Incidence and correlates of catastrophic maternal health care expenditure in India

Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine ß The Author 2009; all rights reserved. Advance Access publication 17 August 2009 Health Policy and Planning 2009;24:445 456 doi:10.1093/heapol/czp032 Incidence and correlates of catastrophic maternal health care in India Sekhar Bonu, 1 * Indu Bhushan, 2 Manju Rani 3 and Ian Anderson 4 Accepted 5 May 2009 Keywords Using data from the 60 th round of the National Sample Survey of India (2004), the study investigates the incidence and correlates of catastrophic maternal (ME) in India. Data on ME come from 6879 births that took place during 365 days prior to the survey. The study adapts earlier definitions and methods for catastrophic total health care to measure catastrophic ME as: (i) maternal health care more than 10% of the annual normative household consumption (ME-1), and (ii) maternal health care more than 40% of the annual capacity to pay (ME-2). The capacity to pay was derived by subtracting state-wise poverty-line household from household consumption. The average maternal varied by place of delivery: US$9.5, US$24.7 and US$104.3 for birth at home, in a public facility and in a private facility, respectively. Sixteen per cent of households incurred ME of more than 10% of total household consumption (ME-1), while 51% households incurred ME of more than 40% of household capacity to pay (ME-2). While incidence of ME-1 increased with income decile, the reverse was observed for ME-2, reflecting higher non-utilization of institutional maternal care and its non-affordability among poorer households. All the households from the poorest decile and 99% from the second poorest decile paid more than 40% of their capacity to pay. Multivariate regression results indicate that antenatal care and delivery care in private facilities increased the chances of ME-1 and ME-2 (P < 0.001). Measuring maternal against capacity to pay (ME-2) may be better than measuring it as a proportion of overall household when assessing financial constraints in the use of maternal services. Improving the performance of the public sector, appropriate regulation of and partnership with the private sector, and effective direct cash transfers to pregnant women in the poorest households may increase utilization of maternal services and reduce the financial distress associated with ME. India, maternal health,, catastrophic, utilization 1 Principal Urban Development Specialist, South Asia Department, Asian Development Bank, Manila, Philippines. 2 Chairman, Health Community of Practice & Director, Strategy and Policy Department, Asian Development Bank, Manila, Philippines. E-mail: ibhushan@adb.org 3 Scientist, Western Pacific Regional Office, World Health Organization, Manila, Philippines. E-mail: ranim@wpro.who.int 4 Principal Advisor, Regional and Sustainable Development Department, Asian Development Bank, Manila, Philippines. E-mail: ianderson@adb.org * Corresponding author. Principal Urban Development Specialist, South Asia Department, Asian Development Bank, 6 ADB Avenue, Mandaluyong City, Manila, Philippines. Tel: þ632-632 5628. Fax: þ632-636 2293. E-mail: sbonu@adb.org 445

446 HEALTH POLICY AND PLANNING KEY MESSAGES Prevalence of catastrophic maternal health care is high in India. Calculating maternal in relation to household capacity to pay is likely to provide better information on the proportion of households that may need direct assistance for maternal services than an indicator that measures maternal as a proportion of total household consumption. Direct cash support may be required for the poorest households to overcome their financial constraints to access maternal services and to give women from poor households the choice to seek care from either private or public facilities. Introduction Notwithstanding decades of global and national efforts, maternity remains a high-risk event for too many women and their newborn children in developing countries (Starrs 2006). This has led to renewed calls for action to improve the plight of pregnant women in developing countries (Donnay 2000; Borghi et al. 2006a; Filippi et al. 2006). Millennium Development Goal 5 bolstered the global resolve to reduce maternal mortality by 75% between 1990 and 2015 (United Nations 2000). The risk of maternal mortality and morbidity and its links to poverty is an important challenge in India. Despite accelerated economic growth in the past decade, maternal mortality remains stubbornly high in India, accounting for 136 000 maternal deaths and 1 million neonatal deaths each year (World Health Organization 2005). Utilization of institution and home-based antenatal care (ANC), delivery care, and postnatal care (PNC) continues to be low, especially in remote and scattered habitations (Peters et al. 2002; International Institute for Population Sciences 2007a, 2007b). The Report of the Working Group on Health of Women and Children for the Government of India highlighted the missed targets for reduction of the maternal and infant mortality rates in the 10th five-year plan (2002 07) (Government of India 2006a). The Government of India has reaffirmed its commitment to reduce maternal mortality from 4 per 1000 births to 1 per 1000 births in the 11th plan period (2008 12) (Government of India 2006b). A number of factors help to explain low maternal health care use and poor maternal health outcomes in India. Rural health care, in most states, is marked by absenteeism of doctors/health providers, low levels of skills (among nursing and paramedical staff and other formal and informal providers of maternal care), shortage of medicines, inadequate supervision/monitoring and callous attitudes of public health care providers (Government of India 2006a). Factors outside the health sector (e.g. poor maternal nutrition, short birth spacing, early age at marriage, low levels of education in women and skewed gender relations) also adversely impact maternal health