Is demand-side financing equity enhancing? Lessons from a maternal health voucher scheme in Bangladesh

Social Science & Medicine 72 (2011) 1704e1710

Contents lists available at ScienceDirect

Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed

Is demand-side financing equity enhancing? Lessons from a maternal health voucher scheme in Bangladesh Shakil Ahmed a, *, M. Mahmud Khan b a b

The University of Melbourne, Australia Tulane University School of Public Health & Tropical Medicine, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 7 April 2011

Demand-side financing (DSF) is used in the less-developed countries of the world to improve access to healthcare and to encourage market supply. Under DSF, households receive vouchers that can be used to pay for healthcare services. This study evaluated the effects of a universal DSF on maternal healthcare service utilization in Bangladesh. A household survey was conducted in and around the voucher scheme area one year after the initiation of the project. Women who gave birth within a year prior to the survey were interviewed. The utilization rates of maternal health services were found to be higher for all socioeconomic groups in the project area than in the comparison areas. Voucher recipients in the project area were 3.6 times more likely to be assisted by skilled health personnel during delivery, 2.5 times more likely to deliver the baby in a health facility, 2.8 times more likely to receive postnatal care (PNC), 2.0 times more likely to get antenatal care (ANC) services and 1.5 times more likely to seek treatment for obstetric complications than pregnant women not in the program. The degree of socioeconomic inequality in maternal health service utilization was also lower in the project area than in the comparison area. The use of vouchers evidenced much stronger demand-increasing effects on the poor. Poor voucher recipients were 4.3 times more likely to deliver in a health facility and two times more likely to use skilled health personnel at delivery than the non-poor recipients. Contrary to the inverse equity hypothesis, the voucher scheme reduced inequality even in the short run. Despite these improvements, socioeconomic disparity in the use of maternal health services has remained pro-rich, implying that demand-side financing alone will be insufficient to achieve the Millennium Development Goal for maternal health. A comprehensive system-wide approach, including supply-side strengthening, will be needed to adequately address maternal health concerns in poor developing countries. Ó 2011 Elsevier Ltd. All rights reserved.

Keywords: Demand-side financing Equity Utilization of maternal health services Maternal health voucher Bangladesh

Introduction In many developing countries of the world, policy makers are actively considering the adoption or scaling up of demand-side financing (DSF) as a means to improve access to health services (Anwar et al., 2008; Behrman & James, 1998; Bhatia & Gorter, 2007; Bhatia, Yesudian, & Gorter, 2006). The traditional approach for expanding access has been to invest in supply-increasing interventions such as geographic coverage of service delivery and ensuring the availability of trained personnel, equipment, supplies, drugs, etc. By the early 1990s, it became clear that supply-side

* Corresponding author. Tel.: þ61 3 9035 5540; fax:þ61 3 8344 9140. E-mail address: [email protected] (S. Ahmed). 0277-9536/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2011.03.031

policies alone will not sufficiently address low utilization and inequality concerns (Bhatia & Gorter, 2007; Bhatia et al., 2006; Ensor & Cooper 2004; Koblinsky et al., 2006; Thaddeus & Maine 1994). Despite the capacity expansion of healthcare facilities, access to maternal health services remained low, especially for poorer sections of the community, due to relatively high out-ofpocket expenses in both the public and private sectors (Bonu, Bhushan, Rani, & Anderson, 2009; Borghi, Sabina, Blum, Hoque, & Ronsmans, 2006, Borghi, Ensor, Somanathan, Lissner, & Mills, 2006; Dhar et al., 2009). To improve the utilization of maternal health services and to help achieve the Millennium Development Goal (MDG) for maternal health, the focus has shifted to reducing financial barriers to access (Bhatia & Gorter, 2007; Bhatia et al., 2006; Witter, Adjei, Armar-Klemesu, & Graham, 2009). In recent years, maternal health voucher-based DSFs have been adopted in a number of low-income countries, e.g., Bangladesh,

