Correlates of Caesarean Section Deliveries: Evidence from Indonesia

The data for this study comes from the seventh round of the IDHS conducted in 2012 by Statistics Indonesia ( Badan Pusat Statistik – BPS) in collaboration with the National Population and Family Planning Board (BKKBN) and the Indonesia Ministry of Health (MoH)


Introduction
Caesarean section (henceforth CS), a surgical incision performed as a method of childbirth, is usually chosen based on medical indications (Mander, 2007).However, there are cases when CS demanded or elective (Mander, 2007).The World Health Organisation (WHO) suggest that the rate to CS should not exceed 10-15 % of all deliveries (WHO, 2015).Beyond such figures, there may be an indication of caesarean deliveries based on non-clinical grounds and may indicate its overuse (WHO, 2015), which has health and financial consequences.Medically unnecessary CS pose a higher risk of post-partum morbidity (Souza et al., 2010) and also a loss of economic resources which could be used for other more beneficial objectives (Gibbons et al., 2012).
There is an upward trend of CS deliveries in developing countries (Stanton & Holtz, 2006).In Indonesia, the trend is the in a similar direction.The study by Hatt et al. (2007) assessed the trends in rates of CS delivery in Indonesia from 1986 to 2002 by wealth quintile and found an increasing trend of CS delivery especially among wealthiest women.However, the authors of that study did not address other determinants as their objective was to evaluate the village midwife programme in Indonesia (Hatt et al., 2007).Hence, determinants of CS delivery are not well understood.Therefore, this study attempts to investigate spatial, obstetric, and sociodemographic correlates of caesarean section delivery in Indonesia.We do this by analysing the recent Indonesia Demographic and Health Survey 2012 (henceforth 2012 IDHS).

Data Source
The data for this study comes from the seventh round of the IDHS conducted in 2012 by Statistics Indonesia (Badan Pusat Statistik -BPS) in collaboration with the National Population and Family Planning Board (BKKBN) and the Indonesia Ministry of Health (MoH).The 2012 IDHS covered a nationally representative sample from households sampled using a multistage stratified cluster sampling design (Statistics Indonesia, BKKBN, MOH, & ICF International, 2013).Although previous studies have demonstrated the limitations of CS delivery data from DHS (Holtz & Stanton, 2007;Stanton, Dubourg, De Brouwere, Pujades, & Ronsmans, 2005), this is one of the most reliable sources of information of nationally representative CS rate in Indonesia.survey to reduce recall bias.The initial sample size is 15,262 births (Statistics Indonesia et al., 2013).Births with missing information on the dependent and independent variables were excluded from the analysis (n = 383).As such, the final analytic sample included 14,879 most recent births (97.49% of the initial sample).

Variables
The outcome variable, CS delivery, is measured by a binary variable that takes the value of one if the delivery method for the latest pregnancy was a caesarean section, and zero otherwise.The independent variables in this study consist of socio-demographic and spatial characteristics.As socio-demographic characteristics, maternal age, parity, any complications during pregnancy, number of prenatal care (PNC) visits, delivery took place in a private health facility, mother's and father's years of formal schooling (in years), ownership of health insurance, and quintiles of wealth index have been included in the analysis.We use the provided wealth index that was created through three steps (Statistics Indonesia et al., 2013).Moreover, we also use two spatial characteristics, namely region (Java, Sumatera, Bali and Nusa Tenggara, Kalimantan, Sulawesi, and Maluku and Papua) and place of residence (urban/rural).

Statistical Analysis
We estimate bivariate and multivariate logit regressions, presenting average marginal effects (AMEs) and 95 % confidence intervals.All regressions were based on weighted data using sampling weights and sampling design of the 2012 IDHS.All hypotheses are tested using two-tailed p values <0.05.As for the descriptive analysis, we calculate and present descriptive statistics as percentages or means in Table 1.We conducted all of the analyses using Intercooled STATA version 13.1 (StataCorp LP, College Station, Texas).

