- Split View
-
Views
-
Cite
Cite
Prabhat Lamichhane, Anurag Sharma, Ajay Mahal, Impact evaluation of free delivery care on maternal health service utilisation and neonatal health in Nepal, Health Policy and Planning, Volume 32, Issue 10, December 2017, Pages 1427–1436, https://doi.org/10.1093/heapol/czx124
- Share Icon Share
Abstract
Nepal introduced free delivery services for births in public facilities in 2005 in 25 districts with the intervention initially restricted to women with less than two living children and/or women with obstetric complications. After November 2007, eligibility conditions were relaxed to include all women, and the programme was later expanded to cover an additional 50 districts in December 2008. We exploit the phased expansion of the free birth delivery programme to identify its impact on place of delivery, the presence of skilled birth attendants (SBAs) and neonatal mortality using difference-in-difference methods, on data for 4457 live-births reported between 2001 and 2008 from Nepal Demographic and Health Surveys for 2006 and 2011. Programme impacts were estimated for: (1) initial implementation until the relaxation of eligibility criteria to include all women in November 2007 (early phase); and (2) initial implementation until the programme was expanded nationwide in December 2008 (longer phase). Early implementing districts were treatment districts, while late implementing hill districts were control districts. In the early phase, the likelihood of delivery by SBAs was 5.6 percentage points higher (95%CI 0.002, 0.111) and the likelihood of delivery in a public facility was 5.1 percentage points higher (95%CI −0.003, 0.106) in treatment districts compared with control districts. The programme lowered the likelihood of neonatal mortality by 4.0 (−0.072, −0.009) percentage points for women with less than two living children and by 6.9 percentage points (95%CI −0.104, −0.035) for women from lower castes and indigenous groups in treatment districts compared with women in control districts, during the early phase. Programme effects on use of public facilities for births and deliveries attended by SBAs were not sustained over a longer exposure period. The results on neonatal mortality persisted with longer programme exposure, although the effects were smaller in magnitude.
We evaluated the impact of fee-removal in Nepal using demographic health surveys.
Free delivery care increased the likelihood of using public sector maternity services in early phase.
Free delivery care lowered the likelihood of neonatal mortality.
Programme benefits flowed to women among lower caste or indigenous ethnicity.
Introduction
User charges in public facilities were initially introduced to lessen moral hazard and improve quality of care in low-income countries in 1980s (Gertler and Hammer 1997; Dzakpasu et al. 2013). Subsequent analyses found that user-fees resulted in low-income households incurring large out of pocket health expenses and lowered their health service use, leading to the abolition and/or lowering of such fees in several countries (World Health Organization 2010). Later reviews noted that removal of user-fees increased health service use including maternal and child health services (Lagarde and Palmer 2008; Hatt et al. 2013). However, these reviews also found that many studies of the impact of user-fees removal on healthcare use, maternal and child health services in particular, had weak study designs, relying on before–after assessments that did not account for underlying temporal trends, or used single period cross-sectional data that could not adequately address difficulties introduced by potential heterogeneity across programme areas and temporal trends (Lagarde and Palmer 2008; Hatt et al. 2013).
Partly as a response to methodological shortcomings in previous work, there has been a recent upsurge of empirical analyses on the impact of user-fees. Recent studies have used difference-in-difference (DID) methods to assess the impact of user-fee removal on maternal and child health (McKinnon et al. 2014; Tanaka 2014). Tanaka (2014) found that the removal of user-fees for pregnant women and under-6 children at public facilities improved the nutritional status of children in South Africa. McKinnon et al. (2014) analysed Demographic and Health Survey data from 10 sub-Saharan African countries and found that user-fee removal led to a 9% reduction in neonatal mortality rates. The authors also found an increase in facility-based deliveries, but it is unclear whether their study adequately controlled for confounding due to other programmes simultaneously ongoing in these countries. In contrast, in Gujarat, India, a DID analysis of data from the Chiranjeevi programme (which covers the cost of a facility-based delivery in the private sector) found no change in facility-based births and birth-related complications (Mohanan et al. 2014).
This article estimates the impact of free provision of birth delivery services in the public sector health facilities in Nepal. Free provision was implemented as part of the Safe Delivery Incentive Programme (SDIP). The government of Nepal announced free delivery care in public health facilities, along with cash transfers targeted to specific populations under the SDIP in July 2005 (Family Health Division 2005). Two components of the programme, no charges for delivery services at public facilities and direct reimbursement by the Ministry of Health to public health facilities for the estimated cost of service provision (US$9.3, Nepali Rupees 1000), were initially announced for 25 low human development index (HDI) mountain and hill districts (Figure 1). Eligible users were exempt from user-fees, including registration charges, and fees for diagnostics and medicines in public facilities. Eligibility was initially limited to women with no more than two children and/or those visiting a health facility with an obstetric complication. The eligibility condition was relaxed to include all women in November 2007; and free delivery was further expanded to include all 75 districts [including the 50 hill and terai (lowland) districts not included initially] in December 2008.
In contrast, other components of SDIP (cash transfers to women and incentives for health workers) were launched nationwide from the start in July 2005 (all 75 districts). The cash transfer paid to women varied by ecological region—equivalent to US$13.9 (Nepali Rupees 1500) in the mountain districts, US$9.3 (Nepali Rupees 1000) in hill districts and US$4.6 (Nepali Rupees 500) in terai districts (southern lowland) to reflect differences in transportation expenses in these areas.
The SDIP was renamed as Aama with nationwide implementation of the free delivery care component in January 2009.
We evaluate the impact of free delivery care on use of public facilities for births and deliveries by skilled birth attendants (SBAs), and neonatal mortality, using data on live-births reported between 2001 and 2008 in Nepal Demographic Health Survey (NDHS) survey rounds of 2006 and 2011. Two periods of programme exposure are evaluated: (1) from initial implementation until the relaxation of the eligibility criterion to include all women in November 2007; and (2) from initial implementation until the programme was expanded from 25 districts to 75 districts (nationwide) in December 2008 (Figure 1). Similar to recent literature evaluating free services using DID, we use DID methods to assess the programme impacts, including for specific sub-groups—all women, women characterized by different eligibility criteria, caste and ethnicity. We isolate the impacts of free delivery care (user-fees removal) by taking advantage of the phased introduction of the free-delivery component even as other programme components were introduced simultaneously in all districts. In other words at baseline, neither treatment nor control districts had access to cash transfers. However, with the implementation of the SDIP, treatment districts had free birth delivery care and cash transfers, whereas the control districts had only cash transfers. The goal of cash transfers was to compensate for travel expenses to facilities, so that effectively the difference between treatment and control districts was free delivery at public facilities.
We also directly controlled for programmes other than SDIP that might have impacted outcomes. This study extends the work of Powell-Jackson and Hanson (2012) which evaluated the impact of SDIP as a whole (free delivery care as well as cash transfer) on six districts using propensity score matching. The authors found that women with knowledge of SDIP were more likely to deliver in a public health facility, deliver with the assistance of a SBA, have a caesarean-section and also switch from non-governmental facilities to public hospitals (Powell-Jackson and Hanson 2012). In contrast to Powell-Jackson and Hanson (2012), our analysis is based on data from the entire country and not just these seven districts (out of 75 districts nationally) and focuses on one specific component of SDIP, namely free delivery services. In addition, we assess programme effects on neonatal mortality, and for different programme periods and sub-groups.
Methodology
Data and variables
We used NDHS data to assess impacts of the free delivery care component of SDIP on health service use and neonatal mortality (Ministry of Health and Population et al. 2007, 2012). Data from NDHS 2006 and NDHS 2011 were used for our DID analysis, with NDHS 2001 data being utilized to assess pre-intervention trends in the treatment and control districts.
NDHS, a two-stage, stratified, nationally representative survey, collects information on the use of maternal and child health services, child mortality and a range of other indicators from households containing women 15–49 years old. The survey data also include information on live-births among women who were 15–49 years of age in the preceding 5-years: NDHS 2001 (6931 live-births), NDHS 2006 (5783 live-births) and NDHS 2011 (5306 live-births). Live-births reported in NDHS 2011 that occurred after the programme was expanded to cover all districts (2282 births) were excluded from our analysis.
