The Effect of A Demand Side Financing Program on the Continuum of Maternal and Child Health Care in India

Background The Indian Government launched a demand-side nancing program, ‘Janani Suraksha Yojana’ (JSY) in 2005 with the aim of reducing maternal and neonatal mortality through increased access to institutional delivery care service. This paper analyses the effects of the JSY on the uptake of maternal and child health (MNCH) care services intending to understand the overall impact of the program on the continuum of care. Methods Using the 2013-14 round of the District-level Household Survey (DLHS) surveyed in high performing states, the average treatment effect on the treated (ATT) was estimated by using the Propensity Score Matching (PSM). Results are reported regarding both ATTs and deviations from the theoretical continuum of care line, which represents 100% uptake, i.e., all women availing all the MNCH services. Results Overall, JSY effects on MNCH components ranged between 0.7% and 12%. As expected, the highest impact of the JSY was on institutional delivery (ATT: 0.12; 95% CI: 0.104-0.131) and the lowest for breastfeeding more than six months (ATT: 0.007; 95% CI: -0.014-0.027). Deviation from the complete continuum of care line ranged from 2.3% to 80.9%. The highest deviation was for three or more Tetanus Toxoid (100%-19.1%=80.9%) injections and the lowest for Polio given at any time (100%-97.7%=2.3%). Conclusions The program had high effects on those MNCH care services, the uptakes of which were already high without the program (low deviations from the continuum of care line), and the program had low effects on those MNCH care services, which had low uptakes in the absence of the program (high deviations from the continuum of care line). The program should also incentivize the utilization of those MNCH care services, which have low uptakes in the absence of the program. child health (MNCH) is already known. But there is no research on the impact of JSY on all the MNCH components as continuum. This research provides new knowledge on the impact of JSY on all components of MNCH. the of care would Any from the continuum of to the level of uptake of a service/component indicates a gap in the uptake. The gap could on the JSY intervention or lack of it. It is hypothesized here that the JSY intervention will increase the uptake of a service and decrease the gap. The objective of this paper is to measure gaps between uptakes with and without JSY and also gaps between the continuum of care line and uptakes without JSY. These measures together will demonstrate the effect of the JSY on the MNCH continuum of care.


Introduction
India has made signi cant progress in maternal and child health (MNCH) indicators in the last two decades. Maternal and neonatal death rates are still higher than that of economies of similar magnitude. For example, the maternal mortality ratio (per 100,000 live births) was 174 in 2015, which was higher than corresponding gures for comparable developing countries in Asia, such as Indonesia (126), and Malaysia (79) (WHO, 2015). Access to and/or demand for maternal and child health services were identi ed as barriers to progress in maternal and child health in many developing countries, including India. To overcome these demand-side barriers, following multiple countries (Anwar et al. 2008 JSY is one of the largest DSF programs in the world. It was launched in April 2005 by the prime minister of India. The key objective of the JSY is reducing maternal and neonatal mortality by promoting institutional delivery in public or accredited private healthcare institutions. The JSY initially targeted poorer women in selected poorer states. However, JSY was extended to all states in 2007. As of January 2017, all pregnant women are eligible for JSY in 10 low performing states (LPS) where the rates of institutional delivery were below 25%. These are: Uttar Pradesh, Uttaranchal, Bihar, Jharkhand, Madhya Pradesh, Chhattisgarh, Assam, Rajasthan, Orissa, and Jammu and Kashmir. All other states were high performing states (HPS), those with more than 25% institutional delivery rates. Only the poor and marginal women, such as scheduled caste and tribal women, are eligible for the program in HPS.
