Improving Access to Delivery Care and Reducing the

0 downloads 0 Views 673KB Size Report
ernment of Bangladesh introduced a DSF scheme known as Maternal Health Voucher ..... Chiranjeevi Yojana voucher program in Gujarat, India showed that the ...
Program Evaluation

Improving Access to Delivery Care and Reducing the Equity Gap Through Voucher Program in Bangladesh: Evidence From Difference-in-Differences Analysis

International Quarterly of Community Health Education 0(0) 1–9 ! The Author(s) 2018 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/0272684X17749568 journals.sagepub.com/home/qch

Kaji Tamanna Keya1, Benjamin Bellows2, Ubaidur Rob3, and Charlotte Warren1

Abstract To test a statistically significant change in delivery by medically trained providers following introduction of a demand-side financing voucher, a population-based quasi-experimental study was undertaken, with 3,300 mothers in 2010 and 3,334 mothers at follow-up in 2012 in government-implemented voucher program and control areas. Results found that voucher program was significantly associated with increased public health facility use (difference-in-differences (DID) 13.9) and significantly increased delivery complication management care (DID 13.2) at facility although a null effect was found in facility-based delivery increase. A subset analysis of the five wellfunctioning facilities showed that facility deliveries increased DID 5.3 percentage points. Quintile-based analysis of all facilities showed that facility delivery increased more than threefold in lower quintile households comparing to twofold in control sites. The program needs better targeting to the beneficiaries, ensuring available gynecologist–anesthetist pair and midwives, effective monitoring, and timely fund reimbursements to facilities. Keywords equity, access, delivery care, voucher, Bangladesh, difference-in-differences

Background Underutilized and low-quality maternity services at health facilities are common in low- and middle-income countries (LMIC) and over the past decade, policymakers have been revising policies to address the persistent challenge by adopting demand-side financing (DSF), such as vouchers and conditional cash transfers that encourage consumers to seek care from qualified service providers.1–7 In voucher initiatives, the beneficiaries, usually disadvantaged and underserved populations, are given a subsidy to seek care from skilled healthcare providers. Available evidence indicates that vouchers can improve service utilization and quality among target populations.1,4,8–12 The strategy can reduce consumer out-of-pocket costs of healthcare and increases demand for services. In Bangladesh, the government implemented a DSF program, enrolling mostly public facilities progressively across subdistricts or upazilas. The previous studies conducted in voucher areas showed that voucher was associated with increase in facility-based deliveries.13–15 Although there is evidence indicating that voucher programs can increase the use of subsidized health services, there is a lack of evidence regarding the impact of the voucher on reducing inequitable utilization trends between rich and poor.4,10,16–19

Primary maternal health services continue to favor wealthier households in LMIC.20–22 In Bangladesh, 90% of deliveries in the lowest quintile occur at home, whereas only 40% of deliveries in the highest quintile occur at home.23 In Bangladesh, public facilities are underutilized due to low-quality services, the out-of-pocket burden of medicines and surgical procedures, and transportation cost24,25 with the result in an estimated annual 7,000 maternal deaths.26 To increase safe delivery among poor rural women, the government of Bangladesh introduced a DSF scheme known as Maternal Health Voucher Scheme in 2006.15,27,28 The program promotes maternal health services including institutionalized delivery and home-based delivery assisted by medically trained providers (MTP) to the poor and disadvantaged population in selected subdistricts (upazilas). The MTP cadre

1

Population Council, Washington, DC, USA Population Council, Lusaka, Zambia 3 Population Council, Dhaka, Bangladesh 2

Corresponding Author: Kaji Tamanna Keya, Population Council, 4301 Connecticut Avenue, Washington, DC 20008, USA. Email: [email protected]

2 includes medical doctors, nurses, family welfare visitors, midwives, paramedics, and community skilled birth attendants.23 Using preset criteria, government field workers identify poor pregnant mothers. Criteria are monthly income under Taka 3000 (US$37.50), the amount of land not more than 15 decimal (also spelled decimal, it is a unit of area in Bangladesh approximately equal to 1/100 acre) and not more than one child.27 The program distributes vouchers to poor pregnant women entitling them to three free antenatal care (ANC) visits, delivery (normal and cesarean), management of complications, emergency referral, and postnatal care (PNC) services; free medicine for complications and delivery; and cash stipends for transportation. In addition, mothers can receive a conditional cash transfer of Taka 2,000 (US$29) and an in-kind incentive (gift box) if she delivers with a designated qualified service provider at home or facility.11,27 The DSF scheme also reimburses facilities, which receive funds to be proportionately divided among designated staff and a facility maintenance fund. Generally, 50% of the reimbursement funds are deposited in the facility’s ‘‘seed fund’’ from where associated expendable costs are incurred. Thus, the DSF for maternal healthcare in Bangladesh is a combination of supply-side incentives for providers and demand-side cash transfer and subsidies for clients. The DSF program covered 35 upazilas by the end of the first two phases and in the third phase in 2010, another 11 upazilas were added. Population Council supported by the Bill and Melinda Gates Foundation undertook a comprehensive evaluation of the 11 third phase DSF upazilas with two rounds of surveys in 2010 and 2012.11 The aim of the study was to empirically examine the effect of the DSF program on increasing delivery by MTPs either at home or facilities and reducing the equity gap in facility use between rich and poor. Specifically, when the voucher is targeted for the poor, does DSF increase delivery service use among the poorest women?

