88665 Muthukumara Mani Limin Wang Copyright © 2014 The International Bank for Reconstruction and Development / The World Bank Group 1818 H Street, NW Washington, DC 20433, USA All rights reserved The findings, interpretations, and conclusions expressed in this report are entirely those of the authors and should not be attributed in any manner to the World Bank, or its affiliated organizations, or to members of its board or executive directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do not imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. Muthukumara Mani Limin Wang CONTENTS FOREWORD 7 ACKNOWLEDGMENTS 9 EXECUTIVE SUMMARY 11 Organization of the report 11 A summary of the key findings 12 CHAPTER 1: HEALTH AND CLIMATE CHANGE PATHWAYS AND THE BANGLADESH CONTEXT 14 Pathways from Climate to Health  15 The Bangladesh Context  16 Climate conditions and climate change projections  16 Major health issues  17 CHAPTER 2: QUANTIFYING THE HEALTH IMPACT OF CLIMATE VARIABILITY ON CHILDHOOD ILLNESSES 20 Key Data Sources  21 Weather data  21 The Bangladesh national health surveys  22 Quantifying the Health Impact of Climate Variability  24 The analytical model  24 Seasonality of disease incidence  24 The impact of climate on disease incidence  26 Impact of household environmental conditions on health 29 Summary of Results 30 Projecting Future Health Burden 30 How Much Climate Impact Can Be Mitigated through Development? 32 Key Messages  33 CHAPTER 3: VECTOR-BORNE DISEASES: HOTSPOTS AND CLIMATE LINKAGES 34 The Epidemiology of Vector-Borne Diseases 36 Key Data Sources and Issues for VBDs 36 The Temporal and Spatial Distribution of VBDs  37 Temporal and seasonal trends 37 Disease hotspots 38 Statistical Correlation between Climate Conditions and VBDs  39 Dengue fever 39 Malaria39 Kala-azar40 Summary of Results  41 Key Messages 41 CHAPTER 4: POPULATION DYNAMICS AND SPATIAL TARGETING: IMPLICATIONS FOR CLIMATE CHANGE AND HEALTH  42 Key Features of Population Dynamics  43 Measuring Health Adaptation Capacity  45 Assessing the Efficiency of Geographic Targeting  46 Key Messages 48 CHAPTER 5: COST-EFFECTIVENESS ANALYSIS OF HEALTH ADAPTATION INTERVENTIONS  50 Why Cost-Effectiveness Analysis Rather Than Cost-Benefit Analysis?  51 An Application of Cost-Effectiveness Analysis in Health Adaptation Interventions  52 Benefit estimation  52 Cost estimation  53 Cost-effectiveness analysis: An illustration  54 Key Messages 55 CHAPTER 6: CONCLUSIONS AND WAY FORWARD 56 Key Messages 57 Way Forward 57 APPENDIX A: DATA SOURCE FOR WATER- AND VECTOR-BORNE DISEASES 58 APPENDIX B: CLIMATE AND NATURAL DISASTER DATA  60 APPENDIX C: NATIONAL HEALTH AND FACILITY SURVEY 61 Health Facility Survey 61 Urban Health Survey 61 APPENDIX D: ANALYTICAL RESULTS 62 APPENDIX E: POLICY SIMULATION OF HEALTH IMPAC 68 REFERENCES 70 Boxes Table 2.8  Health Impact Simulation by 2030: Development vs. Climate Change Box 1.1 Weather and Climate Variability vs. Climate Change Table 3.1 Key Data Sources for VBDs Box 2.1 The Model Table 3.2  Spearman Rank Correlation of Dengue Cases and Climatic Box 2.2 Key Assumptions for Projection of Disease Burden by 2050 Factors Box 3.1 Major VBDs in Bangladesh Table 3.3  Spearman Correlation between Malaria Cases and Box 4.1 Issue of Sampling Bias Climate Conditions Box 5.2 Cost of Adaption Options Table 3.4  Spearman Rank Correlations of Kala-azar Cases and Box 5.3 Two Major Nutrition Projects of Inclusive Impact Climatic Factors Table 4.1  Spatial Analysis: Disease Incidence vs. Local Capacity (2011) Figures Potential Bias of Health Indicators of Urban Population: Table 4.1.1  An Example of Dhaka Figure 1.1 Pathways from Climate Change to Health Table 4.2  Pearson Correlation Coefficient between Adaptation Figure 2.1 Average Monthly Weather Variables by Region and Month Capacity and Disease Incidence: 2011 Figure 2.2 Maps of 2004 and 2007 DHS Survey Locations Table 5.1 Summary of Methods for Benefit Quantification Figure 2.3 Spatial Distribution of Disease Incidence, Temperature, Table 5.2 Major Cyclones (1970-2009) and Floods by 2050 Table 5.3  Cost Benefit Analysis: Comparison of Three Adaptation Figure 3.1 Trends and Monthly Disease Patterns of VBDs Options Figure 3.2 Geographic Distributions of VBDs in 2011 Table A.1 Health MIS: Diseases and Coverage Figure 4.1 Population Density and Growth Table A.2 IEDCR Data Sources Figure 4.2 Spatial Distribution of Overall Health Adaptation Capacity Table A.3 Diarrheal Disease and Enteric Infection Surveillance Figure 4.3 Correlation between Health Facility and Disease Table A.4 Bangladesh Demographic and Health Survey (DHS) Incidence: 2010 District Level Data Table B.1 Weather and Disaster Management Data System Table C.1  The Number of Health Facilities Selected for the 2011 Tables Health Facility Survey Table D.1 Logit Regression Results Using Full Sample (Odds Ratio) Table 1.1 Estimated Total Deaths by Cause: Bangladesh Table D.2  Logit Regression Results Using Urban Households (Odds Table 1.2 Child