Study setting

Matlab is a rural sub-district under Chandpur district located about 55 km Southeast of Dhaka, the capital of Bangladesh. International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b) has been maintaining a Health and Demographic Surveillance System (HDSS) in this area covering a population of 230,000 (in 2013) since 1966 [20]. Matlab HDSS area is divided into 2 equal halves: icddr,b service area (SA) and government service area (Fig 1). The present study was conducted in Matlab icddr,b service area (divided further into 4 blocks, A to D) where under a maternal, neonatal and child health (MNCH) program, free basic emergency obstetric, neonatal and child health care services are provided from four icddr,b sub-centres and one central icddr,b Matlab hospital (Fig 1) [15]. In addition to icddr,b services, basic and comprehensive emergency obstetric and neonatal care (EmONC) services are available from public sector subdistrict (Matlab) and district hospitals (Chandpur), and from a growing number of for-profit private sector hospitals and clinics situated mostly in Chandpur district town [15].

Download:

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

Fig 1. Study area in Matlab with population level C-section rate.

https://doi.org/10.1371/journal.pone.0188074.g001

Study design and data collection method

This is a retrospective study where all women who delivered in icddr,b service area during 2013 were included in the analysis. Several existing HDSS databases were linked to retrieve required socio-demographic, service use and the birth outcome data to perform the statistical analysis to look into the indications and socio-demographic determinants of C-section. We used the following HDSS data files: a) birth file, b) pregnancy registration file, c) the socioeconomic census 2014, and d) verbal autopsy database. Women who delivered in 2013 were identified from the HDSS birth file. Service utilization and birth-outcome data were available from pregnancy registration file, birth file and icddr,b facility records (four sub centres and icddr,b Matlab hospital).While information on household construction materials, source of drinking water, type of latrine use and other asset variables were available from the Matlab Icddr,b 2014 socio economic census, which were subsequently used to construct wealth quintiles using Principal Component and Factor Analysis Methods [21, 22]. The number and causes of maternal deaths were gathered from routine verbal autopsy forms of the HDSS. All monitoring, surveillance and socioeconomic survey databases were linked together using HDSS assigned unique registered identification number (RID) of pregnant women to perform required analysis.

All women who delivered in icddr,b service area during 2013 were identified reviewing the birth and pregnancy files. Among them, who delivered by C-section, their indications were collected reviewing health facility records where the procedure took place. One project research physician identified the public and private facilities where C-section actually took place and visited all these facilities that include Chandpur district hospital and seven private clinics situated in Matlab and Chandpur to collect information on clinical indication of caesarean sections. Facility records review included patient admission file, in-patient registrar, operation theatre registrar, and bed-head tickets. However, the indications of C-sections for 195 women could not be retrieved following this procedure, therefore ‘obstetric complications’ recorded as referral indications in icddr,b service registrars (maintained electronically) were accepted as proxy indications for C-sections [23]. Even though, data were missing for 22 women who probably bypassed icddr,b service delivery system. Indications of C-section for these 22 women were gathered through direct interview with mothers. A female study investigator with obstetric background conducted all these 22 interviews. A structured check-list was used to guide the women to recall their birth events to retrieve true clinical indication of C-section undertaken. Interview data were validated using other proxy sources such as referral notes, doctor’s prescriptions and discharge certificates (if any). All C-section indications data were entered in a data base and linked with other HDSS databases using unique HDSS identifier (RID of the mother). WHO, International Classification of Disease version 10 (ICD 10) [24] was used to classify the indication of C-sections. Field data collection was completed in 4 months during September to December 2014. One programmer assisted in data entry, cleaning and linking with other HDSS databases.

Data analysis

Indications of C-sections were first presented as frequency distribution of individual clinical conditions (ICD-10). These indications (of C-section) were further divided into different groups following three distinct classification systems: 1) Absolute vs. Non-Absolute Maternal Indications 2) Primary vs. Repeat C-sections, and 3) Maternal vs. Fetal indications; to identify the specific risk-groups for C-section.

ICD-10 codes for indication of C-section were grouped into 10 subclasses: hypertensive disorder, mal presentation, disorder of amniotic fluid, antepartum haemorrhage (APH) including placenta praevia, post-dated pregnancy, prolonged & obstructed labour, fetal distress, previous caesarean delivery, maternal disorder related to pregnancy, and general disease complicating pregnancy. The fetal distress group includes non-specified fetal distress and fetal distress due to meconium-stained liquor. Hypertensive disorder covers gestational hypertension, pre-eclamptic toxaemia (PET) and eclampsia. Amniotic fluid disorder category includes oligo and poly-hydramnion. However, cephalo-pelvic disproportion (CPD) and failed induction of labor are included in the category of prolonged or obstructed labor. Finally, C-section indications documented as Maternal distress, Rh negative mothers and retention of urine were grouped under ‘pregnancy-related maternal disorder’, while thalassemia, anaemia, asthma belong to ‘general disease complicating pregnancy’ category. Later on, C-section indication documented under APH including placenta praevia’ ‘pregnancy-related maternal disorder’ and ‘general disease complicating pregnancy’ were re-categorised under ‘other indication’ (Fig 2)

