Policy Research Working Paper 9783 Drivers of Utilization, Quality of Care, and RMNCH-N Services in Bangladesh A Comparative Analysis of Demand and Supply-Side Determinants Using Machine Learning for Investment Decision-Making Saji Gopalan Rianna Mohammed-Roberts Hellen Chrystine Zanetti Matarazzo Health, Nutrition and Population Global Practice September 2021 Policy Research Working Paper 9783 Abstract Amid noticeable improvements and achievements in the rounds of the Bangladesh Health Facility Survey and the reproductive, maternal, neonatal, child health, and nutri- Demographic and Health Survey (2014 and 2017) were tion landscape in Bangladesh, existing evidence suggests that analyzed. The findings indicate that the relative impor- further accelerated progress hinges on strategic investment tance of the demand-side determinants (except wealth decision making. Addressing the top service utilization and education status) have recently declined. Conversely, determinants that are both context- and time-specific is investments in key supply-side determinants (for example, one cost-effective way of improving the unmet reproductive, availability of skilled staff, readiness for care, and quality maternal, neonatal, child health, and nutrition outcomes in of care) could provide a thrust toward further increases in a short timeframe. Against this backdrop, using machine utilization. Immediate attention is needed to address the learning analysis, the overall aim of this study was to help regressive role of wealth status on utilization through, for Bangladesh identify priority investment areas that could example, demand-side financing that goes beyond user fee accelerate reproductive, maternal, neonatal, child health, exemptions. Further, developing strategies to improve the and nutrition utilization, quality, and outcomes over engagement of community health workers in reproductive, the short run, by comparing the relative importance of maternal, neonatal, child health, and nutrition utilization demand- and-supply-side determinants of key reproductive, and tapping into the potential of mobile health technology maternal, neonatal, child health, and nutrition indica- to support community health workers’ performance and tors over the past decade (across two time points). Two women’s awareness could help to boost utilization patterns. This paper is a product of the Health, Nutrition and Population Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at sajisaraswathyg@gmail.com, rmohammed@worldbank.org, and hzmatarazarro@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Drivers of Utilization, Quality of Care, and RMNCH-N Services in Bangladesh A comparative analysis of demand and supply-side determinants using machine learning for investment decision-making Saji Gopalan Rianna Mohammed-Roberts Hellen Chrystine Zanetti Matarazzo JEL: I10, I12, I14, I19 Keywords: maternal and child health, family planning, nutrition, utilization determinants, investment decision-making, machine learning, Bangladesh Acknowledgments This paper was prepared by a World Bank team led by Rianna Mohammed-Roberts (Senior Health Specialist), Saji Gopalan (Senior Economist Consultant), and Hellen Chrystine Zanetti Matarazzo (Data Scientist Consultant), under the overall supervision of E. Gail Richardson (Practice Manager). Insightful comments, suggestions and guidance are gratefully acknowledged from Mersedeh Tariverdi (Senior Data Scientist), Manuela Villar Uribe (Health Specialist) and Asib Nasim (Senior Health Specialist Consultant). Any errors or omissions are entirely the responsibility of the authors. 1. Introduction 1.1 Why is investing in priority determinants important for RMNCH-N utilization and outcomes in low- and middle-income countries? Globally, the performance trajectory of countries in improving reproductive, maternal, neonatal, child health, and nutrition (RMNCH-N) outcomes over the last few decades highlights a number of exemplary pathways.1–5 First, global evidence shows that timely and appropriate RMNCH-N service utilization is the foremost key to improving the RMNCH-N outcomes.5 Specifically, countries that invested in enhancing utilization have achieved further improvements in RMNCH-N outcomes. A second pivotal step in improving RMNCH-N service utilization is by addressing its determinants. Third, both the demand- and supply-side determinants are equally relevant in boosting RMNCH-N service utilization in a timely manner. The relative importance of the various determinants, however, changes at different phases of the progress trajectory. Fourth, the determinants of RMNCH-N service utilization and outcomes are context-specific (i.e., country or region-specific) and time-specific.1 For instance, certain demand-side determinants (e.g., cultural factors) are more dominant in the early stages of progress and sometimes, new determinants may arise (e.g., mass media influence). However, it is difficult to predict the role and magnitude of such determinants unless they are regularly assessed. Fifth, global evidence shows that countries can make more rapid progress when they make specific investments in priority determinants of RMNCH-N service utilization, rather than relying on generic strategies and interventions. For example, a recent global assessment has shown that interventions specifically addressing the priority (i.e., context and time-specific) RMNCH-N service determinants could reduce the fertility rate to 2.5 in certain Asian settings compared to generic interventions.1 Another study of diarrhea and pneumonia interventions observed that 15 highly cost- effective interventions addressing the top RMNCH-N service utilization determinants – if implemented at scale – would prevent 95 percent of deaths from diarrhea and 67 percent of deaths from pneumonia in under-5 children by 2025.6 Such specific interventions addressing priority utilization determinants are also proven to be cost effective in improving the RMNCH-N outcomes in a short timeframe. In summary, the existing evidence shows that regular tracking, coupled with prioritized investments in context-specific utilization determinants, can improve RMNCH-N service utilization. 1.2 Reproductive, maternal, child health, and nutrition landscape in Bangladesh Bangladesh has witnessed significant improvements across key maternal, child health and nutrition indicators over the last few years, with multi-faceted and sector-wide approaches to improving the health of pregnant women and under-five children.7 Changing fertility behavior, and increased access to and utilization of health services, for example, have been major contributors to the consistent decline in under- five and maternal mortalities.8 Notwithstanding progress, however, Bangladesh is still far away from reaching a key Sustainable Development Goal (SDG) – “maternal mortality ratio target is 70 deaths per 100,000 live births” (currently at 165).8 The uptake of certain key maternal and child health (MCH) services has been consistently sub-optimal. For instance, institutional deliveries are only 50% and skilled birth attendance is at 53%.9 As per verbal autopsies and household surveys, two-thirds of total maternal deaths occur during the postpartum period, but postnatal care visits are still low at 65%.10 There are also persistent inequities in service utilization and health status among different socio-economic groups. Health indicators and service utilization, for example, are relatively better among the rich, urban and educated groups compared to their counterparts.11 For 1 instance, under-5 mortality was 55 (out of 1,000 live births) among the lowest quintile compared to 36 among the highest, and perinatal mortality was 52 and 39 respectively in 2017. Women in urban areas were more likely to receive quality ANC services than those in rural areas (27% versus 14%) in 2017, and more women with a secondary-level education or higher (33%) reported receiving quality ANC services compared to less-educated women (6%). Only 7% of women from the lowest wealth quintile received quality ANC, compared to 37% from the highest wealth quintile. Similarly, only 21.4% of rural women received quality ANC compared to urban counterparts (34.6%). The status of key RMNCH-N indicators are summarized in Table 1. Table 1: Key reproductive, maternal, child and nutrition indicators Indicators Value Maternal mortality ratio per 100,000 live births (2017) 173 Neonatal Mortality Rate per 1,000 live births (2019) 15 Infant Mortality Rate per 1,000 live births (2019) 34 Under-five Mortality Rate per 1,000 live births (2019) 40 Under-five stunting prevalence moderate & severe (2019) 31 Under-five underweight prevalence moderate & severe (2019) 22.6 Under-five wasting prevalence moderate & severe (2019) 9.8 Under-five overweight prevalence moderate & severe (2019) 2.4 Early initiation of breastfeeding (2019) 46.6% Exclusive breastfeeding up to six months (2019) 62.6% Children weighed at birth (2019) 51.9% Delayed bathing (> 24 hours after birth) (2019) 80% Under-five received all vaccination (2017) 86% Care-seeking for fever (2019) 55.6% Contraceptive prevalence rate (2019) 70.3% 4+ antenatal care visits by any provider (2019) 36.9% Institutional deliveries (2019) 53% Skilled birth attendance (2019) 59% Postnatal health check for the mother at least one (2019) 65% Contraceptive prevalence rate among 15-49 age women 62% Sources: DHS201712; MICS 20199; The World Bank 201713 1.3 Context for the study With considerable investments and noticeable improvements, Bangladesh currently stands at a crucial phase in its RMNCH-N landscape. In order to accelerate achievement of improved RMNCH-N outcomes, global evidence suggests the need for smart and strategic health sector investments and interventions.1 Further, global evidence also shows that the path to achieving the top RMNCH-N milestones (e.g. SDGs) is more complex and difficult for countries, compared to the earlier milestones (e.g. millennium development goals).1 Consequently, for Bangladesh, strategic