Institutional delivery in public and private sectors in South Asia: a comparative analysis of prospective data from four demographic surveillance sites

We sourced information on maternity care from the surveillance systems of four cluster randomized controlled trials conducted in socio-economically disadvantaged communities between 2005 and 2011. They were run in three underserved rural districts in Bangladesh (Bogra, Maulvibazaar, and Faridpur) [32], in Dhanusha district in the southern Nepal Terai [33], in rural Jharkhand and Odisha states in eastern India [34], and in informal settlements in Mumbai [35]. Table 1 describes the characteristics of each study, including the types of health facilities available.

Nepal attained the MDG 4 target for reduction in child mortality and marginally attained the MDG 5 target for reduction in MMR. There has been some consideration of how this was achieved, given a widespread breakdown in governance during the insurrection, but a national Maternity Incentive Scheme (MIS) was implemented in 2005 and was operational during data collection. Dhanusha district had one zonal hospital, a private medical college hospital equipped for comprehensive obstetric care, public health posts, and a range of small private facilities. At the time of data collection the MIS was available only through public health providers.

India almost attained the MDG 4 and MDG 5 targets. The rural data from India were collected between 2005 and 2008. The JSY was launched in 2005 and beginning to be operational in study areas towards the end of 2008, so that its impact was unlikely to be reflected. In the urban setting, mothers with Below Poverty Line cards were eligible for the JSY cash incentive for delivery at a health facility [36]. The urban India study included informal settlements in Mumbai with a wealth of public and private providers [37]. Public facilities included primary, secondary, and tertiary institutions run by the municipal corporation. Municipal and government health infrastructure together constituted half of the inpatient care available in the city, the rest being provided by private hospitals and clinics.

The participants were women of reproductive age (15–49 years) who delivered in the study areas during the data collection period and who consented to be interviewed around six weeks after delivery. All the sites operated vital registration systems that monitored births, stillbirths, and neonatal deaths identified by female community-based informants covering 250–350 households each. Reports were verified by trained interviewers and informants remunerated for correct identification. In Bangladesh, rural India, and urban India, women were visited six weeks after delivery for a postpartum interview. In Nepal, births in the study area were registered and interviews conducted after all births in small clusters and in a random sample of 10 births per month in larger clusters. Interviews were predominantly based on closed questions about antenatal, delivery, and postnatal events.

The outcome of interest was institutional delivery in the public or private sector, based on the reported type or name of the institution.

We chose variables purposively from the available dataset, to reflect socioeconomic position (household asset index, maternal schooling) and demography (maternal age, parity). Household economic position was defined by tertiles of an asset index developed from standardized weights of the first component of a principal components analysis [38, 39]. Scores were calculated separately for each site. Maternal education was categorized as none, primary, secondary, or higher. Maternal age was categorized as under 20, 20–24, 25–29, or 30+ years. Gravidity was represented by a binary variable: primigravid or multigravid.

The original trials were designed to evaluate the impact of participatory women’s groups on maternal and neonatal health outcomes [32‐35, 40], and we restricted our analysis to residents of control areas who reported a birth in the trial period. We summarized sites of delivery and choice of provider by socio-economic and socio-demographic position, with frequencies and percentages. The denominators were all deliveries for institutional delivery, and institutional deliveries for private sector delivery. We entered private sector institutional delivery as the dependent variable in a series of univariable logistic regression models, including a random effect for cluster, for each of the four independent variables. A multivariable model included all four of them as covariates. The analysis was run for each trial separately in Stata13 (Stata, College Station, TX, USA).

Data for the study originated with trials that had received ethical approval from the ethics committee of the Diabetic Association of Bangladesh, an independent ethics committee in Jamshedpur, Jharkhand, India, the Nepal Health Research Council, the Municipal Corporation of Greater Mumbai and the Independent Ethics Committee for Research on Human Subjects (Mumbai, India), and the ethics committee of the Institute of Child Health, University College London.

Our analysis of data from 52 750 births in three South Asian countries over six years showed broad variation in rates of institutional delivery and public-private split across study sites and between socio-economic and socio-demographic groups. There were substantial differences in institutional delivery proportion between rural and urban areas, by a factor of about three. There were also socio-economic and socio-demographic disparities in institutional maternity care, with women from lower socio-economic and educational backgrounds less likely to have institutional delivery. Use of the private sector differed between sites. In Bangladesh and urban India, it was greater for wealthier, better educated, and older women. The opposite was observed in rural India and, perhaps, in rural Nepal (the numbers were small), in which greater proportions of poorer, less educated, multigravid women delivered at private rather than public facilities.

