Prevalence and determinants of caesarean section in private and public health facilities in underserved South Asian communities: cross-sectional analysis of data from Bangladesh, India and Nepal

Objectives To describe the prevalence and determinants of births by caesarean section in private and public health facilities in underserved communities in South Asia. Design Cross-sectional study. Setting 81 community-based geographical clusters in four locations in Bangladesh, India and Nepal (three rural, one urban). Participants 45 327 births occurring in the study areas between 2005 and 2012. Outcome measures Proportion of caesarean section deliveries by location and type of facility; determinants of caesarean section delivery by location. Results Institutional delivery rates varied widely between settings, from 21% in rural India to 90% in urban India. The proportion of private and charitable facility births delivered by caesarean section was 73% in Bangladesh, 30% in rural Nepal, 18% in urban India and 5% in rural India. The odds of caesarean section were greater in private and charitable health facilities than in public facilities in three of four study locations, even when adjusted for pregnancy and delivery characteristics, maternal characteristics and year of delivery (Bangladesh: adjusted OR (AOR) 5.91, 95% CI 5.15 to 6.78; Nepal: AOR 2.37, 95% CI 1.62 to 3.44; urban India: AOR 1.22, 95% CI 1.09 to 1.38). We found that highly educated women were particularly likely to deliver by caesarean in private facilities in urban India (AOR 2.10; 95% CI 1.61 to 2.75) and also in rural Bangladesh (AOR 11.09, 95% CI 6.28 to 19.57). Conclusions Our results lend support to the hypothesis that increased caesarean section rates in these South Asian countries may be driven in part by the private sector. They also suggest that preferences for caesarean delivery may be higher among highly educated women, and that individual-level and provider-level factors interact in driving caesarean rates higher. Rates of caesarean section in the private sector, and their maternal and neonatal health outcomes, require close monitoring.


INTRODUCTION
Access to comprehensive emergency obstetric care, including caesarean section, is key to preventing the estimated 287,000 maternal and 2.9 million neonatal deaths that occur worldwide every year. 1 2 Although debate continues about how to quantify the need for life-saving obstetric surgery, a 1985 World Health Organization (WHO) report suggested that the optimal population range for caesarean section rates is between 5% and 15%, and this endures as a reference. 3 4 Caesarean section rates are increasing worldwide, albeit unequally: a recent analysis of Demographic and Health Survey (DHS) data in 26 South Asian and sub-Saharan African countries found that rates were highest among the 'urban rich' in all countries, and lowest among the 'rural poor' in 18. 5 In all countries, fewer than 5% of mothers in the poorest wealth quintile delivered by caesarean. 6 Caesarean sections conducted without clinical need can have adverse consequences for mothers and children. A 2008 WHO survey of 373 facilities across 24 countries found that unnecessary caesareans were associated with an increased risk of maternal mortality and serious outcomes for mothers and newborn infants, compared with spontaneous vaginal delivery. 7 Recent ecological analyses also highlighted strong associations between caesarean delivery and increased neonatal mortality in countries with low and medium caesarean section rates. 8 Unnecessary caesareans lead to considerable costs for families and health systems: an estimated 6.2 million unnecessary procedures were performed in 2008, costing approximately US$ 2.32 billion. 9 Several South Asian countries have recorded substantial increases in caesarean section rates over the past decade. In Bangladesh, rates rose from 2% (2000) to 17% (2011); in India, from 3% (1992) to 11 % (2006); and in Nepal, from 1% (2000) to 5% (2011).  Several studies in these countries have raised concerns about high caesarean rates in private facilities, and a recent DHS analysis speculated that national increases in caesarean section rates in South Asian countries could be driven in part by higher rates among deliveries in private sector facilities . 6 16-18 Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to desire caesareans. 19 20 There are few large, recent community-based studies from South Asia quantifying differences in caesarean section rates between public and private facilities. Such studies are necessary in order to examine whether increases in caesareans in these settings are indeed likely to be driven by the private sector, demand from wealthier and more educated mothers, or a combination of the two. We conducted a cross-sectional analysis of data from Bangladesh, India, and Nepal to explore the prevalence and determinants of caesarean section delivery by type of facility and location, and to examine the

Study populations
We used data collected through vital events surveillance systems established during four clusterrandomised controlled trials conducted between 2005 and 2011. The trials were done with communities that can be considered socio-economically disadvantaged: in Bangladesh and Nepal, they took place in four rural, underserved districts (Bogra, Maulvibazaar, and Faridpur in Bangladesh, and the Terai district of Dhanusha, Nepal); in rural India, most participants were from Scheduled Tribes in two states of eastern India (Jharkhand and Odisha); in urban India, data came from slum communities in Mumbai. Table 1 describes the characteristics of each study and its population, including the background maternal mortality ratio and types of facilities present in the study areas. The original trials were designed to evaluate the impact of participatory women's groups on maternal and neonatal health outcomes. [21][22][23][24][25] We used data from the control areas of these trials only, because the women's group intervention led to changes in mortality and practices in several locations. 26 Health system contexts All three study countries have experienced substantial increases in institutional delivery rates over the last two decades: births in health facilities increased from 4% (1993) to 29% (2011) in Bangladesh; from 26% (1992-3) to 47%  in India; and from 8% (1996) to 35% (2011) in Nepal. [10][11][12][13][14][15] All three countries have implemented incentive schemes to promote institutional delivery, though these have varying coverage. In 2010, Bangladesh's pilot maternity voucher scheme reached an estimated 10.4 million people across 31 sub-districts, around 7% of the country's population. 27 28 Mothers receive a cash incentive for antenatal care and delivery in a public or private facility, or at home with a skilled birth attendant. Both government and private facility staff also receive cash incentives, including 3000 Bangladeshi Taka (US $38.5) for a caesarean section and 300 Taka for a normal delivery. 28   and its impact is unlikely to be reflected here. The Indian urban area included in our study spanned informal settlements (slums) with a wealth of public and private providers. 30 Mothers with Below Poverty Line cards in such areas are eligible for JSY and can receive Rs 500 towards the costs of delivering in a health facility. 31 Nepal began a safe delivery incentive scheme in 2005 and free deliveries have been available in government facilities since 2009 through the Aama Surakshya programme. 32 Dhanusha district, from which data for this study come, has one zonal tertiary hospital and a private medical college hospital equipped for c-sections, and a variety of small public and private health facilities without comprehensive obstetric care.

Data collection
All data were collected using surveillance systems to monitor births and deaths prospectively. [21][22][23][24][25] In all study locations, a female community-based key informant reported births and deaths in her area, which covered a population ranging from 250 to 350. A trained interviewer then verified these reports and paid the informant an incentive for each correct identification. In Bangladesh, rural and urban India, the interviewer administered a structured questionnaire to all eligible mothers around six weeks after delivery; in Nepal, all births in the study area were registered, and interviews were conducted on all births in small clusters and on a random sample of 10 births per month in the larger clusters. In each study location, mothers were interviewed using a questionnaire to collect information about events in the antenatal, delivery and postnatal periods. In each trial and in this subsequent study, participants were women of reproductive age  who delivered in the study area during the data collection period, and who consented to be interviewed six weeks after delivery, and their infants.

Study sample
The initial sample for this analysis included 46,393 births in Mumbai, rural India, rural Bangladesh, or rural Nepal, of whom 17,565 (38%) were delivered in health care facilities. We excluded the 2348 births occurring outside Mumbai, because we collected limited information on delivery location for these.  The primary outcome in these analyses was caesarean section delivery, identified by self-report from the mother or another household member around six weeks after giving birth. The main covariate of interest was the type of delivery facility, coded as public (e.g. district hospitals), private (individual private clinics or hospitals), or NGO/charitable (e.g. Christian missionary hospitals in rural India; BRAC clinics in Bangladesh). We grouped private and NGO/charitable facilities together due to the small number of births in NGO/charitable facilities (6% of institutional deliveries in Bangladesh, 7% in rural India, and 0 in urban India and rural Nepal; data on request). Additional covariates in the models included measures of the characteristics of the pregnancy and delivery, maternal socio-demographic characteristics, and the location in which the delivery occurred. We created an indicator variable to identify women experiencing serious problems during pregnancy or delivery, with women reporting symptoms of pre-eclampsia (blurred vision or swelling of face and hands), symptoms of eclampsia (fits or seizures during pregnancy or delivery), haemorrhage during delivery, or labour lasting more than 24 hours considered to have serious complications. Other characteristics of the pregnancy and delivery considered in the analysis included full utilisation of antenatal care, number of prior pregnancies, and multiple pregnancy. Full utilization of antenatal care was defined as four or more visits to antenatal care, with at least one visit to a skilled provider. Number of pregnancies was entered as an ordered categorical variable, with categories of one (first), two, three, or four or more pregnancies.
Maternal socio-demographic characteristics included in the models were: mother's age at delivery, her educational attainment, and household assets. Mother's age was entered into the model as a categorical measure in 10-year groups. Educational attainment was entered as a categorical variable using the following categories: no formal education, primary education, secondary education, or bachelor's degree or higher. To develop an asset index, we used polychoric factor analysis on data on common assets and amenities found in the mother's household, and grouped the resulting factor scores into quartiles. 33 Assets and amenities included electricity, radio or cassette player, electric fan (Bangladesh and India only), television, refrigerator (Bangladesh and India only), telephone (Bangladesh and Nepal only), generator (India only), and bicycle. All models were additionally adjusted for location (Bangladesh, India, Nepal), and year of interview in three-year groups. The data were collected in a stratified, clustersampled survey, and we accounted for survey design in the analysis using a fixed effect for stratum and a random effect for cluster. socio-economic and socio-demographic characteristics, pregnancy and delivery characteristics, and year of delivery. Table 3 shows crude and adjusted measures of association between type of delivery facility (private or public) and caesarean section for each location. Delivering in a private health facility was associated with an increased odds of caesarean section in all but one location (rural India). The relative odds of a caesarean in a private facility were greatest in Bangladesh (OR: 6 Women who had four or more antenatal check-ups were significantly more likely to have a caesarean delivery in rural Bangladesh and Nepal, with positive but non-significant trends in all other locations. Parity was not associated with caesarean delivery in any location. Having a serious health complication during pregnancy and delivery was associated with caesarean delivery in all locations except rural Bangladesh, where we observed a negative association (AOR: 0.87, 95% CI: 0.76, 1.00). Multiple birth was associated with an increased odds of caesarean section only in urban India and in rural Nepal.
Higher maternal age was only associated with an increased odds of caesarean section in urban India.
Maternal education was only associated with an increased odds of caesarean delivery in rural  education having four times greater odds of caesarean delivery in public facilities than women with no formal education (AOR 4.55, 95% CI: 2. 22, 9.33). There was also an educational gradient among women delivering in private facilities in urban India. Educated mothers delivering in private facilities were more likely to deliver by caesarean than were mothers in other groups (AOR 2.10; 95% CI: 1.61, 2.75).
However, there was no apparent gradient by educational attainment among women delivering in public facilities.

