Introduction

Over the past 30 years, there has been a significant increase in the use of caesarean section (CS) in many parts of the world [1–3]. From an global rate of 6.7% in 1990 to 21.1% in 2018, some regions of the world have seen a spectacular increase in their CS rates, such as East Asia (from 4.9% in 1990 to 33.7% in 2018) and Latin America (from 22.8% in 1990 to 42.8% in 2018) [1,2]. This trend affects many low- and middle-income countries (LMICs), where it is common to observe both an abusive practice of CS among women at low obstetrical risk and predominantly socioeconomically advantaged, and an under-use of this procedure among the most disadvantaged, due to a lack of access to care [4,5]. However, there is no benefit in performing a CS in the absence of a medical indication, and particularly in resource-limited settings where it may increase morbidity and mortality [6,7]. The increase in CS represents a serious challenge for LMICs because of the increase in maternal and perinatal morbidity associated with its excessive use, the health inequalities it causes and the resulting diversion of otherwise limited resources [4,8–10].

There are many reasons for the excessive use of CS [8]. As summarized by the ecological model of Betrán et al., many non-clinical factors related to clinicians’ practices (e.g., loss of skills in conducting vaginal deliveries, clinicians’ preference for CS) and to health system (e.g., financial incentives, lack of institutional guidelines) influence the use of CS [8]. Besides, mothers’ demand for this mode of birth has often been cited to justify the overall increase in CS rates [8,11]. Indeed, this demand seems to be associated with non-medically justified use of CS in certain contexts [12]. Women’s demand for CS is often motivated by fear of pain and complications associated with vaginal birth, as well as by the perceived advantages of CS, such as the possibility of planning the birth and the supposed safety of this mode of delivery [8,13–19]. According to Panda et al., maternal request for CS may then encourage clinicians to perform the procedure, mainly due to the fear of litigation from the women and their families or personal convictions regarding women’s rights and their autonomy to choose their mode of delivery [11].

Since the number of CS performed on maternal request is often difficult to measure—particularly when this practice is not officially authorized and documented in medical records—previous studies have explored the relationship between women’s preference for CS and their actual mode of birth [20–24]. Considering the preference for CS as an indicator of maternal demand, these studies aimed to assess its influence on the decision for mode of birth [20–23]. In Norway, nulliparous women who preferred cesarean delivery were significantly more likely to undergo antepartum CS, but also, to a lesser extent, intrapartum CS [23]. In China, Deng et al. estimated that the probability of giving birth by CS on maternal request increased when the woman had a preference for CS in late pregnancy [22]. However, this link may vary depending on the context, and on whether women’s preference is taken into account in the decision on the mode of delivery [25,26].

In addition, several studies have concluded that women’s preference alone cannot explain the dramatic increase in CS rates, as the proportion of women expressing this preference remains low [13,25–27]. In 2011, a meta-analysis revealed that approximately 22% of women in LMICs preferred CS [27]. More recently, a systematic scoping review indicated that the proportion of women preferring CS varied from 1.4% to 50% in these countries [13], highlighting the heterogeneity of contexts and the different methodologies used to measure women’s preferences for this mode of delivery [13]. Thus, questions remain regarding the actual impact of women’s preference and demand for CS on the increasing overuse of CS observed in many LMICs.

Materials & methods

Results

A total of 1,827 low-risk women were included in this analysis. A flowchart describes the recruitment and selection of the study sample in Fig 1. In this sample, 10.4% of women preferred CS (14.5% in Argentina, 1.8% in Burkina Faso, 19.1% in Thailand and 5.1% in Viet Nam) and the average CS rate was 24.5% (448/1827), ranging from 18.2% in Burkina Faso to 27.6% in Viet Nam (Fig 2 and Table 1). The results show that CS rates were higher among women who preferred a CS than among those who preferred a vaginal birth (Fig 2). The Chi2 test was significant in each country (Fig 2). This difference was particularly marked in Viet Nam, where 93% of women who indicated a preference for CS had a CS delivery, compared with 23% of women who indicated a preference for vaginal birth (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of women included in this analysis and institutional factors according to maternal preference and mode of delivery (Quali-Dec).

