Bulletin of the World Health Organization

Trends in caesarean delivery by country and wealth quintile: cross-sectional surveys in southern Asia and sub-Saharan Africa

Francesca L Cavallaro a, Jenny A Cresswell a, Giovanny VA França b, Cesar G Victora b, Aluísio JD Barros b & Carine Ronsmans a

a. London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, England.
b. Postgraduate Program in Epidemiology, Federal University of Pelotas, Pelotas, Brazil.

Correspondence to: Francesca L Cavallaro (e-mail: francesca.cavallaro@lshtm.ac.uk).

(Submitted: 24 January 2013 – Revised version received: 30 May 2013 – Accepted: 20 June 2013 – Published online: 09 August 2013.)

Bulletin of the World Health Organization 2013;91:914-922D. doi: http://dx.doi.org/10.2471/BLT.13.117598

Introduction

Caesarean sections, when adequately indicated, can prevent poor obstetric outcomes and be life-saving procedures for both the mother and the fetus.1 However, at a time when the caesarean delivery rate – as a percentage of live births – has been rising globally,1 there is growing concern about unnecessary caesarean sections.2 Unnecessary caesarean sections can increase the risk of maternal morbidity, neonatal death and neonatal admission to an intensive care unit.2 At the same time, there is also concern that – in low-income countries in general and among the poorer sections of the populations in such countries in particular – caesarean sections are not always accessible, even when they are clearly indicated.3

There is no consensus on the “optimal” rate of caesarean delivery at the population level. Although values between 5% and 15% of live births have been suggested, the basis on which these thresholds have been proposed is not clear.4 Some historical studies indicate that low maternal mortalities can be achieved when the caesarean delivery rate is far below 15% of live births. In the Netherlands, for example, maternal mortality had fallen below 20 deaths per 100 000 live births by 1950, when caesarean sections were associated with less than 2% of live births.5,6 The results of some ecological studies indicate not only that no further reductions in mortality occur when caesarean delivery rates increase above 10%, but also that rates above 15% may be associated with additional mortality.7,8 The World Health Organization (WHO) has suggested that a caesarean delivery rate of 15% should be taken as a threshold that should not be exceeded – rather than a target to be achieved.4

The lower threshold for an “acceptable” rate of caesarean delivery has received much less attention than the upper threshold. Extremely low rates are indications that access to surgical care is poor and that, in consequence, women, fetuses and neonates are dying unnecessarily. As 1 to 2% of all births are associated with conditions that absolutely require caesarean sections to save the mothers’ lives – such as obstructed labour and complete placenta praevia – caesarean delivery rates of less than 1% or less than 2% are thought to reflect a real deficit in access to life-saving obstetric care and to be associated with excess maternal mortality.912 Rates of at least 5% are thought to be necessary to save the greatest numbers of both mothers and neonates, although there is little evidence to support such a cut-off.4

National rates of caesarean delivery can mask substantial within-country variation in the rates of such surgery. For example, urban rates are consistently found to be higher than rural rates13 and the rates for the poorest sections of the population often fall well below the national mean. In a retrospective analysis of data from Demographic and Health Surveys (DHSs) conducted in 42 developing countries, caesarean delivery rates were often found to fall below 1% either in the poorest quintile of the population (20 countries) or in all but the richest quintile (six countries).3 Only in five countries included in this analysis did the rate of caesarean delivery in the poorest quintile exceed 5% of live births.3

With Millennium Development Goals (MDGs) 4 and 5 nearing their target date of 2015, it is timely and necessary to assess recent progress in improving access to caesarean sections. In this paper we analyse trends in caesarean delivery rates in southern Asia and sub-Saharan Africa over the past 15 years. We focused on countries in southern Asia and sub-Saharan Africa because such countries account for 85% of all maternal deaths14 and 73% of all intrapartum neonatal deaths globally.15 We examined caesarean delivery rates over time and by wealth quintile and estimated, for each country, how many and which of the five wealth quintiles were experiencing caesarean delivery rates below 1%, 2% and 5%.