DISCUSSION
Our analysis of data from over 45,000 births confirms the findings of other studies identifying differences in caesarean rates between public and private facilities, and suggests that, even in underserved areas in South Asia, caesareans without medical indication are of concern. In three of four locations, rates of caesarean section were higher in private/charitable facilities than in public facilities. The findings from Bangladesh are particularly noteworthy as they show much greater odds of caesarean section in private facilities, concurring with previous analyses. 37 This was also the only location where serious complications in pregnancy and delivery were not associated with caesarean delivery, suggesting that obstetric surgery was performed over and above clinical need.
Our findings confirm the results of earlier studies of the prevalence of caesarean delivery in South Asia, and indicate that high rates can be found in underserved rural areas. In India, a recent analysis of 2010-11 Annual Health Survey (AHS) data from 284 districts in nine States, including Jharkhand, found that the median caesarean section rate in the private sector was 28%, compared with 5% in the public sector. 38 There appear to be strong financial incentives for surgical procedures in the private sector. 39 In a rural, largely indigenous part of eastern India where more mothers delivered in private/charitable facilities than in public facilities, more caesarean sections were performed in the public sector. In this particular setting, our field experience suggests that only women with serious complications would go to a facility and have a caesarean section, especially as JSY was not yet available in the study areas at the time of data collection. Such women are likely to have experienced multiple referrals from either ill-equipped public or private facilities not wanting to take the risk of admission. It is however possible that preference for the private sector has changed in Jharkhand since the advent of JSY, and further disaggregated analyses of Annual Health Survey data would allow a more contemporary exploration of State-level variations in caesarean section rates between public and private sectors.
In Mumbai slums with a high uptake of institutional births and a 60/40 split in favour of the public sector for delivery care, we found no difference in caesarean section rates between the two sectors, which is somewhat surprising given the high rates of caesareans observed in the private sector in nine other states of India (excluding Maharashtra). At least two scenarios may be relevant in this setting. First, public sector hospitals are able to provide caesarean section, while smaller private maternity homes may not be able to. Second, the costs of caesarean section borne by private sector providers are higher in Mumbai, and families may be unwilling to bear these additional costs. This provides motivation for private providers not to provide caesareans, and makes it likely that women experiencing complications move into the public sector. The variety of facility options in an urban environment means that mothers are able to choose between different types of facilities, while in rural areas only one provider may be accessible.
Analyses using the most recent Bangladesh DHS found that, in 2011, three in five facility births were delivered by caesarean section. This reflects a historical trend: in 2001-2003, nearly half of deliveries in private facilities in Bangladesh were already by caesarean section. 18 Our data from rural, socioeconomically disadvantaged communities in three districts of Bangladesh confirm these population-level findings, but also suggest that high caesarean section rates in private facilities are not merely an issue for wealthy urban mothers. Although some research from Bangladesh suggests that mothers may have a preference for caesarean delivery because of fear of labour pain or a desire to select an auspicious date for the birth, other studies also highlight women's fears of caesarean section and their distrust of health providers who recommend them, mainly because of the high costs associated with the procedure. 37 40 41 A qualitative study involving twenty women who had experienced obstetric complications in Matlab in 2008-9 found that most of the fourteen women who had undergone caesareans had spent over Tk 14,999 (US$ 217) on the procedure, which was approximately one-third of GDP per capita at the time. 41 The lack of association between caesareans and complications in pregnancy or delivery and multiple births, coupled with the high financial incentives given to providers for performing caesarean sections and the requirement for junior doctors to 'practise' their surgical skills, further suggest that obstetric surgery is being used over and above clinical need. A possible explanation for the significant interactive associations between maternal education and caesarean delivery in both public and private facilities in Bangladesh is that well-educated women may be delivering in more expensive or highly rated institutions, which may in turn be more likely to do caesarean sections for financial reasons and if they act as training centres for junior doctors. seek care in private facilities may also be more willing or able to pay for caesarean sections, and providers may conduct more of them to increase their income.
A 2010 analysis of DHS data examining the role of the private sector in maternity care in 16 countries found evidence of a trend towards privatisation in delivery care between the 1990s and mid-2000s, but with strong differences between countries, which might reflect the heterogeneous nature of this sector both between and within countries. 42 This DHS study highlighted the need for more context-specific data on the nature of the private sector in low and middle-income countries, and its role in maternity care. Further research might focus on understanding the motivations and experiences of women undergoing caesarean sections in private facilities in South Asian settings, pathways for switching between public and private sectors in the event of obstetric complications, a more comprehensive tally of the financial incentives (official or non-official) that motivate private providers to carry out caesareans in each setting, and the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.
In our study, locations with higher prevalence of caesarean deliveries also showed a positive educational gradient even after adjusting for wealth measured by household assets. Moreover, there was an interactive association between education and type of facility, with highly educated women particularly likely to receive caesarean deliveries in private facilities in Bangladesh and urban India. Literature exploring the determinants of caesarean delivery emphasises that multiple influences drive a woman's decision to deliver by caesarean. Profit considerations at the facility level may prompt some providers to urge women to receive unnecessary procedures, while women themselves may prefer caesareans for cultural reasons, fear of painful deliveries, or because they believe it to be safer. 36 43-45 Within the clientprovider interaction, providers may be more likely to acquiesce to a request for caesarean from a highly educated woman; and such women may be more likely to accept advice from a provider. 19 46 Our findings lend support to the hypothesis that, while provider-level factors are probably partially responsible for the rapid increase in caesarean deliveries, it is also necessary to consider women's own preferences and decision-making processes and how they are shaped by social and cultural factors. previous delivery was by caesarean, which limits the completeness of our analysis of determinants. In addition, levels of serious complications in pregnancy and delivery varied considerably between locations and were often higher than expected, suggesting potential over-reporting and limited reliability as an indicator of complications. Finally, some variables had small denominators, and others had high levels of missing data (for example, maternal age in rural India).

Conclusions
Our study found that delivering in a private health facility was associated with an increased odds of caesarean section in three of four South Asian locations, and that the associations persisted after adjustments for maternal, pregnancy and delivery characteristics, and year of delivery. We also found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and urban India (private facilities only). These results lend support to the hypothesis that increased caesarean section rates in these three South Asian countries may in part be driven by the private sector, but also suggest that, in some settings such as Bangladesh and urban India, demand from more educated mothers may play a part. These findings call for greater, local understanding of the role of private providers in maternity care, together with careful examination of the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.

Health facility characteristics
Public health facility 68.

Study design 4
Present key elements of study design early in the paper 1, 5-6 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection 5 and Table 1 Participants 6 (a) Cohort study-Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up Case-control study-Give the eligibility criteria, and the sources and methods of case ascertainment and control selection. Give the rationale for the choice of cases and controls Cross-sectional study-Give the eligibility criteria, and the sources and methods of selection of participants 6 (b) Cohort study-For matched studies, give matching criteria and number of exposed and unexposed

Key messages
• Caesarean sections were more common in private than in public health facilities in three of four study locations (rural Nepal, rural Bangladesh, and informal settlements in urban India).
• We found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and in informal Indian urban settlements (private facilities only).
• Future research should monitor caesarean section rates in the private sector in South Asia and subsequent maternal and neonatal health outcomes

Strengths and limitations
• This study had a large sample size and focused on underserved communities, which are a priority for public health interventions in South Asia.
• Our data were not nationally or sub-nationally representative, which limits the generalisability of our findings. Other limitations included small denominators and missing data for some variables.
We did not have data on some known predictors of caesarean section, which would have enhanced the completeness of our determinants analysis.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  Although debate continues about how to quantify the need for life-saving obstetric surgery, a 1985 World Health Organization (WHO) report suggested that the optimal population range for caesarean section rates is between 5% and 15%, and this endures as a reference. 3 (2000) to 5% (2011).  Several studies in these countries have raised concerns about high caesarean rates in private facilities, and a recent DHS analysis speculated that national increases in caesarean section rates in South Asian countries could be driven in part by higher rates among deliveries in private sector facilities . 6 16-18 Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to desire caesareans. 19 20 There are few large, recent community-based studies from South Asia quantifying differences in caesarean section rates between public and private facilities. Such studies are necessary in order to examine whether increases in caesareans in these settings are indeed likely to be driven by the private sector, demand from wealthier and more educated mothers, or a combination of the two. We conducted a cross-sectional analysis of data from Bangladesh, India, and Nepal to explore the prevalence and determinants of caesarean section delivery by type of facility and location, and to examine the  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   5 interactions between maternal education and caesarean section delivery in private and public facilities.

Study populations
We used data collected through vital events surveillance systems established during four clusterrandomised controlled trials conducted between 2005 and 2011. The trials were done with communities that can be considered socio-economically disadvantaged: in Bangladesh and Nepal, they took place in four rural, underserved districts (Bogra, Maulvibazaar, and Faridpur in Bangladesh, and the Terai district of Dhanusha, Nepal); in rural India, most participants were from Scheduled Tribes in two states of eastern India (Jharkhand and Odisha); in urban India, data came from informal settlements (slums) in Mumbai. Table 1 describes the characteristics of each study and its population, including the background maternal mortality ratio and types of facilities present in the study areas. The original trials were designed to evaluate the impact of participatory women's groups on maternal and neonatal health outcomes. [21][22][23][24][25] We used data from the control areas of these trials only, because the women's group intervention led to changes in mortality and practices in several locations. 26 Health system contexts Social Health Activists (ASHAs) also receive 600 Indian Rupees ($9.6) for identifying pregnant women and helping them get to a health facility. 29 Although the rural Indian data included in this study were  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Data collection
All data were collected using surveillance systems to monitor births and deaths prospectively. [21][22][23][24][25] In all study locations, a female community-based key informant reported births and deaths in her area, which covered a population ranging from 250 to 350. A trained interviewer then verified these reports and paid the informant an incentive for each correct identification. In Bangladesh, rural and urban India, the interviewer administered a structured questionnaire to all eligible mothers around six weeks after delivery; in Nepal, all births in the study area were registered, and interviews were conducted on all births in small clusters and on a random sample of 10 births per month in the larger clusters. In each study location, mothers were interviewed using a questionnaire to collect information about events in the antenatal, delivery and postnatal periods. In each trial and in this subsequent study, participants were women of reproductive age  who delivered in the study area during the data collection period, and who consented to be interviewed six weeks after delivery, and their infants.

Measures used
The primary outcome in these analyses was caesarean section delivery, identified by self-report from the mother or another household member around six weeks after giving birth. Maternal socio-demographic characteristics included in the models were: mother's age at delivery, her educational attainment, and household assets. Mother's age was entered into the model as a categorical measure in 10-year groups. Educational attainment was entered as a categorical variable using the following categories: no formal education, primary education, secondary education, or bachelor's degree or higher. To develop an asset index, we used polychoric factor analysis on data on common assets and amenities found in the mother's household, and grouped the resulting factor scores into quartiles. 33 Assets and amenities included electricity, radio or cassette player, electric fan (Bangladesh and India only), television, refrigerator (Bangladesh and India only), telephone (Bangladesh and Nepal only), generator (India only), and bicycle. All models were additionally adjusted for location (Bangladesh, India, Nepal), and year of interview in three-year groups. The data were collected in a stratified, clustersampled survey, and we accounted for survey design in the analysis using a fixed effect for stratum and a random effect for cluster.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 8

Statistical analysis
We used frequencies to describe caesarean section rates by delivery location at each site. We used the Generalized Linear Latent And Mixed Models (GLLAMM) procedure in Stata 13.1, with adaptive quadrature for binary outcomes, to estimate the crude association between type of delivery facility and caesarean section. 34 35 We identified other maternal, pregnancy, and delivery characteristics potentially associated with caesarean section using existing literature, especially studies resulting from the WHO multi-country surveys, and entered these in adjusted models to explore how they modified the association between type of delivery facility and caesarean section, and their individual association with caesarean section. Some South Asian and Latin American studies have detected a strong association between maternal education and caesarean delivery. 7 36 37 We fitted models including indicator variables for each group of mothers by education and type of delivery facility to explore differences in the strength of association between caesarean delivery and private facility by mother's education. To account for the sampling procedure used in rural Nepal, models were adjusted using pweights (probability of selection within cluster); these weights were rescaled to reflect the total number of institutional deliveries.

Ethical approval
The trials that provided data for this study received ethical approval from the following committees: the

