Accuracy of Ultrasound Estimation of Fetal Weight at Three Teaching Hospitals in Addis Ababa

Ultrasound examination and measurement of fetal weight have become an integral part of modern obstetric care. Knowing the accuracy of the ultrasound estimation of fetal weight (EFW) and what factors affect it helps make a more informed decision. Fetal weight estimation is thought to help predict fetal survival and make management decisions in the low birth weight group and in managing the delivery of the large baby, where complications may occur. To The study design was a facility-based prospective cross-sectional study comparing sonographic EFW with actual birth weight, involving 384 women admitted to the study hospitals. And assessed 12 factors that might affect accuracy (place of delivery/scan, level of the operator, time of scan (working vs. non-working hours), fetal presentation, liquor volume, placental location, days passed between scan and delivery, time of the scan to labor (antepartum vs. intrapartum), fetal sex, maternal BMI, gestational age, and fetal weight. Data were collected using structured questionnaires, developed based on study variables and available literature, and analyzed using SPSS software.


Abstract Background
Ultrasound examination and measurement of fetal weight have become an integral part of modern obstetric care. Knowing the accuracy of the ultrasound estimation of fetal weight (EFW) and what factors affect it helps make a more informed decision. Fetal weight estimation is thought to help predict fetal survival and make management decisions in the low birth weight group and in managing the delivery of the large baby, where complications may occur.

Objective
To assess the accuracy of ultrasound estimation of fetal weight.

Methods
The study design was a facility-based prospective cross-sectional study comparing sonographic EFW with actual birth weight, involving 384 women admitted to the study hospitals. And assessed 12 factors that might affect accuracy (place of delivery/scan, level of the operator, time of scan (working vs. nonworking hours), fetal presentation, liquor volume, placental location, days passed between scan and delivery, time of the scan to labor (antepartum vs. intrapartum), fetal sex, maternal BMI, gestational age, and fetal weight. Data were collected using structured questionnaires, developed based on study variables and available literature, and analyzed using SPSS software.

Conclusion and recommendation
The EFW of R4's was signi cantly different from that of R1's and R2's, but not R3's we recommend that all ultrasound scans whose weight estimation can affect the management plan of the patient to be done by the senior residents rather than the junior residents i.e. the R3's and R4's.

Background
Obstetrics took a great leap forward with the advent of ultrasound in the second half of the 20th century with the ability to image the pregnant uterus and its contents. EFW in late pregnancy or before delivery is often an important variable in clinical decision-making in obstetrics. Fetal weight estimation is of key importance in the decision-making process for obstetric planning and management (1).
Ultrasonographically, EFW is obtained from measurements of fetal parts and the use of these measurements in a regression formula to derive birth weight. Obstetric ultrasound has turned from a luxuries instrument that is used seldom into an almost basic and integral part of the obstetric examination.
When scans are done by well-trained sonographers, most studies show a mean absolute error in birth weight prediction of approximately 8-10% of the birth weight (2). Perhaps a more telling statistic is that in about 30% of cases, the absolute error in estimated weight is greater than 10% of the birth weight. In most of the remaining cases, the error is less than 20% of the birth weight. Antenatal care has focused on the diagnosis of fetal growth restriction, which may be associated with iatrogenic premature delivery, intrauterine fetal compromise, or intrauterine demise, which will all result in decreased fetal weight for the gestational age. On the other hand, the delivery of macrosomic infants is equally important and is associated with higher rates of adverse outcomes for both mothers and infants in comparison to the delivery of normal-weight infants. Risks to the large infant include shoulder dystocia, brachial plexus injury, perinatal asphyxia, and neonatal death. Adverse maternal outcomes include prolonged labor, genital tract trauma, postpartum hemorrhage, and a higher risk of cesarean section (4,5). Accurate prediction of both small and large infants plays an important role in obstetric clinical practice.
The accuracy of EFW is documented as the percentage of correctly estimated fetal weights (usually stated as < 10% difference from the actual birth weight), and the studies put the results indicate how number of the scans were within 10% of the actual birth weight, and some also assessed that many scans was within 5%, 10%, or 20% of the actual birth weight after delivery. Some just put the mean absolute errors ± standard deviation of ultrasound fetal weight estimations.
There have been A large number of studies in which the accuracy of ultrasound for predicting birth weight has been evaluated (6,7).
Ultrasound results in the ± 10% of actual birth weight were 62.9% in a study done in Italy in 2008 involving 589 pregnant women using 35 different formulas. Only patients delivered within 48 hours were included and concluded that most formulas are relatively accurate at predicting birth weight up to 3,500 g, and all algorithms tended to underestimate large fetuses (8).
When we see studies done on our continent with residents doing the scan, one study in Tunisia in 2015 involving 500 pregnancies was scanned on day of delivery by Ob-Gyn residents showed a result in the range of ± 10% in 75.2% of the cases. Linear regression analysis revealed a good correlation between EFW and birth weight (r = 0.79, p < 0.0001). In conclusion, the EFWs calculated by residents were as accurate as those calculated by experienced sonographers. Nevertheless, predictive performance remains limited, with low sensitivity in the diagnosis of macrosomia (9).
There are also different studies done to assess the factors affecting the accuracy of the EFW. One of the studies concluded that advancing training among residents signi cantly decreased the percent error and reached acceptable levels of more than 70% of estimates within 10% of birth weight after 24 months of ultrasonographic experience (10), while a study comparing Ob-Gyn residents to maternal-fetal medicine subspecialists showed more experienced maternal-fetal medicine subspecialists to be more accurate (11), while there was one study in our university that assessed only sonographic accuracy of fetal head circumference measurement and concluded that the experience of US performers was not found to be associated with the accuracy of the measurement (12).
In another study, a systemic literature review was done on factors in uencing EFW. The review was done in Switzerland in 2014 and involved 29 studies and investigated 14 variables and found that GA, maternal BMI, AFI, and ruptured membranes, presentation of the fetus, location of the placenta, and the presence of multiple fetuses do not seem to have an impact on EFW accuracy. The in uence of the examiner's grade of experience and that of fetal gender were discussed controversially. Fetal weight, the time interval between estimation and delivery, and the use of different formulas seem to have an evident effect on EFW accuracy. No results were obtained on the impact of active labor (13).
Knowing the EFW accurately will help in the decision process. Data on how the experience of the sonographer affects the accuracy of ultrasound EFWs are sparse. There are even fewer studies that assess the accuracy of ultrasound EFWs made by obstetrics and gynecology (Ob-Gyn) residents alone (10,11,14,15). When we come to our setup, there are no studies that access the accuracy of EFW done by any level of professionals. There is only one study done that assessed the accuracy of Johnson's formula and the palpation method of EFW done in 2004 (16), and another study was done in 2014 assessing "Sonographic accuracy of Fetal Head Circumference Measurements" (12).