Outcome variables
We consider four different outcome variables: (1) delivery in any facility, (2) delivery in any government facility, (3) delivery assisted by an SBA and (4) neonatal death. For each live-birth the surveyed women were asked “Where did you give birth (Name).” The variable “delivery in any facility” was taken as 1 if the birth place was reported to be a government facility or a non-government facility (private or not-for-profit). The variable “delivery in government facility” was defined as 1 if the birth place was a government facility, 0 otherwise. The variable “delivery assisted by SBA” was defined as 1 if a doctor, nurse or midwife was reported to have assisted birth, 0 otherwise. The variable “neonatal death” was equal to 1 if the woman reported that the child died within 0–30 days of birth, 0 otherwise. We included deaths up to 30 days because previous literature suggests age heaping (rounding of at 1 month for children who died on 28–29 days old) in fertility and birth history due to recall error (Beckett et al. 2001).
Treatment variables
A treatment variable was created to ascertain programme implementation in the district of residence. The variable took the value of 1 if district had free delivery care programme in place and 0 otherwise. Two indicator variables were created to capture the timing of births, one for the births in the intervention period and another for the pre-intervention period.
Explanatory variables
We accounted for other programmes in place at the time of the intervention: improvement in the provision of emergency obstetric care (EOC), the presence of comprehensive emergency obstetric care (CEOC) facilities and a Female Community Health Volunteer (FCHVs) programme. At the district level, a dummy variable was created which took the value 1 if the EOC strengthening programme had been implemented. The programme initially started in four districts in late 1990s through the Nepal Safer Motherhood Project and has been gradually expanded (Barker et al. 2007; Department of Health Services 2009). The EOC was intended to improve physical facilities, availability of equipment and capacity development of hospital staff to handle obstetric emergencies and for monitoring of EOC programme in public facilities. Similarly, a dummy variable took the value of 1 if the district had a CEOC facility available in referral hospitals, zonal hospitals, medical colleges and referral mission hospitals (not-for-profit hospitals), 0 otherwise.
An indicator variable was also generated to capture the activity level of FCHVs at the district level, as indicated by activity reports received from FCHVs by the district health office. The FCHVs are required to submit activity reports that include information on iron tablet distribution, organization of mothers’ group meetings, confirming breast feeding within 1 h, and neonatal and post-natal visits to health facilities (Department of Health Services 2006). The ratio of reports received from FCHVs to the total required reports was estimated from the annual report of department of health services (Department of Health Services 2006, 2007, 2008, 2009). The FCHV activity indicator was equal to 1 if the FCHV reporting was >80% for the district.
In addition, an SBA policy was introduced in 2006 to train SBAs to provide obstetric first aid and to identify, manage and refer complications and manage normal pregnancies. Under this policy, 406 personnel were provided training from different districts from Fiscal Year 2004/05 to Fiscal Year 2007/08 (Department of Health Services 2006, 2007, 2008). Although it would have been helpful to have information on the districts where the trained SBAs were ultimately deployed, the number trained was a small fraction of the overall SBAs needed (about 8000 nationwide).
We also included dummy variables indicating the ecological region (hill, terai and mountain) as the cash amount given to women varied ecologically. We also used two variables for district level characteristics—HDI and road density. The HDI for each district for the year 2001 was included in our analysis (UNDP 2004). We used road density data for each district for two points in time, before (2006) and after the programme (2011) (Department of Roads 2015). We controlled for year-specific effects using “year dummy” for year of birth. We also controlled for the length of the recall period (months between birth and interview) to control for the possible recall bias.
Study districts
Nepal has 16 mountain districts, 39 hill districts and 20 terai (southern plains) districts (Supplementary Appendix S3). The free delivery component of SDIP was initially implemented (phase I) in 25 districts—16 mountain and 9 hill districts (Family Health Division 2005). Low HDI, low health service utilization and poor health status in the region were the reasons for implementation of free delivery care in the 25 districts (UNDP 2004; Family Health Division 2005). An equal proportion (18%) of treatment and hill control districts had an EOC programme in place during the study period. About half of the districts in hill-terai control districts had CEOC facilities, i.e. C-section or blood transfusion facility in the study period while none of the treatment districts had CEOC facilities.
Empirical strategy
We used the phased expansion of the free delivery component of SDIP to undertake a DID analysis. Early implementation districts (hills and mountain districts) were treatment districts, while late implementation (hill) districts were used as controls. The timing of live-births as reported by mothers, the district of residence, the number of living children and the number of births was used to determine eligibility for programme benefits. We used two eligibility criteria:
Eligible criteria I: Women who had two or fewer living children
Eligible criteria II: Women who had two or fewer living children and/or women who had experienced five or more births. Since women with five or more births are at greater risk of complications, this modification was expected to capture the complicated delivery cases.
Sub-group analysis was also performed, by higher caste (brahmin or chhetri) and lower or indigenous caste (dalit or janajati) women. Brahmin (priest) are on top of the social hierarchy closely followed by chhetri (warriors) while dalits are the bottom placed. Dalits are considered “impure” or “untouchables” and janajati are the indigenous groups. Compared with brahmin–chhetris, dalits and janajatis have poor health care seeking behaviour, educational status and socioeconomic status (Bennett et al. 2006; Bennett et al. 2008).
Two programme periods were evaluated for all live-births occurring after the introduction of free delivery in treatment districts for two periods of programme exposure: (1) from implementation until the eligibility criteria was changed to all women in November 2007 (February 2006–November 2007), hereafter referred to as early phase; and (2) from implementation until the programme was expanded nationwide in January 2009 (February 2006–December 2008), hereafter called the “longer phase”. Since the cash transfer component of SDIP was implemented nationwide we could not isolate the impacts of the cash transfers in our study.
Here, denotes the binary outcome(s) related to live-births corresponding to individual I, in district D, at time T. D is an indicator variable for treatment (1 = if treatment district, 0 otherwise), T is a time dummy which takes the value 1 if birth was after the intervention was introduced and 0 if occurred pre-intervention. is the vector for different individual and household characteristics and is the error term. In probit models, the interaction coefficient in (1) is not directly interpretable as a DID estimate. The impact estimates are instead calculated as the differences in predicted probabilities of outcomes for treatment districts compared with control districts over time (Karaca-Mandic et al. 2012; Puhani 2012). Sampling weights were used and standard errors were clustered at the district level.
Programme period of analysis
As the official launch of SDIP did not translate into immediate on-ground implementation, we estimated Equation (1) after excluding information related to live-births for the first 7 months of the programme (i.e. we excluded live-births from July 2005 to January 2006 from our analysis). Fund transfers from the donor agency (Department for International Development (UK)) created substantial delays in allocation of programme funds in the first year (Powell-Jackson et al. 2008). Moreover, fund disbursements from the central government to local facilities involves two steps, which delayed implementation: from programme announcement in July 2005 to actual budgetary allocations to district funds or hospitals (a period of 4–6 months) and subsequent allocations from districts to lower-level facilities, a period of 0–1 months.
Robustness checks included estimating Equation (1) after excluding live-births in the Kathmandu valley districts (capital city with largest tertiary hospitals) and using different sets of controls (hills and terai districts). This was done because tertiary hospitals in Kathmandu did not implement SDIP for the first 2 years of the programme, and tertiary hospitals that account for a large number of deliveries are located in Kathmandu valley. We re-estimated the results without dropping the live-births for the period from July 2005 to January 2006 (the period where on-ground implementation was limited). We also re-estimated our results using DID analysis on a matched dataset of births, constructed using coarsened exact matching (Iacus et al. 2011; G King et al. 2011). Beginning with a pooled dataset of births in NDHS surveys from the years 2001, 2006 and 2011, we identified women in 2001 who delivered in a health facility. To this group, we then matched women in years 2006 and 2008 by age, educational status, ethnicity, household size and region of residence. The set of matched women/births from 2006 and 2008 constituted the matched dataset.