DISCUSSION
Our analysis of data from over 45,000 births confirms the findings of other studies identifying differences in caesarean rates between public and private facilities, and suggests that, even in underserved areas in South Asia, caesareans without medical indication are of concern. In three of four locations, rates of caesarean section were higher in private/charitable facilities than in public facilities. The findings from Bangladesh are particularly noteworthy as they show much greater odds of caesarean section in private facilities, concurring with previous analyses. 37 This was also the only location where serious complications in pregnancy and delivery were not associated with caesarean delivery, suggesting that obstetric surgery was performed over and above clinical need.
Our findings confirm the results of earlier studies of the prevalence of caesarean delivery in South Asia, and indicate that high rates can be found in underserved rural areas. In India, a recent analysis of 2010-11 Annual Health Survey (AHS) data from 284 districts in nine States, including Jharkhand, found that the median caesarean section rate in the private sector was 28%, compared with 5% in the public sector. 38 There appear to be strong financial incentives for surgical procedures in the private sector. 39 In a rural, largely indigenous part of eastern India where more mothers delivered in private/charitable facilities than in public facilities, more caesarean sections were performed in the public sector. In this particular setting, our field experience suggests that only women with serious complications would go to a facility and have a caesarean section, especially as JSY was not yet available in the study areas at the time of data collection. Such women are likely to have experienced multiple referrals from either ill-equipped public or private facilities not wanting to take the risk of admission. It is however possible that preference for the private sector has changed in Jharkhand since the advent of JSY, and further  18 Our data from rural, socioeconomically disadvantaged communities in three districts of Bangladesh confirm these population-level findings, but also suggest that high caesarean section rates in private facilities are not merely an issue for wealthy urban mothers. Although some research from Bangladesh suggests that mothers may have a preference for caesarean delivery because of fear of labour pain or a desire to select an auspicious date for the birth, other studies also highlight women's fears of caesarean section and their distrust of health providers who recommend them, mainly because of the high costs associated with the procedure. 37 40 41 A qualitative study involving twenty women who had experienced obstetric complications in Matlab in 2008-9 found that most of the fourteen women who had undergone caesareans had spent over Tk 14,999 (US$ 217) on the procedure, which was approximately one-third of GDP per capita at the time. 41 The lack of association between caesareans and complications in pregnancy or delivery and multiple births, coupled with the high financial incentives given to providers for performing caesarean sections, further suggest that obstetric surgery is being used over and above clinical need. A possible explanation for the significant interactive associations between maternal education and caesarean delivery in both public and private facilities in Bangladesh is that well-educated women may be delivering in more expensive or highly rated institutions, which may in turn be more likely to do caesarean sections for financial reasons. Our own experience also suggests that such facilities have additional incentives to encourage caesareans in order to allow junior doctors to practice their surgical skills. In our sample of institutional deliveries from rural Nepal, 5% of all births were by caesarean, which is similar to the national average: the 2011 Nepal DHS found an overall national caesarean section rate of 4.6%, with sections more commonly performed for births to highly educated mothers (13%) and mothers in the highest wealth quintile (14%). 26 A 2010 analysis of DHS data examining the role of the private sector in maternity care in 16 countries found evidence of a trend towards privatisation in delivery care between the 1990s and mid-2000s, but with strong differences between countries, which might reflect the heterogeneous nature of this sector both between and within countries. 42 This DHS study highlighted the need for more context-specific data on the nature of the private sector in low and middle-income countries, and its role in maternity care. Further research might focus on understanding the motivations and experiences of women undergoing caesarean sections in private facilities in South Asian settings, pathways for switching between public and private sectors in the event of obstetric complications, a more comprehensive tally of the financial incentives (official or non-official) that motivate private providers to carry out caesareans in each setting, and the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.
In our study, locations with higher prevalence of caesarean deliveries also showed a positive educational gradient even after adjusting for wealth measured by household assets. Moreover, there was an interactive association between education and type of facility, with highly educated women particularly likely to receive caesarean deliveries in private facilities in Bangladesh and informal settlements in urban India. Literature exploring the determinants of caesarean delivery emphasises that multiple influences drive a woman's decision to deliver by caesarean. Profit considerations at the facility level may prompt some providers to urge women to receive unnecessary procedures, while women themselves may prefer caesareans for cultural reasons, fear of painful deliveries, or because they believe it to be safer. 36 43-45 Within the client-provider interaction, providers may be more likely to acquiesce to a request for caesarean from a highly educated woman; and such women may be more likely to accept advice from a provider. 19 46 Our findings lend support to the hypothesis that, while provider-level factors are probably partially responsible for the rapid increase in caesarean deliveries, it is also necessary to consider women's own preferences and decision-making processes and how they are shaped by social and cultural factors.

Strengths and limitations
The strengths of the study were its large sample size and focus on underserved communities, which are a priority for public health interventions in South Asia. It had five main limitations. It was not a nationally or sub-nationally representative study, and districts or clusters were sampled purposively from previous cRCTs. This limits the generalisability of our findings to geographical settings outside the study areas. The study was cross-sectional, and therefore only able to suggest associations rather than causal relationships. The data also did not include key predictors of caesarean section such as breech presentation, and whether the previous delivery was by caesarean, which limits the completeness of our analysis of determinants. In addition, levels of serious complications in pregnancy and delivery varied considerably between locations and were often higher than expected, suggesting potential overreporting and limited reliability as an indicator of complications. Finally, some variables had small denominators, and others had high levels of missing data (for example, maternal age in rural India).

Conclusions
Our study found that delivering in a private health facility was associated with an increased odds of caesarean section in three of four South Asian locations, and that the associations persisted after adjustments for maternal, pregnancy and delivery characteristics, and year of delivery. We also found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and informal settlements in urban India (private facilities only). These results lend support to the hypothesis that increased caesarean section rates in these three South Asian countries may in part be driven by the private sector, but also suggest that, in some settings such as Bangladesh and informal settlements in urban India, demand from more educated mothers may play a part. These findings urgently call for greater monitoring of the role of private providers in maternity care, together with action to mitigate the negative consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes. It remains the explicit responsibility of the medical profession and associated regulatory bodies to ensure that surgical intervention in delivery care is undertaken exclusively in the interests of the mother and unborn child, and not those of care    * Nepal numbers weighted using women's probability of selection within each cluster, scaled to total number of institutional births.
** At least one visit with skilled provider ***Includes: symptoms of eclampsia (fits, seizures, convulsions, or unconsciousness during pregnancy or delivery); reduced or no fetal movement; labour lasting more than 24 hours. ****For the Nepal data, two respondents with bachelor's degrees (2 respondent total, 1 delivering in institution) were combined with respondents with secondary education.

24
All analyses additionally adjusted for survey design using fixed effect of stratum and random effect of cluster.

Study design 4
Present key elements of study design early in the paper 1, 5-6 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection 5 and Table 1 Participants 6 (a) Cohort study-Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up Case-control study-Give the eligibility criteria, and the sources and methods of case ascertainment and control selection. Give the rationale for the choice of cases and controls Cross-sectional study-Give the eligibility criteria, and the sources and methods of selection of participants 6 (b) Cohort study-For matched studies, give matching criteria and number of exposed and unexposed Cross-sectional study-Report numbers of outcome events or summary measures Table 3 Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (eg, 95% confidence interval). Make clear which confounders were adjusted for and why they were included

"Prevalence and determinants of caesarean section in private and public health facilities in underserved South Asian communities: cross-sectional analysis of data from Bangladesh, India, and Nepal"
Supplemental

Health facility characteristics
Public health facility 68. . We found significant interactive associations between maternal education and private facility delivery in two of four locations (p=0.025 in rural Bangladesh and p<0.001 in informal settlements in urban India), with highly educated women particularly likely to deliver by caesarean in private facilities.

Conclusions:
Our results lend support to the hypothesis that increased caesarean section rates in these South Asian countries may be driven in part by the private sector. They also suggest that preferences for caesarean delivery may be stronger among highly educated women, and that individual-and provider-level factors interact in driving caesarean rates upwards. Increasing rates of caesarean section in the private sector require close monitoring and action.

Article focus
• Caesarean sections are on the rise in South Asia. Research suggests that private health facilities may be contributing to this by conducting caesareans over and above clinical need.
• Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to want caesareans.
• This study examined the prevalence and determinants of caesarean sections in private and charitable facilities compared with public health facilities for over 45,000 births in underserved communities of rural Nepal, rural Bangladesh, rural India, and informal settlements in urban Indiaurban India.

Key messages
• Caesarean sections were more common in private than in public health facilities in three of four study locations (urban India, rural Nepal, and rural Bangladesh, and informal settlements in urban India).
• We found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and in informal Indian urban settlements urban India (private facilities only).
• Future research should monitor caesarean section rates in the private sector in South Asia and subsequent maternal and neonatal health outcomes

Strengths and limitations
• This study had a large sample size and focused on underserved communities, which are a priority for public health interventions in South Asia.
• Our data were not nationally or sub-nationally representative, which limits the generalisability of our findings. Other limitations included small denominators and missing data for some variables. We did not have data on some known predictors of caesarean section, which would have enhanced the completeness of our determinants analysis.

INTRODUCTION
Access to comprehensive emergency obstetric care, including caesarean section, is key to preventing the estimated 287,000 maternal and 2.9 million neonatal deaths that occur worldwide every year. 1 2 Although debate continues about how to quantify the need for life-saving obstetric surgery, a 1985 World Health Organization (WHO) report suggested that the optimal population range for caesarean section rates is between 5% and 15%, and this endures as a reference. 3 4 Caesarean section rates are increasing worldwide, albeit unequally: a recent analysis of Demographic and Health Survey (DHS) data in 26 South Asian and sub-Saharan African countries found that rates were highest among the 'urban rich' in all countries, and lowest among the 'rural poor' in 18. 5 In all countries, fewer than 5% of mothers in the poorest wealth quintile delivered by caesarean. 6 Caesarean sections conducted without clinical need can have adverse consequences for mothers and children. A 2008 WHO survey of 373 facilities across 24 countries found that unnecessary caesareans were associated with an increased risk of maternal mortality and serious outcomes for mothers and newborn infants, compared with spontaneous vaginal delivery. 7 Recent ecological analyses also highlighted strong associations between caesarean delivery and increased neonatal mortality in countries with low and medium caesarean section rates. 8 Unnecessary caesareans lead to considerable costs for families and health systems: an estimated 6.2 million unnecessary procedures were performed in 2008, costing approximately US$ 2.32 billion. 9 Several South Asian countries have recorded substantial increases in caesarean section rates over the past decade. In Bangladesh, rates rose from 2% (2000) to 17% (2011); in India, from 3% (1992) to 11 % (2006); and in Nepal, from 1% (2000) to 5% (2011).  Several studies in these countries have raised concerns about high caesarean rates in private facilities, and a recent DHS analysis speculated that national increases in caesarean section rates in South Asian countries could be driven in part by higher rates among deliveries in private sector facilities . 6 16-18 Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to desire caesareans. 19 20 There are few large, recent community-based studies from South Asia quantifying differences in caesarean section rates between public and private facilities. Such studies are necessary in order to examine whether increases in caesareans in these settings are indeed likely to be driven by the private sector, demand from wealthier and more educated mothers, or a combination of the two. We conducted a cross-sectional analysis of data from Bangladesh, India, and Nepal to explore the prevalence and determinants of caesarean section delivery by type of facility and location, and to examine the

Study populations
We used data collected through vital events surveillance systems established during four clusterrandomised controlled trials conducted between 2005 and 2011. The trials were done with communities that can be considered socio-economically disadvantaged: in Bangladesh and Nepal, they took place in four rural, underserved districts (Bogra, Maulvibazaar, and Faridpur in Bangladesh, and the Terai district of Dhanusha, Nepal); in rural India, most participants were from Scheduled Tribes in two states of eastern India (Jharkhand and Odisha); in urban India, data came from informal settlements (slums) slum communities in Mumbai. Table 1 describes the characteristics of each study and its population, including the background maternal mortality ratio and types of facilities present in the study areas. The original trials were designed to evaluate the impact of participatory women's groups on maternal and neonatal health outcomes. [21][22][23][24][25] We used data from the control areas of these trials only, because the women's group intervention led to changes in mortality and practices in several locations. 26 Health system contexts Social Health Activists (ASHAs) also receive 600 Indian Rupees ($9.6) for identifying pregnant women and helping them get to a health facility. 29 Although the rural Indian data included in this study were

Data collection
All data were collected using surveillance systems to monitor births and deaths prospectively. [21][22][23][24][25] In all study locations, a female community-based key informant reported births and deaths in her area, which covered a population ranging from 250 to 350. A trained interviewer then verified these reports and paid the informant an incentive for each correct identification. In Bangladesh, rural and urban India, the interviewer administered a structured questionnaire to all eligible mothers around six weeks after delivery; in Nepal, all births in the study area were registered, and interviews were conducted on all births in small clusters and on a random sample of 10 births per month in the larger clusters. In each study location, mothers were interviewed using a questionnaire to collect information about events in the antenatal, delivery and postnatal periods. In each trial and in this subsequent study, participants were women of reproductive age (15-49) who delivered in the study area during the data collection period, and who consented to be interviewed six weeks after delivery, and their infants.

Study sample
The initial sample for this analysis included 46,393 births in Mumbai, rural India, rural Bangladesh, or rural Nepal, of whom 17,565 (38%) were delivered in health care facilities. We excluded the 2348 births occurring outside Mumbai, because we collected limited information on delivery location for these.

Measures used
The primary outcome in these analyses was caesarean section delivery, identified by self-report from the mother or another household member around six weeks after giving birth. The main covariate of interest was the type of delivery facility, coded as public (e.g. district hospitals), private (individual private clinics or hospitals), or NGO/charitable (e.g. Christian missionary hospitals in rural India; BRAC clinics in Bangladesh). We grouped private and NGO/charitable facilities together due to the small number of births in NGO/charitable facilities (6% of institutional deliveries in Bangladesh, 7% in rural India, and 0 in informal settlements in urban India and rural Nepal; data on request). While we refer to all results from Mumbai as being from 'informal settlements', is worth noting that the health facilities used by women in informal settlements for delivery are not always in the settlements themselves: women often travel to other parts of the city to seek care with a recommended or preferred provider, usually private. Additional covariates in the models included measures of the characteristics of the pregnancy and delivery, maternal socio-demographic characteristics, and the location in which the delivery occurred. We created an indicator variable to identify women experiencing serious problems during pregnancy or delivery, with women reporting symptoms of pre-eclampsia (blurred vision or swelling of face and hands), symptoms of eclampsia (fits or seizures during pregnancy or delivery), haemorrhage during delivery, or labour lasting more than 24 hours considered to have serious complications. Other characteristics of the pregnancy and delivery considered in the analysis included full utilisation of antenatal care, number of prior pregnancies, and multiple pregnancy. Full utilization of antenatal care was defined as four or more visits to antenatal care, with at least one visit to a skilled provider. Number of pregnancies was entered as an ordered categorical variable, with categories of one (first), two, three, or four or more pregnancies.
Maternal socio-demographic characteristics included in the models were: mother's age at delivery, her educational attainment, and household assets. Mother's age was entered into the model as a categorical measure in 10-year groups. Educational attainment was entered as a categorical variable using the following categories: no formal education, primary education, secondary education, or bachelor's degree or higher. To develop an asset index, we used polychoric factor analysis on data on common assets and amenities found in the mother's household, and grouped the resulting factor scores into quartiles. 33 Assets and amenities included electricity, radio or cassette player, electric fan (Bangladesh and India only), television, refrigerator (Bangladesh and India only), telephone (Bangladesh and Nepal only), generator (India only), and bicycle. All models were additionally adjusted for location (Bangladesh, India, Nepal), and year of interview in three-year groups. The data were collected in a stratified, clustersampled survey, and we accounted for survey design in the analysis using a fixed effect for stratum and a random effect for cluster.