decision-making and investments are necessary.14 It is already known that investments targeting appropriate RMNCH-N utilization would lead to improved RMNCH-N outcomes.15,16 However, the success of such investments largely depends on how efficiently and timely (i.e. immediate to short-term) the service utilization determinants are addressed.14 While there could be several determinants of RMNCH-N utilization, for example, addressing the key ones could be one of the most strategic approaches to ensuring timely progress in the short-term. For instance, while it is known that the quality of maternal, neonatal and child health is low in many countries, it is also known that improving the quality of care would reduce maternal and neonatal deaths. Current global evidence, however, does not establish that improving the quality of care alone would necessarily motivate women to 2 seek timely care. This is because RMNCH-N care-seeking also depends on certain demand-side factors, and utilization patterns are largely context- and time-specific. This situation necessitates that the determinants of RMNCH-N utilization be identified and addressed to ensure prioritized investment decision-making. In Bangladesh, the evidence-base is inconclusive on the demand- and supply-side determinants that should be addressed urgently for improving RMNCH-N utilization. While the prevailing evidence largely points to socio-economic, cultural and regional factors as the key drivers of RMNCH-N utilization,8 little is known about the role of service delivery status and quality of care in RMNCH-N uptake. Similarly, the relative significance of supply-side determinants over the demand-side determinants is not clear. For instance, if quality of care is the key determinant of service uptake, improving quality versus sizably investing in addressing certain demand-side determinants (e.g., cultural barriers) would be an immediate priority. While most existing studies have used DHS data and demand-side determinants alone to examine the patterns of RMNCH-N utilization, assessments which combine both demand- and supply-side determinants of RMNCH-N utilization, are limited. It is against this backdrop that the World Bank, in collaboration with the Ministry of Health and Family Welfare (MoHFW), carried out this technical assistance study to ascertain the most important determinants that need to be addressed, on a priority basis, to improve RMNCH-N outcomes further. A comparative assessment of the drivers of RMNCH-N performance across time periods (2014 and 2017 surveys) reflects the success, as well as gaps, in the government strategies and interventions for further policy navigation. Study findings would also be influential to RMNCH-N policy making in the post-COVID-19 context. 1.4 Aim The overall aim of this study was to inform strategic RMNCH-N investment decision-making, through a three-pronged approach that involved: 1. Identifying the top ten determinants that need to be addressed on a priority basis to improve RMNCH-N utilization, quality of care and outcomes nationally and for divisions (or states) 2. Comparing the relative importance of demand- and-supply-side determinants on key RMNCH-N indicators in the last decade (across two time points) in Bangladesh 3. Proposing key strategic steps for improving the RMNCH-N utilization and quality in the country. This study used machine learning analysis as the selected data sources (given below) were high-dimensional and large. 2.Methodology 2.1 Data sources Two sources of secondary data were used in this study: the Bangladesh Health facility Survey (BHFS) and the Demographic and Health Survey (DHS). Two rounds (2014 and 2017) of BHFS and DHS survey data were used. Data sources and timelines were selected considering the need for a comparative analysis across two time points using multiple rounds of household and facility surveys. Both DHS and BHFS cover similar timeframes, enabling such a comparison. Averages of BFHS indicators by district and location (urban- rural) were merged with the DHS data. 3 BHFS BHFS is a nationally representative survey of registered health facilities; it includes public facilities, NGO static clinics/hospitals, and private for-profit hospitals. The BHFS 2014 included 1,596 facilities,17 while the BHFS 2017 included 1,600 facilities.18 The BHFS provides information on the availability of basic and essential health care services, and the readiness of health facilities for care provision. It uses standardized questionnaires from the service provision assessment (SPA) component of the Demographic and Health Surveys (DHS) Program. Specifically, it provides information on child health, maternal and newborn care and family planning. For each of these services, BHFS assesses whether necessary elements are present and functioning. DHS DHS is a nationally-representative survey of households and ever-married women aged 15-49 years.12 The DHS 2014 covered 17,300 households and 17,863 women,17 while the 2017 survey covered 19,457 households and 20,127 women.12 DHS collects information on health and living conditions of women and children, household assets and wealth status, socio-economic and demographic characteristics, maternal and child health status, nutrition, fertility, family planning, mortality and care seeking for maternal and child health. This study only considered the RMNCH-N care-seeking, perceptions and health outcomes for the most-recent childbirth and child health episode from each DHS survey. 2.2 Variables Nineteen outcome variables (Table 2) related to service utilization, quality of care and nutrition status were selected from the DHS survey. Demand-side explanatory variables (or potential determinants) selected from the DHS were socio-economic and demographic characteristics and location of service seekers (Table 3). Supply-side explanatory variables chosen from the BFHS and the DHS include – type of provider, quality of care, availability of services (infrastructure, drugs, equipment and skilled staff) and readiness for care provision (Table 4). For the BFHS data, where possible, similar variables were combined and aggregated to create indices (e.g., medical equipment index and drug availability index) to reduce the variance of similar variables based on the categorization or criteria of indexing provided by the BFHS survey (more details can be found in BFHS reports). For utilization of reproductive and maternal care and quality of ANC, the unit of analysis was an individual woman. The unit of analysis for child related indicators was a child under-5. All study variables were selected based on the contextual literature and compatibility for a comparative analysis across time points.16,19–22 Table 2: Outcome variables Domain Indicator Utilization Antenatal care (ANC) • 4+ ANC visits • ANC from medically trained provider • ANC in public sector Delivery • Skilled birth attendance (i.e., delivery assisted by skilled staff such as doctor, nurse or midwife) • Delivery in public sector Postnatal care (PNC) • Availed at least one PNC (for mother) within 42days 4 • PNC for mother received from a medically trained provider Family Planning (FP) • Use of any FP methods among married women (15- 49 years) • Use of Modern FP methods among married women (15- 49 years) Vaccination coverage • 12–23 months old children receiving all basic vaccinations (i.e., BCG, three doses of pentavalent (DPT-HepB-Hib), three doses of oral polio vaccine, and one dose of measles vaccine) Care-seeking for sick under-5 children • Children with diarrhea treated at a facility • Children with Acute Respiratory Infection (ARI) treated at a facility • Children with fever treated at a facility Quality of care 1 Adequate ANC quality (adequate quality indicates – a woman reported of receiving all given below services in one visit) • Weight measured during ANC • BP measured during ANC • Urine sample taken for ANC • Blood sample taken during ANC • Ultra-sonogram conducted during ANC • Informed of signs of pregnancy complications during ANC Adequate quality for newborn PNC (adequate quality indicates – a woman reported of her baby receiving all given below services in one visit). • Examined the cord • Counselled on danger signs • Assessed temperature • Counselled mother on feeding • Observed feeding • Assessed weight Adequate essential newborn care at birth (adequacy indicates – a woman reported of her baby receiving all given below services at birth). • use of a safe delivery kit/ bag or boiled blade • nothing applied to the cord or only chlorhexidine applied • after the cord is cut and tied, cord dried within 5 minutes after birth • bathing delayed 72 hours or more • immediate breastfeeding Health/nutrition Under-five Child nutrition outcomes • Stunting (i.e., more than 2 standard deviations below the median for height-for-age) • Wasting (i.e., more than 2 standard deviations below the median for weight-for-height) • Underweight (i.e., less than 2 standard deviations below the median for weight-for-age) Source: DHS Table 3: Demand-side determinants* Variables Household access to improved sanitation facilities Household access to improved drinking water Household access to hand washing Household wealth status in quintile Age of mother at childbirth Age at marriage Education status Employment status 1 Quality of care indices consisted of equally weighted averages of their component variables. 