S. Ahmed, M.M. Khan / Social Science & Medicine 72 (2011) 1704e1710

Bolivia, Cambodia, India, Kenya and Pakistan. The main objectives of these programs are to increase access to essential maternal health services and to enhance equity in the utilization of these services (Ahmed & Khan, 2010; Anastasia, 2009; Bashir, Kazmi, Eichler, Beith, & Brown, 2010; Bhat, Mavalanker, Singh, & Singh, 2009; Ir, Horemans, Souk, & Damme, 2010; Ensor, 2004; Lenel & Griffith, 2007). In a maternal health voucher scheme, pregnant women are given vouchers which can be used to pay for targeted health services either in part or in full (Ahmed & Khan, 2010; Anastasia, 2009; Ir et al., 2010). Although preliminary evidence indicates that the vouchers do increase the use of maternal health services, little is known about the equity impact of these programs (Gorter & Bellows, 2008; Gupta, Joe, & Rudra, 2010; United Nations Population Fund, 2006). A systematic evaluation of an existing program would thus be very useful in demonstrating the potential impact of these vouchers on equity in service utilization. This study attempted to answer this important question, i.e., whether the DSF programs are successful in reducing socioeconomic disparity in medical care utilization, one of the three main goals (namely, utilization, equity and quality) of the DSF approaches (Gorter & Bellows, 2008; Gupta et al., 2010; Lenel & Griffith, 2007; United Nations Population Fund, 2006). The Maternal Health Voucher Scheme (MHVS) initiated by the Government of Bangladesh (Government of the People’s Republic of Bangladesh, 2007) in 2007 offers a unique opportunity to conduct this analysis. The MHVS in Bangladesh was adopted to lessen the financial barriers faced by pregnant women in accessing medical care and to increase the demand for maternal health services (Government of the People’s Republic of Bangladesh, 2007). The ultimate goal of the program is to decrease maternal mortality, particularly among poor socioeconomic groups. In 2007, the Ministry of Health and Family Welfare, Government of Bangladesh, started the implementation of MHVS in 33 subdistricts (Government of the People’s Republic of Bangladesh, 2007). The scheme used two different approaches for targeting, i.e., the universal program and the targeted program. The universal program was implemented in the nine poorest subdistricts, where all pregnant women of Parity 1 or 2 (first or second pregnancy), regardless of poverty status, were offered vouchers. The targeted program was implemented in 24 subdistricts, where means-testing was used to identify the eligible beneficiaries. The means-testing used the following criteria for inclusion: the recipient must be a resident of the subdistrict, currently pregnant with their first or second child, owning less than 6534 square feet of land, receiving a household income less than US$38.50 per month, and lacking ownership of other productive assets. Under the MHVS, targeted women receive vouchers for three ANC check-ups, safe delivery at a facility or at home by skilled birth attendants, one postnatal checkup within six weeks of delivery and the management of complications, including Caesarean section, from designated providers. They also receive transportation costs for accessing the covered services. As additional incentives, the scheme provides a gift box and a cash payment to women who deliver at a health facility. Details of the implementation approaches along with several initial lessons learned have been described elsewhere (Ahmed & Khan, 2010). The aim of this study was to empirically analyze the effect of MHVS on the utilization of maternal health services by the poor in one subdistrict of Bangladesh. This study was based on data collected in the area at the end of first year of implementation of the pilot project. Although one year is not long enough to identify the effect of the project on final outcomes such as maternal mortality, many of the process indicators should show improvements in the short run with respect to reducing maternal mortality and reducing morbidity in the long run.

1705

Methods Study area The MHVS was launched in April, 2007 in the Sarishabari subdistrict, located in the Jamalpur district of Bangladesh. Although this subdistrict was selected to implement the targeted version of the scheme, the field-level managers mistakenly implemented the universal program, and vouchers were offered to all pregnant women (Ahmed & Khan, 2010). Similar problems with MHVS implementation have been documented in a recent evaluation report (Hatt et al., 2010). Sarishabari is one of the first 21 subdistricts in which the MHVS was started as a pilot project. This subdistrict is considered to be one of the poorest among the pilot subdistricts. Jamalpur district is divided into six rural subdistricts and one urban subdistrict, and the household survey was carried out in all six rural subdistricts. The survey was conducted during May and June of 2008, exactly one year after the implementation of the MHVS in the area. Because the scheme was implemented in only one subdistrict, the other five subdistricts were used as the comparison area for this analysis. The research was approved by the Ethical Review Committee of the Bangladesh Medical & Research Council, Dhaka, Bangladesh. Sampling The survey collected information from a total of 3600 women, 600 in each of the six subdistricts. The sample size was deemed sufficiently large to yield reliable estimates of maternal health service utilization by socioeconomic groups. The households in the survey were selected using a two-stage sampling procedure. In the first stage, 20 villages were selected from each of the subdistricts, with the probability of selection proportional to village size. The village list of the 2001 population census (Govt. of the People’s Republic of Bangladesh , 2004) was used as the sampling frame for village selection. The survey collected data from women who had delivered a baby (live birth) within the year preceding the survey. At least 30 women were interviewed in each of the villages. The interviewers, on average, visited 185 households to identify 30 eligible women. The interviewers undertook a systematic enumeration of the selected villages by starting from the northwest corner and visiting every household in turn to identify the eligible women. If 30 women could not be interviewed from a single village, the interviewer made up the shortfall by visiting as many additional households as required in the neighboring village. Data collection and analysis Data were collected on age, parity, the educational status of women and their husbands, the socioeconomic status of households including household expenditures and the use of health services during the last pregnancy. Data were analyzed using SPSS (SPSS Inc., 2008) version 17.0. The Sarishabari subdistrict, the area selected by the pilot project for voucher distribution, was considered the project area (n ¼ 600 respondents), and the remaining five rural subdistricts in the survey constituted the comparison area (n ¼ 3000 respondents). The household asset index was constructed using a principal component analysis for all households in the survey. The construction of a single asset index allowed for the comparison of household socioeconomic status across all the six rural subdistricts. The asset index has been widely applied to analyze socioeconomic differentials in the use of health services (Bollen, Glanville, & Stecklov, 2001; Filmer & Pritchett, 2001, 1999; Rutstein & Johnson, 2004; Wagstaff & Watanabe 2003). The construction of