outcomes (Filippi et al. 2006; Ronsmans et al. 2006). In addition, health care cost and financing of maternal services can be key determinants of the utilization of maternal services. Maternal can lead to financial distress for poor households. A number of studies have explored the levels of maternal in South Asia including India (Nahar and Costello 1998; McCord et al. 2001; Balaji et al. 2003; McCord and Chowdhury 2003; Afsana 2004; Borghi et al. 2004; Duggal 2004; Sharma et al. 2005; Borghi et al. 2006b; Borghi et al. 2006c). In Bangladesh and Nepal, household maternal health care was found to range from US$20 to US$400 (Nahar and Costello 1998; McCord and Chowdhury 2003; Afsana 2004; Borghi et al. 2004; Borghi et al. 2006b; Borghi et al. 2006c). Likewise, wide variation in maternal health care was reported by studies in India (McCord et al. 2001; Balaji et al. 2003; Duggal 2004; Sharma et al. 2005). Reasons for the wide range of maternal in the studies from India, among others, relate to the different study periods, study settings, client and provider background, and package of services covered by the study. The studies from India suffer from two major limitations. First, most of these studies are based on small, localized geographical areas, which are hard to generalize to other areas given the geographic variability and complexity of India. Second, maternal was calculated in a standalone manner without reference to overall household. The important question that still needs to be answered empirically is whether maternal s are large enough to require consumption of other goods to be sacrificed, possibly including basic needs, or to prevent use of maternal health care, especially when an emergency arises. The idea behind the concept is that such high spending implies an excessive opportunity cost of other consumption in the short or long term and may lead to under-estimation of overall levels of poverty measured on the basis of total household consumption (Berki 1986; Wyszewianski 1996; Murray et al. 2000; Wagstaff and van Doorslaer 2003; Xu et al. 2003). Such levels of health care s are labelled as catastrophic in health finance literature. Generally, the assessment for catastrophic health care s has been limited to total health care, and more specifically to curative health care in a defined reference period (i.e. 1 year). The catastrophic health care s have usually been defined by a rather arbitrary fraction of total household (10%) or of total net of subsistence (40%) during the same reference period (Berki 1986; Wyszewianski 1996; Murray et al. 2000; Kawabata et al. 2002; Wagstaff and van Doorslaer 2003; Xu et al. 2003; Bonu et al. 2007). The second definition partially overcomes the limitation of the first definition in which severity of budget constraint means that most resources are absorbed by the items essential to

CATASTROPHIC MATERNAL EXPENDITURE IN INDIA 447 sustenance, such as food, leaving little to spend on health care. Almost no studies have examined and applied the same concept to maternal health care s exclusively to assess the extent of catastrophic maternal s. Using data from a nationally representative sample survey and the same underlying conceptual framework as used for catastrophic total health, with some modifications and adaptations (as explained in the methods section), the study attempts to investigate the incidence and correlates of catastrophic maternal, and to assess its policy implications. Methods Data The study uses data from the 25 th schedule of the 60th round of the National Sample Survey (NSS), conducted from January to June 2004 in India. The survey was nationally representative except for the exclusion of a few interior areas of India. The NSS followed a stratified two-stage design with sampling of census villages in the rural areas and the NSS urban frame survey blocks in the urban areas in the first stage, followed by sampling of households in the second stage (NSSO 2006). A total of 73 868 households were sampled. Information on household consumer was elicited through a short set of five questions on consumption aggregates, rather than through a detailed listing of consumption items, in the 30 days prior to survey. The five items used for calculating the household consumption were: (i) purchases; (ii) home-produced stock; (iii) receipts from exchange of goods and services; (iv) gifts and loans; and (v) free collection. This rather crude determination of household total from relatively few consumption aggregates over a 30-day reference period remains a limitation and is discussed later. Data on total maternal health care (ANC, delivery care and PNC) were collected for all the ever-married women aged 15 49 years who were pregnant during the 365 days prior to the survey. Information was elicited on the aggregate s incurred on different types of services (ANC, childbirth and PNC), and the source of care public or private for ANC and PNC, and home, public or private for delivery care. No information was elicited on for abortion care. For both household and maternal s, efforts were made to interview all the adult male members personally in addition to the women to take care of potential underestimation if only women were interviewed and if all financial transactions were dealt with by the household head (NSSO 2006). In all, 9664 women reported being pregnant during the 365 days prior to survey, and 6879 women reported giving birth in the reference period. The study was restricted to the sub-sample of the women (6879) who gave birth during the reference period to capture full maternal health care. Outcome variables The study uses two outcome variables to model catastrophic maternal using two different measurement methods borrowed and adapted from the literature on catastrophic health s (Berki 1986; Wyszewianski 1996; Murray et al. 2000; Kawabata et al. 2002; Wagstaff and van Doorslaer 2003; Xu et al. 2003; Bonu et al. 2007). However, the methods used in this study differ in two key