Intervention setting: ReMiND intervention and theory of change
The intervention scenario comprised of routine maternal and child health care services plus the ReMiND intervention. Out of Kaushambi district’s eight community development blocks, the ReMiND was implemented in two blocks: Mooratganj and Manjhanpur. In 2012, mHealth application was implemented through 259 ASHAs in two intervention blocks serving a population of about 300,000 individuals. The ASHAs in ReMiND program were provided with basic Java-based mobile phones operating on an open-source Comm-Care software [
19]. It had a tailored content, which guided the ASHA through the course of a woman’s pregnancy and newborn child care. More specifically, it was used to register the pregnant woman; update her ANC record during subsequent home visits; track her utilization of services from pregnancy into the postpartum period; and track the health of the newborn and the immunization until 2 years of age. They were given extensive trainings on use of mobile phone which also provided audio-visual support to the ASHA workers in order to counsel the pregnant woman at each of these steps [
18,
19].
Data entered about the pregnancy guided ASHAs in providing timely and appropriate health information to pregnant women; and helped them to prioritize home visits. It contained algorithms to assist in the early identification, treatment, and rapid referral for appropriate care of any danger signs among pregnant women or neonates [
18,
19]. Data on services which are due and those utilized by pregnant women, recorded by ASHAs through the mHealth application, were pooled on a common server. The sector facilitators used the data to monitor all the ASHAs working in their area. Data were also shared with the health education officer at the primary health centre during monthly meetings. Thus, the implementing NGO partners—Catholic Relief Services (CRS) and Vatsalya worked in coordination with the district health system to monitor the performance of ASHAs using data generated by the mHealth application.
The purpose of the mHealth application was to improve the quality of counselling by ASHA worker, which in turn was aimed to improve the knowledge of pregnant women. Ultimately, this was intended to generate demand for seeking antenatal and natal services; and for timely care of complications during pregnancy, after delivery and during neonatal period [
19]. The increased utilization of preventive health services, as a result of demand generation and better supply-side monitoring, is likely to result in lower illnesses and as a result reduction in mortality and disability. Similarly, improved care-seeking can also bring about a reduction in fraction of illnesses which are fatal or which result in long-term complications.
Counterfactual: routine care
The routine care scenario comprised of delivery of preventive and curative maternal and child health services, including implementation of the flagship program—National Health Mission [
27]. These comprised of all the set of basic demand and supply side services which are recommended for maternal, newborn, child and adolescent health care as envisaged under the Reproductive, Maternal, Newborn, Child and Adolescent Health Care program (RMNCHA+) [
28]. It was launched in year 2013 to provide a continuum of care right from the start of reproductive age of a girl child to the adolescent health of her offspring. It envisages health system strengthening for providing antenatal care, intra-partum, postpartum care to women for safe maternity, essential new born, early identification and referral services in case of any complication, immunization, prevention and treatment of childhood morbidities, family planning along with interventions for improving physical and psychological health in adolescents. The only difference between the intervention and the control area was the rollout of mHealth application which was used by ASHA workers.
General model overview
A decision tree (Additional file
1: Figure S1) was parameterized on MS-Excel spreadsheet to estimate the incremental cost effectiveness of implementing ReMiND. A time horizon of 10 years starting from base year of 2011 was considered appropriate to cover all costs and effects comprehensively on grounds of intervention characteristics and theory of change for effectiveness mediation. This time horizon was justified based on several reasons. Firstly, the m-health software is unlikely to change in this period as the broad nature of services will remain same. Secondly, based on expert opinion, even if the software has to be edited based on revisions in the program package, such changes are unlikely to have any major cost implications. Thirdly, while several costs of capital nature are incurred during the early years of implementation, however, the consequences of those investments i.e. health benefits continue to occur till many years later. These health effects are likely to occur during pregnancy (such as reduction of high-risk pregnancies, and their early detection and appropriate management), childbirth (such as reduction in post-partum haemorrhage), neonatal (reduction of low-birth weight, and prevention and management of neonatal illnesses), infancy and childhood period (such as prevention of vaccine preventable diseases) up to 5–10 years of age. Finally, economic evaluations of similar m-health packages have also relied on a similar 10–12 year time horizon [
29,
30].
