IMPACT OF HEALTH SECTOR SUPPORT PROJECT ON ESSENTIAL NUTRITION SERVICES DISCUSSION PAPER SEPTEMBER 2022 Wameq Raza Deepika Chaudhery / Impact of Health Sector Support Project on Essential Nutrition Services Evidence from Bangladesh Wameq Raza Deepika Chaudhery September 2022 i Health, Nutrition, and Population (HNP) Discussion Paper This series is produced by the Health, Nutrition, and Population Global Practice of the World Bank. The papers in this series aim to provide a vehicle for publishing preliminary results on HNP topics to encourage discussion and debate. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors, or to the countries they represent. Citation and the use of the material presented in this series should take into account this provisional character. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of the World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. For information regarding the HNP Discussion Paper Series, please contact the Editor, Jung-Hwan Choi at jchoi@worldbank.org or Erika Yanick at eyanick@worldbank.org. Rights and Permissions The material in this work is subject to copyright. Because the World Bank encourages the dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street, NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. © 2022 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW, Washington, DC 20433 All rights reserved. ii Health, Nutrition, and Population (HNP) Discussion Paper Impact of Health Sector Support Project on Essential Nutrition Services: Evidence from Bangladesh (2018–2020) Wameq Razaa Deepika Chaudheryb a Health and Nutrition Specialist (HSAHP), HNP Global Practice, World Bank, Dhaka, Bangladesh b Senior Health and Nutrition Specialist (HSAHN), HNP Global Practice, World Bank, New Delhi, India Abstract: Despite progress over the past two decades, poor nutrition remains a significant public health challenge in Bangladesh. Stunting among children under five years declined from 43 to 31 percent between 2007 and 2018, while 42 percent of women between 15 to 49 years are anemic. The Ministry of Health and Family Welfare is implementing the Health Sector Support Project (HSSP) with financing and technical assistance from the World Bank in the Sylhet and Chattogram divisions of the country. The project considers improving nutrition outcomes as a core priority. Leveraging administrative data from 13,855 community clinics (CCs) from 2018 to 2020 and a difference-in-difference approach, the analysis finds that HSSP led to improvements in the delivery of both maternal and child nutrition services. The proportion of eligible pregnant women who received requisite antenatal services (receipt of at least 30 iron and folic acid tablets, nutrition counseling, and weight measurement) increased by 2.7 percent over the duration. Similarly, the proportion of children between 0 and 23 months, who received age-appropriate nutrition counseling, increased by 8.9 percent over the same period. The paper identifies several factors that led to these improvements and notes the impediments. The HSSP renewed focus on the importance of delivering quality nutrition services, and the technical assistance provided through the HSSP has strengthened capacity, not only around the delivery of services but also in improving the data ecosystem and quality of project monitoring and results verification. There are, however, issues impeding service delivery of nutrition services. The Community Health Care providers (CHCPs) are often faced with competing priorities, as nutrition is one of the many services they provide. Similarly, the CHCPs have been found to lack the required capacity and skills in delivering services and are also burdened with poor Information Technology (IT) equipment. Keywords: Nutrition, Bangladesh, HSSP, impact evaluation, administrative data Disclaimer: The findings, interpretations, and conclusions expressed in the paper are entirely those of the authors, and do not represent the views of the World Bank, its Executive Directors, or the countries they represent. Correspondence Details: Wameq Raza, HNP Global Practice, World Bank, Dhaka, Bangladesh; +1-202-390-5682; wraza@worldbank.org. 4 Table of Contents ACKNOWLEDGMENTS ............................................................................................................ 7 PART I – INTRODUCTION AND BACKGROUND .................................................................... 8 HEALTH SECTOR SUPPORT PROJECT ACTIVITIES .......................................................... 9 PART II – DATA AND METHODS ............................................................................................10 DATA.....................................................................................................................................10 ANALYTICAL TECHNIQUES ................................................................................................11 DESCRIPTIVE STATISTICS .................................................................................................12 Maternal Nutrition Services ................................................................................................12 Child nutrition services .......................................................................................................14 PART III – IMPACT OF HSSP ON ESSENTIAL NUTRITION SERVICES ................................16 MATERNAL NUTRITION SERVICES ....................................................................................16 CHILD NUTRITION SERVICES ............................................................................................17 STUDY LIMITATIONS ...........................................................................................................18 PART IV – DISCUSSION AND CONCLUSION ........................................................................19 REFERENCES .........................................................................................................................21 ANNEXES ................................................................................................................................23 5 6 ACKNOWLEDGMENTS The authors are grateful to the World Bank for publishing this report as an HNP Discussion Paper. The authors acknowledge and are grateful for the insightful feedback from Ajay Tandon (Lead Economist, HSAHN), Patrick Hoang-Vu Eozenou (Senior Economist, HHNGF), Mickey Chopra (Lead Health Specialist, HSAHP), Shiyong Wang (Senior Health Specialist, HSAHP), and Iffat Mahmud (Senior Operations Officer, HSAHP). The authors acknowledge contributions from Md. Ferdous Ripon (Consultant, HSAHP), Khodeja Zaman (Consultant, HSAHP), and Md. Saimumuzzaman (Consultant, HSAHP) in acquiring the data used in this report and their insights into understanding the results. Lastly, the authors appreciate the comments and insights on the draft from Dr. Mustafizur Rahman (Line Director, National Nutrition Services) and Dr. Md M. Bulbul Islam (Program Manager, National Nutrition Services). 