Disentangling the contributions of maternal and fetal factors to estimate stillbirth risks for intrapartum adverse events in Tanzania and Uganda

Abstract Objective To estimate the stillbirth risk associated with intrapartum adverse events, controlling for fetal and maternal factors. Methods The present study was an analysis of cross‐sectional patient‐record and facility‐file data from women with viable fetuses who experienced obstetric adverse events at 23 hospitals and 38 health centers in Tanzania (between December 2015 and October 2016), and 22 hospitals, 16 level‐4 health centers, and five level‐3 health centers in Uganda (between May 2016 and September 2017). Adverse events were categorized in three severity groups (postpartum, intrapartum non‐near‐miss, and intrapartum near‐miss) to calculate stillbirth rates and adjusted prevalence ratios. Results Data from 3816 women in Tanzania and 8305 in Uganda were included. Compared with postpartum adverse events, intrapartum near‐miss was associated with a 3.73‐ and 4.55‐fold higher prevalence of stillbirth in Uganda and Tanzania, respectively. Most women who experienced near‐miss had organ dysfunction on arrival or developed it soon after. The risk of stillbirth was higher among preterm deliveries compared with term deliveries, and was 42% and 59% lower in Tanzania and Uganda, respectively, for cesarean deliveries compared with vaginal deliveries after intrapartum non‐near‐miss adverse events. Conclusion Stillbirth risk increased with severity of complications and was higher among premature deliveries. Survival was higher for cesarean deliveries in intrapartum non‐near‐miss complications, identifying the opportunity to prevent deterioration by timely actions.


| INTRODUCTION
Perinatal mortality and morbidity are intimately linked to maternal mortality and morbidity. In recent years, the severe end of maternal morbidity, often referred to as maternal near-miss, has received recognition as an important indicator to assess healthcare service and program performance. 1 Since standardization of the maternal morbidity and near-miss concepts, 1-3 facility-based incidents of maternal morbidity have been reported from high-, middle-, and low-income countries. Partly linked to measurement challenges in definition and data capturing, 4 however, there is a paucity of studies analyzing stillbirth or quantifying the risk among near-miss women; this is despite the general understanding that delivery outcomes are particularly sensitive to the quality of intrapartum care and management of complications.
Stillbirth is associated with obstetric complications such as prepartum hemorrhage, ruptured uterus, pre-eclampsia, eclampsia, maternal anemia, and infection. 5,6 Fetal conditions associated with stillbirth include post-term pregnancy, small for gestational age, low birthweight, prematurity, and multiple gestations. [6][7][8] Maternal complications explain many severe fetal morbidities and mortalities during delivery, and thus appropriate and timely management of these complications has the potential to avoid many adverse fetal outcomes. 9 The large disparity in stillbirth rates seen in maternal near-miss cases across different countries, from 3.8% in Finland to as high as 46% in low-to middle-income countries, [10][11][12][13][14] is probably attributable to substandard management of complications. However, measurement challenges prevail in low-resource settings, 15 restricting a direct comparisons of rates across different settings.
The aim of the present study was to clarify the role of the severity of maternal complications or maternal near-miss in stillbirth, after taking into account associations between maternal (obstetric and reproductive) factors and fetal conditions in datasets from two low-income countries, Tanzania and Uganda. Maternal and fetal factors could be causal factors or consequences of maternal near-miss events; disentangling the relationships will enable the risk of stillbirth to be estimated in maternal near-miss events.