Methods

All of the data that we analysed – retrospectively – came from DHSs, which are nationally representative cross-sectional household surveys in which detailed birth histories for women of reproductive age are collected. All of the datasets that we used were downloaded from the MEASURE DHS website.16 In such surveys, socioeconomic status is evaluated – using principal components analysis – as a relative wealth index that is based on household assets. These indices then allow each surveyed household to be assigned to one of five wealth quintiles. The data that we used came from the countries in southern Asia or sub-Saharan Africa that were included in the “Countdown to 2015” initiative,17 although only data from the 26 countries where there had been at least two DHSs were analysed. The countries that we investigated were categorized into three regions – eastern and southern Africa, southern Asia, and western and central Africa – according to the classification of the United Nations Children’s Fund.18

We merged all available surveys for each country and pooled the data for all deliveries associated with a live birth in the 5 years preceding each survey whenever possible. In a few surveys, data on deliveries were only collected for the 3 or 4 years preceding the survey. We investigated the mode of delivery for each singleton birth and for the neonate who was born last in each multiple birth.

Deliveries that had been recorded as caesareans even though they had occurred in locations where caesarean sections were implausible – such as homes, dispensaries and health posts – were recoded as vaginal deliveries.19 The data on deliveries in higher-level facilities were excluded if information on mode of delivery was missing. However, the proportion of deliveries included in a survey that had missing information on mode of delivery never exceeded 3.3% – recorded in a survey in the United Republic of Tanzania in 1996 – and generally fell below 1%. The response rate in each of the surveys that we investigated was at least 90%.

We used three types of analysis. All analyses took account of sampling weights, in addition to clustering and stratification where appropriate. First, we calculated caesarean delivery rates by country and survey year. These rates were calculated as percentages of the deliveries that ended in live births – excluding, in multiple births, the deliveries of all but the last born neonates. We tested for time trends in these rates by using a binomial log–linear regression model20 to calculate annual rates of increase – as crude risk ratios (RRs) per year. Since caesarean sections are no longer a rare outcome in several of the countries that we investigated, odds ratios obtained with logistic regression would have overestimated the RRs. For each study country, annual rates of increase in caesarean deliveries were calculated for all the women and for the women who fell in the two lowest wealth quintiles combined – that is, for the poorest 40% of the women in the country. We also calculated caesarean delivery rates by wealth quintile and survey year within each country.

Finally, we categorized each delivery according to whether the mother lived in a rural or urban area and whether her household’s wealth index fell above the national median value – indicating that the mother was “richer” – or below it – indicating that the woman was “poorer”. This allowed us to evaluate caesarean delivery rates separately for relatively poor and wealthy urban women and relatively poor and wealthy rural women. All of the data analyses were performed using Stata SE version 12 (StataCorp LP, College Station, United States of America).

Results

Data were available for 80 surveys, which had been conducted in four countries in southern Asia, 11 countries in western and central Africa and 11 countries in eastern and southern Africa. The median number of surveys per country was three, with a range of two to four. In the surveys, data on births in the previous 5 years (n = 68), 4 years (n = 1) or 3 years (n = 11) had been collected. The total sample consisted of 686 789 deliveries – each of which had ended in a live birth – that had occurred between 1985 and 2011.

Table 1 presents the caesarean delivery rates recorded in the 3 to 5 years preceding each survey, by country and survey year, and the corresponding annual rates of increase. Statistically significant increases in caesarean delivery rates – varying from 2 to 19% per year – were observed in seven of the 11 study countries in western and central Africa, nine of the 11 study countries in eastern and southern Africa and all four of the study countries in southern Asia. However, only 12 of the study countries – three in western and central Africa, five in eastern and southern Africa and the four in southern Asia – showed evidence of an increase in caesarean delivery rates among the two lowest wealth quintiles. The crude RRs for the annual rates of increase in these 12 countries varied from 1.03 in Madagascar to 1.30 in Bangladesh. We were not able to calculate an annual rate of increase for the poorest 40% in Chad because caesarean deliveries had only been reported in one year in the surveys from Chad that we investigated.

