Out-of-pocket expenditure and catastrophic health spending on maternal care in public and private health centres in India: a comparative study of pre and post national health mission period

Background The National Health Mission (NHM), one of the largest publicly funded maternal health programs worldwide was initiated in 2005 to reduce maternal, neo-natal and infant mortality and out-of-pocket expenditure (OOPE) on maternal care in India. Though evidence suggests improvement in maternal and child health, little is known on the change in OOPE and catastrophic health spending (CHS) since the launch of NHM. Aim The aim of this paper is to provide a comprehensive estimate of OOPE and CHS on maternal care by public and private health providers in pre and post NHM periods. Data and method The unit data from the 60th and 71st rounds of National Sample Survey (NSS) is used in the analyses. Descriptive statistics is used to understand the differentials in OOPE and CHS. The CHS is estimated based on capacity to pay, derived from household consumption expenditure, the subsistence expenditure (based on state specific poverty line) and household OOPE on maternal care. Data of both rounds are pooled to understand the impact of NHM on OOPE and CHS. The log-linear regression model and the logit regression models adjusted for state fixed effect, clustering and socio-economic and demographic correlates are used in the analyses. Results Women availing themselves of ante natal, natal and post natal care (all three maternal care services) from public health centres have increased from 11% in 2004 to 31% by 2014 while that from private health centres had increased from 12% to 20% during the same period. The mean OOPE on all three maternal care services from public health centres was US$60 in pre-NHM and US$86 in post-NHM periods while that from private health center was US$170 and US$300 during the same period. Controlling for socioeconomic and demographic correlates, the OOPE on delivery care from public health center had not shown any significant increase in post NHM period. The OOPE on delivery care in private health center had increased by 5.6 times compared to that from public health centers in pre NHM period. Economic well-being of the households and educational attainment of women is positively and significantly associated with OOPE, linking OOPE and ability to pay. The extent of CHS on all three maternal care from public health centers had declined from 56% in pre NHM period to 29% in post NHM period while that from private health centres had declined from 56% to 47% during the same period. The odds of incurring CHS on institutional delivery in public health centers (OR .03, 95% CI 0.02, 06) and maternal care (OR 0.06, 95% CI 0.04, 0.07) suggest decline in CHS in the post NHM period. Women delivering in private health centres, residing in rural areas and poor households are more likely to face CHS on maternal care. Conclusion NHM has been successful in increasing maternal care and reducing the catastrophic health spending in public health centers. Regulating private health centres and continuing cash incentive under NHM is recommended.


Background
Reduction of maternal mortality, neonatal and under-five mortality and financial risk protection are three key health related targets of sustainable development goals (SDGs) [1]. Achieving health related SDGs required significant investment in maternal and child health to protect households from high out-of-pocket expenditure (OOPE) and catastrophic health spending (CHS). High OOPE is positively associated with CHS and reduced access to health services, increases untreated morbidity, reduce consumption of goods and services, and lead to longterm impoverishment [2]. The level of CHS varies across countries, among socio-economic groups and by nature and type of health services. Cross country studies suggest that households with low educational attainment, lower economic status, without health insurance and residing in rural areas are more likely to incur CHS [3][4][5][6]. The OOPE and CHS are high for maternal services in many developing countries including India [7,8].
Estimates of OOPE and CHS on health care are gaining increasing research and programmatic attention. A growing number of studies from developing countries suggest that the health care payment has increased the poverty level and affect the poor most [9][10][11]. Globally, OOPE studies on maternal care addressed socioeconomic and demographic differentials and estimated the incidence and correlates of CHS. The general findings suggest higher OOPE for caesarean delivery, complicated delivery, deliveries in private health centres and for higher socioeconomic groups [12][13][14][15][16]. The global progress in improvement of maternal and child health and reduction of CHS is contingent on India's success on these indicators.