Pre-intervention trends
Results
Descriptive results
Table 1 shows descriptive statistics for maternal health service utilisation and neonatal death prior to programme implementation before (February 2001–June 2005) and after the programme was introduced in its early phase and longer phase. In the early phase, deliveries in any facility increased from 66 per 1000 live-births to 112 per 1000 live-births, deliveries in government facilities increased from 48 per 1000 live-births to 89 per 1000 live-births and deliveries by SBAs increased from 73 per 1000 live-births to 127 per 1000 live-births in treatment districts. In control districts, delivery in any facility and delivery by SBAs also increased, from 238 per 1000 live-births to 259 per 1000 live-births and 261 per 1000 live-births to 265 per 1000 live-births, respectively. However, deliveries in government facilities decreased from 205 per 1000 live-births to 198 per 1000 live-births. Neonatal deaths declined from 49 per 1000 live-births to 25 per 1000 live-births in treatment districts while increasing from 16 per 1000 live-births to 35 per 1000 live-births in control districts during the early phase of the programme. During the longer phase as well, maternal health service utilization continued to increase in treatment and control districts. Neonatal deaths declined from 49 per 1000 live-births to 35 per 1000 live-births in treatment districts while increasing from 16 per 1000 live-births to 39 per 1000 live-births in control districts.
. | Treatment . | Hill control . | ||||||
---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |
Early phase | ||||||||
Delivery in any facility | 66(7) | 1325 | 112(12) | 641 | 238(12) | 1231 | 259(19) | 555 |
Delivery in government facility | 48(6) | 1325 | 89(11) | 641 | 205(12) | 1231 | 198(17) | 555 |
Delivery by SBA | 73(7) | 1325 | 127(13) | 641 | 261(13) | 1231 | 265(19) | 555 |
Neonatal death | 49(6) | 1325 | 25(6) | 641 | 16(4) | 1231 | 35(8) | 555 |
Longer phase | ||||||||
Delivery in any facility | 66(7) | 1325 | 135(11) | 1021 | 238(12) | 1231 | 288(15) | 880 |
Delivery in government facility | 48(6) | 1325 | 111(10) | 1021 | 205(12) | 1231 | 222(14) | 880 |
Delivery by SBA | 73(7) | 1325 | 146(11) | 1021 | 261(13) | 1231 | 290(15) | 880 |
Neonatal death | 49(6) | 1325 | 35(6) | 1021 | 16(4) | 1231 | 39(7) | 880 |
. | Treatment . | Hill control . | ||||||
---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |
Early phase | ||||||||
Delivery in any facility | 66(7) | 1325 | 112(12) | 641 | 238(12) | 1231 | 259(19) | 555 |
Delivery in government facility | 48(6) | 1325 | 89(11) | 641 | 205(12) | 1231 | 198(17) | 555 |
Delivery by SBA | 73(7) | 1325 | 127(13) | 641 | 261(13) | 1231 | 265(19) | 555 |
Neonatal death | 49(6) | 1325 | 25(6) | 641 | 16(4) | 1231 | 35(8) | 555 |
Longer phase | ||||||||
Delivery in any facility | 66(7) | 1325 | 135(11) | 1021 | 238(12) | 1231 | 288(15) | 880 |
Delivery in government facility | 48(6) | 1325 | 111(10) | 1021 | 205(12) | 1231 | 222(14) | 880 |
Delivery by SBA | 73(7) | 1325 | 146(11) | 1021 | 261(13) | 1231 | 290(15) | 880 |
Neonatal death | 49(6) | 1325 | 35(6) | 1021 | 16(4) | 1231 | 39(7) | 880 |
Note: (1) Standard errors in parentheses. (2) Estimations are sample weighted.
. | Treatment . | Hill control . | ||||||
---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |
Early phase | ||||||||
Delivery in any facility | 66(7) | 1325 | 112(12) | 641 | 238(12) | 1231 | 259(19) | 555 |
Delivery in government facility | 48(6) | 1325 | 89(11) | 641 | 205(12) | 1231 | 198(17) | 555 |
Delivery by SBA | 73(7) | 1325 | 127(13) | 641 | 261(13) | 1231 | 265(19) | 555 |
Neonatal death | 49(6) | 1325 | 25(6) | 641 | 16(4) | 1231 | 35(8) | 555 |
Longer phase | ||||||||
Delivery in any facility | 66(7) | 1325 | 135(11) | 1021 | 238(12) | 1231 | 288(15) | 880 |
Delivery in government facility | 48(6) | 1325 | 111(10) | 1021 | 205(12) | 1231 | 222(14) | 880 |
Delivery by SBA | 73(7) | 1325 | 146(11) | 1021 | 261(13) | 1231 | 290(15) | 880 |
Neonatal death | 49(6) | 1325 | 35(6) | 1021 | 16(4) | 1231 | 39(7) | 880 |
. | Treatment . | Hill control . | ||||||
---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |
Early phase | ||||||||
Delivery in any facility | 66(7) | 1325 | 112(12) | 641 | 238(12) | 1231 | 259(19) | 555 |
Delivery in government facility | 48(6) | 1325 | 89(11) | 641 | 205(12) | 1231 | 198(17) | 555 |
Delivery by SBA | 73(7) | 1325 | 127(13) | 641 | 261(13) | 1231 | 265(19) | 555 |
Neonatal death | 49(6) | 1325 | 25(6) | 641 | 16(4) | 1231 | 35(8) | 555 |
Longer phase | ||||||||
Delivery in any facility | 66(7) | 1325 | 135(11) | 1021 | 238(12) | 1231 | 288(15) | 880 |
Delivery in government facility | 48(6) | 1325 | 111(10) | 1021 | 205(12) | 1231 | 222(14) | 880 |
Delivery by SBA | 73(7) | 1325 | 146(11) | 1021 | 261(13) | 1231 | 290(15) | 880 |
Neonatal death | 49(6) | 1325 | 35(6) | 1021 | 16(4) | 1231 | 39(7) | 880 |
Note: (1) Standard errors in parentheses. (2) Estimations are sample weighted.
Table 2 presents descriptive statistics for different sub-groups of women before and after the programme was introduced in its early phase. In all groups, maternal health service utilization increased in treatment as well as control districts following introduction of free delivery. For instance, delivery in any facility increased from 121 per 1000 live-births to 188 per 1000 live-births and delivery by SBAs increased from 128 per 1000 live-births to 213 per 1000 live-births among women in eligibility criteria-I in treatment districts. For all the sub-groups, neonatal death rates declined in treatment districts and increased in control districts in the early phase.