| MATERIALS AND METHODS
The present study used cross-sectional data collected as part of a trial evaluating the effects of 1-day competency-based "Helping Mothers Survive Bleeding After Birth" (HMS BAB) training to reduce postpartum hemorrhage (PPH)-related morbidity and mortality in Tanzania  was waived by the boards that approved the study because the data were anonymized, and identification of individuals not possible.
Details of the HMS BAB trial are described elsewhere. 16 In Tanzania, 23 hospitals and 38 health centers in 20 districts were included. All were government-owned facilities except for six mission hospitals or clinics. In Uganda, 22 hospitals, 16 level IV health centers, and five high-volume level III health centers offering some emergency obstetric care services were included. All were government-owned except for eight faith-based facilities and one NGO facility.
The WHO near-miss tool 2 was adapted for the study (File S1).
Maternity staff received a short training, after which they reviewed the prenatal, delivery, and postnatal registries, and patient case notes on a daily basis to identify women with complications. Data were abstracted by using the tool to capture information on obstetric complications (PPH, severe pre-eclampsia, eclampsia, sepsis/severe infection, ruptured uterus, severe complications of induced and spontaneous abortions, and prepartum hemorrhage), critical interventions, organ dysfunctions, maternal outcome, mode of delivery or end of pregnancy, and vital status of the neonate at delivery. The tool was similar in both countries, but minor adaptations were made to address country differences in data collection and practices. For example, the timing of a stillbirth (before or during delivery) was estimated by appearance of the skin in Uganda (fresh or macerated) but not in Tanzania. Stillbirths were not weighed in Tanzania and hence birthweight data were not collected.
Women were included in the study if they delivered a potentially viable fetus (≥1000 g or, if birthweight was unknown, ≥28 weeks of pregnancy) and experienced PPH, prepartum hemorrhage, eclampsia/ pre-eclampsia, sepsis, or ruptured uterus.
To classify the degree of severity and its potential effect on stillbirth, women were categorized on the basis of complications into three mutually exclusive risk groups (Table S1): a low-risk group, including those experiencing "postpartum complications" only (in the present study, PPH only); a medium-risk group including those experiencing prepartum or intrapartum complications without organ dysfunction or those without management-based criteria indicating severity (blood transfusion or hysterectomy) (collectively termed "intrapartum non-near-miss complications"); and a high-risk group including those experiencing prepartum or intrapartum complications and any organ dysfunction and/or management-based criteria indicating severity (termed "intrapartum near-miss complications"). As suggested by Nelissen et al., 17 the threshold of blood transfusion was lowered from 5 units to 2 units to define coagulation/hematologic dysfunction. 17 Analysis followed the conceptual framework (Fig. S1), which was based on previous studies. 4,5 Underlying causes contributing to stillbirth, factors associated with (but not directly contributing to) stillbirth, and factors on the causal pathway to stillbirth were considered separately in the analysis. Factors associated with stillbirth, in particular maternal factors (age and parity), were considered potential confounders because they might also relate to maternal complications.
Place and mode of delivery, and a fetal factor (gestational age) were on the causal pathway between complications and stillbirth, and therefore stratified analyses were conducted. Immediate causes of stillbirth (e.g., asphyxia or infection) were not measured and thus not included in analysis.   T A B L E 2 (Continued) models were used to estimate the adjusted prevalence ratio (aPR) of stillbirth in risk groups for ease of interpretation. 18 Stratified analysis was used to explore the mediating effects on stillbirth of factors on the causal pathways. Outcomes were imputed by using multiple imputation techniques of 20 data sets and the estimates were combined by using Rubin rules. 19 A sensitivity analysis was used to compare the results between observed data and imputed data (Fig. S2 and Tables S3-S5). P<0.05 was considered to be significant.

| RESULTS
During the study period, 83 520 and 163 559 deliveries were reported. Because the data were obtained from routine recording systems, delivery outcomes were missing for 730 and 459 deliveries in Tanzania and Uganda, respectively. There was no significant difference in maternal, delivery, and fetal factors between those with delivery outcomes and those without in Tanzania, except that mode of delivery was more likely to be unknown for women missing an outcome, and the majority of those without outcomes had intrapartum non-near-miss complications (Table S2). In Uganda, the majority of missing outcomes were from data collected in health centers and for women who had intrapartum near-miss complications. The mode of delivery for most women with missing outcomes was unknown (Table 2). In total, 3816 (654 high-risk, 1416 medium-risk, and 1746 low-risk) and 8305 (2374 high-risk, 1644 medium-risk, and 4287 low-risk) deliveries with obstetric complications in Tanzania and Uganda, respectively, were included in the study (Fig. 1). In both countries, the stillbirth rate was significantly higher in the  (Table S6).
Stratified analysis showed that the prevalence of stillbirth more than doubled for preterm deliveries as compared with term deliveries in the postpartum and intrapartum non-near-miss groups and increased by 68% in the near-miss group, after adjustment for types of complication, maternal age, and parity in Tanzania (Table 3).
The prevalence ratio of stillbirth for preterm deliveries relative to term deliveries varied between 1.27 and 3.99 across the three risk groups in Uganda.
After adjustment, a 42% (aPR 0.58, 95% CI 0.39-0.85) and a 59% (aPR 0.41, 95% CI 0.30-0.56) reduced risk of stillbirth in cesarean as compared with vaginal deliveries was observed in the intrapartum non-near-miss group in Tanzania and Uganda, respectively.
In Uganda, a 26% (aPR 0.74, 95% CI 0.60-0.92) reduced risk of stillbirth in cesarean as compared with vaginal deliveries was observed in the intrapartum near-miss group after adjustment. No significant difference in stillbirth rates were observed between hospital-based and health-center-based deliveries among intrapartum complication groups.