Caesarean delivery rates were found to be very low in the sub-Saharan African study countries. In the most recent survey for each country, for example, 10 of the study countries in sub-Saharan Africa had national rates of less than 2% and only five countries – Ghana, Kenya, Lesotho, Rwanda and Uganda – had national rates of more than 5%. The corresponding rates recorded in the most recent survey in each of three of the study countries in southern Asia were much higher. Nepal was the only southern Asian study country in which the most recently recorded, national, caesarean delivery rate was less than 5%.

Table 2 (available at: http://www.who.int/bulletin/volumes/91/12/13-117598) and Fig. 1 present the caesarean delivery rates stratified by wealth quintile and survey. The rates were extremely low among the poorest quintile in every survey. In the most recent survey for each country, for example, the caesarean delivery rates among the poorest quintile were less than 1% in 12 of the study countries – all in sub-Saharan Africa – and they were less than 2% in all of the study countries except Lesotho, Malawi, Rwanda, Uganda and Zimbabwe. Caesarean delivery rates among the richest quintile were much higher in all of the study countries but exceeded 15% only in Bangladesh, India and Pakistan.

Fig. 1. Caesarean delivery ratesa by country, survey year and wealth quintile, southern Asia and sub-Saharan Africa, 1990–2011
Fig. 1. <b>Caesarean delivery rates<sup>a</sup> by country, survey year and wealth quintile, southern Asia and sub-Saharan Africa, 1990–2011</b>
a As percentages of the deliveries that ended in live births.
Note: The wealth quintile to which each surveyed household belonged was categorized as 1, 2, 3, 4 or 5. Quintile 1 comprised the poorest 20% of households and Quintile 5 comprised the richest 20%.

In the most recent survey for each of 17 of the study countries, caesarean delivery rates increased monotonically from the lowest quintile for wealth to the highest (Table 2). In the other nine study countries, the between-quintile variation in the rates was very small. In 10 of the study countries in sub-Saharan Africa – seven in western and central Africa and three in eastern and southern Africa – caesarean delivery rates of less than 1% had been recorded among the poorer 40% or 60% of women. In eight of these countries – Chad, Ethiopia, Guinea, Madagascar, Mali, Mozambique, Niger and Nigeria – the poorer 80% of women had caesarean delivery rates of less than 1%. The poorest quintile in three of the study countries in southern Asia had caesarean delivery rates of more than 1%. In Nepal, however, the corresponding rate for the two lowest quintiles for wealth combined was less than 1%. In seven of the eight study countries that had national rates above 5%, the overall rate for the three lowest wealth quintiles combined was less than 5%.

In the most recent surveys, caesarean delivery rates were highest among the “urban richer” in all 26 study countries and lowest among the “rural poorer” in 18 of the study countries (Table 3). In all four study countries in southern Asia, the caesarean delivery rate was higher among the “rural richer” than among the “urban poorer”; the absolute difference ranged from 2.6% in Nepal (95% confidence interval, CI: −2.0 to 7.2) to 10.2% in Bangladesh (95% CI: 7.7 to 12.7). Of the study countries in sub-Saharan Africa, however, only Ghana and Kenya had markedly higher caesarean delivery rates in the “rural richer” than in the “urban poorer” – with absolute differences of 5.0% (95% CI: −0.3 to 10.3) and 6.7% (95% CI: 3.0 to 10.4), respectively. In six western African and two eastern African countries, the rural women – whether “richer” or “poorer” – had caesarean delivery rates of less than 2%.