Download:

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

Fig 2. Distribution of C-section indications under eight broad categories according to ICD-10 code.

https://doi.org/10.1371/journal.pone.0188074.g002

AMIs include 4 distinct life-threatening obstetric complications: uncontrolled bleeding, unstable lie or presentations (transverse lie, face or brow presentation), gross cephalo-pelvic disproportion (CPD) and uterine rupture [25].

Primary C-section group included women having C-section for the first time, while repeat C-section group included women with one or more C-section prior to current birth.

Maternal causes included clinical conditions such as hypertensive disorder, amniotic fluids disorders, post-dated pregnancy, maternal distress, Rh-negative mother, psychological disorder and general diseases complicating pregnancies such as thalassemia, anaemia, asthma, and retention of urine. Fetal causes included multiple gestation, big baby, and fetal distress. Problems related to both mother and fetus such as prolonged or obstructed labor, CPD, failed induction, placenta praevia, and malpresentation were grouped as combined cause.

Further analyses focused on exploring socio-demographic determinants of C-section delivery. The exposure variables included; maternal age, religion, education, employment status, wealth quintile, birth order, number of Ante Natal Care (ANC), history of fetal loss, delivery outcome, sex of the baby, fetal number, and gestational age while the outcome variable was ‘delivery by C-section (0 = delivery by other method; 1 = delivery by C-section). At bivariate level, Chi-square tests were performed to explore the relationship between outcome and exposure variables. Multicollinearity between exposures variables were checked using Variants Inflation Factors (VIF). The exposure variables showing significant relationship with the outcome variable at bivariate level were included in the binary logistic regression model. We used Mantel-Haenszel test to examine the independent effects of exposure variables upon C-section delivery after controlling the confounding effects of all covariates in the model. The strength of association between exposure and outcome variables were measured as adjusted odds ratios (OR) with 95% confidence intervals (CI) of odds ratios. All analysis was performed in STATA 13 (STATA Corp.)

Characteristics of women and newborn

A total of 2549 women delivered in the icddr,b service area with a known birth-outcome during 2013 were included in the study. Of them 87% were Muslims and their mean age was 26 years. More than 3/4^th of them completed secondary level education and only 2% were engaged in any formal job. On average 76% of this study cohort attended at least three antenatal (ANC) visits and 19% had bad obstetric histories (fetal loss in earlier pregnancies) (Table 1). Most of the newborns were live births (98.9%), singleton (99.1%), and born full-term (89.4% (Table 2).

Download:

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

Table 1. Sociodemographic characteristics of study participants.

https://doi.org/10.1371/journal.pone.0188074.t001

Download:

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

Table 2. Birth outcome data of the study participants.

https://doi.org/10.1371/journal.pone.0188074.t002

Delivery patterns

Out of 2549 deliveries, 84% were conducted in health facilities: 43% in icddr,b facilities, 33% in for-profit private facilities and 8% in government facilities. During study period, only 16% deliveries were conducted at home (Table 1). Of all deliveries 35% were conducted by C-section and the rate was highest in block A which includes Matlab town (Fig 1). Majority of C-sections (81.7%) were conducted in for-profit private sector facilities.

Indications of C-section

Based on ICD-10 classification, ‘previous history of C-section’ was the most common indication (24.1%) for doing C-sections. Other indications included: ‘fetal distress’ 20.6%, ‘prolonged and obstructed labor’ 15.9%, ‘amniotic fluid disorder’ 14.3%, ‘post-dated pregnancy’ 13.1%, ‘maternal disorder related to pregnancy’ 4.5%, ‘fetal mal-presentation’ 3.5%, ‘hypertensive disorder in pregnancy’ 2.5%, ‘placenta praevia’ 0.78%, and ‘general disease complicating pregnancy’ 0.7% (Fig 2).

Of all C-sections only 1.44% was conducted for AMIs. In another way, we found that 58.3% C-sections were conducted for maternal causes, 21.5% for fetal causes and the rest 20.2% were conducted for both maternal and fetal causes. Of 214 (24%) repeat C-section cases, there was even no secondary indication for repeating this surgical procedure. (Fig 3)

Download:

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

Fig 3. Distribution of C-section indications by three different categories.

