We conducted this study at Butaro, Kirehe and Rwinkwavu District Hospitals, three rural facilities that are operated by the Rwandan Ministry of Health with support from Partners In Health, a Boston-based non governmental organization. Rwinkwavu and Kirehe District Hospitals are located in the Eastern Province and Butaro District Hospital is located in the Northern Province of Rwanda. Together, the catchment area of these hospitals includes 46 health centers and serves more than 800,000 people . These districts are in the rural settings where 49% of the population make 30 min travel to get to clean water source, and only 12% have access to electricity. The population in this setting is likely to work in small farming agricultural sector and 77% of poorer households work in agricultural sector compared to 21% of richer ones that work in the same sector .
According to Rwanda’s protocol on labor and delivery, pregnant women first present to the health center for delivery led by nurses. If the delivery is complicated to be managed at the health center, i.e. nurses trained in emergency obstetric care detect obstetrical emergencies such as mal-presentation, cord prolapse and others, then the mother is referred to the district hospital and transported using a Ministry of Health ambulance. District hospitals are equipped with obstetrical theatres, trained personnel and equipment to perform cesarean delivery, the recommended procedure for high-risk labor. Some health centers have their own ambulance (including 11 out of 46 health centers in the three study districts) and the remaining health centers call to the district hospital for assistance. After being admitted to the district hospital, a midwife or nurse will assess the pregnant woman and determine whether a cesarean section should be considered. A general practitioner makes the final decision for a cesarean section and leads the procedure.
In 2015, there were 7569 deliveries at the three district hospitals. Of all deliveries, 39% were through cesarean section and were considered high risk as they were referred from health centers due to complications or previous complications. More than 90% of deliveries are performed by skilled doctors, nurses and midwives . These hospitals are staffed with 11–20 midwives, 73–90 nurses and 9–14 general practitioners. In the maternity department, two nurses or midwives staff the delivery room during the day or night. One general practitioner is assigned to delivery room during the day, but at night, one general practitioner services the whole of maternity department. Each of the three district hospitals is equipped with at least two operating tables, basic anesthesia equipment and supplies to deliver regional and general anesthesia, four obstetrics tables, thirty ward beds, six incubators and two lamps. A mother who has community-based health insurance is responsible for 10% of the total cost, with the remaining 90% paid by the government. Otherwise, the mother is responsible for all related delivery costs.
Study design and population
This retrospective cross-sectional study included neonates born to mothers undergoing emergency cesarean section between 01 January and 31 December 2015 at Butaro, Kirehe and Rwinkwavu District Hospitals. Due to the large population of cesarean deliveries at the district hospitals and insufficient resources to collect data on all deliveries, we randomly sampled 200 cesarean deliveries per hospital.
The study’s inclusion criterion was emergency cesarean delivery without intrauterine death prior to the decision for cesarean section. Women who received elective cesarean section and those with non-emergent indications for cesarean section were excluded. For sample selection, because the majority of cesarean sections at district hospitals are due to the referral of complicated cases from health centers, we first assumed all cesarean section deliveries were emergency and used segmented sampling to sample 200 women in each hospital. Authors NN and YL led the sampling procedures. All cesarean section charts were grouped by month, a sampling interval determined and charts selected by skipping the interval. During sampling, women who had scheduled or requested for cesarean delivery (charts marked elective) and women with neonatal intrauterine death prior to the decision to conduct the cesarean section were removed and randomly replaced with other women who met inclusion criteria from same month of hospital admission. A total of 597 cesarean deliveries were sampled. During data cleaning, we excluded women with no indication that the cesarean delivery was truly emergent (this included women whose only indication was prior cesarean delivery and women who were post-term or with twins but no other emergent indication). The following diagnoses were considered emergent and included in the study: uterine rupture, fetal distress, cord prolapse, abruption placenta, preeclampsia prolonged rupture of membranes, cephalo-pelvic disproportion, prolonged labor, and mal-presentation.
Data collection and analysis
Demographic and clinical characteristics of mothers and neonates as well as neonatal outcomes were extracted from charts into a pretested data collection form. Data were then transferred into an Access electronic database. Travel time from health center to district hospital was calculated using an average estimate of the time for an ambulance to travel from the mother’s health center of first presentation to the district hospital. We could not obtain an actual measure for a specific mother’s travel time as that information is not routinely collected. The severity of the indication for cesarean section was classified into two categories, very severe and severe, based on complication to the fetus and through consultation with a local obstetrician. Very severe indication included uterine rupture, fetal distress, cord prolapse, and abruption placenta and severe indication included preeclampsia, prolonged rupture of membranes, cephalo-pelvic disproportion, prolonged labor, and mal-presentation. In the case of more than one indication, we categorized according to the most severe indication. Neonatal outcome was dichotomized into poor neonatal outcome, defined as either death within 24 h or an APGAR score less than 7 at 5 min, or good neonatal outcome, defined as alive with an APGAR greater than or equal to 7 at 5 min.
We described mothers’ demographic and clinical characteristics using frequencies and percentages for categorical data. We tested each variable’s relationship to neonatal outcome using a Chi-squared test or in the case of small cell counts, a Fisher’s exact test. Factors significant at α = 0.20 level were retained for consideration in the multivariate logistic regression model. We developed the final model using backward stepwise selection, such that only factors significant at α = 0.05 significance level were retained in the final model. We report odds ratios (ORs), 95% confidence intervals (95%CIs) and p-values. For univariate and bivariate analyses, missing data were ignored, but we reported a reduced sample size in the case of missing data. For multivariate analysis, if a variable was missing for 15% or more observations, then we included a missing category in the model. All analyses considered the neonate as the unit of analysis and were completed in Stata v14.0 (College Station, TX: StataCorp LP).