https://doi.org/10.1371/journal.pone.0339007.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart representing the recruitment and selection of included women.

https://doi.org/10.1371/journal.pone.0339007.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Mode of delivery among low-risk women, according to their preference and by country (n = 1,827).

https://doi.org/10.1371/journal.pone.0339007.g002

Table 1 presents the socio-demographic and clinical characteristics of the women in the sample, as well as the institutional characteristics of the hospitals in which they gave birth, according to the exposure variable (preference in late pregnancy) and the outcome (mode of delivery). The calculation of the propensity score was based on 12 variables: maternal age, wealth index, residence, maternal profession, level of education, BMI, birthweight of the newborn, presence of complications, type of antenatal care, presence of a dedicated anaesthetist in the delivery room and capacity for care during labour (ratio between the average number of deliveries per day and the total number of beds in the delivery room).

Table 2 presents the results of the evaluation of the matching quality. The distribution of variables between the two comparison groups did not differ after matching. The mean absolute bias was less than 5%, Rubin’s B was less than 25 and Rubin’s R was between 0.5 and 2, as recommended. Therefore, the two comparison groups were considered sufficiently balanced.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Assessment of matching quality by propensity score.

https://doi.org/10.1371/journal.pone.0339007.t002

Table 3 presents the association between women’s preference and mode of delivery, in the overall sample and by country, based on the matched sample. The results show that, on average, preference for CS increased the probability of having a CS by 32% (CI95% [0.23–0.41]; p < 0.001). The relative risk was estimated at 2.69 (CI95%: 2.43; 2.95) (Table 4). Given the low proportion of women preferring CS in this sample (10.4%), the fraction of CS associated with women’s preference was estimated at 15% (CI95%: 12.9% − 16.9%) (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Average treatment effect (ATT) of women’s preference for CS on the mode of delivery after matching, in the total sample and by country (Quali-Dec).

https://doi.org/10.1371/journal.pone.0339007.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Relative risk* (RR) and fraction of CS associated with women’s preference for CS.

https://doi.org/10.1371/journal.pone.0339007.t004

Analysis by country showed that preference for CS increased the probability of delivery by CS by 34% (CI95% [0.17–0.51]; p < 0.001) in Argentina; by 31% (CI95% [0.19–0.43]; p < 0.001) in Thailand; and by 58% (CI95% [0.33–0.82]; p < 0.001) in Viet Nam (Table 3). The proportion of CS associated with women’s preference for CS ranged from 12% in Viet Nam to 23% in Thailand (Table 4). In other words, we estimated that 12% of CS (CI95%: 10.2%; 14.4%) in the sample were performed due to a preference for CS in Viet Nam, while this proportion was 17.8% (CI95%: 10.8%; 23.9%) in Argentina and 23% (CI95%: 18.2%; 28.0%) in Thailand (Table 4).

Discussion

In the 32 participating hospitals across Argentina, Burkina Faso, Thailand and Vietnam, the preference for CS among low-risk women appears to increase the probability of having a CS by 32% (ranging from 31% in Thailand to 58% in Vietnam). Nevertheless, the proportion of CS associated with preference for this mode of birth was calculated to be 15% (ranging from 12% in Vietnam to 23% in Thailand), which is relatively low given the low disposition of women to prefer CS and it is due to the small proportion of women preferring CS in these populations.

