Menu
Menu
Menu
Menu
Background: Low birthweight (LBW) (< 2500 g) is a significant determinant of infant morbidity and mortality worldwide. In low-income settings, the quality of birthweight data suffers from measurement and recording errors, inconsistent data reporting systems, and missing data from non-facility births. This paper describes birthweight data quality and the prevalence of LBW before and after implementation of a birthweight quality improvement (QI) initiative in Amhara region, Ethiopia. Methods: A comparative pre-post study was performed in selected rural health facilities located in West Gojjam and South Gondar zones. At baseline, a retrospective review of delivery records from February to May 2018 was performed in 14 health centers to collect birthweight data. A birthweight QI initiative was introduced in August 2019, which included provision of high-quality digital infant weight scales (precision 5 g), routine calibration, training in birth weighing and data recording, and routine field supervision. After the QI implementation, birthweight data were prospectively collected from late August to early September 2019, and December 2019 to June 2020. Data quality, as measured by heaping (weights at exact multiples of 500 g) and rounding to the nearest 100 g, and the prevalence of LBW were calculated before and after QI implementation. Results: We retrospectively reviewed 1383 delivery records before the QI implementation and prospectively measured 1371 newborn weights after QI implementation. Heaping was most frequently observed at 3000 g and declined from 26% pre-initiative to 6.7% post-initiative. Heaping at 2500 g decreased from 5.4% pre-QI to 2.2% post-QI. The percentage of rounding to the nearest 100 g was reduced from 100% pre-initiative to 36.5% post-initiative. Before the QI initiative, the prevalence of recognized LBW was 2.2% (95% confidence interval [CI]: 1.5–3.1) and after the QI initiative increased to 11.7% (95% CI: 10.1–13.5). Conclusions: A QI intervention can improve the quality of birthweight measurements, and data measurement quality may substantially affect estimates of LBW prevalence.
A comparative pre-post study was conducted to determine the effect of QI on birthweight data quality and the proportion of newborns with LBW in rural Amhara, Ethiopia. Amhara region is subdivided into 13 administrative zones and 140 districts, and it has an estimated population size of over 21,000,000. Nearly 85% of the population live in rural areas. According to the 2019 Ethiopian Mini Demographic and Health Survey (EMDHS) report, in Amhara region, the percentage of institutional delivery and those receiving antenatal care from a skilled provider were 54.2% and 82.6%, respectively [13]. Approximately 23% of women of reproductive age had a body mass index of below 18.5 kg/m2 [14]. A retrospective facility record review was performed in 14 health centers located in seven districts of three zones in Amhara, namely South Gondar, West Gojjam and North Wollo, from February to May 2018 (Additional file 1). The health centers were chosen as prospective sites for the parent ENAT study, and the populations were selected based on high rates of maternal malnutrition, risk of fetal growth restriction, and need for nutrition interventions. Geomorphological maps were used to identify drought and famine-prone districts, and local government and partners were consulted. Two health centers from each district were selected based on the current antenatal care volume, using data from the region’s Health Management and Information System. After selection of the health centers for the parent ENAT study within the zones above (South Gondar, West Gojjam), birthweight QI was implemented in 12 study health centers and birthweight data were prospectively collected from late August to early September 2019, and December 2019 to June 2020. These study health facilities were selected in consultation with the Amhara Regional Health Bureau, Amhara Public Health Institute, and partners, with preference for health centers with higher ANC volume and access to transportation (Additional file 1). The phase 2 study was conducted with the objective of not only improving the quality of anthropometric measurements and recording, but also for testing study tools and procedures of the ENAT parent study. Birthweight measurement quality improvement initiative package A high-quality digital infant scale (ADE-M112600, Germany) (precision 5 g) was provided to each study health center for measuring the newborn weight at birth. Adequate weighing stations were set up within all health centers; scales were placed on a clean flat surface with the display clearly visible and calibrated every morning with the 1000 g and 2000 g standard weights. If the weight did not read between ± 10 g of the standardized weights, proper adjustments with the placement of scales were made until the correct measures were displayed. Health center staff (n = 1 to 4 midwives per health center) were trained on how to accurately measure birthweight using digital weight scales. Methods to weigh the baby with clothing using the tare function were developed, based on methods proposed by the International Fetal and Newborn Growth Consortium for the 21st Century (INTERGROWTH-21st) [15]. Where weight measurement of a naked newborn was not possible, a scale was first tared with a blanket or cloth, and weight of the baby with the clothing or blanket was taken. Training on recording the full precision of birthweight values, without any rounding to the nearest number was given to all personnel involved in measuring anthropometry. We developed job aids demonstrating how to properly measure and document birth weight using step-by-step pictorial procedures and posted them on the delivery room walls (Additional file 2). Ongoing field supervision was performed in all health centers by the field team (study physicians and field coordinators) throughout the post-QI phase. To ensure the standard operating procedures were followed, the field team observed the midwives measuring and recording the newborns’ weight, checked facility birthweight records (labor/delivery registers) for heaping and rounding, and provided feedback when appropriate. In addition, a review meeting with all health center midwives and directors was held after 2 weeks of implementing the initiative (phase 2 study). Quality of birthweight data extracted from health center records, lessons learnt and ways to forward were discussed. Birthweight data were extracted from delivery registers by trained research staff using the Survey Solutions® electronic data collection software (version 20.10, World Bank, Washington DC, USA). Routine field supervision was conducted in all sites, and collected data were reviewed and cleaned by the data management team daily. During phase 1, about 1400 health center delivery registers were retrospectively reviewed. All labor/delivery registers recorded from February to May 2018 (n ~ 100 records from each health center) were included. Birthweight was measured using fully and semi-functional analog/spring weighing scales available in the health center (Fig. 2), with routine delivery room measuring practices and record keeping. Birthweight scales used before and after initiative. A) Birthweight scale commonly used pre-QI. B) Birthweight scale used post-QI In phase 2, health center midwives measured the weight of newborns at birth (n ~ 100 per health center) using precise digital infant weight scales (Fig. 2). In line with the routine practice, health center midwives weighed the newborn only once and documented on labor/delivery registers. Birthweight data quality was assessed for implausibility, heaping and rounding. The presence of implausible values defined as extreme or unlikely birthweight values, i.e., 6000 g, was checked. Percentages of heaping exactly at multiples of 500 g (e.g., 1500 g, 2000 g, 2500 g, 3000 g, 3500 g, etc.) were calculated. Heaping index (HI) was also computed by dividing the number of exact weight values (e.g., 3000 g) by all weights within the adjacent 250 g brackets, excluding the exact values (e.g., 2750–2999 + 3001–3249) [16]. Proportions of birthweights rounded to the nearest 100 g increments were calculated. Histograms were constructed to visually inspect the birthweight distribution. To calculate the prevalence of LBW, the number of live births with a birthweight of less than 2500 g was divided by the total number of liveborn babies with reported birthweights. To examine the effect of birthweight heaping, prevalence of LBW was computed by including 50% of newborns who had exactly 2500 g as LBW [16, 17]. Data analysis was done using STATA v.15 (StataCorp LLC, College Station, TX, USA).
