Rising trends and persisting inequalities in cesarean section rates in Nepal: evidence from demographic and health surveys 2006-2016

Background Sustainable development goals require member countries to reduce maternal mortality ratio below 70 per 100,000 live births by 2030. Addressing inequalities in emergency obstetric care is crucial for reducing the maternal mortality ratio. This study was undertaken to examine the time trends and socio-demographic inequalities in the utilization of cesarean section (CS) in Nepal during the period of 2006 and 2016. Methods Data from the Nepal Demographic and Health Surveys (NDHS) 2006, 2011 and2016 were sourced for this study. Women who had a live birth in the last five years of the survey (most recent birth if there were two or more childbirths) were the unit of analysis for this study. Absolute and relative inequalities in CS rates were expressed in-terms of rate difference and rate ratios, respectively. We used binary logistic regression models to assess the rate of cesarean sections by background socio-demographic characteristics of women. Results Age and parity adjusted CS rates were found to have increased almost three-fold (from 3.2%,95% CI:2.1-4.3 in 2006 to 10.5%;95% CI:8.9-11.9 in 2016) over the decade. In 2016, women from Mountain region (3.0%;95% CI:1.1-4.9), those from poorest wealth quintile (2.4%,95% CI:(1.2-3.7) and those living in province 6(2.4%,95% CI:1.3-3.5) had CS rate below 5%. Whereas, women from the richest income quintile (25.1%,95% CI :20.2-30.1), with higher education (21.2%,95% CI:14.7-27.8) and those delivering in private facilities (37.1%,95% CI:30.5-43.7) had CS rate above 15%. Women from the richest income quintile (OR-3.3,95% CI: 1.6-7.0) and those delivered in private/NGO-run facilities (OR-3.6;95% CI:2.7-4.9) were more than three times more likely to deliver by CS compared


facilities.
Background Caesarean section(CS) is one of the vital interventions to save lives of mothers and babies at the time of life threatening complications during pregnancy and child birth (1). There is a growing public health concern that population-based CS rates in some countries have reached above the WHO recommended level (2,3). Non-medical factors such as monetary incentives for health care providers and patient preferences are suggested to be key factors for unnecessary CS (4).
Government of Nepal(GoN) aims to reduce maternal and child mortality at the critical time of childbirth by ensuring basic emergency obstetric and newborn care and Comprehensive emergency obstetric and newborn care services to all women who need it (5). Nepal Health Sector strategy has highlighted equitable access to health services and leaving no one behind as one of the major foci of health sector (6). Additionally, in line with sustainable development goals, GoN strives to reduce Maternal mortality ratio to less than 70 per 100,000 live births and to reduce newborns and children deaths due to preventable causes to below one per cent (7). It is critical to have an easy access to emergency obstetric and newborn care to achieve these goals.
Nepal's Safe Motherhood Program has been operational since 1997 with the aim of reducing maternal and newborn mortality. Nepal introduced maternity incentive scheme in 2005 (popularly known as Aama Program) to reduce financial barriers to reach health facility for institutional delivery.
Additionally, GoN has rolled out free delivery care including CS nationwide by abolishing all user fees since 2009 from accredited facilities (8). Furthermore, current Aama Program provides a cash incentive of NPR 3,000 (USD 30 approximately) in mountain districts, NPR 2,000 (USD 20 approximately) in Hill districts, and NPR 1,000 (USD 10 approximately) in Terai districts to women delivering at health institutions (9). This also includes NPR 800 (USD 8 approximately) for women who completed four antenatal care as per GoN protocol. Health facilities also receive a case-based payment for providing free delivery according to type of delivery (normal deliveries: USD 10 for HFs below 25 beds, USD 15 for HFs above 25 beds, complicated deliveries: USD 30 and CS: USD 70) (10)(11)(12). Moreover, GoN has established an emergency referral fund in order to airlift women from geographically remote regions to referral centers in case of complications (5). As a result of these demand-side and supply-side policies, institutional deliveries in Nepal have considerably increased from 9% in 1996 (13) to 57% in 2016 (14). However, given the geographical diversity and sociocultural barriers, reaching poor, marginalized, uneducated and women living in the remote and rural areas is still a big concern (15,16). Furthermore, Mehata et al showed significant inequalities in CS in Nepal by place of residence, wealth quintile, age of the mother, educational status and caste/ethnicity (17).
CS are not easily accessible to women from marginalized sections of the communities and those living in remote areas even when there is a strong medical indication (18). Simultaneously , unnecessary use is common in private hospitals (3), urban areas (19) and among higher educated mothers (20). Although increasing number of private and community hospitals are now implementing Aama program, charging fees for CS is common (21). In resource-poor settings such as Nepal, poor quality of obstetric care (lack of support and monitoring of child birth process, lack of pain management) resulting in less confidence to health care is thought to have increased demand for CS among upper and middle class women (22,23). Since, private hospitals in Nepal are hitherto largely unregulated, high CS rate in private hospitals may be the result of 'on-demand' or 'providerinitiated' for monetary incentives (16,24).
Although both underuse and overuse were associated with poor health outcomes for mothers and newborn, determining optimal CS rate at the population level is a herculean task (25). WHO's new statement released in 2015 recommended that the CS rates should be below 10% (25). Besides poor health outcomes, CS is associated with high cost to the health care system and to the families (26).
Furthermore, as in other settings (27), surge in CS rate has received increasing attention in Nepal on the grounds of women's rights, over-medicalization of birth and abuse of technology (24). Therefore, it is imperative to have an optimal CS rate at a population level while protecting the rights of women to have CS when there is a medical indication. Hence, it is important to monitor service utilization pattern and inequalities by population groups to assess program and policy outcomes. Therefore, by utilizing the most recent household level data from last three rounds of DHS surveys we present the trend and social determinants of inequalities in utilization of CS in Nepal.

