Caesarean section performed by medical doctors and associate clinicians in Sierra Leone

Background: Many countries lack sufficient medical doctors to provide safe and affordable surgical and emergency obstetric care. Task-sharing with associate clinicians (ACs) has been suggested to fill this gap. The aim of this study was to assess maternal and neonatal outcomes of caesarean sections performed by ACs and doctors. Methods: All nine hospitals in Sierra Leone where both ACs and doctors performed caesarean sections were included in this prospective observational multicentre non-inferiority study. Patients undergoing caesarean section were followed for 30 days. The primary outcome was maternal mortality, and secondary outcomes were perinatal events and maternal morbidity. Results: Between October 2016 and May 2017, 1282 patients were enrolled in the study. In total, 1161 patients (90⋅6 per cent) were followed up with a home visit at 30 days. Data for 1274 caesarean sections were analysed, 443 performed by ACs and 831 by doctors. Twin pregnancies were more frequently treated by ACs, whereas doctors performed a higher proportion of operations outside office hours. There was one maternal death in the AC group and 15 in the doctor group (crude odds ratio (OR) 0⋅12, 90 per cent confidence interval 0⋅01 to 0⋅67). There were fewer stillbirths in the AC group (OR 0⋅74, 0⋅56 to 0⋅98), but patients were readmitted twice as often (OR 2⋅17, 1⋅08 to 4⋅42). Conclusion: Caesarean sections performed by ACs are not inferior to those undertaken by doctors. Task-sharing can be a safe strategy to improve access to emergency surgical care in areas where there is a shortage of doctors. This was a prospective observational multicentre non‐inferiority study of women who underwent caesarean section, including laparotomy for uterine rupture. All hospitals in Sierra Leone where both ACs trained in surgery and doctors were performing caesarean section at the start of the study interval were invited to participate in the study. Women who had caesarean sections done by either an AC or doctor as the primary surgical provider were eligible for inclusion. Patients were excluded if the fetus weighed less than 500 g or if essential data were missing. After oral explanation of the study, written consent was obtained either before, or as soon as possible after, the procedure. The study protocol (Appendix S1, supporting information) was approved by the Sierra Leone Ethics and Scientific Review Committee and the Regional Committees for Medical and Health Research Ethics in central Norway (ethical clearance number 2016/1163), and registered at the International Clinical Trial Registry (ISRCTN16157971). In each hospital, anaesthesia team members were trained to enrol patients in the study and to collect the in‐hospital data. The primary investigator collected and reviewed the data by undertaking hospital visits at 1–3‐week intervals, at which time the anaesthesia nurses were also mentored in enrolment and data collection. The data were entered into a Microsoft Excel® (Microsoft, Redmond, Washington, USA) database on location. Missing or inconsistent data were supplemented from operation logbooks or patient files. Financial incentives were given to the anaesthesia nurses based on the number of patients included in the study. Follow‐up home visits were done from 30 days after the caesarean section by one of four trained research nurses, who also assisted the anaesthesia team members with the collection of in‐hospital data. The research nurses were supervised by the primary investigator biweekly. During the home visits, women received an incentive in the form of a health promotion package with basic sanitary items. In‐hospital outcome data were validated during the follow‐up home visits. For patients lost to follow‐up, only the data collected during hospital admission were analysed. The primary outcome of the study was perioperative maternal mortality, defined as maternal death during caesarean section or within 30 days after the operation. Perioperative maternal mortality was subdivided into intraoperative death, in‐hospital death and death after discharge. Secondary outcomes were perinatal events and maternal morbidity parameters. Perinatal events included stillbirth, perinatal death and neonatal death. Stillbirth was classified as macerated where the fetus showed skin and soft tissue changes suggesting death occurred before the start of the delivery, and fresh where the fetus lacked such skin changes15. Neonatal deaths were divided into early (within 7 days after delivery) and late (between 8 and 28 days after delivery) deaths. Perinatal deaths were defined as the sum of fresh stillbirths and early neonatal deaths. Maternal morbidity parameters included: blood loss exceeding 600 ml, reoperation, readmission, wound infection and postoperative pain. Presence of persistent postoperative abdominal pain and readmission were surveyed during home visits. Wound infections and reoperations were either reported during admission or assessed during the home visit. In addition, duration of operation (interval from incision to final closure) and duration of hospital stay (excluding readmission) were recorded. The sample size calculation was based on the non‐inferiority assumption that caesarean sections performed by ACs are non‐inferior to those done by doctors for the primary outcome perioperative maternal mortality. Comparable studies reported a maternal mortality rate after caesarean section between 0·8 and 2·0 per cent16, 17. As Sierra Leone has the world’s highest maternal mortality rate, the upper limit of 2·0 per cent was used. In a previous meta‐analysis9, the lower bound of the confidence interval was an odds ratio (OR) of 2·75, which, with an average mortality rate of 2·0 per cent, led to a suggested non‐inferiority margin of 5·5 per cent. By applying a conservative approach and taking into account the importance of the outcome measure mortality, the non‐inferiority margin was set at 2·5 per cent (equivalent to an OR of 2·31 with a 2·0 per cent mortality rate)18. With α = 0·05 and β = 0·10, an expected success rate in both groups of 98 per cent and a non‐inferiority limit of 2·5 per cent, the total required sample size was calculated to be 107619. With an anticipated loss to follow‐up of 10 per cent, inclusion of a total of 1195 patients was required. Baseline and operative characteristics are presented as numbers with percentages and mean(s.d.) values. Missing data are indicated in the tables. Student’s t test was used for comparison of numerical means and Fisher’s exact test to compare categorical data. ORs were calculated by exact logistic regression and presented with 90 per cent confidence intervals, corresponding to a significance of 0·05 (α) for testing in a non‐inferiority analysis18. For the primary outcome, perioperative maternal death, both crude ORs and ORs adjusted for clusters using exact logistic regression are presented. P < 0·050 was considered statistically significant for equality tests. Statistical analyses were performed with Stata® 15.1 (StataCorp, College Station, Texas, USA). The primary data are available from the corresponding author on request.