ways from the traditional literature on catastrophic health s. First, the reference period for the maternal was different from that for household and could theoretically spread over 9 months prior to the survey depending upon the timing of the birth. Maternal was not included in household ; rather the relationship of maternal was assessed against longer-run normative household. By comparison, other crosssectional studies on catastrophic health in the literature include total health in total household with the same reference period. Our method overestimates the incidence compared with the traditional method of calculating incidence of catastrophe where health s are included in both numerator and denominator. Second, most of the earlier studies investigated catastrophic health s mainly incurred for curative services. This study focuses on maternal health care services, which could be considered as preventive services where the option of home delivery can be considered as foregoing preventive institutional care. (Maternal health care if perceived as a high cost and unaffordable service is likely to lead to non-utilization of services, leading to home delivery.) Accordingly, the two outcome variables are: Maternal Expenditure-1 (ME-1) is a binary outcome variable coded as 1 if total maternal health care is more than 10% of annual household consumption ; and coded as 0 if equal to or less than 10%. The 10% cut-off for this outcome variable is based on the widely used cut-off for defining catastrophic health, i.e. when a household s annual health care exceeds 10% of total annual household (Berki 1986; Wyszewianski 1996; Murray et al. 2000; Wagstaff and van Doorslaer 2003; Xu et al. 2003). Maternal Expenditure-2 (ME-2) is a binary outcome variable comparing maternal health care to household capacity to pay, calculated as total annual household consumption minus subsistence such as on food and other basic needs. The capacity to pay of the household was derived by subtracting state-wise household poverty-line adjusted to household size from household consumption (NSSO 2002a; NSSO 2002b). Capacity to pay was adjusted to zero for households below the poverty line. Wagstaff and van Doorslaer (2003) and Xu et al. (2003) suggested subtracting actual observed food instead of aggregate poverty-line to overcome negative capacity to pay problems. However, in the absence of disaggregated data on individual items (e.g. food and non-food items) in this particular survey, as explained in the earlier section, using the poverty line as a proxy for subsistence to derive capacity to pay was the best alternative closest to the suggested definition. Inflation-adjusted national poverty line data for the year 2004 05, as published by Government of India, was multiplied by household size (giving equal weight to all

448 HEALTH POLICY AND PLANNING household members) to obtain the household-level subsistence (Bonu et al. 2007). ME-2 is coded as 1 if total maternal health care is more than 40% of the annual capacity to pay ; and is coded 0 if it is equal to or less than 40%. Construction of this outcome variable is based on the cut-off used for defining catastrophic health in the existing health finance literature on capacity to pay (Kawabata et al. 2002; Xu et al. 2003). Independent variables The selection of independent variables was guided by existing literature as well as social, cultural, political and administrative aspects specific to India (Berki 1986; Kawabata et al. 2002; Xu et al. 2003; Russell 2004; Bonu et al. 2005; Su et al. 2006). The independent variables used in the study can be classified as individual, household, community and maternal health care service variables. The individual level variables include age and education of the woman. The household variables include social group (as measured by caste), religion and income (as measured by household consumption ), which are important markers of social and economic disparities in India (Agarwal 1997; Dyson and Moore 1983; Galanter 1984; Ghurye 1992; Bonu et al. 2005). Women were classified into household consumption quintiles and deciles for examining the relationship of ME with income status. Urban-rural residence and states (or provinces) were included as the community level variables. States in India have an important role in the provision of health services, with varying policies and programme implementation. Hence, a dummy variable for state was included to control for differentials across the various states in the multivariate analysis. The maternal health service variables include source of care (public/private/ home). Statistical methods Bivariate analysis examined the unadjusted association of various independent variables with both the outcome variables defined earlier. Multivariate logistic regression models were fitted to assess the independent association of the outcome variables with the selected independent variables as follows: Logitðp ij Þ¼ þ I ij i þ M ij m þ Hh ij h þ C j j P ij is the odds of high maternal health care for the ith women who delivered in the jth community. I ij is a vector of individual-level characteristics for the ith mother in the jth community, M ij is a vector of maternal health service characteristics for the ith mother in the jth community, Hh ij is a vector of household characteristics of the ith mother in the jth community, and C j is a vector of community characteristics for the jth community. All the estimates and the standard errors were adjusted for the multistage sampling design and clustering at the primary sampling unit, and were weighted at national level to give results that are unbiased and representative of the population (White 1980; White 1982). Stata version 10 was used for the analysis (StataCorp 2007). Results Overall, 73% and 64% of the women obtained ANC and PNC, respectively. Only 43% of women gave birth in a health facility (21% in a public and 22% in a private facility). Table 1 gives the results of bivariate analysis. The