Sample Characteristics
Table 1 presents the characteristics of the sample by selected background variables.The proportion of women who had CS method for their latest delivery during the five years preceding the survey was 12.8 % (95 % CI: 11.89, 13.77).The descriptive statistics revealed that the women predominantly live in Java region with over half of the sample reside there.As for the place of residence, the sampled women are roughly equally distributed.Concerning maternal age, most of the mothers reported giving their latest birth when they were 25 to 29 years old.Regarding parity, every three in five women had given birth before their latest delivery but had no previous CS delivery.As for complication, almost nine in ten women reported having no complication during their pregnancy.In regards to PNC, every three in five women had, at least, eight PNC visits during their latest pregnancy.As for the place of delivery, the majority of women reported giving birth in a public health facility or at home.The average years of schooling for the mothers were 9.29 years while that of the fathers was slightly higher at 9.38 years.Regarding health insurance, almost two-thirds of the women reported having no insurance cover at all and social security was the highest source of health insurance.As for wealth index, the women are roughly equally distributed in each quintile.

Regression Analyses
The final multivariate LRM consists of ten explanatory variables.The model was statistically significant (F (25,1782) = 32.56;p< 0.001).Table 2 presents the results of bivariate and multivariate regressions.We exclude father's schooling in the multivariate logit model due to its high correlation mother's schooling.

Spatial Variables
Many studies have demonstrated spatial inequalities of CS deliveries across regions within a country, such as in Egypt (Khawaja, Kabakian-Khasholian, & Jurdi, 2004), rural China (Klemetti et al., 2010), Bangladesh (Kamal, 2013), and Nepal (Prakash & Neupane, 2014).Higher CS rate in urban areas has also been shown in previous studies (Arrieta, 2011;Collin, Anwar, & Ronsmans, 2007;Kamal, 2013;Magadi, Agwanda, Obare, & Taffa, 2007;Prakash & Neupane, 2014;Ronsmans, Holtz, & Stanton, 2006).Contrary to previous literature, however, the present study does not indicate significant spatial inequalities in CS delivery both across regions and between urban and rural areas.In terms of region, Java region was chosen as the reference category due to its large population size (Statistics Indonesia, 2010).Women residing in the residing in Sumatera region have a higher probability of CS delivery compared to those living in Java region.It is observed that CS deliveries in other regions were not statistically different from that in the reference region.
In terms of place of residence, the probability of CS in urban women is not significantly higher than that in rural women.
Moreover, women who gave birth at the age between 25 and 29 have a higher probability of CS delivery compared to those in the reference category.Women who gave birth at the age between 30 and 34 also have a higher probability of CS delivery compared to those who gave birth at the age between 20 to 24 years.Lastly, women who were aged 35 years and over when giving birth have a higher probability of CS delivery compared to women in the reference category.These are coherent with what previous studies suggest (Arrieta, 2011;Klemetti et al., 2010;Liu et al., 2007).

Parity
In this study, parity was operationalised in a three-category variable following that of Bragg and others' study in 2010 (Bragg et al., 2010) where women who had no children prior to their latest delivery acts as the reference category.Bragg et al. (2010) demonstrated that multiparous women with (no) history of CS delivery have higher (lower) odds of undergoing CS procedure than nulliparous women.Similarly, this study observed that multiparous women with (no) history of CS delivery have a higher (lower) probability of CS delivery compared to reference women.These findings are consistent with that of previous studies as it has been shown that women with previous CS delivery are more likely to have another one (Liu et al., 2007;Ma et al., 2010).

Prenatal Complication
Existing studies agree that women with prenatal complications are at higher risk of CS delivery (Bragg et al., 2010;Ma et al., 2010).The result of this study is in line as it is observed that women who reported having any complications during their pregnancy have a higher probability of CS delivery than those who reported having none.

