Health care services use, stillbirth, and neonatal and infant survival following implementation of the Maternal Health Voucher Scheme in Bangladesh: A difference-in-differences analysis of Bangladesh Demographic and Health Survey data, 2000 to 2016.

BACKGROUND: Starting in 2006 to 2007, the Government of Bangladesh implemented the Maternal Health Voucher Scheme (MHVS). This program provides pregnant women with vouchers that can be exchanged for health services from eligible public and private sector providers. In this study, we examined whether access to the MHVS was associated with maternal health services utilization, stillbirth, and neonatal and infant mortality. METHODS AND
FINDINGS: We used information on pregnancies and live births between 2000 to 2016 reported by women 15 to 49 years of age surveyed as part of the Bangladesh Demographic and Health Surveys. Our analytic sample included 23,275 pregnancies lasting at least 7 months for analyses of stillbirth and between 15,125 and 21,668 live births for analyses of health services use, neonatal, and infant mortality. With respect to live births occurring prior to the introduction of the MHVS, 31.3%, 14.1%, and 18.0% of women, respectively, reported receiving at least 3 antenatal care visits, delivering in a health institution, and having a skilled birth attendant at delivery. Rates of neonatal and infant mortality during this period were 40 and 63 per 1,000 live births, respectively, and there were 32 stillbirths per 1,000 pregnancies lasting at least 7 months. We applied a difference-in-differences design to estimate the effect of providing subdistrict-level access to the MHVS program, with inverse probability of treatment weights to address selection into the program. The introduction of the MHVS program was associated with a lagged improvement in the probability of delivering in a health facility, one of the primary targets of the program, although associations with other health services were less evident. After 6 years of access to the MHVS, the probabilities of reporting at least 3 antenatal care visits, delivering in a health facility, and having a skilled birth attendant present increased by 3.0 [95% confidence interval (95% CI) = -4.8, 10.7], 6.5 (95% CI = -0.6, 13.6), and 5.8 (95% CI = -1.8, 13.3) percentage points, respectively. We did not observe evidence consistent with the program improving health outcomes, with probabilities of stillbirth, neonatal mortality, and infant mortality decreasing by 0.7 (95% CI = -1.3, 2.6), 0.8 (95% CI = -1.7, 3.4), and 1.3 (95% CI = -2.5, 5.1) percentage points, respectively, after 6 years of access to the MHVS. The sample size was insufficient to detect smaller associations with adequate precision. Additionally, we cannot rule out the possibility of measurement error, although it was likely nondifferential by treatment group, or unmeasured confounding by concomitant interventions that were implemented differentially in treated and control areas.
CONCLUSIONS: In this study, we found that the introduction of the MHVS was positively associated with the probability of delivering in a health facility, but despite a longer period of follow-up than most extant evaluations, we did not observe attendant reductions in stillbirth, neonatal mortality, or infant mortality. Further work and engagement with stakeholders is needed to assess if the MHVS has affected the quality of care and health inequalities and whether the design and eligibility of the program should be modified to improve maternal and neonatal health outcomes.

Introduction

Progress in population health in Bangladesh has been called a paradox. Despite high rates of poverty, low per capita gross domestic product (GDP), and modest health systems investments [1,2], the country has made substantial progress in health since its independence in 1971, including dramatic reductions in under 5 and maternal mortality [1,3]. Despite its impressive progress, rates of maternal and neonatal mortality, estimated at 173 per 100,000 live births in 2017 and 17 per 1,000 live births in 2018 [4,5], respectively, remain high in Bangladesh compared to 2030 Sustainable Development Goal (SDG) targets. Similarly, many women are not receiving essential maternal health services, with especially low rates of coverage for poorer households in rural areas [6-8]. According to the 2017 to 2018 Demographic and Health Surveys (DHS), among women ages 15 to 49 years who reported a live birth in the preceding 3 years, only 44% received 4 or more antenatal care visits, 49% delivered in a health facility, and 53% were assisted at delivery by medically trained personnel [9]. Further improvements in maternal and newborn health necessary to achieve the SDGs in these areas will require expanded coverage of quality reproductive and maternal health interventions that reach underserved populations [10,11].

Supply-side funding has been the primary strategy for increasing access to health services in Bangladesh, with the government providing funds to public facilities to provide free or highly subsidized maternal care [12]. Nonetheless, access is not universal, due in part to the inadequate supply of skilled health professionals [13], unofficial user fees, and discrimination by public providers against poor women [14,15], with many turning to private facilities [16]. Further supply-side investment is unlikely to reach socioeconomically disadvantaged groups without accompanying efforts that address demand-side barriers [2,8,11]. These barriers include distance to facilities, as well as travel costs and other out-of-pocket expenditures (e.g., medicines) that are not covered [16,17], and can be substantial even at public facilities [18,19]. Cultural beliefs and norms, lack of awareness about sources of care, and social stigma also shape demand for maternal health services [12].

Demand-side funding schemes are designed to increase the quantity and quality of maternal health services used by underserved populations by reducing financial barriers to access and increasing competition in the market. One example is Bangladesh’s Maternal Health Voucher Scheme (MHVS), which provides pregnant women with vouchers that can be exchanged for health services from eligible public and private sector providers who are later reimbursed.

The MHVS was implemented by the Bangladesh Ministry of Health and Family Welfare (MOHFW), with technical and financial support from several multilateral and bilateral organizations, and aims “to increase demand and utilization of maternal health services, to improve access to and utilization of safe delivery, to encourage institutional delivery, and to improve equity in the utilization of maternal health services,” with the ultimate goal of improving maternal and neonatal health outcomes [20,21]. It was introduced in 2006 to 2007 as a pilot program in 21 upazilas, which are subdistrict administrative units with a local governance structure that direct development activities within their boundaries [22]. By 2010, the MHVS expanded to cover 53 of Bangladesh’s 545 upazilas (based on 2011 census boundaries), with another 2 upazilas gaining access in 2017 to 2018. According to the program’s design, the MHVS was prioritized to upazilas with greater need based on their sociodemographic profile (e.g., literacy rate, poverty level), as well as the presence of health workers to help administer the program. However, in practice, the program was not geographically targeted based on poverty rates [21]. Within upazilas selected to participate in the program, individual-level access was initially universal in some districts and means-tested in others, but gradually shifted to a means-tested program in all upazilas. To enroll beneficiaries, community health workers identified pregnant women in their first trimester and applied to a local committee to determine if eligibility criteria were met based on residence in the upazila, parity, income, and land and asset ownership (). These criteria, however, were not consistently enforced. For example, vouchers were sometimes given to women who were not pregnant for the first or second time and poverty criteria were difficult to verify, resulting in imperfect targeting to poorer households [21,23]. Community health workers assisted with the distribution of program information, vouchers, referrals, subsidies, and cash incentives.