The overall eligibility criteria for JSY in LPS and HPS are listed in Table 1. However, in HPS, some of the criteria were amended in 2013 by removing restrictions regarding age and the number of children. As earlier (in Table 1), JSY was only open to women living below the poverty line, scheduled caste, or scheduled tribe categories. But the earlier version excluded those women who were 19 years and less and had more than the second birth. This amendment was based on the review that found women aged below 19 years and with higher numbers of children were at higher risk of maternal mortality and their children experienced higher rates of neonatal mortality. Although these restrictions were amended in May 2013, we are unable to report when they were implemented in various states. However, these changes have no impact on the present study, as our data comes from the survey happened in 2013-2014 and the times of the last births of women used in this study fall within the period from 2008 to just before the survey (that means the survey included those women who already gave their last births). So, this study uses the eligibility criteria mentioned in Table 1. JSY is implemented through the public healthcare system. Accredited social health activist (ASHA) and grassroots level health volunteers work as an intermediary between the health system and pregnant women to implement the JSY. In both LPS and HPS, the roles of the ASHA associated with the JSY are as follows (Yojana 2006): (1) to identify the eligible pregnant woman for JSY and to report or facilitate registration for Antenatal Care (ANC); (2) to assist the pregnant woman in obtaining the necessary certi cates of the eligibility for JSY; (3) to provide and/or help the pregnant woman in receiving at least three ANC check-ups, including tetanus injections and Iron Folic Acid tablets; (4) to identify a functional Government healthcare centre or an accredited private health institution for referral and delivery; (5) to counsel for institutional delivery; (6) to escort the bene ciary woman to the pre-determined health centre and stay with her until discharged; (7) to arrange to immunize the new born until the age of 14 weeks; (8) to inform of the birth or death of the child or mother to the auxiliary nurse and midwife ; (9) to arrange a postnatal visit within 7 days of delivery to track mother's health after delivery and facilitate in obtaining care, wherever necessary; (10) to counsel for initiation of breastfeeding of the new-born within one-hour of delivery and its continuance until 3-6 months and promote family planning. ASHA receive nancial incentives for undertaking the above activities under the JSY.
When the eligible pregnant woman visits a public health facility for the rst time to register her entitled bene ts/services -three antenatal care services including iron-folic acid and tetanus injection, institutional delivery and one postnatal care service for both mother and the new born, she receives her total entitled cash as payee cheque from that facility. If she goes rst time to an accredited private health facility for those entitled bene ts' registrations, that facility gives her the third-fourth of her entitled cash at the time of that visit and the rest at the time of her discharge after delivery. Without those services' registrations, an eligible woman will not receive her entitled cash at all. The ASHA gets her entitled money also from the facility where the JSY recipient woman gave her birth but only at the time of her discharge after delivery. As JSY directly works to increase the uptake of the above registered services, these are called intended outcomes. The program may have effects not only on intended outcomes but also on unintended outcomes such as immunization and birth rate because of a spill over effect. There are several studies (Carvalho et Modugu et al. 2012), which show the effects of JSY on both intended and unintended outcomes.. However, these studies were limited to discussing the effects of JSY on some speci c (intended or/and unintended) MNCH care components/services. Such treatment effect analyses are not able to provide su ciently a broad picture regarding pathways towards reducing maternal and child mortality rates (Angrist and Pischke 2008). For this, a method that provides information about how the program is performing with regard to all (possible intended and unintended) service components is needed. The proposed approach is consistent with the current understanding that gains in reproductive, maternal, newborn, child and adolescent health can best be realised by ensuring services across the continuum of care, rather than concentrating on one element or some elements and the neglect of others.
The hypothetical/theoretical Fig. (1) illustrates the concept of the continuum of care. The horizontal axis of the gure shows four hypothetical MNCH care services, such as MNCH1-MNCH4. For the time being, we consider that these are all possible intended and unintended service components under the continuum of care required to reduce maternal and child mortality. The vertical axis of the gure shows the utilization rates (or uptakes) of those four service components with JSY and without JSY. That axis also shows the continuum of care line (parallel to the vertical line) at where the utilization rates of those service components are 100% (e.g., 1). If all mothers in an area or a country received all the components in the gure, the continuum of care line would have been obtained. Any deviation from the straight (e.g., continuum of care) line to the level of uptake of a service/component (indicated in red with JSY and green without JSY) indicates a gap in the uptake. The gap could depend on the JSY intervention or lack of it. It is hypothesized here that the JSY intervention will increase the uptake of a service and decrease the gap. The objective of this paper is to measure gaps between uptakes with and without JSY and also gaps between the continuum of care line and uptakes without JSY. These measures together will demonstrate the effect of the JSY on the MNCH continuum of care.

Data
The study used wave 4 (2013-14) of the Indian District Level Household Survey (DLHS-4). DLHS is the most suitable for the present study, and earlier waves of DLHS have already been widely considered for JSY evaluation (Carvalho et  Bengal) and three high performing Union Territories (Andaman and Nicobar Islands, Chandigarh, and Pondicherry) where the institutional delivery rates are more than 25% (MoHFW 2013). Data are not available for other states and Union Territories. However, these 21 high performing states and Union Territories contain around two-thirds of the Indian population. In this study, we use data from all of these 21 high performing states and Union Territories.