Methodology Study Design Bangladesh’s DSF program was piloted in 2004 in 21 upazilas but did not start until 2006 and expanded in a second phase to an additional 12 upazilas in 2007.15 For the third phase, the Directorate General of Health Services identified 11 administrative upazilas that had comparatively lower healthcare service utilization and a high maternal mortality rate. These 11 upazilas were added to the voucher program in 2010, which presented an opportunity to conduct a robust external evaluation. For the control sites of this study, 11 non-DSF matched upazilas were selected from the same or a nearby district based on several characteristics, for example, availability of comprehensive or emergency basic obstetric care services, number of available service providers and support

International Quarterly of Community Health Education 0(0) staff, number of beds, presence of an anesthesiologist and gynecologist pair, and literacy rate of the area as a proxy for socioeconomic status.11 To evaluate the impact of the DSF scheme, a quasiexperimental design was used. The national figure of 14.6% of births at the facility was assumed as the baseline level of facility-based births in the voucher areas. To detect a 12% increase in the proportion of facility-based births, 1,650 experimental subjects and 1,650 control subjects were required to be able to reject the null hypothesis that the proportion of facility-based births for experimental and control subjects are equal with probability (power) of 0.8.

Survey Design From each of the 22 upazilas, 150 respondents were selected through multistage sampling. Three of nine unions from each upazila were selected through probability proportional to size (PPS) to get the required number of samples, that is, 50 respondents per union. The next stage comprised the selection of three villages from each union through PPS. Finally, from each village, the required numbers of respondents were selected at random from the list of pregnant mothers prepared by fieldworkers. A total of 3,300 respondents were interviewed in 2010 and 3,334 were interviewed in 2012. The key dependent variables for this study were both delivery in a health facility and delivery by MTP in the community. The key independent variable was the economic status as measured by a survey-specific index of household assets (i.e., wealth index). Data were collected on respondent’s age, education, voucher utilization, parity, and the use of maternal health services. To measure economic status, a study-specific wealth index was calculated for 3,300 households in 2010 and 3,334 households in 2012. The wealth index is used as a background characteristic in tables and figures and has been tested in a number of studies to test for variation in inequalities in household income, use of health services, and health outcomes.13,29–32 As an indicator of the level of relative wealth, it is consistent with expenditure and income measures.

Analysis The key independent variable, the wealth index, was constructed from household asset data using principal components analysis. Asset information was collected in the household questionnaire, including information on household ownership of consumer items ranging from a mobile phone and radio to a bicycle or boat, as well as dwelling characteristics like building materials and land ownership. Each asset was assigned a weight (factor score) generated through principal component analysis, and the resulting asset scores were standardized on a normal distribution with a mean of zero and standard deviation of one.33 Each household was then assigned a score for each asset, and the scores were summed

Keya et al. for each household. Individuals were ranked according to the total score of the household in which they resided. The sample was then divided into quintiles (five groups) from one (lowest) to five (highest). The DSF voucher intervention was not implemented at the same pace and same time in all 11 intervention facilities. Therefore, to understand the effect of the voucher based on fidelity to the DSF voucher model, a subset analysis was done with the household survey data collected nearby five wellfunctioning DSF facilities. The matched control facility was considered as a control facility for the subset analysis. To identify a study area as a well-functioning area, two things were considered: the number of facility deliveries that took place in an upazila utilizing voucher and the number of voucher beneficiaries available for interview during our follow-up data collection in 2012. To evaluate the effects of vouchers on socioeconomic disparity in delivery care, two different measures were examined: 1. Whether the respondent delivered in a health facility for her most recent pregnancy 2. Whether the respondent was assisted by an MTP for her most recent pregnancy either in facility or in home

Statistical Analysis A difference-in-differences (DID) was estimated to evaluate the impact of the voucher program on the utilization of maternal delivery care. DID is calculated by subtracting changes in delivery services between 2012 and 2010 in voucher areas minus the difference in changes in the outcome in control areas. This DID calculation model was followed both in 11 voucher areas and 5 well-functioning areas       d^ ¼ Y2012 Voucher  Y2010Voucher  Y2012Control  Y2010Control

