STRENGTHS AND LIMITATIONS OF THIS STUDY

This study analysed weighted nationally representative Demographic and Health Survey (DHS) data from 21 sub-Saharan Africa (SSA) countries.

This study adopted weighted statistical methods at the countries and SSA scale.

The main limitation is that DHS data currently do not differentiate demanded caesarean section (CS) that are not medically indicated (for convenience purposes) among elective CS.

Introduction

Caesarean section (CS) is a critical surgical procedure carried out to give birth when vaginal birth delivery would lead to maternal or perinatal complications. CS can either be elective or emergency, each with its specific indications. Elective CS should be medically indicated in the case of any condition that compromises maternal or fetal survival and requires early fetal extraction for both mother and child safety (eg, severe pre-eclampsia, intrauterine growth retardation).1 As well as any situation at risk of dystocia during childbirth or infectious transmission in the event of vaginal birth delivery. Elective CS is usually planned before the due date of delivery and spontaneous labour, following multidisciplinary consultation involving neonatologists, perinatologists and gynaecologists. While emergency CS is carried out in an emergency when labour leading to vaginal delivery has already started and posed a threat to the mother or baby’s life, for example, this is when there is severe bleeding or severe eclampsia.2 However, among elective CS cases, there are those with non-medical indications that are considered as medically unnecessary CS.3 These medically unnecessary CS procedures are typically elective CS cases mostly performed in private health facilities for income purposes. Other medically unnecessary CS cases are among CS on demand, where women or their partners request a planned CS delivery for convenience purposes.4 These medically unnecessary CS procedures create additional challenges for health systems, since CS requires more financial, material and human resources, which are already limited in poor settings such as in SSA countries. Beyond these aspects, medically unnecessary CS can lead to avoidable perinatal complications and morbidities, for example, it was found that this CS type increase rates of neonatal respiratory distress, especially before 39 weeks of gestation.5 However, the emergency CS has seemed to have more complications than elective CS for both mother and newborn such as infection, fever, urinary tract infection and disseminated intravascular coagulation.6

CS rates keep on increasing in all regions around the world and would lead to additional challenges for poor health systems such as in sub-Saharan Africa (SSA) and Southern Asia regions.7 If performed at adequate rates, CS can prevent cases of maternal and neonatal mortality.8 However, the optimal CS rate for reducing maternal and neonatal mortality remains difficult to estimate, unclear and constitutes a debate in the existing literature. Currently, WHO does not have a recommended optimal CS rate at the country level. However, their studies suggest that a rise in CS rate towards 10% may lead to a reduction in neonatal and maternal mortality, but above 10%, there is no evidence of the relationship between CS rate and neonatal and maternal mortality.9 Molina et al8 estimated the optimal CS rate to be approximately around 19%, while Ye et al10 found 10%.

There is a broad literature on CS, Boerma et al11 provided an overview of the global epidemiology of the CS. The review found that the SSA region with an average rate of 5% (95% CI 3.5–6.6), has the lowest rate among all the other regions.12 However, there are generally disparities and variations in CS rates, with countries such as Rwanda with a very high rate (64.2%) and Sao Tome with a very low rate (0%).13 In addition to inter-countries disparities, recent studies have investigated the associated socioeconomic factors of CS access in SSA. Islam et al14 found that wealth index, place of residence, maternal and paternal education, were associated with CS delivery in SSA. Yaya et al13 performed a finer analysis by including sociodemographic, socioeconomic, sociocultural and some clinical variables in the investigation of the determinants of CS in SSA, and it was found that CS was associated with socioeconomic factors. Chu et al15 investigated clinical variables and indications of CS in SSA countries, and found that the main indication of CS in SSA countries include: obstructed labour, poor presentation, previous CS, fetal distress, uterine rupture and antepartum haemorrhage, which led to a better understanding of the clinical practice of CS in the SSA region.

Almost all these studies did not differentiate the two types of CS, which is important to investigate. Issues have been raised regarding the rising cases of medically unnecessary CS cases globally including in the SSA region.3 16 17 Betran et al7 highlighted the challenges in controlling CS use in SSA countries in the CS steadily increasing dynamic, which would be inevitability challenging for health systems. Together, with the global concern of medically unnecessary CS among elective cases, emergency CS cases are also a concern, especially in settings of poor health systems such as in SSA countries as most health facilities in these settings have shortcomings in emergency obstetric and newborn care.18 However, there is limited knowledge about the different types of CS in SSA countries, including the prevalence of each type, the profiles of women who undergo the different types of CS, and the reasons for the use of the type of CS.