5 Marital status Mass media access (access to at least one media once a week out of radio, television and newspaper) Women empowerment in health index (i.e., earning woman able to spend money as they wish, and able to participate in the decision-making for the health care of herself and child) Residence (rural/urban) Division Women aware of community clinic nearby residence Gender of sick child Birth order Exclusive breast feeding up to six months Initiation of breastfeeding within 1 hour of birth Child’s weight at birth Under-5 received curative care at a facility for diarrhea or fever or acute respiratory infection Source: DHS; *represents mothers who delivered in the last five years preceding the survey Table 4: Supply-side determinants Variables Definition Proportion of amenities available out of - national electricity grid, regular electricity, improved water source, visual and auditory privacy, Basic amenities index functioning improved client latrine, computer with internet, emergency transport, and functional communication equipment in a facility. Proportion of equipment available out of - child scale, adult scale, infant Basic equipment index scale, thermometer, blood pressure apparatus, stethoscope and light source in a facility. Proportion of lab tests available out of - tests for hemoglobin, blood Diagnostic laboratory capacity index glucose, urine protein, urine glucose, stool microscopy, urine pregnancy test and functional microscope in a facility. Standard precautions for infection Proportion of amenities available out of - necessary guidelines, equipment control index 2 and supplies for infection control in a facility. Proportion of diagnostic services available out of - diagnostic capacity for Diagnostic capacity for antenatal hemoglobin, urine protein, urine glucose, blood grouping and Rhesus care services index factor and syphilis in a facility. Medicines for routine antenatal care Proportion of medicines available out of - iron tablets, folic acid tablets, index combined iron and folic acid and iron or folic acid tablets. Essential medicines and commodities Proportion of medicines and commodities available out of - essential and for normal delivery index 3 priority medicines and commodities for normal delivery in a facility. Essential Medicines for newborn Proportion of medicines available out of - essential medicines for newborn care index care in a facility. Essential and priority medicines and Proportion of medicines and commodities available out of - essential and commodities for childcare index priority medicines and commodities for childcare in a facility. Proportion of commodities available out of - oral pills, injectable, intra Family planning commodities index uterine contraceptive device and condoms in a facility. 2 Standard infection control measures include- sterilization equipment; equipment for high level disinfection and safe final disposal of sharps waste and infectious waste; appropriate storage of sharps waste and infectious waste; disinfectant, syringes, and needles; soap, running water, alcohol-based hand disinfectant, latex gloves, medical masks, gowns, eye protection and guidelines for standard precautions. 3 Essential medicines and commodities for delivery care include - injectable uterotonic (oxytocin), injectable antibiotic, injectable magnesium sulphate, injectable diazepam, skin disinfectant, intravenous fluids with infusion set, sodium chloride injectable solution, injectable calcium gluconate, ampicillin powder for injection, injectable metronidazole, misoprostol capsules or tablets, azithromycin capsules or tablets or oral liquid, cefixime capsules or tablets, benzathine benzyl penicillin powder for injection, injectable bethamethasone/ dexamethasone, nifedipine capsules or tablets. 6 Proportion of amenities available out of - ANC guidelines, staff trained in Readiness for antenatal care services ANC, equipment, hemoglobin and urine protein testing capacity, and drugs in a facility. Readiness for normal delivery Proportion of amenities available out of - guidelines, staff trained in services delivery, equipment, and drugs in a facility. Proportion of amenities available out of – integrated management of Readiness for child curative care childhood illness (IMCI) guidelines, IMCI staff, growth chart, equipment services and drugs in a facility. Readiness for family planning Proportion of amenities available out of - guidelines, staff trained in FP, service equipment and supplies in a facility. Staff trained in ANC Facilities with staff trained in ANC Staff trained in delivery care Facilities with staff trained in normal delivery Staff trained in IMCI Facilities with staff trained in IMCI Staff trained in growth monitoring Facilities with staff trained in growth monitoring Adequate quality for ANC (adequate quality indicates - a woman reported of receiving all given below services in one visit). • Weight measured during ANC • BP measured during ANC • Urine sample taken for ANC • Blood sample taken during ANC • Ultra-sonogram conducted during ANC • Informed of signs of pregnancy complications during ANC Adequate quality for newborn PNC (adequate quality indicates - a woman reported of her baby receiving all given below services in one visit). • Examined the cord • Counsel on danger signs Quality of care* • Assessed temperature (reported by recent mothers) • Counselled mother on feeding • Observed feeding • Assessed weight • Height for age • Weight for age Adequate quality for essential newborn care at birth (adequacy indicates - a woman reported of her baby receiving all given below services at birth). • Use of a safe delivery kit/ bag or boiled blade • Nothing applied to the cord or only chlorhexidine applied • After the cord is cut and tied, cord dried within 5 minutes after birth • Bathing delayed 72 hours or more • Immediate breastfeeding Type of provider (public, private for- Providers of RMNCH-N care as reported by recent mothers profit, NGO) * ANC by medically trained provider* Reported by recent mothers PNC for mothers by medically Reported by recent mothers trained provider* ANC 4+*(completed at least 4 ANC Reported by recent mothers visits) Skilled birth attendance * Reported by recent mothers Institutional delivery* Reported by recent mothers # Married woman aged 15-49 years NGO workers include paid and unpaid community health workers (CHW) visited by NGO worker* deployed by the NGOs # Married woman aged 15-49 years Government FP worker is a CHW and includes - Family Welfare visited by government FP worker * Assistants (FWA) deployed by the Directorate General of Family Planning (DGFP) 7 # Married woman aged 15-49 years Government health worker is a CHW and includes - Health Assistants visited by government health (HA) deployed by the Directorate General of Health Services (DGHS) worker* Availability of contraceptives Availability of any modern contraceptive at facilities Sources: BHFS & DHS; * relevant only for a few outcome indicators; # represents community health workers (CHW) 2.3 Analysis Why machine learning analysis? This study applied the machine learning technique using R programming language.23 Machine learning was found to be useful for this comparative analysis since it allowed large demand- and supply-side data sets to be combined. Machine learning was applied over conventional analysis due to the following specific advantages that it offers: 1) it helps to identify the trends and patterns in large data sets easily, and with minimal human biases; and 2) machine-learning algorithms continuously learn from data to help addressing the potential errors in the data. This increases the efficiency of the analysis and the accuracy and quality of predicted outcomes or trends.24 Machine learning consists of a group of artificial intelligence techniques, in which the algorithms learn the patterns in the data without being explicitly programmed to carry out specific applications.24 In this study, a gradient boosting algorithm was used to predict the outcomes based on the predictors listed in Tables 3 and 4. Gradient boosting combines shallow and successive decision trees.25 Decision trees consist of recursively partitioning (also known as splitting) predictors. The goal of each tree in our model was to maximize the accuracy of prediction. Each decision tree learns successively and improves upon the previous (learning rate). Eventually, predictions are based on a weighted combination of these trees.26 The pre- processed data (variable scaling and removal of highly correlated predictors) was split into training (80%) and test segments (20%) for all outcome variables. First, the algorithm was trained on the training segment and then validated on the test segment to determine predictions. The data was 10-fold cross-validated, with the data split into 80% training and 20% test observations randomly, ten times for all algorithms. The average of the cross-validations was taken as the final result. The relative influence of the predictors with the output was estimated using mean decrease Gini impurity coefficient (MDG) in the gradient boosting algorithm.25 MDG is the average across all trees of the decrease in Gini impurity for a predictor.25 The mean decrease in Gini coefficient is a measure of how each variable contributes to the homogeneity of the nodes and leaves in the resulting hierarchy of decision trees. A higher MDG value shows that the degree of impurity arising from a variable category could be reduced farthest by one variable, and therefore suggests an important associated factor.25 The ranking of the top 10 predictors was plotted based on the MDG. Predictors with a higher relative importance value are more likely to affect the outcomes than with the lower values. For instance, the predictor on the top of the relative importance chart (given in the results) has a relatively higher relative importance value and the highest influence on the outcome indicator among all predictors considered. Similarly, the predictor on the bottom of the relative importance chart (given in the results) has the least relative importance value and the least influence on the outcome among all predictors considered. 