. | Treatment . | Hill control . | ||||||
---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |
Delivery in any facility | ||||||||
Eligibility criteria-I | 121(13) | 590 | 188(24) | 273 | 350(18) | 706 | 386(28) | 306 |
Eligibility criteria-II | 83(9) | 917 | 133(16) | 437 | 280(15) | 935 | 303(23) | 399 |
Lower caste or indigenous ethnicity | 60(11) | 496 | 112(20) | 262 | 154(14) | 660 | 164(21) | 308 |
Higher caste | 66(9) | 767 | 106(16) | 372 | 301(21) | 462 | 334(33) | 209 |
Delivery in government facility | ||||||||
Eligibility criteria-I | 90(12) | 590 | 147(21) | 273 | 300(17) | 706 | 281(26) | 306 |
Eligibility criteria-II | 60(8) | 917 | 105(15) | 437 | 242(14) | 935 | 224(21) | 399 |
Lower caste or indigenous ethnicity | 50(10) | 496 | 94(18) | 262 | 136(13) | 660 | 130(19) | 308 |
Higher caste | 44(7) | 767 | 80(14) | 372 | 243(20) | 462 | 251(30) | 209 |
Delivery by SBA | ||||||||
Eligibility criteria-I | 128(14) | 590 | 213(25) | 273 | 378(18) | 706 | 389(28) | 306 |
Eligibility criteria-II | 89(9) | 917 | 148(17) | 437 | 304(15) | 935 | 306(23) | 399 |
Lower caste or indigenous ethnicity | 66(11) | 496 | 114(20) | 262 | 170(15) | 660 | 171(21) | 308 |
Higher caste | 73(9) | 767 | 128(17) | 372 | 329(22) | 462 | 332(33) | 209 |
Neonatal death | ||||||||
Eligibility criteria-I | 55(9) | 590 | 37(11) | 273 | 20(5) | 706 | 49(12) | 306 |
Eligibility criteria-II | 60(8) | 917 | 29(8) | 437 | 19(4) | 935 | 45(10) | 399 |
Lower caste or indigenous ethnicity | 60(11) | 496 | 6(5) | 262 | 22(43) | 660 | 43(12) | 308 |
Higher caste | 44(7) | 767 | 37(10) | 372 | 12(5) | 462 | 19(9) | 209 |
. | Treatment . | Hill control . | ||||||
---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |
Delivery in any facility | ||||||||
Eligibility criteria-I | 121(13) | 590 | 188(24) | 273 | 350(18) | 706 | 386(28) | 306 |
Eligibility criteria-II | 83(9) | 917 | 133(16) | 437 | 280(15) | 935 | 303(23) | 399 |
Lower caste or indigenous ethnicity | 60(11) | 496 | 112(20) | 262 | 154(14) | 660 | 164(21) | 308 |
Higher caste | 66(9) | 767 | 106(16) | 372 | 301(21) | 462 | 334(33) | 209 |
Delivery in government facility | ||||||||
Eligibility criteria-I | 90(12) | 590 | 147(21) | 273 | 300(17) | 706 | 281(26) | 306 |
Eligibility criteria-II | 60(8) | 917 | 105(15) | 437 | 242(14) | 935 | 224(21) | 399 |
Lower caste or indigenous ethnicity | 50(10) | 496 | 94(18) | 262 | 136(13) | 660 | 130(19) | 308 |
Higher caste | 44(7) | 767 | 80(14) | 372 | 243(20) | 462 | 251(30) | 209 |
Delivery by SBA | ||||||||
Eligibility criteria-I | 128(14) | 590 | 213(25) | 273 | 378(18) | 706 | 389(28) | 306 |
Eligibility criteria-II | 89(9) | 917 | 148(17) | 437 | 304(15) | 935 | 306(23) | 399 |
Lower caste or indigenous ethnicity | 66(11) | 496 | 114(20) | 262 | 170(15) | 660 | 171(21) | 308 |
Higher caste | 73(9) | 767 | 128(17) | 372 | 329(22) | 462 | 332(33) | 209 |
Neonatal death | ||||||||
Eligibility criteria-I | 55(9) | 590 | 37(11) | 273 | 20(5) | 706 | 49(12) | 306 |
Eligibility criteria-II | 60(8) | 917 | 29(8) | 437 | 19(4) | 935 | 45(10) | 399 |
Lower caste or indigenous ethnicity | 60(11) | 496 | 6(5) | 262 | 22(43) | 660 | 43(12) | 308 |
Higher caste | 44(7) | 767 | 37(10) | 372 | 12(5) | 462 | 19(9) | 209 |
Note: (1) Standard errors in parentheses. (2) Estimations are sample weighted.
. | Treatment . | Hill control . | ||||||
---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |
Delivery in any facility | ||||||||
Eligibility criteria-I | 121(13) | 590 | 188(24) | 273 | 350(18) | 706 | 386(28) | 306 |
Eligibility criteria-II | 83(9) | 917 | 133(16) | 437 | 280(15) | 935 | 303(23) | 399 |
Lower caste or indigenous ethnicity | 60(11) | 496 | 112(20) | 262 | 154(14) | 660 | 164(21) | 308 |
Higher caste | 66(9) | 767 | 106(16) | 372 | 301(21) | 462 | 334(33) | 209 |
Delivery in government facility | ||||||||
Eligibility criteria-I | 90(12) | 590 | 147(21) | 273 | 300(17) | 706 | 281(26) | 306 |
Eligibility criteria-II | 60(8) | 917 | 105(15) | 437 | 242(14) | 935 | 224(21) | 399 |
Lower caste or indigenous ethnicity | 50(10) | 496 | 94(18) | 262 | 136(13) | 660 | 130(19) | 308 |
Higher caste | 44(7) | 767 | 80(14) | 372 | 243(20) | 462 | 251(30) | 209 |
Delivery by SBA | ||||||||
Eligibility criteria-I | 128(14) | 590 | 213(25) | 273 | 378(18) | 706 | 389(28) | 306 |
Eligibility criteria-II | 89(9) | 917 | 148(17) | 437 | 304(15) | 935 | 306(23) | 399 |
Lower caste or indigenous ethnicity | 66(11) | 496 | 114(20) | 262 | 170(15) | 660 | 171(21) | 308 |
Higher caste | 73(9) | 767 | 128(17) | 372 | 329(22) | 462 | 332(33) | 209 |
Neonatal death | ||||||||
Eligibility criteria-I | 55(9) | 590 | 37(11) | 273 | 20(5) | 706 | 49(12) | 306 |
Eligibility criteria-II | 60(8) | 917 | 29(8) | 437 | 19(4) | 935 | 45(10) | 399 |
Lower caste or indigenous ethnicity | 60(11) | 496 | 6(5) | 262 | 22(43) | 660 | 43(12) | 308 |
Higher caste | 44(7) | 767 | 37(10) | 372 | 12(5) | 462 | 19(9) | 209 |
. | Treatment . | Hill control . | ||||||
---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |
Delivery in any facility | ||||||||
Eligibility criteria-I | 121(13) | 590 | 188(24) | 273 | 350(18) | 706 | 386(28) | 306 |
Eligibility criteria-II | 83(9) | 917 | 133(16) | 437 | 280(15) | 935 | 303(23) | 399 |
Lower caste or indigenous ethnicity | 60(11) | 496 | 112(20) | 262 | 154(14) | 660 | 164(21) | 308 |
Higher caste | 66(9) | 767 | 106(16) | 372 | 301(21) | 462 | 334(33) | 209 |
Delivery in government facility | ||||||||
Eligibility criteria-I | 90(12) | 590 | 147(21) | 273 | 300(17) | 706 | 281(26) | 306 |
Eligibility criteria-II | 60(8) | 917 | 105(15) | 437 | 242(14) | 935 | 224(21) | 399 |
Lower caste or indigenous ethnicity | 50(10) | 496 | 94(18) | 262 | 136(13) | 660 | 130(19) | 308 |
Higher caste | 44(7) | 767 | 80(14) | 372 | 243(20) | 462 | 251(30) | 209 |
Delivery by SBA | ||||||||
Eligibility criteria-I | 128(14) | 590 | 213(25) | 273 | 378(18) | 706 | 389(28) | 306 |
Eligibility criteria-II | 89(9) | 917 | 148(17) | 437 | 304(15) | 935 | 306(23) | 399 |
Lower caste or indigenous ethnicity | 66(11) | 496 | 114(20) | 262 | 170(15) | 660 | 171(21) | 308 |
Higher caste | 73(9) | 767 | 128(17) | 372 | 329(22) | 462 | 332(33) | 209 |
Neonatal death | ||||||||
Eligibility criteria-I | 55(9) | 590 | 37(11) | 273 | 20(5) | 706 | 49(12) | 306 |
Eligibility criteria-II | 60(8) | 917 | 29(8) | 437 | 19(4) | 935 | 45(10) | 399 |
Lower caste or indigenous ethnicity | 60(11) | 496 | 6(5) | 262 | 22(43) | 660 | 43(12) | 308 |
Higher caste | 44(7) | 767 | 37(10) | 372 | 12(5) | 462 | 19(9) | 209 |
Note: (1) Standard errors in parentheses. (2) Estimations are sample weighted.
In the longer phase, a 2-fold increase in maternal health service utilization indicators was observed across most of sub-groups of women in treatment districts (Table 3). Delivery in any facility increased from 83 per 1000 live-births to 156 per 1000 live-births in treatment districts and delivery by SBA increased from 89 per 1000 live-births to 166 per 1000 live-births among women under eligible criteria-II. Neonatal deaths declined in treatment districts for all sub-groups except for higher caste women for whom neonatal deaths were stagnant at 44 per 1000 live-births. In control districts neonatal death rates increased across all of the sub-groups.