7 PART I – INTRODUCTION AND BACKGROUND Bangladesh has made great strides in improving maternal and child nutrition outcomes over the past decades. Chronic undernutrition, measured by stunting levels among children under-five years, declined from 43 to 31 percent between 2007 and 2018 (NIPORT, Mitra and Associates, and ICF International 2008, 2019). The effects are largely reflective of improvements in health service delivery, food security and dietary diversity, as well as educational attainment, women’s empowerment, and environmental conditions. Despite the progress, poor nutrition remains a significant public health challenge for Bangladesh. This is reflected by the trends in the Human Capital Index that suggest that the probability of a child surviving until the age of five is 97 percent, while 31 percent of children will be stunted (Kraay 2019). Bangladesh has one of the highest low-birth-weight rates (14.5 percent), reflecting a high burden of fetal growth restriction and preterm births (Ahmed 2021). Nearly 4.5 million children under five years are stunted (31 percent), while a further 3.2 million (22 percent) are acutely malnourished (NIPORT, Mitra and Associates, and ICF International 2019). While the period from birth to two years of age is critical for physical and cognitive growth, exclusive breastfeeding during the first six months has stagnated at around 65 percent between 2011 and 2017–18. For children between six months to two years, age-appropriate feeding practices were followed for only a third of the children (NIPORT, Mitra and Associates, and ICF International 2008, 2019). Approximately 42 percent of women of reproductive age (15 to 49 years) are anemic, and one quarter of nonpregnant, nonlactating Bangladeshi women are calcium-deficient, increasing the risks of preventable deaths and complications during pregnancy and delivery (Ash et al. 2020; NIPORT, Mitra and Associates, and ICF International 2013; Black et al. 2013). Poor maternal nutrition also has implications for the growth and development of newborns. Malnutrition in childhood and pregnancy can lead to many adverse consequences for child survival and long- term developmental potential. The COVID-19 pandemic is expected to increase undernutrition in vulnerable households due to diminished food security and impeded access to curative and preventive health services, thus perpetuating the intergenerational cycle of malnutrition (FAO et al. 2020). The World Bank–financed Health Sector Support Project (HSSP) supports the government of Bangladesh’s (GoB) Fourth Health, Population, and Nutrition Sector Program (HPNSP) (2017– 22), which encompasses nutrition services delivered through the government health system. While the HPNSP is a national-level program, the HSSP primarily focuses on Sylhet and Chattogram, where many indicators are lower than national averages (Figure 1). The Fourth HPNSP is implemented through 29 Operational Plans (OPs). Each of the OPs focus on a specific technical area with a prescribed set of activities, budgets, and intermediate indicators necessary to achieve the higher‐level objectives of the program “to ensure that all citizens of Bangladesh enjoy health and well‐being by expanding access to quality and equitable health care in a healthy and safe living environment” through health systems strengthening and provision of quality services. The HSSP provides support through a results-based mechanism, whereby disbursement of funds is tied to achievement of annual results under 16 Disbursement-Linked Indicators (DLIs), which reflect improvements in both system development and service delivery and coverage. DLIs reflecting service delivery improvements are focused on the 42 million people living in the two 8 divisions. Among the results supported by HSSP are improvements in nutrition services, focusing on maternal nutrition interventions provided through antenatal care services, and on expansion and intensification of infant and child Figure 1: Implementation Areas of the nutrition interventions through primary care HSSP services. The two nutrition-specific DLIs are (i) increase in the percentage of registered women receiving specified maternal nutrition services (DLR 13.4); and (ii) increase in the percentage of registered children aged under-two years receiving specified nutrition services (DLR 14.4), in Sylhet and Chattogram divisions. In conjunction with the DLIs, the project also supports effective implementation of individualized monitoring and case management through a system of individual records for registered pregnant women , infants, and children through the District Health Information System-2 (DHIS-2). Using annualized administrative data from 13,855 Community Clinics (CCs) over a three-year period (2018–2020), this paper evaluates the impact of the HSSP on the programmatic indicators; that is, proportion of registered eligible pregnant women and children who are provided requisite services. Source: Authors’ rendition The paper is organized as follows: the following section describes the modus operandi of the HSSP’s nutrition intervention; Part II describes the data and analytical techniques, and descriptive statistics; Part III presents the impact of the program; Section 4 discusses the findings and draws conclusions. HEALTH SECTOR SUPPORT PROJECT ACTIVITIES The HSSP’s nutrition component is implemented by the National Nutrition Services (NNS) Operational Plan based in the Institute of Public Health and Nutrition (IPHN), Directorate General of Health Services (DGHS), in collaboration with the Community Based Healthcare (CBHC) Operational Plan. The Management Information System (MIS) manages the flow of administrative data through the management of relevant Information Technology (IT) infrastructure across the country. At the district and subdistrict levels, the program is overseen by a CBHC health management team comprising Civil Surgeons (CS), Upazila Health and Family Planning Officers (UH&FPOs) and district- and subdistrict-level statisticians. At the field level, the program is implemented by CBHC’s network of CCs, the core public primary health care providers in the country. The CCs are typically operated by Community Health Care Providers (CHCPs) with 9 support from a Family Welfare Assistant and a Health Assistant. In addition to a host of other responsibilities, the CHCPs are primarily responsible for the outreach and provision of essential nutrition services in their respective catchment areas, including the collection and submission of service data into the District Health Information System-2 (DHIS-2). The HSSP provides technical assistance (TA) to the NNS through the Nutrition Information and Planning Unit (NIPU) team. The NIPU team, comprising and leveraging expertise from select development partners, 1 provides support in two key areas—capacity-development and improving monitoring and reporting of nutrition services by strengthening and mainstreaming nutrition data into the DHIS-2. The capacity-development aspect spans several fronts including enhancing understanding of the core programmatic indicators (DLIs 13.4 and 14.4) and data literacy, which includes aspects such as data entry skills of the CHCPs, extraction, analysis and