Data Source
We used data from three rounds of the Nepal Demographic and Health Survey (NDHS) conducted in 2006, 2011 and 2016. New ERA, a local research firm in Nepal was responsible in implementing all these surveys under the leadership of Ministry of Health and population (MoHP) and technical assistance from ICF. Briefly, NDHS collected nationally representative data every 5 years on a broad range of issues including fertility, reproductive and maternal health, nutrition, and child health. In each round, a nationally representative sample of households was obtained using stratified cluster sampling approach. NDHS 2006 and 2011 used two-stage stratified cluster sampling to select households. Stratification was done by urban/rural place of residence. In the first stage, Primary Sampling Units (PSUs) were selected using probability proportional to size. In the second stage, households were selected using systematic sampling from individual PSU in rural areas. However, three-stage stratified cluster sampling technique was used for selecting households in urban areas for NDHS 2016 i.e in the first stage, PSUs were chosen by probability proportional to size followed by random selection of enumeration areas (EA) from PSUs in second stage and in third stage, households were selected systematically from selected EAs. Details of the survey design and data collection procedure is available in respective survey reports (14,28,29). Women aged 15-49 years who had a live-birth within 5 years preceding the survey comprised the unit of analysis for the current study. In the case when there was more than single birth within five years' period, we considered the most recent birth in our study.

Measures
The dependent variable for this study was whether the last live birth of a woman aged 15-49 years was conducted by C-section. The DHS surveys used same questions: "Was the (name of the last child) delivered by caesarian section?' The responses were recorded as a binary variable either 1 meaning 'Yes' or 0 meaning 'No.' Based on the available literature (30)(31)(32), the independent variables selected for this study include: women's age categories, parity(2 or less, 3-4, more than 4), women's education, women's current working status(Currently working, not working), wealth index, ecological region(Mountain, Hill, Terai), province(Province 1-7), place of residence(urban/rural), ethnicity and place of delivery. The wealth index used for this study were calculated by principal component analysis based on the easy-tocollect data on a household's ownership of selected assets, such as televisions and bicycles; materials used for housing construction; and access to water and sanitation facilities. Households were then categorized to the poorest (Q1), poorer, middle, richer, or the richest (Q5) group (30,33).