. | Treatment . | Hill control . | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |||||
Delivery in any facility | ||||||||||||
Eligibility criteria-I | 121(13) | 590 | 218(19) | 454 | 350(18) | 706 | 412(22) | 506 | ||||
Eligibility criteria-II | 83(9) | 917 | 156(14) | 709 | 280(15) | 935 | 332(18) | 650 | ||||
Lower caste or indigenous ethnicity | 60(11) | 496 | 127(16) | 417 | 154(14) | 660 | 192(17) | 509 | ||||
Higher caste | 66(9) | 767 | 125(14) | 585 | 301(21) | 462 | 394(28) | 316 | ||||
Delivery in government facility | ||||||||||||
Eligibility criteria-I | 90(12) | 590 | 172(18) | 454 | 300(17) | 706 | 310(21) | 506 | ||||
Eligibility criteria-II | 60(8) | 917 | 126(12) | 709 | 242(14) | 935 | 253(17) | 650 | ||||
Lower caste or indigenous ethnicity | 50(10) | 496 | 103(15) | 417 | 136(13) | 660 | 156(16) | 509 | ||||
Higher caste | 44(7) | 767 | 104(13) | 585 | 243(20) | 462 | 298(26) | 316 | ||||
Delivery by SBA | ||||||||||||
Eligibility criteria-I | 128(14) | 590 | 237(20) | 454 | 378(18) | 706 | 412(22) | 506 | ||||
Eligibility criteria-II | 89(9) | 917 | 166(14) | 709 | 304(15) | 935 | 332(18) | 650 | ||||
Lower caste or indigenous ethnicity | 66(11) | 496 | 131(17) | 417 | 170(15) | 660 | 192(17) | 509 | ||||
Higher caste | 73(9) | 767 | 137(14) | 585 | 329(22) | 462 | 393(28) | 316 | ||||
Neonatal death | ||||||||||||
Eligibility criteria-I | 55(9) | 590 | 50(10) | 454 | 20(5) | 706 | 58(10) | 506 | ||||
Eligibility criteria-II | 60(8) | 917 | 43(8) | 709 | 19(4) | 935 | 49(8) | 650 | ||||
Lower caste or indigenous ethnicity | 60(11) | 496 | 22(7) | 417 | 22(6) | 660 | 43(9) | 509 | ||||
Higher caste | 44(7) | 767 | 44(8) | 585 | 12(5) | 462 | 24(9) | 316 |
. | Treatment . | Hill control . | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |||||
Delivery in any facility | ||||||||||||
Eligibility criteria-I | 121(13) | 590 | 218(19) | 454 | 350(18) | 706 | 412(22) | 506 | ||||
Eligibility criteria-II | 83(9) | 917 | 156(14) | 709 | 280(15) | 935 | 332(18) | 650 | ||||
Lower caste or indigenous ethnicity | 60(11) | 496 | 127(16) | 417 | 154(14) | 660 | 192(17) | 509 | ||||
Higher caste | 66(9) | 767 | 125(14) | 585 | 301(21) | 462 | 394(28) | 316 | ||||
Delivery in government facility | ||||||||||||
Eligibility criteria-I | 90(12) | 590 | 172(18) | 454 | 300(17) | 706 | 310(21) | 506 | ||||
Eligibility criteria-II | 60(8) | 917 | 126(12) | 709 | 242(14) | 935 | 253(17) | 650 | ||||
Lower caste or indigenous ethnicity | 50(10) | 496 | 103(15) | 417 | 136(13) | 660 | 156(16) | 509 | ||||
Higher caste | 44(7) | 767 | 104(13) | 585 | 243(20) | 462 | 298(26) | 316 | ||||
Delivery by SBA | ||||||||||||
Eligibility criteria-I | 128(14) | 590 | 237(20) | 454 | 378(18) | 706 | 412(22) | 506 | ||||
Eligibility criteria-II | 89(9) | 917 | 166(14) | 709 | 304(15) | 935 | 332(18) | 650 | ||||
Lower caste or indigenous ethnicity | 66(11) | 496 | 131(17) | 417 | 170(15) | 660 | 192(17) | 509 | ||||
Higher caste | 73(9) | 767 | 137(14) | 585 | 329(22) | 462 | 393(28) | 316 | ||||
Neonatal death | ||||||||||||
Eligibility criteria-I | 55(9) | 590 | 50(10) | 454 | 20(5) | 706 | 58(10) | 506 | ||||
Eligibility criteria-II | 60(8) | 917 | 43(8) | 709 | 19(4) | 935 | 49(8) | 650 | ||||
Lower caste or indigenous ethnicity | 60(11) | 496 | 22(7) | 417 | 22(6) | 660 | 43(9) | 509 | ||||
Higher caste | 44(7) | 767 | 44(8) | 585 | 12(5) | 462 | 24(9) | 316 |
Note: (1) Standard errors in parentheses. (2) Estimations are sample weighted.
. | Treatment . | Hill control . | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |||||
Delivery in any facility | ||||||||||||
Eligibility criteria-I | 121(13) | 590 | 218(19) | 454 | 350(18) | 706 | 412(22) | 506 | ||||
Eligibility criteria-II | 83(9) | 917 | 156(14) | 709 | 280(15) | 935 | 332(18) | 650 | ||||
Lower caste or indigenous ethnicity | 60(11) | 496 | 127(16) | 417 | 154(14) | 660 | 192(17) | 509 | ||||
Higher caste | 66(9) | 767 | 125(14) | 585 | 301(21) | 462 | 394(28) | 316 | ||||
Delivery in government facility | ||||||||||||
Eligibility criteria-I | 90(12) | 590 | 172(18) | 454 | 300(17) | 706 | 310(21) | 506 | ||||
Eligibility criteria-II | 60(8) | 917 | 126(12) | 709 | 242(14) | 935 | 253(17) | 650 | ||||
Lower caste or indigenous ethnicity | 50(10) | 496 | 103(15) | 417 | 136(13) | 660 | 156(16) | 509 | ||||
Higher caste | 44(7) | 767 | 104(13) | 585 | 243(20) | 462 | 298(26) | 316 | ||||
Delivery by SBA | ||||||||||||
Eligibility criteria-I | 128(14) | 590 | 237(20) | 454 | 378(18) | 706 | 412(22) | 506 | ||||
Eligibility criteria-II | 89(9) | 917 | 166(14) | 709 | 304(15) | 935 | 332(18) | 650 | ||||
Lower caste or indigenous ethnicity | 66(11) | 496 | 131(17) | 417 | 170(15) | 660 | 192(17) | 509 | ||||
Higher caste | 73(9) | 767 | 137(14) | 585 | 329(22) | 462 | 393(28) | 316 | ||||
Neonatal death | ||||||||||||
Eligibility criteria-I | 55(9) | 590 | 50(10) | 454 | 20(5) | 706 | 58(10) | 506 | ||||
Eligibility criteria-II | 60(8) | 917 | 43(8) | 709 | 19(4) | 935 | 49(8) | 650 | ||||
Lower caste or indigenous ethnicity | 60(11) | 496 | 22(7) | 417 | 22(6) | 660 | 43(9) | 509 | ||||
Higher caste | 44(7) | 767 | 44(8) | 585 | 12(5) | 462 | 24(9) | 316 |
. | Treatment . | Hill control . | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Before . | After . | Before . | After . | |||||||||
Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | Per 1000 live-births . | n . | |||||
Delivery in any facility | ||||||||||||
Eligibility criteria-I | 121(13) | 590 | 218(19) | 454 | 350(18) | 706 | 412(22) | 506 | ||||
Eligibility criteria-II | 83(9) | 917 | 156(14) | 709 | 280(15) | 935 | 332(18) | 650 | ||||
Lower caste or indigenous ethnicity | 60(11) | 496 | 127(16) | 417 | 154(14) | 660 | 192(17) | 509 | ||||
Higher caste | 66(9) | 767 | 125(14) | 585 | 301(21) | 462 | 394(28) | 316 | ||||
Delivery in government facility | ||||||||||||
Eligibility criteria-I | 90(12) | 590 | 172(18) | 454 | 300(17) | 706 | 310(21) | 506 | ||||
Eligibility criteria-II | 60(8) | 917 | 126(12) | 709 | 242(14) | 935 | 253(17) | 650 | ||||
Lower caste or indigenous ethnicity | 50(10) | 496 | 103(15) | 417 | 136(13) | 660 | 156(16) | 509 | ||||
Higher caste | 44(7) | 767 | 104(13) | 585 | 243(20) | 462 | 298(26) | 316 | ||||
Delivery by SBA | ||||||||||||
Eligibility criteria-I | 128(14) | 590 | 237(20) | 454 | 378(18) | 706 | 412(22) | 506 | ||||
Eligibility criteria-II | 89(9) | 917 | 166(14) | 709 | 304(15) | 935 | 332(18) | 650 | ||||
Lower caste or indigenous ethnicity | 66(11) | 496 | 131(17) | 417 | 170(15) | 660 | 192(17) | 509 | ||||
Higher caste | 73(9) | 767 | 137(14) | 585 | 329(22) | 462 | 393(28) | 316 | ||||
Neonatal death | ||||||||||||
Eligibility criteria-I | 55(9) | 590 | 50(10) | 454 | 20(5) | 706 | 58(10) | 506 | ||||
Eligibility criteria-II | 60(8) | 917 | 43(8) | 709 | 19(4) | 935 | 49(8) | 650 | ||||
Lower caste or indigenous ethnicity | 60(11) | 496 | 22(7) | 417 | 22(6) | 660 | 43(9) | 509 | ||||
Higher caste | 44(7) | 767 | 44(8) | 585 | 12(5) | 462 | 24(9) | 316 |
Note: (1) Standard errors in parentheses. (2) Estimations are sample weighted.
