Data sources

The present study was based on data from the MICS 2019 conducted by the Central Bureau of Statistics (CBS) from May to November 2019 in collaboration with the United Nations Children’s Fund (UNICEF). The survey applied a multistage (2-stage) sampling technique. The urban and rural areas within each province were made the main sampling strata. Kathmandu valley urban was included as a separate stratum. Within each stratum, primary sampling units (PSUs) were selected systematically with probability proportional to size. Furthermore, a household listing was carried out within the selected PSUs, identifying the households with and without children under five years. In total, 25 households with and without children under five were selected in each PSU through a systematic random sampling method. The households with children under five were oversampled for the assurance of strong representativeness of children under five in the sample, where 13 households with children under five and 12 households without were selected from the listing in each Enumeration Areas (EAs). The detailed methodology of the survey can be accessed from the Nepal MICS 2019 report [9]. The website of the MICS program makes the anonymized survey dataset freely available under request [13].

Participants and sample size

A total sample of 512 PSUs and 12,800 households was selected for the survey where 12,655 were interviewed. A total of 14,805 women aged 15–49 years were interviewed which included 6,658 mothers/caretakers of children under five years. After merging two different data sets i.e., mother and household, a total of 14,805 observations were found. 1,950 observations of married women aged 15–49 years with a live birth in the last two years by place of delivery of the most recent live birth were selected and retrieved. After removing five observations with missing values: the place of delivery (n = 1), health insurance (n = 3) and household head education (n = 1), a total of 1,945 observations remained. Finally, 13 observations which did not fit the criteria for the place of delivery were removed and 1,932 observations were considered in the final analysis.

Conceptual framework

We categorized the determinants of institutional delivery into external environment, household contextual and household individual groups based on extensive literature review and expert opinion. The conceptual framework for the study is shown in Fig 1.

Download:

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

Fig 1. Conceptual framework for institutional delivery service utilization in Nepal.

https://doi.org/10.1371/journal.pone.0292054.g001

Outcome variable

Our outcome variable was the place of delivery, that is, institutional delivery or home delivery. Delivery in the government hospital, government clinic/health centre, health post, private hospital, private clinic and maternity home or private medical institute was considered institutional delivery, whereas delivery in the respondent’s home or other’s home was considered home delivery.

Covariates

Altogether 13 characteristics relevant to the study were selected and divided into three different categories. External environment variables included the province of residence and area of residence. Household contextual variables included religion, ethnicity, household wealth index quintile, household head’s age, household head’s sex and household head’s educational status. Likewise, household individual variables included women’s media exposure, enrollment into health insurance, women’s age, women’s educational status and antenatal care (ANC) visits. The description and measurements of the explanatory variables are presented in Table 1.

Download:

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

Table 1. Description and measurements of study variables.

https://doi.org/10.1371/journal.pone.0292054.t001

Data management and analysis

After approval of the request by UNICEF MICS team, data was downloaded from MICS website survey page and received in Statistical Package for Social Science version 23 format. Two datasets namely the household dataset and the women dataset were used for this study. Required variables from household were merged into the women’s dataset. The variables were then created and recoded according to the objectives of the study using Statistical Analysis System OnDemand for Academics and syntax and output files were well documented.

Descriptive statistics were reported for all the variables. Women weight, cluster and strata were adjusted to calculate the descriptive statistics. At first, bivariate analyses were conducted where an unadjusted odds ratio with a 95% confidence interval was calculated, and variables significant at P<0.05 in bivariate analysis were considered as candidate variables for the multilevel analysis. Furthermore, the variables were categorized as external environment factors, contextual factors and individual factors for further analysis in multilevel regression. The multilevel nested structure of the analysis comprised 1,932 individuals grouped (level 1) nested into 504 PSU (level 2). The estimation of generalized linear mixed models, including random effects and model fitness, was carried out using the GLIMMIX procedure. Weight was adjusted and the cluster was placed at the subject statement. A parsimonious final model for institutional delivery was obtained where the significant variables in the previous models were carried over to the subsequent models as per the conceptual framework (Fig 1); predictors that remained statistically significant at 5% (P ⩽ .05) were retained in the analysis for adjustment in the next model. The first model consisted of external environment factors. Furthermore, household contextual factors were added in the second model, while individual factors were added in the third model. The covariance parameters were significant for all the models. The multicollinearity of the variables was tested before entering them into the models. No multicollinearity was observed among the independent variables. Adjusted odds ratios (aOR) with 95% confidence intervals (CI) were calculated to identify the variables with a higher likelihood of institutional delivery.