3. Results 8 3.1 ANC (ANC 4+visits, ANC by medically trained provider and ANC in the public sector) For all three ANC indicators – the level of utilization consistently across two time points (2014 and 2017) was directly related to reported ANC quality among pregnant women, wealth status and media access (Figures 1- 3). These results indicate that the level of ANC utilization was higher for richer women, those having a higher media access, and those who reported that facilities provide a better quality of ANC. Compared to 2014, the relative role of education status was less in 2017 (except for 4+ANC). The chance of completing 4+ANC was higher among more educated women across years. However, women irrespective of their education level had a high chance of receiving ANC from medically trained staff. Meanwhile, the chance of availing ANC in public facilities was higher among relatively better educated women. The relative importance of residence and divisions declined in 2017, indicating that inequalities in ANC utilization based on rural-urban status and divisions, have reduced in recent years. The demand-side determinants (e.g., handwashing access and age at marriage) had a higher influence on ANC usage than supply-side determinants in 2014. This shows that while women who had access to handwashing facilities in their households had a better chance of availing ANC earlier, this trend was reversed in 2017. The relative role of supply-side determinants (e.g., quality of care, availability of staff trained in ANC, amenities and equipment) became stronger in 2017, reflecting the emerging role of supply-side determinants in influencing improvements in ANC utilization in recent times. 4+ANC visits - the role of both public and NGO facilities became one of the top determinants of completing 4+ ANC in 2017 (Figure 1), indicating that women who availed ANC from these facilities had a higher chance of completing 4+ ANC visits. These results reflect the progressive influence of public and NGO facilities on 4+ ANC. Similarly, the role of supply-side determinants was stronger in 2017. For instance, the availability of staff trained in ANC became a top determinant (4th position) and readiness of facilities for ANC became the 7th most important determinant. ANC by a medically trained provider - the role of demand-side determinants (e.g., education and age at marriage) decreased in 2017, while wealth remained a top determinant in both 2014 and 2017 (Figure 2). The role of supply-side determinants increased in 2017. For instance, the availability of staff trained in ANC and ANC medicines emerged as top determinants, indicating that they could both positively increase the chance of availing ANC from a medically trained staff. The role of CHWs (i.e., NGO worker and government health worker) also became a top determinant, reflecting their positive influence on encouraging women to seek ANC from medically trained staff. ANC in the public facilities - the importance of supply-side determinants increased in 2017 (Figure 3). Findings show that women considered the availability of medically trained staff and quality in choosing the public sector. Women who were aware of the presence of community clinics in their locality had a higher chance of availing ANC from the public sector. The visits of CHWs (i.e., NGO worker and government health worker) were also the top determinants, positively influencing the use of the public facilities for ANC. 9 Figure 1: Top determinants of ANC 4+ visits 4,5 Figure 2: Top determinants of ANC by medically trained provider 4 Determinants are defined as follows: (1) ancquality - women reported specific procedures during the ANC care (weight and BP measured; urine and blood sample taken; ultra-sonogram conducted; and informed of signs of pregnancy complications); (2) wealth - wealth status in quintile for women; (3) ancPrivate - ANC received in private for-profit; (4) equipment_basic - basic equipment available for ANC in the facility; (5) residence - rural or urban; (6) mediaAccess - access to at least one media once a week; (7) amenities-basic - basic amenities such as electricity in a facility; (8) labtests - generic laboratory services available in a facility; (9) medicines- anc - availability of medicines for ANC in a facility; and (10) ancNGO - ANC received in NGO sector. 5 For this, and all other figures in this study, determinants on the top indicate higher influence on the outcome. The figure also shows the comparative importance of determinants between 2014 and 2017. 10 Figure 3: Top determinants of any ANC in public sector 6 3.2 Delivery (SBA and delivery in the public sector) For the two delivery indicators – the top determinants consistent across 2014 and 2017 were reported ANC quality, wealth and education status (Figures 4 and 5). Findings reflect that the delivery indicators were equally influenced by both demand- and supply-side determinants. This finding is different from that for ANC indicators (as shown above), which showed an emergence and prominence of supply-side determinants in 2017. Skilled birth attendance - women getting ANC from a medically trained provider and completing 4+ ANC visits had a higher chance of SBA (Figure 4) across years. Availing ANC from a private for-profit provider increased the chance of SBA slightly better than having ANC in a public health facility. Wealth and education were consistently the top determinants across years, indicating that the utilization of SBA has been regressive in the country, as wealthier and more educated women have a better chance of SBA. Delivery in public facilities - supply-side determinants were consistently the top determinants in 2014 and 2017 (Figure 5). Top determinants that could positively influence the chance of delivery in the public facilities were – availing any ANC in public sector, reported quality of ANC and readiness of facilities for delivery. 6 Determinants are defined as follows: (1) labtests - generic laboratory services available in a facility; (2) diagnosis_anc - diagnostic capacity for ANC in a facility; (3) age - age at birth for women; (4) ageMarriage - the age at marriage; (5) anc4plus– women completed 4 or more visits; (6) medicines-anc– essential medicines for ANC available in a facility; (7) staff_ anc – facilities with staff trained in ANC; and (8) awareComClinic– women aware of community clinics nearby residence. 11 Figure 4: Top determinants of Skilled Birth Attendance Figure 5: Top determinants of institutional delivery in public sector 7 7 The determinant - Staff_delivery- indicates facilities with staff trained in delivery. 12 3.3 Postnatal care for mothers (PNC within 42 days and PNC by medically trained provider) For the two PNC indicators – SBA was the top determinant in both 2014 and 2017. This indicates that women having SBA had a better chance of availing PNC within 42 days of delivery and receiving PNC from a medically trained provider (Figures 6 and 7). PNC for mother – the relative importance of supply-side determinants increased in 2017 (Figure 6). For example, the readiness of facilities for delivery and the availability of lab tests at facilities became the top determinants. These results indicate that facilities with such amenities increased the possibility of PNC utilization for mothers. In 2017, the chance of PNC was higher among women who gave birth in public facilities. The role of the private for-profit and NGO sectors, however, was not visible in both 2014 and 2017. The relative role of demand-side determinants decreased in 2017. For example, the position of employment status declined significantly, showing that women, irrespective of their employment status, had a chance of accessing PNC services in recent times. PNC for mother by a medically trained provider - the role of both demand- and supply-side determinants was consistently observed across years (Figure 7). In fact, the role of wealth became stronger in 2017, indicating that wealthier women had a higher chance of availing PNC from a medically trained provider. However, the role of supply-side determinants also became stronger in 2017. For instance, the readiness of facilities for delivery became a determinant in 2017, reflecting that improved service delivery capacity of facilities could increase the chance of PNC by a medically trained provider. Women delivering in the private for-profit sector had a higher chance of accessing PNC services in 2014, but this was replaced by deliveries in public facilities in 2017, highlighting the emerging influence of public facilities in influencing PNC utilization. Figure 6: Top determinants of PNC for mother in 42 days 8 8 Determinants are defined as follows: (1) ready-del– readiness of facilities to provide normal delivery care; and (2) awareComClinic – women aware of community clinics nearby residence. 13 Figure 7: Top determinants of any PNC for mother by medically trained provider 3.4 Family planning (receipt of any FP method and receipt of modern FP) For the two FP indicators – the determinants were nearly the same for both indicators (Figures 8 and 9). Both the demand- and supply-side determinants played somewhat equal role in driving FP utilization across years. For instance, relatively older women had a higher chance of FP usage compared to their younger counterparts. Interestingly, the visits of CHWs (i.e., government health and FP workers) significantly influenced FP uptake across years, but the role of the former was slightly more prominent. The role of location, as measured by division, declined in 2017, reflecting recent reductions in inequalities in FP uptake among divisions. Receipt of any FP method – the receipt of any FP method was higher among relatively older women and those visited by the CHWs (i.e., government health worker and FP workers). Inequities in the receipt of any FP method among divisions were lesser in 2017. Receipt of modern FP - the role of employment status and women empowerment increased in 2017 for utilization of modern FP. This shows that employed women, and those with autonomy in spending and health care decision-making, had a higher chance of availing modern FP. Media access was not a top determinant, like it was for utilization of maternal care services. 14 Figure 8: Top determinants for use of any family planning method 9 Figure 9: Top determinants of receipt of modern FP 9 The determinant - Contraccep_avlbl indicates – availability of any modern contraceptives at a facility. 