We analyzed costs and effects from both health system and societal perspective. Health system costs included the resources spent by the department of health and the implementing partners in delivering the intervention. These included resources such as building, space, staff salaries, equipment, software for m-health intervention, medicines, consumables, overheads etc. While measuring the societal costs, in addition to the health system cost, we also measured the out-of-pocket expenditures (OOPE) incurred by households. These OOPE were incurred for purchasing medicines, medical or surgical procedures, boarding, lodging, and transportation as a result of any health care sought during the pregnancy, intra-partum care, or neonatal period. We did not include the measurement of indirect costs in terms of productivity loss to the household as a result of absenteeism due to illness. Effect was measured in terms of illness episodes averted, maternal and neonatal deaths prevented, life years gained and DALYs averted. Both costs and effects were discounted at 3% to account for time preference of cost and utility. The choice of discount rate is justified on the following grounds. First, as per World Health Organization’s Choosing Health Interventions for Cost Effectiveness guidelines (WHO-CHOICE), it has been recommended to discount all future costs and consequences at 3% for international comparability [
31]. This is in coherence with the recent guidelines released by Disease Control Priority 3 [
22] and the reference case developed for low middle income countries by International Development Support Initiative and the Gates Foundation [
32]. Also a recent systematic review of the economic evaluations done in India revealed that 82% of the studies which reported the value of discount rate, 3% rate was used to discount future costs and benefits [
26]. To account for uncertainty in value of discount rate, we varied it up to 8% in probabilistic sensitivity analysis. The concept of discounting incorporates both the components i.e. time preference and inflation rate. Time preference represents the opportunity cost of an investment.
We report our findings as incremental cost of implementing ReMiND intervention per DALY averted, per illness episode prevented and per infant death averted as compared to routine care services [
33]. An incremental cost-effectiveness ratio (ICER) is a summary measure in economic evaluations to represent the economic value of an intervention in comparison with an alternative or no alternative (comparator). The ICER is expressed as the ratio of the difference in costs between two strategies to the difference in effectiveness. For preference based outcomes, disability rates were taken from the Global Burden of Disease data [
34,
35].
There are several thresholds which could be used for decision making in a cost-effectiveness analysis [
36]. It could be a supply-side threshold, demand-side threshold or GDP based thresholds. Supply-side threshold is a measure of health benefits forgone due to reduced funding for current interventions as a result of allocating resources for a new intervention from provider’s perspective. A demand-side threshold describes the willingness to pay of an individual to gain additional health benefits in view of other competing demand of his resources. Third, the per capita GDP of a country recommended by several guidelines in the absence of evidence on other threshold measures [
31]. The approach suggested by the commission for Macroeconomics on Health (2001) is that interventions with an incremental cost per DALY averted less than the per capita GDP in low middle income countries (LMICs) are “very cost effective”, and those costing less than triple the per capita GDP are “cost- effective”. In India, till date, there is a scarcity of evidence on supply-side and demand-side thresholds. Hence, per capita GDP is the most commonly used threshold in economic evaluations done in India [
37‐
39].
The standard guidelines for conducting and reporting an economic evaluation survey (CHEERS) were adhered to and details are available as Additional file
2: Appendix S1.
Costing
We analysed the costs from both health system and societal perspective. The health system costs comprised of four distinct components—firstly, it included the cost of implementing the mHealth application, i.e. development of software, training of ASHA workers, mobile phones and data transmission charges etc. [
40]. These costs were obtained in US dollars which were converted into Indian rupees using dollar exchange rates given by Internal Revenue Service for year 2015 (1US$ = INR 63.35) [
41]. The converted rates were then inflated from the year of purchase to the current value of product in 2015 by applying Consumer Price Index in India [
42]. Secondly, we considered the incremental time spent for monitoring and supervision of this additional activity by implementing partners and state health department. Thirdly, introduction of intervention could have resulted in change in the allocation of time for provision of services by ASHA workers and hence affecting the staff costs. However, ASHA workers are not full-time paid staff, and are instead paid a performance-based incentive. As a result, determining the time allocation was not meaningful in this scenario. Instead, we estimated the total performance based payment paid to ASHA workers in intervention and control area to compute the difference. The mHealth application was intended to increase the counseling skills of ASHAs and better understanding of the beneficiaries; we found that there was no significant increase in the utilization of incentive based services which largely includes institutional deliveries. Hence, the incremental costs related to ASHAs incentive were not included in the analysis.