1706

S. Ahmed, M.M. Khan / Social Science & Medicine 72 (2011) 1704e1710

asset scores was based on household ownership of durable goods and dwelling characteristics. The durable goods used were the following: radios, televisions, mobile telephones, non-mobile telephones, refrigerators, armoires, tables/chairs, watches/clocks, bicycles, and motorcycles. The dwelling characteristics were the following: electrical power; water source; type of toilet; and the construction materials of the floor, walls, and roof. Using the asset scores, households were categorized into three equal groups (terciles T1, T2, T3, with T1 being the poorest). It was found that 31.0% of the top tercile (T3) had radios, 94.0% had tables and 10.8% had watches, compared to 2.8%, 15.0% and 10.8%, respectively, for the poorest tercile. To evaluate the effects of vouchers on socioeconomic disparity in maternal health service utilization, five different measures of maternal health service use were defined: 1. ANC from trained health professionals used at least three times during pregnancy 2. Treatment received for obstetric complications from trained health professionals, if needed 3. Delivery conducted and/or assisted by skilled health personnel 4. Delivery in a health facility 5. PNC from a trained health professional used at least once in six months after delivery Concentration Indices (CI) were calculated to measure the degree of inequality (van Doorslaer, Koolman, & Jones, 2004; van Doorslaer & Masseria, 2004; Lu et al., 2007; O’Donnnell, van Doorslaer, Wagstaff, & Lindelow, 2007; Wagstaff & van Doorslaer, 2000) in the use of maternal health services by wealth index and by project and comparison areas. The CI value ranged from 1 to þ1. A negative value indicates that the inequality is pro-poor, i.e., poorer households use a higher level of healthcare than the non-poor. Similarly, positive value indicates a pro-rich inequality. To isolate the effects of MHVS on maternal healthcare utilization, five logistic regression models were estimated. The dependent variables are the five measures of maternal health service utilization mentioned above, i.e., whether or not the respondent received one of the five maternal healthcare services during the previous year. The dependent variable was coded as one if the respondent received, for example, ANC at least three times from a trained provider and zero otherwise. The selection of independent variables was guided by previous research studies examining the factors associated with the use of maternal health services (Celik & Hotchkiss, 2000; Mekonnen & Mekonnen, 2003). The variables included in the regression models include household size, average per capita monthly expenditure of the household, age and education of the woman, parity, age and education of the husband, participation of the woman in income-earning activities, being Table 1 Comparison of socioeconomic and demographic characteristics of individuals in the survey by program and comparison areas. Average values for the variables

Program area (n ¼ 600)

Comparison area (n ¼ 3000)

P value

Age of women in the sample (years) Parity Years of education of women in the sample Proportion of the women involved in income-earning activities Age of husbands (years) Years of education of husbands

23.61

23.65

0.872

2.29 4.39

2.61 2.93

0.000 0.000

0.23

0.13

0.000

31.74 3.89

32.28 2.72

0.110 0.000

poor and whether or not the woman was a voucher recipient. Because the distribution of monthly average per capita expenditure showed a distribution skewed to the right, a logarithmic transformation of this variable was used in the regression models. An interaction term between voucher receipt and being poor was used to determine if the voucher had a differential impact on the poor and the non-poor. Findings Sample characteristics Table 1 presents the basic descriptive statistics for the surveyed women. The average ages of the women were similar in the project and comparison areas, but the areas differed significantly in terms of parity, the working status of women, and the education of women and their husbands. Women in the comparison area had higher parity than women in the project area. Twenty-three percent of the women were involved in income generating activities in the project area versus only 13% in the comparison area. The average years of education of women and their husbands were also higher in the project area than those in the comparison area. The socioeconomic status of women in the project area was found to be better than that of the socioeconomic status of women in the comparison area. Among the six rural subdistricts of Jamalpur, only the project area, the Sarishabari subdistrict, had a Communitybased Skilled Birth Attendant (CSBA) training program, and the presence of this program was the reason for selecting this area for the pilot MHVS. Because women in the comparison area had a poorer socioeconomic status than in the project area, it was important to introduce a number of socioeconomic controls in the model to isolate the pure effect of the vouchers. Rich-poor differences in maternal healthcare utilization Table 2 compares the use of maternal health services by women in the poorest and the richest terciles. Rich-to-poor ratios for all the service utilization measures were greater than 2.0 in the comparison area, and all the rich-to-poor differences were statistically significant except for the variable “percent of pregnancy cases experiencing complications”. In the project area, differences between rich and poor were significant for ANC use, use of services for obstetric complications and institutional delivery. Testing the differences in the utilization rates of maternal services between the project and comparison areas for the poorest and the richest terciles indicated significantly greater use of ANC, PNC, institutional delivery and delivery assisted by skilled personnel by the poor in the MHVS area. Utilization rates in the project area were also higher for all types of maternal health services for the richest tercile. With respect to the percentage of pregnancy cases experiencing complications, the differences were not statistically significant for the poorest terciles in the comparison and project areas. In the project area, 21.6% of the women belonging to the poorest tercile received at least three ANC sessions from a trained provider, whereas this use was only 6.9% in the comparison area. More than 71% of the women in the MHVS area, irrespective of socioeconomic status, reported experiencing pregnancy-related obstetric complications compared with 61% in the comparison area. Women reporting obstetric complications were asked if they sought treatment to address their complications and, interestingly, the rich-topoor ratios were almost the same in both areas. The rich-to-poor ratios for the proportions of delivery assisted by skilled personnel were 1.3 and 4.2 in MHVS area and comparison area, respectively. In the project area, 11.7% of poor women delivered at a health facility, whereas the rate was only 0.7% in the comparison area. The rich-to-

