V a r i a b i l i t a ' B r e v e Periodo: V a r i a b i l i t a ' L u n g o Periodo: Interval Index: Delta: Numero Movimenti: Nuraero Contrazioni:

5 9,7 34,3 9,3 54,3 msec 17 7

-

!

90

0

I

0

g

1

II 5 l

Accel,~r.,io'.l Decel!J~azxoni

II

]

l

I

I

l

1

P

J3 0

T

I

II

I l

•

[

I

P

l | l l m l l

14 S ]

I ]

1

1

I

I

I

16

I

I 60

0

75

1

I

1

Contrazioni

l lll I l l l

I

I

I

9O

I

0

I

I

t

[lS

g

I

I

I

130

I

[

I

]

145

]

I

I

0

I

I

I

I 75

Fig. I. Example of output print of a 2CTG tracing. The variables considered in this study are: baseline FHR (Fbr Base), number of accelerations (Accelerazioni), number of decelerations (Decelerazioni), and one of the long-term variability index (Delta ms).

> 10 bpm lasting > 60 rain) and the type of decelerations (early, variable, late). Four observers were independently asked to analyze each strip for the same variables: baseline heart rate categorized in 10 bpm, long-term variability (amplitude <5 bpm, between 5 and 10 bpm, > 10 bpm), number of large accelerations and number and type of decelerations. Two of the observers were consultants with experience of CTG readings ('experts'); the other two were residents with 1 year of experience ('non-experts'). To make possible the comparison between observers and computer readings the computer baseline heart rate was categorized in 10 bpm and the variability as 6 < 5 bpm, between 5 and 10 bpm, > 10 bpm. The decelerations were classified by one of the authors, who did not read the tracings, on the basis of the computer description (number of bpm below the baseline, duration, onset compared to the onset of the contraction), according to the following criteria: early deceleration (no delay between onset of contraction and onset of deceleration and maximum descent from the baseline < 30 bpm), late decelerations (delay of 30 min between onset of contraction and onset of deceleration and maximum descent from the baseline a 20 bpm), variable decelerations (descent from baseline > 30 bpm, variable onset).

The r statistic [12] was used to assess the reproducibility among observers and between each observer and the computer readings for each of the 5 variables (baseline heart rate, variability, accelerations, number of decelerations and type of decelerations). The coefficient measures the extent of agreement beyond that expected by chance alone. A r value <0.40 indicates poor agreement, between 0.40 and 0.75 fair to good agreement and >0.75 excellent agreement. The underlying assumption is that none of the methods used (computer reading or expert reading) is a measure of truth.

Table I Reproducibility among observers Variable

• value

FHR Variability Accelerations Decelerations

0.65 0.38 0.58

-- N u m b e r

0.67 0.05

- - Type

T. Todros et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 68 (1996) 83-86

and decelerations) the agreement is almost always fair to good. Only 17 tracings showed decelerations. The K value between observers and 2CTG for the type of decelerations is very low, ranging from 0.01-0.25.

Table 2 Kappa values between observers and computer readings

EXPi vs. 2CTG EXP2 vs. 2CTG NEI vs. 2CTG NE2 vs. 2CTG

FHR Variability Accelerations (number)

Decelerations (number)

0.48 0.74 0.18 0.16 0.24 0.65 0.36 0.69

0.45 0.41 0.54 0.54

0.58 0.64 0.37 0.48

4. Discussion

Our results are further evidence that interobserver reproducibility in CTG reading is, at best, good. The data about decelerations must be interpreted with caution, because they were few in our tracings. At variance with other studies [3,5,13] the agreement on the number of decelerations was good; but our study confirms the poor agreement on the type of decelerations reported by the same authors [3,13]. The agreement between 'experts' and computer and the agreement between 'non-experts' and computer readings is similar. This underscores the fact that experience does not affect reproducibility. The small effect of experience, measured as number of years of CTG use or number of patients tested per months, on reliability had

EXP, expert; NE, non-expert.

3. Results

The overall reproducibility among observers ranges from fair to good for most of the variables (Table 1). There are no significant differences when the 'experts' or the 'non-experts' readings are considered. Reproducibility between observers and computer readings is shown in Table 2. The agreement between any of the observers and 2CTG is poor for FHR baseline. For the other variables (variability, accelerations

Ist.

Gineoolo=ia

• Ostetrlola

200

80

......

~- ......

~ .......

! .......

~" ......

~ .......

i .......

~ ......

~ .......

:" ......

? .......

: .......

dl

G.O.

- Unive:slt~

Gestazione:

? ......

~i

Torlno

39 Serf. e 4 gg.

~

.......

:" ......

; ......

~ .......

" ......

~ .....................

................................. $ - - ~ ....................................................... k z ~ A ...... , ~ - a - ........... ~ ....................... i-:;J~ . . . . . . . . . ~

z ~ , o -F . . . . . .

,o

" .......

- cart.

del 13-Jan-1993

Esame di:

85

......

~.

.

i ..............

I, 25

.

.

.