Our analysis had some limitations. All the participants could be classified as poor according to international metrics. We used principal components analysis to generate asset indices for each site, but the index distributions were highly positively skewed to a degree to which normalization by transformation was not possible. For this reason, quantiles of asset index did not yield groups of equal size from some of the sites. Secondly, classification of institutions as public or private was not possible using the raw data in <1 % of cases. We should also be aware of the variable nature of private care, particularly in terms of its categorization as for-profit or not-for-profit. Institutions such as charitable trusts and non-government organisations were classified as private when in some cases – Bangladesh being a good example – the cost to clients may have been similar to the costs of public sector care. Finally, there have been increases in institutional delivery since the study period. More recent data from the rural India site, for example, suggest that the proportion had reached 50 % by 2011 [41].

The finding of low prevalence of institutional delivery in South Asian countries is consistent with previous research. At three of the four sites, less than one-third of births were institutional. In rural Bangladesh, institutional delivery rates in the first and second phases were low and corroborate national Demographic and Health Survey (DHS) data for the corresponding periods (11 % in 2007 and 23 % in 2011) [4, 42]. In Nepal, home birth continues to predominate, especially in rural areas, and our findings validate the low proportion of institutional delivery reported by the Ministry of Health and Population (25 %) [8]. Neither 24-h comprehensive obstetric care nor the Maternity Incentive Scheme were available at private health facilities in the Nepal site at the time of the study, which may explain the lower usage in this area. In rural India, one-in-five births was institutional, compared with a District Level Household and Facility Survey (DLHS) national figure of 38 % in 2007–8 [43]. This difference is probably due to the fact that the Jharkhand and Odisha sites were home to a high proportion of underserved tribal communities with poor access to health services. In contrast, residents of urban India have access to a range of health services in both public and private sectors. Most of the births in urban Mumbai were institutional (92 %), and this agrees with roughly concomitant data from the third National Family Health Survey (83 %) [44]. The urban-rural differences were somewhat predictable. A systematic review of determinants of maternal health service use found that urban, wealthier women were more likely to have institutional deliveries [45]. Explanations include the fact that women in urban areas tend to be better educated, live nearer to health services, have access to more extensive transport systems, and may have greater autonomy and access to information [46, 47].

Our findings support those of studies that show economic conditions to be an important predictor of institutional delivery [48, 49]. Women of low economic position are less likely to use maternal health services, and we observed a systematic increase in institutional delivery with increasing wealth at all sites. Our analysis also corroborates the findings of studies which suggest that maternal education [50], household wealth, and urban-rural residence have important and consistent effects on utilization of health services for delivery [23, 28, 51‐53]. Education is a key determinant of access and utilization. Fewer women had schooling in rural areas of Nepal (26 %) and India (32 %) and we saw fewer institutional deliveries in these areas. In comparison, a higher proportion of women reported institutional delivery in Bangladesh, where educational attainment was higher (72 % of women reported some schooling in phase 1 and 82 % in phase 2). The proportion of institutional delivery was also higher in urban India, where 56 % of women reported completing secondary education. Other determinants of healthcare uptake include age and parity. Most of our study participants were young and one-third were primigravid. Several South Asian studies have shown that women in their first pregnancies are more likely to deliver at a health facility and with a skilled birth attendant [8, 23, 27, 54]. Our findings suggested a substantial influence of parity on utilization of delivery care, with younger women more likely to have their first birth in a health facility.

Private sector delivery is increasing over time, with positive associations with wealth and education, previous private sector delivery, and – to a lesser extent – primiparity [55]. In rural Nepal, Bangladesh, and urban India, the public sector was still a major provider of care for women who chose institutional delivery. Availability of better public health infrastructure combined with accessibility may have led to increased utilization compared with the private sector in urban informal settlements in India. A shift from public sector to private sector delivery was seen in the second phase of the Bangladesh study.