Study population characteristics at baseline
Before implementation of the programme in treatment districts, 88% of the women had primary or no education, 92% were Hindu, 63% were from higher castes (brahmin or chhetri), 35% were from lower or indigenous castes (dalit or janajati), most (98%) were living in rural areas, the mean age of a household head was 40.2 years and 48% of households had a size of 5 or less. In control districts at baseline, 70% of the women had primary or no education, 78% were Hindu, 33% were from higher castes (brahmin or chhetri), 55% were from lower or indigenous castes (dalit or janajati), 81% were living in rural areas, the mean age of a household head was 41.2 years and 54% were living in households with size of 5 or less.
DID estimates for period until change in eligibility—early phase
Impact estimates of the programme before and after free delivery care for early phase of the programme are reported as the difference in predicted probabilities in Table 4 (e.g. Puhani 2012). Provision of free delivery care lowered neonatal deaths in the early phase of the programme, with the probability of neonatal death in treatment districts being 4.5 percentage points (95%CI −0.066, −0.024) lower compared with control districts. The probability of delivery by SBAs was 5.6 percentage points (95%CI 0.002, 0.111) higher in treatment districts. Likewise the probability of delivery in a public facility was 5.1 percentage points (95%CI −0.003, 0.106) higher compared with control districts.
. | Delivery in any facility . | Delivery in government facility . | Delivery by SBA . | Neonatal death . |
---|---|---|---|---|
All women | 0.027(−0.025, 0.080) | 0.051(−0.003, 0.106) | 0.056(0.002, 0.111) | −0.045(−0.066, −0.024) |
Observations | 3,752 | 3,752 | 3,752 | 3,741 |
Eligibility criteria-I | 0.006(−0.067, 0.079) | 0.057(−0.026, 0.140) | 0.055(−0.022, 0.133) | −0.040(−0.072, −0.009) |
Observations | 1875 | 1875 | 1875 | 1869 |
Eligibility criteria-II | 0.027(−0.036, 0.090) | 0.065(−0.006, 0.134) | 0.061(−0.001, 0.124) | −0.061(−0.089, −0.033) |
Observations | 2688 | 2688 | 2688 | 2680 |
Lower caste/indigenous ethnicity | 0.069(−0.014, 0.152) | 0.063(−0.007, 0.135) | 0.082(−0.008, 0.173) | −0.069(−0.104, −0.035) |
Observations | 1726 | 1726 | 1726 | 1721 |
Higher caste | 0.014(−0.057, 0.085) | 0.040(−0.024, 0.106) | 0.065(−0.008, 0.138) | −0.017(−0.043, 0.007) |
Observations | 1808 | 1808 | 1808 | 1802 |
. | Delivery in any facility . | Delivery in government facility . | Delivery by SBA . | Neonatal death . |
---|---|---|---|---|
All women | 0.027(−0.025, 0.080) | 0.051(−0.003, 0.106) | 0.056(0.002, 0.111) | −0.045(−0.066, −0.024) |
Observations | 3,752 | 3,752 | 3,752 | 3,741 |
Eligibility criteria-I | 0.006(−0.067, 0.079) | 0.057(−0.026, 0.140) | 0.055(−0.022, 0.133) | −0.040(−0.072, −0.009) |
Observations | 1875 | 1875 | 1875 | 1869 |
Eligibility criteria-II | 0.027(−0.036, 0.090) | 0.065(−0.006, 0.134) | 0.061(−0.001, 0.124) | −0.061(−0.089, −0.033) |
Observations | 2688 | 2688 | 2688 | 2680 |
Lower caste/indigenous ethnicity | 0.069(−0.014, 0.152) | 0.063(−0.007, 0.135) | 0.082(−0.008, 0.173) | −0.069(−0.104, −0.035) |
Observations | 1726 | 1726 | 1726 | 1721 |
Higher caste | 0.014(−0.057, 0.085) | 0.040(−0.024, 0.106) | 0.065(−0.008, 0.138) | −0.017(−0.043, 0.007) |
Observations | 1808 | 1808 | 1808 | 1802 |
Note: 1. Confidence Interval in parentheses. 2. Estimates are difference in predicted probabilities as described in Puhani (2012) calculated after probit regression 3. Regression controlled for maternal, household and district characteristics (age, educational status of mother, occupation, religion, ethnicity, parity, use of radio, sex of household head, age of household head, household size, educational status of husband, household asset, presence of EOC programme, presence of CEOC facility, district level FCHV activity, year of birth, HDI of districts, recall period, road density).
. | Delivery in any facility . | Delivery in government facility . | Delivery by SBA . | Neonatal death . |
---|---|---|---|---|
All women | 0.027(−0.025, 0.080) | 0.051(−0.003, 0.106) | 0.056(0.002, 0.111) | −0.045(−0.066, −0.024) |
Observations | 3,752 | 3,752 | 3,752 | 3,741 |
Eligibility criteria-I | 0.006(−0.067, 0.079) | 0.057(−0.026, 0.140) | 0.055(−0.022, 0.133) | −0.040(−0.072, −0.009) |
Observations | 1875 | 1875 | 1875 | 1869 |
Eligibility criteria-II | 0.027(−0.036, 0.090) | 0.065(−0.006, 0.134) | 0.061(−0.001, 0.124) | −0.061(−0.089, −0.033) |
Observations | 2688 | 2688 | 2688 | 2680 |
Lower caste/indigenous ethnicity | 0.069(−0.014, 0.152) | 0.063(−0.007, 0.135) | 0.082(−0.008, 0.173) | −0.069(−0.104, −0.035) |
Observations | 1726 | 1726 | 1726 | 1721 |
Higher caste | 0.014(−0.057, 0.085) | 0.040(−0.024, 0.106) | 0.065(−0.008, 0.138) | −0.017(−0.043, 0.007) |
Observations | 1808 | 1808 | 1808 | 1802 |
. | Delivery in any facility . | Delivery in government facility . | Delivery by SBA . | Neonatal death . |
---|---|---|---|---|
All women | 0.027(−0.025, 0.080) | 0.051(−0.003, 0.106) | 0.056(0.002, 0.111) | −0.045(−0.066, −0.024) |
Observations | 3,752 | 3,752 | 3,752 | 3,741 |
Eligibility criteria-I | 0.006(−0.067, 0.079) | 0.057(−0.026, 0.140) | 0.055(−0.022, 0.133) | −0.040(−0.072, −0.009) |
Observations | 1875 | 1875 | 1875 | 1869 |
Eligibility criteria-II | 0.027(−0.036, 0.090) | 0.065(−0.006, 0.134) | 0.061(−0.001, 0.124) | −0.061(−0.089, −0.033) |
Observations | 2688 | 2688 | 2688 | 2680 |
Lower caste/indigenous ethnicity | 0.069(−0.014, 0.152) | 0.063(−0.007, 0.135) | 0.082(−0.008, 0.173) | −0.069(−0.104, −0.035) |
Observations | 1726 | 1726 | 1726 | 1721 |
Higher caste | 0.014(−0.057, 0.085) | 0.040(−0.024, 0.106) | 0.065(−0.008, 0.138) | −0.017(−0.043, 0.007) |
Observations | 1808 | 1808 | 1808 | 1802 |
Note: 1. Confidence Interval in parentheses. 2. Estimates are difference in predicted probabilities as described in Puhani (2012) calculated after probit regression 3. Regression controlled for maternal, household and district characteristics (age, educational status of mother, occupation, religion, ethnicity, parity, use of radio, sex of household head, age of household head, household size, educational status of husband, household asset, presence of EOC programme, presence of CEOC facility, district level FCHV activity, year of birth, HDI of districts, recall period, road density).