S. Ahmed, M.M. Khan / Social Science & Medicine 72 (2011) 1704e1710

1707

Table 2 Rich-to-poor ratios for maternal health service utilization indicators by program and comparison areas. Indicators

Percentage of women receiving ANC at least three times from a trained provider Percent of pregnancy cases experiencing complications Percentage of complication cases receiving treatment for obstetric complicationsa Percentage of delivery assisted by skilled health personnel Percentage of babies delivered at a health facility Percentage receiving PNC from a trained provider within 42 days of delivery

Program area (n ¼ 600)

Comparison area (n ¼ 3000)

Poorest tercile (n ¼ 111)

Richest tercile (n ¼ 294)

Rich-to-poor Ratio

Poorest tercile (n ¼ 1090)

Richest tercile (n ¼ 907)

21.6**

33.3

1.5

6.9***

20.3

67.6

69.7

1.0

61.1

14.7**

33.7

2.3

24.3

32.0

11.7* 28.8

Difference between program and comparison areas: P values Rich-to-poor Ratio

Poorest terciles

Richest terciles

2.9

0.000

0.000

59.2

0.9

0.182

0.001

11.1***

24.0

2.2

0.360

0.008

1.3

3.0***

12.6

4.2

0.000

0.000

18.7

1.6

0.7***

8.4

12.0

0.000

0.000

31.3

1.1

4.5***

14.9

3.3

0.000

0.000

Difference from the richest tercile at *p<¼0.10, **p<¼0.05, ***p<¼0.001. a This analysis was based on those women who reported experiencing obstetric complications during pregnancy.

poor ratio for the rate of delivery at a health facility was very high (12.6) in the comparison area, higher than the national ratio of 10; however, the ratio was only 1.6 in the project area (Government of the People’s Republic of Bangladesh, 2009). The percentages of women in the poorest terciles receiving PNC were 29% and 4.5% for the project and comparison areas, respectively. Table 3 presents the CIs and the standard errors of the CIs along with their t-statistics to show the degree of inequality in the utilization of maternal health services. All the CIs were positive for both the project and comparison areas, indicating that the use of maternal health services in Bangladesh remains pro-rich. The magnitudes of the CIs were higher in the comparison area than in the project area for all maternal health services. The highest CI value in the table is that for facility delivery in the comparison area (0.43), implying that socioeconomic inequality in the use of facility utilization is very high in the absence of an MHVS. In contrast, the CI for this maternal service utilization was only 0.14 in the project area. Clearly, the scheme has significantly reduced the inequality in facility use by increasing the utilization rate of the poor. The CI index for the use of PNC services in the project area was not significantly different from zero. Therefore, the voucher project has effectively eliminated the pro-rich bias in the use of PNC services within a year after its implementation. The relatively high rich-to-poor ratios and concentration indices in the comparison area reflect the underlying high degree of inequality that exists in the utilization of maternal health services in Bangladesh. Although the differences between project and comparison areas may not be due to the voucher scheme alone, vouchers clearly played an important role here. Assuming that the voucher scheme has no effect on the utilization rate of the richest in

the project area (as they are unlikely to face significant financial barriers to accessing maternal health services), the difference in the observed utilization rates for the richest terciles in the project and comparison areas should reflect underlying inherent differences in utilization. Assuming that the relative utilization rates between project and comparison areas for the poorest terciles are similar to those for the richest terciles, the inherent difference between the areas can be corrected for the poorest group by inflating their utilization rate in the comparison area in proportion to the richto-rich ratios between the project and comparison areas. This correction still implies much higher utilization rates among the poor in the project area. Clearly, the introduction of demand-side financing significantly improved access to maternal health services for the poor households in Bangladesh. Predictors of maternal healthcare utilization Logistic regression models were constructed for each of the five measures to identify the factors affecting the utilization of maternal health services. The results of the five logistic regressions are presented in Table 4. Effect of individual characteristics The age and education of the woman were positively associated with utilization of all maternal health services. A one-year increase in a woman’s education increased the odds of receiving any type of maternal health services by nine percent. The husband’s education also showed positive effects on the utilization of all types of maternal health services, except seeking treatment for obstetric complications. Involvement in income-earning activities by women had

Table 3 Concentration indices in the use of maternal healthcare services in the program and comparison areas. Indicator

Received ANC at least three times from a trained provider Sought treatment for obstetric complicationsa Delivery assisted by skilled health personnel Delivered at a health facility Received PNC from a trained provider within 42 days of delivery

Program area (n ¼ 600)

Comparison area (n ¼ 3000)

Concentration Index (CI)

Standard error of CI

t-value

Concentration Index (CI)

Standard error of CI

t-value

0.086* 0.163*** 0.080* 0.142** 0.051

0.033 0.041 0.035 0.051 0.036

2.57 3.98 2.29 2.76 1.42

0.234*** 0.177*** 0.321*** 0.433*** 0.295***

0.026 0.030 0.034 0.039 0.032

9.00 5.91 9.46 11.0 8.98

*p<¼0.05, **p<¼0.01, ***p<¼0.001. a This analysis was based on those women who reported experiencing obstetric complications during pregnancy.