The DLHS-4 adopted a multi-stage strati ed systematic sampling design. Data were collected through interviews using bilingual questionnaires. The survey covered 378,487 households. In this study we used data from 73,073 mothers in the age group 15-49 years who had provided information about their pregnancies since 2008. In this study only mothers who have responded to all the continuum of care variables are included. As JSY was implemented in 2007 in the HPS, all pregnancies captured by DLHS-4 fall within the time frame of the program implementation.
Propensity score matching Women are selected for JSY based on eligibility criteria, i.e. observed characteristics. In such cases, Propensity Score Matching (PSM) is a widely used method, which can turn data so that random distribution of treatment and control groups, under unconfoundedness and overlapping assumptions, can be availed when there was actually no randomization (Angrist and Pischke 2008). However, for unbiased results, matching variables should be chosen carefully so that the problem of incomplete (inadequate matching variables) and inexact (unrelated matching variables) matching is minimal. We have chosen matching variables based on eligibility criteria chosen by the JSY and self-selection criteria of women. We assume that there is no self-selection bias or endogeneity bias because, after matching, unobserved factors also become random between treatment and control groups along with observed factors.
In the online appendix, we have described which matching variables are used to estimate propensity score. We have also checked whether two crucial assumptions, unconfoundedness and overlap, are satis ed or not. Which algorithm is used is also mentioned in the online appendix. Overall, we have followed mostly the guidelines of Yao et al. (2017) to explain the propensity score method brie y.
The list of outcome variables is shown in Table 4. All of them are the continuum of care components, which are the dummy variables. They have a common sample size, which is 29,858. As we mentioned earlier, the sample size drastically falls from around 70,000 women to 29,858 women, because many women did not respond about child immunizations because probably their children did not reach the ages of immunizations and therefore did not take immunizations. In other words, mostly huge numbers of missing values in immunizations left the common sample as low. However, we have to focus on these common women, because, for the continuum of care analysis, we need to know how much a group of women utilized all continuum of care services. In other words, we need to understand how the same women behave differently in different components of the continuum of care. If we take different samples for different components of the continuum of care, the gaps between the continuum of care line and the actual uptakes of those care components will not be comparable. Table 4 provides ATTs with 95% con dence intervals and associated probabilities. It also provides means of MNCH variables in the treatment group (JSY) and control group. ATT is the difference between the means of MNCH services received by the treatment group and the service taken by the same group if they were not treated i.e. if they were not JSY recipients (Rubin 1974;Rubin and John 2001). ATT is obtained from the matched sample obtained through PSM. The means of MNCH service indicators of the treated group under both unmatched and matched conditions are the same. In the controls column, the means of MNCH services in the unmatched and matched categories are different. This is because, in the unmatched sample, we estimated means of MNCH indicators of the non-recipients (control group) of the JSY. In the matched sample, we calculated the means of MNCH services of JSY recipients (treatment group) as if they were non-JSY recipients. This is done by imputing MNCH indicators of the treated group after nding the closest matches from the untreated/control group. So, ATT is the difference between the mean of actual values of a MNCH service of the treatment group and mean of the imputed values of that service of the same group, but those imputed values are taken from the control group through matching. Institutional delivery, postnatal care and advice for the mother

Postnatal care for children
The JSY effect on the postnatal care for children are given in Fig. 2(d) and Table 4 (Sect. 4). Among the measures taken in this section, the highest impact of JSY was on weight taken at birth, the effect was 10% (0.929 − 0.833 = 0.096). The effect of JSY on check-up after delivery was 6% (85.9-79.9 = 0.060). It appears that not all children whose weight was taken at birth had a check-up after birth. Breastfeeding for more than 6 months was not signi cantly different between the treated and control group, suggesting no effect of JSY on breastfeeding (0.466 − 0.459 = 0.007). The JSY had around a 5% effect (0.573 − 0.522 = 0.052) on increasing women's knowledge about managing diarrhoea for the new-born. Deviation from the line of continuum of care to the line of the treated group varied between 2.3 to 25.7%. The highest deviation was women's knowledge about phenomena (69.5%) and the lowest was for weight taken at birth (7.1%).

Immunisation and Vitamin A for children
This section examines the effect of JSY on immunisation, although JSY had no mandate to increase immunisation. It considers 6 immunisations and the results are given in Fig. 2(e) and Table 4 (Sect. 5). Overall, immunisation coverage in the study population without the JSY ranged between 68%-98%. Therefore, the JSY had a relatively lesser scope to further increase uptake of some of the immunisation, particularly those close to Overall, the JSY effect on MNCH indicators varied between 1% and 12%. No woman in the study had taken all the 45 MNCH services and no MNCH service was taken by all the women. It is important to note that not all elements of MNCH are equally important for maternal and child survival, but they all are important to ensure longer-term health and well-being of the mother and child. Therefore, it is necessary to utilise all the MNCH services recommended for mother and child health, which is still at risk in developing countries, including India.