Data Analysis
Initially, we computed crude CS rates by women's socio-demographic characteristics. Then, we computed the adjusted CS rates and their 95% confidence intervals standardizing for age and parity using direct standardization method using the sample of NDHS 2016 as the standard population. We then created the standard weight of age and parity separately and combined them for each NDHS.
Further, we adjusted them for each covariate using command "stdize" and "stdweight". To measure the inequalities, two inequality indicators-the ratio between the highest rank and the lowest rank (ratio of Q1 to Q5 for health indicators and the ratio of Q5 to Q1 for indicators of health care) (34,35) and, Rich Poor difference were calculated for this study. The Rate ratio indicator links the level of health or use of health services between the highest and lowest rank. Rate difference provides an absolute difference in prevalence of caesarian section between the highest and lowest rank. To some extent, these two indicators i.e absolute inequality in terms of rate difference (Q5-Q1) and relative inequality in terms of rate ratio(Q5/Q1) was done for each of the survey periods (35)(36)(37)(38). Presenting both relative and absolute measures of inequalities is important for increasing transparency, and to provide unbiased evidence for policy making (39) however these two frequently used measures of inequality are easy to understand, but comparisons are limited to two extreme groups rather than covering the full population variety (35,40). We conducted binary logistic regression analysis adjusting for women's age and parity to estimate the odd ratios of CS according to the women's characteristics. All independent variables were tested for collinearity before running logistic regression. Since no collinearity was detected, all variables were included in the regression model. The level of statistical significance was set to 0.05. Although recent guideline has suggested CS rate of 10% as adequate at population level(25), we used 5% as minimum rate and 15% as maximum rate to interpret underuse and overuse of CS based on previously accepted standards (41,42). We used sampling weights to adjust for variations in the selection probabilities and interviews among respondents and "svyset" command was used to account for complex survey design and to provide unbiased estimate. All analyses were performed using STATA version 15.0.

Ethical Considerations
Nepal DHS surveys were reviewed and approved by Nepal Health Research Council (NHRC) and the Institutional Review Board of ICF. The interviewers pursued informed consent from the women before the interviews as per the guideline of NHRC. We used de-identified publicly available data upon the request from DHS program (https://dhsprogram.com/Data/terms-of-use.cfm). This is the secondary analysis and the publicly available datasets did not include individual identifiers and thus did not require ethics approval.

Results
The  Table 1). Crude C-section increased from 2.7% in 2006 to 9.0% in 2016, with large and increasing absolute disparities in C-section according to women's socio-demographic characteristics ( Table 2). A higher education level was associated with a markedly higher C-section rate in all three surveys, specifically; in 2016 this rate was at 19.5%. Furthermore, higher C-section rates (28.2% in 2016) was observed among the women from the richest income quintile and the lowest rate (2.4%) among women from the poorest income quintile. Provincial analysis showed, province 3 had persistently higher C-section rate (6.0%, 8 (Table 3). Cesarean section was determined for the most recent birth of women aged 15-49 years who had a live-birth five years preceding the surveys. All percentages are weighted, so the absolute number of participants doesn't perfectly correspond to percentages. The question on C-section was asked only of women who delivered in a health facility while analysis was based on the number of total births. NA: not available Age-and parity-adjusted C-section rates increased from 3.2% in 2006 to 10.5% in 2016 representing almost a threefold increase during 10 years' period ( Table 3). The greatest increase in CS was seen among women with a higher educational level, those in the richest wealth quintile and women who delivered in private facility. Similar patterns of socioeconomic differences were observed among women living in province 1, province 3 and province 4. The absolute inequality in CS rate by maternal educational status increased from 9.   Table 4 shows age and parity-adjusted odds ratios of C-section using binary logistic regression. The

Discussion
In this study, we examined the trend of CS rate in Nepal and inequalities in its utilization by geography (Mountain, Hill and Terai); place of residence (urban/rural, Province); population characteristics (educational status, income quintile, caste/ethnicity, occupational status) and place of delivery. We found that overall CS rate in Nepal in 2016 (9%) was below WHO's maximum limit of 15% and was much lower than in China (43,44), Bangladesh (26)and Brazil (27). The current analysis revealed more than three-fold increase (from 3.2% ;95% CI: 2.1-4.3 in 2006 to 10.5%, 95% CI: 8.9-11.9 in 2016) in age and parity adjusted CS rate over the decade. The rising trend in CS found in our study is consistent with the experience of other low and middle income countries (45,46). This and improvement in CS procedure may be associated with increase in CS rates as seen in other settings (49,50).  (51), and the cost was significantly higher to deliver at private hospital.
Furthermore, women with higher educational level (21.2%;95% CI:14.7-27.8) had a higher CS rate than women with no education (6.3%;95% CI:3.5-9.0) after adjusting for maternal age and parity (Table 3). However, the association disappeared in multivariate regression analysis (Table 4). We previously found inequalities in full vaccination of children was larger by household wealth quintiles than by maternal educational status (52). Altogether, these findings suggest economic barriers for service utilization to be stronger than maternal educational related barriers in Nepal for maternal health and immunization service utilization.