utilization of ANC, delivery care and PNC by type of provider varied significantly by community, household and individual factors. Private providers were preferred over public providers in urban areas for all maternity services ANC, delivery and PNC. In rural areas, this was the case only for PNC. Use of ANC in public facilities varied from as low as 9% in Bihar to 85% in Himachal Pradesh, while ANC in private facilities varied from 8% in Himachal Pradesh to 68% in Kerala & Goa. Likewise, use of public sector delivery care varied from 4% in Uttar Pradesh and Bihar to 70% in Union Territories. Public sector PNC varied from as low as 10% in Bihar to 64% in Himachal Pradesh, and private sector PNC varied from 10% in Himachal Pradesh to 60% in Kerala & Goa. Women from households in the richest decile sought ANC more often from private facilities than public, and the reverse was the case for women from poorer deciles. Women from the poorest three deciles gave birth more often in public facilities. However, women in all the deciles used private providers more than public providers for PNC services. Among the social groups, the scheduled tribes, which are the poorest in India, utilized public providers services for ANC, delivery and PNC more than private providers. The average maternal was estimated at 2272 Indian rupees (Rs) (US$50.5), which included Rs467 ($10.4) for ANC, Rs1519 ($33.8) for birth care, and Rs286 ($6.3) for PNC. (A consistent exchange rate of 1 US$ ¼ Rs45 is used in the paper.) Average maternal health care incurred in a public health facility was Rs1650 ($36.7), including Rs276 ($6.1) for ANC, Rs1111 ($24.7) for birth care and Rs261 ($5.8) for PNC. Average maternal health care in a private facility was almost four times that in a public facility at Rs6690 ($149), including Rs1402 ($31.2) for ANC, Rs4692 ($104.7) for birth care and Rs596 ($13.3) for PNC. Average maternal health care on home birth care was Rs428 ($9.5). Expenditure on ANC, delivery and PNC was higher in urban areas compared with rural areas, except for delivery care in public facilities, where rural households spent more. The ANC, delivery and PNC also varied widely by state, household and the women s background characteristics. ANC for public providers varied from Rs80 ($1.8) in the poorest decile to Rs1094 ($24.3) in the richest decile, compared with Rs447 ($9.9) to Rs1782 ($39.6) for private providers, respectively. Delivery care varied from Rs352 ($7.8) for the poorest decile to Rs668 ($18.4) in the richest decile for home delivery; Rs717 ($15.9) to Rs1574 ($35.0), respectively, for delivery in public facilities; and Rs2449 ($54.4) to Rs7008 ($155.7), respectively, for delivery in private facilities. For PNC care, varied from Rs108 ($2.4) in the poorest decile to Rs752 ($16.7) in the richest decile for public providers, and from Rs321 ($7.1) to Rs1067 ($23.7), respectively, for private providers (Table 1). There was wide variation in the ANC, delivery and PNC by states. For example, on home delivery varied from Rs183 ($4.0) in Orissa to Rs747 ($16.6) in

CATASTROPHIC MATERNAL EXPENDITURE IN INDIA 449 Table 1 Use of antenatal care (ANC), delivery and postnatal care (PNC) by type of facility; percentage of maternal -1 and maternal -2; and average for ANC, delivery and PNC ANC Delivery PNC Maternal ANC Delivery PNC Pub Pri Pub Pri Pub Pri 1 a 2 a Pub Pri Home Pub Pri Pub Pri % % % % % % % % Rs Rs Rs Rs Rs Rs Rs Overall 42 31 21 22 28 36 16 51 276 1091 428 1111 4692 263 596 Residence Rural 42 27 18 17 28 34 14 52 250 957 414 1165 4137 232 541 Urban 38 45 31 43 30 41 23 47 382 1402 552 994 5480 367 762 State Union Territories 55 27 70 17 41 16 16 37 530 927 402 998 6189 84 1021 Jammu & Kashmir 63 17 53 3 42 23 17 34 915 1867 747 1222 7345 500 1211 Himachal Pradesh 85 8 53 8 64 10 31 35 527 1727 730 3267 7882 506 646 Punjab 37 30 21 35 30 26 23 27 881 1770 727 3011 5778 668 621 Haryana 55 25 5 28 16 31 15 31 239 1084 411 1781 5040 249 1256 Delhi 50 28 49 12 50 25 13 34 636 2048 774 910 7021 431 1017 Rajasthan 56 18 20 13 27 32 11 53 328 1114 474 2054 4838 605 1067 UP and Uttarakhand 30 26 4 12 18 44 9 56 121 665 512 1684 4153 201 402 Bihar 9 46 4 14 10 44 10 51 497 468 407 2227 2072 253 339 North East 53 12 41 6 35 10 7 19 277 1517 214 1205 6354 413 267 Assam 55 23 34 8 41 30 16 44 241 863 347 1238 2649 316 444 West Bengal 58 31 43 13 36 30 14 56 295 856 345 884 5038 174 389 Jharkhand 23 36 8 21 21 39 4 63 252 443 350 904 2922 141 229 Orissa 66 13 25 6 47 22 15 73 218 683 183 1493 3487 232 486 Chhattisgarh 56 17 10 10 36 21 5 69 131 502 337 945 3678 128 271 Madhya Pradesh 44 18 27 15 40 21 13 54 285 1442 431 1403 6341 335 687 Gujarat 47 33 15 39 25 30 20 36 133 1899 385 1040 4006 169 1401 Maharashtra 38 46 27 40 27 40 20 46 246 1289 325 721 4771 185 644 Andhra Pradesh 43 48 30 40 32 42 30 40 456 1506 340 832 3717 251 494 Karnataka 60 28 32 31 51 28 18 45 152 1153 212 388 5170 193 564 Kerala and Goa 28 68 31 69 25 60 51 50 1570 1631 400 1991 6369 971 1413 Tamil Nadu 64 33 53 38 49 27 34 50 121 1852 234 448 5862 82 654 Per capita decile Poorest 43 21 11 8 29 33 11 100 88 467 352 717 2449 108 321 2 45 22 19 9 26 36 11 99 187 455 403 866 2267 119 333 3 41 19 23 8 29 27 8 77 217 810 389 578 2795 204 449 4 42 29 20 17 32 35 15 46 121 807 454 1059 3049 182 419 5 45 28 24 17 30 30 14 29 233 912 395 1105 3496 334 620 6 40 32 20 22 25 34 19 33 267 1116 502 1487 3188 221 544 7 44 31 28 27 28 36 16 23 427 1216 528 1026 3935 386 621 8 45 37 26 32 35 33 21 18 305 1178 467 1184 6220 258 758 9 41 44 24 49 24 48 26 12 465 1615 540 1941 5747 495 963 Richest 25 65 16 74 23 62 36 6 1300 1851 668 1574 7008 752 1067 Religion Hindu 43 31 21 22 29 36 16 50 256 1124 426 1114 4665 252 598 Islam 38 29 19 20 24 38 15 56 261 935 440 831 4503 229 535 Christian 35 43 31 35 34 30 27 32 903 1034 206 2010 6431 567 1114 Others b 36 36 27 35 31 30 13 36 654 885 558 1406 4586 522 601 (continued)