15 3.5 Under-5 care-seeking (diarrhea, fever and acute respiratory infection (ARI)) For the three indicators – although both demand- and supply-side determinants consistently played a significant role across years, the role of some supply-side factors increased in prominence in 2017 (Figures 10 to 12). For instance, the availability of IMCI staff increased the chance of care-seeking in 2017. The role of improved sanitation facilities at households declined in 2017, perhaps giving more prominence to supply- side determinants. However, among the demand-side determinants – being younger increased the chance of availing care for all three health conditions across years. Education, however, became a top determinant in 2017 showing that children of mothers with a relatively higher education had a better chance of care- seeking. Division also became a top-determinant in 2017, reflecting that the differences in care-seeking among divisions has increased in recent times. Care-seeking for diarrhea - awareness by women of community clinics increased the odds of care-seeking. Among the demand-side determinants, mother’s education became a top determinant in 2017. Care-seeking for fever - wealth was a top determinant consistently across years, reflecting that the inequality based on wealth status is persistent. Care-seeking for ARI - awareness by women of community clinics increased the odds of care-seeking. Interestingly, the role of certain facility characteristics (i.e., availability of growth monitoring staff, basic equipment, and amenities) became the top determinant. Figure 10: Top determinants of - seeking Treatment for Under-5 Diarrhea 16 Figure 11: Top determinants of - Seeking Treatment for Under-5 Fever Figure 12: Top determinants of - Seeking Treatment for Under-5 ARI 3.6 Receipt of full vaccination (12 to 23 months old children receiving all basic vaccinations) Interestingly, results showed that demand-side determinants are consistently top determinants, across years (Figure 13), in receipt of full vaccination. The chance of completing all required vaccinations was higher among children of wealthier, more educated, and younger mothers. In addition, location (i.e. division) became a top determinant in 2017, reflecting that differences in vaccination, among divisions, have increased in recent times. 17 Figure 13: Top determinants of receiving full vaccination (12 to 23 months old children) 3.7 Quality of care (ANC, PNC for newborn and essential newborn care at birth) For the three quality indicators - the role of the private for-profit sector was dominant for the quality of ANC and newborn PNC, where the reported quality was better for such providers (Figures 14 and 15). However, this observation was not relevant for essential newborn care practice, where the role of the private for-profit sector was not a top determinant. Supply-side determinants were the top determinants (e.g., readiness for delivery and availability of staff trained in delivery, equipment, and amenities) consistently across years. The role of wealth status declined in 2017 (except for ANC quality). Conversely, the role of location (i.e. division) was prominent (except for ANC), indicating that significant differences in quality exist across divisions. ANC quality - the role of supply-side determinants increased in 2017 (Figure 14). Consistently across years, women reported receiving better-quality ANC if the provider was medically trained in ANC. The readiness of facilities for ANC became a determinant in 2017, reflecting that woman reported better-quality in those facilities having a higher readiness for ANC. Interestingly, the role of the public facilities became a determinant in 2017; earlier it was only the private for-profit sector. The influence of public facilities, however, was very minor, indicating that a greater number of women who availed care from the private for-profit sector reported a better quality of care, while this proportion was lesser for the public facilities. Wealthier and educated women reported receiving better-quality ANC across years. Newborn PNC quality - the reported quality for newborn PNC was consistently better for the private for- profit sector (Figure 15). Location (i.e. division) became a top determinant in 2017, reflecting an increase in differences in quality, across divisions, in recent times. Conversely, the role of wealth status declined in 18 2017, indicating that, irrespective of wealth status, women reported that their newborns received satisfactory-quality PNC, with inequalities based on wealth status declining in recent times. Figure 14: Top determinants of Quality of ANC Figure 15: Top determinants of Quality of PNC for Newborn Essential newborn care at birth - supply-side determinants were top determinants consistently across years (Figure 16). For instance, reported quality was better for those facilities having a higher readiness for delivery and availability of staff trained in delivery, equipment, and amenities. The role of wealth and education status declined in 2017, reflecting that quality of care did not vary much based on such demand- 19 side characteristics. Division was a top determinant across years, reflecting persistent inequalities in the quality of essential newborn care across divisions. Figure 16: Top determinants of - Quality of Essential Newborn Care Practice at Birth 3.8 Under-5 nutrition status (stunting, wasting and underweight) For the three nutrition indicators – the demand-side determinants (e.g., child’s age and mother’s wealth and education status) were more influential to nutrition indicators (Figures 17 to 19). However, certain supply- side factors also emerged as top determinants in 2017. For instance, the availability of growth monitoring staff and IMCI staff have positively influenced nutrition status in recent years. Location, being a strong determinant in 2017, for stunting and underweight, showed that persistent differences in nutrition indicators exist across divisions. Under-5 stunting - although both demand- and supply-side determinants were found to be the key determinants, the latter had a more pronounced role consistently across the two periods (Figure 17). Child’s age (i.e., being older), wealth (relatively poor) and mother’s education (less educated) increased the chance of stunting. Media access – which had a positive influence on stunting - was a strong determinant in 2014 and 2017. Location became a top determinant in 2017, indicating that persistent differences exist in stunting across divisions. However, the relative significance of growth monitoring staff and IMCI staff increased in 2017, indicate that their presence could reduce stunting prevalence. Under-5 wasting - some supply-side determinants emerged as top determinants in 2017. For instance, the availability of staff trained in growth monitoring positively influenced (i.e. reduced) wasting prevalence (Figure 18). Child’s age (i.e., younger children less wasted) was the top determinant across the two time periods. Meanwhile, the role of wealth relatively declined in 2017, showing that the chance of wasting was, to some extent, irrespective of wealth status in recent years. 20 Figure 17: Top determinants of - Under-5 Stunting 10 Figure 18: Top determinants of - Under-5 Wasting Under-5 underweight - mother’s wealth status (relatively poor) and education (less educated) were the top determinants consistently across the two time periods (Figure 19). Location became a top determinant in 10 The Determinant curativecare_child indicates – under-5 received care at a facility for diarrhea or fever or ARI. 21 2017, indicating persistent differences for stunting prevalence across divisions. However, the relative significance of growth monitoring staff and IMCI staff increased in 2017, showing that their presence could reduce stunting prevalence. Figure 19: Top determinants of - Under-5 underweight 3.9 Divisional results Divisional results were mostly similar to the national-level results for all indicators across years. One major difference was that residence was a major determinant in two provinces (Dhaka and Chattogram) for ANC 4+ visits in 2017, indicating that pregnant women in urban areas had a higher chance of completing 4+ ANC visits. Similarly, residence was a determinant for reported ANC quality in all provinces and more visibly for Chattogram, Khulna and Sylhet in 2017, reflecting the prevailing better performance of urban areas on quality of ANC across provinces. Secondly, some provinces showed the positive role of CHWs (government health worker) for some indicators, while this was not observed nationally. For instance, visits of government health worker positively influenced SBA in Khulna, Barisal and Sylhet and PNC in Rajshahi, but their visit was not a top determinant for SBA and PNC nationally. 4. Discussion The study findings draw attention to the determinants that have greater influence on the RMNCH-N service landscape in Bangladesh, and those that need to be prioritized and addressed for further improvements to RMNCH-N outcomes. All findings discussed below are also validated through the existing literature, both for Bangladesh and the South Asian region. 1. The role of individual and household determinants is reducing; wealth and education status are exceptions, being the top determinants for some outcomes. 