In LMICs, the majority of studies reporting clinicians’ views describe maternal demand for CS as a ‘key factor’ in the observed increase in CS rates [53]. However, our findings do not confirm the perception of clinicians and suggest that women’s preference for this procedure is a relatively minor contributor to the overuse of CS. Three other literature reviews, based on studies of women’s demand or preference rather than clinicians’ views, also suggest that women’s preference for CS is only a secondary factor which, according to Gamble and Creedy, “may divert attention away from physician-led influences on the continuing high cesarean section rates” [25–27]. The high CS rates observed among low-risk women in the selected hospitals may be explained by other factors linked to the quality of care and clinicians‘practices, or to the organization of the healthcare system [8,11,16,19,54–57]. In Burkina Faso, for example, general practitioners who lacked obstetric skills but worked in maternity wards due to a shortage of specialized staff were more likely to perform unnecessary CS [58]. In Thailand, obstetricians describe vacuum or forceps extraction as requiring “a high level of expertise and skill” that they believe they do not have and consider CS to be the best option to avoid adverse events [17]. In addition, CS is perceived by some doctors as a way to better balance their clinical and academic responsibilities, and personal time, as well as to ensure the baby’s safety. In contrast, vaginal delivery is often perceived as more unpredictable in terms of neonatal outcomes [17]. In Viet Nam, healthcare professionals have reported experiencing considerable pressure and being targeted on social media. This results in CS conducted for fear of litigation as CS is seen as a form of protection [16]. This finding is also described in Thailand and Argentina [17,59]. Financial incentives to perform CS rather than vaginal deliveries are regularly mentioned in the literature [8,11,54,59–62]. This is probably the case in Thailand and Viet Nam, where private practice within public hospitals, observed during the research but not yet documented, is widespread. Finally, certain organizational factors may contribute to the use of CS. A previous QUALI-DEC study showed that emergency intrapartum CS was significantly associated with a higher bed occupancy level [48], revealing clinicians’ preference for performing CS in order to facilitate the organization of care and, in their opinion, ensure optimal quality of care [11,63,64]. Furthermore, in Argentina, shortage of midwives in some hospitals or their limited role during labor and delivery, while obstetricians manage labor in most public hospitals, even for low-risk pregnancies, are described as factors that may contribute to excessive use of CS [59]. The association observed in our study between women’s preference and mode of delivery is in line with the findings of previous studies in both high-income countries (HICs) [20–23] and in LMICs [20,22]. Our finding may suggest that women who prefer CS are more likely to request it, and some doctors are more inclined to abide by this preference in the absence of medical indications. However, we have no data to indicate whether women’s preference resulted in a request for a CS in our study because the practice of CS on demand is not officially accepted in most of the four participating countries. Based on our findings, we assume that in Viet Nam, and to a lesser extent in Argentina and Thailand, women’s preference for CS tends to be considered by doctors when deciding on the mode of delivery. In Thailand, for example, from providers’ view, CS meets the needs of women and their families while preventing litigation in the event of adverse outcomes in vaginal childbirth [17]. In Argentina, women’s preference is supported by a law which, since 2015, guarantees women the right to participate in the decision-making process throughout pregnancy and birth [65].

Although the proportion of CS associated with women’s preference for this mode of delivery is low, it varies between countries. The highest proportion of CS associated with women’s preference is observed in Thailand and can be explained by the high prevalence of women’s preference for CS in that country (19%) compared to other countries. This result had already been mentioned in a previous publication [36], in which we also observed that preference for CS was more pronounced among nulliparous women and was linked to the fear of pain, especially in Thailand [36]. In this country, pain management is nearly non-existent, and nulliparous women account for more than half (51.6%) of our sample, which could explain the higher prevalence of women’s preference for CS. Similarly, there is strong belief in Thailand that the date of birth should be chosen at an auspicious time to ensure the destiny of the child and the family [17].

In Viet Nam, where the proportion of women who prefer CS is low (5%), the strong statistical association between preference and actual mode of birth in Viet Nam could be misleading as doctors were probably the most influential factor in shaping women’s preference for CS [36]. The influence of the doctor in the preference for CS is well documented in the literature [13,14,22,53,66]. Several studies in Iran, Brazil and Lebanon, have shown that women who requested a CS were influenced in their decision by their doctor, who reinforced their feeling that a CS was the safest option [13,66–69]. Especially in Viet Nam, this raises questions about the quality of information given by clinicians to their patients. According to the principles of ethical care, healthcare professionals must provide valid information that reflects scientific progress, and is free from personal beliefs or interests that may influence patients’ decision [70].