450 HEALTH POLICY AND PLANNING Table 1 Continued Jammu and Kashmir; public facility delivery varied from Rs388 ($8.6) in Karnataka to Rs3267 ($72.6) in Himachal Pradesh; and private facility delivery varied from Rs2072 ($46.0) in Bihar to Rs7882 ($175.2) in Himachal Pradesh. Figure 1 shows household annual, household annual capacity to pay and total maternal in households that had a birth in the reference period, by different deciles. It shows that households from the poorest two deciles have maternal health care higher than their annual capacity to pay, which clearly indicates the scale of financial distress that the poorest households may suffer due to relating to maternal health care. The incidence of ME-1 is 16% and incidence of ME-2 is 51% (Table 1). Figure 2 shows the incidence of ME-1 and ME-2 in different deciles, along with percentage of home, public and private hospital deliveries. ME-1 increases with decile, while ME-2 reduces sharply after the third decile, and continues to reduce further. Interestingly, the trend for home delivery is closely correlated with ME-2, while private facility delivery is correlated with ME-1. Table 2 shows the results of the multivariate logistic regression for ME-1 and ME-2. The probability of ME-2 was more likely in urban areas (P < 0.001) compared with rural Maternal ANC Delivery PNC ANC Delivery PNC Pub Pri Pub Pri Pub Pri 1 a 2 a Pub Pri Home Pub Pri Pub Pri % % % % % % % % Rs Rs Rs Rs Rs Rs Rs Social group Scheduled tribes 48 16 17 7 33 19 7 62 175 437 269 704 3644 163 366 Scheduled castes 43 27 23 13 27 32 11 56 171 859 418 879 3796 215 502 Other backward castes 40 32 19 24 29 36 18 50 268 1203 467 1125 4278 261 619 Forward castes 40 39 23 33 26 44 21 42 425 1188 458 1414 5555 356 663 Age category (years) 15 19 49 28 24 20 30 32 19 53 260 1209 334 963 4376 220 616 20 24 42 32 23 24 29 36 17 47 281 1137 444 1056 4557 250 632 25 29 43 32 22 24 29 35 15 50 255 1014 450 1099 4710 315 574 30 34 40 30 18 20 25 38 17 55 294 1132 428 1154 5147 275 616 35 39 35 25 11 11 26 34 9 57 171 1120 380 1650 5553 159 382 40 44 29 28 13 10 20 30 10 71 789 551 388 3065 5767 162 649 45 49 14 27 14 12 17 35 6 49 726 619 398 1107 2431 157 314 Education status Illiterate or <primary 40 23 16 11 26 32 9 57 175 614 398 958 2852 187 427 Primary completed 51 29 29 21 35 31 20 53 235 1488 486 853 4473 241 643 Middle completed 47 38 30 32 33 39 24 41 369 1687 555 1163 5013 297 833 Secondary/Sn. Sec. 37 53 27 50 29 45 33 38 483 1228 518 1779 4898 463 744 Diploma/Degree 29 65 16 75 24 60 31 18 1101 1494 479 1320 7508 669 936 Notes: 1 US$ was approximately equal to Rs45 in 2004. n ¼ 6879 women who delivered during the survey period. ANC ¼ antenatal care; PNC ¼ postnatal care; Pri ¼ private facility; Pub ¼ public facility; Rs ¼ Indian rupees. a Maternal Expenditure-1 (ME-1) is maternal health care more than 10% of the annual household consumption ; Maternal Expenditure-2 (ME-2) is maternal health care more than 40% of the capacity to pay (obtained after subtracting poverty-line subsistence allowances from household ). b Other religions include Sikhism, Jainism, Buddhism, Zoroastrianism and other religions not covered elsewhere. areas after controlling for all other factors in the model. After controlling for other variables in the model, no significant association was seen for either ME-1 or ME-2 with social group or women s age. However, there was significance for household quintile, woman s educational status, religion and state. Compared with women from the poorest quintile, women from the richer quintiles have lower odds of incurring ME-1 and ME-2 after controlling for other variables in the model. Compared with women who do not seek ANC, women who seek ANC at private facilities are more likely to have higher odds of ME-1 and ME-2. Compared with women who delivered at home, the odds of ME-1 or ME-2 are higher in women who delivered in public or private facilities; and likewise, compared with women who did not use PNC, women who used PNC in public or private facilities had higher odds of ME-1. Compared with illiterate and less than primary educated women, the likelihood of ME-1 was higher in women with primary to secondary-level education, while ME-2 was more likely in women with primary-level education. Compared with women of Hindu religion, women from other religions (including Sikh, Jains and Parsis) had lower odds of ME-1. Among the states, compared with the control states (Uttar Pradesh and

Expenditure Deciles (1=Poorest, 10=Richest) 10 9 8 7 6 5 4 3 2 1 Percentage 1,058 723 1,006 45 839 0 100 Average maternal in a household with a delivery during the survey period Household annual capacity to pay Household annual 2,532 2,007 90 80 70 60 50 40 4,976 3,425 1,522 2,883 7,984 8,015 1,763 5,772 11,939 17,243 19,462 26,479 24,693 27,748 30,644 36,573 34,031 40,764 46,558 Maternal Expenditure-1 Maternal Expenditure-2 % Home deliveries 57,956 56,609 % Deliveries in public hospital % Deliveries in private hospital 85,438 0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 80,000 90,000 Expenditure in Rs Figure 1 Total annual household, household annual capacity to pay, and average maternal (in households with a delivery during the survey period), by deciles Source: Authors analysis of National Sample Survey 2004 (NSSO 2006). Notes: The annual household capacity to pay was derived by subtracting state-specific poverty-line annual from annual household. 30 20 10 0 1 2 3 4 5 6 7 8 9 10 Poorest <------------- Expenditure deciles -------------> Richest Figure 2 Percentage of households that incurred high maternal (ME-1 and ME-2) and type of delivery care (home, public or private) by deciles Source: Authors analysis of National Sample Survey 2004 (NSSO 2006). Note: Maternal Expenditure-1 is maternal health care more than 10% of the annual household consumption ; Maternal Expenditure-2 is maternal health care more than 40% of the capacity to pay (obtained after subtracting subsistence allowances from household consumption ). Maternal health care includes antenatal, delivery and postnatal health care.

