Sources of data and estimation of national CS rates

We obtained national CS rates from several data sources: (1) CS rates from routine statistical surveillance system reports or national surveys from government health offices were considered to provide nationwide estimates (12 countries); (2) CS rates from the World Health Organization (WHO) Health Indicators database, the WHO European Health for all database, or the 2005 WHO World Health Report were assumed to be national CS rates, unless stated otherwise (52 countries); (3) CS rates from national surveys that included the Demographic and Health Surveys (DHS), which reports from surveys that have been conducted since 1990, which were considered to be nationally representative (59 countries); and (4) CS rates published in the literature (13 countries) and personal communication by the ministry of health (1 country) were considered to provide country-level estimates if they specifically stated that the figures represented country rates.

In published articles that reported hospital CS rates (only considering births that occurred at hospital level), we considered them to be national rates if the country had a proportion of deliveries at health facilities of >90%. For countries with a proportion of hospital deliveries of <90%, the same assumption would result in overestimates of CS national rates. Thus, in those cases, we adjusted the rate by multiplying the CS rate by the proportion of births in health facilities. When the proportion of hospital deliveries was not available, we used the proportion of births that were attended by skilled health personnel (4 countries).

When country data were available for several years or several sources, the most recent data were retrieved. In cases in which data from different sources differed, the most reliable source was used at the authors’ judgment. Sources of data for each included country are shown in the Supplementary Table .

Estimation of the worldwide number of CSs in countries with low and high CS rates

The annual number of CSs that were performed in each country was calculated by the multiplication of the CS rate by the annual number of births. The number of births was obtained from health statistics provided by UNICEF for 2008. Data by country are available in the Supplementary Table .

The adequate range for the CS rate in a country remains a matter of debate. We based our decisions on the following assumptions: (1) The recommended minimum CS rate at population level to avoid death and severe morbidity in the mother lies from 1-5%, according to the WHO and other researchers. Regarding neonatal outcomes, studies that have evaluated the association of CS rates with neonatal death have shown outcome improvements in CS rate of 10%. Thus, the minimum threshold for a population level CS rate could be considered to lie between 5% and 10%. (2) Regarding the upper level, the best-known recommended upper limit is 15%, which was proposed by the WHO in 1985. Although these figures are based on theoretic estimates, 2 recent observational studies have supported that recommendation. Both studies assessed the association between CS rates and mortality and morbidity rates in mothers and neonates and found no reductions in those indicators when the frequency of CSs was >15%. Moreover, 1 study showed that an increased rate of intervention was associated with higher mortality and morbidity rates in mothers and neonates. Until further research gives new evidence, rates of >15% may result in more harm than good.

On the basis of these 2 assumptions, we primarily classified countries in 3 groups according the national rates of CS: (1) countries with CS rates of <10%, (2) countries with rates of 10-15%, and (3) countries with CS rates of >15%. In a secondary more conservative analysis, we expanded the range of category 2 to 5-20% CS rates.

In countries with CS rates of <10%, we calculated the number of additional CSs that would be needed to increase the national rate to 10%; to achieve this rate, we multiplied the annual number of births by 10, minus the CS rate. In countries with CS rates of >15%, we calculated the number of CSs above this value by multiplying the annual number of births by the CS rate minus 15. We followed the same approach for the secondary analysis with the 5-20% CS rate category.

Estimating the cost

An “ingredients approach” was used to measure the costs of CS. This approach requires information about the quantities of the physical inputs that are needed and their cost per unit. That is, a standard set of inputs that are required to perform CSs and related procedures was defined. These inputs were then cost analyzed in the different settings, with consideration that account determinants of cost vary by setting. Only the marginal resources that are associated directly with the CS and related procedures were cost analyzed; in other words, none of the routine costs that are associated with antenatal care visits were included nor were other services that would be considered part of normal vaginal delivery (such as the costs of skilled birth attendants, tetanus prophylaxis, or clean cord practices) or the costs of care that are associated with morbidity and death that result from a failure to perform CS.

The quantities of inputs that are required at the point of care were estimated from various sources, which included expert opinion and treatment practice guidelines. A standardized profile for CS inputs at point of care was used for all countries; these included the initiation of labor at referral level, the diagnosis of obstructed labor and referral, CS-associated devices and medicines, operative facility time, medical human resources time, management of shock that includes hysterectomy and blood transfusion (assumed for 1% of CSs performed), and postoperative hospital stay for stabilization.

The point-of-care input profile was further augmented by standardized estimates of the resources that are required to establish and maintain these point-of-care services, which include program administration, training, the corresponding office space, electricity and other services, and a variety of standard consumables and equipment.

Only economic costs were considered; thus, mark-ups that are due to monopoly power or corruption or losses that are due to wastage were not considered. The costs that were considered, therefore, will be underestimates of the actual financial costs in many settings and possibly overestimates in others if procedures are performed at substandard care levels. The estimated costs should represent the costs of performance of the procedure according to recognized practice guidelines under idealized market conditions.

For point-of-care inputs, the cost for countries with CS rates of <10% was calculated as the cost of the resources that are required to bring the country’s CS rate up to 10% (as a proportion of live births in that country); the cost in those countries with CS rates of >15% was calculated as the cost of the resources that are involved in the performance of CSs of >15% of live births in that country. For the costs of program administration and related services, which are not incurred at the point of care, only the proportional component of these costs that are attributable to CS in excess of 15% of deliveries or that are required to bring the CS rate up to 10% of deliveries, respectively, was included in the estimates of total costs.

Unit costs for the inputs that were identified were derived from a search of published and unpublished literature and databases and from consultation with costing experts. For goods that are traded internationally, the most competitive international price that could be identified was used. For example, drug prices were estimated on the basis of the median supply price that was published in an international price guide, with a standardized mark-up applied to account for transportation and distribution. For goods that are available only locally (eg, human resources, inpatient bed days), costs have been shown to vary substantially across countries, so cross-country regressions that account for national income levels and local characteristics of the supply of health care were used to generate estimates of costs per unit.

Ethical approval for this analysis was not required.