Background
Worldwide, approximately 15% of live births (21 million) were low birthweight (LBW) in 2015; 91% from low- and middle-income countries (LMICs) [
1]. LBW, commonly caused by preterm birth and/or intrauterine growth restriction [
2], leads to a range of short- and long-term health impacts, including respiratory distress and feeding intolerance, growth impairment, developmental delay, and higher risk of diabetes and cardiovascular diseases [
3]. LBW newborns are twenty times more likely to die in the first year than normal weight newborns [
4].
Though recent estimates have shown that some progress has been made in reducing the risk of LBW [
1], intensified efforts are needed to meet the World Health Assembly global target of decreasing the proportion of infants with LBW by 30% (~ 14 million) by the end of 2025. Ethiopia set a goal to reduce the prevalence of LBW to 7% by 2025 from 10.8% in 2012 [
5]. Accurately measuring newborn weight at birth is crucial to provide special care for LBW infants, monitor the burden of LBW in the population, evaluate access to interventions aimed to improve antenatal care, and planning appropriate actions to accelerate the reduction of neonatal morbidity and mortality [
6]. In LMICs, however, the quality of birthweight data suffers from measurement and recording errors, inconsistent data reporting systems, and missing data from non-facility births [
1]. Improving measurement, recording, and reporting of birthweight are therefore warranted to target interventions and track progress toward the global nutrition target [
7]. Strengthening the existing routine health systems of LMICs has been recommended as an essential strategy to improve birthweight data quality [
8]. The
Every Newborn Action Plan endorsed by the World Health Assembly prioritizes measurement improvement, with a focus on strengthening routine facility-based data, to track the national 2030 milestones (≤ 12 neonatal deaths and ≤ 12 stillbirths per 1000 live births) [
9]. In LMICs, health system strengthening measures that include training of healthcare staff and supportive supervision showed to improve health facility data quality [
10,
11]. The Ethiopian National Newborn and Child Survival Strategy (2015/16—2019/20) aims to strengthen the existing health information management system for improving the percentage of live births with a reported birthweight from 5.2% in 2013 to 95% by 2020 [
12]. Notwithstanding, the recent global and national LBW estimates conducted by Blencowe and colleagues could not provide an estimated LBW prevalence from Ethiopia, due to lack of adequate birthweight data quality [
1].
The Enhancing Nutrition and Antenatal Infection Treatment (ENAT) study is a pragmatic effectiveness study testing the impact of optimizing prenatal nutrition status and infection control on birth outcomes in rural Amhara (ISRCTN15116516). Prior to the study, we introduced a birthweight quality improvement (QI) initiative in all study sites with the objective of improving birthweight data quality. Herein, we present the impact of QI on birthweight data quality measures and the prevalence of LBW before and after the implementation of the initiative.
Discussion
Improving quality of birthweight data is essential for identifying infants at high risk for medical and neurodevelopmental problems, determining the burden of LBW, and monitoring access and usage of interventions aimed to enhance prenatal care and newborn survival. We observed a reduction in percentages of heaping and rounding and an increase in the proportion of LBW by about 10 percentage points after the implementation of birthweight QI initiative (provision of high-precision digital weight scales, practical training and supportive supervision).
We found that introduction of QI decreased birthweight heaping exactly at multiples of 500 g (e.g., 1500 g, 2000 g, 2500 g, 3000 g, 3500 g, etc.). Heaping, commonly observed when scales with low precision are used or continuous data are rounded, represents misclassification [
18]. A recent study conducted in Bangladesh, Nepal and Tanzania showed significant heaping (19–67%) at 2500 g and 3000 g in health facilities [
16]. The amount of heaping on 2500 g can have a significant effect on LBW estimation. Some newborns whose birthweights rounded to exactly 2500 g may have been lighter and should actually be classified in the LBW category, while some may have been heavier. The effect of this type of misclassification is critical in determining the proportion of LBW [
18]. In this study, although pre- and post-initiative data appear symmetrical, the highest percentage of heaping observed at 3000 g (26%) pre-QI declined by three-quarters after the initiative. Heaping exactly at 2500 g also reduced from 5.4 pre-QI to 2.2% post-QI. Furthermore, 100% of data rounded to the nearest 100 g pre-initiative dropped by two-thirds post-initiative. In South Africa, a similar data improvement intervention, including training on data collection and providing feedback to healthcare staff with monthly data reviews and audits, increased the accuracy and completeness of health facility data [
10]. Health system strengthening measures such as performance review feedback activities and enhanced supervision have also shown to increase ownership of data among healthcare workers in Rwanda [
11]. The tendency to round birthweight may be due to the fact that there are no formal standards in recording birthweight within the health systems [
19], low precision and common use of spring birthweight scales [
20], and/or high workload of healthcare providers. Proper recording of birthweight data on health facility records, which can be used as a source of data for regional and national estimates, may improve the quality of birthweight data and decrease the need for statistical adjustments [
1].
Based on data from pre-initiative phase, the percentage of newborns with LBW was about 2%. However, the prevalence increased to 12% after birthweight QI implementation, which was more consistent with a meta-analysis that included 4105 participants from nine observational studies showing 16% of newborns were LBW in Amhara region [
21]. The 2016 Ethiopian Demographic and Health Survey report indicated a LBW prevalence of 22% in Amhara region, which was the highest nationally [
14]. Recent studies performed at Debre Tabor hospital in South Gondar zone, and at Dangla Primary hospital in West Gojjam zone found a LBW prevalence of 10–12% among facility births [
22,
23]. Estimation of the percentage of newborns with LBW is dependent on how newborns with a reported birthweight of exactly 2500 g are classified [
19]. Data from LMICs showed an increase in the prevalence of LBW from 1.7% to 7.2% after reallocating 50% of newborns with a birthweight of exactly 2500 g to the LBW category [
16]. In our study, when we included half of newborns with exactly 2500 g as LBW in the calculation, the percentage increased by 2.7% pre-QI but declined to 1.2% after the initiative. In addition to recording errors, the change in estimated LBW rates due to differences in instrument precision is also possible [
17]. A recent study reported lower heaping indices using digital scales compared to analog [
24]. Compared to the traditional method focusing on control, audit and examination, the new approach of supportive supervision with a focus on strengthening routine health system, problem solving and training healthcare providers shows superior results in improving essential newborn care [
25]. Our findings suggest that provision of digital weight scales together with periodic supervision may help to improve birthweight data quality and estimates of the true burden of LBW.
The present study has some limitations. Our findings may not be truly generalizable because women who give birth at home may differ in health and socioeconomic status from those who give birth in facilities, which may lead to different rates of LBW. Since the main aim was to improve quality of birthweight data in study health centers prior to commencing the parent ENAT study, we did not conduct proper sample size calculation and collect birthweight data from all health centers or home births throughout the study phases. Stillbirths were not routinely measured or reported and hence not included in the analysis. Given that the trial was underway, we did not repeat observations after an interval of time from the initial QI initiative, hence we could not assess whether the improvements were sustained. Despite these limitations, we were able to assess birthweight within hours of birth, prior to significant weight loss, using high-precision digital weight scales at the health facilities.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.