Finally, the intervention could have brought about change in utilization of heath care services, which entails a cost. The benefits reaped as a result of improved knowledge and treatment-seeking in the intervention population were inherent in overall benefits measurement and therefore, we included its associated increase in cost (i.e. cost of increased utilization of healthcare services). As a result, we estimated the cost of delivering extra services—preventive or curative. We used the unit cost of providing mHealth intervention under ReMiND in two blocks of Kaushambi as estimated in the cost analysis—i.e. INR 31.4 (US $ 0.49) per capita and INR 1294 (US $ 20.5) per pregnant woman [
40]. The cost per pregnant woman is most appropriate which incorporate not only population distribution but also fertility levels and thus, allows modeling of costs in a scale up scenario in most appropriate way. The cost per woman of reproductive age captures the age distribution of fertile women in the population but it does not capture the level of fertility and its effect on cost of program implementation. The unit cost capita is also inappropriate for use in the economic evaluation, as it neither captures the effect of population demographics, nor fertility.
The cost of delivering preventive and curative health services at different levels of the health care delivery system in another study from North Indian states was used [
43,
44]. The preventive services included in our analysis were maternal and child health services like provisioning of antenatal care (consumption of iron folic acid tablets, tetanus toxoid vaccine, number of ANC visits), postnatal care, essential newborn care and full immunization till 1 year of age. The curative services included institutional deliveries; treatment of complications during pregnancy, after delivery, among newborn and infants in either outpatient or inpatient setting at various levels of health care facilities. These studies had employed bottom up costing methods to comprehensively estimate the cost of delivering services in a representative sample of sub-centers, primary health centre, community health centre and district hospitals. Unit costs for antenatal care, postnatal care, and immunization were INR 525 (USD 10) per full ANC care, INR 767 (USD 14) per PNC case registered, and INR 97 (USD 1.8) per child immunized in routine immunization respectively [
43]. Similarly, the cost incurred on per outpatient consultation at PHC and CHC was taken as INR 120 (95% CI 90–151) and 126 (95% CI 92–160) respectively while the unit cost per hospitalization was INR 1156 (95% CI 343–2140) at PHC and INR 1115 (95% CI 400–2188) at CHC level [
45]. The cost per OPD consultation and bed day hospitalization for gynaecology (INR 165; 997) and paediatrics (INR 137; 1028) department at district hospital respectively were taken for our model [
46]. The detailed cost analysis is provided in Additional file
3: Appendix S2.
In order to assess the change in utilization of health care services, we analysed the care seeking behaviour of pregnant women for illnesses/complications during delivery and after child-birth. This was assessed based on analysis of a household survey—CEAHH (cost effectiveness analysis household survey) survey, which was used to determine care seeking for illnesses reported in pregnancy, after child-birth and during neonatal period [
18]. The out of pocket expenditure estimates for seeking outpatient and inpatient care for various maternal and childhood illnesses were given in Table
1. Expenditures by households’ in the form of OOPE were included along with health system costs to estimate cost to the society from a societal perspective. Details for the household survey are available elsewhere in the protocol and impact assessment papers [
18,
20].
Table 1
Demographic and epidemiological parameters
Demographic parameters |
Total population of Uttar Pradesh State | 199,812,341 | 169,840,490 | 229,784,192 | Lognormal | Census 2011 |
Birth rate (per 1000 population) | 27.8 | 27.5 | 29.1 | Lognormal | Census 2011 |