| DISCUSSION
The present large study, including 3816 and 8305 complicated deliveries in Tanzania and Uganda, respectively, found that there was a 4.55-and 3.73-fold higher risk of stillbirth risk when intrapartum complications developed into a near-miss situation. The risk of stillbirth was significantly lower for term than for preterm deliveries, and for cesarean than for vaginal deliveries, particularly when the complication did not develop into a near-miss situation.
The association between maternal near-miss and stillbirth has rarely been quantified in low-and middle-income countries. One multi-country study from Latin America reported an almost fourfold higher risk of stillbirth for women experiencing any near-miss complication, as compared with non-near-miss (including uncomplicated) deliveries, similar to the present finding. However, the Latin American study reported much lower stillbirth rates for maternal near-miss deliveries (37 per 1000 near-miss deliveries) as compared with the present estimate. 20 The present high stillbirth rates are consistent with those of a previous Ugandan study, reporting 120 stillbirths per 1000 deliveries with severe complications in a referral hospital, but lower than those of a Nigerian study, which documented 211 stillbirths per 1000 deliveries with severe maternal morbidity in a tertiary hospital. 12,21 Furthermore, the present study supports earlier findings that preterm delivery is a risk factor for stillbirth. 22 Considering the high prevalence of preterm deliveries among women with complications, the number of stillbirths associated with preterm delivery may be greater than suggested previously. The global prevalence of preterm delivery among the general population in low-income countries is estimated to be 11.8%. 23 In the present study, 37% and 28% of intrapartum near-miss women delivered preterm in Tanzania and Uganda.
Access to timely, high-quality intrapartum care is essential for the prevention of intrapartum stillbirth. Although the study hospitals were equipped to provide comprehensive emergency obstetric and neonatal care, the survival of neonates was poor in hospital deliveries. In maternal near-miss events, most women had a near-miss condition on arrival or within 12 hours of admission, suggesting that delays before or on admission contributed to the high stillbirth rate. Moreover, three-quarters of the stillbirths in Uganda were reported to have occurred during delivery.
As expected, the postpartum complication group had a lower risk than the intrapartum groups because complications occurred after delivery. Nevertheless, the stillbirth rate was still higher than estimates among the general population, 24 which may support the hypothesis that there is a pathological link between PPH and other obstetric complications such as undiagnosed hypertensive disorders (i.e., HELLP syndrome) and gestational diabetes, which leads to placenta disorders. In addition, in two-thirds of the patients with reported PPH, delivery was by cesarean. Although the reason for the cesarean was not recorded, it is likely that many procedures were performed for obstetric complications, including prolonged labor or fetal distress, which might explain the high number of stillbirths. In conclusion, the risk of stillbirth was found to be higher among patients who experienced intrapartum near-miss events than among patients who experienced intrapartum complications without nearmiss events. Prompt action to prevent the development of organ dysfunction in the mother, coupled with early management of complications, has the potential to save many newborns.

AUTHOR CONTRIBUTIONS
AH and CH conceived the study. In Tanzania, FA, CH, and ABP supervised the trial and data collection with support from JLM. In Uganda, SA, CH, and FK supervised the trial and data collection with support from JLM. AH performed statistical analysis with support from GM and CH. AH drafted the manuscript. All authors revised and approved the final manuscript.

ACKNOWLEDGMENTS
The HMS BAB trial was funded by the Laerdal Foundation (Number 40070) through a grant to the International Federation of Gynecology and Obstetrics (FIGO), which acts as a sponsor. The funder had no role in the study design, data collection, management, or analysis of the study.

JLM is employed by the International Federation of Gynecology and
Obstetrics (FIGO). The authors have no other conflicts of interest.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article. Figure S1. Conceptual framework.       File S1. Data collection tool.