Specific services and fees covered by vouchers included: consultation fees for up to 3 antenatal care checkups, delivery by a skilled birth attendant at the home or a health facility, management of complications including eclampsia management and delivery by cesarean section, and a single postnatal care checkup within 6 weeks of delivery. Services were covered if received from accredited public, nongovernmental, or private sector providers with the capacity to provide emergency obstetric care. Providers were reimbursed through locally assigned banks after submitting evidence of service provision. Beneficiaries were reimbursed up to 500 Taka (Tk.) (7.20 US dollars (USD) based on a 2010 annual average exchange rate of 0.0144 USD per Tk.) for transport to antenatal care visits, delivery, and postnatal care services and to cover referral fees from subdistrict to district-level facilities. Additionally, beneficiaries received a cash payment of Tk. 2,000 (USD 28.80) for delivering in a health facility or Tk. 500 (USD 7.20) for delivering at home with a skilled attendant, as well as a baby-care gift box valued Tk. 500 (USD 7.20). Although it is primarily a demand-side funding scheme, the program also aims to address the supply-side through payments to public sector facilities to increase the availability and quality of maternal health services. Further details regarding the design and operation of the MHVS are available elsewhere [20,21].

Subnational, cross-sectional comparisons of voucher recipients versus nonrecipients, as well as mothers who had recently given birth in upazilas where the program was offered versus not, suggest that the MHVS was associated with greater use of maternal health services and more comprehensive maternal care, with inconsistent evidence on out-of-pocket expenditures [18,20,24-27]. However, this research cannot inform inference regarding the causal effect of the MHVS. The only quasi-experimental evaluation, a differences-in-differences analysis that compared changes in outcomes for 11 upazilas before and after they were added to the program in 2010 to corresponding changes from 11 matched control upazilas, suggests that the program increased use of public facilities, but did not influence the proportion of women receiving maternal health services unless the treated group was restricted to 5 high-performing upazilas [28,29].

In this study, we used national information on pregnancies and live births reported by women surveyed as part of the 2004, 2007, 2011, 2014, and 2017 to 2018 Bangladesh Demographic and Health Surveys (BDHS) to evaluate the association between upazila-level access to the MHVS and maternal health services utilization, stillbirth, and neonatal and infant mortality. Our evaluation of the MHVS was designed to address several knowledge gaps concerning the impact of maternal voucher programs, including: (1) if short-term increases in the use of priority maternal health services are sustained over time [30,31]; and (2) whether they affect maternal and neonatal health outcomes [30-32], where population-level evidence is limited [30,33].

Methods

Study design

We used a difference-in-differences (DD) approach [34-36] to estimate the “intention-to-treat” (ITT) effect of providing upazila-level access to the MHVS (the treatment) on our outcomes of interest. The standard DD estimate compares outcome trends before versus after an intervention with corresponding trends among a “control” group, which represents the counterfactual. Since access to the MHVS was expanded to treated upazilas over the course of approximately 4 years (i.e., variation in treatment timing), we used event study models to: (1) assess whether outcome trends in treated and control areas were parallel in the preintervention period; and (2) estimate the effect of providing access to the MHVS in treated upazilas (i.e., the average treatment effect on the treated) [37,38]. A prospective analysis plan, available in , was used in designing the study, but was not preregistered or published. In contrast to the analysis plan, we added stillbirth as an outcome and did not evaluate whether the program has affected socioeconomic inequalities in our primary outcomes, due to inadequate sample sizes and the imprecision of stratified estimates. In addition, we incorporated event study models and inverse probability weights in our DD analyses to assess evidence for parallel preintervention trends and construct more comparable treatment groups. This study is reported as per the Reporting of studies Conducted using Observational Routinely-collected Data (RECORD) guidelines ().

Data

Individual-level data were derived from the 2004, 2007, 2011, 2014, and 2017 to 2018 rounds of the BDHS, which are conducted in coordination with the worldwide DHS program [39]. The BDHS are repeated cross-sectional household surveys that provide information on the sociodemographic, health, and nutritional profile of the population [40]. The BDHS uses a stratified, multistage cluster sampling scheme based on the Bangladesh census and covers the entire population residing in noninstitutional dwellings. In the first stage, enumeration areas (EAs), typically city blocks in urban areas and villages in rural areas, are selected with the probability of selection proportional to EA size. In the second stage, 30 households were randomly sampled from selected EAs. Based on the household roster, all ever-married women (ages 15 to 49 in 2014 and 2017 to 2018, 12 to 49 in 2011, 10 to 49 in 2007, and 10 to 49 in 2004) who were usual members of a selected household were asked about their contraceptive use, reproductive history, and maternal health services use, among other factors. Comparability of data across waves is enhanced through interviewer training and standardized survey methods. Response rates for eligible women ranged between 97.9% and 98.8% across the 5 waves. Further details regarding the sampling, survey, and quality control procedures are available elsewhere [9,40-43]. For our analyses, we used information reported by women ages 15 to 49 to create samples of pregnancies and live births during the study period from 2000 to 2016. Details regarding sample selection are reported in . The Institutional Review Board of McGill University reviewed and approved this study (IRB Study Number A05-E42-17B).

Measures

Treatment

The treatment, measured at the upazila level, was gaining access to the MHVS. The list of the 55 treated upazilas that introduced the MHVS during the study period was available through public reports [20,21,29]. However, the month and year of implementation was not recorded for all treated upazilas. Through personal communication (AQ) with government officials with direct knowledge of the MHVS, we obtained the years of implementation for each upazila. We were advised to assume a starting month of December, which is the midpoint of the financial year, since it provides time for funds allocated to the program to be dispersed and the program to be implemented. Because access to the program was not randomized across upazilas, we used upazila-level administrative data () from the World Bank’s Bangladesh Interactive Poverty Maps [44] to model the probability of treatment and create more exchangeable control groups.

We used ArcGIS software to map the location of each respondent’s EA, as a proxy for the location of their residence, to its corresponding upazila based on Global Positioning System coordinates from the center of the EA, which are provided by the BDHS. This allowed us to identify if each respondent lived in a treated or control upazila and, if treated, the timing of recorded pregnancies and live births in relation to program implementation. shows MHVS implementation across upazilas. We added a buffer of 2.5 km around the geographic boundaries of treated upazilas, in order to account for the random displacement of EAs in publicly available DHS data and potential spillovers, as discussed in Appendix A in .

