Caesarean section rate in Nigeria between 2013 and 2018 by obstetric risk and socio‐economic status

Caesarean section (CS) can be life‐saving for both mother and child, but in Nigeria the CS rate remains low, at 2.7% of births. We aimed to estimate the rate of CS and early neonatal mortality in Nigeria according to obstetric risk and socio‐economic background and to identify factors associated with CS.


Introduction
Caesarean sections (CS) can be life-saving for mother and child during childbirth. However, the procedure is associated with complications [1] and should be performed based on medical indications [2]. Large variations in CS rates exist globally, ranging from 0.6% (South Sudan), indicating lack of access to this live-saving intervention, to 58.1% (Dominican Republic) where overuse is likely [3]. The smallest increase in CS rate during the last decades (1990-2014) was found in sub-Saharan Africa (from 2.3 to 3.5%) [4].
The optimal CS rate to save lives is debated but levels between 9% and 16% have been suggested in previous studies [5,6]. To enhance the monitoring of CS use, the application of the Robson classification is recommended by WHO [7,8]. This classification groups births into 10 totally inclusive and mutually exclusive groups based on a few commonly documented obstetric parameters (e.g. single/multiple pregnancy, term/preterm) for easy comparisons [8]. Few studies using this approach have been published from low-income countries (LIC) [9,10].
According to the latest Nigeria DHS, the maternal mortality ratio is still high at 512 per 100 000 live births (95% confidence interval (CI): 447-578) [11]. Further, the prevalence of stillbirths is estimated at 39.6/1000 births, and 32% of those stillbirths could be associated with asphyxia [12]. A study showed that infants born by CS in Nigeria were roughly three times more likely to die than those born vaginally [13], indicating that the intervention might be used too late. Other factors associated with increased risk of adverse neonatal outcome during labour in Nigeria include referral status, parity, gestational age [12], male gender and rural residence [13].
The percentage of facility-based births estimated by the 2018 DHS was 39%; 26% of births took place in the public sector and 13% in a private health facility. The percentage of women reporting receiving any ANC was 67% and women receiving four or more visits was 57% [11]. Further, access to Emergency Obstetric Care (EmOC) in Nigeria is insufficient [14] and financial and geographical access barriers prevail [15,16]. Cultural beliefs and negative attitudes towards CS are other possible explanations for a delay in care-seeking behaviour and low CS rate [17,18]. As a result, the population CS rate remains low at 2.7% [11] and between 11 and 49% in facilities [19][20][21][22], with positive associations between likelihood of CS and urban residence [23], wealth and higher education level [24,25]. Extremely low CS rates of 0.4% are found among the poorest women [24].
While the evidence of the effect of socio-economic status and place of residence on CS rates in Nigeria is well established, little is known about CS rates in relation to obstetric needs. A few facility-based studies have shown increased odds of having a CS for women with overweight, previous CS [21] and obstructed labour [22]. Only one population-based study has investigated the association between obstetric factors and CS in Nigeria using the 2013 Demographic and Health Survey (DHS) data, showing a positive association between CS and maternal age, large birth size, multiple pregnancy, primiparity, health insurance coverage, antenatal care (ANC) and maternal obesity [26].
To improve the evidence on the intersection of socioeconomic inequalities and obstetric risk factors in Nigeria, we aimed to estimate the rate of CS and early neonatal mortality rate (ENMR) by applying an innovative approach with newly collected DHS variables, which allowed the concurrent analysis of obstetric risk (based on the Robson classification) and socio-economic factors.

Methods
We used the 2018 Nigeria DHS to obtain information on obstetric factors, socio-economic background, CS rate and ENMR [27]. DHS are cross-sectional, nationally representative household surveys, typically conducted in 5year intervals. Data on maternal health and care-seeking are based on self-reports of women in reproductive age (15-49 years) within sampled households. Standardised questionnaires are used for data collection by trained enumerators. The surveys were conducted in Hausa, Yoruba, Igbo and English, which are the 4 main languages in Nigeria, and the questionnaires were pre-tested for comprehensibility.
The Nigeria DHS used two-stage sampling; after stratification of rural/urban setting, 1400 clusters were selected from the enumeration areas, and from each cluster, 30 households were randomly chosen. All women of childbearing age were interviewed in each sampled household, yielding a total size of 41 821 women, corresponding to a response rate of 99% [11]. All live births to these women within five years prior to the survey were considered in this study (Figure 1).