reporting of DHIS-2 data for district- and subdistrict-level statisticians. Capacity-development efforts also include the District Action Plan (DAP), which supports each of the 15 districts in the project to develop annual targets for a set of nutrition indicators. The DAP is further supported through capacity development on tracking the progress towards these targets. On the monitoring front, NIPU team members consistently follow up on the progress of nutrition indicators using the DHIS- 2 data, assess its quality, and submit monthly and annual progress reports. The team members regularly follow up with low-performing district- and subdistrict-level health managers and CCs through field visits or through telephonic calls to support their efforts. PART II – DATA AND METHODS DATA Annualized administrative data (from DHIS-2) pertaining to essential nutrition services were extracted for all 13,855 CCs in Bangladesh from 2018 and 2020. While essential nutrition services are being implemented across the country, the support through the HSSP is focused on Sylhet and Chattogram, given these two regions are the most lagging in the country in terms of health outcomes. As such, the 3,485 CCs from these two divisions are designated as the treated, while the 10,370 CCs from the remaining six divisions 2 are demarked as control. For nutrition services, the CHCPs typically collect and report service-level data for each CC; that is, the number of eligible individuals who have been registered and have received the requisite services. On the maternal front, outcomes of interest include the programmatic indicator DLI 13.4, 3 reflecting the proportion of registered pregnant women who have received specified nutrition services, and each of the indicator’s components; that is, eligible pregnant women (i) registered for nutrition services; (ii) weighed during antenatal care (ANC) visits; (iii) received more than 30 iron and folic acid (IFA) tablets; and (iv) received nutrition counseling. 1 In addition to the World Bank, members of the NIPU team currently include representation and support from UNICEF, Global Alliance for Improved Nutrition (GAIN), and Nutrition International (NI). Over time, the roles of the organizations have evolved to support NNS. 2 Dhaka, Mymensingh, Khulna, Barisal, Rajshahi, and Rangpur. 3 In the way of programmatic reporting, the DLRs 13.4 and 14.4 are typically calculated at the project scale; that is, the number of (child or maternal) nutrition services in both Sylhet and Chattogram divisions are divided by the aggregated number of eligible individuals registered for a specific period. This approach, however, does not lend itself to CC-level estimations. Consequently, the figures presented in this paper are calculated at the CC level. 10 On the child front, several outcomes are evaluated. The first relates to the program indicator DLR 14.4, reflecting the proportion of registered children who have been provided age-appropriate nutrition services. Next, driven by the age-disaggregated reporting patterns in the DHIS-2 (0 to 5 months and 6 to 23 months), we separately report on the proportion of children who have been registered to receive services and have received nutrition counseling for two age groups. Next, we report on the proportion of children who have been screened for wasting, stunting, and overweight for ages between 0 to 23 months. While the CCs have comparable catchment areas, the population density is heterogeneous. To account for this, this paper normalizes the CC-level data from DHIS-2 using projected district- level population from 2020 in two steps. First, the population for each cohort of interest is calculated using proportions derived from the District Nutrition Profile published by the National Nutrition Services at the CC level. 4 Next, as services from the CCs are typically accessed by the poor, the size of the eligible cohorts are adjusted accordingly (Arifeen et al. 2013; Riaz et al. 2020). Given the national poverty headcount ratio of 23.4 percent (BBS 2019), we uniformly impose this distribution on the cohorts of interest to arrive at a denominator for normalization.5 The reported figure for each CC is then divided by this figure to arrive at a tentative coverage rate. Next, given the presence of significant outliers for both maternal and child nutrition service indicators, stratified by location and time, the top-five percentile of the observations are truncated from our estimations (Blaine 2018). ANALYTICAL TECHNIQUES Given the censored nature 6 of the dependent variables and high dispersion (see Figure A1), we use the following linear-censored (two-limit) Tobit difference-in-difference specification (Cameron and Trivedi 2010; Long 1997; Honoré 1992) to estimate the impact of HSSP on essential nutrition services: [ ∗ | 0 ≥ ≤ 1] = + 1 + 2 + + + (1) = 2018, 2020 = 1 … where ∗ is the expected value of the latent outcome of interest , which represents the ℎ proportion of eligible population registered and served by the CC in year . 2 is the coefficient of interest, representing the average treatment effect of HSSP for each outcome in the ℎ year. represents the time invariant treatment status; represents times trends. As the health care system is managed centrally at the district level, we opt to include district-level fixed effects in the model, denoted by . represents the robust idiosyncratic error term. To measure the heterogeneity of effects across the two treated divisions (Chattogram versus Sylhet), we extend Equation (1) to estimate the following specification: [∗ | 0 ≥ ≤ 1] = + 1 + 2 + + ∗ Φ + + (2) where Φ equals to 1 if the CC is in Chattogram, and zero otherwise. 4 Size of the cohort in 2018 have been retrospectively deflated using the population growth between the two years (World Bank 2021). 5 In the absence of more granular data, the results should be interpreted with the caveat that the blanket approach to applying the national poverty line to calculate the CC-level eligible population can either over- or under-estimate its size, thereby biasing our estimates. 6 The dependent variables are a proportion that can only take overserved values between 0 to 1. 