Implementation of the Bangladesh MHVS by year across subdistricts (upazilas) and locations of BDHS EAs, represented by points.

Base layer extracted from the GADM database (www.gadm.org) version 3.6 (April 2018), available here: https://geodata.ucdavis.edu/gadm/gadm3.6/shp/gadm36_BGD_shp.zip. BDHS, Bangladesh Demographic and Health Survey; EA, enumeration area; MHVS, Maternal Health Voucher Scheme.

Outcomes

Indicators of maternal health services utilization, with respect to live births, included whether women reported: at least 3 antenatal care visits, which is consistent with the number of visits covered by the MHVS; having a skilled birth attendant at delivery, including a qualified doctor, nurse, midwife, paramedic, community skilled birth attendant, or family welfare visitor [45,46]; delivering in an institution or health facility outside the home, including public, nongovernmental, and private medical hospitals, clinics, and health centers; and whether the delivery was by cesarean section. We defined stillbirth as a pregnancy lasting at least 7 months that did not result in a live birth [47]. We created binary indicators for neonatal and infant mortality to record deaths occurring within the first 28 days and 1 year of life, respectively, among live births.

Other covariates

We included household and individual-level characteristics that might be associated with the outcomes of interest. Demographic characteristics included women’s age at marriage, household size (categorized as 1 to 4, 5 to 6, or 7 or more people), rural/urban residence, and division of residence. Socioeconomic characteristics included women’s and husbands’ educational attainment (categorized as none, primary, or secondary or higher) and a measure of household wealth (dichotomized as above or below the median value), which is based on ownership of specific assets, environmental conditions, and housing characteristics and was constructed by DHS using the method proposed by Filmer and Pritchett [48,49]. Pregnancy and birth-related characteristics included women’s age at the time of the stillbirth or live birth, measured continuously, a binary indicator for whether the interval between the index birth or pregnancy outcome and a prior birth was short (<24 months) or not (24+ months and first births), and the number of prior stillbirths.

Statistical analyses

Event study models are an extension of the standard 2-way fixed effects DD model in which the treatment effect is estimated by a series of leads and lags representing the timing of each observation (recorded pregnancies and live births) in relation to when the MHVS was introduced in the respondent’s upazila of residence:

In the linear probability model above, we estimated for observation i in upazila c in year t the effect of upazila-level access to the MHVS on the probability of each outcome, Y, where the treatment, , is modeled as a series of 2-year event periods; the indicators for these periods, l, included 4 leads (i.e., 6+, 4 to 6, 2 to 4, and 0 to 2 years prior to the introduction of the MHVS) and 4 lags (i.e., 0 to 2, 2 to 4, 4 to 6, and 6+ years after the introduction of the MHVS), with the 2-year period prior to implementation taken as the reference. Observations from upazilas that did not receive access to the MHVS during the study period were assigned zeros for each indicator.

To account for unmeasured time-fixed differences between treatment groups, we included a fixed effect, τ, that indicated whether the observation came from an upazila that gained access to the MHVS during the study period or not. Because the BDHS resamples upazilas in each round, we could not include upazila-level fixed effects. However, as described in Appendix B in , we estimated the probability that an upazila gained access to the MHVS program during the study period, conditional on upazila-level sociodemographic characteristics, restricted the sample to the region of common support, and weighted models by the stabilized inverse probability of treatment weight [50,51]. We included fixed effects for 6 divisions (based on the 6 divisions represented in the 2004 and 2007 surveys), γ, to account for inter-division differences that might be correlated with the treatment and outcomes of interest. We also included fixed effects for year of birth (or the pregnancy outcome in analyses of stillbirth), ω, to account for secular trends in outcomes during the study period shared between treated and control upazilas. We controlled for a vector of time-varying individual-level covariates, X, taken at the time of survey and assigned to each observation in year t, including socioeconomic, pregnancy, and birth-related characteristics. Conditional on the other covariates in the model, the coefficient for each event period, μ, represents the difference in the probability of the outcome between treated and control observations in the event period relative to that difference in the reference period. Model (1) was used to examine potential violations of the parallel trends assumption [37,38]. In addition, our study design assumes no unmeasured time-varying confounding and no residual confounding, after weighting, by fixed differences between upazilas within divisions [36,52].

We also estimated the overall DD effect using a standard 2-way fixed effects model that replaced the event periods in Eq 1 with a time-varying indicator, A, equal to 1 if the observation was from a treated upazila after it had gained access to the MHVS, and lagged by 9 months since observations measured directly after the upazila gained access to the MHVS would not have been eligible for the program.

To assess the robustness of our main findings, we compared our main event study estimates with a 2.5 km buffer around each treated upazila to those with no buffer and a larger 5 km buffer. Additionally, we compared our main results from weighted event study models to unweighted estimates. All models incorporated cluster robust standard errors to account for clustering of observations within upazilas [53].

Results

Among women 15 to 49 years of age at the time of their interview, we restricted our samples to pregnancies and live births occurring in the 5 years prior to interview during the study period from 2000 to 2016; sample sizes ranged from 24,699 live births for analyses of antenatal care to 37,909 pregnancies lasting at least 7 months for analyses of stillbirth (), from women in 508 upazilas (52 treated and 456 control). After excluding observations with missing outcome or covariate information (ranging from 2.8% to 3.3% across outcomes), sample sizes for complete case analyses ranged from 23,947 live births for analyses of antenatal care to 36,677 pregnancies for analyses of stillbirth.

Sample characteristics for live births and pregnancies prior to December 2006, when the MHVS was first introduced in treated upazilas, are presented in Tables and , respectively. In general, women from areas that gained access to the MHVS were more socioeconomically disadvantaged, in terms of educational attainment and household wealth, and were more likely to come from rural areas. Women from treated areas were also less likely to report use of health services than women in control areas, including at least 3 antenatal care visits (22.2% versus 32.8%), delivering in an institution or health facility (7.2% versus 15.3%), and having a skilled birth attendant at delivery (10.8% versus 19.2%). Treated and control groups were similar in their distributions of neonatal mortality (4.0% in both groups), infant mortality (5.8% among treated versus 6.4% among control), and stillbirth (3.5% among treated versus 3.2% among control). Trends in maternal health services utilization, stillbirth, and neonatal and infant mortality before, during, and after MHVS implementation are shown in .