Statistical analysis
We used frequencies to describe caesarean section rates by delivery location at each site. We used the Generalized Linear Latent And Mixed Models (GLLAMM) procedure in Stata 13.1, with adaptive quadrature for binary outcomes, to estimate the crude association between type of delivery facility and caesarean section. 34 35 We identified other maternal, pregnancy, and delivery characteristics potentially associated with caesarean section using existing literature, especially studies resulting from the WHO multi-country surveys, and entered these in adjusted models to explore how they modified the association between type of delivery facility and caesarean section, and their individual association with caesarean section. Some South Asian and Latin American studies have detected a strong association between maternal education and caesarean delivery. 7 36 37 We fitted models including indicator variables for each group of mothers by education and type of delivery facility to explore differences in the strength of association between caesarean delivery and private facility by mother's education. To account for the sampling procedure used in rural Nepal, models were adjusted using pweights (probability of selection within cluster); these weights were rescaled to reflect the total number of institutional deliveries.

Ethical approval
The trials that provided data for this study received ethical approval from the following committees: the

DISCUSSION
Our analysis of data from over 45,000 births confirms the findings of other studies identifying differences in caesarean rates between public and private facilities, and suggests that, even in underserved areas in South Asia, caesareans without medical indication are of concern. In three of four locations, rates of caesarean section were higher in private/charitable facilities than in public facilities. The findings from Bangladesh are particularly noteworthy as they show much greater odds of caesarean section in private facilities, concurring with previous analyses. 37 This was also the only location where serious complications in pregnancy and delivery were not associated with caesarean delivery, suggesting that obstetric surgery was performed over and above clinical need.
Our findings confirm the results of earlier studies of the prevalence of caesarean delivery in South Asia, and indicate that high rates can be found in underserved rural areas. In India, a recent analysis of 2010-11 Annual Health Survey (AHS) data from 284 districts in nine States, including Jharkhand, found that the median caesarean section rate in the private sector was 28%, compared with 5% in the public sector. 38 There appear to be strong financial incentives for surgical procedures in the private sector. 39 18 Our data from rural, socioeconomically disadvantaged communities in three districts of Bangladesh confirm these population-level findings, but also suggest that high caesarean section rates in private facilities are not merely an issue for wealthy urban mothers. Although some research from Bangladesh suggests that mothers may have a preference for caesarean delivery because of fear of labour pain or a desire to select an auspicious date for the birth, other studies also highlight women's fears of caesarean section and their distrust of health providers who recommend them, mainly because of the high costs associated with the procedure. 37 40 41 A qualitative study involving twenty women who had experienced obstetric complications in Matlab in 2008-9 found that most of the fourteen women who had undergone caesareans had spent over Tk 14,999 (US$ 217) on the procedure, which was approximately one-third of GDP per capita at the time. 41 The lack of association between caesareans and complications in pregnancy or delivery and multiple births, coupled with the high financial incentives given to providers for performing caesarean sections and the requirement for junior doctors to 'practise' their surgical skills, further suggest that obstetric surgery is being used over and above clinical need. A possible explanation for the significant interactive associations between maternal education and caesarean delivery in both public and private facilities in Bangladesh is that well-educated women may be delivering in more expensive or highly rated institutions, which may in turn be more likely to do caesarean sections for financial reasons and if they act as training centres for junior doctors. Our own experience also suggests that such facilities have  In our sample of institutional deliveries from rural Nepal, 516% of all facility births were by caesarean, which is similar to higher than the national average: the 2011 Nepal DHS found an overall national caesarean section rate of 4.65%, with sections more commonly performed for births to highly educated mothers (13%) and mothers in the highest wealth quintile (14%). 26 It is possible that women with complications are more likely to deliver in facilities, and also that, as in Bangladesh, private providers are motivated by financial incentives to conduct caesareans more frequently than strictly necessary. Women who can afford to seek care in private facilities may also be more willing or able to pay for caesarean sections, and providers may conduct more of them to increase their income.
A 2010 analysis of DHS data examining the role of the private sector in maternity care in 16 countries found evidence of a trend towards privatisation in delivery care between the 1990s and mid-2000s, but with strong differences between countries, which might reflect the heterogeneous nature of this sector both between and within countries. 42 This DHS study highlighted the need for more context-specific data on the nature of the private sector in low and middle-income countries, and its role in maternity care. Further research might focus on understanding the motivations and experiences of women undergoing caesarean sections in private facilities in South Asian settings, pathways for switching between public and private sectors in the event of obstetric complications, a more comprehensive tally of the financial incentives (official or non-official) that motivate private providers to carry out caesareans in each setting, and the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.
In our study, locations with higher prevalence of caesarean deliveries also showed a positive educational gradient even after adjusting for wealth measured by household assets. Moreover, there was an interactive association between education and type of facility, with highly educated women particularly likely to receive caesarean deliveries in private facilities in Bangladesh and informal settlements in urban Indiaurban India. Literature exploring the determinants of caesarean delivery emphasises that multiple influences drive a woman's decision to deliver by caesarean. Profit considerations at the facility level may prompt some providers to urge women to receive unnecessary procedures, while women themselves may prefer caesareans for cultural reasons, fear of painful deliveries, or because they believe it to be safer. 36 43-45 Within the client-provider interaction, providers may be more likely to acquiesce to a request for caesarean from a highly educated woman; and such women may be more likely to accept advice from a provider. 19 46 Our findings lend support to the hypothesis that, while provider-level factors are probably partially responsible for the rapid increase in caesarean deliveries, it is also necessary to  13 consider women's own preferences and decision-making processes and how they are shaped by social and cultural factors.

Strengths and limitations
The strengths of the study were its large sample size and focus on underserved communities, which are a priority for public health interventions in South Asia. It had five main limitations. It was not a nationally or sub-nationally representative study, and districts or clusters were sampled purposively from previous cRCTs. This limits the generalisability of our findings to geographical settings outside the study areas. The study was cross-sectional, and therefore only able to suggest associations rather than causal relationships. The data also did not include key predictors of caesarean section such as breech presentation, and whether the previous delivery was by caesarean, which limits the completeness of our analysis of determinants. In addition, levels of serious complications in pregnancy and delivery varied considerably between locations and were often higher than expected, suggesting potential overreporting and limited reliability as an indicator of complications. Finally, some variables had small denominators, and others had high levels of missing data (for example, maternal age in rural India).

Conclusions
Our study found that delivering in a private health facility was associated with an increased odds of caesarean section in three of four South Asian locations, and that the associations persisted after adjustments for maternal, pregnancy and delivery characteristics, and year of delivery. We also found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and informal settlements in urban India urban India (private facilities only). These results lend support to the hypothesis that increased caesarean section rates in these three South Asian countries may in part be driven by the private sector, but also suggest that, in some settings such as Bangladesh and informal settlements in urban Indiaurban India, demand from more educated mothers may play a part. These findings urgently call for greater monitoring of the role of private providers in maternity care, together with action to mitigate the negative consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes. It remains the explicit responsibility of the medical profession and associated regulatory bodies to ensure that surgical intervention in delivery care is undertaken exclusively in the interests of the mother and unborn child,

Conclusions:
Our results lend support to the hypothesis that increased caesarean section rates in these South Asian countries may be driven in part by the private sector. They also suggest that preferences for caesarean delivery may be higher among highly educated women, and that individual-and provider-level factors interact in driving caesarean rates higher. Rates of caesarean section in the private sector, and their maternal and neonatal health outcomes, require close monitoring. • Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to want caesareans.
• This study examined the prevalence and determinants of caesarean sections in private and charitable facilities compared with public health facilities for over 45,000 births in underserved communities of rural Nepal, rural Bangladesh, rural and urban India.

Key messages
• Caesarean sections were more common in private than in public health facilities in three of four study locations (urban India, rural Nepal, and rural Bangladesh).
• We found significant interactive associations between maternal education and private facility delivery on caesarean in Bangladesh (both private and public facilities) and urban India (private facilities only).
• Future research should monitor caesarean section rates in the private sector in South Asia and subsequent maternal and neonatal health outcomes

Strengths and limitations
• This study had a large sample size and focused on underserved communities, which are a priority for public health interventions in South Asia.
• Our data were not nationally representative, which limits the generalisability of our findings.
Other limitations included small denominators and missing data for some variables. We did not have data on some known predictors of caesarean section, which would have enhanced the completeness of our determinants analysis. Although debate continues about how to quantify the need for life-saving obstetric surgery, a 1985 World Health Organization (WHO) report suggested that the optimal population range for caesarean section rates is between 5% and 15%, and this endures as a reference. 3 (2000) to 5% (2011).  Several studies in these countries have raised concerns about high caesarean rates in private facilities, and a recent DHS analysis speculated that national increases in caesarean section rates in South Asian countries could be driven in part by higher rates among deliveries in private sector facilities . 6 16-18 Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to desire caesareans. 19 20 There are few large, recent community-based studies from South Asia quantifying differences in caesarean section rates between public and private facilities. Such studies are necessary in order to examine whether increases in caesareans in these settings are indeed likely to be driven by the private sector, demand from wealthier and more educated mothers, or a combination of the two. We conducted a cross-sectional analysis of data from Bangladesh, India, and Nepal to explore the prevalence and determinants of caesarean section delivery by type of facility and location, and to examine the interactions between maternal education and place of delivery on prevalence of caesarean section.

Study populations
We used data collected through vital events surveillance systems established during four clusterrandomised controlled trials conducted between 2005 and 2011. The trials were done with communities that can be considered socio-economically disadvantaged: in Bangladesh and Nepal, they took place in four rural, underserved districts (Bogra, Maulvibazaar, and Faridpur in Bangladesh, and the Terai district of Dhanusha, Nepal); in rural India, most participants were from Scheduled Tribes in two states of eastern India (Jharkhand and Odisha); in urban India, data came from slum communities in Mumbai. Table 1 describes the characteristics of each study and its population, including the background maternal mortality ratio and types of facilities present in the study areas. The original trials were designed to evaluate the impact of participatory women's groups on maternal and neonatal health outcomes. [21][22][23][24][25] We used data from the control areas of these trials only, because the women's group intervention led to changes in mortality and practices in several locations. 26 Health system contexts Social Health Activists (ASHAs) also receive 600 Indian Rupees ($9.6) for identifying pregnant women and helping them get to a health facility. 29 Although the rural Indian data included in this study were

Data collection
All data were collected using surveillance systems to monitor births and deaths prospectively. [21][22][23][24][25] In all study locations, a female community-based key informant reported births and deaths in her area, which covered a population ranging from 250 to 350. A trained interviewer then verified these reports and paid the informant an incentive for each correct identification. In Bangladesh, rural and urban India, the interviewer administered a structured questionnaire to all eligible mothers around six weeks after delivery; in Nepal, all births in the study area were registered, and interviews were conducted on all births in small clusters and on a random sample of 10 births per month in the larger clusters. In each study location, mothers were interviewed using a questionnaire to collect information about events in the antenatal, delivery and postnatal periods. In each trial and in this subsequent study, participants were women of reproductive age (15-49) who delivered in the study area during the data collection period, and who consented to be interviewed six weeks after delivery, and their infants.