CS for AMI (Yes & No); Repeat CS (Yes & No); Underlying cause (Fetal, Maternal & combined).

https://doi.org/10.1371/journal.pone.0188074.g003

Determinants of C-section

In multi-logistic regression model, when the effects of all co-variates were controlled statistically, probability for using C-section delivery increased with better socio-economic status, higher education of women, increasing age, decreasing birth order, and with presence of bad obstetric history. Women from richest quintile households were two and half times more likely to deliver by C-section compared to women from the poorest quintile households (OR: 2.47; 95% CI: 1.78–3.34). Similarly women with higher secondary and above education were two times (OR: 2.06 95% CI: 1.24–3.25) more likely to deliver by C-section than women with no education. Women with three or more birth order were 68% less likely to deliver by C-section (OR: 0.32 95% CI: 0.23–0.44) than women with 1^st birth order. Women attending more than three ANC sessions were 2 times more likely to deliver by C-section than women having ‘0–2’ ANC sessions (OR: 2.19 95% 1.67–2.82). Women having history of fetal loss were about one and half times more likely to deliver by C-section (OR: 1.38; 95% CI: 1.10–1.73) than women without bad obstetric histories. (Table 3)

Download:

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

Table 3. Determinant of caesarean delivery.

https://doi.org/10.1371/journal.pone.0188074.t003

Reported adverse maternal outcome

Verbal autopsy data reveals that during study period, a total of eight maternal deaths took place in the study area. Of them, seven took place during postpartum period and one during antepartum period due to complication of heart disease. Of seven postpartum maternal deaths, five took place after C-section and in private hospitals. None of these five deaths had AMI for undergoing C-section. The cause of post caesarean death for five mothers were as follows: two from post-partum haemorrhage, one from severe anaemia and kerosene poisoning, and the rest two from post-partum infection and pleural effusion.

Main findings

In this rural area facility delivery rate was 84% and C-section rate was 35% of all deliveries. C-section rate was higher than national average (23%) [10] and even more higher than global recommendations [26]. Only 1.44% C-sections were conducted for AMIs. The top five indications were repeat C-section (24%), fetal-distress (21%), prolonged labor (16%), oligohydramnios (14%) and post-maturity (13%). In the repeat C-section group, there was no other valid indications that contribute to save lives. Higher age, lower birth order, higher education, better socioeconomic status, bad obstetric history, and three or more ANC uptake were significant predictors for C-section delivery. Of total eight maternal deaths, seven took place during postpartum period, five were among women who underwent C-section in private sector facilities.

Interpretations

Similar to the findings in other settings [27], the population based C-section rate in the study population (35%) was higher than recommended upper limit of 15%. One recent ecological study has argued that the safe upper limit of population based C-section rate could be 19% [19]. Referring to this findings, WHO in their recent strategic document [28] has emphasized more on monitoring indications rather than concentrating on appropriate C-section rates [29, 30]. The five most common indications explored in this study have also contributed for global excess of C-section rates. One study has explored that in 2001, 30% of C-sections in developed countries were for repeat C-sections [31]. Another study in Bangladesh reported that in government hospitals 35% of C-sections were actually repeat C-sections [32]. National Institute for Health and Clinical Excellence [33] and the American College of Obstetricians and Gynaecologists [34] have clearly instructed that previous C-section should not be an indication in absence of any obstetric emergencies. There is also proof that the success rate for vaginal birth after C-section (VBAC) was much higher (80%) and morbidity rate was much lower than those with repeat C-section [29, 31]. Thus the priority should focus on reducing primary C-section rate through engaging both client and providers in informed decision making process [29, 32].

Fetal distress, identified as the second leading cause of C-section (20.6%) in this study population, has a reported global prevalence of about 20% [35] and was accounted for about 16% C-section at tertiary level hospitals in Bangladesh [32, 36]. There is a range of medical intervention, from simple maternal left lateral position, oxygen inhalation, to para-cervical amnio-infusion [37] for restoration of fetal heart rate. The reported success rate for para cervical amnio-infusion was about 70% even in limited resource setting[35] and therefore it is imperative to encourage care-givers to practice these interventions before opting for emergency C-section for fetal distress.

Prolonged labor was the next most common cause of C-section (15.9%) in the current study. It also contributed for the rising C-section rate in other population-based studies, with prevalence as high as 30% [31, 38]. Active management of second stage of labour by augmentation was found to be effective to manage prolonged labour [29, 39] without causing severe birth asphyxia[40, 41]. However, over-reporting of prolonged labour as indication of C-section could be avoided if the failed induction of labour is reported separately. In addition to that monitoring labour through plotting a simple graph such as partograph was found to reduce C-section rate by around 31%, as documented by studies done in different settings in United States, United Kingdom and South Africa [40]. In contrast to that, partograph use rate in Bangladesh is documented to be very low. A need assessment study done among 24 districts of Bangladesh has reported the partograph user rate as three percent [42]. Another study conducted in the health facilities of Khulna and Sylhet divisions of Bangladesh has conferred that partograph user rate was 5% to 33% [14].

