Maternal, Adolescent, Reproductive & Child Health (MARCH) Centre, London School of Hygiene & Tropical Medicine, Keppel Street, London, WC1E 7HT, UK.
Florence Nightingale Faculty of Nursing, Midwifery & Palliative Care, King's College London, London, UK.
BMC Pregnancy Childbirth. 2021 Mar 26;21(Suppl 1):226. doi: 10.1186/s12884-020-03238-7.
An estimated >2 million babies stillborn around the world each year lack visibility. Low- and middle-income countries carry 84% of the burden yet have the least data. Most births are now in facilities, hence routine register-recording presents an opportunity to improve counting of stillbirths, but research is limited, particularly regarding accuracy. This paper evaluates register-recorded measurement of hospital stillbirths, classification accuracy, and barriers and enablers to routine recording.
The EN-BIRTH mixed-methods, observational study took place in five hospitals in Bangladesh, Nepal and Tanzania (2017-2018). Clinical observers collected time-stamped data on perinatal care and birth outcomes as gold standard. To assess accuracy of routine register-recorded stillbirth rates, we compared birth outcomes recorded in labour ward registers to observation data. We calculated absolute rate differences and individual-level validation metrics (sensitivity, specificity, percent agreement). We assessed misclassification of stillbirths with neonatal deaths. To examine stillbirth appearance (fresh/macerated) as a proxy for timing of death, we compared appearance to observed timing of intrauterine death based on heart rate at admission.
23,072 births were observed including 550 stillbirths. Register-recorded completeness of birth outcomes was > 90%. The observed study stillbirth rate ranged from 3.8 (95%CI = 2.0,7.0) to 50.3 (95%CI = 43.6,58.0)/1000 total births and was under-estimated in routine registers by 1.1 to 7.3 /1000 total births (register: observed ratio 0.9-0.7). Specificity of register-recorded birth outcomes was > 99% and sensitivity varied between hospitals, ranging from 77.7-86.1%. Percent agreement between observer-assessed birth outcome and register-recorded birth outcome was very high across all hospitals and all modes of birth (> 98%). Fresh or macerated stillbirth appearance was a poor proxy for timing of stillbirth. While there were similar numbers of stillbirths misclassified as neonatal deaths (17/430) and neonatal deaths misclassified as stillbirths (21/36), neonatal deaths were proportionately more likely to be misclassified as stillbirths (58.3% vs 4.0%). Enablers to more accurate register-recording of birth outcome included supervision and data use.
Our results show these routine registers accurately recorded stillbirths. Fresh/macerated appearance was a poor proxy for intrapartum stillbirths, hence more focus on measuring fetal heart rate is crucial to classification and importantly reduction in these preventable deaths.
全球每年仍有超过 200 万婴儿死亡，但这些婴儿的情况并未得到充分关注。中低收入国家承担了 84%的负担，但这些国家的数据却最少。现在大多数分娩都在医疗机构中进行，因此常规登记记录为提高死产计数提供了机会，但相关研究有限，特别是关于准确性方面的研究。本文评估了医院死产的登记记录测量、分类准确性以及常规记录的障碍和促成因素。
EN-BIRTH 混合方法、观察性研究在孟加拉国、尼泊尔和坦桑尼亚的五家医院进行（2017-2018 年）。临床观察员收集了围产期护理和分娩结果的时间戳数据，作为金标准。为了评估常规登记记录的死产率的准确性，我们比较了产房登记记录的分娩结果与观察数据。我们计算了绝对率差异和个体水平验证指标（灵敏度、特异性、百分比一致性）。我们评估了将死产误诊为新生儿死亡的情况。为了检查死产外观（新鲜/木乃伊化）作为死亡时间的替代指标，我们将外观与根据入院时的心率观察到的宫内死亡时间进行了比较。
观察到 23072 例分娩，包括 550 例死产。出生结局的登记记录完整性>90%。观察到的研究死产率范围为 3.8（95%CI=2.0，7.0）至 50.3（95%CI=43.6，58.0）/1000 例总分娩，而在常规登记中低估了 1.1 至 7.3/1000 例总分娩（登记：观察比值 0.9-0.7）。登记记录的分娩结局特异性>99%，不同医院的灵敏度差异较大，范围为 77.7-86.1%。所有医院和所有分娩方式的观察者评估的分娩结局与登记记录的分娩结局之间的百分比一致性非常高（均>98%）。新鲜或木乃伊化的死产外观是死产时间的一个很差的替代指标。虽然有相同数量的死产被误诊为新生儿死亡（17/430）和新生儿死亡被误诊为死产（21/36），但新生儿死亡更有可能被误诊为死产（58.3%比 4.0%）。更准确地记录出生结局的促成因素包括监督和数据使用。
我们的结果表明，这些常规登记准确地记录了死产。新鲜/木乃伊化的外观是分娩时死产的一个很差的替代指标，因此更注重测量胎儿心率对于分类和重要的预防这些可预防的死亡至关重要。