Study sample
The initial sample for this analysis included 46

Measures used
The primary outcome in these analyses was caesarean section delivery, identified by self-report from the mother or another household member around six weeks after giving birth. Maternal socio-demographic characteristics included in the models were: mother's age at delivery, her educational attainment, and household assets. Mother's age was entered into the model as a categorical measure in 10-year groups. Educational attainment was entered as a categorical variable using the following categories: no formal education, primary education, secondary education, or bachelor's degree or higher. To develop an asset index, we used polychoric factor analysis on data on common assets and amenities found in the mother's household, and grouped the resulting factor scores into quartiles. 33 Assets and amenities included electricity, radio or cassette player, electric fan (Bangladesh and India only), television, refrigerator (Bangladesh and India only), telephone (Bangladesh and Nepal only), generator (India only), and bicycle. All models were additionally adjusted for location (Bangladesh, India, Nepal), and year of interview in three-year groups. The data were collected in a stratified, clustersampled survey, and we accounted for survey design in the analysis using a fixed effect for stratum and a random effect for cluster.

Statistical analysis
We used frequencies to describe caesarean section rates by delivery location at each site. We used the Generalized Linear Latent And Mixed Models (GLLAMM) procedure in Stata 13.1, with adaptive quadrature for binary outcomes, to estimate the crude association between type of delivery facility and  34 35 We identified other maternal, pregnancy, and delivery characteristics potentially associated with caesarean section using existing literature, especially studies resulting from the WHO multi-country surveys, and entered these in adjusted models to explore how they modified the association between type of delivery facility and caesarean section, and their individual association with caesarean section. Some South Asian and Latin American studies have detected a strong association between maternal education and caesarean delivery. 7 36 37 We fitted models including indicator variables for each group of mothers by education and type of delivery facility to explore differences in the strength of association between caesarean delivery and private facility by mother's education. To account for the sampling procedure used in rural Nepal, models were adjusted using pweights (probability of selection within cluster); these weights were rescaled to reflect the total number of institutional deliveries.

Ethical approval
The trials that provided data for this study received ethical approval from the following committees: the    However, there was no apparent gradient by educational attainment among women delivering in public facilities.

DISCUSSION
Our analysis of data from over 45,000 births confirms the findings of other studies identifying differences in caesarean rates between public and private facilities, and suggests that, even in underserved areas in South Asia, caesareans without medical indication are of concern. In three of four locations, rates of caesarean section were higher in private/charitable facilities than in public facilities. The findings from Bangladesh are particularly noteworthy as they show much greater odds of caesarean section in private facilities, concurring with previous analyses. 37 This was also the only location where serious complications in pregnancy and delivery were not associated with caesarean delivery, suggesting that obstetric surgery was performed over and above clinical need.
Our findings confirm the results of earlier studies of the prevalence of caesarean delivery in South Asia, and indicate that high rates can be found in underserved rural areas. In India, a recent analysis of 2010-11 Annual Health Survey (AHS) data from 284 districts in nine States, including Jharkhand, found that the median caesarean section rate in the private sector was 28%, compared with 5% in the public sector. 38 There appear to be strong financial incentives for surgical procedures in the private sector. 39 In a rural, largely indigenous part of eastern India where more mothers delivered in private/charitable facilities than in public facilities, more caesarean sections were performed in the public sector. In this particular setting, our field experience suggests that only women with serious complications would go to a facility and have a caesarean section, especially as JSY was not yet available in the study areas at the time of data collection. Such women are likely to have experienced multiple referrals from either ill-equipped public or private facilities not wanting to take the risk of admission. It is however possible that preference for the private sector has changed in Jharkhand since the advent of JSY, and further disaggregated analyses of Annual Health Survey data would allow a more contemporary exploration of State-level variations in caesarean section rates between public and private sectors.
In Mumbai slums with a high uptake of institutional births and a 60/40 split in favour of the public sector for delivery care, we found no difference in caesarean section rates between the two sectors, which is  18 Our data from rural, socioeconomically disadvantaged communities in three districts of Bangladesh confirm these population-level findings, but also suggest that high caesarean section rates in private facilities are not merely an issue for wealthy urban mothers. Although some research from Bangladesh suggests that mothers may have a preference for caesarean delivery because of fear of labour pain or a desire to select an auspicious date for the birth, other studies also highlight women's fears of caesarean section and their distrust of health providers who recommend them, mainly because of the high costs associated with the procedure. 37 40 41 A qualitative study involving twenty women who had experienced obstetric complications in Matlab in 2008-9 found that most of the fourteen women who had undergone caesareans had spent over Tk 14,999 (US$ 217) on the procedure, which was approximately one-third of GDP per capita at the time. 41 The lack of association between caesareans and complications in pregnancy or delivery and multiple births, coupled with the high financial incentives given to providers for performing caesarean sections and the requirement for junior doctors to 'practise' their surgical skills, further suggest that obstetric surgery is being used over and above clinical need. A possible explanation for the significant interactive associations between maternal education and caesarean delivery in both public and private facilities in Bangladesh is that well-educated women may be delivering in more expensive or highly rated institutions, which may in turn be more likely to do caesarean sections for financial reasons and if they act as training centres for junior doctors.
In our sample of institutional deliveries from rural Nepal, 16% of facility births were by caesarean, which is higher than the national average: the 2011 Nepal DHS found an overall national caesarean section rate of 5%, with sections more commonly performed for births to highly educated mothers (13%) and mothers in the highest wealth quintile (14%). 26 It is possible that women with complications are more likely to deliver in facilities, and also that, as in Bangladesh, private providers are motivated by financial incentives to conduct caesareans more frequently than strictly necessary. Women who can afford to seek care in private facilities may also be more willing or able to pay for caesarean sections, and providers may conduct more of them to increase their income.
A 2010 analysis of DHS data examining the role of the private sector in maternity care in 16 countries found evidence of a trend towards privatisation in delivery care between the 1990s and mid-2000s, but with strong differences between countries, which might reflect the heterogeneous nature of this sector both between and within countries. 42 This DHS study highlighted the need for more context-specific data on the nature of the private sector in low and middle-income countries, and its role in maternity care. Further research might focus on understanding the motivations and experiences of women undergoing caesarean sections in private facilities in South Asian settings, pathways for switching between public and private sectors in the event of obstetric complications, a more comprehensive tally of the financial incentives (official or non-official) that motivate private providers to carry out caesareans in each setting, and the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.
In our study, locations with higher prevalence of caesarean deliveries also showed a positive educational gradient even after adjusting for wealth measured by household assets. Moreover, there was an interactive association between education and type of facility, with highly educated women particularly likely to receive caesarean deliveries in private facilities in Bangladesh and urban India. Literature exploring the determinants of caesarean delivery emphasizes that multiple influences drive a woman's decision to deliver by caesarean. Profit considerations at the facility level may prompt some providers to urge women to receive unnecessary procedures, while women themselves may prefer caesareans for cultural reasons, fear of painful deliveries, or because they believe it to be safer. 36 43-45 Within the clientprovider interaction, providers may be more likely to acquiesce to a request for caesarean from a highly educated woman; and such women may be more likely to accept advice from a provider. 19 46 Our findings lend support to the hypothesis that, while provider-level factors are probably partially responsible for the rapid increase in caesarean deliveries, it is also necessary to consider women's own preferences and decision-making processes and how they are shaped by social and cultural factors.

Strengths and limitations
The strengths of the study were its large sample size and focus on underserved communities, which are a priority for public health interventions in South Asia. It had five main limitations. It was not a nationally representative study, and districts or clusters were sampled purposively from previous cRCTs. This limits the generalisability of our findings to geographical settings outside the study areas. The study was crosssectional, and therefore only able to suggest associations rather than causal relationships. The data also did not include key predictors of caesarean section such as breech presentation, and whether the previous delivery was by caesarean, which limits the completeness of our analysis of determinants. In addition, levels of serious complications in pregnancy and delivery varied considerably between locations and were often higher than expected, suggesting potential over-reporting and limited reliability as an indicator of complications. Finally, some variables had small denominators, and others had high levels of missing data (for example, maternal age in rural India).

Conclusions
Our study found that delivering in a private health facility was associated with an increased odds of caesarean section in three of four South Asian locations, and that the associations persisted after adjustments for maternal, pregnancy and delivery characteristics, and year of delivery. We also found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and urban India (private facilities only). These results lend support to the hypothesis that increased caesarean section rates in these three South Asian countries may in part be driven by the private sector, but also suggest that, in some settings such as Bangladesh and urban India, demand from more educated mothers may play a part. These findings call for greater, local understanding of the role of private providers in maternity care, together with careful examination of the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.

Competing interests
None to declare     * Nepal numbers weighted using women's probability of selection within each cluster, scaled to total number of institutional births.

Contributors
** At least one visit with skilled provider ***Includes: symptoms of eclampsia (fits, seizures, convulsions, or unconsciousness during pregnancy or delivery); reduced or no fetal movement; labour lasting more than 24 hours. ****For the Nepal data, two respondents with bachelor's degrees (2 respondent total, 1 delivering in institution) were combined with respondents with secondary education.
All analyses additionally adjusted for survey design using fixed effect of stratum and random effect of cluster.  All results adjusted for number of antenatal care visits, parity, medical indication for caesarean, multiple birth, maternal age (10-yr group), household assets, stratum, and cluster (random effect). * Nepal numbers weighted using women's probability of selection within each cluster, scaled to total number of institutional births. ** p-value from -2 log-likelihood test comparing nested models with and without interaction terms

Conclusions:
Our results lend support to the hypothesis that increased caesarean section rates in these South Asian countries may be driven in part by the private sector. They also suggest that preferences for caesarean delivery may be higher among highly educated women, and that individual-and provider-level factors interact in driving caesarean rates higher. Rates of caesarean section in the private sector, and their maternal and neonatal health outcomes, require close monitoring. • Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to want caesareans.
• This study examined the prevalence and determinants of caesarean sections in private and charitable facilities compared with public health facilities for over 45,000 births in underserved communities of rural Nepal, rural Bangladesh, rural and urban India.

Key messages
• Caesarean sections were more common in private than in public health facilities in three of four study locations (urban India, rural Nepal, and rural Bangladesh).
• We found significant interactive associations between maternal education and private facility caesarean delivery on caesarean in Bangladesh (both private and public facilities) and urban India (private facilities only).
• Future research should monitor caesarean section rates in the private sector in South Asia and subsequent maternal and neonatal health outcomes

Strengths and limitations
• This study had a large sample size and focused on underserved communities, which are a priority for public health interventions in South Asia.
• Our data were not nationally representative, which limits the generalisability of our findings.
Other limitations included small denominators and missing data for some variables. We did not have data on some known predictors of caesarean section, which would have enhanced the completeness of our determinants analysis. Although debate continues about how to quantify the need for life-saving obstetric surgery, a 1985 World Health Organization (WHO) report suggested that the optimal population range for caesarean section rates is between 5% and 15%, and this endures as a reference. 3 (2000) to 5% (2011).  Several studies in these countries have raised concerns about high caesarean rates in private facilities, and a recent DHS analysis speculated that national increases in caesarean section rates in South Asian countries could be driven in part by higher rates among deliveries in private sector facilities . 6 16-18 Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to desire caesareans. 19 20 There are few large, recent community-based studies from South Asia quantifying differences in caesarean section rates between public and private facilities. Such studies are necessary in order to examine whether increases in caesareans in these settings are indeed likely to be driven by the private sector, demand from wealthier and more educated mothers, or a combination of the two. We conducted a cross-sectional analysis of data from Bangladesh, India, and Nepal to explore the prevalence and determinants of caesarean section delivery by type of facility and location, and to examine the

Study populations
We used data collected through vital events surveillance systems established during four cluster-  Table 1 describes the characteristics of each study and its population, including the background maternal mortality ratio and types of facilities present in the study areas. The original trials were designed to evaluate the impact of participatory women's groups on maternal and neonatal health outcomes. [21][22][23][24][25] We used data from the control areas of these trials only, because the women's group intervention led to changes in mortality and practices in several locations. 26 Health system contexts

Data collection
All data were collected using surveillance systems to monitor births and deaths prospectively. [21][22][23][24][25] In all study locations, a female community-based key informant reported births and deaths in her area, which covered a population ranging from 250 to 350. A trained interviewer then verified these reports and paid the informant an incentive for each correct identification. In Bangladesh, rural and urban India, the interviewer administered a structured questionnaire to all eligible mothers around six weeks after delivery; in Nepal, all births in the study area were registered, and interviews were conducted on all births in small clusters and on a random sample of 10 births per month in the larger clusters. In each study location, mothers were interviewed using a questionnaire to collect information about events in the antenatal, delivery and postnatal periods. In each trial and in this subsequent study, participants were women of reproductive age  who delivered in the study area during the data collection period, and who consented to be interviewed six weeks after delivery, and their infants.

