Individual and institutional determinants of caesarean section in referral hospitals in Senegal and Mali: A cross-sectional epidemiological survey

Background: Two years after implementing the free-CS policy, we assessed the non-financial factors associated with caesarean section (CS) in women managed by referral hospitals in Senegal and Mali.Methods: We conducted a cross-sectional survey nested in a cluster trial (QUARITE trial) in 41 referral hospitals in Senegal and Mali (10/01/2007-10/01/2008). Data were collected regarding women’s characteristics and on available institutional resources. Individual and institutional factors independently associated with emergency (before labour), intrapartum and elective CS were determined using a hierarchical logistic mixed model.Results: Among 86 505 women, 14% delivered by intrapartum CS, 3% by emergency CS and 2% by elective CS. For intrapartum, emergency and elective CS, the main maternal risk factors were, respectively: previous CS, referral from another facility and suspected cephalopelvic-disproportion (adjusted Odds Ratios from 2.8 to 8.9); vaginal bleeding near full term, hypertensive disorders, previous CS and premature rupture of membranes (adjusted ORs from 3.9 to 10.2); previous CS (adjusted OR=19.2 [17.2-21.6]). Access to adult and neonatal intensive care, a 24-h/day anaesthetist and number of annual deliveries per hospital were independent factors that affected CS rates according to degree of urgency. The presence of obstetricians and/or medical-anaesthetists was associated with an increased risk of elective CS (adjusted ORs [95%CI] = 4.8 [2.6-8.8] to 9.4 [5.1-17.1]).Conclusions: We confirm the significant effect of well-known maternal risk factors affecting the mode of delivery. Available resources at the institutional level and the degree of urgency of CS should be taken into account in analysing CS rates in this context. © 2012 Briand et al.; licensee BioMed Central Ltd. This secondary analysis included data extracted from a cluster-randomized controlled trial (QUARITE trial) in referral hospitals in Senegal and Mali. The protocol of the trial was approved by the ethics committee of Sainte-Justine Hospital in Montreal, Canada, and by the national ethics committees in Senegal and in Mali. The study protocol of the QUARITE trial and data collection procedures have already been published [15]. Briefly, the trial aimed to assess the effectiveness of the multifaceted Advances in Labour and Risk Management (ALARM) International Program – based on maternal death reviews – to reduce maternal mortality. Secondary goals included evaluation of the relationships between effectiveness and resource availability, service organization, medical practices that included CS rates, and satisfaction among health personnel. The trial was conducted in 46 out of a total of 49 eligible referral hospitals – 26 in Senegal and 23 in Mali – spread across both countries. A hospital was eligible for the trial if it had functional operating rooms and carried out >800 deliveries annually. Three eligible hospitals were excluded for the trial: two already had a structured programme for carrying out maternal-death audits before the project began, and the other hospital did not receive written consent from the local authorities. For the current analysis, we used the data collected during the first year of the trial – from October 2007 to October 2008 – when the ALARM intervention had not yet been implemented (i.e. pre-intervention phase of the trial). Therefore, there were no constraints or guidelines regarding investigations, treatments, admission and discharge decisions. Five hospitals out of the 46 included in the trial were excluded because four did not carry out any CS during the study period, and one only had data from mid-2008 (Figure ​(Figure1).1). All women who delivered in the 41 centres during the study period were included in the analyses, except those who lived outside Senegal or Mali, had a spontaneous abortion, and if the delivery date or mode of delivery was unknown. Flow chart. A total of 91,028 women delivered in the 46 referral hospitals selected for the QUARITE trial during the first year of the trial (from October 2007 to October 2008). Five hospitals were excluded from the analysis: four did not carry out any caesarean deliveries during the study period and one had data from mid-2008 only. $ Spontaneaous abortion was defined as birth weight less than 500 grams. Trained midwives who were supervised by the national coordinators of the survey collected data from medical records. In each country, data were collected on a daily basis on every woman who gave birth in every selected facility. It included: maternal demographic characteristics, obstetric history, prenatal care, management of labour and delivery, complications, and the vital status of both mother and child until hospital discharge. Pathologies during the current pregnancy and CS indications were reported using open questions and a pre-defined list of diagnoses or CS indications. The national coordinators of the study regularly verified that data collection was exhaustive (by comparing the number of eligible patients on the hospital’s birth register with the number of forms collected) and also checked data quality in a random sample of forms [15]. Between October 2007 and October 2008, 99% of the eligible women were included in the trial. The concordance rate – defined as the proportion of patient forms whose information was concordant with the hospital registers and medical records – was of 88% during the study period. Missing data for all variables accounted for <1% of cases, except for oxytocin use, which was missing for 5% of cases. For each institution, available resources were recorded in September 2007 and October 2008. A standardized