452 HEALTH POLICY AND PLANNING Table 2 Adjusted odds of high maternal -1 and maternal -2 Maternal -1 Maternal -2 OR SE OR SE Urban (0 ¼ Rural) Urban 0.84 0.12 11.98*** 2.22 State (0 ¼ UP & Uttarakhand) Union Territories 1.08 0.43 1.37 0.48 Jammu and Kashmir 2.44* 1.29 3.53* 2.31 Himachal Pradesh 4.18*** 2.06 5.35* 5.20 Punjab 3.99*** 1.67 1.20 0.71 Haryana 1.24 0.50 1.61 0.57 Delhi 1.39 1.28 2.39 1.78 Rajasthan 1.18 0.39 1.91*** 0.47 Bihar 0.77 0.31 0.35*** 0.11 North East 1.02 0.41 0.64 0.23 Assam 1.75 0.82 0.78 0.26 West Bengal 1.11 0.32 1.17 0.28 Jharkhand 0.16*** 0.09 0.61 0.23 Orissa 1.88* 0.63 0.72 0.30 Chhattisgarh 0.54 0.26 0.46 0.23 Madhya Pradesh 1.32 0.37 0.43 0.15 Gujarat 1.13 0.32 0.44*** 0.13 Maharashtra 0.74 0.22 0.75 0.19 Andhra Pradesh 1.28 0.31 0.24*** 0.07 Karnataka 0.71 0.22 0.27*** 0.09 Kerala & Goa 1.49 0.42 2.75*** 0.89 Tamil Nadu 1.65* 0.43 0.64 0.19 Religion (0 ¼ Hindu) Muslim 0.92 0.19 1.34* 0.23 Christian 1.13 0.43 1.26 0.54 Others 0.30*** 0.12 1.29 0.52 Social group (0 ¼ Scheduled tribes) Scheduled castes 0.93 0.28 0.66 0.19 Other backward castes 1.42 0.41 0.68 0.18 Others 1.26 0.39 0.65 0.18 Expenditure quintile (0 ¼ Poorest) Second poorest 0.66** 0.14 0.01*** 0.00 Middle 0.76 0.16 0.00*** 0.00 Second richest 0.47*** 0.10 0.00*** 0.00 Richest 0.48*** 0.11 0.00*** 0.00 Age (Continuous) 0.99 0.01 1.00 0.01 Education (0 ¼ illiterate/primary not completed) Primary 1.75*** 0.32 1.74*** 0.30 Middle 1.61** 0.30 1.18 0.22 Secondary/sn. secondary 1.65*** 0.31 1.32 0.27 Diploma/degree 1.06 0.26 0.64* 0.17 ANC care (0 ¼ none) Public facility 1.73** 0.43 1.07 0.20 Private facility 3.70*** 0.93 1.96*** 0.38 Delivery (0 ¼ home) Public facility 4.12*** 0.80 2.27*** 0.50 Private facility 12.59*** 2.43 7.64*** 1.52 PNC care (0 ¼ none) Public facility 1.44** 0.23 1.01 0.17 Private facility 2.20*** 0.35 1.36* 0.21 Population size 16 684 820 16 684 820 F(43, 4149) 15.12 18.71 Prob > F 0.00 0.00 OR ¼ odds ratio; SE ¼ standard error. *P < 0.1; **P < 0.05; ***P < 0.01.

CATASTROPHIC MATERNAL EXPENDITURE IN INDIA 453 Uttarakhand), women from Punjab and from Himachal Pradesh have higher odds (P < 0.001) of ME-1, while those in Jharkhand had lower odds. Compared with the control states, Rajasthan and Kerala & Goa had higher odds of ME-2, while Bihar, Gujarat, Andhra Pradesh and Karnataka had lower odds. Discussion The study suffers from some limitations that may bias the estimates of average ME and annual household and the relationship between them. Collection of the data for both ME and household as aggregate consumption items may underestimate the level of, in comparison with the collection of information disaggregated by each cost item (NSSO 2006). However, this method is still expected to give a reasonable proxy for relative ranking of households according to the level of standard of living (NSSO 2006). In addition, the observed birth rates in the NSS are lower than the birth rates reported in other surveys and census, suggesting some under-reporting of births (NSSO 2006). Differential under-reporting of births across different population groups may lead to overestimation or underestimation of average ME depending on which socio-economic groups have had higher under-reporting. While some of the methodological issues encountered in studies of catastrophic health in the literature are also applicable to this study (e.g. use of ad hoc cut-offs, lack of data on source of financing, and absence of longitudinal data), others have been overcome to some extent. The potential endogeneity problem (health spending appearing on both the right- and left-hand sides of the regression equation) in assessing the relationship between total household consumption and health s was overcome by using different reference periods for total household and maternal heath and the exclusion of maternal health from total household s. Secondly, one of main issues in estimating catastrophic health s is the exclusion from the analysis of households that forgo health care to avoid the catastrophic consequences of such s. Generally, the sample for such estimation includes only those households where at least one member sought hospitalization or other formal health care. However, all the households that had a birth were included in this study, allowing better assessment of the role of financial constraints among households that forgo institutional maternal care and choose home delivery. The increase in maternal health with rising household consumption is in line with other studies showing a high degree of elasticity between health and household consumption (O Donnell et al. 2008), but having information on source of care helps to explain it better. The explanations for the observed elasticity in this study are most likely to be: lack of heath insurance; constraints in access to public health facilities (Roy and Howard 2006); the inability to control prices; and the changing preference for private health care over public health care. The lowest incidence of ME-1 and highest incidence of ME-2 in the poorest two deciles indicates not only that high-cost institutional care is foregone to avoid immediate financial distress, but also the little income available to afford even home delivery. Two of the main reasons given for not seeking institutional delivery care in the National Family Health Survey III were that institutional delivery is not necessary (71%) and that costs were too much (26.2%) (International Institute for Population Sciences 2007b). Compared with catastrophic curative health care (e.g. in sudden acute illness or accidents), where households may make an effort to meet the expenses through different coping strategies (asset sale or borrowing), for maternal health care the costs may most likely lead to non-utilization of institutional care and preference for home delivery. Comparing data on catastrophic maternal health care and catastrophic health care Using a similar household survey conducted in 1995 96, Flores et al. (2008) estimated that 30% of households with at least one hospitalized member spent more than 10% of total household on inpatient care. Bonu et al. (2007) estimated that 13.1% of households incurring health (on both outpatient and inpatient care) spent more than 10% of total household ; and 5.1% of households made health payments exceeding 40% of their capacity to pay. Though not directly comparable due to some differences in methodology, this does suggest that while the incidence of catastrophic ME (16%) is lower than for catastrophic inpatient care (which is in line with the higher overall average cost of one hospitalization episode than the cost of one normal delivery), it is higher than that for overall health care payments. This indicates that households incurring maternal health care in a particular year are likely to spend more than the amount an average household in India would spend on overall health care, and this on maternal health care is likely to put the household in deeper financial distress, or more likely discourage institutional care. Which indicator captures financial distress better? For people below the poverty line, the capacity to pay is extremely constrained, and hence even small amounts of extra health can cause financial distress and more importantly can lead to non-utilization of institutional care. In India, households below the poverty line are those unable to purchase 2400/2100 calories per capita per day in rural/urban areas, respectively (see Sharma 2004 for more details on poverty line calculation in India). While the incidence of ME-2 declined with increasing wealth quintile, the reverse was observed for ME-1, reflecting increasing levels of utilization of institutional maternity care by the rich. All the households in the poorest decile incurred maternal of more than 40% of their capacity to pay, while the incidence of ME-1 in the poorest decile was only 11%, showing non-utilization of services. Hence, in developing countries, where a large number of households live below the poverty line, measuring maternal health care as a proportion of capacity to pay may be a more relevant indicator than ME-1 (proportion of total household consumption ) to capture the potential financial distress it causes.