22 Findings clearly indicate that, compared to 2014, the relative importance of individual (e.g., mother’s age, birth order) and household (sanitation and handwashing) determinants have declined considerably in 2017. Similarly, differences in utilization based on divisional location (except for nutrition status) and residence (rural-urban) also declined in 2017. The existing literature shows that Bangladesh has been successful in reducing inequalities in RMNCH-N service utilization based on location (i.e. division), residence and socio- demographic groups; the role of individual determinants (e.g., age and birth order) has less influential after 2015.27,28 The persistent efforts of the government in integrating community needs into the safe motherhood programs, identifying local needs through community-based platforms and mainstreaming NGOs into community health, are worth acknowledging here as success factors.27 The study found two exceptions to the above findings. First, in the case of FP usage, age and birth order were found to be top determinants. This, however, was a positive observation, indicating that older women, and those with a greater number of children, had a higher chance of using FP methods. This is promising finding for maintaining the low fertility rate in Bangladesh, and reflects the success of FP strategies. A second exception was increased availability of improved sanitation and handwashing facilities, which has improved household care-seeking behavior for the under-5 age-group. Perhaps, households with such amenities were more aware of appropriate care-seeking and financially better-off to afford care. Further, it could be postulated that wealth status is the ultimate determinant here, as wealth status considerably increases the possibility of having sanitation and handwashing facilities in households.27 Additionally, wealth and education status affected utilization across most RMNCH-N service indicators. For example, and as expected, wealth and education status of mothers were the top determinants of nutrition outcomes. Inequalities in utilization, based on wealth status, can adversely exacerbate rural-urban gaps in RMNCH-N service usage. A recent study reported noticeable inequalities in RMNCH-N service utilization, based on wealth, even within urban areas.28 As such, prevailing user fees exemption (with only nominal charges) in public facilities may have limited scope in improving utilization by poor women for two major reasons: 1) functional readiness of public facilities is limited, causing women to seek care in the private for-profit sector and/ or purchase drugs and supplies outside of facilities, thereby incurring sizable out-of- pocket expenses; and 2) remote and less-developed regions have limited presence of public facilities. Resultingly, women must depend on private providers.29 This study, however, also observed that a greater number of wealthier women used public facilities for PNC compared with their poorer counterparts. This is a common phenomenon in the South Asia region, where the wealthier have a better chance of availing public goods and social benefits.30 This study did not assess the financial catastrophic spending (where out- of-pocket spending exceeds a certain threshold of monthly household income or monthly total consumption) related to RMNCH-N care-seeking. Heavy reliance on the private for-profit sector, and the indirect costs of care (e.g. transportation), can contribute to severe financial catastrophies.31 Results also showed that a better educated woman had a higher chance of completing 4+ ANC and SBA, and her under-5 child(ren) also had a relatively higher chance of accessing treatment. Education status also influenced the chance of child-care seeking and full vaccination. Access to mass media could be instrumental here. Although this study did not assess the impact of any specific health awareness strategies through mass media platforms, women having better access to mass media, also showed higher RMNCH- N service utilization. Some pilot studies have shown that behavior change interventions through mobile technology are cost effective for maternal and neonatal health in Bangladesh.32,33 Education and wealth status are also directly related and, perhaps, low RMNCH-N utilization among the less-educated could be due to both their inability to pay, and a lack of knowledge and awareness. In the South Asia context, poor 23 awareness has greater potential to reduce FP utilization, than maternal care usage.28 This is because there is a greater prioritization of maternal care at the household-level, as childbirth is considered to be more of a household event.34 However, this study observed a contrary scenario in which utilization was better for FP than that for maternal care. 2. Availability of skilled staff, amenities and equipment at facilities could positively influence RMNCH-N service utilization and outcomes. All RMNCH-N service utilization indicators were positively affected by the availability of skilled staff, amenities, drugs, and equipment in both public and private sectors. Availability of growth monitoring staff and childcare medicines positively influenced nutrition status, care-seeking and vaccination, with this positive influence perhaps could linked to their delivery of sensitization messages to mothers.35 It is a welcome trend that most women are willing to seek care if facilities are functional with medically trained staff, and necessary equipment and amenities. 3. Quality of care could positively influence RMNCH-N service utilization. This study observed that the utilization rate was higher among those women who reported receiving better quality of ANC, PNC and essential newborn birth care practice. However, quality of care in the private for- profit sector had a relatively better influence on RMNCH-N service utilization. This scenario reflects sub- optimal and less-satisfactory quality experienced by women in public facilities. Women reported receiving better quality care if they received ANC from a medically trained staff. This is likely linked to adherence to relevant treatment guidelines, by medically trained staff .31 Similarly, women reported better quality care if basic equipment and amenities were available in facilities. While this study did not assess the motivation and performance of health workers, reported quality was proportionately higher based on the availability of medically trained staff, equipment, and amenities. This could perhaps be interpreted as a scenario in which medically trained providers were motivated to perform their duties in the for-profit private sector. And given the low influence of quality in the public facilities on RMNCH-N service utilization, it could be interpreted that work motivation and performance is relatively lower for public facilities’ providers.28 Existing studies in Bangladesh show that the quality of RMNCH-N care in Bangladesh is particularly poor in the public sector,36 with some divisions and urban areas far better- off in terms of quality of care.36 4. The role of public facilities is emerging strongly in influencing RMNCH-N service utilization, but poor quality of care is a constraint here. Women who availed ANC from the private for-profit sector had a better chance of completing 4+ ANC in 2014. This trend was reversed in 2017, with an emerging role for the public sector. This is a welcome trend, and reflects substantive improvements in the availability of services at public facilities and the performance of staff in motivating women to avail appropriate care. Perhaps the pivotal role of community clinics in linking the community with the primary health care system needs to be highlighted.28 5. The role of CHWs was obvious in driving FP utilization, but not maternal and childcare utilization. Although primary data collection was not conducted, the study findings show that CHWs had a limited role in influencing maternal and child service utilization unlike for FP. With regards to FP use, the motivation of CHWs (i.e., FP and health workers) were shown to have a greater influence than mass media. Not having 24 a qualitative study is a limitation in validating this finding. However, the existing literature does not provide substantial contradictory evidence to refute this finding. While studies do reflect the promising role of CHWs in FP utilization, there is limited evidence of influence on other utilization indicators.37 Nonetheless, this limited role is not promising, especially when compared with other countries in the South Asia region that have shown the impressive role of CHWs in driving maternal and child service uptake. The Accredited Social Health Activist (ASHA) program in India, for example, has emerged as one of the largest and most effective CHW programs globally for RMNCH-N services.38 It is worth noting that the Bangladesh National Community Health Worker Strategy 2019 also calls for reinforcing the scope of CHWs in community health through multi-faceted approaches.37 Study Limitations This study has certain limitations. First, the selection of determinants and outcome indicators was limited by both data availability and compatibility for a comparative analysis across two time points. Therefore, certain crucial health outcomes (e.g., maternal and under-five mortalities) and child feeding practices could not be included when considering the nutrition outcomes. Also, data for Mymensingh was not separately included in DHS 2014 and therefore this division was excluded from division-specific analyses. Second, this study inherits the data quality issues and biases of both DHS and BHFS. Utilization and quality of care data were self-reported without verification, with a chance of recall bias, information bias and social desirability bias. BHFS did not physically verify the availability of services in facilities, as it collected information from the facility records. Validation studies in LMICs, however, have found a moderate to high sensitivity, and a moderate validity for self-reported coverage of RMNCH-N service utilization and quality of care in household surveys.39,40 For quality of care, DHS uses objective questions on standardized medical procedures and thus, the typical subjectivity concerns for psychometric assessments may not be applicable here as shown by validating studies.39,40 Also, while assessing the determinants of RMNCH-N service utilization, the service seekers’ experience is more meaningful than a facility-based empirical assessment of quality.40 Third, this study did not have a primary data component, and, in particular, a qualitative component to triangulate why and how certain determinants influenced each outcome in a particular