Trends in measures of maternal health services utilization, stillbirth, and neonatal and infant mortality among Bangladesh MHVS respondents, 2000–2016.

MHVS, Maternal Health Voucher Scheme.

After excluding upazilas outside the region of common support (Appendix B in ), weighted analyses included 338 upazilas (50 treated, 288 control) that were balanced in terms of measured upazila-level sociodemographic characteristics (the average standardized mean difference was reduced from 46.2% in the unweighted sample to 3.2% in the weighted sample, as shown in ). Weighted DD estimates from event study models are shown in , with corresponding estimates in . Each point represents the difference in the probability of the outcome comparing treated and control observations in the event period (e.g., 2 to 4 years after implementation) relative to the reference period immediately prior to implementation, along with robust 95% confidence intervals (CI).

Weighted event study estimates of the effect of gaining access to the MHVS in specified periods before and after the implementation of the program; BDHS data, 2000–2016.

Notes: Models included fixed effects for 6 divisions, fixed effects for year of birth (or the pregnancy outcome in analyses of stillbirth), and a vector of time-varying individual-level covariates, including household size, maternal age at marriage, women’s age at the time of the stillbirth or live birth, rural residence, women’s and husband’s educational attainment, household wealth (above or below median), whether the interval between the index pregnancy or birth outcome and a prior birth was short (<24 months) or not (24+ months and first births), and prior stillbirth (in analyses of stillbirth). Weights accounted for the following upazila-level characteristics described in : age structure, rural population, literacy, educational attainment, school attendance, poverty, employment, and household characteristics. Event periods included 4 leads (i.e., 6+, 4–6, 2–4, and 0–2 years prior to the introduction of the MHVS) and 4 lags (i.e., 0–2, 2–4, 4–6, and 6+ years after the introduction of the MHVS), with the 2-year period prior to implementation taken as the reference. Each coefficient represents the difference in the probability of the outcome (percentage point differences) comparing treated and control observations in the event period relative to the reference period; 95% confidence intervals account for the clustering of observations within upazilas. Sample sizes are reported in , and the estimation of weights is described in Appendix B in . BDHS, Bangladesh Demographic and Health Survey; MHVS, Maternal Health Voucher Scheme.

Preintervention trends were approximately parallel for all outcomes except delivery by cesarean section, which are shown in Appendix C in . In the postintervention period, there was evidence of increases in the use of maternal health services that materialized several years after the introduction of the MHVS, suggesting a lagged association. For example, after 6 years of access to the MHVS, the probabilities of delivering in a health facility and having a skilled birth attendant present increased by 6.5 (95% CI = −0.6, 13.6) and 5.8 (95% CI = −1.8, 13.3) percentage points, respectively, in treated areas compared to the control group. There was less evidence for a sustained increase in the receipt of antenatal care; 6 years after program implementation, the probability of reporting at least 3 antenatal care visits increased by 3.0 (95% CI = −4.8, 10.7) percentage points. We did not observe evidence consistent with the program having an impact on the probabilities of stillbirth, neonatal mortality, or infant mortality, with probabilities of stillbirth, neonatal mortality, and infant mortality decreasing by 0.7 (95% CI = −1.3, 2.6), 0.8 (95% CI = −1.7, 3.4), and 1.3 (95% CI = −2.5, 5.1) percentage points, respectively, after 6 years of access to the MHVS. Unweighted estimates were qualitatively similar (), as were weighted estimates with different buffers ( and ) and standard DD estimates (.

DD estimates comparing changes in the treated group before and after receiving access to the MHVS relative to corresponding trends in the control group; BDHS data, 2000–2016.

Notes: Models included fixed effects for 6 divisions, fixed effects for year of birth (or the pregnancy outcome in analyses of stillbirth), and a vector of time-varying individual-level covariates, including household size, maternal age at marriage, women’s age at the time of the stillbirth or live birth, rural residence, women’s and husband’s educational attainment, household wealth (above or below median), whether the interval between the index pregnancy or birth outcome and a prior birth was short (<24 months) or not (24+ months and first births), and prior stillbirth (in analyses of stillbirth). Weights accounted for the following upazila-level characteristics described in : age structure, rural population, literacy, educational attainment, school attendance, poverty, employment, and household characteristics; 95% confidence intervals account for the clustering of observations within upazilas. Sample sizes are reported in , and the estimation of weights is described in Appendix B in . BDHS, Bangladesh Demographic and Health Survey; DD, difference-in-difference; MHVS, Maternal Health Voucher Scheme.

Discussion

Bangladesh’s MHVS, introduced in 2007 and subsequently expanded to roughly 10% of the nation’s subdistricts, is one of many health care voucher programs that have proliferated across low- and middle-income countries (LMICs) over the past 15 years [30]. Our analyses support 2 main conclusions. First, the introduction of the MHVS was positively associated with selected maternal health services, particularly the probability of delivering in a health facility, which is consistent with the literature on the impact of voucher schemes in LMICs [33], including Bangladesh [20,23,24,27,28]. Second, despite having a longer period of follow-up than most extant evaluations [31,33], we did not observe attendant reductions in stillbirth, neonatal mortality, or infant mortality.

Our results cohere with the few empirical studies that have examined impacts of voucher programs on measures of utilization and health outcomes [33], with prior research indicating that access to the MHVS was associated with greater use of maternal health services, including antenatal care, having a skilled attendant at delivery, delivering in a health facility, receiving postnatal care, and more comprehensive maternal care [20,24,25,27]. One cross-sectional study comparing voucher recipients to nonrecipients suggested that recipients were more likely to seek medical assistance in case of obstetric complications [54]. However, as with the broader literature on maternal voucher programs, the evidence for impacts of voucher programs on perinatal, infant, and maternal health outcomes, including mortality, is limited [30,33].

One potential explanation for why robust increases in institutional delivery and modest increases in antenatal care and skilled delivery have not been clearly associated with improved health outcomes relates to the targeting and implementation of the MHVS. Although awareness of the MHVS was high in some studies [20], it was lower in others [21,29], which may have reduced demand and utilization of the program. National reports suggest that the program distributed vouchers to 72% of the estimated number of potential beneficiaries between 2011 and 2016 [21]. However, although eligibility criteria related to pregnancy and household income were not strictly enforced [20,21,23], by design, the program may have excluded women at greater risk of adverse pregnancy and birth outcomes, including those lacking access to family planning [55].