1708

S. Ahmed, M.M. Khan / Social Science & Medicine 72 (2011) 1704e1710

a significant positive impact on the use of ANC, treatment for obstetric complications and facility delivery. In contrast, parity was inversely associated with utilization, i.e., women with higher parity were less likely to use maternal health services. This finding is consistent with the results obtained by other studies (Fotso, Ezeh, & Essendi, 2009; Mekonnen & Mekonnen, 2003). The age of the husband did not show any significant influence on service utilization. Effect of household characteristics Per capita expenditure was a major predictor of maternal health service utilization. Wealthy households were more likely to use

maternal health services. Poor households showed a significantly lower likelihood of using ANC and facility delivery. Effect of vouchers After controlling for various individual and household level characteristics, women who received vouchers were more likely to use services than the women not receiving the vouchers. The largest effect size was obtained for delivery service assisted by skilled health personnel; voucher recipients were 3.58 times more likely to be assisted by skilled health personnel during delivery than others. Voucher program participants were also 2.5 times more likely to deliver baby in a health facility and 2.8 times more

Table 4 Effects of vouchers on maternal healthcare utilization after controlling for various individual and household characteristics. Independent variables Age of woman Beta Exp(B) SE Parity Beta Exp(B) SE Education of woman in years Beta Exp(B) SE Working status of woman Beta Exp(B) SE Age of husband Beta Exp(B) SE Education of husband in years Beta Exp(B) SE Household size Beta Exp(B) SE Ln per capita expenditure Beta Exp(B) SE Poor household Beta Exp(B) SE Voucher recipient Beta Exp(B) SE Poor household* Voucher recipient Beta Exp(B) SE Intercept Beta Exp(B) SE Log likelihood Model chi-square Significance Goodness-of-fit test (Hosmer-Lemeshow chi-square) Significance N

Received ANC at least three times from a trained provider

Sought treatment for obstetric complications

Delivery assisted by skilled health personnel

Delivered at a health facility

Sought postnatal care

0.074**** 1.077 0.017

0.060**** 1.061 0.018

0.091**** 1.095 0.021

0.125**** 1.134 0.028

0.091**** 1.096 0.019

0.338**** 0.713 0.068

0.119* 0.887 0.066

0.425**** 0.653 0.085

0.649**** 0.523 0.127

0.446**** 0.640 0.080

0.085**** 1.088 0.019

0.091**** 1.095 0.023

0.081**** 1.085 0.023

0.090*** 1.095 0.031

0.087**** 1.090 0.021

0.393*** 1.481 0.133

0.083 0.920 0.165

0.304* 1.356 0.161

0.351* 1.420 0.216

0.210 1.233 0.154

0.017 0.983 0.012

0.003 0.997 0.013

0.007 0.993 0.014

0.001 1.001 0.019

0.010 0.990 0.013

0.043*** 1.044 0.014

0.021 1.021 0.017

0.070**** 1.072 0.018

0.073*** 1.076 0.023

0.040*** 1.041 0.016

0.039 0.962 0.026

0.061** 1.062 0.029

0.016 1.016 0.028

0.049 1.050 0.035

0.002 0.998 0.028

0.409** 1.505 0.170

0.871**** 2.389 0.194

0.629*** 1.876 0.206

1.283**** 3.609 0.263

0.683**** 1.980 0.193

0.348** 0.706 0.149

0.036 1.037 0.161

0.307 0.735 0.214

0.775** 0.461 0.391

0.165 0.848 0.181

0.649**** 1.913 0.126

0.429*** 1.536 0.149

1.276**** 3.582 0.140

0.932**** 2.539 0.182

1.025**** 2.787 0.136

0.423 1.527 0.294

0.325 0.723 0.382

0.765** 2.148 0.328

1.455*** 4.283 0.518

0.825*** 2.281 0.298

5.433**** 0.004 1.193

9.403**** 0.000 1.374

8.613**** 0.000 1.442

14.780**** 0.000 1.858

8.401**** 0.000 1.351

2654.754 337.886 0.000 6.248

1918.716 157.623 0.000 7.947

1870.628 433.25 0.000 6.666

1098.900 362.552 0.000 5.898

2127.069 377.899 0.000 8.679

0.619 3600

0.439 2258

0.573 3600

0.659 3600

0.370 3600

*p<¼0.10, **p<¼0.05, ***p<¼0.01,****p<¼0.001. SE ¼ Standard Error.