Ethical considerations

The survey protocol for MICS was approved by the Central Bureau of Statistics (CBS) as per the Statistical Act (1958) in September 2018. The Statistical Act enables CBS to carry out surveys according to the government’s ethics protocol without involving an institutional review board. Questionnaires were administered with comprehensive introductions and obtained verbal consent from respondents. Participants were explicitly informed of the survey’s voluntary nature, their right to decline answering specific questions or discontinuing the interview at any point, and the assurance of confidentiality and anonymity for their information.

Socio-economic and demographic characteristics of respondents

The socio-economic and demographic characteristics of respondents are presented in Table 2. The percentage of women attending institutional delivery was lower in Karnali and Madhesh provinces, namely 62.9% and 64.6%, respectively, while it was highest in Bagmati province (89.2%). Slightly over 84% of women had institutional delivery in urban areas, compared to 66.9% of women in rural areas. Both external environmental factors, such as province and area of residence, were significantly associated with the place of delivery.

Download:

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

Table 2. Socio-economic and demographic characteristics of respondents.

https://doi.org/10.1371/journal.pone.0292054.t002

For household contextual factors, approximately 80% of Hindus and 70% of non-Hindus delivered in institutions. Similarly, the percentage of institutional delivery was higher among women from relatively advantaged ethnic groups, at 87.5%, while only 72.5% of women from relatively less advantaged ethnic groups delivered in health facilities.

About 58% of women from the poorest wealth quintile delivered in institutions, while the percentage gradually increased with each wealth quintile. Almost 97% of women from the richest wealth quintile delivered in health facilities. The median age of the household head was 47 years for women who had health institution-based delivery. About 83% of women whose household head’s sex was female delivered in health facilities, whereas the percentage was lower among women whose household head’s sex was male (76.9%). The percentage of women attending institutional delivery was approximately 83% when the household head had received formal education, but it was 71.8% among women whose household head was uneducated. Religion, ethnicity, wealth quintile, and household head’s education showed a significant association with the place of delivery.

Likewise, for individual factors, almost 86% of the women who had adequate media exposure delivered their babies in health facilities, whereas about nearly 66% of the women who didn’t have adequate media exposure attended health institution-based delivery. Nearly 95% of the women who were covered with insurance had institutional delivery, however, the percentage was lower for the women who were not covered by health insurance (77.4%). Both media exposure and insurance were significantly associated with the place of delivery. The median age of women attending institutional delivery and home delivery was 25 years each. Regarding education, 84.6% of women who had attended formal education had institutional delivery, whereas only more than half of the women who were uneducated delivered in health facilities. The median number of ANC visits for women delivering in health facilities was four, whereas it was three for women delivering at home. Women’s age, women’s education and the number of ANC visits were significantly associated with the place of delivery. The percentage of women delivering in health institutions was 78.3% (95% CI: 75.3–81.1) while it was 21.7% (95% CI: 18.8–24.6) for home delivery.

Multilevel logistic regression between the place of delivery and contextual and individual factors

The result of the multilevel logistic regression model for the place of delivery is presented in Table 3. A parsimonious final model for institutional delivery was obtained by carrying the significant variables in the previous model to adjust in the subsequent model. External environment factors were fitted in the first model, while household contextual factors and individual factors were added in the second and the third models respectively.

Download:

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

Table 3. Multilevel logistic regression between the place of delivery with contextual and individual factors.

https://doi.org/10.1371/journal.pone.0292054.t003

In the first model, Province 1, Madhesh Province, Lumbini Province, Karnali Province and rural area were negatively associated with institutional delivery. Likewise, in the second model, Madhesh Province, rural area, non-Hindu religion, relatively less-advantaged ethnicity, and poor wealth groups were negatively associated with institutional delivery. In the final model (third), Madhesh Province, rural area, relatively less-advantaged ethnicity, and poor wealth groups were negatively associated with institutional delivery, while women education, and the number of ANC visits were positively associated with institutional delivery.