11 We conduct two sensitivity tests with alternative specifications to ensure the robustness of the results. The first adopts a linear model with CC-level fixed, while the latter repeats the exercise with district-level fixed effects. Both sets of results suggest that the results are robust to alternate specifications. Results are presented in Tables A1 and A2 in Annex. Results presented from Equations 1 and 2 are interpretable as percentage changes, that is, a coefficient of 0.011 is interpretable as a 1.1 percent increase. Results from the linear model are interpretable as percentage point changes, that is, a coefficient of 0.071 is interpretable as an increase of 7.1 percentage points. All analysis are conducted using STATA v17.1 and Tableau v2020.3.2. DESCRIPTIVE STATISTICS Maternal Nutrition Services The proportion of coverage and provision of maternal nutrition services in 2018 and 2020, stratified by the treatment status, are presented in Table A3. In 2018, 28.0 and 20.6 percent of the eligible registered pregnant women (DLR 13.4) in treated and control CCs, respectively, were provided the three ANC services. The proportions increased notably by 2020 to 80.8 and 73.9 percent among the treated and control CCs, respectively. Approximately 6.6 and 9.2 percent of eligible pregnant women were registered for services in 2018 by treated and control CCs, respectively. By 2020, these proportions increased to 39.7 and 41.7 percent among the treated and control CCs, respectively. Approximately 5.4 percent of eligible registered women were weighed by treated CCs compared to 7.3 percent among control CCs in 2018 during the ANC visit. The proportion increased to 34.8 and 36.8 percent among the treated and control, respectively, in 2020. Approximately 5.4 percent of eligible pregnant women received at least 30 IFA tablets from treated CCs compared to 7.5 percent among the control CCs in 2018. In contrast, the proportion who received at least 30 IFA tablets increased among both the treated and control CCs in 2020 to 35.5 and 37.5 percent, respectively. The proportion of eligible pregnant women who received nutrition counseling during ANC visits were 5.4 and 7.2 among the treated and control CCs, respectively. In 2020, 34.5 percent of the treated CCs provided nutrition counseling compared to 36.2 percent among the control CCs. Division-level estimates are presented in Figure A2. 12 We next explore whether the increasing rate of eligible registered women resulted in an increase in the delivery of requisite services across treated and control locations (Figure 2). We find that Figure 2: Receipt of Nutrition Services by Eligible Registered Pregnant Women Source: Authors’ calculations Note: Figure shows absolute numbers of registered individuals by each CC, against the number of nutrition services delivered across the three calendar years, stratified by their treatment status. Trends are represented by the LOWESS curves. A steeper trend line suggests a higher proportion of nutrition services provided for each registered mother. there has been progress across both treated and control areas. Overall, the rates of registration and service delivery were low in 2018 for both treated and control CCs. The trend lines for 2020 are steeper for both, meaning a higher proportion of registered women received the three services 13 across the country, than in 2018. A steeper curve among the treated CCs than the control in 2020, however, suggests that the CCs in treated locations provided more services to registered pregnant women than CCs in control locations. This is likely driven by the fact that while nutrition services are rolled out across the country, the technical support and supervisory services of the HSSP focus on the Sylhet and Chattogram divisions. Child Nutrition Services The proportion of coverage and provision of child nutrition services between 2018 and 2020, stratified by treatment status, are presented in Table A4. Approximately 15.0 and 14.6 percent of the registered eligible children (0-23 months) in the treated and control CCs, respectively, were provided age-appropriate nutrition counseling in 2018 (DLR 14.4). The figure increased notably for both the treated and control CCs in 2020, to 80.5 and 61.1 percent, respectively (Table A4, Panel A). At baseline, the treated CCs registered 39.2 percent of the eligible 0 to 5-month–old children, while the control CCs registered 48.2 percent in their respective catchment areas. In 2020, the proportion for the treated CCs remained comparable (38.3 percent), while it declined for those in control areas to 43.4 percent. Approximately 9.7 and 12.9 percent of caregivers of eligible children were provided nutrition counseling services on exclusive breastfeeding by the treated and control CCs, respectively, in 2018. The proportion increased significantly for CCs in both treated and control locations to 26.0 and 28.9 percent, respectively. Treated and control CCs registered approximately 23.1 and 28.6 percent of eligible 6 to 23– month-old children in 2018 in their respective catchment areas and remained generally comparable in 2020 (23.2 and 25.1 percent among the treated and control CCs). Further, 3.3 and 4.6 percent of children from this cohort received age-appropriate complementary feeding counseling in 2018. The proportion increased significantly by 2020, to 20.0 and 20.5 percent for CCs in the treated and control areas, respectively. The rates of screening generally remained low over the duration. Approximately 4.0 and 5.1 percent of eligible children were screened for wasting in treated and control areas, respectively, in 2018. While the figure increased marginally to 4.5 percent among the treated CCs, it remained unchanged for those in control areas. The rate of screening for stunting of eligible children in treated and control CCs is comparable to that for wasting. The average rate of screening for stunting in 2018 was approximately 4.1 and 5.3 percent of children among the treated and control CCs. The rates remained largely comparable across CCs in both groups in 2020. Similarly, 4.4 and 5.7 percent of eligible children were screened for being overweight in treated and control CCs, respectively, in 2018. Following trends of other screening indicators, the coverage increased marginally for CCs in treated groups and generally remained the same for those in control areas in 2020. Division-level means are presented in Figure A3. 14 Figure 3: Distribution of Child Nutrition Services by Registered Eligible Children Source: Authors’ calculations Note: Figure shows absolute numbers of registered individuals by each of the CCs, against the number of nutrition services delivered across the three calendar years, stratified by their treatment status. Trends are represented by the LOWESS curves. A steeper trend line suggests a higher proportion of nutrition services provided for each registered child. 15 Like maternal nutrition services, we next explore whether improvement in the rate of registration of eligible children (0–5 months and 6–23 months) is associated with an increased delivery of relevant services across treated and control CCs (Figure 3). Overall, the ratio of registered children to services provided is low for both treated and control CCs in 2018. The trends are notably steeper in 2020 for both groups, suggesting that the ratio has improved across the country. The slopes are steeper for the treated CCs, meaning a higher proportion of registered children received requisite services than children in control CCs in 2020. Like maternal nutrition services, this is likely driven by the fact that while nutrition services are rolled out across the country, the technical support and supervisory services of the HSSP focus on the Sylhet and Chattogram divisions. PART III – IMPACT OF HSSP ON ESSENTIAL NUTRITION SERVICES MATERNAL NUTRITION SERVICES Panel A in Figure 4 shows the impact of HSSP on maternal nutrition services between 2018 and 2020 and Panel B shows the heterogeneity of effects; that is, how CCs in Chattogram performed compared to those in Sylhet. We find that the HSSP led to a 2.7 percent increase in the proportion of registered eligible pregnant women who have been weighed and received at least 30 IFA Figure 4: Impact of HSSP on Maternal Nutrition Services (2018–2020) Source: Authors’ calculations Notes: IFA = Iron and folic acid. Figures show the impact (Panel A) and heterogeneity of effects (Panel B) of HSSP on maternal nutrition services between 2018 and 2020, derived using administrative data from the DHIS-2. Dependent variables are a normalized percentage of eligible pregnant women served by each CC. Heterogeneity of effects are calculated as the impact in Chattogram versus those in Sylhet. Coefficients are derived from a linear-censored (two-limit) Tobit difference-in-difference model using district-level fixed effects, interpretable as percentage changes; that is, a coefficient of 0.089 is interpretable as a 8.9 percent increase. 