Outcomes and explanatory variables
We adapted the Conceptual framework for factors affecting Infant Mortality, to guide our analysis [28,29] (Figure S1). Our main outcomes were self-reported CS and early neonatal mortality (death at age <7 days) for all live births. Women who gave birth at a healthcare facility were asked 'Was (name) delivered by caesarean section, that is, did they cut your belly open to take the baby out?'. The reason for CS was clarified by 'What was the reason for taking the decision to have the CS?' [11].
We categorised births into 'modified Robson groups' using parity (nulliparous; 0 previous live birth before index pregnancy, or multiparous;>=1 previous live birth before index pregnancy), singleton/multiple pregnancy (twins or triplets categorised as multiple), full-term or preterm birth using birthweight as a proxy, breech position and whether the CS was planned or emergency. Usually, information regarding spontaneous onset of labour vs. induction is considered when assessing Robson group, as is previous CS. However, this information was not collected by the DHS. The original and modified Robson groups are presented in Table 1. Data about whether the CS was planned were used to create Robson groups 2 and 4. Information about breech position (n = 150) was only available in case of CS, since it was a follow-up question about the reason why CS was performed. The two most common Robson groups 1 and 3 are considered low-risk groups with rates typically below 10 %, while CS rates among groups 5 and 8 are usually fundamentally higher with rates up to 60-70% and for group 10 around 33-37% [30].
Of all live births, 98.95% were reported as full-term (born in pregnancy month 9 or 10) based on the contraceptive calendar data. Such high proportion of full-term births is improbable compared to global estimates of preterm births [31]. Therefore, the baby's birthweight was used as a proxy measurement when creating the modified Robson groups [32,33]. The variable 'birthweight' was generated by combing two variables, since few of the infants had their actual weight recorded at birth. For babies with a recorded birthweight (22.8%), we categorised those weighing <2500 g as preterm/low birthweight. Babies with a missing birthweight were categorised based on mother's recall of baby's size (very small, small, average, larger than average). 'Very small' and 'small' were considered preterm/low birthweight, while 'average' or 'larger than average' were categorised as normal birthweight (full-term).

Missing data
There were 514 cases (1.5%) with missing birthweight (neither recorded birthweight nor mother's report of birth size). These observations were excluded from the modified Robson analyses. For the other determinants, there were no cases with missing data.

Statistical analyses
We estimated the percentage (95% CI) of live births by CS and ENMR (per 1000 live births). Further, the relative size of each modified Robson group and the corresponding CS rate and ENMR were calculated. Thereafter, the CS rates were estimated for each modified Robson group, stratified by socio-economic background (education, wealth, residence, place of birth). When comparing CS rates between different socio-economic factors within the Robson groups, only Robson 1, 3, 8 and 10 were kept, since groups 2, 4 and 6 all had a CS rate of 100%. Logistic regression analyses were used to explore determinants for CS. First, each independent variable was examined in bivariate analysis and variables significantly associated (P < 0.05) with CS were retained in multivariable analysis. When assessing the determinants for CS, only the woman's most recent live birth in the recall period was included. The rationale for this was to reduce the risk of recall bias [26] and because some variables were only available for most recent birth as well as to align with previous studies on CS [4,37]. Referral status during childbirth was assessed in a multivariable subanalysis, restricting the population to most recent births at a healthcare facility. The svyset command in Stata was used to account for sampling design (sample weights, clustering and stratification). All analyses were performed in Stata version 16.

Ethical considerations
The DHS surveys were approved by ICF Institutional Review Board (IRB) and by national IRB. Informed consent was obtained by each participant [38]. This secondary data analysis of anonymised data was exempt from ethical review at Karolinska Institutet.