inventory, developed by Villar [16], based on the WHO’s Complexity Index was used. This reflects the availability of different categories of resources required to provide high quality emergency obstetric care: basic services, screening tests, basic emergency obstetric resources, intrapartum care, general medical services, anaesthesiology resources, human resources, academic resources, and clinical protocols. Because resources changed during the study period, we split the study into period 1, from October 2007 to March 2008, and period 2, from April to October 2008. Women who delivered during periods 1 and 2 were assumed to have access to resources recorded in the first and second inventories, respectively. Regarding human resources, we created a categorical variable to distinguish between four different levels based on the number and qualifications of the medical staff: level I (‘reference’ group): general practitioners (GPs) trained in obstetrics, with nurse-anaesthetist(s) and two or less midwives; level II: trained GP(s), with nurse-anaesthetist(s) and three or more midwives; level III: at least one obstetric/gynaecology specialist, +/− trained GP(s), with nurse-anaesthetist(s) and three or more midwives; level IV: at least one obstetrics/gynaecology specialist, +/− trained GP(s), with at least one medical anaesthetist and three or more midwives. Mode of delivery was the main outcome of interest. Because the factors associated with CS differed according to the degree of urgency [14], we performed three distinct analyses, i.e. (i) emergency CS before labour (referred to as “emergency”) vs. all other deliveries, (ii) emergency intrapartum CS (“intrapartum”) vs. all vaginal deliveries, and (iii) elective CS vs. all delivery births with a trial of labour, which included both vaginal and intrapartum caesarean deliveries. No distinction was made between spontaneous vs. operative vaginal deliveries. For each type of CS, analysis was performed using a two-step procedure. As the first step, we assessed only individual factors, as they were expected to have the highest impact on CS likelihood. Potential individual risk factors were selected according to results obtained from previous studies in low- and middle-income countries [11,16-19]: age, parity, previous CS, multiple pregnancy (vs. single pregnancy), hypertensive disorders, vaginal bleeding near full term, suspected cephalopelvic-disproportion, suspected intrauterine death, premature rupture of the membranes, referral from another hospital, premature labour and oxytocin use. We considered that women did not have a condition if it had not been reported by a midwife. Obstetric complications that occurred during labour (i.e. obstructed labour or foetal distress) were not included in the analyses because they closely affected the decision regarding CS. Referral from another hospital was considered as a potential marker for more severe conditions because of delays due to large travel distances or lack of transportation. Both tri-variate (i.e., adjusted for the country and the period) and multi-variable analyses were performed. All variables, regardless of their association with CS in tri-variate analyses, were included in the multivariable model. They were all kept in the final model as they were independent and highly significant determinants of outcome (P<0.01). We used a conservative significance level to account for multiple analyses, and a very large sample size implied that any clinically relevant association was very significant. As the second step of analyses, we assessed which institutional factors were independently associated with CS, while adjusting for individual factors that were selected into the final multivariable model of the first step. Institutional factors considered for analysis were all items recorded in the standardized inventories (see the list of factors in Additional file 1). We did not use the Complexity Index, which aggregates the information on all available resources, but we tested each factor to determine which specifically influenced the decision for CS. Then, as in step one, all variables, regardless of their association with CS in tri-variate analysis, were considered for the multivariable analysis. In the final model, only those variables with a P<0.01, after a forward-stepwise procedure, were selected. We used a forward elimination procedure to account for very high sample size and high correlation between institutional variables. The level of qualification of the medical staff and the time period were forced into the final multivariable model. We used a logistic mixed model to account for the dependence of observations within hospital [20]. Indeed, including a random intercept to the model, assumed that women who delivered in the same hospital were more likely to have the same mode of delivery – because of common individual characteristics and shared institutional resources – than women who delivered in different hospitals. Also, we estimated the relative contribution of individual and institutional factors to the variability of each outcome (i.e., elective, emergency and intrapartum CS) between hospitals. In that purpose, we used the ratios of the random intercept variances [21]. To determine the effect of medical-staff configuration, we calculated the variation of elective CS rates between hospitals in women with either a low risk for CS (primiparous, <35 years old, with singleton pregnancy and no hypertensive disorders) or a high risk for CS (multiparous, >35 years old, with previous caesarean section and hypertensive disorders) in level I and IV hospitals [22]. All statistical analyses were performed using SAS system software (SAS Institute Inc., Cary, NC, USA). Hierarchical logistic mixed-regression models were estimated using the PROC NLMIXED procedure.