16 tablets and nutrition counseling (DLR 13.4). A higher proportion of CCs in Sylhet provided eligible pregnant women with all three services (by 6.2 percent) than in Chattogram. Next, assessing the impact on each component of the programmatic indicator, we find that the HSSP led to a 3.4 percent increase in the number of new eligible pregnant women who were registered for services between 2018 and 2020. CCs in Chattogram outperformed those in Sylhet by 7.5 percent. The HSSP led to 2.8 percent more women from treated areas being weighed during an ANC visit than in control ones. Similarly, the proportion of women receiving at least 30 IFA tablets and nutrition counseling was higher among the treated CCs by 3.0 and 2.7 percent, respectively, than among the control CCs. In all three instances, treatment effects were stronger in Chattogram than in Sylhet by 4.3, 5.3, and 4.3 percent, respectively. Overall, HSSP is seen to have positive effects on the programmatic indicator and each of its components between 2018 and 2020. Analysis of the heterogeneity of impact across the two treated divisions show that CCs in Chattogram provided more services than those in Sylhet for nearly all outcomes. CHILD NUTRITION SERVICES Impact estimates of HSSP on child nutrition services are presented in Figure 5. Estimates presented are coefficients of a linear-censored (two-limit) Tobit difference-in-difference model using district-level fixed effects. Results in Figure 5, Panel A show the impact between 2018 and 2020, while those in Panel B show the heterogeneity of impact; that is, how CCs in Chattogram performed in comparison to those in Sylhet. Participation in the HSSP led to a significant improvement in the programmatic indicator DLR 14.4, representing the proportion of registered eligible children below 24 months, who have been provided age-appropriate nutrition counseling, by 8.9 percent between 2018 to 2020. Analysis of heterogeneity suggests that overall, CCs in Chattogram provided services to 3.7 percent more registered children than CCs in Sylhet. Next, we assess the impact of HSSP on each of the components that make up the programmatic indicator separately. Participation in the HSSP led to a 5.9 percent increase in the rate of registration of eligible children aged below six months, between 2018 and 2020. Overall, CCs in Sylhet registered 5.2 percent more children below six months over the same duration than did CCs in Chattogram. In terms of receiving nutrition counseling, however, children from this cohort did not experience a significant increase in HSSP locations over the control ones. On average, the rate of registration of eligible children aged between 6 to 23 months rose among the treated by 4.6 percent over the control between 2018 to 2020. During this time, CCs in Sylhet registered 3.7 percent more children than CCs in Chattogram. Similarly, 1.5 percent more children of this age group received age-appropriate nutrition counseling in HSSP locations over control ones. CCs in Sylhet served 0.8 percent more children than did CCs in Chattogram. Next, we find that participation in the HSSP led to a higher proportion of children being screened for stunting, wasting, and overweight by 0.8 percent each, between 2018 and 2020. CCs in Sylhet performed better than those in Chattogram. 17 Overall, child nutrition services for all age groups in treated areas experienced significant gains between 2018 and 2020. Significant impact was seen on the core programmatic indicator reflecting the proportion of eligible registered children who were provided age-appropriate nutrition services. Effects on the rate of registration and provision of pertinent services to children Figure 5: Impact of HSSP on Child Nutrition Services (2018–2020) Source: Authors’ calculations Notes: Figures show the impact (Panel A) and heterogeneity of effects (Panel B) of HSSP on child nutrition services between 2018 to 2020, derived using administrative data from the DHIS-2. Dependent variables are normalized percent of eligible children served by each CC. Heterogeneity of effects are calculated as the impact in Chattogram versus those in Sylhet. Coefficients derived from a linear-censored (two-limit) Tobit difference-in-difference model using district-level fixed effects, interpretable as percentage changes; that is, a coefficient of 0.089 is interpretable as a 8.9 percent increase. between 0 and 5 months were stronger than for children aged between 6 and 23 months. Similarly, the treatment effects on screening for stunting, wasting, and overweight for children aged below 24 months were positive and significant. Analysis of the heterogeneity of impact suggests that CCs in Chattogram, in general, outperformed those in Sylhet. STUDY LIMITATIONS This paper is not without limitations. First, the absence of a randomized rollout and lack of data beyond service records (thereby preventing the construction of a synthetic control group or using matching techniques as an identification strategy) weakens the claims to causality to a certain extent. However, we control for district-level heterogeneity in the models and employ a difference- in-difference approach to manage potential biases. Second, administrative data, especially in resource and capacity-scarce settings, typically suffer from quality issues due to a variety of reasons. However, as these issues are qualitatively expected to remain consistent across both treated and control areas, we assume that controlling for time-trends in our models will negate this issue to a large extent. Third, as the background activities to operationalize the HSSP began 18 in 2017 in the treated areas, data from 2018 are, in essence, not the true baseline. The estimates presented here, therefore, are likely lower-bound effects of the HSSP. Lastly, in parallel to improving the quality and quantity of nutrition service delivery, the HSSP also incentivizes higher rates of reporting. Therefore, some of the impact detected may be driven by higher reporting of services. However, the adoption of a conservative approach; that is, winsorizing the top-five percentile of the service data can be expected to mitigate the overreporting bias to a certain extent. PART IV – DISCUSSION AND CONCLUSION