way. Therefore, drawing conclusions from the findings requires a cautionary approach. Fourth, the machine learning analysis provides the top determinants that influenced the study outcomes. But a careful approach (substantiated by literature) is needed while concluding the direction of association between a specific determinant and an outcome. This study used a vast contextual literature including the DHS reports for triangulation. Fifth, this study exclusively focused on the pre-pandemic (COVID-19) phase and used one of the most recent household and facility surveys data. Findings may not be appropriate for any decision making during the pandemic but would certainly be relevant for the post-COVID phase. The existing evidence indicates that the determinants of RMNCH-N service utilization would remain largely the same in the post-pandemic phase. Nonetheless, low service provision, poor quality and financial constraints of service seekers perhaps would be multiplied and remain as the top determinants in the post-pandemic phase.41 This study, however, also shows that service availability, quality and household wealth status are the top determinants even in the pre-pandemic phase. Lastly, this study did not assess the impact of any ongoing policy interventions. Thus, findings are not robust enough to make a conclusion on the ultimate impact of any initiatives. However, findings certainly reflect the key determinants that could be addressed further, on a priority basis. Findings can also indicate the need, as well as the direction, for mainstreaming some of the current policy strategies and interventions. 25 5. Policy Implications The findings clearly indicate that demand-side awareness has improved substantially, with women more influenced by service availability, and quality, than by cultural barriers. This shift could be considered progressive, as women choose to seek care depending on the quality of services available at facilities. A caution here is that the readiness of facilities for RMNCH-N care and quality of services need to be improved further, especially in public facilities to keep up the current momentum of care-seeking. Otherwise, either this momentum may not persist, or the care-seeking behaviors of pregnant women and mothers may not translate into proportionate improvements in MCH status and reduction in mortality. Improved health facility readiness and quality of care, however, does not exclude the need to significantly increase the financial access of poor women, and further increase awareness among less-educated women. Unless targeted and specific strategies are implemented for these two groups (poor and less educated), the country’s remarkable progress in improving RMNCH-N outcomes may be limited.42 Similarly, study findings also show that some divisions will require additional focused attention. Overall RMNCH-N achievements in the country over the last two decades, and some obvious improvements in the determinants of RMNCH-N (as shown in the study) point to the need to deepen and mainstream: a) targeted population-specific interventions to enhance affordability and awareness; b) community-based approaches; and c) specific management strategies to improve facility performance and quality of care. A few strategies that can be considered are outlined below: 1. Improving funding, functional readiness and management capacities of facilities – public facilities in remote and some rural areas require constant funding, supportive supervision, and governance support. A recent study in Bangladesh showed that public facilities cannot improve their efficiency further without additional investments on specific inputs (e.g. staff performance).43 In this regard, a specific health sector intervention involving results-based financing or pay-for-performance, which could potentially improve the capacity and performance of facilities in remote areas, could be piloted.42 Results-based financing has shown to improve functional readiness of health facilities in several low- and middle-income countries (LMIC).42 2. Improving clinical quality of care – improving the clinical quality further would be a requirement, particularly for the rural areas in all divisions. The prevailing evidence shows that measures such as continuous capacity building needs assessment of staff could help to improve the clinical quality of care.44 Some countries have also experienced improvements in quality through pay-for-performance inventives.45 3. Strengthening the use of mass media or mobile technology in RMNCH-N service provision a. Targeted behavior change interventions using mass media platforms– targeted interventions using mass media have proven to be more effective for behavior change than regular awareness strategies.46 Behavior-change strategies targeting specific groups (e.g. less educated women) have also been shown to remove cultural barriers in MCH care.46 This platform could potentially help in elevating the consistently low PNC usage from the current 50% and 4+ ANC among less educated women.46 b. Mobile technology for improving the capacity and performance of health and FP workers in rural and urban areas– mobile applications could help in the identification of maternal and neonatal danger signs; sending of reminders about ANC, PNC and vaccination timeline; spreading awareness on nutrition; and growth monitoring of children. 26 4. Mainstreaming the scope, capacity, and influence of CHWs (health, FP and NGO workers)– improving the scope, capacity, and performance of CHWs could be a wise strategy for improving RMNCH-N service utilization, particularly PNC. Existing evidence shows that a regular capacity building needs assessment of CHWs would help to identify the gaps in skills relative to local RMNCH- N service needs. Introducing measures to improve CHWs’ work motivation and performance could be relevant here. Studies show that motivation of CHWs is pivotal to their satisfactory performance. In addition, challenges such as transportation and non-acceptance from the community, which are shown to affect performance, should also be addressed.38 Findings also encourage a full-fledged implementation of the Bangladesh National CHW Strategy as a means of enhancing CHW performance. 5. Deepening the public-private (i.e. NGO) engagement at the community-level– mainstreaming the role of CHWs in RMNCH-N service delivery. First, NGOs could be deployed to train CHWs. Evidence from some of the pilot interventions in Bangladesh have shown successes with this approach. For instance, a public-private partnership pilot was effective in early identification of maternal danger signs, and community mobilization efforts, by training the community skilled birth attendants (CSBA) in 10 sub-districts and 50 unions of Sunamganj.47 Second, NGOs could be involved in the supportive- supervision of CHWs in rural and remote areas. Evidence from the South Asia region shows promising results from the provision of supportive supervision to CHWs, by NGOs.48 6. Improving affordability for poor women– as mentioned earlier, user fees exemption in public facilities will have limited impact on poor women, given their dependence on private providers (due to limited public facilities in remote areas and their low functional readiness). In this context, demand- side vouchers (cashless system) and cash incentives could be potential options explored, with the possibility of national scalability. Vouchers have been shown to be effective in improving RMNCH-N service utilization across several LMICs and in addressing the indirect costs of care.49 Similarly, cash incentives are known to positively impact a wide range of RMNCH-N behaviors including dietary behaviors in LMICS.49 7. Additional explorations through analytic studies– given findings that less educated women had better FP utilization and less MCH utilization, there is a need to explore: (i) the role of affordability and awareness in driving MCH utilization through a qualitative study; (ii) the household financial catastrophic spending from RMNCH-N care; (iii) a benefit-incidence analysis to understand equity in public health spending for RMNCH-N services; and (iv) the positive role of community-based mechanisms (e.g., community clinics, CHWs and NGOs) in improving FP usage, in order to replicate such exemplary strategies for MCH care. 6. Conclusions Overall, this study found that the health system of Bangladesh stands at a positive juncture currently, where increasing the functional readiness of facilities and quality of care would increase RMNCH-N service utilization, compared to earlier times, given the reduced dominance of demand-side determinants (e.g., age, birth order, education status) in recent years. Findings clearly indicate that demand-side awareness has improved substantially, with women, on the whole, being more influenced by service availability, and quality, than cultural barriers. This shift could be considered progressive, since women choose to seek care if quality services are perceived to be available at health facilities. However, quality of care was better in urban areas across almost all divisions. A caution is that the readiness of facilities for RMNCH-N care provision, and quality services, needs to be improved further, especially in the public facilities to keep up 27 with the current momentum of care-seeking, as the influence of the public facilities was relatively lower (except for PNC). Otherwise, either this momentum may not persist, or the sincere care-seeking of pregnant women and mothers may not result in proportionate improvements in RMNCH-N status and reductions in mortality. Supply-side improvements, however, do not preclude the significant need to increase the affordability of poor women and further improve awareness among less-educated women. Wealth status, being a top determinant across years, indicates that perhaps the prevailing user fees exemption alone is not adequate to improve RMNCH-N service uptake. Additionally, given low service delivery status in