This study aimed at estimating the prevalence rates of emergency and elective CS for SSA countries individually and globally, then, identifying socioeconomic factors associated with each type of CS in SSA countries, and finally, evaluating the association between both types of CS with early neonatal mortality.

MethodsStudy design and setting

This was a multi-countries analysis based on Demographic and Health Surveys (DHSs) data, which used cross-sectional designs. All 46 SSA countries were eligible to be included in the study. However, countries that had not captured the CS timing in their recent DHS were excluded (a total of 25 countries). Thus, a total of 21 countries were included in the study for analysis. All SSA subregions were represented (Western, Central, Southern and Eastern). Table 1 provides a list of counties included in the study and the respective surveys, grouped by subregions according to the United Nations geographical grouping.19

Table 1

Included countries with the survey years grouped by region

Data sources, sampling procedure and participants

Data were freely downloaded from the DHS programme website (https://dhsprogram.com/) after an approved request. DHS uses standardised data collection tools and sampling procedures among all countries where they are carried out. They are stratified two-stage random sampling. In the first stage, the enumeration areas (EAs) from the general census of the population are sampled proportionally to the number of households in clusters after stratification in rural and urban EA, respectively. Then, at the second stage, the households are sampled within the sampled EA in the first stage systematically and with equal probability, more details regarding the sampling procedures of DHS could be found in.20 DHS targets various population which include women of reproductive age (15–49 years) with their live births during the last 5 years preceding the survey, men and women aged 15–64 years old, adolescents of 15–19 years old and children. Our study population was children born in the last 5 years before the survey. For each included country, we downloaded the ‘Children’s Data’, which contains children of women born in the last 5 years (0–59 months) before the survey. In the analysis, only institutional delivery was considered, these include children born in a health institution.

Study variables

The outcomes of this study were the prevalence of elective and emergency CS and early neonatal mortality. CS type was captured using variable m17 (binary variable defined by delivery by CS or not) and m17a (a variable that assesses the timing on decision for CS: before or after the labour started). The timing of CS was used to define CS type. CS carried out before the start of labour was classified as elective CS, while the cases carried out after the start of labour were classified as emergency CS. Such categorisation has been previously adopted by Muhammad et al21 in a recent study investigating CS type in Bangladesh. Regarding the early neonatal mortality outcome, it was defined as a death event within the first 7 days of life. Variable b6 was used to classify early neonatal mortality as a binary variable by grouping on one hand, children that died between 0 and 7 days after birth, and another hand, alive children or those dead after 7 days of life. In the investigation of associated factors of both types of CS, the main independent variable was the wealth quintile index (richest, richer, average, poorer and poorest) of children’s households, which is a variable constructed as a composite variable using household amenities (such as radio station, television and air conditioner). Other independent variables included the mother’s education level at the time of child birth (none, primary, secondary and plus), the place of residence (urban vs rural), mother age at the birth (15–19, 20–24, 25–29, 30–34, 35–39, 40–44 and 45–49), baby size at the birth (very large, larger than average, average, small than average and very small), birth order (first, second, third and plus), child gender (male or female), a baby is a twin (yes or not), the number of antenatal care (ANC) (none, less than four, four and plus and do not know).

Regarding the investigation of the association between both types of CS with early neonatal mortality, the birth delivery mode (vaginal, emergency CS and elective CS) and the postnatal care received (postnatal check within 2 days after birth by a skilled provider) were the main independent variables. The confounders were all the independent variables considered in the analysis of associated factors of both types of CS, that were included in the analysis after collinearity checking.

Patient and public involvement statement

None.

Statistical analysesCountry and global estimates

The elective and emergency CS prevalence rates with their 95% CIs were estimated at the country levels and for the overall SSA region. At the country level, the prevalence rates were estimated as weighted proportions using the household sampling weights provided by the DHS. For the regional SSA estimates, first, the weights of each country were evaluated by dividing the total live births in the SSA region by the total of live births in each respective country.22 Then, the SSA regional estimates of each CS type were estimated as a weighted average of countries’ prevalence rates, using the calculated country’s weights.