Using annual administrative data from 13,855 Community Clinics (CCs) from across the country between 2018 and 2020, we find a significant improvement in the coverage of maternal and child nutrition services in Chattogram and Sylhet in comparison to the rest of the country. The proportion of eligible pregnant women who received requisite antenatal services increased by 2.7 percent due to the HSSP over this duration in comparison to the control. This is driven by a positive impact on each of the components, the proportion of eligible women who are registered (3.4 percent) and have received each of the three antenatal services (weighing [2.7 percent], nutrition counseling [2.7 percent], and iron and folic acid tablets [3.0 percent]). Analysis of the heterogeneity of impact across the two treated divisions shows that CCs in Chattogram, in general, provided more services than those in Sylhet. This is likely driven by the fact that two of the four districts in Sylhet are located in the country’s wetland area (Haor region), typically characterized by their remote and sequestered nature, thereby hindering access to the facilities and services. A large proportion of eligible women of the remaining two districts live and work in tea plantations, generally situated in the outskirts of the region, further dampening the demand for essential nutrition services. In contrast, only 3 of 11 districts in the Chattogram region are in the remote hilly terrains. Similarly, we also find a significant improvement in the coverage of child nutrition services. The proportion of registered eligible children aged between 0 and 23 months who received age- appropriate nutrition counseling has increased by 8.9 percent between 2018 and 2020 in comparison to the control. Like maternal nutrition services, the impact on the indicator is driven by the positive effect on some of its components. While the registration of eligible children among both age groups (0 to 5 months and 6 to 23 months of age) has increased significantly by 5.9 and 4.6 percent, respectively, the provision of nutrition counseling was only significant for the latter group (by 1.5 percent). In parallel, the proportion of children who were screened for stunting, wasting, or underweight showed a significant increase in the HSSP divisions (by 0.8 percent each, respectively). With regard to the heterogeneity of effects, in contrast to the maternal nutrition services, CCs in Sylhet, in general, outperformed those in Chattogram. While this is a conundrum, discussions with field experts suggest that while a large proportion of the women working in remote locations such as tea plantations do not seek services for themselves, they are significantly more proactive when it comes to their children. This dichotomy, however, merits further investigation. There are several factors that contributed to the impact on nutrition service indicators. First, the HSSP’s DLI approach brings renewed focus and priority to service delivery and promotion of behavior change in maternal and child nutrition services within the broader context of the delivery 19 of the health service package. Second, the HSSP provides technical assistance to NNS through the NIPU team on two aspects—capacity development and improving monitoring and reporting of nutrition services by strengthening and mainstreaming nutrition data into the District Health Information System-2 (DHIS-2). The capacity-development component builds knowledge and skills on delivery of nutrition services through regularly scheduled learning and refresher sessions. On the data front, it improves awareness of the importance of data in improving implementation; the TA develops skills of relevant teams in extracting, analyzing, and curating the results for presentation; and provides support to strengthen the data ecosystem—thereby strengthening program capacity for monitoring and results verification. The NIPU supports the system functionaries in following up with low-performing districts, subdistricts, and CCs; flagging operational bottlenecks; and identifying solutions and planning actions to mitigate emerging issues. The HSSP’s approach has evolved over time in the way the program is implemented. For example, given the way the core programmatic indicators were framed, the focus was on maximizing the proportion of registered individuals who were provided services, which led to the CCs fulfilling the target at the expense of maximizing the number of individuals both enrolled and served through the program. This feedback led to corrective action to focus on increasing both the registration rate and service coverage of eligible individuals. There are, however, issues impeding service delivery of nutrition services. Delivering nutrition services is only one of the many responsibilities of the CHCPs. As a result, the CHCPs are often faced with emerging and competing priorities. During the COVID-19 pandemic, a large number of the CHCPs were co-opted to support the vaccination efforts. As the CCs are supported by Health Assistants and Family Welfare Assistants in addition to the CHCPs, engaging their support will help increase the reach of essential nutrition services. Quality assessment reports suggest that there is a lack of community mobilization structures and mechanisms to improve outreach and generate demand for nutrition services among eligible women and children. Soliciting systematic support of additional field cadres, such as the Multi-Purpose Health Volunteers, may notably improve the situation. Data and IT issues persist. A recent report from the Independent Monitoring and Evaluation Division (IMED), an independent Government of Bangladesh monitoring agency, reported that a large proportion of laptops being operated by the CHCP for service data entry are nonfunctional. As a result, the CHCPs use their personal mobile phones in place of the laptops for such purposes. The situation is further exacerbated by unreliable data connections, particularly in the remote locations in the Chattogram Hill Tracts. Therefore, addressing the IT issues, further upskilling the CHCP, and introducing stronger supervision and quality control processes is critical. The HSSP’s explicit focus on delivery of both maternal and child nutrition services through a strong emphasis on data monitoring and its use for planning and action has contributed to improving the outreach to eligible beneficiaries and for coverage of nutrition services in program areas. The gaps identified above need to be actively addressed by the government of Bangladesh to accelerate improvements, which will provide further impetus to enhancing service delivery and reducing malnutrition in Bangladesh. 20 REFERENCES Ahmed, Md. Shabbir. 2021. “Mapping the Prevalence and Socioeconomic Predictors of Low Birth Weight among Bangladeshi Newborns: Evidence from the 2019 Multiple Indicator Cluster Survey.” International Health. Doi: https://doi.org/10.1093/inthealth/ihab048. Arifeen, Shams, Aliki Christou, Laura Reichenback, Afrina F. Osman, Kishwar Azad, Khaled S. Islam, Faruque Ahmed, et al. 2013. “Community-Based Approaches and Partnerships: Innovations in Health-Service Delivery in Bangladesh.” The Lancet 382 (9909): 2012– 2026. Doi: https://doi.org/10.1016/S0140-6736(13)62149-2. Ash, D., Z. Mahmud, K. Kappos, S. Ireen, and T. Forissier. 