Methods
This is a facility-based prospective cross-sectional study that was conducted on all pregnant women who came to deliver in the hospitals and had US estimation of fetal weight in Addis Ababa, who is in the catchment area of the three hospitals, two teaching hospitals of Addis Ababa University-medical faculty hospitals Tikur Anbessa Specialized Hospital (TASH) and Zewditu Memorial Hospital (ZMH) and St. Paul's Hospital Millennium Medical Collage's Hospital (SPHMMC), which are found in Addis Ababa. The study was conducted from April 1-June 30, 2016.
This study aimed to determine the accuracy of ultrasound estimated fetal weight performed by residents at the three teaching hospitals and identify factors affecting the accuracy of ultrasound EFW. All pregnant women who had had the US scans done in the above-mentioned hospitals and had delivered in the same hospital within seven days of the scan and ful lled the inclusion criteria.
Questionnaires were developed based on the study objectives and available literature. Data were collected by the residents assigned to the labor ward during the study period and duty hours. Fetal biometry was performed at all levels of residents was taken. The principal investigator supervised the data collection for the utmost quality.
Data entry and analysis were performed using SPSS version 20. The responses in the completed questionnaires were coded and entered into a data entry template. Descriptive statistics were computed for sociodemographic characteristics and maternal and fetal characteristics. A paired sample t-test was applied between the two means i.e actual birth weight and EFW. Both bivariate and multiple logistic regressions were computed to assess the presence and strength of the association between the dependent and independent variables. P-values less than 0.05 were taken as signi cant in multivariate regression. For each regression, odds ratios (with the accompanying p-values and con dence intervals) of the relationship were reported. Summary tables were used to describe the data.