454 HEALTH POLICY AND PLANNING Public sector maternal care for the poor Average maternal care for a birth at a public sector facility was Rs717 ($16) in the poorest decile. This was higher than the average annual capacity to pay (close to zero) in the same decile, leading to almost 81% women from this decile giving birth at home. Though not assessed in this study, this is likely to result in adverse maternal outcomes (Ganatra et al. 1998). Concerns have been expressed that India s economic growth in recent years has not been inclusive enough (World Bank 2006), with limited increase in capacity to pay among the poorest households. This implies that direct maternal health care interventions, including direct cash transfers to pregnant women, may be necessary to assist the poorest households to overcome their financial constraints to access maternal health care. As perinatal and neonatal mortality gain more significance to reduce child mortality, increasing the use of maternal health care services becomes all the more important not only to reduce maternal mortality but also to reduce child mortality (Claeson et al. 2000; Jones et al. 2003). More effective targeting of public resources Considering the limited public resources and competing priorities, direct interventions to overcome financial constraints to improve maternal health care use should be targeted at specific population groups, as suggested by this study, e.g. scheduled tribe households, certain states with very high incidence of ME-2 (e.g. Rajasthan, Uttar Pradesh, Bihar, Jharkhand, Orrisa and Chattisgarh). The National Rural Health Mission (NRHM), launched by the Government of India in 2005, is making concerted efforts by providing conditional cash transfers for maternal health care and institutional deliveries, and cash incentives to health workers (Paul 2007). The Government of India is paying Rs1400 to women for institutional delivery under the Janani Suraksha Yojana. The NRHM faces a number of challenges (Chatterjee 2006) in ensuring effective utilization of these cash transfers by the poorest women, and in improving the capacity and quality of maternal health services in the public sector to meet the resultant increasing demand. The results from this study may provide some guidance on the level at which cash transfers should be set to encourage use of maternal services by the poorest households. It is important to assess the adequacy of cash transfers and to target them effectively to the poorest women. The cash transfers should be enough to help poor households meet the opportunity costs (such as lost working days) and other direct costs such as transportation costs, accommodation and child care costs, besides expenses at the health facility. Informal fees, where prevalent, should be addressed so that they do not become a barrier to use of maternal care. In addition, a more comprehensive approach is required for ANC, maternal care and PNC to ensure that the package includes neonatal care and nutritional services. Improving the public sector Improving the predictability and quality (both technical and inter-personal) of maternal services in the public sector is essential to increase their utilization, in addition to direct cash transfers. Despite almost three times higher in private facilities compared with public facilities, a substantial proportion of women from the poorest decile opted for private facilities (8% for private, 11% for public). This is in line with previous studies that documented the poor and even very poor bypassing notionally free public facilities in favour of private providers, even at the risk of getting into greater financial distress. Reasons for bypassing the public sector documented in the literature include: lower waiting times in the private sector, higher predictability of the availability of providers, of drugs and of fees charged, better interpersonal care and cleanliness in the private versus public sector (Brugha and Pritze-Aliassime 2003; Waters et al. 2003; Lewis et al. 2005). The overall utilization of public maternity services and average incurred in the public facilities vary significantly across states in India. Given the wide variation in the public sector maternal service use (70% in Union Territories and 53% in Himachal and Jammu and Kashmir on one hand, and 4% in Uttar Pradesh and Bihar on the other), the various reasons responsible for the low public sector performance need to be investigated and remedied at state level. Partnership with the private sector Considering the current inadequate capacity of public health facilities in many states, concurrent strategies need to be adopted for improving the quality and predictability of public maternal health care services, and increasing provider choice for pregnant women through public-private partnership with private sector providers. One way this could be achieved is by providing vouchers for poor women for ANC, delivery and PNC services, which could be utilized either in public or in private facilities. Introduction of contracting-in of health facility management to the private sector through performance-based public-private partnerships can provide alternative models. In situations where institutional failures are too severe in the public sector, contracting-out of maternal and child health services may be an alternative solution to full-scale public sector reforms. However, the limited capacity of the state and district health departments to manage complex public-private partnerships will be a major constraint. The inclusion of private providers would also mean setting higher levels of cash transfers. A number of innovative pilot public-private partnerships including voucher schemes have evolved for maternal health care in India. These need to be carefully evaluated in order to select models for scale-up, though different models may be required for different regions according to the different challenges faced. Regulation of the private sector and controlling costs The high in the private sector may be partly due to over-medicalization (Brugha and Pritze-Aliassime 2003). Regulation of the private sector and other measures to control the costs of essential services like maternity care should be explored. The NRHM s introduction of direct cash transfers to pregnant women and negotiating price ceilings with private sector providers to avail NRHM schemes may help in controlling costs through the collective bargaining power of