2020. “Delivery of Maternal Nutrition Interventions at Scale and Mainstreaming into the Health System in Bangladesh.” Field Exchange Issue 63 (October). www.ennonline.net/fex. BBS (Bangladesh Bureau of Statistics). 2019. “Preliminary Report on the Household Income and Expenditure Survey (HIES) 2016–17.” https://tinyurl.com/yfbuhfoq. Black, Robert E., C. G. Victora, S. P. Walker, Z. A. Bhutta, P. Christian, M. de Onis, et al. 2013. “Maternal and Child Undernutrition and Overweight in Low-Income and Middle-Income Countries.” The Lancet 382 (9890): 427–51. Blaine, Bruce. 2018. “Winsorizing.” The SAGE Encyclopedia of Educational Research, Measurement, and Evaluation (Vols. 1–4). Thousand Oaks, CA: Sage Publications, Inc. https://dx.doi.org/10.4135/9781506326139. Cameron, Colin, A., and Pravin K. Trivedi 2010. Microeconometrics Using STATA. College Station, TX: Stata Press, Stata Corp LP. ISBN: 978-59718-073-3. FAO, IFAD, UNICEF, WFP and WHO (Food and Agriculture Organization, International Fund for Agricultural Development, United Nations Children’s Fund, and World Health Organization). 2020. The State of Food Security and Nutrition in the World 2020. Transforming Food Systems for Affordable Healthy Diets. Rome: FAO. https://doi.org/10.4060/ca9692en. Govindaraj, Ramesh, Dhushyanth Raju, Federica Secci, Sadia Chowdhury, and Jean-Jacques Frere. 2018. “Health and Nutrition in Urban Bangladesh: Social Determinants and Health Sector Governance. Directions in Development.” Washington, DC: World Bank. Hasan M. M., J. Uddin, M. H. Pulok, N. Zaman, and M. Hajizadeh. 2020. “Socioeconomic Inequalities in Child Malnutrition in Bangladesh: Do They Differ by Region?” Int J Environ Res Public Health 817, no. 3 (Feb.): 1079. Hong, R., and M. Ruiz-Beltran. 2007. “Impact of Prenatal Care on Infant Survival in Bangladesh.” Matern Child Health J 11: 199–206. Honoré, Bo E. 1992. "Trimmed Lad and Least Squares Estimation of Truncated and Censored Regression Models with Fixed Effects." Econometrica 60 (3): 533–65. Huda, T. M., A. Alam, and T. Tahsina, M. M. Hasan, A. Iqbal, J. Khan, G. Ara, et al. 2020. “Shonjibon Cash and Counselling: A Community-Based Cluster Randomised Controlled Trial to Measure the Effectiveness of Unconditional Cash Transfers and Mobile Behaviour Change Communications to Reduce Child Undernutrition in Rural Bangladesh.” BMC Public Health 20, 1776 (Nov.). https://doi.org/10.1186/s12889-020- 09780-5. 21 Khan, M. A., N. Khan, O. Rahman, G. Mustagir, K. Hossain, R. Islam, H. T. A. Khan. 2021. “Trends and Projections of Under-5 Mortality in Bangladesh including the Effects of Maternal High-Risk Fertility Behaviours and Use of Healthcare Services.” PLoS ONE 16 (2): e0246210. Kim, S. S., R. Rawat, E. M. Mwangi, R. Tesfaye, Y. Abebe, J. Baker, E. A. Frongillo, et al. 2016. “Exposure to Large-Scale Social and Behavior Change Communication Interventions Is associated with Improvements in Infant and Young Child Feeding Practices in Ethiopia.” PLoS One 11 (10): e0164800. Kraay, A. 2019. The World Bank Human Capital Index: A Guide. Oxford University Press on behalf of the World Bank. https://tinyurl.com/y3pq87ur. Kuhnt, J., and S. Vollmer. 2017. “Antenatal Care Services and Its Implications for Vital and Health Outcomes of Children: Evidence from 193 Surveys in 69 Low-Income and Middle-Income Countries.” BMJ Open 7: e017122. doi: 10.1136/bmjopen-2017-017122. Long, J. S. 1997. “Regression Models for Categorical and Limited Dependent Variables.” Thousand Oaks, CA: Sage Publishing. Mostafa, Ishita, Mahamudul Hasan, Subhasish Das, Soroar Hossain Khan, Md. Iqbal Hossain, Abu Faruque, and Tahmeed Ahmed. 2021. “Changing Trends in Nutritional Status of Adolescent Females: A Cross-Sectional Study from Urban and Rural Bangladesh.” BMJ Open 11 (2): e044339. https://doi.org/10.1136/bmjopen-2020-044339. NIPORT (National Institute of Population Research and Training), Mitra and Associates, and ICF International. 2008. Bangladesh Demographic and Health Survey 2007. Dhaka, Bangladesh, and Calverton, MD. ———. 2013. Bangladesh Demographic and Health Survey 2011. Dhaka, Bangladesh and Calverton, MD. ———. 2016. Bangladesh Demographic and Health Survey 2014. Dhaka, Bangladesh, and Rockville, MD. ———. 2019. Bangladesh Demographic and Health Survey 2019. Dhaka, Bangladesh, and Rockville, MD. Nguyen, P. H., 2017. “Integrating Nutrition Interventions into an Existing Maternal, Neonatal, and Child Health Program Increased Maternal Dietary Diversity, Micronutrient Intake, and Exclusive Breastfeeding Practices in Bangladesh: Results of a Cluster-Randomized Program Evaluation.” J Nutr. 147, no. 12 (Dec.): 2326–37. Riaz, B. K., L. Ali, S. K. Ahmad, Z. M. Islam, R. K. Ahmed, and S. Hossain. 2020. “Community Clinics in Bangladesh: A Unique Example of Public-Private Partnership.” Heliyon 6 (5): e03950. World Bank. 2021. World Bank Microdata Library. https://tinyurl.com/yj8uaahc. 22 ANNEX Figures Figure A1: Dispersion of Outcomes (Child and Maternal Registration) Source: Authors’ calculations Note: The kernel density curves show dispersion of normalized child and maternal registration rates. 23 Figure A2: (Actual) Means and Distribution of Maternal Nutrition Services across Divisions and Years Source: Authors’ calculations Note: Figure shows CC-level distribution of reported child nutrition services with means. Services (Y-axis) are presented in the logarithmic form. Grey line represents the mean for each division by year. Figure A3: (Actual) Means and Distribution of Child Nutrition Services across Divisions and Years Source: Authors’ calculations Note: Figure shows CC-level distribution of reported child nutrition services with means. Services (Y-axis) are presented in the logarithmic form. Grey line represents the mean for each division by year. Annex Tables Table A1: Sensitivity Analysis of Maternal Nutrition Indicators CC-level fixed District-level fixed Outcomes effects effects (1) (2) % of eligible pregnant women Coeff. -0.001 0.017 registered Std. Err. (0.014) (0.013) % of eligible pregnant women Coeff. 0.045*** 0.035*** received all three services Std. Err. (0.008) (0.008) % of eligible pregnant women Coeff. 0.031*** 0.028*** weighed Std. Err. (0.011) (0.010) % of eligible pregnant women Coeff. 0.032*** 0.030*** received >30 IFA tabs Std. Err. (0.011) (0.010) % of eligible pregnant women Coeff. 0.030*** 0.027*** received nutrition counseling Std. Err. (0.011) (0.010) Source: Authors’ calculations Note: IFA = Iron and folic acid. Table shows coefficients from a linear difference-in-difference model derived using administrative data from the DHIS-2. Column 1 presents results using CC-level fixed effects, while Column 2 is estimated using district-level fixed effects. Outcomes are presented as the normalized percentage of pregnant women served by each CC. Standard errors are presented in parenthesis; *** / ** / * represent significance at the 1, 5, and 10 percent levels, respectively. Table A2: Sensitivity Analysis of Child Nutrition Indicators CC-level fixed District-level fixed Outcomes effects effects (1) (2) Panel A: Program indicator Proportion of registered children Coeff. 