Results
The total number of respondents was 397. The mean age of the study participants was 27.92 ± 4.9 years, with the minimum and maximum ages being 18 and 45 years of age respectively. The majority of the study participants 300(, 81.7%) were from Addis Ababa, whereas the rest were out of the city. Table 1 shows the maternal characteristics of the study participants. The mean birth weight of the deliveries was 2880gm with a standard deviation of 721gm, and the mean EFW was 2854gm with a standard deviation of 740gm.  The overall mean error of EFW was 9.1%, with an SD of ±7.1%. In other words, 229 of the 367 scans were within 10% of the actual birth weight and 138 of the scans were outside 10% of the actual birth weight.
The least-underestimated EFW was 30.17% of the actual weight. It was a 1800gm fetus estimated as 1257gm by a 2 nd -year resident, while the most overestimated EFW was 31.82% from the actual weight. It was a 2200gm fetus estimated as 2900gm by a 1 st -year resident.
A paired sample t-test was applied (P= 0.14). The standard error of the difference between the two means was also calculated, which was 17.34. This difference in standard error was not even equal to the actual difference between the two means (i.e., 25.6). So the difference in EFW and actual fetal weight was statistically insigni cant.
The measurement error and degree of inaccuracy of ultrasound EFW within 10% of actual birth weight were found to be the least in ultrasound scans done at SPHMMC.

Discussion
The rst intention of this study was to see the accuracy of our EFW scan. When we compare our results with other studies, the EFW that was within 10% of the actual birth weight is comparable with other studies. Our result was 62.4% within 10%, and among the other results maximum number was 91.6% and the minimum result was 60.6% among 10% and 49.4% within 10% for a subgroup in one of the studies.
The second intention of the study was to assess the factors affecting our scan. We assessed 12 factors that might affect our accuracy, they are the place of delivery (scan), level of the operator, time of scan (working hours vs. duty hours), fetal presentation, liquor volume, placental location, days passed between scan and delivery, time of the scan to labor (antepartum vs. intrapartum), fetal sex, maternal BMI, gestational age, and fetal weight.
When we see the other studies mentioned previously in the literature review, the only factor that persistently affected EFW accuracy was the time interval between estimation and delivery i.e. 0-7 days vs. 8-14 days (13,17). But our study only included scans done within 7 days of delivery and the EFW was not affected among the scans done in this interval.
One factor not assessed by any of the studies and one systemic literature review involving 29 studies stated "No results were obtained on the impact of active labor" (13). In our study, active labor was not found to affect the accuracy of EFW.  (10) in the USA and concluded that there is a learning curve for ultrasonographic estimates of fetal weight, with a signi cant decrease in the percent error seen with advancing training among residents, reaching acceptable levels of more than 70% of estimates within 10% of birth weight after 24 months of ultrasonographic experience. Similarly, a study was done by Pks Yau et al. (11) in Hong Kong showed that more experienced Fetomaternal subspecialists were more accurate than the less experienced Ob-Gyn residents. But in a systemic review of 29 studies done by Juozas et al. (18) in Switzerland stated that experience was found to be controversial among the 29 studies 6 assessed skill/ experience of the examiner and it was discussed controversially. Meanwhile, in a study done by Ivo Markus et al. (17) in Germany, the experience did not affect the accuracy of EFW.

Conclusion
The difference between the mean birth weight of the deliveries and the mean EFW was statistically insigni cant. Showing that on average, the EFW was accurate enough measurement to be used for decision making. The EFW that was within 10% was 62.4%. The overall mean error of EFW was 9.1%, with an SD of ± 7.1%.
Of the 12 factors assessed that might affect the accuracy of EFW, it was found that the only place of delivery/scan and the level of resident experience signi cantly affected the accuracy. In place of delivery, the scans done at SPHMMC were found to be more accurate than scans done either at TASH or ZMH.
When we see the level of residency, we found that the accuracy of EFW goes up when the level of residency increased from R1 to R4. The most accurate being the R4's but the accuracy is not signi cantly different from R3's while the difference is signi cant when compared to R1's and R2's.

Limitation
The limitation of the study was that the study was facility-based and may not be generalized to the population. Besides, the use of only one sonographic model i.e. Hadlock's formula, only to derive estimates of fetal weights, while there was no study that this formula is the best for our population.

Recommendation
Because the EFW of R4's was signi cantly different from R1's and R2's but not R3's we can recommend that all ultrasound scans whose weight estimation can affect the management plan of the patient to be done by the senior residents rather than the junior residents i.e. R3's and R4's.

Declarations
Ethics approval and consent to participate Ethical clearance was obtained from the Research and Publication Committee (RPC) of the Department of Gynecology and Obstetrics of the College of Health Sciences, Addis Ababa University. Permission was also obtained from the study facilities to collect data. Participation in the study was completely voluntary, and informed written consent was acquired from every participant before participation. The study did not involve vulnerable populations.

Consent for publication
Not applicable since data were anonymized and no data on a speci c participant are presented.

Availability of data and materials
All data generated or analyzed during this study are included in this published article (Results section and tables within the manuscript).

Competing interests
The study has no nancial associations. The authors declare that they have no competing interests.