The wealth quintile was cross-tabulated with the place of delivery, type of delivery, type of health facility, and type of provider. Concentration curves plotting the cumulative outcome of delivery by the cumulative percentage of women ranked by wealth were created to graphically present inequality in the use of delivery service by wealth status.34,35 For delivery service, rich-to-poor equity ratios (ERs) were calculated dividing the highest quintile (Q5) by the lowest quintile (Q1). An ER of 1 means that service utilization is the same for the poor and the rich, an ER of more than 1 means service utilization is pro-rich, and an ER of less than 1 means service utilization is pro-poor. The concentration curve was used to show the equity gap in different health outcomes such as facility delivery and delivery service by MTP in voucher and nonvoucher areas. The line of equity shows the equality of health outcomes among different quintiles. In the y axis, the cumulative outcome of

3 facility delivery and delivery by MTP are shown. Women are plotted on the x axis by the wealth quintile starting from poorest to richest (Figures 4 to 7)

Results Although facility deliveries appeared to increase from 2010 to 2012, the change was not statistically significant as the control areas experienced a similar increase (Table 1). The subgroup analysis comparing the five well-functioning voucher facilities against five controls showed that facility deliveries had a net increase of 5.3 percentage points after differencing away the positive trend in controls. There was a statistically significant increase in the use of public facilities (DID 13.9 percentage points, p ¼ .004), a sharp decline in the use of for-profit private facilities (negative DID 17.9 percentage points, p < .001), and a significant increase in deliveries at nongovernment organization (NGO) facilities (DID 3.9 percentage points, p ¼ .018), which suggests that voucher may have encouraged women to go to public and NGO facilities, who instead would have gone to a forprofit private facility. Among the public facilities, delivery at upazila hospitals was more common (81%) in intervention areas compared with control areas (67%). By end line, one third of the births in voucher areas were attended by MTPs with no significant difference between voucher and control areas. Table 2 shows that women who had delivery complications received services for that complication more in the 11 voucher intervention areas compared with matched control areas (DID 3.9 percentage points). This significant improvement in the uptake of delivery complication services at public facilities versus home locations in voucher areas compared with controls (DID 13.2 percentage points). In the voucher areas, a significant number of women received the complication services from doctors (DID 12.7 percentage points), while women in comparison areas mostly received the services from unqualified providers. Figure 1 shows the percentage by wealth quintile of respondents who received a voucher. The highest percentage of respondents belong to the second wealth quintile and altogether just under half of the voucher recipients fall in the first and second quintiles and nearly another half of the voucher recipients fall in the third and fourth wealth quintiles. As these wealth quintiles do not correspond to the beneficiary eligibility criteria, it is difficult to determine whether there were problems with targeting; however, from a simple propoor perspective, one would prefer to see a greater proportion of beneficiaries in the lower quintiles. Figure 2 reveals that, prior to the introduction of the voucher, there was a large equity gap in the proportion of women delivering at facility, for example, in the lowest quintile only 9% delivered at facility and in the highest quintile 40% delivered at facility. After the introduction of the voucher program, the equity gap reduced; among the poor,

4

International Quarterly of Community Health Education 0(0)

Table 1. Changes in the Uptake of Delivery Services. Voucher areas Type of service Place of delivery Home Facility (11 voucher and 11 nonvoucher facilities) N Facility (5 well-functioning voucher and 5 nonvoucher facilities) N Facility (6 low functioning voucher and 6 nonvoucher facilities) N Type of facility Public Private NGO N Public facility type Tertiary hospital UHC MCWC/HFWC/CC N Type of delivery Normal Cesarean Assisted N Type of provider Doctor Nurse/FWV/midwife CSBA Unqualified provider N Delivery by MTP N

Nonvoucher areas

2010

2012

2010

2012

DID

p

81.5 18.5 1,650 22.8 750 15.0 900

68.9 31.1 1,662 41.2 746 23.0 916

79.3 20.7 1,650 21.6 750 20.0 900

68.2 31.8 1,672 34.7 760 29.3 912

1.5 1.5

.489 .489

5.3

.125

1.3

.634

41.2 57.2 1.6 306

50.9 43.3 5.8 517

37.7 60.8 1.5 342

33.5 64.8 1.7 532

13.9 17.9 3.9

.004** .000*** .018*

26.2 65.1 8.7 126

14.1 81.0 4.9 263

25.6 54.3 20.1 129

19.7 66.9 13.4 178

6.2 3.3 2.9

.324 .642 .549

89.3 9.2 1.5 1,650

80.0 17.1 2.9 1,662

85.3 13.0 1.7 1,650

77.7 19.7 2.6 1,672

1.7 1.2 0.5

.345 .500 .439

11.9 8.1 0.7 79.3 1,650 20.7 1,650

18.8 12.9 1.9 66.4 1,662 33.6 1,662

14.7 8.9 0.4 76.0 1,650 24.0 1,650

21.5 12.5 1.1 64.9 1,672 35.1 1,672

0.1 1.2 0.5 1.8

.961 .417 .265 .397

1.8

.397

Note. DID, difference-in-differences; NGO, nongovernment organization; MCWC, Maternal and Child Welfare Center; HFWC, Health and Family Welfare Center; CC, community clinic; FWV, female welfare volunteer; CSBA, community skilled birth attendant; MTP, medically trained providers; UHC, Upazila Health Complex. *p