Estimates of aggregate impacts may hide differential sub-group level effects. Table 4 shows results for programme impacts in population sub-groups. The probability of neonatal death was lower by 4 percentage points (95%CI −0.072, −0.009) among eligible criteria-I women, by 6.1 percentage points (95%CI −0.001, 0.124) among eligibility criteria-II women and by 6.9 percentage points (95%CI −0.104, −0.035) among women from lower castes or indigenous ethnic groups in treatment districts. We also observed a corresponding increase of delivery by SBAs by 6.1 percentage points (95%CI −0.001, 0.124) and 8.2 percentage points (95%CI −0.008, 0.173) among eligibility criteria-II women and women belonging to lower castes or indigenous groups, respectively. The probability of delivery in a public facility increased by 6.5 percentage points (95%CI −0.006, 0.134) and 6.3 percentage points (95%CI −0.007, 0.135) among women under eligibility criteria-II and lower castes or indigenous ethnic groups. Although effect signs remained positive, the results were statistically insignificant for the indicator “delivery in any facility”’ for all sub-groups of women in the early phase.
DID estimates for longer exposure
In Table 5, we report the difference in predicted probabilities for longer phase of programme implementation. We continued to observe lower probability of neonatal mortality in treatment groups in the longer phase of the programme: for all women (3.9 percentage points, 95%CI −0.060, −0.018), women under eligibility criteria-I (4.2 percentage points, 95%CI −0.071, −0.012), eligibility criteria-II (5.0 percentage points, 95%CI −0.078, −0.022) and lower caste women (5.6 percentage points, 95%CI: −0.089, −0.024). The effect magnitudes were however lower than in the early phase of the programme. Unlike the early phase, the results for the likelihood of delivery in a government facility and delivery by SBA were also smaller in magnitude and statistically insignificant, even though mostly positive in sign. As with the early phase, the greatest increase in delivery by SBA was observed among lower caste/indigenous women for whom the predicted probability was 4.6 percentage points (95%CI −0.030, 0.123) higher in treatment districts (statistically insignificant).
. | Delivery in any facility . | Delivery in government facility . | Delivery by SBA . | Neonatal death . |
---|---|---|---|---|
All women | 0.012(−0.045, 0.070) | 0.046(−0.017, 0.110) | 0.032(−0.020, 0.086) | −0.039(−0.060, −0.018) |
Observations | 4457 | 4457 | 4457 | 4446 |
Eligibility criteria-I | −0.002(−0.086, 0.080) | 0.050(−0.040, 0.141) | 0.033(−0.043, 0.110) | −0.042(−0.071, −0.012) |
Observations | 2256 | 2256 | 2256 | 2250 |
Eligibility criteria-II | 0.008(−0.062, 0.080) | 0.050(−0.027, 0.128) | 0.032(−0.032, 0.097) | −0.050(−0.078, −0.022) |
Observations | 3211 | 3211 | 3211 | 3203 |
Lower caste/indigenous ethnicity | 0.033(−0.042, 0.110) | 0.030(−0.041, 0.103) | 0.046(−0.030, 0.123) | −0.056(−0.089, −0.024) |
Observations | 2082 | 2082 | 2082 | 2077 |
Higher caste | 0.002(−0.070, 0.075) | 0.044(−0.019, 0.109) | 0.030(−0.037, 0.098) | −0.021(−0.049, 0.006) |
Observations | 2128 | 2128 | 2128 | 2120 |
. | Delivery in any facility . | Delivery in government facility . | Delivery by SBA . | Neonatal death . |
---|---|---|---|---|
All women | 0.012(−0.045, 0.070) | 0.046(−0.017, 0.110) | 0.032(−0.020, 0.086) | −0.039(−0.060, −0.018) |
Observations | 4457 | 4457 | 4457 | 4446 |
Eligibility criteria-I | −0.002(−0.086, 0.080) | 0.050(−0.040, 0.141) | 0.033(−0.043, 0.110) | −0.042(−0.071, −0.012) |
Observations | 2256 | 2256 | 2256 | 2250 |
Eligibility criteria-II | 0.008(−0.062, 0.080) | 0.050(−0.027, 0.128) | 0.032(−0.032, 0.097) | −0.050(−0.078, −0.022) |
Observations | 3211 | 3211 | 3211 | 3203 |
Lower caste/indigenous ethnicity | 0.033(−0.042, 0.110) | 0.030(−0.041, 0.103) | 0.046(−0.030, 0.123) | −0.056(−0.089, −0.024) |
Observations | 2082 | 2082 | 2082 | 2077 |
Higher caste | 0.002(−0.070, 0.075) | 0.044(−0.019, 0.109) | 0.030(−0.037, 0.098) | −0.021(−0.049, 0.006) |
Observations | 2128 | 2128 | 2128 | 2120 |
Note: 1. Confidence Interval in parentheses. 2. Estimates are difference in predicted probabilities as described in Puhani (2012) calculated after probit regression 3. Regression controlled for maternal, household and district characteristics (age, educational status of mother, occupation, religion, ethnicity, parity, use of radio, sex of household head, age of household head, household size, educational status of husband, household asset, presence of EOC programme, presence of CEOC facility, district level FCHV activity, year of birth, HDI of districts, recall period, road density).
. | Delivery in any facility . | Delivery in government facility . | Delivery by SBA . | Neonatal death . |
---|---|---|---|---|
All women | 0.012(−0.045, 0.070) | 0.046(−0.017, 0.110) | 0.032(−0.020, 0.086) | −0.039(−0.060, −0.018) |
Observations | 4457 | 4457 | 4457 | 4446 |
Eligibility criteria-I | −0.002(−0.086, 0.080) | 0.050(−0.040, 0.141) | 0.033(−0.043, 0.110) | −0.042(−0.071, −0.012) |
Observations | 2256 | 2256 | 2256 | 2250 |
Eligibility criteria-II | 0.008(−0.062, 0.080) | 0.050(−0.027, 0.128) | 0.032(−0.032, 0.097) | −0.050(−0.078, −0.022) |
Observations | 3211 | 3211 | 3211 | 3203 |
Lower caste/indigenous ethnicity | 0.033(−0.042, 0.110) | 0.030(−0.041, 0.103) | 0.046(−0.030, 0.123) | −0.056(−0.089, −0.024) |
Observations | 2082 | 2082 | 2082 | 2077 |
Higher caste | 0.002(−0.070, 0.075) | 0.044(−0.019, 0.109) | 0.030(−0.037, 0.098) | −0.021(−0.049, 0.006) |
Observations | 2128 | 2128 | 2128 | 2120 |
. | Delivery in any facility . | Delivery in government facility . | Delivery by SBA . | Neonatal death . |
---|---|---|---|---|
All women | 0.012(−0.045, 0.070) | 0.046(−0.017, 0.110) | 0.032(−0.020, 0.086) | −0.039(−0.060, −0.018) |
Observations | 4457 | 4457 | 4457 | 4446 |
Eligibility criteria-I | −0.002(−0.086, 0.080) | 0.050(−0.040, 0.141) | 0.033(−0.043, 0.110) | −0.042(−0.071, −0.012) |
Observations | 2256 | 2256 | 2256 | 2250 |
Eligibility criteria-II | 0.008(−0.062, 0.080) | 0.050(−0.027, 0.128) | 0.032(−0.032, 0.097) | −0.050(−0.078, −0.022) |
Observations | 3211 | 3211 | 3211 | 3203 |
Lower caste/indigenous ethnicity | 0.033(−0.042, 0.110) | 0.030(−0.041, 0.103) | 0.046(−0.030, 0.123) | −0.056(−0.089, −0.024) |
Observations | 2082 | 2082 | 2082 | 2077 |
Higher caste | 0.002(−0.070, 0.075) | 0.044(−0.019, 0.109) | 0.030(−0.037, 0.098) | −0.021(−0.049, 0.006) |
Observations | 2128 | 2128 | 2128 | 2120 |
Note: 1. Confidence Interval in parentheses. 2. Estimates are difference in predicted probabilities as described in Puhani (2012) calculated after probit regression 3. Regression controlled for maternal, household and district characteristics (age, educational status of mother, occupation, religion, ethnicity, parity, use of radio, sex of household head, age of household head, household size, educational status of husband, household asset, presence of EOC programme, presence of CEOC facility, district level FCHV activity, year of birth, HDI of districts, recall period, road density).