Increasing institutional delivery uptake has been the key focus of JSY. In this study the effect of JSY is about 12%. That means an additional 12% of women have taken institutional delivery due to JSY, and this leads to about 95% women availing institutional delivery in the HPS state leaving a remaining 5% to achieve 100% institutional delivery. This is higher than the effect shown in an earlier study carried out in HPS using previous rounds of DLHS (Lim et al. 2010). Studies have also shown that the effect of JSY on the low performing states (LPS), as expected, was much higher. In the HPS, clearly the JSY has been successful in accomplishing its mandate to increase institutional delivery. However, the fact that JSY did not increase the uptake of MNCH indicators uniformly across the MNCH services suggests that the spill over effect of JSY was limited, particularly to those services for which uptake has been very low. Some women have taken institutional delivery without being registered for pregnancy or having access to ANC further suggest some leakage of JSY services. It appears that the focus of JSY on institutional delivery in some cases tends to disregard with ANC services. An increased attention towards pregnancy registration and ANC as the pathway to institutional delivery may be emphasised in the JSY to achieve a continuum of care for MNCH.
Iron De ciency Anaemia (IDA) is high in India. Treating IDA is vital for better maternal and child health outcomes and (Peña-Rosas et al. 2015). Iron Folic Acid (IFA) is provided to all pregnant women free of cost through the public healthcare service providers, particularly through ASHA, and is one of the activities to be undertaken as part of the JSY. JSY's effect on the uptake of IFA was about 9% and there is a further 21% increase required to reach 100%. A related issue is blood tests carried out for haemoglobin as part of ANC, which is 73% with JSY. Thus, over 20% of pregnant women do not receive IFA or blood test for haemoglobin despite JSY and this calls for a need to re-examine JSY priorities.
Generally, provision of advice on maternal and child health such as advice on family planning, nutrition, diarrhoea, breastfeeding, pneumonia is lower than other more clinically oriented services; such as ANC, delivery, or immunisation. The spill over effect of JSY on advice on various MNCH indicators was limited. As clinically oriented MNCH service uptake is increasing in the HPS, JSY may reorganize to include MNCH indicators that re ect maternal and child health beyond the pre and postnatal period.
In general, uptake of immunizations for mother and children are high. However, dispensing polio within the rst two weeks, Vitamin A for children, and three or more doses of Tetanus Toxoid for pregnant women were relatively lower compared to other immunisations.

Conclusion
JSY is an effective intervention to increase institutional deliveries in the HPS of India. As many of the MNCH indicators are performing better, it is now time for the JSY to redesign its key objectives to include indicators that are performing poorly. The program may consider staged conditional cash transfer to cover the entire continuum of care spectrum.
Using PSM has possible self-selection/endogeneity/simultaneity bias. The participation in JSY can be affected by unobserved factors along with observed factors. There can be some criteria of the inclusion of a woman into JSY which are uno cially determined by the program administrators. These can be personal relationships between a woman or her family, ASHA worker, geographical location of a woman, and political a liation of a woman's family. These types of variables are absent in the data. There can be some other potential unobserved factors, which are self-selection criteria of women. These include social restrictions, such as husband's disallowance to participate in JSY and reliability on traditional health workers. If unobserved factors affect the JSY participation, the PSM method will give biased results of treatment effects. However, we assumed that along with observed factors, unobserved factors will also be balanced between treatment and control groups. If the assumptions are true (which cannot be known), there will be no biases due to unobserved factors. DLHS data is very rich. However, there are some limitations in data too. There are multiple missing values in immunization variables. The main reason being many babies had not yet reached the age of needing immunization at the time of the survey. The surveyors do not, therefore, have any information about them. If we drop immunization variables from the outcome/care variable list, the sample size increases to more than 20,000. To gain more observations we have dropped immunizations from our analysis, as the main focus of the study is the continuum of care, which requires data from pregnancy registration to immunization. However, after dropping immunizations we have compared the results of other services, and found that results remain very close to those of the previous sample with immunizations. The other limitation of data is that DLHS-4 was not surveyed in LPSs where the program is universal.