Study sample
The initial sample for this analysis included 46

Measures used
The primary outcome in these analyses was caesarean section delivery, identified by self-report from the mother or another household member around six weeks after giving birth. Maternal socio-demographic characteristics included in the models were: mother's age at delivery, her educational attainment, and household assets. Mother's age was entered into the model as a categorical measure in 10-year groups. Educational attainment was entered as a categorical variable using the following categories: no formal education, primary education, secondary education, or bachelor's degree or higher. To develop an asset index, we used polychoric factor analysis on data on common assets and amenities found in the mother's household, and grouped the resulting factor scores into quartiles. 33 Assets and amenities included electricity, radio or cassette player, electric fan (Bangladesh and India only), television, refrigerator (Bangladesh and India only), telephone (Bangladesh and Nepal only), generator (India only), and bicycle. All models were additionally adjusted for location (Bangladesh, India, Nepal), and year of interview in three-year groups. The data were collected in a stratified, clustersampled survey, and we accounted for survey design in the analysis using a fixed effect for stratum and a random effect for cluster.

Statistical analysis
We used frequencies to describe caesarean section rates by delivery location at each site. We used the Generalized Linear Latent And Mixed Models (GLLAMM) procedure in Stata 13.1, with adaptive  34 35 We identified other maternal, pregnancy, and delivery characteristics potentially associated with caesarean section using existing literature, especially studies resulting from the WHO multi-country surveys, and entered these in adjusted models to explore how they modified the association between type of delivery facility and caesarean section, and their individual association with caesarean section. Some South Asian and Latin American studies have detected a strong association between maternal education and caesarean delivery. 7 36 37 We fitted models including indicator variables for each group of mothers by education and type of delivery facility to explore differences in the strength of association between caesarean delivery and private facility by mother's education. To account for the sampling procedure used in rural Nepal, models were adjusted using pweights (probability of selection within cluster); these weights were rescaled to reflect the total number of institutional deliveries.

Ethical approval
The trials that provided data for this study received ethical approval from the following committees: the    However, there was no apparent gradient by educational attainment among women delivering in public facilities.

DISCUSSION
Our analysis of data from over 45,000 births confirms the findings of other studies identifying differences in caesarean rates between public and private facilities, and suggests that, even in underserved areas in South Asia, caesareans without medical indication are of concern. In three of four locations, rates of caesarean section were higher in private/charitable facilities than in public facilities. The findings from Bangladesh are particularly noteworthy as they show much greater odds of caesarean section in private facilities, concurring with previous analyses. 37 This was also the only location where serious complications in pregnancy and delivery were not associated with caesarean delivery, suggesting that obstetric surgery was performed over and above clinical need.
Our findings confirm the results of earlier studies of the prevalence of caesarean delivery in South Asia, and indicate that high rates can be found in underserved rural areas. In India, a recent analysis of 2010-11 Annual Health Survey (AHS) data from 284 districts in nine States, including Jharkhand, found that the median caesarean section rate in the private sector was 28%, compared with 5% in the public sector. 38 There appear to be strong financial incentives for surgical procedures in the private sector. 39 In a rural, largely indigenous part of eastern India where more mothers delivered in private/charitable facilities than in public facilities, more caesarean sections were performed in the public sector. In this particular setting, our field experience suggests that only women with serious complications would go to a facility and have a caesarean section, especially as JSY was not yet available in the study areas at the time of data collection. Such women are likely to have experienced multiple referrals from either ill-equipped public or private facilities not wanting to take the risk of admission. It is however possible that preference for the private sector has changed in Jharkhand since the advent of JSY, and further disaggregated analyses of Annual Health Survey data would allow a more contemporary exploration of State-level variations in caesarean section rates between public and private sectors.
In Mumbai slums with a high uptake of institutional births and a 60/40 split in favour of the public sector for delivery care, we found no difference in caesarean section rates between the two sectors, which is providers who recommend them, mainly because of the high costs associated with the procedure. 37 40 41 A qualitative study involving twenty women who had experienced obstetric complications in Matlab in 2008-9 found that most of the fourteen women who had undergone caesareans had spent over Tk 14,999 (US$ 217) on the procedure, which was approximately one-third of GDP per capita at the time. 41 The lack of association between caesareans and complications in pregnancy or delivery and multiple births, coupled with the high financial incentives given to providers for performing caesarean sections and the requirement for junior doctors to 'practise' their surgical skills, further suggest that obstetric surgery is being used over and above clinical need. A possible explanation for the significant interactive associations between maternal education and caesarean delivery in both public and private facilities in Bangladesh is that well-educated women may be delivering in more expensive or highly rated institutions, which may in turn be more likely to do caesarean sections for financial reasons and if they act as training centres for junior doctors.
In our sample of institutional deliveries from rural Nepal, 16% of facility births were by caesarean, which is higher than the national average: the 2011 Nepal DHS found an overall national caesarean section rate of 5%, with sections more commonly performed for births to highly educated mothers (13%) and mothers in the highest wealth quintile (14%). 26 It is possible that women with complications are more likely to deliver in facilities, and also that, as in Bangladesh, private providers are motivated by financial incentives to conduct caesareans more frequently than strictly necessary. Women who can afford to seek care in private facilities may also be more willing or able to pay for caesarean sections, and providers may conduct more of them to increase their income.
A 2010 analysis of DHS data examining the role of the private sector in maternity care in 16 countries found evidence of a trend towards privatisation in delivery care between the 1990s and mid-2000s, but with strong differences between countries, which might reflect the heterogeneous nature of this sector both between and within countries. 42 This DHS study highlighted the need for more context-specific data on the nature of the private sector in low and middle-income countries, and its role in maternity care. Further research might focus on understanding the motivations and experiences of women undergoing caesarean sections in private facilities in South Asian settings, pathways for switching between public and private sectors in the event of obstetric complications, a more comprehensive tally of the financial incentives (official or non-official) that motivate private providers to carry out caesareans in each setting, and the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.
In our study, locations with higher prevalence of caesarean deliveries also showed a positive educational gradient even after adjusting for wealth measured by household assets. Moreover, there was an interactive association between education and type of facility, with highly educated women particularly likely to receive caesarean deliveries in private facilities in Bangladesh and urban India. Literature exploring the determinants of caesarean delivery emphasizses that multiple influences drive a woman's decision to deliver by caesarean. Profit considerations at the facility level may prompt some providers to urge women to receive unnecessary procedures, while women themselves may prefer caesareans for cultural reasons, fear of painful deliveries, or because they believe it to be safer. 36 43-45 Within the clientprovider interaction, providers may be more likely to acquiesce to a request for caesarean from a highly educated woman; and such women may be more likely to accept advice from a provider. 19 46 Our findings lend support to the hypothesis that, while provider-level factors are probably partially responsible for the rapid increase in caesarean deliveries, it is also necessary to consider women's own preferences and decision-making processes and how they are shaped by social and cultural factors.

Strengths and limitations
The strengths of the study were its large sample size and focus on underserved communities, which are a priority for public health interventions in South Asia. It had five main limitations. It was not a nationally representative study, and districts or clusters were sampled purposively from previous cRCTs. This limits the generalisability of our findings to geographical settings outside the study areas. The study was crosssectional, and therefore only able to suggest associations rather than causal relationships. The data also did not include key predictors of caesarean section such as breech presentation, and whether the previous delivery was by caesarean, which limits the completeness of our analysis of determinants. In addition, levels of serious complications in pregnancy and delivery varied considerably between locations and were often higher than expected, suggesting potential over-reporting and limited reliability as an indicator of complications. Finally, some variables had small denominators, and others had high levels of missing data (for example, maternal age in rural India).

Conclusions
Our study found that delivering in a private health facility was associated with an increased odds of caesarean section in three of four South Asian locations, and that the associations persisted after adjustments for maternal, pregnancy and delivery characteristics, and year of delivery. We also found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and urban India (private facilities only). These results lend support to the hypothesis that increased caesarean section rates in these three South Asian countries may in part be driven by the private sector, but also suggest that, in some settings such as Bangladesh and urban India, demand from more educated mothers may play a part. These findings call for greater, local understanding of the role of private providers in maternity care, together with careful examination of the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.

Contributors
** At least one visit with skilled provider ***Includes: symptoms of eclampsia (fits, seizures, convulsions, or unconsciousness during pregnancy or delivery); reduced or no fetal movement; labour lasting more than 24 hours. ****For the Nepal data, two respondents with bachelor's degrees (2 respondent total, 1 delivering in institution) were combined with respondents with secondary education.

Supplemental tables
Supplemental

Health facility characteristics
Public health facility 68.  * Nepal numbers weighted using women's probability of selection within each cluster, scaled to total number of institutional births.

F o r p e e r r e v i e w o n l y
** At least one visit with skilled provider ***Includes: symptoms of pre-eclampsia (swollen face or limbs or blurred vision [not asked in Nepal]), eclampsia (fits, seizures, convulsions, or unconsciousness during pregnancy or delivery); haemorrhaging during delivery, labour lasting more than 24 hours. ****For the Nepal data, secondary and bachelor's (2 respondent total, 1 delivering in institution) were combined.

Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 4 Objectives 3 State specific objectives, including any pre-specified hypotheses 4

Study design 4
Present key elements of study design early in the paper 1, 5-6 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection 5 and Table 1 Participants 6 (a) Cohort study-Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up Case-control study-Give the eligibility criteria, and the sources and methods of case ascertainment and control selection. Give the rationale for the choice of cases and controls Cross-sectional study-Give the eligibility criteria, and the sources and methods of selection of participants 6 (b) Cohort study-For matched studies, give matching criteria and number of exposed and unexposed   Cross-sectional study-Report numbers of outcome events or summary measures Table 3 Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (eg, 95% confidence interval). Make clear which confounders were adjusted for and why they were included Table 3 (b) Report category boundaries when continuous variables were categorized (c) If relevant, consider translating estimates of relative risk into absolute risk for a meaningful time period Other analyses 17 Report other analyses done-eg analyses of subgroups and interactions, and sensitivity analyses Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based 13 *Give information separately for cases and controls in case-control studies and, if applicable, for exposed and unexposed groups in cohort and cross-sectional studies.  determinants of caesarean section delivery by location.
Results: Institutional delivery rates varied widely between settings, from 21% in rural India to 90% in urban India. The proportion of private and charitable facility births delivered by caesarean section was 73% in Bangladesh, 30% in rural Nepal, 18% in urban India, and 5% in rural India. The odds of caesarean section were greater in private and charitable health facilities than in public facilities in three of four study locations, even when adjusted for pregnancy and delivery characteristics, maternal characteristics,

Conclusions:
Our results lend support to the hypothesis that increased caesarean section rates in these South Asian countries may be driven in part by the private sector. They also suggest that preferences for caesarean delivery may be higher among highly educated women, and that individual-and provider-level factors interact in driving caesarean rates higher. Rates of caesarean section in the private sector, and their maternal and neonatal health outcomes, require close monitoring.

Article focus
• Caesarean sections are on the rise in South Asia. Research suggests that private health facilities may be contributing to this by conducting caesareans over and above clinical need.
• Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to want caesareans.
• This study examined the prevalence and determinants of caesarean sections in private and charitable facilities compared with public health facilities for over 45,000 births in underserved communities of rural Nepal, rural Bangladesh, rural and urban India.

Key messages
• Caesarean sections were more common in private than in public health facilities in three of four study locations (urban India, rural Nepal, and rural Bangladesh).
• We found that highly educated women delivering in private facilities were particularly likely to deliver by caesarean in Bangladesh and urban India.
• Future research should monitor caesarean section rates in the private sector in South Asia and subsequent maternal and neonatal health outcomes

Strengths and limitations
• This study had a large sample size and focused on underserved communities, which are a priority for public health interventions in South Asia.
• Our data were not nationally representative, which limits the generalisability of our findings.
Other limitations included small denominators and missing data for some variables. We did not have data on some known predictors of caesarean section, which would have enhanced the completeness of our determinants analysis.