Taking patients’ preferences into account is one of the fundamental elements of shared medical decision-making [71,72]. Shared decision-making, which limits inappropriate interventions and reduces the heterogeneity of medical practice, is a key component of the quality of obstetric care. Involving patients in decision-making also improves their satisfaction with care and their knowledge [53,70,71,73,74]. However, it is not just a question of meeting women’s expectations, regardless of the risks and benefits of each delivery mode. This shared decision-making process includes the obligation to provide patients with impartial and valid information and to engage in a balanced dialogue that considers the woman’s values and perspectives, to reach consensus of the course of action [71,72,75].

Our findings have implications for clinical practice. Low-risk women who reveal a preference for CS during their pregnancy may be interested to use the QUALI-DEC decision analysis tool (DAT) which will help them to make an informed choice [35,76]. This DAT may improve patient-provider communication. But given the relatively small proportion of CS associated with this preference, its potential impact on CS rates could be limited if this DAT is used alone, not in combination with other interventions targeting healthcare providers or health care systems. Assuming the challenge of its widespread use is met, this DAT could help reduce the overall use of CS by encouraging clinicians to base their decisions on the risks and benefits of each mode of delivery and to consider the preferences of most women for vaginal birth. However, our findings also highlight the need to design and implement multidimensional and context-specific strategies that are consistent with WHO recommendations on non-clinical interventions to reduce unnecessary CS rates [28]. These recommendations emphasize the importance of simultaneously addressing as many non-clinical factors as possible that may influence the excessive use of CS. According to the WHO, providing women with evidence-based information, strengthening clinicians’ adherence to guidelines, and promoting shared decision-making are effective interventions when adapted to the local context. In addition, and aligned with WHO recommendations, our findings confirm that it is essential to ensure that women receive adequate support during pregnancy and access to appropriate methods of pain relief during labor and delivery. These interventions reduce fear and anxiety associated with pain and childbirth, thereby reducing the preference for CS [77].

To our knowledge, this is the first study to quantify the extent of the influence of women’s preference on their mode of birth, measured in a comparable and rigorous manner in four LMICs. The multisite and multicountry design is a strength of this study, as it could allow the results to be generalized to similar contexts, although it should be noted that our non-probabilistic sampling technique limits this generalizability. In addition, we carried out a comprehensive data collection to adjust the analyses on the individual and institutional variables that influence the use of CS. The quality of the survey data was assured based on the WHO data collection and management system for the surveillance of maternal and perinatal health [78]. Finally, the analysis methodology used enabled us to control confounding biases in order to reliably measure the association between women’s preference for CS and their mode of delivery.

There are also several limitations to this analysis. The first is that women were questioned in post-partum about their preference in late pregnancy. This approach presents a risk of recall bias, compared with prospective studies in which women are interviewed during pregnancy [24]. In addition, the stated preference may have been influenced by the outcome of their recent childbirth. Women may have been biased toward reporting a preferred mode of delivery consistent with the type of delivery they had just experienced, thereby inflating the observed association. We think that adjusting the measure on actual mode of delivery may control partially this recall bias. The more appropriate measure of the association between women’s preferences and their mode of delivery would be a prospective study examining women’s preferences during pregnancy. Due to budgetary and research organization constraints in the countries, the option of interviewing women twice (in late pregnancy and then post-partum) was not retained. However, several mechanisms were put in place to minimize the disadvantages of the chosen approach, including the preparation of a survey questionnaire reviewed by many experts, tested in each country and used by specially trained interviewers. In addition, we were not able to analyze according to the type of CS (during or before labour), due to a limited number of women. This stratified analysis would have allowed us to test the hypothesis that women’s preference for a CS is more likely to lead to a planned CS than an emergency CS during labour. Matched propensity score analyses enabled control for confounding bias. This method has been reported to provide a more accurate estimate of the association between an exposure factor and an outcome and more reliable statistical inference than multivariate logistic regression analysis [38]. However, unmatched women, who tend to have ‘extreme’ propensity scores (outside the common support), are excluded from the analysis, resulting in a loss of power. In addition, matched women may have similar propensity scores but very different characteristics, as the matching is not based on the variables themselves but on a score combining all these variables. Nor can we state that all the variables explaining both the exposure (women’s preference) and the outcome (caesarean delivery) were included in the calculation of the propensity score. Indeed, some likely confounding factors could not be measured, in particular the attitudes and clinical practices of providers and institutional norms are not captured in the data. Finally, we calculated the proportion of CS associated with preference for this mode of birth. However, we cannot definitely establish a causal relationship between women’s preferences and their mode of birth, as the data are cross-sectional and do not account for the temporality of events. For this reason, we have not used the term “attributable fraction” but rather “associated fraction”. Therefore, the findings on associated fractions provide an indication of the contribution of women’s preference to the use of CS and should be interpreted with caution.