Women from Madhesh province compared to Bagmati province were less likely to deliver in health institution [aOR: 0.32, 95% CI: 0.17–0.61]. Similarly, women residing in rural areas were less likely to attend institutional delivery compared to women form urban areas (aOR: 0.55, 95% CI: 0.39–0.78). Likewise, women from relatively less-advantaged ethnic groups had 48% (aOR: 0.52, 95% CI: 0.35–0.76) lower chances of giving birth in health institution than women from relatively advantaged ethnic group. Moreover, the results showed that, compared to women from the richest household group, women from the poorest (aOR: 0.09, 95% CI: 0.04–0.22) and second wealth group (aOR: 0.29, 95% CI: 0.13–0.64) had a lower likelihood of giving birth in health institution. The odds of institutional delivery increased with the educational level where women with formal education were nearly twice as likely (aOR: 1.65, 95% CI: 1.16–2.35) to deliver in an institution over uneducated women. Moreover, the odds of having an institutional delivery increased by 59% for each additional ANC visit (aOR: 1.59, 95% CI: 1.43–1.75).

Influence of external environment factors on utilization of institutional delivery

Our study found a provincial disparity in the uptake of institutional delivery which corroborates with previous studies from Nepal, which reported uneven progress in closing the equity gap in the use of maternal health services in seven provinces of Nepal [12, 14, 15]. In support, disparities in the utilization of health institution-based delivery according to geographical regions have been reported from various countries as well [16–18]. The provincial differences in the utilization of delivery services in Nepal may be explained by geographical barriers, unavailability of and inaccessibility to health facilities, socio-economic status, and prevailing traditional beliefs about childbirth. In our study, women from Madhesh province were less likely to deliver in a health institution than the women from Bagmati province. Even though Madhesh province is largely plain in terms of topography and therefore has greater physical access to health services than other provinces, the lower likelihood of health institution delivery can be explained by poor socio-economic indicators. In addition, the underlying sociocultural practices and traditional beliefs could also be a major barrier that hinders women from accessing delivery services. The finding from the 2016 Nepal Demographic and Health Survey (NDHS) showed that nearly 80% of mothers in Madhesh province felt unnecessary to deliver in a health institution [19]. Many women consider the birthing process as a part of natural life events and only consider seeking institutional delivery care services only if complications arise [20].

In our study, women from rural areas were significantly less likely to deliver in a health institution. This finding is consistent with previous studies [21, 22]. The rural areas in Nepal are characterized by harsh geographical conditions, lack of transportation facilities, and unavailability of health services and health workers. A qualitative study by Shah et al. cited traditional socio-cultural norms and values disfavoring institutional deliveries, lack of access to birthing facilities, poor perception regarding quality of health services, insufficient financial incentives, and poor infrastructures and equipment at birthing centres as barriers among women from rural areas which deter them from delivering at health facilities [23]. A study conducted in rural areas of Nepal revealed that the majority of the birthing centres did not meet the required guidelines set by National Safe Motherhood Program (NSMP) and many facilities lacked the basic commodities, medicines and supplies to provide essential ANC, delivery and newborn services [24]. The low levels of availability, readiness, and coverage of obstetric services in rural areas of Nepal could be major reasons for urban-rural differentials in the uptake of institutional delivery services.

Influence of household contextual factors on utilization of institutional delivery

Several studies have identified caste/ethnicity as one of the important predictors of institutional delivery in Nepal [12, 25–27]. A similar finding has been reported in our study where women from relatively disadvantaged castes were less likely to deliver in a health institution. This could be explained by a social discrimination against women of relatively disadvantaged castes by service providers or other members of the community. A study from Nepal revealed that women’s experience of disrespect and abuse during labour and childbirth, was significantly associated with relatively disadvantaged castes [28]. These encounters of mistreatment in the health facilities at any stage in the continuum of care may deter women from seeking maternity care from health facilities and consequently, they might resort to home deliveries. Moreover, lower-caste women are found to be usually deprived of many socioeconomic and educational opportunities as well [29]. The fact that ethnicity has consistently been a factor that influences the uptake of institutional delivery among women suggests that maternal health programs of GoN do not adequately address the preexisting social structures and deeply ingrained cultural issues which are checking women from disadvantaged castes from institutional delivery.