S. Ahmed, M.M. Khan / Social Science & Medicine 72 (2011) 1704e1710

likely to receive PNC than women not in the program. The voucher recipients were two times more likely to use ANC services and 1.5 times more likely to seek treatment for obstetric complications. It is possible that the vouchers affect the maternal health utilization rates of poor and non-poor differently. To test for the presence of a potential differential impact, we introduced an interaction term, i.e., voucher receipt and poor, into the regression models. It is interesting to note that the estimated odds ratio for this interaction term was 4.3 in the facility delivery model, indicating that poor voucher recipients were 4.3 times more likely to deliver in a health facility than the non-poor voucher recipients. Therefore, the provision of vouchers benefits the poor at a much higher rate than the non-poor. Non-poor households did not take full advantage of the vouchers, likely due to the perceived poor quality of the services offered through the health facilities in rural areas. With regard to delivery assisted by skilled health personnel and PNC use, the odds of using these services were more than two times greater for the poor voucher recipients than the non-poor voucher recipients. Discussion Demand-side financing has become increasingly popular in poor developing countries over the last decade. Despite its popularity, little is known about the effects of DSF programs on healthcare service utilization and health outcomes. This study used a relatively large household survey to examine the effects of a maternal health voucher program, a specific type of DSF intervention, on maternal health service utilization in a poor rural community of Bangladesh. Because these vouchers lower the cost of accessing healthcare services, the changes in health-seeking behavior observed in Bangladesh should be generalizable to other developing regions. In rural Bangladesh, the introduction of universal maternal health vouchers increased health service utilization rates by all socioeconomic groups. Although all pregnant women in the project area received vouchers irrespective of their economic status, the effects were significantly higher for poor women than for the non-poor. Because the poor women increased their utilization of health services at a higher rate, the overall inequality in the utilization of services became significantly lower in the project area than in the neighboring comparison areas. The introduction of the MHVS increased skilled personnel assisted delivery, health facility-based delivery and the utilization of PNC among poor women. The only indicator of maternal health that remained problematic after the introduction of vouchers is access to care for obstetric complications. Women experiencing maternal complications may view these outcomes as a direct consequence of the poor quality of care received, and such perceptions clearly make financial constraints less important in decision making when seeking care for complications. Consistent with the results reported by other researchers (Castro-Leal, Dayton, Demery, & Mehra, 2000; Brazier et al., 2009), this study also found that poor women, in general, use very low levels of maternal healthcare services in Bangladesh. However, once the financial barriers were removed, maternal health service utilization by the poor increased at a rapid rate. Improvement in the use of maternal healthcare services by poor pregnant women is essential for achieving MDG 5 (Anwar et al., 2008; Houweling, Ronsmans, Campbell, & Kunst, 2007; Starfield 2007). Our results indicate that appropriate pro-poor policy interventions such as lowering financial barriers to the use of services should improve access to care and reduce socioeconomic inequality in service utilization. The analysis suggests that the MHVS significantly improved the utilization of maternal health services in rural areas of Bangladesh. In the voucher scheme area, women from all socioeconomic groups

1709

used maternal health services at a much higher rate than the women in the comparison area. The effect of vouchers on maternal health service utilization was higher for the poor than the nonpoor. This probably indicates that the barriers faced by nonusers among economically better-off households are related to the quality of care or cultural rather than financial barriers and that the vouchers may have replaced the money households normally use to pay for services out-of-pocket. In addition, voucher schemes need time to gain momentum, and one year may not be long enough to observe effective utilization of the vouchers. In any new program, it takes time for both the recipients and the service providers to become accustomed to vouchers. Often, the non-poor is more sensitive to the quality of services, and provision of vouchers by itself may not generate significant additional demand without significant improvements in the supply environment. Utilization rates for all types of maternal health services were significantly higher in the voucher area than in the comparison area, and the inequality indices were lower. In fact, the socioeconomic inequality in service utilization (as measured by the Concentration Index, CI) for one of the maternal health services became virtually zero in the project area. This appears to contradict the so-called “inverse equity hypothesis”, which argues that providing improved access to care worsens the socioeconomic inequality in healthcare utilization in the initial stage before showing improvements at a later stage (Victora, Vaughan, Barros, Silva, & Tomasi, 2000). The demand-side financing scheme of Bangladesh reduced inequality from the onset, implying that a new intervention would not necessarily increase inequality. It is thus possible to design pro-poor health interventions even when the interventions are generalized in nature rather than narrowly targeted toward the poorer socioeconomic groups. The results herein suggest that the voucher schemes in poor developing countries are likely to be inequality reducing even in the short run. The results of this study also highlight the important problem of relying primarily on demand-side financing to achieve high levels of maternal healthcare use in low-utilization areas. Although the vouchers did expand the utilization of targeted services, most women still did not seek care. Even with the vouchers, two-thirds of pregnant women from the richest tercile and three-fourths from the poorest tercile did not use minimum levels of ANC, safe delivery or PNC. Clearly, demand-side financing alone will not be able to improve the maternal health service utilization rate to the desired level of 90% or more without a comprehensive approach of improving access to care, especially for poorer sections of the population. Despite significant improvements in the healthcare delivery infrastructure, service quality has remained very poor in Bangladesh. The high proportion of women not seeking care for pregnancy-related complications is a reflection of the perceived poor quality of services available in the area. To accelerate the improvement of maternal health outcomes, a comprehensive approach is needed that explicitly addresses financial constraints, gender disparity, social barriers to access and supply-side concerns. Strengthening the health system will be a crucial aspect of this comprehensive approach, and poor developing countries should emphasize system-wide policy changes to improve the utilization of maternal health services. Acknowledgments The authors acknowledge the financial support of the Department for International Development (DFID), Bangladesh through United Nations Population Fund (UNFPA), Bangladesh under the joint United Nations initiative “Accelerating progress toward maternal and neonatal mortality and morbidity reduction” that enabled the collection of the data used in this study. The findings of this study do not necessarily reflect those of the funding agencies.