Health: Incidence of Illnesses and Malnutrition (%) Ratio) (Random-Effect Estimation) Table 1.3 Major Infectious Diseases: Cases and Deaths Table D.3  Logit Regression Results Using Rural Households (Odds Table 2.2 Household Distribution in 2004 and 2007 DHS by Region Ratio) and Survey Month Table D.4  Estimation of Marginal Impact on the Incidence of Table 2.3 Summary of Key Variables in 2004 and 2007 DHS Childhood Illness and Malnutrition: Odds Ratio from 2011 Table 2.4 Disease Incidence by Survey Month and Season: 2004 DHS and 2007 DHS Table D.5  District-Level Population Change between 2001 and 2010 Table 2.5 The Impact of Climate Variables on Disease Incidence: Table D.6  Spatial Analysis: Disease Incidence versus Local Capacity, 2004 and 2007 DHS 2011 Table 2.6 The Impact of Water and Sanitation Facilities on Disease Incidence (2004 and 2007): Odds Ratio Table 2.7 Health Burden Projection by 2050 Abbreviations HFA Height for age AIDS Acquired immunodeficiency syndrome HFS Health Facility Survey ARI Acute respiratory infection HIV Human immunodeficiency virus BCCRF Bangladesh Climate Change Resilience Fund ICDDR,B International Center for Disease and Diarrhoeal Research, BDHS Bangladesh Demographic and Health Survey Bangladesh BINP Bangladesh Integrated Nutrition Project BMD Bangladesh Meteorological Department IEDCR  Institute of Epidemiology, Disease Control, and Research BWDB Bangladesh Water Development Board MDG Millennium Development Goal CBA Cost-Benefit Analysis MIS Management Information System CCHPU Climate Change and Health Promotion Unit CDC Communicable Disease Control MoHFW Ministry of Health and Family Welfare CEA Cost-Effectiveness Analysis NIPORT  National Institute of Population Research and Training CPP Cyclone Preparedness Program NNP National Nutrition Project DALY Disability-adjusted life years DGHS Directorate General of Health Services PCDS Priority communicable disease surveillance DHS Demographic and Health Survey SS Sentinel surveillance DMB Disaster Management Bureau TRMM Tropical Rainfall Measuring Mission DMIN Disaster Management Information Network ENSO El Nino-Southern Oscillation UHS Urban Health Survey ESD Essential Service Delivery USAID U.S. Agency for International Development FFWC Flood Forecasting and Warning Centre VBD Vector-Borne Disease GBM Ganges-Brahmaputra-Meghna GDP Gross domestic product WFH Weight For Height GIS Geographic information system WHO World Health Organization 7 FOREWORD Bangladesh is one of the countries that is most vulnerable to climate change. It is already facing enormous challenges due to extreme events such as droughts, land and coastal flooding, and other extreme weather events. Added to these challenges are demographic and socio-economic factors, such as rapid population growth and urbanization, poor health conditions, water scarcity and inadequate sanitary conditions. In this context, climate change is an additional stressor that is expected to increase the burden of diseases, resulting in increased morbidity and mortality. The Government of Bangladesh has recognized the important challenge facing the country, and is taking a number of steps to address it. The 2008 Bangladesh Climate Change Strategy and Action Plan highlights the need for implementing surveillance systems for existing and new disease risks and ensuring health systems are prepared to meet future demands. Bangladesh became one of the first countries to establish a Climate Change and Health Promotion Unit (CCHPU) under the Ministry of Health and Family Welfare tasked to conduct research and evaluate and monitor programs related to health promotion and climate change. The Ministry of Environment has identified the need to conduct an in-depth nationwide study focusing on climate-sensitive diseases to fill in the important knowledge gap in the area of health in the context of climate change. This study was jointly undertaken by the Climate Change and Health Promotion Unit of the Ministry of Health and Family Welfare, the International Centre for Diarrheal Disease Research, Bangladesh, and the World Bank. This study had two broad objectives: (1) to assess national vulnerability and impact on major diseases of increased climate variability and extreme events in Bangladesh; and (2) to assess existing institutional and implementation capacity, financial resources at the local level, and existing public programs targeted at climate-sensitive diseases. Three key messages emerge from this study: zz First, the health impacts of increased climate variability and extreme weather events are projected to be significant by 2050, but well-targeted development investments can mitigate the excess health burden attributable to climate change. zz Second, rapid urbanization and a growing urban slum population are quickly changing the population dynamics in Bangladesh, and this has implications for climate-induced health risks. zz Third, given the seasonality effects and the role of confounding factors, the allocation