CATASTROPHIC MATERNAL EXPENDITURE IN INDIA 455 a public sector purchaser a role normally played by providers of health insurance in developed countries, which, unfortunately, is still at a nascent stage in India. Other policy level barriers that restrict the use of emergency obstetric care, as documented by previous studies, also need to be addressed (Mavalankar et al. 2005). India s private health sector market is extremely fragmented, with a multitude of small-scale providers. Corporatization of health care provision in India could help in improving the quality of care and control costs by driving greater efficiencies. Regulators have a role in ensuring that the benefits of greater efficiencies are passed on by the corporate sector to the end beneficiaries, controlling health care costs and increasing utilization. Conclusion The study explores two methodological approaches of measuring catastrophic maternal. It concludes that measuring maternal in relation to household capacity to pay may capture the financial barriers to maternal care better than measuring it as a proportion of overall household. The study provides compelling empirical evidence of potential financial distress due to maternal, especially for the poorest women, and argues that catastrophic maternal health care can be a formidable barrier to the use of maternal services. Adequate direct cash transfers with effective pro-poor targeting, along with concurrent efforts to improve public sector performance, regulation of and partnership with the private sector, are some of the broad policy options. Acknowledgements The authors thank the National Sample Survey Organization of India for giving access to the survey data used in the study. The authors thank the two anonymous referees for their valuable suggestions, which helped in improving the manuscript. The views expressed are those of the authors and do not necessarily reflect those of their affiliated institutions. References Afsana K. 2004. The tremendous cost of seeking hospital obstetric care in Bangladesh. Reproductive Health Matters 12: 171 80. Agarwal B. 1997. Gender, environment and poverty inter-links: regional variations and temporal shifts in Rural India, 1971 91. World Development 25: 23 52. Balaji R, Dilip TR, Duggal R. 2003. Utilization of and on delivery care services: some observations from Nashik District, Maharashtra. Regional Health Forum, WHO South-East Asia Region, Volume 7, Number 2. Berki SE. 1986. A look at catastrophic medical expenses and the poor. Health Affairs 5: 139 45. Bonu S, Rani M, Peters DH, Jha P, Nguyen SN. 2005. Does use of tobacco or alcohol contribute to impoverishment from hospitalization costs in India? Health Policy and Planning 20: 41 49. Bonu S, Bhushan I, Peters DH. 2007. Incidence, intensity and correlates of catastrophic out-of-pocket health payments in India. ERD Working Paper Series No. 102. Manila: Asian Development Bank. Online at: http://www.adb.org/documents/erd/working_papers/ WP102.pdf Borghi J, Ensor T, Neupane BD, Tiwari S. 2004. Coping with the burden of the costs of maternal health. Nepal Safe Motherhood Project, Government of Nepal. Borghi J, Ensor T, Somanathan A, Lissner C, Mills A. 2006a. Mobilising financial resources for maternal health. The Lancet 368: 1457 65. Borghi J, Ensor T, Neupane BD, Tiwari S. 2006b. Financial implications of skilled attendance at delivery in Nepal. Tropical Medicine and International Health 11: 228 37. Borghi J, Sabina N, Blum LS, Hoque ME, Ronsmans C. 2006c. Household costs of health care during pregnancy, delivery, and the postpartum period: a case study from Matlab, Bangladesh. Journal of Health, Population and Nutrition 24: 446 55. Brugha R, Pritze-Aliassime S. 2003. Promoting safe motherhood through the private sector in low-and middle-income countries. Bulletin of the World Health Organization 81: 616 23. Chatterjee P. 2006. India s government aims to improve rural health. The Lancet 368: 1483 4. Claeson M, Bos ER, Mawji T, Pathmanathan I. 2000. Reducing child mortality in India in the new millennium. Bulletin of the World Health Organization 78: 1192 9. Donnay F. 2000. Maternal survival in developing countries: what has been done, what can be achieved in the next decade. International Journal of Gynecology and Obstetrics 70: 89 97. Duggal R. 2004. The political economy of abortion in India: Cost and patterns. Reproductive Health Matters 12: 130 7. Dyson T, Moore M. 1983. On kinship structure, female autonomy, and demographic behavior in India. Population and Development Review 9: 35 60. Filippi V, Ronsmans C, Campbell OMR et al. 2006. Maternal health care in poor countries: the broader context and a call for action. The Lancet 368: 1535 41. Flores G, Krishnakumar J, O Donnell O, Van Doorslaer E. 2008. Coping with health-care costs: Implications for the measurement of catastrophic s and poverty. Health Economics 17: 1393 412. Galanter M. 1984. Competing equalities: law and the backward classes in India. Berkley, CA: University of California Press. Ganatra BR, Coyaji KJ, Rao VN. 1998. Too far, too little, too late: a community-based case-control study of maternal mortality in rural west Maharashtra, India. Bulletin of the World Health Organization 78: 591 8. Ghurye GS. 1992. Features of caste system. In: Gupta D. (ed.). Social stratification. New Delhi and New York: Oxford University Press, pp. 28 49. Government of India. 2006a. Report of the working group on health of women and children for the Eleventh Five Year Plan (2007 2012). New Delhi: Planning Commission. Government of India. 2006b. Towards faster and more inclusive growth: an approach to the 11 th five year plan. New Delhi: Planning Commission. International Institute for Population Sciences. 2007a. Fact Sheet: National Family Health Survey, India (2005 2006). Mumbai, India: IIPS. International Institute for Population Sciences. 2007b. National Family Health Survey (NFHS-3), 2005 06: India: Volume I. Mumbai, India: IIPS. Jones G, Steketee RW, Black RE et al. 2003. How many child deaths can we prevent this year? The Lancet 362: 65 71.