Direct standardization method according to Robson classification for comparison of cesarean rates

Global cesarean section (CS) rate is approximately 21.1%, but there are significant differences in the access of CS all over the world. In least developed countries, CS represent about 8.2% of births, including sub-Saharan Africa with the lowest rate (5.0%). Meanwhile, the five countries with highest overall CS rates worldwide are Dominican Republic (58.1%), Brazil (55.7%), Cyprus (55.3%), Egypt (51.8%) and Turkey (50.8%), where the surgical deliveries surpass vaginal births [1].

It is agreed that the comparison of CS rates between populations or in the same population over time is problematic. The different characteristics of the obstetric population must be taken into account to avoid bias. To assist with this issue, many classifications for CS have been proposed over the years. In a systematic review, Torloni et al. [2] suggested the Robson Ten-Group Classification (RTGC) as the best option to meet local and international needs.

The RTGC is based on five parameters (obstetric history, onset of labor, fetal presentation, number of neonates and gestational age) to define ten mutually exclusive and fully inclusive groups (Table 1). The classification is based on a few routinely recorded variables and can be applied prospectively [3].

In 2015, the World Health Organization (WHO) published an Implementation Manual for RTGC use to assess, monitor and compare CS rates [3]. The manual proposes the construction of the Robson Classification Report Table for each population / time lapse and serial steps to interpret it, including data quality, type of population and assessment of CS rates. Therefore, it is a holistic and irreplaceable approach to deal with the complexity of the factors that influence CS rates. However, in certain situations in which it is necessary to work with the overall rate, the presentation and use of Report Table may not be applicable.

In recent years, many papers have been published comparing the overall CS rates as part of the results, even when data for RTGC is available [4‐9]. However, CS rates are expected to be different between RTGC groups, which makes meaningless and biased comparison of overall crude rates among populations with distinct RTGC distributions.

Understanding the complexity of factors that influence CS rates, the aim of this study is to propose a direct standardization method according to Robson’s classification in order to make overall CS rates comparable among different populations or in the same population over time. Researchers will benefit from having a single measure to depict CS rates. It will be useful when comparing overall rates in addition to the Robson Classification Report Table, or for some type of studies such as multi-group analytical ecological studies that deal with the Report Tables which are laborious and hard to interpret.

Figure 1 presents the CS rate for each country, crude and standardized by the RTGC. In addition, countries were ranked from highest to lowest CS according to the crude and standardized estimate. Comparing crude CS rates, the leading country is China with 47.6%, followed by Mexico (47.5%), Brazil (47.0%), Paraguay (46.8%) and Ecuador (45.5%). On the other hand, comparing standardized rates, the leading country is Paraguay with 46.9%, followed by Nicaragua (44.5%), Vietnam (43.0%), China (41.2%) and Mexico (40.9%). Brazil presented the biggest difference regarding its position in CS rank, from third place in the crude rate to ninth place in the standardized rate.

Table 3 shows the difference, for the crude and RTGC standardized rates, between the WHOMCS and the WHOGS data, in order to observe the implication of the standardization of CS rates in the same population over time. All countries, except for Nicaragua, Paraguay and Thailand, had their WHOMCS rates influenced by standardization. All countries, except for Peru, Thailand and Vietnam, had their WHOGS rates impacted by standardization (Table 3). Brazil, DRC, Ecuador, Kenya, Mexico, Nicaragua, Niger, Peru, Philippines and Uganda had their rates difference (WHOMCS rate – WHOGS rate) significantly affected by standardization.

When comparing crude rates, the largest increase in CS over time was observed in Brazil (20.0%) followed by Nicaragua (18.1%). Both countries also presented the greatest increases when assessing the standardized differences (12.5 and 15.6%, respectively). However, the magnitude of the observed increase was significantly lower.

With the standardization, the classification of countries according to CS rates changed considerably. As example, we observed that Brazil and Niger were the countries that presented the greatest difference between crude and adjusted rates, changing from 3rd to 9th place and from 21st to 19th place in the ranking, respectively. In the same way, variations of CS rates over time in the same country were also modified by standardization. As example, the CS rates in Brazil raised 20.0% when considering crude values and only 12.5% when considering the standardized rates.

According to Betrán et al. [17], the increase in the rate of CS without true medical need does not confer a gain on health, but it causes adverse results. In addition, it may increase the demand for CS in future pregnancies that could be avoided. Identifying this profile is important to build efforts to reduce this index. Confirming this point of view, the WHO concludes that efforts should focus on ensuring that CS are performed in cases that are necessary. Only when indicated for medical reasons, and not just aiming to achieve a specific rate [3]. In this context, studies that assess factors associated with CS rates or the impact of CS rates on adverse outcomes are important tools to control and manage of this problem in public health. Fear of pain, the idea that CS are safer for the baby, comfort for health professionals and the mother, and fear of medical litigation are associated with an increase in the CS rate due to non-medical factors. In addition, changes in the obstetric population such as prevalence of obesity, maternal age, health conditions presented during pregnancy (gestational diabetes and HIV infection, number of prenatal consultations and fetal presentation) are characteristics that influence CS rates [17‐19]. Therefore, standardization is necessary so that heterogeneous populations become comparable and enabling the real identification of changes in the prevalence of CS in a global context [20].