Second, increased utilization of maternal health services may be insufficient to bring about improvements in health indicators if the quality of care remains low [56]. At least initially, the program failed to induce nonpublic providers to enroll in the program [20,23]; this may have limited competition between providers, which is a key mechanism through which demand-side funding schemes are theorized to reduce costs and increase quality. Additionally, reports indicate that providers have not been adequately compensated for services rendered or increased workload, which may have reduced the quality of available services [12,20,21,23]. Among enrolled clinics, shortages of facilities, services (i.e., emergency obstetric care), staff, and specialists were recurring issues [21,23]. These challenges, while not unique to the MHVS in Bangladesh [57], may have limited the potential for the program to have its intended effects.

Third, there may be behavioral and sociocultural explanations that warrant further investigation. For example, recent work suggests that women who delivered in health facilities were less likely than those who delivered in their homes to initiate breastfeeding within 1 hour of delivery [58], which may counteract positive effects of increased maternal health services use on infant survival through other mechanisms.

Finally, given our sample size, we were unable to detect small associations with adequate precision. An observational study comparing deliveries in 8 health facilities from treated upazilas with 8 facilities in matched control upazilas found that the incidence of stillbirth was 0.9 percentage points lower in treated facilities and concluded that “the incidence of stillbirths was significantly lower in voucher than in control areas” [20]. We found a similar effect size 6 years postimplementation, although with less precision, since our empirical approach trades off some precision in order to limit confounding [59]. Secondary data analyses that leverage existing demographic data can be viewed as a cost-effective way of generating evidence that can be triangulated with existing research.

There were other limitations to our study. With respect to measurement, the random geographic displacement of EAs may have led to nondifferential misclassification of the treatment, as described in Appendix A in ; in our main analyses, we added a buffer around treated upazilas that would reduce the misclassification of treated upazilas as control units and mitigate potential spillover effects, and tested different buffers from 0 to 5 km in sensitivity analyses. Additionally, the locations of respondents’ EAs were measured at the time of interview, which does not account for migration between the time of each reported birth/pregnancy outcome and the time of interview. Measurement of covariates and outcomes was based on self-report, and there is likely error that is nondifferential by treatment group. For outcomes that are likely underreported, such as stillbirth, this error would tend to bias our estimates only if it were differential between our treatment and control groups. With respect to confounding bias, there is the potential for unmeasured confounding, as in any nonrandomized study. However, we included fixed effects for treatment group, year, and division and accounted for measured upazila-level characteristics using inverse probability weights to create more exchangeable treatment groups. Nonetheless, incorporating weights with event study models likely reduced the precision of our estimates and we cannot rule out the possibility of reverse confounding (e.g., if access was targeted to areas with lower levels of access to maternal health care services) or residual confounding by public, private, or nongovernmental services or initiatives that coincided with the rollout of the MHVS, differentially in treated and control upazilas. If, for example, the existing supply of community health workers in each area influenced the process by which the MHVS was phased-in across upazilas and magnified its impact on health care services use, then the effect of providing access to the program may have been overestimated. Finally, regarding the external validity of our estimates, we selected our analytical sample based on the eligibility criteria of the MHVS, with further restriction in our main analyses to the area of common support based on propensity score distributions. We did not apply DHS survey sampling weights and any inference beyond our sample should be done cautiously [60].

The aim of this quasi-experimental study was to provide a more comprehensive, longer-term, population-level analysis of the MHVS. Caveats considered, our results suggest that the MHVS was associated with increased utilization of maternal health services, particularly delivery in a health facility, which was incentivized by the program. However, we did not detect evidence of improvements in birth, neonatal, and infant health outcomes. Potential explanations include the suboptimal implementation of the program and inadequate supply-side investments to match increased demand and ensure access to quality health services. These challenges should be investigated and addressed [21] if the MHVS is to accelerate Bangladesh’s progress toward achieving its maternal and newborn health goals.

RECORD statement: Checklist of items, extended from the STROBE statement, that should be reported in observational studies using routinely collected health data.

(DOCX)

Click here for additional data file.

Prospective analysis plan.

(PDF)

Click here for additional data file.

Appendix A, Appendix B, and Appendix C.

(DOCX)

Click here for additional data file.

Sample sizes for primary study outcomes.

(DOCX)

Click here for additional data file.

Upazila-level characteristics from the World Bank’s Bangladesh Interactive Poverty Maps.

(DOCX)

Click here for additional data file.

Standardized mean differences comparing upazila-level sociodemographic factors in the unweighted and stabilized inverse probability of treatment weighted samples.

(DOCX)

Click here for additional data file.

Difference-in-differences estimates from events study models.

(DOCX)

Click here for additional data file.

Unweighted event study estimates of the effect of gaining access to the Maternal Health Voucher Scheme in specified periods before and after the implementation of the program; Bangladesh Demographic and Health Survey Data, 2000–2016.

(DOCX)

Click here for additional data file.

Weighted event study estimates of the effect of gaining access to the Maternal Health Voucher Scheme in specified periods before and after the implementation of the program; sensitivity analyses with 0-kilometer buffer around treated upazilas; Bangladesh Demographic and Health Survey Data, 2000–2016.

(DOCX)

Click here for additional data file.

Weighted event study estimates of the effect of gaining access to the Maternal Health Voucher Scheme in specified periods before and after the implementation of the program; sensitivity analyses with 5-kilometer buffer around treated upazilas; Bangladesh Demographic and Health Survey Data, 2000–2016.

(DOCX)

Click here for additional data file.

4 Jun 2021

Dear Dr Nandi,

Thank you for submitting your manuscript entitled "The impact of Bangladesh’s Maternal Health Voucher Scheme on health care services use, stillbirth, and neonatal and infant survival" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by Jun 08 2021 11:59PM.

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Louise Gaynor-Brook, MBBS PhD

Associate Editor

PLOS Medicine

4 Oct 2021

Dear Dr. Nandi,

Thank you very much for submitting your manuscript "The impact of Bangladesh’s Maternal Health Voucher Scheme on health care services use, stillbirth, and neonatal and infant survival" (PMEDICINE-D-21-02429R1) for consideration at PLOS Medicine.

Your paper was evaluated by three independent reviewers, including a statistical reviewer, and was discussed among all the editors here and with an academic editor with relevant expertise. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Oct 25 2021 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Louise Gaynor-Brook, MBBS PhD

Associate Editor, PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

General comments:

Please include line numbers in your revised manuscript; preferably not starting from 1 with each new page.

Please avoid inadvertently using causal language e.g. effect, effectiveness, impact, etc.