Implications
On the basis of current findings and earlier findings by Acharya et al showing significant hidden cost to utilize CS (51) , the maternity incentive program needs to be revised targeting poor women from remote areas /mountain region areas when maternity complications arise. Although, GoN has established emergency referral fund to airlift women from poor, Dalit, Janajati, geographically disadvantaged and from mountainous region to transport to a higher facility in case of life threatening maternity complications, the use of this fund is very low. Only nine women in FY 2017/2018 used this fund to be air lifted (5). There is a need of an in-depth study to understand barriers in the use of emergency referral funds.
It is beyond the scope of this study to determine the cause for high C-section in private facilities.
However, such a high CS rate in private/NGO-run facilities in Nepal warrants an appropriately defined quality assurance mechanism (53) . Appropriate government policies and clinical practice protocols should be enacted to achieve an optimal C-section rates. Increasing rates of C-section has created a huge financial burden for Bangladesh's health care system (26). A study on an economic burden of CS in Nepal may be a starting point to understand it's significance. The findings can be used to develop advocacy tool , eventually to implement a contextual and validated tool such as Robson classification to support health workers to assess need and monitor the use of CS (54). Monitoring of the CS rate is necessary to ensure those who need the service are receiving and equity in health care use is achieved. Further, it can minimize provider induced CS for the sake of money. Additionally, ensuring continuous support (emotional support, information support, and advocacy of her wishes) for women during child-birth (55), and enhancing quality of obstetric care by ensuring adequate pain management, and use of technology can build-up women's confidence in obstetric care and decrease maternal demand for CS (23).
Some studies have indicated that midwife led maternity care can reduce C-section rates including the maternity costs (56). In Nepal, maternity care in hospitals is obstetrician led and it is high time to consider promoting physiological birth through midwife led maternity care (24).There is a need to scale up current midwifery education to improve maternity care as well as to reduce over-burden of normal deliveries in tertiary care facilities.
Although the overall CS rates in Nepal is within WHO standard [15], due attention is necessary to monitor the increasing C-section rate (crude CS rate from 2.7% in 2006 to 9% in 2016) and appropriate measures should be in place to curb the increasing trend. More importantly, women from the poorest income quintiles with maternity complications need special targeting. Furthermore, there is a need to expand C-section in Mountain districts and some provinces (Province 2, Province 5, Province 6 and Province 7) to reduce geographical inequality.

Strengths and limitations
We used high quality nationally representative data to assess trend and inequalities in CS. However, DHS survey lacks clinical data to evaluate appropriateness of CS. However, it is unlikely that clinical factors drive these variations. Furthermore, supply side variables and variables related to sociocultural norms and beliefs were not available. Therefore, future studies can utilize supply side data from national surveys such as Nepal Health Facility Survey(57) for combining with DHS data to get a wider picture of the factors affecting CS utilization (58). Furthermore, qualitative studies are needed to understand service users' and providers' perspectives determining CS utilization.

Conclusion
In this paper, we found age and parity adjusted CS rate in Nepal to have increased almost three fold within ten years' period and is consistently high among women from the richest income quintiles, those with higher education and those delivering in private sector. This study also reveals that inequality persists and is increasing among some population groups (by maternal educational status, household wealth quintile) and provinces. Strategies are needed to support mothers from the poor households, those from mountain region and those from province 2, province 5, province 6 and province 7 to utilize C-section when there is a medical need. Health Research Council, Nepal; consent was sought before beginning the interview as per NHRC ethical review guidelines; data were publicly available and did not include individual identities and thus did not require ethics approval.

Consent for publication
Not Applicable

Availability of data and materials
Data is easily accessible from the DHS program website (https://dhsprogram.com/Data/terms-ofuse.cfm) upon the request.