the public sector, and the heavy reliance on the private for-profit sector for RMNCH-N care, it would be pertinent to consider addressing the direct and indirect costs of care through demand-side financial incentives. Although this study did not assess the direct impact of any awareness strategies through mass media, women having mass media access had a better chance of utilizing RMNCH-N services. This suggests the need for specific targeted interventions using mobile technology to augment awareness and utilization among vulnerable groups (e.g., less educated). The low influence of CHWs in maternal and childcare utilization (unlike FP) also further suggests the need to leverage the potential of mobile technology to improve CHW capacity and connectivity with the communities. If needed, NGOs can be involved in providing supportive supervision of CHWs. Finally, additional analytical studies are suggested to generate further evidence on improving both the financial access of poor women and the role of CHWs in the provision of MCH care. 28 References 1. Stenberg, K., Sweeny, K., Axelson, H., Temmerman, M. & Sheehan, P. Returns on Investment in the Continuum of Care for Reproductive, Maternal, Newborn, and Child Health. in Disease Control Priorities, Third Edition (Volume 2): Reproductive, Maternal, Newborn, and Child Health 299–317 (The World Bank, 2016). doi:10.1596/978-1-4648-0348-2_ch16 2. Richter, L. M. et al. Investing in the foundation of sustainable development: pathways to scale up for early childhood development. Lancet 389, 103–118 (2017). 3. Liu, L. et al. Global, regional, and national causes of under-5 mortality in 2000–15: an updated systematic analysis with implications for the Sustainable Development Goals. Lancet 388, 3027– 3035 (2016). 4. The Lancet Global Health. Progressing the investment case in maternal and child health. Lancet Glob. Heal. 9, e558 (2021). 5. Crear-Perry, J. et al. Social and Structural Determinants of Health Inequities in Maternal Health. J. Women’s Heal. 30, 230–235 (2021). 6. Kerber, K. J. et al. Continuum of care for maternal, newborn, and child health: from slogan to service delivery. Lancet 370, 1358–1369 (2007). 7. The World Bank. The Path to Universal Health Coverage in Bangladesh. (2015). 8. Ministry of Health and Family Welfare, B. Success Factors for Women’s and Children’s Health. (2015). 9. Bangladesh Bureau of Statistics (BBS) and UNICEF. Multiple Indicator Cluster Survey 2019. Progotir Pathey 1, (2019). 10. Roy A and Shengelia L. An Analysis on Maternal Healthcare Situation in Bangladesh: A Review. Divers. Equal. Heal. Care (2016). 11. Rajia, S., Sabiruzzaman, M., Islam, M. K., Hossain, M. G. & Lestrel, P. E. Trends and future of maternal and child health in Bangladesh. PLoS One 14, e0211875 (2019). 12. National Institute of Population Research and Training (NIPORT), and I. Bangladesh Demographic and Health Survey 2017-18: (2019). doi:https://dhsprogram.com/publications/publication-PR104-Preliminary-Reports-Key-Indicators- Reports.cfm 13. The World Bank Group, W. D. Maternal mortality ratio. (2017). Available at: https://data.worldbank.org/indicator/SH.STA.MMRT?locations=BD. 14. Horton S and Levin C. Cost-Effectiveness of Interventions for Reproductive, Maternal, Neonatal, and Child Health. in Reproductive, Maternal, Newborn, and Child Health: Disease Control Priorities, Third Edition (Volume 2) (ed. Black RE, Laxminarayan R, T. M. et al.) (The World Bank, 2016). 15. Zhao, P. et al. Maternal health services utilization and maternal mortality in China: a longitudinal study from 2009 to 2016. BMC Pregnancy Childbirth 20, 220 (2020). 16. Kim, C., Saeed, K. M. A., Salehi, A. S. & Zeng, W. An equity analysis of utilization of health services in Afghanistan using a national household survey. BMC Public Health 16, 1226 (2016). 17. National Institute of Population Research and Training (NIPORT), A. for C. and & Population Research (ACPR), and I. I. Bangladesh Health Facility Survey 2014. (2016). 18. National Institute of Population Research and Training (NIPORT) and ICF. Bangladesh Health Facility Survey 2017. (2017). 19. Mumbare, S. & Rege, R. Ante natal care services utilization, delivery practices and factors affecting them in tribal area of North Maharashtra. Indian Journal of Community Medicine 36, 287 (2011). 20. World Health Organization (WHO). Postnatal Care for Mothers and Newborns Highlights from the World Health Organization 2013 Guidelines. (2015). 29 21. Niswade, A. et al. Neonatal morbidity and mortality in tribal and rural communities in Central India. Indian J. Community Med. 36, 150–158 (2011). 22. Paul, V. K. et al. Reproductive health, and child health and nutrition in India: meeting the challenge. Lancet 377, 332–349 (2011). 23. R Core Team (2021). A language and environment for statistical computing. R Foundation for Statistical Computing. (2021). 24. Panch, T., Szolovits, P. & Atun, R. Artificial intelligence, machine learning and health systems. J. Glob. Health 8, (2018). 25. Natekin, A. & Knoll, A. Gradient boosting machines, a tutorial. Front. Neurorobot. 7, (2013). 26. Huang, Y., Li, W., Macheret, F., Gabriel, R. A. & Ohno-Machado, L. A tutorial on calibration measurements and calibration models for clinical prediction models. J. Am. Med. Informatics Assoc. 27, 621–633 (2020). 27. Haider, M. R. et al. Impact of maternal and neonatal health initiatives on inequity in maternal health care utilization in Bangladesh. PLoS One 12, e0181408 (2017). 28. Khan, M. N., Kumar, P., Rahman, M. M., Islam Mondal, M. N. & Islam, M. M. Inequalities in Utilization of Maternal Reproductive Health Care Services in Urban Bangladesh: A Population- Based Study. SAGE Open 10, 215824402091439 (2020). 29. Atchessi, N., Ridde, V. & Zunzunegui, M.-V. User fees exemptions alone are not enough to increase indigent use of healthcare services. Health Policy Plan. 31, 674–681 (2016). 30. Khan, J. A. M. et al. Benefit incidence analysis of healthcare in Bangladesh – equity matters for universal health coverage. Health Policy Plan. czw131 (2016). doi:10.1093/heapol/czw131 31. Joarder, T., Chaudhury, T. Z. & Mannan, I. Universal Health Coverage in Bangladesh: Activities, Challenges, and Suggestions. Adv. Public Heal. 2019, 1–12 (2019). 32. Rajia, S., Sabiruzzaman, M., Islam, M. K., Hossain, M. G. & Lestrel, P. E. Trends and future of maternal and child health in Bangladesh. PLoS One 14, e0211875 (2019). 33. Tobe, R. G., Haque, S. E., Ikegami, K. & Mori, R. Mobile-health tool to improve maternal and neonatal health care in Bangladesh: a cluster randomized controlled trial. BMC Pregnancy Childbirth 18, 102 (2018). 34. Mukherjee, S. & Singh, A. Has the Janani Suraksha Yojana (a conditional maternity benefit transfer scheme) succeeded in reducing the economic burden of maternity in rural India? Evidence from the Varanasi district of Uttar Pradesh. J. Public health Res. 7, 957 (2018). 35. Deshmukh, P., Dongre, A. & Garg, B. Childhood morbidity, household practices and health care seeking for sick children in a tribal district of Maharashtra, India. Indian Journal of Medical Sciences 64, 7 (2010). 36. Biswas, T. K., Sujon, H., Rahman, M. H., Perry, H. B. & Chowdhury, M. E. Quality of maternal and newborn healthcare services in two public hospitals of Bangladesh: identifying gaps and provisions for improvement. BMC Pregnancy Childbirth 19, 488 (2019). 37. Olaniran, A., Madaj, B., Bar-Zev, S. & van den Broek, N. The roles of community health workers who provide maternal and newborn health services: case studies from Africa and Asia. BMJ Glob. Heal. 4, e001388 (2019). 38. Gopalan, S. S., Mohanty, S. & Das, A. Assessing community health workers’ performance motivation: a mixed-methods approach on India’s Accredited Social Health Activists (ASHA) programme. BMJ Open 2, (2012). 39. Liu, L. et al. Measuring Coverage in MNCH: A Validation Study Linking Population Survey Derived Coverage to Maternal, Newborn, and Child Health Care Records in Rural China. PLoS One 8, e60762 (2013). 40. Hancioglu, A. & Arnold, F. Measuring Coverage in MNCH: Tracking Progress in Health for Women and Children Using DHS and MICS Household Surveys. PLoS Med. 10, e1001391 (2013). 41. Elston, J. W. T. et al. Maternal health after Ebola: unmet needs and barriers to healthcare in rural Sierra Leone. Health Policy Plan. (2019). doi:10.1093/heapol/czz102 30 42. Lassi, Z. S., Middleton, P. F., Bhutta, Z. A. & Crowther, C. Strategies for improving health care seeking for maternal and newborn illnesses in low- and middle-income countries: a systematic review and meta-analysis. Glob. Health Action 9, 31408 (2016). 43. Ahmed, S. et al. Technical efficiency of public district hospitals in Bangladesh: a data envelopment analysis. Cost Eff. Resour. Alloc. 17, 15 (2019). 44. Mayumana, I., Borghi, J., Anselmi, L., Mamdani, M. & Lange, S. Effects of Payment for Performance on accountability mechanisms: Evidence from Pwani, Tanzania. Soc. Sci. Med. (2017). doi:10.1016/j.socscimed.2017.02.022 45. Bonfrer, I., Van de Poel, E. & Van Doorslaer, E. The effects of performance incentives on the utilization and quality of maternal and child care in Burundi. Soc. Sci. Med. 123C, 96–104 (2014). 46. Fatema, K. & Lariscy, J. T. Mass media exposure and maternal healthcare utilization in South Asia. SSM - Popul. Heal. 11, 100614 (2020). 47. Hossain, J. et al. Filling the human resource gap through public-private partnership: Can private, community-based skilled birth attendants improve maternal health service utilization and health outcomes in a remote region of Bangladesh? PLoS One 15, e0226923 (2020). 48. Das, A. et al. Strengthening malaria service delivery through supportive supervision and community mobilization in an endemic Indian setting: an evaluation of nested delivery models. Malar. J. 13, 482 (2014). 49. Hunter, B. M. & Murray, S. F. Demand-side financing for maternal and newborn health: what do we know about factors that affect implementation of cash transfers and voucher programmes? BMC Pregnancy Childbirth 17, 262 (2017). 31