Maternal and child health in pre and post NHM periods in India
India with one-sixth of the world's population, federal nature of governance and sustained economic growth, is undergoing health transition. The healthcare system in the country is characterized by the presence of both public (central, state and local government) and private health care providers, varying delivery structures and multiple systems of medicine [17]. While health is a state subject (within the state government) the central government helps in policy making, planning, guiding, evaluating and providing the funding to implement the national program. The utilization of health services from private health care providers is generally linked to the ability to pay and quality of care.
A decade ago, the state of maternal and child health in India was extremely poor. The maternal mortality ratio (MMR) and the infant mortality rate (IMR) were high. About three-fifths of the mothers were delivering without any medical assistance and women perceived cost as a major barrier to the utilization of maternal care. As a policy response, the Government of India launched the National Rural Health Mission (NRHM) in 2005 to improve the health system by providing universal access to equitable, affordable and quality health care. The NRHM intended to reduce maternal and child mortality and OOPE on maternal care in rural areas of 18 states that had poor health infrastructure and health indicators. Subsequently, the program was extended to all the states and urban areas in the country and renamed as National Health Mission (NHM). The key components of NHM are Janani Suraksha Yojana (JSY) and Janani Shishu Suraksha Karyakram (JSSK). The JSY is a cash incentive scheme provided to mothers for delivering at public health centres or accredited health centers [18]. The incentive to mothers varies in rural and urban areas and in low and high performing states. Besides, it covers the incentive to Accredited Social Health Activist (ASHA). The JSSK, launched in 2011 entitled all pregnant women to deliver in public health institutions absolutely free and free treatment to sick infants up to 1 year. Over ten million pregnant women are provided cash incentives under the scheme annually. Details of the NHM are documented elsewhere [19]. Figure 1 presents the trends in health budget and that of NHM over time at constant prices [1US$ = 65.43]. The NHM accounts for more than half of the health spending of the central government over time. In 2014-15, about US$ 2626 million (Rupees 16,809 crores) were spent on NHM [20]. The recently released National Health Policy, 2017 highlights the success of NHM in the public health system and aim to reduce IMR to 26 per 1000 live births by 2019 and MMR to 100 per 100,000 live births by 2020 [21].
Since the implementation of NHM in 2005, there has been significant progress in maternal and child health in the country. Trends in IMR and MMR suggest substantial improvements in the post NHM period; IMR declined from 58 in 2004 to 40 per 1000 live births in 2013 (31% decline in the post NHM period) [22,23] and MMR declined from 254 in 2004-06 to 167 per 100,000 live births in 2011-13 (34% decline) [24,25]. Process indicators such as prenatal care, institutional delivery and postnatal care have shown significant progress in India [26]. JSY had a significant impact on increasing antenatal and natal care and reducing perinatal and neo-natal deaths [27]. Several small-scale studies have examined the success and constraints of the program [28][29][30][31]. Besides improving maternal and child health survival, the NHM was intended to reduce the OOPE and CHS on delivery, prenatal and post natal care. Though a number of studies have examined the differentials and determinants of maternal care [32][33][34], there are limited studies on the economic burden of maternal care over time in India.
The first systematic attempt on estimating OOPE and CHS on maternal care for India suggest high CHS to poor, rural households and the less educated  [7]. The OOPE on delivery care varies largely across states and household characteristics [35]. Studies also found that maternal care in India placed a high economic burden on households and suggested reduction in OOPE to benefit the poor [36][37][38].
The aim of this paper is to estimate the OOPE and CHS on maternal care before the launch of the National Health Mission in India and a decade later. It addresses the research question whether the OOPE and CHS on maternal care have declined in the post NHM period? The paper has been conceptualized with the following rationale. First, largescale investment in the public health programs of developing countries has large opportunity costs. Though there has been significant reduction in maternal and child mortality and increase in maternal care utilization in public health centres in the post NHM period, there is no study that has examined the OOPE by public and private health care providers in the pre and post NHM period. Understanding the level of OOPE in the pre and post NHM periods will help assess the functioning of the program and will be of immense help to multiple stakeholders. Second, both public and private entities in India are providers of maternal care services and are guided by varying principles. Hence, any systematic analyses in understanding the effectiveness of the program should focus on disaggregated analyses by type of provider (public-private). Third, high OOPE on health care is associated with higher CHS and increasing poverty [39,40]. Studies also suggest that the medical care costs has been rising faster than the overall well-being of the households [41]. Reduction of CHS among the poor and marginalized was one of the main objectives of NHM and analyses on CHS need to be sensitive to account the health spending among the poor.