In the past, some authors have already proposed and tested methodologies to standardize CS rates. Lieberman et al. [21] described the use of direct standardization to compare CS rates between community-based and hospital-based practice settings in a teaching hospital. The percentages of women in each subgroup were determined according to obstetric history and conditions such as number of neonates, fetal presentation, gestational age and medical indication, which had many similarities with the parameters of RTGC. Hanley et al. [22] used logistic regression modeling to standardize CS rates by maternal characteristics (age, body mass index, gestational weight gain, smoking and parity) and conditions (hypertensive disorders in pregnancy, preexisting or gestational diabetes, gestational age, prior fetal or neonatal death, fetal presentation and number of neonates). Bailit and Garret [11] compared risk-adjusted methodologies for CS rates and found substantial agreement in the rankings from direct standardization and logistic regression methods. The authors reinforce that direct standardization is an easier method to comprehend and perform, what makes it a better decision on the condition that there are births in all risk strata.

With that being said, in 2015 the WHO released its statement on CS rates and proposed that the RTGC should be used as a standard instrument worldwide in the assessment, monitoring and comparison of CS rates over time. However, until the present study there is no reference on the RTGC use as a stratifying factor for standardization [23]. Standardization by RTGC merges the practice of standardization, already shown to be essential in comparing CS rates, with the standard instrument adopted by the WHO, taking an important step towards the unification of methods for CS rates comparison.

In the present study, considering only WHOMCS results, Niger and Brazil were the countries that had the greatest variation between crude and standardized rates. With the standardization, Brazil had a 10.5% drop and Niger showed a 12.7% rise in CS rates. This means that the proportion of CS in Brazil is originally high and in Niger is low, but the characteristics of its obstetric population partially justifies these numbers, and must be considered when compared with CS rates of other countries. Moreover, the relative size of Robson groups in these countries are very different from the relative sizes of the standard population, which explains the great variance between crude and standardized rates. In that regard, it is also important to note that the adjusted rate will vary according to the defined standard population [10].

For example, more than 60% of Niger obstetric population is in Robson group 3 (Multiparous [excluding previous CS], single, cephalic, > = 37 weeks, in spontaneous labour) and only 4.8% of this group have CS. For other groups, like 2, 4 and 5, that all together represent only 5.3% of the obstetric population, CS rates reach more than 50% [9]. Contrarily, group 3 of the standard population represents 32.3% of the total, while groups 2, 4 and 5 together are 23.3% of the total. In Paraguay, the country that presented the smallest variation between crude and adjusted rates, more than 30% of the obstetric population in Paraguay is in the Robson group 1 (nulliparous, with a single fetus, cephalic, > = 37 weeks, in labor spontaneous) and 36% of this group has CS. In other groups, such as 8 and 9, which together represent 1.6% of the obstetric population, the CS rates are 100%. However, standardized population group 1 contributes 29.4% to the overall CS rate and groups 8 and 9 together contribute 2.5% to the overall CS rate. In that country, the relative size of Robson’s groups is very close to the relative size of the standardized population, which justifies their smaller variation between rates.

Among the strengths of this study, we can highlight the large sample size studied in several countries and the possibility of comparison over time. As limitations, we cannot control the inconsistency of missing values and correct classification in the Robson groups in the different countries in the sample. The data used for the study is from 2011, but we do not consider it as a limitation, as the objective of the study was not to evaluate CS rates or to bring new data, but to test the methodology of direct standardization.

Finally, it is important to highlight that the standardization is only recommended for the purpose of comparisons. Planning health actions based on standardized CS rates, eliminating specificities such as territorial reality, culture and social conditions would lead to a biased analysis that could negatively affect health services [24‐27]. For issues such as management and planning of health services for specific location and time, standardization is not indicated and considering the real crude numbers is mandatory [24].

Our results highlight the need of standardization for comparison of CS rates. Analysis comparing crude rates of CS are biased, as they ignore the characteristics of each obstetrical population and the demand of necessary CS according to these characteristics.

Although many authors have raised the problem of comparing crude rates of CS [20‐22], the adjusting methods proposed included complex modeling, were dependent of variables not routinely available, being mainly applicable for comparison between hospital units. The RTGC has been widely used all over the world, it is based on a few basic obstetrics variables and its use for a direct standardization is not only easy to apply and interpret, but also is attuned with WHO recommendation in 2015, and may complement the results of studies using the Robson Classification Report Table.