Annual decline in birth rate (%) | − 0.04 | − 0.03 | − 0.05 | Lognormal | |
Maternal mortality ratio (per 100,000 live births) | 258 | 241 | 275 | Lognormal | Annual Health Survey (AHS) Report, 2012–2013 |
Neonatal mortality rate (per 1000 live births) | 40 | 23 | 43 | Lognormal | Census 2011 |
Epidemiological parameters |
Prevalence of anaemia in pregnant women | 0.51 | 0.34 | 0.68 | Beta | NFHS-4, 2015–2016 |
Risk of anaemia: for women taking IFA during pregnancy | 0.25 | 0.21 | 0.28 | Beta | |
Risk of anaemia: for women not taking IFA during pregnancy | 0.75 | 0.60 | 0.90 | Beta |
Risk of postpartum haemorrhage (PPH) among anaemic pregnant women | 0.29 | 0.19 | 0.39 | Beta | |
Risk of prematurity among anaemic pregnant women | 0.63 | 0.33 | 1.01 | Beta | |
Risk of low birth weight (LBW) among anaemic pregnant women | 0.31 | 0.13 | 0.51 | Beta |
Probability of maternal mortality with PPH: with treatment | 0.00038 | 0.00029 | 0.00047 | Beta | |
Probability of maternal mortality with PPH: without treatment | 0.00051 | 0.00044 | 0.00058 | Beta |
Probability of neonatal mortality due to prematurity: with treatment | 0.102 | 0.082 | 0.122 | Beta | |
Probability of neonatal mortality due to prematurity: without treatment | 0.332 | 0.266 | 0.398 | Beta |
Probability of neonatal mortality due to LBW: with treatment | 0.047 | 0.038 | 0.056 | Beta |
Probability of neonatal mortality due to LBW: without treatment | 0.113 | 0.090 | 0.136 | Beta |
Prevalence of hypertension (HTN) in pregnancy | 0.07 | 0.047 | 0.093 | Beta | NFHS-4, 2015–2016 |
Risk of preeclampsia in hypertensive pregnant women | 0.63 | 0.422 | 0.838 | Beta | |
Risk of eclampsia in preeclampsia pregnant women | 0.115 | 0.077 | 0.153 | Beta |
Risk of perinatal complications due to eclampsia | 0.524 | 0.351 | 0.697 | Beta | The Magpie Trial 2007 |
Probability of maternal mortality due to eclampsia among pregnant women: with treatment | 0.18 | 0.144 | 0.216 | Beta |
Probability of maternal mortality due to eclampsia among pregnant women: without treatment | 0.4 | 0.32 | 0.48 | Beta |
Probability of neonatal mortality due to perinatal complications: with treatment | 0.102 | 0.082 | 0.122 | Beta | |
Probability of neonatal mortality due to perinatal complications: without treatment | 0.332 | 0.266 | 0.398 | Beta |
Risk of maternal mortality in home deliveries | 0.02 | 0.016 | 0.024 | Beta | |
Risk of maternal mortality in institutional deliveries | 0.00279 | 0.00223 | 0.00335 | Beta |
Prevalence of sepsis in neonates | 0.03 | 0.02 | 0.04 | Beta | National Neonatal Perinatal Database |
Probability of neonatal deaths due to sepsis: with treatment | 0.18 | 0.14 | 0.22 | Beta | |
Probability of neonatal deaths due to sepsis: without treatment | 0.95 | 0.95 | 0.95 | Beta |
Average length of illness (years): anemia | 0.75 | 0.60 | 0.90 | Lognormal | Expert opinion |
Average length of illness (years): PPH | 0.01 | 0.01 | 0.01 | Lognormal |
Average length of illness (years): HTN/eclampsia | 0.75 | 0.60 | 0.90 | Lognormal |
Average length of illness (years): prematurity | 0.03 | 0.02 | 0.03 | Lognormal |
Average length of illness (years): LBW | 0.03 | 0.02 | 0.03 | Lognormal |
Average length of illness (years): sepsis | 0.03 | 0.02 | 0.03 | Lognormal |
Average length of illness (years): perinatal complications | 0.03 | 0.02 | 0.03 | Lognormal |
Impact parameters |
Increase in coverage of 3 ANC visits (%) | 0.00 | 0.00 | 0.00 | Lognormal | ReMiND-impact assessment study |
Increase in coverage of IFA (%) | 12.70 | 8.70 | 16.70 | Lognormal |
Increase in coverage of TT (%) | 0.00 | 0.00 | 0.00 | Lognormal |
Increase in coverage of care seeking (%) | 25.7 | 13.70 | 41.10 | Lognormal |
Increase in coverage of institutional delivery (%) | 0.00 | 0.00 | 0.00 | Lognormal |
Cost parameters (INR) |
Health system costs |
Unit cost: ANC | 525 | 456 | 619 | Gamma | |
Unit cost: PNC | 767 | 538 | 1092 | Gamma |
Unit cost: immunization | 97 | 77 | 120 | Gamma |
Unit cost: institutional delivery | 1872 | 1080 | 2990 | Gamma | Prinja et al. [ 37] (PLOS one) |