Characteristics of the study population
We included 33 924 live births (referred to as births) in the five-year period before the survey to describe CS rates and ENMR according to socio-economic background and modified Robson groups. When exploring determinants for CS, only the most recent live birth was included (n = 21 792) ( Figure 1). Referral status during childbirth was assessed in a sub-sample of facility births (n = 8909).
Descriptive characteristics of the population (Table 2) show that 46.4% of the babies were born to mothers with no formal education, with a majority (61.5%) living in a rural setting ( Table 2). Almost 40% of the births were among women who had given birth to four or more children before their index pregnancy. Most births occurred at home (60.6%), while around 13% of the births occurred in private health facilities. The overall CS rate was 2.7% (95% CI: 2.4-3.1) and varied across regions with the lowest rate in North-West (0.7%, 95% CI: 0.5-0.9) and highest in South-West (7.0%, 95% CI: 5.7-8.7). 1 Nulliparous, singleton, registered birthweight >2500 g or estimated weight by mother as 'average size' or 'larger than average size' 2 Nulliparous, singleton, cephalic, >37 weeks' gestational age, induced labour or CS before labour 2 Nulliparous, singleton, registered birthweight >2500 g or estimated weight by mother as 'average size' or 'larger than average size'. Planned CS. 3 Multiparous (excluding previous caesarean section), singleton, cephalic, >37 weeks' gestation, in spontaneous labour 3 Multiparous, singleton, registered birthweight >2500g or estimated weight by mother as 'average size' or 'larger than average size' 4 Multiparous without a previous uterine scar, with singleton, cephalic pregnancy, >37 weeks' of gestation, induced or CS before labour 4 Multiparous, singleton, registered birthweight >2500 g or estimated weight by mother as 'average size' or 'larger than average size'. Planned CS. 5 Previous caesarean section, singleton, cephalic, >37 weeks' gestation -Information not available 6 All nulliparous with a single breech 6 Nulliparous and multiparous with single breech position 7 All multiparous with a single breech (including CS) -Combined in group 6 8 All multiple pregnancies (including previous CS) 8 All multiple pregnancies 9 All women with a single pregnancy in transverse or oblique lie (including those with previous CS ) -Information not available 10 All singleton, cephalic, <37 weeks' gestation pregnancies (including previous caesarean section) 10 All babies categorised as small in size or with a birthweight <2500 g CS, caesarean section.

Caesarean section rates in each modified Robson group by socio-economic factors
We observed large disparities in CS rate by household wealth quintile ( Figure 2). In modified Robson 8 (multiple pregnancy), the rate ranged from 0.8% (95% CI: 0.2-3.2) in the poorest quintile to 19.7% (95% CI: 12.5-29.7) in the richest. A similar pattern was seen for maternal education; for example, in modified Robson 1, women with no education had a rate of 0.2% (95% CI: 0.0-0.6) compared to 11.2% (95% CI: 7.9-15.7) among women with the highest education level. In modified Robson 8, the CS rate was 30.7% (95% CI: 21.3-42.2) among births in private health facilities and 9.2% (95% CI: 5.5-14.9) in public hospitals.

Determinants associated with caesarean section
When assessing the determinants for CS, the sample was restricted to the most recent live birth including 21 792 women-baby pairs (Table 4). Women with the highest level of education had an adjusted odds ratio (aOR) of a CS birth of 2.47 (95% CI: 1.49-4.08) compared to women with no formal education. The three richer wealth quintiles were positively associated with CS; the richest quintile compared to the poorest had an aOR of 4.66 (95% CI: 2.28-9.51). Women identifying as Christian/Other religion had 1.5 times higher adjusted odds of CS than Muslim women. Number of ANC visits, especially if more frequent (8+), was strongly associated with CS (aOR 5.71, 95% CI: 3.14-10.39).
Among maternal factors, age was a predictor of CS with 4.61 higher adjusted odds (95% CI: 2.51-8.46) for CS for women age >35 years vs. those <20. Unwanted pregnancy also increased the likelihood for CS, whereas marital status was not associated with CS. Women with multiple pregnancy had 2.55 (95% CI: 1.57-4.15) higher adjusted odds of a CS than those with singletons.
In a sub-analysis, referral during childbirth and place of birth were added to the multivariable analyses. Referral increased the odds for having a CS (aOR 11.97, 95% CI: 7.98-17.94) as did giving birth in a private health clinic (aOR 6.98, 95% CI: 4.83-10.08) compared to public primary options. Many of the other predictors, that is wealth, education and ANC visits lost statistical significance when place of childbirth and referral status were added.