Robustness checks
Our results on the increase in the likelihood of delivery by SBAs and declines in neonatal death rates were robust to the exclusion of Kathmandu valley districts in the early phase; but delivery in a public facility was no longer statistically significant, although the direction of the effect was unchanged (see Supplementary Appendix SD2). The estimated declines in the probability of neonatal death among all women, women under eligibility criterion-II and lower caste women were consistent with our results using hill and terai controls, albeit with slightly smaller effect sizes in the early phase. Our results were also largely robust to the use of the combined hill and terai as controls in the longer phase of the programme.
Our DID analysis for a matched dataset of births based on the characteristics of women giving birth provided results similar to the findings reported in Table 4 (see Supplementary Appendix SF1). Estimates from matched data typically reduce differences in baseline characteristics between treatment and control groups and may better address confounding due to trends. However, as a practical matter, matching of observations reduced sample size and therefore the precision of our estimates. For this reason, we used results from the matching analysis primarily for checking the robustness of our results.
Discussion
Overall, we found a positive and statistically significant effect of user fee removal on delivery by SBAs and deliveries in public facilities in the early phase of the programme. A previous impact evaluation of the whole SDIP (Powell-Jackson and Hanson 2012) conducted in six districts (of which three districts were implementing free care as well as cash transfer and three were implementing only cash transfers) found a significant increase of 4.2% for SBA while our results showed 5.6 percentage points increase in the probability of delivery by SBA. Similarly, we also find an impact on the use of public facilities. The difference could be explained by the fact that Powell-Jackson and Hanson (2012) evaluated the impact of the full SDIP programme, inclusive of the cash transfer and free care components.
The results on the increased probability of health service utilization were however not sustained for longer periods of exposure to the programme. The effect size was also lower in longer phase. This phenomenon of unsustained increase in health service utilization after the removal of user-fees has also been reported in systematic reviews of user-fees evaluations (Lagarde and Palmer 2008; Dzakpasu et al. 2013). One of the reasons for observation of an immediate but unsustained effect in health service utilisation is a likely lowering of the quality of care. In our case, the initial increase in demand for services may have run into supply side bottlenecks and increased waiting times in public facilities. Another reason for lack of effect for longer exposure could be due to delays in timely payments under the cash transfer programme that have been reported in previous evaluations (Powell-Jackson et al. 2008). Low levels of awareness in the larger population could also explain the findings over the longer exposure period. Studies report that only 27% of mothers were aware of the programme while they were pregnant and only 26% of deliveries were provided free of charge (Powell-Jackson et al. 2008; Powell-Jackson and Hanson 2012). A qualitative study of SDIP conducted in ten districts supports these findings, noting multiple challenges in programme implementation, including bureaucratic delays in disbursement of funds, confusion on eligibility criteria and lack of reimbursement to health facilities for providing free delivery care (Powell-Jackson et al. 2009).
We did not find any programme effects on delivery in “any facility.” Although initially puzzling, these results may reflect a shift occurring from deliveries in private facilities to public facilities. Our findings of insignificant effects are similar to Mohanan et al. (2014) for Gujarat in India, who also did not find a significant impact of the Chiranjeevi programme in increasing facility-based deliveries. Although the Chiranjeevi programme differed from Nepal’s free delivery care because it covered the cost of deliveries at private hospitals for below-poverty line households, the principle aim of both programmes was to increase facility-based births (or births attended by SBAs).
We find a significant reduction in probability of neonatal deaths in treatment districts in the early phase that continued in longer phase, albeit with lower magnitude. The decline of the likelihood of neonatal mortality is driven by births among women in lower castes/indigenous groups and eligible criteria-II women in treatment districts so that programme benefits may be equity enhancing. This decline is consistent with increases in deliveries assisted by SBAs and a rise in the use of public facilities. For example, the decline in neonatal mortality in the early phase among lower caste/indigenous women (reduction by 6.9 percentage points) was accompanied by significant increases in SBA assisted deliveries by 8.2 percentage points and use of public facility by 6.3 percentage points, consistent with the literature (Lawn et al. 2005).
We believe that our findings of a decline in neonatal mortality may also reflect a relative increase in the use of public services by women with increased risk of birth complications. Although lacking direct evidence, our results of a significant impact on neonatal mortality among live-births for women eligible under criteria-II (i.e. women with two living children and/or high parity) relative to those for criteria-I women suggest this possibility. The lead author’s discussions with senior officials in the MOH in Nepal suggest that changes in eligibility were undertaken precisely to ensure women with high parity can also utilize health services (S Aryal, personal communication). If true, this would imply that women with a greater likelihood of obstetric complications are more price sensitive than average. Although obstetric complications tend to be unpredictable (Bailey et al. 2009), qualitative results suggest that women tend to visit health facilities only after complications arise (Powell-Jackson et al. 2008; Pradhan et al. 2010). So if there is price sensitivity in seeking care, this would be associated with either a delay in accessing care or not seeking care. Multiple studies have demonstrated that cost is one of the major drivers in delaying the decision to seek care (Manandhar 2000; Borghi et al. 2004; Pradhan et al. 2010).
If institutional delivery or professional care reduces neonatal mortality, then it should be reflected in early neonatal mortality (death within 7 days of birth) (Filippi et al. 2006). We examined early neonatal mortality to assess if quality and services had influence on the reduction of mortality and to ensure post-natal care did not confound our results (Filippi et al. 2006). In general, we observed a decline of 3.1 percentage points and 2.5 percentage points in the likelihood of early neonatal deaths in the early programme phase and longer programme phase respectively.
As in any such analysis, our findings have some limitations. Our eligibility criteria-I was based on the number of living children and eligibility criteria-II was based on number of living children and high parity, which is less than ideal as we do not directly capture birth complications. Our impact estimates are based on Intent to treat and not the Average Treatment Effect on Treated, as we did not have programme coverage data. We were limited to evaluating effects only during the early years of the programme. Our analysis, based on cross-sectional household surveys repeated over time in treatment and control districts requires two key assumptions: that the mean pre-intervention outcomes (conditional on covariates) for individuals living in the treatment post-intervention, is the same as the mean outcomes for individuals living in treatment districts in period 0 (pre-intervention); and the same trend holds on average (in the absence of the intervention) in treatment and control households (conditional on covariates) between pre- and post-intervention periods (Lee and Kang 2006). We were unable to address the first assumption. Although we have sought to partially address the second assumption in our analysis on pre-intervention trends, residual confounding due to trend may remain.
Finally, information on the outcomes of interest was based on respondent recall of events up to 5 years in the past, raising the possibility of recall bias. In general, we do not have any reason to suspect systematic differences in reporting between treatment and control groups, and studies conducted previously using similar methods (McKinnon et al. 2014; Valente 2015) have tended not to consider this issue as being significant. It is possible, of course that media campaigns about the programme may have increased people’s recall of births in public facilities in treatment areas, which would lead to a bias away from the null. On the other hand, available evidence suggests that information about the programme was generally poor among Nepal’s population (Powell-Jackson et al. 2008). NDHS datasets have been extensively used for estimating maternal health service utilization and child health outcomes (Corsi et al. 2012). It is also important to note that a study using Malaysian data from 1970s and 1980s found that fertility history for live-born children is not affected by recall bias (Beckett et al. 2001).
Conclusion
Our results suggest that removal of user fees for deliveries in Nepal’s public facilities increased maternal health service use and lowered the likelihood of neonatal among different sub-groups of women. Moreover, our results suggest that women from lower caste/indigenous ethnicity and women with birth complications are likely to have benefited from this intervention.
The effect on maternal health service utilization however was not sustained over a longer exposure period, which may reflect longer run problems with quality of care (e.g. overcrowding and longer waiting times). The results on the probability of neonatal mortality persisted even with longer programme exposure, although effect magnitudes were smaller. Our results point to the importance of user fee removal for delivery on neonatal health outcomes in Nepal as well as equity in outcomes.
Conflict of interest statement. None declared.
Acknowledgements
P.L. would like to thank Monash University for the Monash Graduate Scholarship and Monash International Postgraduate Scholarship for doctoral studies.
Funding
PL was supported by Monash Graduate Scholarship and Monash International Postgraduate Scholarship for his doctoral studies.