Unit cost: PHC |
OPD | 120 | 90 | 151 | Gamma |
IPD | 1156 | 343 | 2140 | Gamma |
Unit cost: CHC |
OPD | 126 | 92 | 160 | Gamma |
IPD | 1115 | 400 | 2188 | Gamma |
Unit cost: gynaecology and obstetrics |
OPD | 165 | 68 | 274 | Gamma | Prinja et al. [ 55] (IJMR) |
IPD | 997 | 592 | 1412 | Gamma |
Unit cost: paediatrics |
OPD | 137 | 102 | 182 | Gamma | Prinja et al. [ 55] (IJMR) |
IPD | 1028 | 444 | 1703 | Gamma |
Out of pocket expenditures (INR) |
Control: public |
ANC | 406 | 305 | 508 | Gamma | Primary data analysis (CEAAH) |
Institutional delivery | 610 | 548 | 672 | Gamma |
Postpartum care | 1216 | 912 | 1520 | Gamma |
Neonatal illness |
OPD | 283 | 29 | 601 | Gamma |
IPD | 2357 | 852 | 6908 | Gamma |
Control: private |
ANC | 845 | 634 | 1056 | Gamma |
Institutional delivery | 13,000 | 11,154 | 14,846 | Gamma |
Postpartum care | 699 | 524 | 874 | Gamma |
Neonatal illness |
OPD | 769 | 472 | 1066 | Gamma |
IPD | 5164 | 3039 | 7289 | Gamma |
Intervention: public |
ANC | 878 | 659 | 1098 | Gamma |
Institutional delivery | 861 | 713 | 1009 | Gamma |
Postpartum care | 450 | 338 | 563 | Gamma |
Neonatal illness |
OPD | 380 | 145 | 615 | Gamma |
IPD | 1399 | 1049 | 1749 | Gamma |
Intervention: private |
ANC | 1420 | 1065 | 1775 | Gamma |
Institutional delivery | 16,900 | 11,051 | 22,749 | Gamma |
Postpartum care | 1791 | 1343 | 2239 | Gamma |
Neonatal illness |
OPD | 800 | 500 | 1100 | Gamma |
IPD | 1000 | 750 | 1250 | Gamma |
Valuing consequences of ReMiND intervention
ReMiND intervention was intended to improve the quality of counselling of pregnant woman by the ASHA worker. Improved counselling was desired to improve knowledge of pregnant women, and utilization of appropriate maternal and child health care services, during pregnancy, child-birth and during the neonatal period. Secondly, the data entered by the ASHA worker helped in tracking the ASHA worker to track pregnant women and their services utilized; besides being used for supervision of ASHA performance [
18].
We undertook a pre and post quasi experimental study to assess the impact of the intervention. Two blocks other than two intervention blocks were selected as controls after matching for coverage of two indicators at baseline—ante natal care and institutional deliveries from the same district. The pre-intervention data was obtained from the Annual Health Survey 2011 conducted by Ministry of Health and Family Welfare. A household survey was carried out in four blocks of Kaushambi district in year 2015 to observe the post intervention coverage. Propensity score matched sample from intervention and control areas in pre-intervention and post-intervention periods were analysed using difference-in-difference method to estimate the impact of ReMiND program. Overall, the ReMiND led to a statistically significant increase in coverage for IFA consumption (12.7%), abdominal examination during ANC care (18.7%), identification and self-reporting of complication during pregnancy (13.20%) and after (19.5%) delivery and care seeking (25.7%) in the intervention area [
20,
56]. The coverage of three or more ANC visits, tetanus toxoid vaccination, full ANC care and ambulance usage also increased in intervention area by 10.3, 4.3, 1 and 2.5%; however, the difference between the improvements in the intervention and control area was not statistically significant [
20] (Table
1).
The coverage of MNCH services in control area from baseline and end-line surveys was used to interpolate the coverage during intervening years, and extrapolate during the future years from 2015 onwards. Linear change was assumed for the purpose of modelling. In case the coverage for any indicator reached 90%, no further increase was assumed thereafter in the subsequent years. Similarly, the impact estimates of difference in difference for intervention area were used to compute annual rate of change in the intervention area, which was further used to model the coverage of MNCH services in intervention scenario, relative to the counterfactual.