Private sector healthcare is heterogeneous in the South Asian context of medical pluralism. Bangladesh is served by a mix of public and private sector providers in which not-for-profit facilities play a major part. India, by contrast, has a dominant private sector. Around 70 % of total health expenditure in 2012 was private, 61 % was out-of-pocket, and 75 % of outpatient visits and 62 % of inpatient episodes were in the private sector. Out-of-pocket expenditure at public sector facilities was low, at 2 % [56]. Nepal can be classified as having a “stratified private sector shaped by low incomes and public sector characteristics” [57]. Around 61 % of total health expenditure in 2012 was private, 49 % was out-of-pocket, and 65 % of outpatient visits and 46 % of inpatient episodes were in the private sector. Out-of-pocket expenditure at public sector facilities was 7 % [57].

Bangladesh has seen a steady increase in institutional deliveries [50]. DHS analyses suggest a rise from 2 % in 1996 to 17 % in 2011 [20], most of which was accounted for by use of the private sector [55]. Wealth-related inequality in institutional delivery seems to have fallen [20, 50], more rapidly in urban than in rural areas, and the private sector is becoming the major provider of maternity care [58]. Much of this private provision is not-for-profit [59], and the nongovernmental health sector is now double the size of the government sector [60]. Similar trends were observed in rural India, where more than three-quarters of institutional deliveries were in private facilities. DHS estimates suggest an increase in institutional delivery of about 7 % between 1998 and 2005, most of which was accounted for by the private sector [55]. In this case, private care is predominantly for-profit. Unlike in Bangladesh and India, most of the increase in institutional deliveries in Nepal was accounted for by the public sector [55], although private delivery increased from 1 % in 1996 to 10 % in 2011 [20].

Healthcare quality can be seen as a combination of service quality (the user experience) and technical quality (the clinical process). Service quality appears to be higher in the private sector [61‐64], and users generally consider private care to be better than public in terms of responsiveness, longer opening hours, appointment systems, shorter waiting times, confidentiality, cleanliness, and privacy [65, 66]. These qualities may appeal to those who can afford them because they respond to women’s busy schedules and desire for positive experiences of health care [67]. The technical quality of private sector services may not be as good as in the public sector [63, 64], but this does not apply across the board: mapping of health facility signal functions across Bangladesh in 2007–2008 suggested that emergency obstetric services such as caesarean section and blood transfusion were available at a higher proportion of private than of public sector health facilities [68], and recent considerations have emphasized the benefits of pluralism [60]. In rural India, and possibly in rural Nepal, concerns about the availability and quality of care may have exacerbated other geographic, financial and social barriers, deterring women from delivering in health facilities unless they experienced antepartum or intrapartum complications [69, 70]. Comprehensive obstetric care was available in the Nepal site only at the regional public hospital and one or two private facilities. If institutional deliveries were often a response to anticipated or actual complications, women may have preferred the private sector if they thought that skilled care, and particularly emergency obstetric care, would be more readily available there.

Economic status is believed to be the strongest influence on choosing between a private-for-profit or public facility for institutional delivery [71]. In Bangladesh and urban India, women from wealthier backgrounds were more likely to deliver at private facilities [72]. In rural Nepal and India, amongst the subset of the population that delivered at institutions that were less poor, better educated, more likely to be younger and primigravida than those delivering at home, economic status was negatively associated with private care. Wealthier women in rural India may have had access to better quality public services, or have been more empowered to navigate their way through the public system. It is also possible that poorer women and those from tribal and Scheduled Caste communities may have feared discrimination in public facilities and avoided them as a result [62, 73‐75], despite the burden of out-of-pocket expenditure on private care [76]. Incentive programs have increased institutional delivery in the public sector, but there is a pressing need to monitor quality of care [16]. A recent evaluation of the JSY suggests that the scheme has led to an increase in births in public facilities and substitution away from the private sector. Importantly, poorer women were more likely than wealthier women to give birth in a public health facility in response to the scheme [77], although it did not compensate fully for the financial stress of complicated deliveries [78]. In rural Nepal, the Maternity Incentive Scheme was introduced around the beginning of our data collection and at that time less poor and more educated women were more likely to access it [79]. Since the incentive was only available through public providers, it may have been that over most of the data collection period better-educated and less poor women came to know of it and responded to it by choosing public sector institutional delivery, whilst poorer families were less aware of the scheme and sought private care when they perceived deliveries to be high-risk.