Discussion

Although caesarean delivery rates have been rising in almost all of the countries that we investigated in southern Asia and sub-Saharan Africa, they remain astonishingly low. In our analysis, 18 countries still had national rates of less than 5% recorded in their most recent surveys, and none of the study countries had a national rate above 10%. Caesarean sections were extremely rare among the poor: they were below 1% for the poorest 20% of the population in each of 12 countries, the poorest 40% in 11 countries and the poorest 80% in eight countries. They fell below 2% for the poorest 20% in each of 21 countries. Over the study period, the study countries in southern Asia experienced a much greater rise in caesarean delivery rates than the countries that we investigated in sub-Saharan Africa. Nevertheless, in the most recent surveys that we included in our analysis, the rates among the poorest 20% of the populations remained below 2% in all four of the southern Asian study countries.

The low rates of caesarean delivery in sub-Saharan Africa are presumably a reflection of very low levels of access to caesarean sections, which are themselves associated with extremely poor access to emergency surgical care in general.21,22 A recent study in Ghana, Kenya, Rwanda, Uganda and the United Republic of Tanzania – five countries included in our study – revealed massive gaps in the infrastructure for emergency surgical care.23 Fewer than 50% of the hospitals surveyed had dependable running water and electricity, and only 19–50% of the hospitals provided 24-hour emergency care.23 Countries in sub-Saharan Africa generally have few skilled workers able to perform surgery – including caesarean sections – and most of their qualified doctors live in urban areas.22,24 In the present study, caesarean delivery rates were extremely low among both the richer and poorer women who lived in rural areas, where structural and workforce constraints may be the most important barriers to access.

A household’s ability to pay for the surgery is thought to be an important determinant of caesarean deliveries.25,26 The cost of emergency caesarean sections can be catastrophic for households.25,26 Although user fee exemptions have been one of the key strategies to increase access to delivery care in sub-Saharan Africa,27 their impact on caesarean delivery rates has yet to be rigorously evaluated. While such fee exemptions may have contributed to the rises seen in caesarean delivery rates in countries such as Ghana and Senegal,28,29 such rises cannot be categorically attributed to the exemptions. Furthermore, a household’s ability to pay for surgery may not be the main barrier to caesarean sections in settings where the necessary health facilities are sparsely distributed.30

The rapid rises seen in caesarean delivery rates in southern Asia over our study period are somewhat surprising, given that most births in this region still take place at home. In the latest DHSs for Bangladesh, India, Nepal and Pakistan, for example, only 15%, 39%, 37% and 35% of the recorded deliveries occurred in a health facility, respectively (data not shown). However, many of these deliveries probably took place in private hospitals,30 where obstetricians and general practitioners are available to lead delivery care and the incentives to perform caesarean sections may be relatively greater.31 This may explain why such large proportions of the women who delivered in health facilities in Bangladesh, India, Nepal and Pakistan – 51%, 22%, 12% and 20%, respectively – had caesarean sections (data not shown). In the present analysis, caesarean delivery rates in the richest quintile were found to be more than 15% in Bangladesh, India and Pakistan, and the rates among the “rural richer” in all four study countries in southern Asia were found to be substantially higher than those among the “urban poorer”.

In every country that we investigated, caesarean delivery rates among the women in the richest quintile were much higher than the rates seen in the poorest quintile. This difference was particularly noticeable in Bangladesh, India and Pakistan, where the poorest quintile probably receives fewer caesarean sections than are indicated, while the richest quintile receives too many – increasing maternal and neonatal morbidity.32 In general – as postulated by the “inverse equity hypothesis” – the wealthy are more likely to adopt new medical interventions than the poor, often leading to increased health inequalities – at least in the short term.33 In southern Asia, however, the richest mothers appear to be receiving more caesarean sections than are warranted, with potentially adverse effects.