Multilevel analysis

Three levels were considered in the modelling, the first level was the country, the second was the cluster (EA) and the third was children nested within clusters. The country level weights have been calculated as described in the previous section. Within countries, cluster and individual level-weights were estimated based on the DHS methods for approximating level-weights,23 which provides steps and guidance for level-weights estimated for DHS data.

Socioeconomic factors identification of elective and emergency CS

Weighted multinomial multilevel logistic (MML) regression was adopted to estimate and identify socioeconomic factors of the emergency CS and elective CS while accounting for the heterogeneity between countries. The statistical definition of these models and their properties could be obtained in Hartzel et al24 and Grilli and Rampichini.25 The function gsem within the Stata software was used to estimate the MML models in the frame of this work. In this model, the response variable is the delivery mode (emergency CS, elective CS or vaginal delivery), with vaginal delivery considered as the base outcome. The relative risk ratio (RRR) with their 95% CIs were estimated to evaluate the magnitude of associations.

Initially, we estimated the non-adjusted RRRs for each study covariate, then, covariates with significant non-adjusted RRR were selected for collinearities checking. Finally, the selected independent variables, with wealth quintile index as primary covariates were considered together in the multivariate MML to estimate the adjusted RRRs (aRRRs).

Association of elective and emergency CS with early neonatal mortality

To evaluate the association of elective and emergency CS with early neonatal mortality, we have considered early neonatal mortality (defined as newborn death occurring during the first 7 days after birth) as the dependent variable. The delivery mode was considered the main independent variable with a vaginal delivery as the base. Then, a multilevel logistic regression was used to assess the magnitude of the effect of emergency and elective CS on early mortality by evaluating unadjusted and adjusted ORs, with their 95% CI.

Univariable selection method was adopted to identify and select variables for the multivariate analysis. Initially, the crude effect of emergency and elective CS on early mortality by estimating unadjusted OR. Then, the adjusted effect was estimated through multivariate logistic regression adjusted for the selected confounders.

Results

The 21 SSA countries included in the study accounted for a total of 238 281 live births. However, when considering the institutional birth, one remains with 155 172 (65.12% institutional birth delivery) live births included for the analysis. The characteristics of the pooled sample are given in online supplemental material table.

Emergency and elective CS rates estimates

Table 2 provides emergency versus elective CS rate estimates with their 95% CIs by country. Overall, the emergency CS rates in SSA countries were estimated at 4.6% (95% CI 4.4–4.7) and were higher than the elective CS rate estimate (3.4%, 95% CI 3.3–3.6). Only South Africa and Mauritania had observed higher elective CS than emergency CS, estimated at 3.1% (95% CI 2.6–3.6) versus 4.6% (95% CI 4.1–5.3) for Mauritania and 8.6 (95% CI 7.5–9.9) versus 16.6% (95% CI 15.0–18.3) for South Africa. All the other included countries had emergency CS rates greater or equal to the elective CS rate. By disaggregating according to the place of delivery, different patterns were observed as summarised in table 3.

Table 2

Emergency and elective caesarean section (CS) rate estimates with their 95% CI estimated by country

Table 3

Emergency versus elective caesarean section (CS) rate estimates with their 95% CI estimated by country and by place of delivery

Overall, in the SSA region, the global CS rate in the private sector was estimated at 17.9%, with the elective CS rate estimated at 10.2% (95% CI 9.3–11.2) larger than the emergency CS rate estimated at 7.7% (95% CI 7.0–8.5). Generally, for all countries, except Malawi, and Tanzania, the elective CS rates were higher than emergency CS rates in the private sector. However, Mauritania, Rwanda and South Africa had extremely high elective CS rates (25.0%, 38.9% and 52.9% respectively) in the private sector. Countries such as Benin, Burundi, Cameroon, Gambia, Guinea, Liberia, Malawi and Mali had low CS rates in the private sector, with elective and emergency CS rates together not exceeding 10%.

In the public health sector, the CS rate in the SSA region was estimated at 6.7%, with the emergency CS rate (4.0%, 95% CI 3.8–4.2) higher than the elective CS rate (2.7%, 95% CI 2.6–2.8). In all included countries, the emergency CS rates were higher than the elective CS rate in public health hospitals, except in South Africa (13.1%, 95% CI 11.6–14.6 vs 8.7%, 95% CI 7.5–10). The prevalence of both CS rates was generally low in public health hospitals (as they were altogether less than 10%) for all included countries, except for South Africa and Mauritania.