The paper is organized as follows: Section I gives a brief introduction of NHM and delineates the progress of maternal and child health in the post NHM period, Section II presents data and method, while Section III presents the results and Section IV provides discussion and conclusion.

Data
The unit data of the 25th Schedule of the 60th and 71st rounds of the National Sample Survey (henceforth referred to as 60th and 71st respectively) is primarily used in the analyses. Both rounds of the surveys are population based nationally representative surveys, similar with respect to design, content and coverage and provide comprehensive information on morbidity, health care and cost of hospitalization. The 60th round was held from January to June 2004 and the 71st round was held from January to June 2014. These time periods were best suited for analyses as 2004 was the pre NHM period and 2014 marked a decade since the implementation of the NHM (post NHM period). No health survey of NSS was conducted between 2004 and 2014. Information on natal care in 2014 was collected as part of hospitalization and missed the cost of deliveries conducted at home. We have created data at the women's level to compute the OOPE and at the household level to compute CHS. In 2014 sample, there were 236 households that had more than one woman who had availed institutional delivery and there were 96 such households in 2004. On the other hand the CHS is used at household level, derived from household consumption expenditure, subsistence expenditure and household's OOPE. The 60th round covered a total of 73,868 households and 383,338 individuals and the 71st round covered a total of 65,932 households and 333,104 individuals. The NSS uses a stratified multistage sampling design and samples are drawn from all states and union territories of India. Samples drawn are representative and provide robust estimates at the state level by rural and urban and selected characteristics. The sampling methodology and findings from these surveys are available in the respective reports [42,43]. Since the expenditure on maternal care is analyzed for 2004 and 2014, we have adjusted the expenditure at 2014 prices for valid comparison using the price deflator for rural (agricultural labourer) and urban areas (industrial worker) and at 2001 base prices [44]. Estimates on OOPE is presented in US$ at the 2014 exchange rate. Besides, in calculating CHS, we have used the state specific poverty line (for subsistence expenditure) for rural and urban areas as recommended by the Planning Commission for 2004-05 and 2011-12 [45]. The poverty estimates are derived from the household consumption expenditure data based on calories intake. The per capita calories intake of 2400 for rural areas and 2100 for urban areas are demarcated as poverty cut-off point in India and make poverty estimates comparable across states of India. The corresponding money value is labeled as poverty line. The poverty estimates in India are usually provided at 5-year interval and the 60th and 71st rounds of NSS (health surveys) are close to that of time period that estimates poverty in India. Mother is the unit of analyses for estimating OOPE while household is the unit of analyses for CHS. The sampling weights are used in the analyses for representativeness of the sample.

Outcome
OOPE and CHS are two outcome variables, computed for delivery care, and maternal care (pre-natal, natal and post-natal) by type of health care provider (public-private). In computing the maternal care from private and public sources, we consider only those women who availed these three services either from the public or private. The OOPE is defined as the expenditure incurred by the women during pre-natal, institutional delivery and post-natal care net of reimbursement. Expenditure on prenatal care, institutional delivery and postnatal care was directly available in the data set and were summed to obtain expenditure on total maternal care. The CHS is defined as the health spending over 40% of household's capacity to pay. The consumption expenditure data of the household and the state specific poverty line is used in deriving the household's capacity to pay.