Prenatal Care Visits
The number of PNC visits was classified into four categories with none as the reference category.It is claimed that more PNC visits are associated with higher likelihood of CS delivery among women (Khawaja et al., 2004;Neuman et al., 2014).It is observed that women who reported having PNC visits for 1 to 3 times have a higher probability of CS delivery compared to those who reported having none.But it is not statistically significant at all conventional levels of significance (p = 0.149).Moreover, women who reported having PNC visits for 4 to 7 times have a higher probability of CS delivery compared to those who reported having none.Lastly, women who reported having PNC visits for 8 times or more have a higher probability of CS delivery compared to those who reported having none.

Private Health Facility Delivery
The last maternal variable in this study was birthing in private facility.It is claimed in previous literature that women who gave birth in a private health facility are more likely to have CS as their method of delivery than women who gave birth in a public health facility or at home (Ma et al., 2010;Mendoza-Sassi, Cesar, Silva, Denardin, & Rodrigues, 2010;Neuman et al., 2014;Ribeiro et al., 2007).In this study, it is found that women giving birth in a private health facility were of a higher probability of undergoing CS for their last delivery compared to those who gave birth in public health facility or at home.This is consistent with the findings of previous studies.

Maternal Education
Most previous studies found a positive association between education and likelihood of having CS as a mode of delivery (Collin et al., 2007;Klemetti et al., 2010;Magadi et al., 2007;Mendoza-Sassi et al., 2010).However, other studies have suggested the opposite, namely that low-educated women are at higher risk of CS delivery (Hsu et al., 2008;Nilsen et al., 2014); or found no association (Kottwitz, 2014) when access to hospital care is taken into account.In this study, it is observed that education of mother is significantly and positively associated with the probability of CS delivery.For every one year increase in formal education of mother, the probability of getting CS for her latest delivery increases.

Health Insurance
It is observed that women with private health insurance cover have a higher probability of CS compared to those with no cover.This is fairly consistent with previous literature.A study on CS in rural China has shown indication of the positive and significant effect of health insurance cover on the probability of CS delivery (Long et al., 2012).Studies of CS in Taiwan also found a positive relationship, albeit statistically not significant, between health insurance cover and CS delivery (Hsu et al., 2008;Liu et al., 2007).

Household Wealth
In general, the association between wealth and CS delivery in the literature is positive (Arrieta, 2011;Collin et al., 2007;Cresswell, Assarag, Meski, Filippi, & Ronsmans, 2015;Kamal, 2013;Prakash & Neupane, 2014;Ronsmans et al., 2006).The results of the present study confirm previous studies, where being in higher wealth quintiles (second, third, fourth, and highest quintiles) corresponds to the higher probability of CS delivery compared to being in the lowest wealth quintile.A previous study on CS delivery in Indonesia suggests that wealth index is not only associated with the likelihood of CS delivery but also on the trends in rates of CS in women in the wealthiest quintile (Hatt et al., 2007).

Strengths and Limitations of the Study
This study has several strengths.First, the data used is nationally representative.Second, the questionnaire in the survey is internationally standardised and hence comparable to multiple countries.Third, the large sample size of IDHS means larger statistical power.However, this study was not free from limitations.One of which is the cross-sectional form of IDHS.Another limitation is that information sourced from IDHS are mostly retrospective and self-reported.Moreover, elective CS cannot be separated from emergency CS.Given these limitations, causal inference is not warranted.

Conclusion
This study investigated the associates of CS deliveries in Indonesia.It is observed that mother's age at childbirth, parity, complications during pregnancy, the number of PNC visits, and delivery in a private health facility were significantly and positively associated with the probability of CS delivery.Women with a history of CS have a substantially higher probability of undergoing another CS.Moreover, it is also found years of schooling of women was associated with higher likelihood of CS delivery.Furthermore, having private health insurance cover compared to none, and being in a higher wealth index quintile were found to be positively associated with the probability of CS delivery.These findings are coherent with that of existing literature and thus enrich the existing knowledge of the factors associated with CS delivery in Indonesia.
Nevertheless, further studies that update the trend of CS deliveries and its determinants in Indonesia are recommended.