Discussion
To improve the evidence of inequity of access to CS in case of obstetric complications in Nigeria, we applied a modified Robson classification to DHS data, while also integrating socio-economic aspects. Thus, our analysis sheds light on social disparities in health by intersecting social inequalities and obstetric risk factors. We indicated suboptimal CS rates within all the modified Robson groups, ranging from 1.0% (modified Robson 3) to 7.1% (modified Robson 8). In the two high-risk groups multiple pregnancy and preterm/low birthweight, the rates were worryingly low and with corresponding high ENMRs. Higher number of ANC visits, multiple pregnancy, wealth, education, religious affiliation and increasing maternal age were positively associated with CS. Finally, referral during childbirth was strongly associated with CS and the referred group had a high ENMR.
Globally, CS births are increasing in all regions, but levels remain low in sub-Saharan Africa [3]. In this study, the underuse of CS was especially apparent for the modified Robson 8 (multiple pregnancy), with a national average rate of 7.1%. Higher rates for multiple pregnancies were found among women who had the highest education (21.7%), greater wealth (19.7%) and with childbirth in private facilities (30.7%). A study of eight countries in Southern and Eastern Africa showed rates between 9 and 36% [39] while in high-income countries CS rates are often above 60% for multiple pregnancies [30]. Further, multiple pregnancies had an unacceptably high ENMR (175/1000 live births) in this study. Our results are in line with a previous study from Nigeria indicating 6 and 2.54 times higher odds of a low Apgar score and of neonatal death, respectively, when compared to singleton pregnancies [40].
The other obstetric risk group, modified Robson 10 (preterm/low birthweight), was the second largest group (13.6% of all births). The CS rates, with an average of 1.8% and as low as 0.2% among the poorest women, are far beneath previous, yet facility-based, findings for this Robson group with rates around 33-37% [30,41]. Further, as shown by Ezeh at al., a large number of early neonatal deaths can be attributed to low birthweight in Nigeria [42]. Thus, this is an obstetric group with a particular need of extra precaution during childbirth.   There was no apparent overuse of CS at a population level in Nigeria in this study. However, the procedure is unevenly distributed. Large disparities were seen within the wealth quintiles, with the poorest women having a CS rate (0.3%) far below the suggested minimum of 9% to maintain safety for women and child [6]. This uneven distribution of CS across wealth groups has been established before in Nigeria [24] and other LIC [37] and these inequalities highlight the need for removing financial barriers to CS.
Lastly, the modified Robson 4 (multipara, planned CS) was small (0.6% of the obstetric groups) but contributed to a large proportion of all CS (23.2%). Even though the indications for the planned CS were unknown, studies have shown that planned CS is largely driven by previous CS [20,21], thus a large part may belong to the Robson group 5, which we were unable to construct. Repeated CS may become an increasingly important factor to consider in Nigeria, despite the low total rates, not least because of the high fertility rate.
Referral during labour substantially increased the odds for CS. Referral could be a proxy for complications such as obstructed labour and antenatal haemorrhage, explaining the high CS rate in this group (50.5%). The very high ENMR in this group can be interpreted as the interventions being done too late to save the baby's life, as others propose [13]. A timely CS performed with good quality of care can be life-saving for the neonate by reducing the risk of asphyxia during labour. However, while the general recommendation that a CS rate of 9-16% to save neonatal lives apply on a population level [5], these rates are not sufficient for risk pregnancies such a preterm/low birthweight or multiple pregnancies due to higher risk of complications such as bleeding, asphyxia and cord prolapse. This is clearly shown in our results where Robson groups 8 and 10 had unacceptable high ENMR, thus stresses the need of more attention during labour and readiness for CS for these high-risk groups in particular.
After adding referral and place of childbirth to the analysis, only unwantedness of pregnancy, older maternal age and multiple pregnancy remained positively associated with CS. This could be explained by the reduction in population size (from 21 792 to 8909 births) or that some of the predictors, that is number of ANC visits [43] and education, being closely interlinked with giving birth at a health facility, affecting the statistical associations [44].