1710

S. Ahmed, M.M. Khan / Social Science & Medicine 72 (2011) 1704e1710

References Ahmed, S., & Khan, M. (2010). A maternal health voucher scheme: what have we learned from the demand-side financing scheme in Bangladesh? Health Policy and Planning. doi:10.1093/heapol/czq015. Anastasia, M. (2009). Bolivia tackles maternal and child deaths. The Lancet., 374, 442. Anwar, I., Sami, M., Akhter, N., Chowdhury, M. E., Salma, U., Rahman, M., et al. (2008). Inequity in maternal health care services: evidence from home-based skilled birth-attendant programmes in Bangladesh. Bulletin of World Health Organization, 86, 252e259. Bashir, H., Kazmi, S., Eichler, R., Beith, A., & Brown, E. (2010). Pay for performance: Improving maternal health services in Pakistan (Working title of draft). Maternal and child health P4P case studies. Bethesda, MD: Health Systems 20/20, Abt Associates Inc. Behrman, J., & James, C. K. (1998). Population and reproductive health: an economic framework for policy evaluation. Population and Development Review, 24(4), 697e737. Bhat, R., Mavalanker, D. V., Singh, P. V., & Singh, N. (2009). Maternal healthcare financing: Gujarat’s chiranjeevi scheme and its beneficiaries. Journal of Health, Population and Nutrition, 27(2), 249e258. Bhatia, M. R., & Gorter, A. C. (2007). Improving access to reproductive and child health services in developing countries: are competitive voucher schemes an option? Journal of International Development, 19, 975e981. Bhatia, M. R., Yesudian, C. A. K., & Gorter, A. (2006). Thankappan KR. Demand side financing for reproductive and child health services in India. Economic and Political Weekly, 41(3), 279e284. Bollen, K. A., Glanville, J., & Stecklov, G. (2001). Socioeconomic status and class in studies of fertility and health in developing countries. Annual Review of Sociology, 27, 153e185. Bonu, S., Bhushan, I., Rani, M., & Anderson, I. (2009). Incidence and correlates of ‘catastrophic’ maternal health care expenditure in India. Health Policy and Planning, 24, 445e456. Borghi, J., Sabina, N., Blum, L. S., Hoque, M. E., & Ronsmans, C. (2006). Household costs of health care during pregnancy: a case study from Matlab, Bangladesh. JHPN, 24(4), 446e455. Borghi, J., Ensor, T., Somanathan, A., Lissner, C., & Mills, A. (2006). Mobilising financial resources for maternal health. Lancet, 368, 1457e1465. Brazier, E., Andrzejewski, C., Perkins, M. E., Themmen, E. M., Knight, R. J., & Bassane, B. (2009). Improving poor women’s access to maternity care: findings from a primary care intervention in Burkina Faso. Social Science & Medicine, 69, 682e690. Castro-Leal, F., Dayton, J., Demery, L., & Mehra, K. (2000). Public spending on health care in Africa: do the poor benefit? Bulletin of the World Health Organization, 78, 66e74. Celik, Y., & Hotchkiss, D. R. (2000). The socio-economic determinants of maternal health care utilization in Turkey. Social Science & Medicine, 50, 1797e1806. Dhar, R. S. G., Nagpal, J., Singha, S., Bhargava, V. L., Sachdeva, A., & Bhartia, A. (2009). Direct cost of maternity-care services in South Delhi: a community survey. Health, Population and Nutrition, 27(3), 368e378. van Doorslaer, E., Koolman, X., & Jones, A. M. (2004). Explaining income-related inequalities in doctor utilization in Europe. Health Economics, 13, 629e647. van Doorslaer, E., & Masseria, C. (2004). Income-Related inequality in the use of medical care in 21 OECD countries. In OECD, towards high performing health systems, Health Policy Studies (pp. 109e166). Paris: OECD. Ensor, T. (2004). Consumer-led demand side financing in health and education and its relevance for low and middle income countries. International Journal of Health Planning and Management., 19, 267e285. Ensor, T., & Cooper, S. (2004). Overcoming barriers to health service access: influencing the demand side. Health Policy Plan, 19, 69e79. Fotso, J., Ezeh, A. C., & Essendi, H. (2009). Maternal health in resource-poor urban settings: how does women’s autonomy influence the utilization of obstetric care services? Reproductive Health, 6, 9. doi:10.1186/1742-4755-6-9. Filmer, D., & Pritchett, L. (2001). Estimating wealth effect without expenditure data or tears: an application to educational enrollments in States of India. Demography, 38, 115e132.