Covariates
Individual and household level covariates are used in the analyses. The household characteristics pertain to the head of the household while individual characteristics pertain to the woman. The selection of covariates is guided by availability of variables in data set and literatures. The covariates included are age, residence (rural/urban), educational level, caste 1 (Scheduled Tribe, Scheduled Caste, Other Backward Classes and Others), religion (Hindu, Muslims and Others), monthly per capita consumption expenditure (MPCE) quintile and household type (labourer household and others). We have also introduced an interaction term in the regression model, namely, type of health care provider and time (pre and post NHM) to capture the effect of time and source of provider in explaining change in OOPE and CHS in India.

Analytical methods
Descriptive statistics, estimation of CHS, log linear regression and the logit regression models are used in the analyses. In literature, two alternative approaches are used in estimating CHS, both using capacity to pay (CTP). The approach suggested by Berki (1986) and later by Van Doorslar et al. (2007) defines CHS as a proportion of consumption expenditure (usually 10% and more) [46,47]. The limitation of this approach is its inability to account for the CHS of poorer sections as the poor spend less on health due to their lower ability to pay. Studies have demonstrated the limitation of this method in the empirical estimation of maternal expenditure [7]. The second approach by Xu et al. (2003) [3] derives CTP by deducting the subsistence expenditure (SE) and is largely used in literatures [48,49]. It defines CHS; Where X i is the consumption expenditure of i th household and f(X) is the subsistence expenditure of the population. The SE is estimated either using median food expenditure or as poverty line of the specific country/region. Unfortunately, the health surveys in India do not collect detailed consumption expenditure and therefore the food expenditure is not available. In such cases, we have used the state specific poverty line to account for the subsistence expenditure. This approach is sensible where the poor are concerned, as those households below poverty line are classified as incurring CHS if they incurred OOPE for maternal care. The cut-off point of CHS is normative and usually taken as 40% in literature. Thus, a household is said to incur CHS if its health spending exceeds 40% of its capacity to pay.
To understand the impact of NHM on OOPE and CHS, we have pooled the variables of both rounds of health surveys. Two types of regression models are used; a log linear regression model for OOPE and a logit regression model for CHS. The log linear regression model was estimated as OOPE was continuous variable and skewed in nature. The logit model was used as the CHS was dichotomous variable, 0 for not incurring catastrophic health spending and 1 for incurring catastrophic health spending. Both set of models were estimated for delivery care and maternal care. We clustered standard error by first stage sampling unit (village/urban blocks). Each of the regression models were adjusted for state and time (surveyyear) fixed effect. The fixed effects are captured at the state level because the states in India exhibit large variation in demographic, social, economic and health parameters. Also, health is a state subject and the state makes policies, programs and implement uniformly across the state. Besides, the state fixed effect model captures the unobserved factors in the regression model.
The regression models used for OOPE is defined as where α is the intercept, res is residence (rural/urban), age is age of the woman, edu is educational level of the mother, mpceqt is monthly per capita consumption expenditure quintile of the household, caste is the caste of the household, religion is religion of the household, htype is the type of household (labourer or nonlabourer household). INT_NHM_SOU is the interaction term computed based on time (pre / post NHM) and source of providers (public/private). These are public*preNHM, private*preNHM, public*postNHM and private*post NHM. The interaction term helps to understand the role of NHM in reducing the CHS in public/private health centres. A total of 36 states and union-territories are included in the analyses. The subscript i is used for ith woman. Similarly, the regression models used for CHS is defined as where π i is the probability of incurring catastrophic health spending for delivery/maternal care of ith household. The model estimates the log odds of incurring CHS adjusted for a set of explanatory variables. All explanatory variables are same as of Eq. (1) except age and education. Information on age and education of the head of the household is used, as household is the unit of analyses and there are cases where more than one woman had delivered from the same household. Results are presented with the help of regression coefficients, odds ratio and 95% CI.