INTRODUCTION
Access to comprehensive emergency obstetric care, including caesarean section, is key to preventing the estimated 287,000 maternal and 2.9 million neonatal deaths that occur worldwide every year. 1 2 Although debate continues about how to quantify the need for life-saving obstetric surgery, a 1985 World Health Organization (WHO) report suggested that the optimal population range for caesarean section rates is between 5% and 15%, and this endures as a reference. 3 4 Caesarean section rates are increasing worldwide, albeit unequally: a recent analysis of Demographic and Health Survey (DHS) data in 26 South Asian and sub-Saharan African countries found that rates were highest among the 'urban rich' in all countries, and lowest among the 'rural poor' in 18. 5 In all countries, fewer than 5% of mothers in the poorest wealth quintile delivered by caesarean. 6 Caesarean sections conducted without clinical need can have adverse consequences for mothers and children. A 2008 WHO survey of 373 facilities across 24 countries found that unnecessary caesareans were associated with an increased risk of maternal mortality and serious outcomes for mothers and newborn infants, compared with spontaneous vaginal delivery. 7 Recent ecological analyses also highlighted strong associations between caesarean delivery and increased neonatal mortality in countries with low and medium caesarean section rates. 8 Unnecessary caesareans lead to considerable costs for families and health systems: an estimated 6.2 million unnecessary procedures were performed in 2008, costing approximately US$ 2.32 billion. 9 Several South Asian countries have recorded substantial increases in caesarean section rates over the past decade. In Bangladesh, rates rose from 2% (2000) to 17% (2011); in India, from 3% (1992) to 11 % (2006); and in Nepal, from 1% (2000) to 5% (2011).  Several studies in these countries have raised concerns about high caesarean rates in private facilities, and a recent DHS analysis speculated that national increases in caesarean section rates in South Asian countries could be driven in part by higher rates among deliveries in private sector facilities . 6 16-18 Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to desire caesareans. 19 20 There are few large, recent community-based studies from South Asia quantifying differences in caesarean section rates between public and private facilities. Such studies are necessary in order to examine whether increases in caesareans in these settings are indeed likely to be driven by the private sector, demand from wealthier and more educated mothers, or a combination of the two. We conducted a cross-sectional analysis of data from Bangladesh, India, and Nepal to explore the prevalence and determinants of caesarean section delivery by type of facility and maternal characteristics.

Study populations
We used data collected through vital events surveillance systems established during four clusterrandomised controlled trials conducted between 2005 and 2011. The trials were done with communities that can be considered socio-economically disadvantaged: in Bangladesh and Nepal, they took place in four rural, underserved districts (Bogra, Maulvibazaar, and Faridpur in Bangladesh, and the Terai district of Dhanusha, Nepal); in rural India, most participants were from Scheduled Tribes in two states of eastern India (Jharkhand and Odisha); in urban India, data came from slum communities in Mumbai. Table 1 describes the characteristics of each study and its population, including the background maternal mortality ratio and types of facilities present in the study areas. The original trials were designed to evaluate the impact of participatory women's groups on maternal and neonatal health outcomes. [21][22][23][24][25] We used data from the control areas of these trials only, because the women's group intervention led to changes in mortality and practices in several locations. 26 Health system contexts million people across 31 sub-districts, around 7% of the country's population. 27

Data collection
All data were collected using surveillance systems to monitor births and deaths prospectively. [21][22][23][24][25] In all study locations, a female community-based key informant reported births and deaths in her area, which covered a population ranging from 250 to 350. A trained interviewer then verified these reports and paid the informant an incentive for each correct identification. In Bangladesh, rural and urban India, the interviewer administered a structured questionnaire to all eligible mothers around six weeks after delivery; in Nepal, all births in the study area were registered, and interviews were conducted on all births in small clusters and on a random sample of 10 births per month in the larger clusters. In each study location, mothers were interviewed using a questionnaire to collect information about events in the antenatal, delivery and postnatal periods. In each trial and in this subsequent study, participants were women of reproductive age  who delivered in the study area during the data collection period, and who consented to be interviewed six weeks after delivery, and their infants.

Study sample
The initial sample for this analysis included 46 The primary outcome in these analyses was caesarean section delivery, identified by self-report from the mother or another household member around six weeks after giving birth. Maternal socio-demographic characteristics included in the models were: mother's age at delivery, her educational attainment, and household assets. Mother's age was entered into the model as a categorical measure in 10-year groups. Educational attainment was entered as a categorical variable using the following categories: no formal education, primary education, secondary education, or bachelor's degree or higher. To develop an asset index, we used polychoric factor analysis on data on common assets and amenities found in the mother's household, and grouped the resulting factor scores into quartiles. 33 Assets and amenities included electricity, radio or cassette player, electric fan (Bangladesh and India only), television, refrigerator (Bangladesh and India only), telephone (Bangladesh and Nepal only), generator (India only), and bicycle. All models were additionally adjusted for location (Bangladesh, India, Nepal), and year of interview in three-year groups. The data were collected in a stratified, clustersampled survey, and we accounted for survey design in the analysis using a fixed effect for stratum and a random effect for cluster.

Statistical analysis
We used frequencies to describe caesarean section rates by delivery location at each site. We used the Generalized Linear Latent And Mixed Models (GLLAMM) procedure in Stata 13.1, with adaptive quadrature for binary outcomes, to estimate the crude association between type of delivery facility and caesarean section. 34 35 We identified other maternal, pregnancy, and delivery characteristics potentially between maternal education and caesarean delivery. 7 36 37 We fitted models including indicator variables for each group of mothers by education and type of delivery facility to explore differences in the strength of association between caesarean delivery and private facility by mother's education. To account for the sampling procedure used in rural Nepal, models were adjusted using pweights (probability of selection within cluster); these weights were rescaled to reflect the total number of institutional deliveries.

Ethical approval
The trials that provided data for this study received ethical approval from the following committees: the

DISCUSSION
Our analysis of data from over 45,000 births confirms the findings of other studies identifying differences in caesarean rates between public and private facilities, and suggests that, even in underserved areas in South Asia, caesareans without medical indication are of concern. In three of four locations, rates of caesarean section were higher in private/charitable facilities than in public facilities. The findings from Bangladesh are particularly noteworthy as they show much greater odds of caesarean section in private facilities, concurring with previous analyses. 37 This was also the only location where serious complications in pregnancy and delivery were not associated with caesarean delivery, suggesting that obstetric surgery was performed over and above clinical need.
Our findings confirm the results of earlier studies of the prevalence of caesarean delivery in South Asia, and indicate that high rates can be found in underserved rural areas. In India, a recent analysis of 2010-11 Annual Health Survey (AHS) data from 284 districts in nine States, including Jharkhand, found that the median caesarean section rate in the private sector was 28%, compared with 5% in the public sector. 38 There appear to be strong financial incentives for surgical procedures in the private sector. 39 In a rural, largely indigenous part of eastern India where more mothers delivered in private/charitable facilities than in public facilities, more caesarean sections were performed in the public sector. In this particular setting, our field experience suggests that only women with serious complications would go to a facility and have a caesarean section, especially as JSY was not yet available in the study areas at the time of data collection. Such women are likely to have experienced multiple referrals from either ill-equipped public or private facilities not wanting to take the risk of admission. It is however possible that preference for the private sector has changed in Jharkhand since the advent of JSY, and further disaggregated analyses of Annual Health Survey data would allow a more contemporary exploration of State-level variations in caesarean section rates between public and private sectors.
In Mumbai slums with a high uptake of institutional births and a 60/40 split in favour of the public sector for delivery care, we found no difference in caesarean section rates between the two sectors, which is providers who recommend them, mainly because of the high costs associated with the procedure. 37 40 41 A qualitative study involving twenty women who had experienced obstetric complications in Matlab in 2008-9 found that most of the fourteen women who had undergone caesareans had spent over Tk 14,999 (US$ 217) on the procedure, which was approximately one-third of GDP per capita at the time. 41 The lack of association between caesareans and complications in pregnancy or delivery and multiple births, coupled with the high financial incentives given to providers for performing caesarean sections and the requirement for junior doctors to 'practise' their surgical skills, further suggest that obstetric surgery is being used over and above clinical need. A possible explanation for the significant interactive associations between maternal education and caesarean delivery in both public and private facilities in Bangladesh is that well-educated women may be delivering in more expensive or highly rated institutions, which may in turn be more likely to do caesarean sections for financial reasons and if they act as training centres for junior doctors.
In our sample of institutional deliveries from rural Nepal, 16% of facility births were by caesarean, which is higher than the national average: the 2011 Nepal DHS found an overall national caesarean section rate of 5%, with sections more commonly performed for births to highly educated mothers (13%) and mothers in the highest wealth quintile (14%). 26 It is possible that women with complications are more likely to deliver in facilities, and also that, as in Bangladesh, private providers are motivated by financial incentives to conduct caesareans more frequently than strictly necessary. Women who can afford to seek care in private facilities may also be more willing or able to pay for caesarean sections, and providers may conduct more of them to increase their income.
A 2010 analysis of DHS data examining the role of the private sector in maternity care in 16 countries found evidence of a trend towards privatisation in delivery care between the 1990s and mid-2000s, but with strong differences between countries, which might reflect the heterogeneous nature of this sector both between and within countries. 42 This DHS study highlighted the need for more context-specific data on the nature of the private sector in low and middle-income countries, and its role in maternity care. Further research might focus on understanding the motivations and experiences of women undergoing caesarean sections in private facilities in South Asian settings, pathways for switching between public and private sectors in the event of obstetric complications, a more comprehensive tally of the financial incentives (official or non-official) that motivate private providers to carry out caesareans in each setting, and the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.
In our study, locations with higher prevalence of caesarean deliveries also showed a positive educational gradient even after adjusting for wealth measured by household assets. Moreover, there was an interactive association between education and type of facility, with highly educated women particularly likely to receive caesarean deliveries in private facilities in urban India. Literature exploring the determinants of caesarean delivery emphasizes that multiple influences drive a woman's decision to deliver by caesarean. Profit considerations at the facility level may prompt some providers to urge women to receive unnecessary procedures, while women themselves may prefer caesareans for cultural reasons, fear of painful deliveries, or because they believe it to be safer. 36 43-45 Within the client-provider interaction, providers may be more likely to acquiesce to a request for caesarean from a highly educated woman; and such women may be more likely to accept advice from a provider. 19 46 Our findings lend support to the hypothesis that, while provider-level factors are probably partially responsible for the rapid increase in caesarean deliveries, it is also necessary to consider women's own preferences and decision-making processes and how they are shaped by social and cultural factors.

Strengths and limitations
The strengths of the study were its large sample size and focus on underserved communities, which are a priority for public health interventions in South Asia. It had five main limitations. It was not a nationally representative study, and districts or clusters were sampled purposively from previous cRCTs. This limits the generalisability of our findings to geographical settings outside the study areas. The study was crosssectional, and therefore only able to suggest associations rather than causal relationships. The data also did not include key predictors of caesarean section such as breech presentation, and whether the previous delivery was by caesarean, which limits the completeness of our analysis of determinants. In addition, levels of serious complications in pregnancy and delivery varied considerably between locations and were often higher than expected, suggesting potential over-reporting and limited reliability as an indicator of complications. Finally, some variables had small denominators, and others had high levels of missing data (for example, maternal age in rural India).

Conclusions
Our study found that delivering in a private health facility was associated with an increased odds of caesarean section in three of four South Asian locations, and that the associations persisted after adjustments for maternal, pregnancy and delivery characteristics, and year of delivery. We also found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and urban India (private facilities only). These results lend support to the hypothesis that increased caesarean section rates in these three South Asian countries may in part be driven by the private sector, but also suggest that, in some settings such as Bangladesh and urban India, demand from more educated mothers may play a part. These findings call for greater, local understanding of the role of private providers in maternity care, together with careful examination of the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.

Competing interests
None to declare
** At least one visit with skilled provider ***Includes: symptoms of eclampsia (fits, seizures, convulsions, or unconsciousness during pregnancy or delivery); reduced or no fetal movement; labour lasting more than 24 hours. ****For the Nepal data, two respondents with bachelor's degrees (2 respondent total, 1 delivering in institution) were combined with respondents with secondary education.

Conclusions:
Our results lend support to the hypothesis that increased caesarean section rates in these South Asian countries may be driven in part by the private sector. They also suggest that preferences for caesarean delivery may be higher among highly educated women, and that individual-and provider-level factors interact in driving caesarean rates higher. Rates of caesarean section in the private sector, and their maternal and neonatal health outcomes, require close monitoring.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  • Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to want caesareans.
• This study examined the prevalence and determinants of caesarean sections in private and charitable facilities compared with public health facilities for over 45,000 births in underserved communities of rural Nepal, rural Bangladesh, rural and urban India.