Oligohydramnios was the indication for 14% of C-section in this study. However, a prospective cohort study done in Pakistan has documented that isolated oligohydramnios is not associated with adverse perinatal outcome such as high rate of birth asphyxia or admission to intensive care unit compare to women having normal amniotic fluid. Thus elective C-section for possible perinatal morbidity due to oligohydramnios is not recommended for any instance[43]. Moreover, it has been reported that estimated rate of oligohydramnios could be up to six percent [44] and majority of the cases the superfluous results of low amniotic fluid comes from subjective variation on sonography [45] Higher C-section for oligohydramnios in this study raises issues regarding validity of ultra-sonographic results which is common in current practice context in Bangladesh. Nevertheless, serial sonography in all four World Health Organization (WHO) recommended ANC visits are crucial to detect congenital fetal anomaly responsible for oligo hydramnions and also to identify potential threat for many fetal complication in advance. [46]. But the data from recent national demographic survey suggest that only 31% of the women in Bangladesh are receiving 4+ ANC.[9]. Furthermore, several studies have reported the poor quality of ANC mostly due to lack of physical infrastructure, more waiting time, less supportive behaviour from service providers and lack of Standard Operating Procedure (SOPs) and evidence based practice like counselling, health education [47, 48].

Similar to the finding of this study, advance maternal age has been documented as influencing factor for high C-section rates in other studies [49–53]. Certain biological changes occur with the advancement of age during pregnancy, such as mal-position, increased risk of hypertension, eclampsia, and diabetes [50]. Maternal preference together with these risks might have increased the caesarean delivery among older mother [50].

As in many other studies Socio Economic Status (SES) was found to be positively associated with C-section delivery [54]. However, the opposite trend has also been observed in developed countries, where higher education and economic status were protective against C-section as awareness and knowledge of childbirth are expected to be high among this group of women [53]. However number of ANC visit was positively associated with higher C-section delivery in this study and other studies that may raise question about its efficacy in controlling C-section epidemic [52, 55–57]. Thus the quality of ANC need to be prioritised along with the number of ANC taken with particular emphasis on engaging women on sensitive discussion on risk and benefit of both normal delivery and C-section and to provide emotional support on taking decision[58]. Though Bangladesh has already prioritised the quality ANC in their policy agenda, the inequity in service uptake has observed. The most recent study confers that 4+ ANC uptake is mostly prevalent in wealthier and educated women, and the group of women residing in urban area[59]

Majority of C-sections were conducted in for-profit private sector facilities that supports the view that private provision of care may increase C-section rate for incentive mechanism associated with the procedure [60, 61]. One study in Brazil has revealed that the prevalence of C-section delivery particularly in private sector hospitals did not correlate with obstetric risk factors [53]. In this context, clinical audit and feedback expected to have supportive role to guide the health professional in analysing and modifying their practice around evidence based clinical guideline on safe delivery care. One meta- analysis from Canadian studies has shown 13% reduction of caesarean section rate upon introduction of clinical audit and feedback and showed more positive result up to 27% C-section reduction with multifaceted intervention including second opinion and positive cultural change [62].

However, it is also evident that C-sections done regardless of obstetric risk factors, could be a means of defensive obstetric practices and that may increase the risk of maternal death particularly in low resource settings [1]. Our study supports the view as of eight maternal deaths among the study population five were related to C-section delivery with the clear evidence of poor quality of care. In relation to that WHO in their strategic document “Beyond the number” urged for verbal autopsy for all maternal deaths as every delivery has a story to tell which can provide relevant and important information for strengthening health system response [63].Bangladesh has been implementing Maternal and Perinatal Death Review (MPDR) at community and facility level through social and verbal autopsy respectively since 2010. One evaluation has shown improved quality of care and increased response to any obstetric emergency at community and facility level[64]. Currently MPDR has been implemented in the 17 out of 64 districts under the Quality improvement Initiative of government of Bangladesh [65].

Strength and limitations

The strength of the study is that the data comes from an established and long standing health and demographic surveillance system which has enabled retrieval of entire birth cohort data for with no missing values for required analysis. All C-section indications could be captured from multiple sources. All these are strengths of the study. There was the chance for recall bias for getting indication from women where data were unavailable from facility records, given that interviews occurred one year after the birth events. However interviews were conducted by trained obstetrician with similar experiences and collected data using well-designed interview guideline and were validated with referral notes and discharge certificates available with women to overcome this weakness.