of public resources to deal with climate health risks in the future should be spatially targeted to reach locations that are likely to be at high climate and health risk to ensure cost-effectiveness. Overall, climate change imposes a considerable additional burden on Bangladeshi society, and this burden falls dispropor- tionately on the vulnerable poorer groups of population having lower adaptive capacity. It is my hope that this study contrib- utes to a sound understanding of the health impacts of climate change in the context of Bangladesh and supports policy- makers in their efforts to address these impacts. This study would not have been possible without valuable inputs from the Government of Bangladesh, non-governmental organizations, and research and academic institutions in Bangladesh, and we are grateful for their contributions. Johannes Zutt World Bank Country Director, Bangladesh 9 ACKNOWLEDGMENTS This study was undertaken jointly by the Climate Change and Health Promotion Unit of the Ministry of Health and Family Welfare, the International Centre for Diarrhoeal Disease Research, Bangladesh, and the World Bank. The study was led by Muthukumara Mani (task team leader and senior environmental economist), Disaster Risk Management and Climate Change Unit in the South Asia Sustainable Development Department (SASDC) under the guidance of Bernice K. Van Bronkhorst (sector manager, SASDC). The core team comprised Limin Wang, Shaphar Selim, Ahmed-Al Sabir, Sharif Hossain, Mohammad Shafiul Alam, and Wasif Ali Khan. The team greatly appreciates guidance provided by Kabir Iqbal, coordinator of the Climate Change and Health Promotion Unit of the Ministry of Health and Family Welfare, and Peter Kim Streatfield, head of the Population Programme, ICDDR,B, at various stages and all the seminar participants in the World Bank Dhaka office and Washington, DC. John Henry Stein, sector director of the South Asia Sustainable Development Department, and Johannes Zutt, country director for Bangladesh, guided the overall effort. In preparing the study, the team benefited greatly from detailed comments received from peer reviewers Kirk Hamilton, Anil Markandya, and Kanta Kumari Rigaud. The team would especially like to thank Jin Di, Hrishi Patel, Keith Miao, Mahmud Khan and Sushenjit Bandyopadhyay for their excellent guidance and assistance in data preparation and mapping work. The team was ably assisted by Angie Harney, Lilian MacArthur, and Marie Florence Elvie. Finally, the team gratefully acknowledges the financial support provided by the Bangladesh Climate Change Resilience Fund (BCCRF), a multi-donor trust fund supported by the governments of the United Kingdom, Denmark, Sweden, Switzerland, Australia, the United States, and the European Union. 11 EXECUTIVE SUMMARY Bangladesh is one of the most climate- adaptation policies and programs in area of climate change and health. vulnerable countries in the world. Bangladesh in the next 10 to 15 years. Chapter 2 carries out an empirical Climate variability and extreme weather The key objectives of this study are (1) analysis to estimate the impact of events, such as inland and coastal to identify the temporal and spatial climate variability on the incidence floods, droughts, tropical cyclones, and distribution of major water-borne and of three major childhood illnesses storm surges, are projected to become vector-borne diseases; (2) to assess (diarrhea, acute respiratory infection more frequent and severe as average the health impact of both climate or ARI, and fever), controlling temperatures rise with climate change variability and socio-environmental confounding factors. These childhood (MoEF 2008). Added to the climate risks conditions and project the future health illnesses are major causes of under- are rapid but unplanned urbanization, burden of climate variability; and (3) to five deaths, accounting for about with a growing slum population, identify cost-effective health adaptation a quarter of total child deaths in inadequate access to safe drinking interventions. The study benefitted Bangladesh. Future vulnerability water and sanitary facilities, high levels greatly from the input provided by the and health burdens associated with of poverty and population density, BCCRF Development Partners, and it was these illnesses are projected, and a and high prevalence of malnutrition finalized after due review and clearance simulation exercise is carried out to and disease incidence among children. from the BCCRF Management Committee. quantify how much investment in Without doubt, the adverse health health adaptation interventions can consequences of increased climate risks Given that a significant amount mitigate the adverse impacts of climate are likely to worsen the situation if of financial resources will be on health. well-targeted and cost-effective health committed to the area of climate adaptation measures are not put in change adaptation and resilience Chapter 3 presents the seasonal