Our analysis has several limitations. First, we only had data for 26 of the 48 countries in sub-Saharan Africa and southern Asia that were included in the “Countdown to 2015” initiative.17. Second, the dates of the most recently published survey varied substantially between countries, and some countries may have made more progress since their most recent survey. The last available survey data for seven of the 11 study countries in western and central Africa were collected before 2007. Third, the caesarean delivery rates estimated in household surveys – generally from the statements of women of reproductive age – tend to be higher than the rates estimated from the records of the corresponding health facilities where caesarean deliveries may be performed.19 However, the facility-derived estimates tend to fall within the 95% CIs of the corresponding household survey estimates.19 Fourth, the wealth index used in the DHSs has several inherent biases that require careful scrutiny. The type of household assets investigated varies between the surveys, and the wealth index – which represents a household’s wealth relative to other households in one particular country at the time of the survey – should not be used to compare absolute levels of wealth between surveys. The association between household wealth and residence in an urban or rural area may be complex.34 Although those who live in urban areas are typically richer than their rural counterparts, the intrinsic meaning of the underlying wealth associated with many assets differs according to the area. We used national wealth indices – rather than urban- and rural-specific wealth indices – to enable direct comparisons between the richer and poorer halves of the populations in rural and urban areas. We were unable to analyse caesarean delivery rates according to wealth quintiles separately for urban and rural residents because the sample was too small, particularly in terms of the number of women from “urban poorer” households. Fifth, some women may have contributed more than one birth to the sample. However, restricting the analysis to only one birth per woman did not alter our findings (data not shown). Lastly, when computing annual rates of increase, we assumed that caesarean delivery rates increased in log–linear fashion. Our conclusions were, however, unaltered when RRs for the increases were calculated by comparing one survey to the next (data not shown).

Programmes to reduce maternal and neonatal mortality should have clear indicators to identify need, monitor implementation and change the course of action, as required.4 There has been a reluctance to include caesarean delivery rates as a core indicator for the monitoring of safe motherhood programmes, partly because the thresholds for “acceptable” or target rates are so uncertain, and partly because such an indicator may be perceived as promoting the unnecessary medicalization of obstetric care. However, this reluctance is unjustified, particularly when very low thresholds are set for the minimum rate. While caesarean delivery rates cannot be a substitute for the measurement of levels of maternal mortality, caesarean rates among the poor should be a key indicator for measuring progress towards achieving MDG 5.35 In the post-2015 health agenda – where the focus is shifting towards measuring the coverage for essential interventions – rates of caesarean delivery among the poor will be critical indicators of access to emergency obstetric care. In addition, as general childhood mortality is reduced, neonatal deaths become relatively more important and access to caesarean sections – when indicated to save the fetus – increases in relative importance as well. Although estimates of the caesarean delivery rate required for indications related to the fetus are imprecise,7 this rate is unlikely to be less than 5% of all births.

Despite the encouraging progress made in increasing national rates of caesarean delivery, large sections of the population in sub-Saharan Africa still lack access to life-saving caesarean sections, and women and children – particularly poor women and their children – are dying as a consequence. Improvements in access to caesarean sections will require massive investments in health system strengthening, particularly in terms of addressing shortages in the health workforce and the infrastructure gaps in rural hospitals.36 The human resource challenge could be partly addressed by allowing clinical officers to perform caesarean deliveries,37 although the sustainability of this strategy when implemented on a large scale remains uncertain. However, as long as hospitals lack the core infrastructure to perform surgery safely – including access to water and electricity – one cannot begin to address the emergency obstetric needs of pregnant women in sub-Saharan Africa.


Acknowledgements

We thank the Pelotas Equity Center for help in preparing the datasets and the Department of Infectious Disease Epidemiology at the London School of Hygiene & Tropical Medicine for input on the statistical analyses.

Funding:

This study was funded by the Bill & Melinda Gates Foundation, The World Bank and the Governments of Australia, Brazil, Canada, Norway, Sweden and the United Kingdom of Great Britain and Northern Ireland. FC was funded by a PhD studentship from the United Kingdom’s Economic and Social Research Council, via grant ES/I903224/1.

Competing interests:

None declared.

References

Share