Filmer, D., & Pritchett, L. (1999). The effect of household wealth on educational attainment: evidence from 35 countries. Population and Development Review, 25, 85e120. Gorter, A. C., & Bellows, B. (2008). Do competitive voucher schemes improve the provision of health care to underserved and/or vulnerable population groups? Experiences from Nicaragua, India and Africa. In Seminar given at Department of Social Medicine, University of Bristol. Government of the People’s Republic of Bangladesh. (2004). Bangladesh population census 2001. Bangladesh Bureau of Statistics. Dhaka: Planning Division, Ministry of Planning. Government of the People’s Republic of Bangladesh. (2007). Demand side financing pilot maternal health voucher scheme proposal. Dhaka: Ministry of Health and Family welfare, Government of the People’s Republic of Bangladesh. Government of the People’s Republic of Bangladesh. (2009). Bangladesh demographic and health survey 2007. Dhaka: National Institute of Population Research and Training. 118. Gupta, I., Joe, W., & Rudra, S. (2010). Demand side financing in health: how far can it address the issue of low utilization in developing countries? Geneva: World Health Organization. World Health report 2010, Background Paper # 27. Hatt, L., Nguyen, H., Sloan, N., Miner, S., Magvanjav, O., et al. (2010). Economic evaluation of demand-Side financing (DSF) for maternal health in Bangladesh [Draft]. Bethesda, MD: Review, Analysis and Assessment of Issues Related to Health Care Financing and Health Economics in Bangladesh. Abt Associates Inc. Houweling, T. A., Ronsmans, C., Campbell, O. M. R., & Kunst, A. E. (2007). Huge poorrich inequalities in maternity care: an international comparative study of maternity and child care in developing countries. Bulletin of the World Health Organization, 85(10), 745e754. Ir, P., Horemans, D., Souk, N., & Damme, W. V. (2010). Using targeted vouchers and health equity funds to improve access to skilled birth attendants for poor women: a case study in three rural districts in Cambodia. BMC Pregnancy and Childbirth, 10(1). Koblinsky, M., Matthews, Z., Hussein, J., Mavalankar, D., Mridha, M., Anwar, I., et al. (2006). Going to scale with professional skilled care. Lancet, 368, 1377e1386. Lenel, A., & Griffith, D. (2007). Voucher schemes in the health sector: The experience of German Financial Cooperation. Working Paper. Frankfurt am Main: KfW Bankengruppe, Corporate Communication. Lu, J. F., Leung, G. M., Kwon, S., Tin, K. Y., Van Doorslaer, E., & O’Donnell, O. (2007). Horizontal equity in health care utilization evidence from three high-income Asian economies. Social Science & Medicine, 64, 199e212. Mekonnen, Y., & Mekonnen, A. (2003). Factors influencing the use of maternal health care services in Ethiopia. Journal of Health, Population and Nutrition, 21(4), 374e382. O’Donnnell, O., van Doorslaer, E., Wagstaff, A., & Lindelow, M. (2007). Analyzing health equity using household survey data: A guide to techniques and their implementation. Washington, DC: The International Bank for Reconstruction and Development/The World Bank. Rutstein, S. O., & Johnson, K. (2004). The DHS wealth index. DHS Comparative report No. 6. Calverton, MD: ORC Macro. SPSS Inc. (2008). SPSS Base 17.0 for Windows user’s guide. Chicago IL: SPSS Inc. Starfield, B. (2007). Pathways of influence on equity in health. Social Science & Medicine, 64, 1355e1362. Thaddeus, S., & Maine, D. (1994). Too far to walk: maternal mortality in context. Social Science & Medicine, 38, 1091e1110. United Nations Population Fund. (2006). Rapid assessment of Chiranjeevi Yojana in Gujarat. New Delhi: UNFP. Victora, C. G., Vaughan, J. P., Barros, F. C., Silva, A. C., & Tomasi, E. (2000). Explaining trends in inequities: evidence from Brazilian child health studies. The Lancet, 356, 1093e1098. Wagstaff, A., & Watanabe, N. (2003). What difference does the choice of SES make in health inequality measurement? Health Economics, 12, 885e890. Wagstaff, A., & van Doorslaer, E. (2000). Equity in health care finance and delivery. In A. J. Culyer, & J. P. Newhouse (Eds.), Handbooks in economics (pp. 1803e1862). Amsterdam: Elsevier, North-Holland. Witter, S., Adjei, S., Armar-Klemesu, M., & Graham, W. (2009). Providing free maternal health care: ten lessons from an evaluation of the national delivery exemption policy in Ghana. Global Health Action. doi:10.3402/gha.v210.1881.