The current study showed that women from poorer household are less likely to deliver in a health institution. Furthermore, the women from poorest households were 91% less likely to deliver in a health institution compared to the women from the richest households which indicates a huge gap in institutional delivery service utilization between the poorest and the richest wealth groups. Differentials in the utilization of delivery services by the household wealth status of women have been evident from previous studies conducted in some South Asian countries as well as Sub–Saharan Africa [30–32]. A recent study by Bhusal also concluded that utilization of institutional delivery was disproportionately higher among women from wealthy groups [33].

GoN introduced the maternity incentive scheme (MIS) in 2005, the Safe Delivery Incentive Scheme (SDIP) in 2006 and the Aama Program in 2009 to reduce the financial barriers to women seeking institutional delivery. Aama Program included free institutional delivery services as well as cash incentives upon completion of four ANC visits [34]. A study reported that institutional delivery increased by five folds after the implementation of the Aama program compared to no free delivery care (FDC) policy [35]. However, a study by Ensor et al. suggested that the beneficial impact of maternal financing policies in Nepal is skewed towards areas and households that are accessible and wealthy [36]. Therefore, the national program which aims to reduce financial barriers may have disproportionately benefited the non-poor household whilst only partly mitigating the financial burden faced by the women from the poorest households. Institutional delivery care costs more than the incentive offered and may not be enough to compensate full range of expenses incurred as a result of delivery which includes transportation costs, opportunity costs, and food and lodging for women from poor households [23, 37]. This existing financial barrier might still be prohibiting the women from the poorest households to seek delivery care services. Moreover, poorer women are more likely to experience distance-related barriers which further exacerbate their situation and deter them from seeking institutional delivery care services [38].

Influence of individual modifiable factors on utilization of institutional delivery

This study established women’s education as a significant predictor of institutional delivery. This finding aligns with the previous studies [12, 14, 15, 39]. A study by Mehata et al. highlighted that out of all socio-demographic variables, maternal education was the most powerful predictor of the use of maternal health services [27]. Schools are an indispensable source whereby female youths can learn about reproductive and maternal health care. They are also more likely to have good knowledge on the importance of skilled birth attendance at delivery which makes them more likely to utilize institutional delivery services. Furthermore, educated women are likely to have a better status in their households along with economic independence which enhances their decision-making power to demand, seek and utilize delivery care services [27].

Our study found the likelihood of uptake of institutional delivery increased with an increase in the number of ANC visits. This finding aligns with previous studies [12, 21]. Women are suggested about the benefits of institutional delivery to mother and child and informed about the cash incentives for institutional delivery during ANC visits. However, a qualitative study by Bhandari et al. found that during regular antenatal visits, a significant proportion of women were not clearly advised to seek institutional delivery care and were not given adequate information about the available cash incentives for institutional delivery [19]. Furthermore, a study by Dixit et al. emphasizes that mere antenatal visits do not ensure the reception of appropriate ANC and therefore decent quality and appropriate delivery of information must be ensured to encourage women to uptake the next level of services [40]. It is suggested that when women are satisfied with the content of care they receive during their ANC visits; they are more likely to utilize institutional delivery services.

Education of the household head was not found to be a significant predictor of the place of delivery in our study. Similar findings were observed in another study from Nepal [12]. However, a plethora of studies have determined husband’s education as a significant predictor for the utilization of maternal health services [17, 41–43]. While our study could not explicitly explore the role of husbands in institutional delivery uptake, their significance as decision-makers, especially in financial matters, and as companions accompanying women to healthcare facilities should not be underestimated. In certain instances, husbands may be the sole available support for transporting their wives to healthcare institutions for delivery. Consequently, their perception of maternal needs is of critical importance as it can influence delays in seeking maternal health services [44, 45]. Given that husband’s involvement directly affects access to and utilization of maternal healthcare services by pregnant women, further investigation of this dimension is imperative in future research, particularly within predominantly patriarchal societies like Nepal.

Women’s exposure to mass media was not significantly associated with the place of delivery as well. However, exposure to mass media can promote their utilization of maternal health services as they are an important source of information about health services and their outcomes, which in turn may influence people’s decision to seek health care services. Mass media campaigns play a key role in disseminating maternal health care information to mothers having low education level, which is especially useful in the context of South Asia as many mothers have limited formal education or are illiterate. A study among four South Asian countries showed mothers’ exposure to mass media increased the likelihood of delivery by a skilled birth attendant by 24–53% [46].