0.070*** 0.065*** provided services (DLR 14.4) SE (0.005) (0.005) Panel B: Children 0–5 months Coeff. 0.078*** 0.059*** % of eligible children registered Std. Err. (0.010) (0.010) % of registered children received Coeff. 0.006 0.019 counseling on exclusive Std. breastfeeding Err. (0.026) (0.021) Panel C: Children 6–23 months Coeff. 0.054*** 0.046*** % of eligible children registered Std. Err. (0.005) (0.005) % of registered children received Coeff. 0.012 0.015 counseling on complementary Std. feeding Err. (0.010) (0.009) Panel D: Children 0–23 months Coeff. 0.013*** 0.008*** % of eligible children screened for stunting Std. Err. (0.002) (0.002) Coeff. 0.012*** 0.008*** % of eligible children screened for wasting Std. Err. (0.002) (0.002) Coeff. 0.013*** 0.009*** % of eligible children screened for overweight Std. Err. (0.002) (0.002) Source: Authors’ calculations Note: Table shows coefficients from a linear difference-in-difference model derived using administrative data from the DHIS-2. Column 1 presents results using CC-level fixed effects, while Column 2 is estimated using district-level fixed effects. Outcomes are presented as the normalized percentage of children served by each CC. Standard errors are presented in parenthesis; *** / **/ * represent significance at the 1, 5, and 10 percent levels, respectively. Table A3: Summary Statistics and Means Tests of Maternal Nutrition Service Indicators Differenc Differenc 2018 2020 e e Treate (Ctrl - Contro Treate (Ctrl - Control d Trt) l d Trt) (1) (2) (3) (4) (5) (6) Panel A: Program indicator Proportion of registered pregnant women provided 0.206 0.280 -0.074 0.739 0.808 -0.068 services (DLR 13.4) Panel B: Enrollment and service provision % of eligible pregnant 0.092 0.066 0.026*** 0.417 0.397 0.019*** women registered % of eligible pregnant 0.073 0.054 0.018*** 0.368 0.348 0.020*** women weighed % of eligible pregnant women received >30 IFA 0.075 0.054 0.021*** 0.375 0.355 0.020*** tabs % of eligible pregnant women received nutrition 0.072 0.054 0.019*** 0.362 0.345 0.017*** counseling N 4,059 1,277 8,986 3,179 Source: Authors’ calculations Notes: IFA = Iron and folic acid. Table shows proportions of outcomes of interest at the CC level across treated and control divisions from 2018, 2019, and 2020, presented as a percentage of coverage of eligible population. Significance of differences (Columns 3, 6, and 8) are calculated using a t-test; *** / ** / * represent significance at the 1, 5, and 10 percent levels, respectively. Table A4: Summary Statistics and Means Tests of Child Nutrition Service Indicators Differen Differen 2018 2020 ce ce Contr Treate (Ctrl - Contr Treate (Ctrl - ol d Trt) ol d Trt) (1) (2) (3) (4) (5) (6) Panel A: Program indicator Proportion of registered children (0– 0.146 0.150 -0.003 0.611 0.805 -0.194 23m) provided services (DLR 14.4) Panel B: Children 0–5 months % of eligible children registered 0.482 0.392 0.090*** 0.434 0.383 0.050*** % of registered children received counseling on exclusive 0.129 0.097 0.033*** 0.289 0.260 0.029*** breastfeeding Panel C: Children 6–23 months % of eligible children registered 0.286 0.231 0.055*** 0.251 0.232 0.019*** % of registered children received counseling on complementary 0.046 0.033 0.013*** 0.205 0.200 0.005 feeding Panel D: Children 0–23 months % of eligible children screened for 0.053 0.039 0.012*** 0.053 0.046 0.007*** stunting % of eligible children screened for 0.051 0.040 0.012*** 0.051 0.045 0.006*** wasting % of eligible children screened for 0.057 0.044 0.014*** 0.058 0.050 0.008*** overweight N 8,336 2,383 7,072 2,609 Source: Authors’ calculations Note: Table shows proportions of outcomes of interest at the CC level across treated and control divisions from 2018, 2019, and 2020, presented as a percentage of coverage of eligible population. Significance of differences (Columns 3, 6, and 8) are calculated using a t-test. *** / ** / * represent significance at the 1, 5, and 10 percent levels. Despite progress over the past two decades, poor nutrition remains a significant public health challenge in Bangladesh. Stunting among children under five years declined from 43 to 31 percent between 2007 and 2018, while 42 percent of women between 15 to 49 years are anemic. The Ministry of Health and Family Welfare is implementing the Health Sector Support Project (HSSP) with financing and technical assistance from the World Bank in the Sylhet and Chattogram divisions of the country. The project considers improving nutrition outcomes as a core priority. Leveraging administrative data from 13,855 community clinics (CCs) from 2018 to 2020 and a difference-in-difference approach, the analysis finds that HSSP led to improvements in the delivery of both maternal and child nutrition services. The proportion of eligible pregnant women who received requisite antenatal services (receipt of at least 30 iron and folic acid tablets, nutrition counseling, and weight measurement) increased by 2.7 percent over the duration. Similarly, the proportion of children between 0 and 23 months, who received age-appropriate nutrition counseling, increased by 8.9 percent over the same period. The paper identifies several factors that led to these improvements and notes the impediments. The HSSP renewed focus on the importance of delivering quality nutrition services, and the technical assistance provided through the HSSP has strengthened capacity, not only around the delivery of services but also in improving the data ecosystem and quality of project monitoring and results verification. There are, however, issues impeding service delivery of nutrition services. The Community Health Care providers (CHCPs) are often faced with competing priorities, as nutrition is one of the many services they provide. Similarly, the CHCPs have been found to lack the required capacity and skills in delivering services and are also burdened with poor Information Technology (IT) equipment. ABOUT THIS SERIES: This series is produced by the Health, Nutrition, and Population Global Practice of the World Bank. The papers in this series aim to provide a vehicle for publishing preliminary results on HNP topics to encourage discussion and debate. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations or to members of its Board of Executive Directors or the countries they represent. Citation and the use of material presented in this series should take into account this provisional character. For free copies of papers in this series please contact the individual author/s whose name appears on the paper. Enquiries about the series and submissions should be made directly to the Editor Martin Lutalo (mlutalo@ worldbank.org) or HNP Advisory Service (askhnp@worldbank.org, tel 202 473- 2256). For more information, see also www.worldbank.org/hnppublications. 1818 H Street, NW Washington, DC USA 20433 Telephone: 202 473 1000 Facsimile: 202 477 6391 Internet: www.worldbank.org E-mail: feedback@worldbank.org