Socioeconomic factors of the elective and emergency CS

Table 4 presents the results of the multilevel multinomial logistic regression for the identification of the associated socioeconomic factors of elective and emergency CS. The focus was on socioeconomic factors such as wealth index, and the mother’s educational level, while adjusting for clinical variables such as baby size at birth, preceding birth interval, baby gender, is baby a twin. Overall, women from the richest households had a significantly higher likelihood of both elective and emergency CS than those from other wealth quintiles, while no differences were found across other wealth quintiles. Women from the richest households had up to 3.58 (95% CI 1.09–11.7) and 1.93 (95% CI 1.16–3.22) RRR of delivering by elective and emergency CS respectively than other women.

Table 4

Associated factors of elective CS and emergency CS versus normal delivery, respectively

Maternal educational level was associated with elective CS, women with secondary and plus educational levels were more likely to deliver by elective CS than vaginal delivery (aRRR=2.13, 95% CI 1.40–3.25). However, the mother’s educational level was not associated with emergency CS. Children born in rural areas had less chance to be delivered by elective CS, with an aRRR of 0.73 (95% CI 0.55–0.96) compared with those in urban areas. In private health facilities, children had up to 2.07 (95% CI 1.60–2.67) the RRR delivered by elective CS than those being birth delivered in public health facilities.

Association of early neonatal mortality and elective/emergency CS

Table 5 displays the unadjusted and adjusted effects of emergency and elective CS on early neonatal mortality. Generally, elective CS was not associated with early neonatal mortality. While, emergency CS was positively associated with early neonatal mortality, with newborns born by emergency CS having an OR of 2.37 (95% CI 1.64–3.41) of early neonatal mortality than those born by standard vaginal delivery. Children who had poor postnatal care quality (postnatal check carried out above 2 days after delivery and/or by a non-professional health worker) had an OR of 1.91 (1.07–3.42) of early neonatal mortality compared with those who had a good quality of postnatal care.

Table 5

Association of early neonatal mortality with delivery mode in sub-Saharan Africa countries

Discussion

It was found that the emergency CS was significantly higher than the elective CS in the SSA region. This finding may indicate, on the one hand, poor quality of ANC in this region which in some cases could prevent the detection of planned or elective CS. On the other hand, one could point to the insufficient coverage of ANC or inadequate ANC visits to allow for better pregnancy monitoring, increasing the chance of emergency CS. A recent study by Amponsah-Tabi et al in Ghana found the quality of ANC was significantly associated with pregnancy outcomes which included anaemia during pregnancy, pre-eclampsia with severe features or delivering a low birth weight baby.26

However, Mauritania and South Africa had lower emergency CS rates compared with elective CS, which could indicate better quality in antenatal and postnatal care. Since 1998 Mauritania has launched an obstetrical plan to improve access to quality maternal care. That consists of voluntary insurance paid by women at their initial ANC visit which cost between €15 and €17 per woman for prenatal and postnatal visits, laboratory tests, ultrasound, assisted delivery, transportation and coverage of any complications.27 This plan has shown a significant impact in improving the overall maternal health outcome in the country.28 29 Thus, with better quality in ANC, medical staff would anticipate birth and delivery complications that could turn into emergency CS for planned or elective CS. However, the overall CS rate in this country remains lower than the optimal minimum threshold, implying that there would remain challenges in CS access in this country. Concerning the South African case, the very high rate of elective CS, particularly for deliveries carried out in private healthcare would reveal many cases of medically unnecessary CS carried out in this setting.30 However, there is no current knowledge on the explanation of the overall high rate of CS in South Africa and this has been argued by Solanki et al.31

One of the most interesting findings of this study was that the pattern of the type of CS varied according to the type of healthcare facility. Private healthcare facilities carried out more elective CS than emergency CS (except Malawi and Tanzania), and the converse was found in the public health facilities (except in South Africa). These results could be explained by a possible overuse of CS in private healthcare facilities for profitability purposes as suggested by Solanki et al.31 Private health facilities are generally profit-oriented organisations, and in some cases due to financial rentability pressure, some medical acts could be indirectly influenced by hidden financial gain purposes. Currently, there are no comprehensive data in most SSA countries to understand the difference in patterns of elective versus emergency CS in private and public healthcare facilities. Investigating the indications of CS in private and public healthcare facilities separately could be an interesting topic for better-informing health systems policies.