Throughout the paper, please adapt reference call-outs to the following style: "... under five and maternal mortality [1,3]." (noting the absence of spaces within the square brackets).

Title: Please revise your title according to PLOS Medicine's style. "Impact of" should be used only if causality can be inferred, i.e., from an RCT. Please place the study design in the subtitle (ie, after a colon). We suggest “Health care services use, stillbirth, and neonatal and infant survival following implementation of the Maternal Health Voucher Scheme in Bangladesh: A difference-in-differences analysis of Bangladesh Demographic and Health Survey data, 2000-2013” or similar

Abstract Background: The final sentence should clearly state the study question.

Abstract Methods and Findings:

Please provide brief demographic details of the study population

Please include the number of participants

Please report the main findings of your study, in keeping with the main outcome measures specified e.g. stillbirth, etc.

Please include the actual numbers / probabilities, in addition to the differences reported.

In the last sentence of the Abstract Methods and Findings section, please describe 2-3 of the main limitations of the study's methodology.

Abstract Conclusions:

Please begin your Abstract Conclusions with "In this study, we observed ..." or similar, to summarize the main findings from your study, without overstating your conclusions. Please emphasize what is new and address the implications of your study, being careful to avoid assertions of primacy.

Author Summary:

At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

In the final bullet point of ‘What Do These Findings Mean?’, please describe the main limitations of the study in non-technical language.

Introduction:

Please address causal language in paragraph beginning ‘In this study…’

Please remove subheadings within your Introduction

Please relocate the section beginning “The MHVS was implemented…” to precede the paragraph beginning “Sub-national, cross-sectional comparisons of voucher recipients vs. non-recipients, …”, followed finally by the paragraph beginning ‘In this study…” (which should the the final paragraph of your Introduction)

Methods:

Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section. If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript. If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and if/when reported analyses differed from those that were planned. Changes in the analysis-- including those made in response to peer review comments-- should be identified as such in the Methods section of the paper, with rationale. If a reported analysis was performed based on an interesting but unanticipated pattern in the data, please be clear that the analysis was data-driven.

Please provide more information regarding ethical approval of your study e.g. reference number

Please ensure that the study is reported according to the RECORD guideline, and include the completed RECORD checklist as Supporting Information. Please add the following statement, or similar, to the Methods: "This study is reported as per the REporting of studies Conducted using Observational Routinely-collected Data (RECORD) guideline (S1 Checklist)." The RECORD guideline can be found here: https://www.record-statement.org/ When completing the checklist, please use section and paragraph numbers, rather than page numbers which will likely no longer correspond to the appropriate sections after copy-editing.

Please note: Authors are responsible for keeping documentation of the nature and course of the personal communication with government officials.

Results:

Please provide a single table showing the baseline characteristics of the study population.

Please quantify more of the results in the main text of your Results section.

Please provide the actual numbers of events for the outcomes where possible, in addition to summary statistics.

For the probabilities reported in your results, please specify the comparison group.

Discussion:

Please present and organize the Discussion as follows: a short, clear summary of the article's findings; what the study adds to existing research and where and why the results may differ from previous research; strengths and limitations of the study; implications and next steps for research, clinical practice, and/or public policy; one-paragraph conclusion.

Please avoid the use of causal language

Figures:

Where weighted estimates are provided, please indicate which factors are adjusted for / taken into consideration, in the respective figure legend.

Tables:

Please define abbreviations used in the table legend of each table (including Supplementary files)

Where weighted estimates are provided, please indicate which factors are adjusted for / taken into consideration, in the respective figure legend.

Tables 2 & 3 - please separate columns ‘Total’ and ‘Control’. Please present numerators and denominators for percentages. Please clarify whether the results presented are indeed Mean and SD for characteristics such as Rural Residence, Mother’s Education, Birth Interval <24 months, etc, as % would make more sense in these (and other) instances.

References:

Please ensure that journal name abbreviations match those found in the National Center for Biotechnology Information (NCBI) databases, and are appropriately formatted and capitalised.

Please also see https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references for further details on reference formatting. For example, 6 authors should be cited prior to ‘et al’

Where website addresses are cited, please specify the date of access.

Supplementary files:

Please provide titles and legends for each individual table and figure in the Supporting Information.

Please see https://journals.plos.org/plosmedicine/s/supporting-information for our guidelines regarding file names.

Comments from the reviewers:

Reviewer #1: Alex McConnachie, Statistical Review

Nandi's paper looks at data from the Bangladesh Demographic and Health Surveys, to assess whether the introduction of the Maternal Health Voucher Scheme to some areas of the country had any impact on maternal health service use and outcomes. This review considers the use of statistics in the paper.

The analysis methods are generally very good. The IPTW approach is appropriate for this analysis, and the model used to assess the lagged impact of the MHVS seems good.

The main conclusion is that the MHVS led to increased use of health services, but not health outcomes. Whilst I agree that this appears to be the trend, it is notable that none of the effect estimates reported are statistically significant, so I would question the strength of the evidence at this stage. Should the wording be a little more cautious?

I feel these results are supportive of the continued rollout of the scheme, and of further analyses along similar lines in the future. This would be supported even more if there was data on the health economics of the scheme. Is that planned at some stage?

My only real question about the analysis comes from looking at Figure 2. I assume that we cannot read too much into these, as the MHVS is introduced at various times in different regions, and the curve for the "treated" line is a mixture of regions that have and have not yet received the scheme. However, the Institutional Delivery and Skilled Birth Attendant figures have a look of the intervention group lagging behind the control group, and catching up over time. This made me wonder whether the MHVS was delivered to regions partly on the basis of past/current health service utilization? If so, could some of the results simply represent regression to the mean? Were past levels of the outcomes of interest included as predictors in the propensity model used to derive the weights for the final analysis? Would this be an appropriate thing to do?

Tables 2 and 3 report the mean and standard deviation for various data items. This is unusual for categorical variables, as most of these are. I think most readers would feel more comfortable seeing these reported as frequencies and percentages.