Implications and future research
Our study indicates that inequitable access to CS is exacerbated in relation to obstetric risk factors. Still, our study also confirms the known socio-economic and financial barriers to access. We note that health insurance coverage in this population was low at 3% [11]. Improving universal health coverage, thus, remains essential. Removing user feesas done in Benin and Maliremains a potentially effective measure to improve accessibility of CS [45], albeit transport cost may continue to be a major barrier. This was confirmed by a study investigating the user fee exemption in some regions in Nigeria, showing lower CS rates for women with low income despite free maternal health care [46]. Thus, increased availability of EMOC in rural areas is also needed.
The inequity identified in relation to obstetric risks demands new thinking and programming. Better risk detection during ANC through, for example, ultrasound to detect multiple pregnancies and for accurately estimating gestational age, intensified counselling of danger signs and early effective referral could address these inequities. Finally, ANC could also provide a platform for encouraging childbirth in a healthcare facility, especially in case of multiple pregnancy. We believe that the opportunity of reducing socio-economic barriers through ANC by focusing on those most in need is given too little attention [47].
Moreover, we believe that our approach to apply the Robson classification to population-level data provided important new knowledge. While there are limitations, these modified Robson groups could be useful for continuous monitoring of CS rates stratified by regions, facilities or socio-economic groups to evaluate equity in access. Further research comparing trends over time in relation to the Robson groups would be useful for examining the drivers of underuse and possible, future overuse of CS in Nigeria and other LIC settings. It would be useful to collect some basic obstetric factors in the DHS, for example previous history of CS, to improve the evaluation of CS rates on a population-based level. However, more medical aspects like comorbidities and complications have limited validity in such surveys, particularly specificity [48].

Strengths and limitations
The DHS includes a large sample size and has a high participation rate (99%). The questionnaires are well tested and with a low number of missing data. The DHS is population-based and includes all states in the sampling processes, which makes the results generalisable to the whole country. This study is the first of its kind to assess CS and ENMR in Nigeria within modified Robson groups, stratified by socio-economic factors.
The data were collected with a five-year recall period; thus, recall bias cannot be ruled out. However, some questions were restricted to only to the most recent birth, for example regarding ANC. Moreover, for some of the data, the accuracy might be substandard; that is, birthweight where the perceived birth size reported by the mother was used for the majority of the cases.
We missed some variables to categorise births by the Robson classification, such as previous CS and induction of labour. However, the majority of women with a previous CS are most likely found in modified Robson 4, multipara with planned CS, as discussed above. Further, information about breech position was obtained by asking about the reason for CS and it is possible that the women were not correctly informed by the provider or did not remember. Moreover, data regarding breech position among vaginal births were not available; thus, proportional CS rate of adverse foetal position could not be calculated. In this study, the relative group size with breech position was 0.4%; ten-fold lower than the expected rate around 3-4% [30], indicating that most of the children with breech position were probably delivered vaginally. Overall, self-reported data on indication for CS might not be as accurate as hospital-derived data, potentially resulting in over-and/or underreporting of some obstetric factors. Additionally, this DHS did not capture mode of birth for stillbirths and since stillbirths can be due to intrapartum complications, some cases of perinatal mortality are not captured [12].

Conclusion
The severe underuse of CS, especially for multiple and preterm births among socio-economically disadvantaged women, emphasises insufficient access to health care in Nigeria. Our analysis highlights the importance of disaggregating data to indicate underserved groups, in order to plan strategies targeted at-risk pregnancies to improve neonatal outcome. We suggest that women with multiple pregnancy or with a high risk of preterm/low birthweight would benefit from specific attention already during ANC. The referral system also needs to be strengthened and striving towards UHC is important to assure access to CS for the most underprivileged women.