In this paper, we model the effect of increased utilization of health care services on reduction in illnesses or complications during pregnancy and after child-birth. Together, these two contributed to reduction in maternal and neonatal deaths—ultimately resulting in averting years of life lost (YLL) to premature mortality and reduction of disability adjusted life years (DALY). In terms of maternal complications, we primarily modelled the effect of changes in antenatal services on two major illnesses during pregnancy—anaemia and hypertension. These two were particularly considered in view of their prevalence in the targeted population [
57], as well as evidence linking reduction in occurrence of these medical conditions with better ANC care [
58]. Baseline prevalence of 51.4 and 5.8% was assumed for anaemia and hypertension during pregnancy [
59]. Subsequently, we modelled the effect of improvement in coverage of complete IFA supplementation as a result of mHealth intervention—risk of anaemia is 75 and 25% without and with IFA supplementation respectively [
47]. Similarly, an 11.5% risk of developing eclampsia was assumed among hypertensive pregnant women [
51]. In turn, we assumed that anaemia results in complications during pregnancy and after child-birth, such as post-partum haemorrhage (29%), and is also associated with adverse neonatal health outcomes such as prematurity (63%) and low birth weight (31%) [
49]. Finally, reduction of post-partum haemorrhage is associated with reduced maternal mortality [
60]; while both prematurity and low birth weight results in a higher risk of neonatal mortality [
50] (Table
1).
Secondly, the ReMiND intervention resulted in improved recognition of the danger signs during pregnancy and after child birth. The care seeking for any illness during pregnancy was higher in the intervention area (71.9%) as compared to control are (46.2%). We modelled the impact of improved care seeking on maternal and neonatal survival. For example, the risk of maternal mortality with post-partum haemorrhage is 25% less with treatment than without [
60,
48].
Besides an increase in YLL as a result of reduction in mortality, we also estimated the reduction in years of life lived in disability (YLD) as a result of reduced illnesses during pregnancy, after child-birth and during neonatal period. We used the disability weights as provided in the Global Burden of Disease, 2010, for computing YLD [
61]. For calculating YLL in case of an infant death, we estimated that the mean age of infant death is 26 days. This estimation was based on the assumption that 60% of infant deaths occur in neonatal period, 60% of neonatal deaths are early neonatal deaths (within first 7 days of birth) [
62‐
64]. We also assumed that mean age of early neonatal, late neonatal and post-neonatal death is 3, 20 days and 6 months respectively; i.e. the mid-point of class interval. Similar assumptions have also been used by another study evaluating cost effectiveness of IMNCI program in India [
37]. We computed the percentage reduction in maternal and neonatal deaths by comparing the number of deaths in intervention where ReMiND program was implemented relative to control using a decision tree model (Additional file
1: Figure S1). Table
1 cites the estimates of effectiveness on proximal outputs i.e. coverage of services, and assumptions for modeling impact on long term outcomes such as morbidity and mortality derived from the Indian studies.
Sensitivity analysis
We undertook a probabilistic sensitivity analysis to test the effect of parameter uncertainty on the findings of the analysis and to estimate the effect of joint uncertainty in all parameters [
65‐
67]. While base analysis is valid for UP state, there is significant variability in values for various parameters from Indian perspective which was important to test in the PSA analysis. For several parameters related to unit cost of health care services, effectiveness estimates of mHealth interventions, some demographic parameters, and service coverage in counterfactual scenario etc., 95% confidence intervals were available from primary analysis as part of the current study or secondary literature [
40,
44,
45,
54,
55,
68]. For other parameters, such as demographics and epidemiological parameters such as risk of various morbidities with or without use of preventive interventions etc., we varied the base estimate obtained from literature 20% on either side. For certain parameters, such as risk of mortality with and without treatment we varied the base estimate by 50% on either side, since this is heavily dependent on other supply side inputs which could vary significantly across different parts of the country. In case of prevalence of risk factors such as anaemia and hypertension during pregnancy, or low birth weight babies, we varied the base parameter by 33% on either side. Finally, we varied the cost of mHealth intervention by 50% on lower side and 20% on higher side. The same was done as we expect that implementation of mHealth intervention through support from a donor partner would be higher than when it is implemented through public sector health system which has relatively lower salary structures.
Probability of ReMiND intervention to remain cost effective at a willingness to pay threshold equal to per capita gross domestic product (GDP) was estimated, using a health system perspective. For undertaking PSA analysis, we assumed a lognormal distribution for unit costs. In case of parameters where 95% confidence interval was available, a beta distribution was used; while uniform distribution was applied where an upper and lower bound were available. Monte Carlo method was used for simulating the results 999 times. Median was computed along with 2.5th and 97.5th percentile to estimate 95% confidence interval.