Reviewer #2: This manuscript analyzes maternal service utilization and outcome data for delivery and newborn health across the implementation period of a national maternal health voucher system in Bangladesh. The authors have done an excellent job synthesizing the data and considered a range of potential scenarios in their sensitivity analysis and weighting approaches to counter potential gaps in the data. However, the most recent data included in this analysis is from 2014 while there was a DHS conducted in Bangladesh in 2018 - why were these data omitted from consideration as the MHVS appears to be an ongoing program? Also, while I appreciate that the data are limited to those from DHS surveys, there is mention of the MHVS program not being implemented with fidelity in all areas or by all community health volunteers in the description of the program. How were these potential irregularities considered in the analysis? How might individual CHW variation in program implementation potentially impact findings? Last, what other projects/programs were underway in Bangladesh through this period to either strengthen the supply side of the system or to further enhance patient demand? The analysis period is lengthy and in this time, there have been some notable shifts in care seeking and use in Bangladesh, principally for higher uptake of private services and a tremendous increase in Cesarean section rates. Could the authors please add these considerations to the Discussion in greater detail?

In addition to these considerations, there is a minor revisions needed: Please change the word "unawareness" to "lack of awareness" or similar on the bottom of page 3.

Reviewer #3: This is an important evaluation of a longstanding voucher scheme in Bangladesh to promote maternal health and care seeking as well as birth outcomes. Its findings are broadly commensurate with other past evaluations in that there is a small increase in coverage (and hence likely care seeking) among recipients but that there are no tangible benefits on mortality.

Methodologically the authors did what was feasible in terms of identifying appropriate matched recipients and non-recipients for comparison. One key point that needs further highlighting is if the sample size was sufficient for mortality estimation?

The key limitation in the analysis is the supply side, so acknowledged by the authors themsleves. It would really help understand the findings if there were concomitant reasonable findings from the health systems for maternal and newborn health. These could be health workforce related, such as skilled birth attendants, facilities for transportation, newborn care etc. The distribution and coverage by CHWs or BRAC workers woukd be an additional variable. These data are available from other country analyses, and would help understand the quality of care issues that likely limited the benefits of this voucher scheme.

Any attachments provided with reviews can be seen via the following link:

[LINK]

19 Nov 2021

Submitted filename: response to reviewers 20111117.docx

Click here for additional data file.

11 Jan 2022

Dear Dr. Nandi,

Thank you very much for submitting your manuscript "Health care services use, stillbirth, and neonatal and infant survival following implementation of the Maternal Health Voucher Scheme in Bangladesh: A difference-in-differences analysis of Bangladesh Demographic and Health Survey data, 2000-2013" (PMEDICINE-D-21-02429R2) for consideration at PLOS Medicine.

Your paper was re-reviewed by three reviewers, including the statistical reviewer, and discussed among all the editors here and with an academic editor with relevant expertise. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

I am afraid that we are not able to accept the manuscript for publication in the journal in its current form. In light of the comments of Reviewer 3, we will only be able to consider a revised version which includes analysis of data from the most recent 2018 Bangladesh Demographic and Health Survey. We plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Feb 01 2022 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Louise Gaynor-Brook, MBBS PhD

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

In light of the comments of Reviewer 3, please include analysis of data from the most recent 2018 Bangladesh Demographic and Health Survey in your revision, which will we require in order to reconsider your manuscript for publication.

Comments from the reviewers:

Reviewer #1: Alex McConnachie, Statistical Review

I thank the authors for their responses to my original points.

I am generally happy, though I still feel that the authors are overstating the results very slightly. The conclusion of the abstract states: "we found that the introduction of the MHVS was positively associated with the use of maternal health services". Whilst it looks likes this is probably the case, the fact remains that all three confidence intervals presented in support of this statement include zero. Ultimately, under the fitted models, the data are consistent with there being no associations. All I suggest is that this should be recognised more openly in the paper.

Reviewer #2: Thank you for these additional considerations and edits, the manuscript is now acceptable for publication.

Reviewer #3: This is a revised version of a extensively reviewed study of the potential impact of a maternal health voucher scheme (MHVS) in Bangladesh on coverage and utilization of maternal health services between 2006 (the year of the introdyuction of the scheme) and 2013. The statistical methods for assessing lagged differences using a difference in differences approach are well described, even though the sample size was likely limited for a robust assessment of impact on mortality.

The major limitation in the usefulness of this analysis for policy is lack of a longer term assesment of impact by inclusion of data from the recent most BDHS 2018 data and I found the arguments against including this latter period in the analysis, rather facile. inclusion of a longer time period of follow up could have answered key questions on program maturity and quality as well as secular trends. The policy relevance of data that are already almost a decade old is questionable.

Any attachments provided with reviews can be seen via the following link:

[LINK]

31 Mar 2022

Submitted filename: response to reviewers 20220331.docx

Click here for additional data file.

11 May 2022

Dear Dr. Nandi,

Thank you very much for re-submitting your manuscript "Health care services use, stillbirth, and neonatal and infant survival following implementation of the Maternal Health Voucher Scheme in Bangladesh: A difference-in-differences analysis of Bangladesh Demographic and Health Survey data, 2000-2016" (PMEDICINE-D-21-02429R3) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by two reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

Please note, when your manuscript is accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you've already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosmedicine@plos.org.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by May 13 2022 11:59PM.

Sincerely,

Louise Gaynor-Brook, MBBS PhD

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

Abstract Methods and Findings:

Line 69 - please revise to ‘After six years of access…’

Abstract Conclusions:

Please briefly address the specific implications of your study, substantiated by the results.

Introduction:

Line 167 - please clarify ‘ unlikely to reach socioeconomically disadvantaged groups absent efforts’

Methods:

Thank you for providing your prospective analysis plan and further details of changes to planned analyses in your cover letter. Please include these details of where reported analyses differed from those planned in the Methods section of the paper (including those made in response to peer review comments), with rationale.

Results:

Line 387 - please clarify ‘3.5% among treated vs. 3.2% among treated’

Figures:

Please confirm that the GADM map is compatible with CC BY licensing

Tables:

Table S4 - please define RD, UCL and LCL in the table legend.

References:

Your reference list appears to have been duplicated; please revise.

Comments from Reviewers:

Reviewer #1: Alex McConnachie, Statistical Review

I thank the authors once again for considering my comments, and I am happy with this latest version. I have no further comments to make.

Reviewer #3: This revised version is much improved and the inclusion of the BDHS 2017-18 data in my opinion strengthens the analysis significantly in terms of the key finding that there does NOT seem to be any impact of the Voucher Scheme on any of the key outcomes. Irrespective of any statistical nuancing, Figure 2 is very clear in terms of lack of any substantive difference between voucher scheme recipients and counterfactuals. The final version of the paper still talks about a notable lagged effect on improved facility births, which I frankly don't see. The authors may consider a more explicit recognition of this finding not being as robust as they make out.