; .......

.

~"~

i .......

............

!~i~,~---~---T~~

~- ......

~ .......

i .......

; ......

,I

,I

I

I

- .......

I

............................

~- ......

,I

30

'

i .......

i .......

~ ......

4 .......

i .......

; ......

,I

,I

I

,I

I

'I

35

~

|

'-

,~ .......

--~ . . . . . .

~- ......

i .......

i

i

40 ~ a l W I

Accelerazlonl Dece~azioni Contrazioni

T~;"¥~-"

I

;

i .......

i ......

.~ .......

', 45

I

ww

7

~

"

~

i- ......

i ......

i

zz

1 50

am

6

............

9

BB

I0

ii

......................................................................................................................................................................................

'°

i

i

2O 0

25

30

"~5

~0

~5

50

Fig. 2. Example of a tracing in which the computer interpretation of the baseline FHR is incorrect: as a result a large number of decelerations is identified by 2CTG reading (D = deceleration).

86

T. Todros et al. /European Journal of Obstetrics & Gynecology and Reproductive Biology 68 (1996) 83-86

already been pointed out by Hage et al. [8]. The agreement between observers and computer readings is similar to the agreement among observers. In this scenario, the use of a computer system should give one important advantage: it overcomes the problem of intra and interobserver reproducibility. We did not take into account intraobserver variations, but it is known from other studies [5,7] that it is only a little better than interobserver reproducibility. The advantage is independent from the degree of experience of the clinical staff. It is expected that improving reproducibility will positively affect the performance of the test. Our study did not address this issue. One study [14] reports only a slight improvement in sensitivity and specificity using a computerized system. Unlike observers, 2CTG also computes short-term variability. At the moment there is no demonstration that short-term variability provides more useful clinical information than long-term variability. However this is a field that warrants further investigation. The 2CTG system does not have a rating scheme and the interpretation of the results is up to the clinician. It must be underlined that the tracings should always be observed carefully, since sometimes there are artifacts that can heavily affect the computer reading (Fig. 2). In conclusion, the use of computerized CTG can give some advantages over the traditional electronic fetal heart rate readings. Whether and to which extent it will improve the effectiveness of electronic fetal monitoring has to be proved.

Acknowledgements We are grateful to Dr. M. Ferrero and M.G. Ball for reading the tracings. This study was supported by a Grant of the Italian National Research Council (CNR) no. PF 41-9500639.

References [I] Thaeker SB, Bcrelman RL. Assessing the diagnostic accuracy and efficacy of selected antepartum fetal surveillance techniques. Obstet Gynecol Surv 1986; 41: 121-141. [2] Mohide P, Keirse MJNC. Biophysical assessment of fetal wellbeing. In: Chalmers I, Enkin M, editors. Effective care in pregnancy and childbirth. Oxford University Press, 1989; 477-492. [3] Donker DK. Interobserver variation in the assessment of fetal heart rate recordings. Thesis. Amsterdam: VU University Press, 1991. [4] Trimbos JB, Keirse MJNC. Observer variability in assessment of antepartum cardiotocograms. Br J Obstet Gynaecol 1978; 85: 900-906. [51 Lotgering FK, Wallenburg HCS, Schouten HJA. Interobserver and intraobserver variation in the assessment of antepartum cardiotocograms. Am J Obstet Gynecol 1982; 144: 701-705. [6] Flynn AM, Kelly J, Matthews K, O'Conor M, Viegas O. Predictive value of, and observer variability in, several ways of reporting antepartum cardiotocographs. Br J Obstet Gynaecol 1982; 89: 434-440. [7] Borgatta L, Shrout PE, Divon MY. Reliability and reproducibility of nonstress test readings. Am J Obstet Gynecol 1988; 159: 554-558. [8] Hage ML. Interpretation of nonstress tests. Am J Obstet Gynecol 1985; 153: 490-495. [9] Stigsby B, Nielsen PV, Docker M. Computer description and evaluation of cardiotocograms: a review. Eur J Obstet Gynecol Reprod Biol 1986; 21: 61-86. [10] Dawes GS, Mouiden M, Redman CWG. The advantages of computerized fetal heart rate analysis. J Perinatal Med 1991; 19: 19-45. [11] Arduini D, Rizzo D, Piana G, Bonalumi A, Brambilla P, Romanini C. Computerized analysis of fetal heart rate: I. Description of the system (2 CTG). J Matern Fetal Invest 1993; 3: 159-163. [12] Cohen J. A coefficient of agreement for nominal scale. Educ Psychol Meas 1960; 20: 37-40. [13] Arduini D, Rizzo G, Giannini F, Garzetti G-G, Romanini C. Computerized analysis of fetal heart rate: II. Comparison with the interpretation of experts. J Matern Fetal Invest 1993; 3: 165-168. [14] Hiett AK, Devoe LD, Youssef A, Gardner P, Black M. A comparison of visual and automated methods of analyzing fetal heart rate tests. Am J Obstet Gynecol 1993; 168: 1517-1521.