Results
The mean age of women was 25. 52  Prenatal, natal and postnatal care in pre and post NHM periods  post-natal care. Since 2004, all three maternal care has recorded increase in public health centres. For example, prenatal care from public health centres has increased from 43% in 2004 to 55% in 2014 while that in private health centres has increased from 32% to 35% during the same period. One interesting pattern that emerges is the continuity of maternal care services and decline in switching of services from public health centres in the post NHM period. Among those women who had availed all three services in 2004, 11% availed these services only from public health centres, 12% availed services only from private health centres and 77% switched from public to private or vice versa (Fig. 3). In 2014, among mothers who availed maternal care services, 31% availed only from public health centres, 20% availed only from private health centres and 49% switched services between private and public health centres.
OOPE on prenatal, natal and postnatal care in pre and post NRHM periods Appendix 1 presents the differentials in OOPE on institutional delivery (natal care) and all three maternal care services in 2004 and 2014 by type of health care providers. The OOPE on total maternal care is presented for those who availed all three services either from public or private health centres exclusively, and those who switched between public and private service providers during pregnancy and childbirth. Among women who availed all three services in public health centres, the mean OOPE had increased   (0.05)-1)). The OOPE on institutional delivery and all three maternal care increases significantly with the economic well-being of the households. Compared to the poorest MPCE quintile, the OOPE on institutional delivery among the richest MPCE quintile was 55% higher (exp(0.44)-1)). The OOPE was higher among women with higher educational attainment and other than labourer households. The OOPE in post NHM period from public health center was 3% lower than that of pre NHM period (not statistically significant). The OOPE for institutional delivery from private health center in post NHM period was 5.62 times higher than that of public health center in pre NHM period. The OOPE for institutional delivery from private health center in post NHM period was even higher than that from private health center in pre NHM period. In post NHM period, the OOPE for all three maternal care from public health center had increased by 32% ((exp(0.28)-1) and statistically significant. In case of private health centers in post NHM period, the increase in OOPE on all three maternal care was about 5.62 times compared to that from public health center in pre NHM period. In general, the pattern on OOPE in maternal care is similar to that of institutional delivery.
Catastrophic health spending in pre and post NHM periods Table 3 presents the differentials of CHS on maternal care and delivery care by public and private health centres and socio demographic characteristics in the pre and post NHM periods. Among mothers who availed all three maternal care from public health centres, the CHS has declined from 56% in the pre NHM period to 29% in the post NHM period and that from private health centres declined from 56% to 47% during the same period. The decline in CHS was quite substantial among all economic groups (except the poorest MPCE quintile) and all educational groups. The CHS on delivery care in public health centres had declined from 51% in 2004 to 20.5% in 2014 and that in private health centres has declined from 52% to 34% during the same period. Overall, the level of CHS on delivery care has declined from 52% in 2004 to 25% in 2014. The CHS on delivery care and maternal care reduced for many sociodemographic characteristics. The CHS on maternal care and delivery care is negatively associated with the economic status of the household, irrespective of the type of service and time. Table 4 presents the odds of incurring CHS on all three maternal care services and institutional delivery. Age of the head of the household is a significant determinant of CHS for institutional delivery and all three maternal care. Urban households are less likely to incur CHS compared to rural households for institutional delivery (OR 0.29, 95% CI 0.26, 0.32) and all three maternal care services (OR 0.27, 95% CI 0.24, 0.31). The educational attainment of head of the household does not show consistent pattern for all three maternal care services. On the other hand, economic gradient is strong, negative and significant suggesting that CHS declines with economic wellbeing of the household. Type of households is not significant for maternal care services and institutional delivery. Maternal care (all three) in public health center in the post NHM period (OR 0.06, 95% CI 0.04, 0.07) are less probable to be catastrophic compared to public health center in pre NHM period. Similarly, all three maternal care services at private hospitals are significantly more likely to be catastrophic in the post NHM period compared to public health centres in the pre NHM period (OR 1.34, 95% CI 1.08, 1.67). In case of institutional delivery the odds of CHS from public (OR 0.03, 95% CI 0.02, 0.03) and private (OR 0.54, 95% CI 0.46, 0.63) health centers in post NHM period was significantly lower than public health centers in pre NHM period.