If this analysis is published, as I hope it might given the importance of the findings, I would strongly recommend an accompanying commentary to discuss these counterintuitive findings and their implications for policy.

Any attachments provided with reviews can be seen via the following link:

[LINK]

13 May 2022

Submitted filename: response to reviewers 20220512.pdf

Click here for additional data file.

16 May 2022

Dear Dr Nandi,

On behalf of my colleagues and the Academic Editor, Prof. Zulfiqar Bhutta, I am pleased to inform you that we have agreed to publish your manuscript "Health care services use, stillbirth, and neonatal and infant survival following implementation of the Maternal Health Voucher Scheme in Bangladesh: A difference-in-differences analysis of Bangladesh Demographic and Health Survey data, 2000-2016" (PMEDICINE-D-21-02429R4) in PLOS Medicine.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper.

Sincerely,

Louise Gaynor-Brook, MBBS PhD

PLOS Medicine

Table 1

Bangladesh MHVS eligibility criteria and benefits.

Eligibility	Benefits
• Residence in program (treated) upazila	• Three antenatal care checkups
• Pregnancy is the first or second, with a family planning method used between pregnancies	• Delivery by skilled attendant (home or facility); beneficiaries received a Tk. 500 (USD 7.20)2 incentive for delivery in home and Tk. 2,000 (USD 28.80) incentive for delivery in a facility
• Household income Tk. 3,100 (USD 44.64) or less per month	• Management of complications including cesarean section by designated providers
• Functionally landless (<6,534 ft2 or approximately 0.15 acres of land)	• Postnatal care checkup within 6 weeks
• Household does not own income-earning assets (e.g., livestock, rickshaw)	• Transportation costs up to Tk. 500 (USD 7.20)
	• Gift box worth Tk. 500 (USD 7.20)

1Information from Khan and Khan (2016) [21].

2Based on a 2010 annual average exchange rate of 0.0144 USD for each Tk.

MHVS, Maternal Health Voucher Scheme; Tk., Bangladesh Taka; USD, US dollar.

Table 2

Sample characteristics for analyses of live births from the period prior to the introduction of the MHVS in treated upazilas (before December 2006); BDHS data.

	Total			Control			Treated
	N	Mean	SD	n	Mean	SD	n	Mean	SD
Household size	11,315	6.5	3.2	9,698	6.5	3.2	1617	6.5	3.1
Maternal age at marriage	11,315	15.8	2.8	9,698	15.9	2.8	1617	15.7	2.6
Respondent’s age at event1	11,315	23.8	6.1	9,698	23.8	6.1	1617	23.9	6.2
	N	Frequency	%	N	Frequency	%	N	Frequency	%
Rural residence	11,315	7,746	68.5	9,698	6,410	66.1	1,617	1,336	82.6
Mother’s education
No schooling	11,315	3,649	32.2	9,698	3,061	31.6	1,617	588	36.4
Primary	11,315	3,577	31.6	9,698	3,034	31.3	1,617	543	33.6
Secondary or more	11,315	4,087	36.1	9,698	3,601	37.1	1,617	486	30.1
Father’s education
No schooling	11,315	4,131	36.5	9,698	3,470	35.8	1,617	661	40.9
Primary	11,315	3,159	27.9	9,698	2,665	27.5	1,617	494	30.6
Secondary or more	11,315	4,016	35.5	9,698	3,555	36.7	1,617	461	28.5
Household wealth >median	11,315	5,351	47.3	9,698	4,797	49.5	1,617	554	34.3
Birth interval <24 months2	11,315	1,278	11.3	9,698	1,082	11.2	1,617	196	12.1
3+ antenatal care visits	8,978	2,809	31.3	7,714	2,528	32.8	1,264	281	22.2
Institutional delivery	11,283	1,596	14.1	9,671	1,480	15.3	1,612	116	7.2
Skilled birth attendant	11,306	2,032	18.0	9,690	1,857	19.2	1,616	175	10.8
Neonatal mortality	11,248	448	4.0	9,639	384	4.0	1,609	64	4.0
Infant mortality	9,481	595	6.3	8,125	517	6.4	1,356	78	5.8

1Women’s age at the time of the live birth, measured continuously.

2Binary indicator for whether the interval between the index birth and a prior birth was short (<24 months) or not (24+ months and first births).

BDHS, Bangladesh Demographic and Health Survey; MHVS, Maternal Health Voucher Scheme; SD, standard deviation.

Table 3

Sample characteristics for analyses of pregnancies lasting at least 7 months from the period prior to the introduction of the MHVS in treated upazilas (before December 2006); BDHS data.

	Total			Control			Treated
	N	Mean	SD	n	Mean	SD	n	Mean	SD
Household size	11,977	6.5	3.2	10,264	6.5	3.2	1,713	6.4	3.1
Maternal age at marriage	11,977	15.8	2.8	10,264	15.9	2.8	1,713	15.7	2.6
Respondent’s age at event1	11,977	23.7	6.1	10,264	23.7	6.1	1,713	23.8	6.1
	N	Frequency	%	N	Frequency	%	N	Frequency	%
Rural residence	11,977	8,192	68.4	10,264	6,774	66.0	1,713	1,418	82.8
Mother’s education
No schooling	11,977	3,815	31.9	10,264	3,190	31.1	1,713	625	36.5
Primary	11,977	3,778	31.5	10,264	3,209	31.3	1,713	569	33.2
Secondary or more	11,977	4,382	36.6	10,264	3,863	37.6	1,713	519	30.3
Father’s education
No schooling	11,977	4,363	36.4	10,264	3,658	35.6	1,713	705	41.2
Primary	11,977	3,330	27.8	10,264	2,812	27.4	1,713	518	30.2
Secondary or more	11,977	4,273	35.7	10,264	3,784	36.9	1,713	489	28.5
Household wealth >median	11,977	5,916	49.4	10,264	5,283	51.5	1,713	633	37.0
Birth interval <24 months2	11,977	1,334	11.1	10,264	1,133	11.0	1,713	201	11.7
Prior stillbirth	11,977	250	2.1	10,264	220	2.1	1,713	30	1.8
Stillbirth	11,977	386	3.2	10,264	326	3.2	1,713	60	3.5

1Women’s age at the time of the stillbirth or live birth, measured continuously.

2Binary indicator for whether the interval between the index pregnancy outcome and a prior birth was short (<24 months) or not (24+ months and first births).

BDHS, Bangladesh Demographic and Health Survey; MHVS, Maternal Health Voucher Scheme; SD, standard deviation.