The wealth index was found to be associated with CS delivery, as the richest women had more likelihood of both elective and emergency CS delivery. This is consistent with findings obtained by Islam et al and Ronsmans et al32 for the overall CS in SSA. It is not surprising that women from the richest households had more likelihood of having elective CS than others. These women could be those accessing better health facilities which are generally expensive in SSA countries, but provide better services, with adequate staffing including well-trained and skilled birth attendants and better care facilities. Also, due to tokophobia, and fear of childbirth by normal vaginal delivery, some women may deliberately choose to give birth by CS.33 In this case, on one hand, delivery CS cost could be a preventing factor for poor or less wealthy women to deliberately choose CS as their chosen mode of delivery, on the other hand, these costs would not be a limiting factor for wealthy women to deliberately choose CS. However, the finding that the richest women had more chances of emergency CS may be surprising and worrying. This finding may suggest that less wealthy women and/or their children would have died at the time of delivery due to non-access to this type of CS, as, being carried out in emergencies, the CS is an ultimate act of life-saving for both mother and child.

Regarding the association of emergency CS with early neonatal mortality, this is an innovative finding which allows situating where is concerned in CS in SSA countries. It was already known from a study by Althabe et al34 that an increased rate of CS in low-income and middle-income countries corresponds to an increase in maternal and neonatal mortality. However, the type of CS in concern was not investigated. The data from this work suggested the vulnerability of newborns after emergency CS in the African region. This may reflect structural deficiencies that allow antenatal detection of conditions eligible for preterm fetal extraction such as low birth weight observed in our study and elsewhere.35 Such results, if confirmed, should lead to recommendations aimed at increasing the rate of elective CSs, particularly in cases of intrauterine growth retardation or similar situations. Another possible explanation of this positive link between emergency CS and early neonatal mortality in SSA could be due to poor labour care, poor surgical care at the time of CS, and poor neonatal care post-CS that may be observed in most SSA countries’ health facilities. Added to these, the skills level of staff members that attend birth delivery and postnatal care check quality could be pointed out as our results demonstrated.

Not having considered the cases of women who have elected themselves for a planned CS with non-medical indication could be viewed as a limitation of this study. Considering this category level would have allowed a better understanding of the high rates of elective CS in countries such as South Africa. Also, the association of elective CS with the wealth quintile index, as well as the association of delivery mode with early neonatal mortality would have been well captured by considering that extra factor level of delivery mode. Therefore, future DHSs could consider adding that factor level among the CS carried out before labour has started. The classification of the type of CS was built from a question asked to women, even though simple (was the CS carried out before or after labour started?), would have brought misclassification in some cases. This misclassification bias could be viewed as well as a limitation, however, this could be marginal and could have not that much affected our results, since we assumed there would have been a tiny number of such cases. Finally, we believe that the set of the variable included in the analysis of this study would have not fully captured the socioeconomic differences between women, and this could be viewed as a limitation of this work. However, performing multilevel analysis as we did could have mitigated this limitation, as such analysis account for the socioeconomic variability at the neighbourhood level.

Conclusions

Overall, in SSA countries, emergency CS is more carried out than elective CS, except in South Africa and Mauritania. This result would suggest for public health policies that more attention should be put on improvement of ANC quality for better monitoring of pregnancies and therefore better delivery preparation in the countries of this region. It has been found that the pattern of the type of CS varied according to the type of place of delivery. Private healthcare facilities carry out more elective CS than emergency CS, and conversely in public health facilities. These suggest that procedures, practices, and recommendations regarding antenatal and delivery care are drastically different from private to public healthcare facilities in SSA countries. Public health authorities need to work on harmonising CS indications in public and private healthcare facilities. Finally, the positive association of emergency CS with early neonatal mortality suggests a need for the improvement of emergency obstetric and newborn care units, especially labour care, surgery care and postnatal care for adequate management emergency CS including post-emergency CS.

Data availability statement

Data may be obtained from a third party and are not publicly available.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Not applicable.

Acknowledgments

The authors acknowledge DHS programme for data request approval for this study.