Discussion
The NHM in India was intended to increase the utilization of maternal care, reduce maternal mortality, neo-natal and infant mortality and reduce the OOPE and catastrophic spending on maternal care. Though evidence suggests an increase in the utilization of maternal care services and reduction in maternal and infant mortality in the post NHM period [50][51][52], there is no study that has examined the effectiveness of the program in reducing the OOPE on maternal care. Earlier studies were confined to a point of time or addressed particular services and none of these provided comparable estimates of OOPE in pre and post NHM by public and private health providers [7,35,38]. This study provides comprehensive estimates of OOPE and CHS in pre and post NHM periods in public and private health centres. The data set we have used is publicly available and the time period is best suited for analyses. The salient findings are as follows. First, the OOPE on delivery care in public health centres (unadjusted The OOPE on institutional delivery in private health centres was nearly four times higher than that in public health center in the pre-NHM period and increased by 6.5 times in the post NHM period. On controlling for socio-economic and demographic confounders and time, the OOPE on institutional delivery from public health center had not shown any significant increase in post NHM period. In case of private health centers, the increase in OOPE for institutional In the case of institutional deliveries, reduction of CHS in public health centres was experienced across all educational and economic classes. The decline in CHS was also noticeable in private health centres. With regard to all three maternal care services, the reduction of CHS was lower compared to that in delivery care but the level remained high. Fourth, the multivariate analyses confirmed decline in CHS in the post NHM period and a larger decline in the share of CHS in public health centres. Households residing in rural areas and with lower economic status are more likely to incur CHS and these findings are consistent with other studies [5,6,35,38]. The interaction of time and NHM suggests that compared to those who delivered in public health centres during pre NHM period, those delivered or availed all there maternal care services in the post NHM period from public health centers were less likely to incur CHS in India. However, in case women availed all three maternal services from private health center in post NHM period, they are more likely to incur CHS. These findings suggest that the NHM has been successful in increasing deliveries in public health centres and reduced the CHS in India. Increase in continuation of services in public health centres and reduction in switching from public to private healthcare providers is indicative of improvement in public health services on maternal care in India. The decline in OOPE from public health centres for the less educated and poor mothers may be attributed to the JSY under NHM. However, the use of pre-natal and postnatal has not recorded similar increase as that of institutional delivery in the post NHM period. This is possibly because the program priority under JSY was on increasing institutional delivery. The reduction of CHS is a reflection of the success of the program. Decline in CHS may be attributed both to the NHM and improvement in the economic well-being of the households. On the other hand, the differentials in OOPE among public and private health centres are large and there has been overall increase in OOPE on delivery care. This is possible because of increasing incidence of caesarian deliveries across the states of India. We acknowledge some limitations of the study. First, a time series analyses could not be feasible due to data constraint. The NSS health survey was conducted in interval of 10 years; 2004 and 2014. No health survey of NSS was conducted in intervening period and so the time series analysis is ruled out. Similarly, the difference in difference analyses is not feasible as NHM was implemented in 18 poor performing states for initial few years and then expanded to advanced states. Getting control group may not feasible, as there were large differentials between poor performing and advanced states of India. Second, the analyses could not be performed separately for caesarian deliveries and normal delivery as the type of delivery (normal/caesarian) was not recorded in the survey. Third, the quality of maternal care that is often linked to cost were not collected in the survey and could not be analyzed. The indirect cost on hospitalization has not been estimated. Besides, improvement in health infrastructure in the post NHM period is beyond the scope of the study.

Conclusion
Based on these findings, we conclude that the NHM is effective in increasing in utilization, continuation of services in public health centres and reducing OOPE and CHS in public health centres on maternal care. We suggest that the cash incentive under NHM should continue and private health care providers should be regulated with respect to pricing and quality of care. The program should focus on improving the quality of services in public health centres. Besides, we recommended that the forthcoming health survey (NSS) should integrate an abridged version of the consumption schedule, question on expenditure on home delivery and a separate code for caesarian and normal delivery is recommended.