Conclusion

Although women’s preference plays a role in the use of CS in the QUALI-DEC participating hospitals, it likely accounts for only a small proportion of CS performed. A more nuanced understanding of this influence is crucial for designing targeted interventions and shaping their focus. These findings reinforce the need for multidimensional interventions that engage all relevant stakeholders. The results suggest the need to provide women with appropriate, evidence-based information, to improve clinicians’ adherence to guidelines, and to strengthen shared decision-making, in order to reduce the overuse of CS. It is also necessary to address other determinants, in particular the systemic factors that influence the use of CS. Finally, it is necessary to improve women’s experience of childbirth so that they can feel confident in choosing and envisaging vaginal birth.

Supporting information

Acknowledgments

We thank all women who participated to the survey and healthcare providers who facilitated the data collection in participating hospitals. We also thank Andrainolo Ravalihasy, a statistical engineer at CEPED unit who advised the first author (C.E.) on statistical analyses, and all the members of the QUALI-DEC consortium group. The composition of this group is as follows: Karolinska Institutet (Stockholm, Sweden): Claudia Hanson, Helle Molsted-Alvesson, Kristi Sidney Annerstedt, Amanda Cleeve; University College Dublin, National University of Ireland (Dublin, Ireland): Michael Robson; World Health Organization (Geneva, Switzerland): Ana Pilar Betrán, Newton Opiyo, Meghan Bohren; Centro Rosario de Estudios Perinatales Asociacion (Rosario, Argentina): Guillermo Carroli; Liana Campodonico; Celina Gialdini; Berenise Carroli; Gabriela Garcia Camacho; Daniel Giordano; Hugo Gamerro; CEDES (Buenos Aires, Argentina): Mariana Romero; Khon Kaen University (Khon Kaen, Thailand): Pisake Lumbiganon, Dittakarn Boriboonhirunsarn, Nampet Jampathong, Kiattisak Kongwattanakul, Ameporn Ratinthorn, Olarik Musigavong; Fundacio Blanquerna (Barcelona, Spain): Ramon Escuriet, Olga Canet; Centre national de recherche scientifique et technologique—Institut de Recherche en sciences de la santé (Ouagadougou, Burkina Faso): Charles Kabore, Yaya Bocoum Fadima, Simon Tiendrébéogo, Zerbo Roger; Pham Ngoc Thach University of Medicine (Ho Chi Minh, Viet Nam): Mac Quoc Nhu Hung, Thao Truong, Tran Minh Thien Ngo, Bui Duc Toan, Huynh Nguyen Khanh Trang, Hoang Thi Diem Tuyet; Research Institute for Sustainable Development (Paris, France): Alexandre Dumont, Laurence Lombard, Myriam de Loenzien, Marion Ravit, Delia Visan. The lead author for the QUALI-DEC group is Alexandre Dumont (contact: alexandre.dumont@ird.fr)