Key messages
• Caesarean sections were more common in private than in public health facilities in three of four study locations (urban India, rural Nepal, and rural Bangladesh).
• We found significant interactive associations between found that highly educated women maternal education and delivering in private facilitiesy delivery on were particularly likely to deliver by caesarean in Bangladesh (both private and public facilities) and urban India (private facilities only).
• Future research should monitor caesarean section rates in the private sector in South Asia and subsequent maternal and neonatal health outcomes

Strengths and limitations
• This study had a large sample size and focused on underserved communities, which are a priority for public health interventions in South Asia.
• Our data were not nationally representative, which limits the generalisability of our findings.
Other limitations included small denominators and missing data for some variables. We did not have data on some known predictors of caesarean section, which would have enhanced the completeness of our determinants analysis.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  Although debate continues about how to quantify the need for life-saving obstetric surgery, a 1985 World Health Organization (WHO) report suggested that the optimal population range for caesarean section rates is between 5% and 15%, and this endures as a reference. 3 (2000) to 5% (2011).  Several studies in these countries have raised concerns about high caesarean rates in private facilities, and a recent DHS analysis speculated that national increases in caesarean section rates in South Asian countries could be driven in part by higher rates among deliveries in private sector facilities . 6 16-18 Literature from other settings indicates that increases in caesarean sections are shaped by both supply and demand pressures: providers often have financial incentives to intervene surgically, and women of higher socio-economic status are also more likely to desire caesareans. 19 20 There are few large, recent community-based studies from South Asia quantifying differences in caesarean section rates between public and private facilities. Such studies are necessary in order to examine whether increases in caesareans in these settings are indeed likely to be driven by the private sector, demand from wealthier and more educated mothers, or a combination of the two. We conducted a cross-sectional analysis of data from Bangladesh, India, and Nepal to explore the prevalence and determinants of caesarean section delivery by type of facility and maternal characteristics. location,  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Study populations
We used data collected through vital events surveillance systems established during four clusterrandomised controlled trials conducted between 2005 and 2011. The trials were done with communities that can be considered socio-economically disadvantaged: in Bangladesh and Nepal, they took place in four rural, underserved districts (Bogra, Maulvibazaar, and Faridpur in Bangladesh, and the Terai district of Dhanusha, Nepal); in rural India, most participants were from Scheduled Tribes in two states of eastern India (Jharkhand and Odisha); in urban India, data came from slum communities in Mumbai. Table 1 describes the characteristics of each study and its population, including the background maternal mortality ratio and types of facilities present in the study areas. The original trials were designed to evaluate the impact of participatory women's groups on maternal and neonatal health outcomes. [21][22][23][24][25] We used data from the control areas of these trials only, because the women's group intervention led to changes in mortality and practices in several locations. 26 Health system contexts  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Data collection
All data were collected using surveillance systems to monitor births and deaths prospectively. [21][22][23][24][25] In all study locations, a female community-based key informant reported births and deaths in her area, which covered a population ranging from 250 to 350. A trained interviewer then verified these reports and paid the informant an incentive for each correct identification. In Bangladesh, rural and urban India, the interviewer administered a structured questionnaire to all eligible mothers around six weeks after delivery; in Nepal, all births in the study area were registered, and interviews were conducted on all births in small clusters and on a random sample of 10 births per month in the larger clusters. In each study location, mothers were interviewed using a questionnaire to collect information about events in the antenatal, delivery and postnatal periods. In each trial and in this subsequent study, participants were women of reproductive age (15-49) who delivered in the study area during the data collection period, and who consented to be interviewed six weeks after delivery, and their infants.

Measures used
The primary outcome in these analyses was caesarean section delivery, identified by self-report from the mother or another household member around six weeks after giving birth. Maternal socio-demographic characteristics included in the models were: mother's age at delivery, her educational attainment, and household assets. Mother's age was entered into the model as a categorical measure in 10-year groups. Educational attainment was entered as a categorical variable using the following categories: no formal education, primary education, secondary education, or bachelor's degree or higher. To develop an asset index, we used polychoric factor analysis on data on common assets and amenities found in the mother's household, and grouped the resulting factor scores into quartiles. 33 Assets and amenities included electricity, radio or cassette player, electric fan (Bangladesh and India only), television, refrigerator (Bangladesh and India only), telephone (Bangladesh and Nepal only), generator (India only), and bicycle. All models were additionally adjusted for location (Bangladesh, India, Nepal), and year of interview in three-year groups. The data were collected in a stratified, clustersampled survey, and we accounted for survey design in the analysis using a fixed effect for stratum and a random effect for cluster.

Statistical analysis
We used frequencies to describe caesarean section rates by delivery location at each site. We used the Generalized Linear Latent And Mixed Models (GLLAMM) procedure in Stata 13.1, with adaptive  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   8 quadrature for binary outcomes, to estimate the crude association between type of delivery facility and caesarean section. 34 35 We identified other maternal, pregnancy, and delivery characteristics potentially associated with caesarean section using existing literature, especially studies resulting from the WHO multi-country surveys, and entered these in adjusted models to explore how they modified the association between type of delivery facility and caesarean section, and their individual association with caesarean section. Some South Asian and Latin American studies have detected a strong association between maternal education and caesarean delivery. 7 36 37 We fitted models including indicator variables for each group of mothers by education and type of delivery facility to explore differences in the strength of association between caesarean delivery and private facility by mother's education. To account for the sampling procedure used in rural Nepal, models were adjusted using pweights (probability of selection within cluster); these weights were rescaled to reflect the total number of institutional deliveries.

DISCUSSION
Our analysis of data from over 45,000 births confirms the findings of other studies identifying differences in caesarean rates between public and private facilities, and suggests that, even in underserved areas in South Asia, caesareans without medical indication are of concern. In three of four locations, rates of caesarean section were higher in private/charitable facilities than in public facilities. The findings from Bangladesh are particularly noteworthy as they show much greater odds of caesarean section in private facilities, concurring with previous analyses. 37 This was also the only location where serious complications in pregnancy and delivery were not associated with caesarean delivery, suggesting that obstetric surgery was performed over and above clinical need.
Our findings confirm the results of earlier studies of the prevalence of caesarean delivery in South Asia, and indicate that high rates can be found in underserved rural areas. In India, a recent analysis of 2010-11 Annual Health Survey (AHS) data from 284 districts in nine States, including Jharkhand, found that the median caesarean section rate in the private sector was 28%, compared with 5% in the public sector. 38 There appear to be strong financial incentives for surgical procedures in the private sector. 39 In a rural, largely indigenous part of eastern India where more mothers delivered in private/charitable facilities than in public facilities, more caesarean sections were performed in the public sector. In this particular setting, our field experience suggests that only women with serious complications would go to a facility and have a caesarean section, especially as JSY was not yet available in the study areas at the time of data collection. Such women are likely to have experienced multiple referrals from either ill-equipped public or private facilities not wanting to take the risk of admission. It is however possible that preference for the private sector has changed in Jharkhand since the advent of JSY, and further disaggregated analyses of Annual Health Survey data would allow a more contemporary exploration of State-level variations in caesarean section rates between public and private sectors.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  In Mumbai slums with a high uptake of institutional births and a 60/40 split in favour of the public sector for delivery care, we found no difference in caesarean section rates between the two sectors, which is somewhat surprising given the high rates of caesareans observed in the private sector in nine other states of India (excluding Maharashtra). At least two scenarios may be relevant in this setting. First, public sector hospitals are able to provide caesarean section, while smaller private maternity homes may not be able to. Second, the costs of caesarean section borne by private sector providers are higher in Mumbai, and families may be unwilling to bear these additional costs. This provides motivation for private providers not to provide caesareans, and makes it likely that women experiencing complications move into the public sector. The variety of facility options in an urban environment means that mothers are able to choose between different types of facilities, while in rural areas only one provider may be accessible.
Analyses using the most recent Bangladesh DHS found that, in 2011, three in five facility births were delivered by caesarean section. This reflects a historical trend: in 2001-2003, nearly half of deliveries in private facilities in Bangladesh were already by caesarean section. 18 Our data from rural, socioeconomically disadvantaged communities in three districts of Bangladesh confirm these population-level findings, but also suggest that high caesarean section rates in private facilities are not merely an issue for wealthy urban mothers. Although some research from Bangladesh suggests that mothers may have a preference for caesarean delivery because of fear of labour pain or a desire to select an auspicious date for the birth, other studies also highlight women's fears of caesarean section and their distrust of health providers who recommend them, mainly because of the high costs associated with the procedure. 37 40 41 A qualitative study involving twenty women who had experienced obstetric complications in Matlab in 2008-9 found that most of the fourteen women who had undergone caesareans had spent over Tk 14,999 (US$ 217) on the procedure, which was approximately one-third of GDP per capita at the time. 41 The lack of association between caesareans and complications in pregnancy or delivery and multiple births, coupled with the high financial incentives given to providers for performing caesarean sections and the requirement for junior doctors to 'practise' their surgical skills, further suggest that obstetric surgery is being used over and above clinical need. A possible explanation for the significant interactive associations between maternal education and caesarean delivery in both public and private facilities in Bangladesh is that well-educated women may be delivering in more expensive or highly rated institutions, which may in turn be more likely to do caesarean sections for financial reasons and if they act as training centres for junior doctors.
In our sample of institutional deliveries from rural Nepal, 16% of facility births were by caesarean, which is higher than the national average: the 2011 Nepal DHS found an overall national caesarean section rate of 5%, with sections more commonly performed for births to highly educated mothers (13%) and  26 It is possible that women with complications are more likely to deliver in facilities, and also that, as in Bangladesh, private providers are motivated by financial incentives to conduct caesareans more frequently than strictly necessary. Women who can afford to seek care in private facilities may also be more willing or able to pay for caesarean sections, and providers may conduct more of them to increase their income.
A 2010 analysis of DHS data examining the role of the private sector in maternity care in 16 countries found evidence of a trend towards privatisation in delivery care between the 1990s and mid-2000s, but with strong differences between countries, which might reflect the heterogeneous nature of this sector both between and within countries. 42 This DHS study highlighted the need for more context-specific data on the nature of the private sector in low and middle-income countries, and its role in maternity care. Further research might focus on understanding the motivations and experiences of women undergoing caesarean sections in private facilities in South Asian settings, pathways for switching between public and private sectors in the event of obstetric complications, a more comprehensive tally of the financial incentives (official or non-official) that motivate private providers to carry out caesareans in each setting, and the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.
In our study, locations with higher prevalence of caesarean deliveries also showed a positive educational gradient even after adjusting for wealth measured by household assets. Moreover, there was an interactive association between education and type of facility, with highly educated women particularly likely to receive caesarean deliveries in private facilities in Bangladesh and urban India. Literature exploring the determinants of caesarean delivery emphasizes that multiple influences drive a woman's decision to deliver by caesarean. Profit considerations at the facility level may prompt some providers to urge women to receive unnecessary procedures, while women themselves may prefer caesareans for cultural reasons, fear of painful deliveries, or because they believe it to be safer. 36 43-45 Within the clientprovider interaction, providers may be more likely to acquiesce to a request for caesarean from a highly educated woman; and such women may be more likely to accept advice from a provider. 19 46 Our findings lend support to the hypothesis that, while provider-level factors are probably partially responsible for the rapid increase in caesarean deliveries, it is also necessary to consider women's own preferences and decision-making processes and how they are shaped by social and cultural factors.

Strengths and limitations
The strengths of the study were its large sample size and focus on underserved communities, which are a priority for public health interventions in South Asia. It had five main limitations. It was not a nationally representative study, and districts or clusters were sampled purposively from previous cRCTs. This limits  13 the generalisability of our findings to geographical settings outside the study areas. The study was crosssectional, and therefore only able to suggest associations rather than causal relationships. The data also did not include key predictors of caesarean section such as breech presentation, and whether the previous delivery was by caesarean, which limits the completeness of our analysis of determinants. In addition, levels of serious complications in pregnancy and delivery varied considerably between locations and were often higher than expected, suggesting potential over-reporting and limited reliability as an indicator of complications. Finally, some variables had small denominators, and others had high levels of missing data (for example, maternal age in rural India).

Conclusions
Our study found that delivering in a private health facility was associated with an increased odds of caesarean section in three of four South Asian locations, and that the associations persisted after adjustments for maternal, pregnancy and delivery characteristics, and year of delivery. We also found significant interactive associations between maternal education and caesarean delivery in Bangladesh (both private and public facilities) and urban India (private facilities only). These results lend support to the hypothesis that increased caesarean section rates in these three South Asian countries may in part be driven by the private sector, but also suggest that, in some settings such as Bangladesh and urban India, demand from more educated mothers may play a part. These findings call for greater, local understanding of the role of private providers in maternity care, together with careful examination of the consequences of increased caesarean sections in the private sector for maternal and neonatal health outcomes.

Contributors
Objectives 3 State specific objectives, including any pre-specified hypotheses 4

Study design 4
Present key elements of study design early in the paper 1, 5-6 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection 5 and Table 1 Participants 6 (a) Cohort study-Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up Case-control study-Give the eligibility criteria, and the sources and methods of case ascertainment and control selection. Give the rationale for the choice of cases and controls Cross-sectional study-Give the eligibility criteria, and the sources and methods of selection of participants 6 (b) Cohort study-For matched studies, give matching criteria and number of exposed and unexposed  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48 Cross-sectional study-If applicable, describe analytical methods taking account of sampling strategy (e) Describe any sensitivity analyses N/A Results