and place now. in Bangladesh, CEA should be used spatial patterns of vector-borne routinely in allocating resources to diseases (dengue fever, malaria, Over the past few decades, with programs and projects. The results of and kala-azar) and investigates the support from donor agencies, CEA should be disseminated widely statistical links between climate Bangladesh has invested extensively among stakeholders as a means to conditions and cases of vector- in health and education, basic improve transparency as well as to borne disease in high-prevalence environmental services, and disaster develop an evidence-based decision- regions in Bangladesh. These three risk reduction measures, all of which making process for selecting of projects vector-borne diseases, although less are central to improving health or programs and identifying new areas significant as a share of total health outcomes. However, one pertinent of policy interventions. burden, are predicted to become question is whether major investment larger health concerns in the future decisions will be cost-effective and The methodology includes multilevel as climate change affects the patterns well-targeted, particularly in areas (hierarchical) modeling and econometric of seasonality, temperature, and that are most vulnerable to health and analysis using household survey data, precipitation, which are likely to climate risk in the coming decades. correlation analysis of disease incidence become more conducive to disease and climate variables, vulnerability outbreaks. The health impact of climate variability assessment using geographic targeting, and extreme weather events is a and cost-effective analysis of possible Chapter 4 analyzes the evolution of major area highlighted in the 2008 adaptation interventions. population dynamics over the past Bangladesh Climate Change Strategy decade and assesses vulnerability and Action Plan. This study was using geographic targeting to compare requested by the Government of Organization of the report health adaptation capacity with Bangladesh and endorsed by the This study consists of five chapters. disease incidence. Chapter 5 presents BCCRF Management Committee as a Chapter 1 provides an overview of an illustration of cost-effectiveness priority research project to inform the pathways between health and climate analysis for ranking different health design and implementation of health and the Bangladesh context in the adaption interventions, including 12 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? disaster and risk management gap on cost-effectiveness of resources. One of the key features of investment, expansion of access to various control and management population dynamics in Bangladesh safe water and sanitation services, programs. Using monthly surveillance is the large rural-urban migration and investment in nutrition-focused data in regions with a high incidence and the fast growth of urban slum programs. The final chapter presents of VBDs, this study reveals strong areas where about a third of the conclusions and recommendations. patterns of seasonality for VBDs, but urban population currently reside. no clear trends over the past decade. The spatial assessment conducted A summary of the key findings The strong statistical correlation in this study, which uses data from between short-term climate variability population censuses, provides Impact of climate variability on and VBD cases is also established. In strong evidence of poor targeting of childhood diseases is significant and the case of dengue fever and malaria, public investment, with districts of varies by season, but investment in all climate variables (temperature, high disease prevalence particularly traditional areas of development rainfall, and humidity), both current lacking in access to health and basic can mitigate to a great extent the and lagged, have a strong correlation environmental services. excessive health burden attributable with the disease caseload. However, to climate change. The results from The total population of Bangladesh is only temperature is significantly projected to increase by 64.6 million a combined data set that integrates correlated with kala-azar cases. data from geographically referenced between 2010 and 2030, reaching national health surveys with data from about 217.9 million people in 2030, Although VBDs are projected to with three-fourths of that growth local weather stations located across increase with climate change in the country, spanning more than 30 expected to occur in urban areas. Bangladesh, as highlighted in Therefore, the health implications years, confirm a significant impact of the government’s 2008 Climate climate variability on the incidence of of population dynamics and rapid Change Strategy and Action Plan, it population growth in urban slums three childhood illnesses (diarrhea, is important to recognize that VBDs fever, and ARI). The health impact areas should be rigorously assessed. account for a significantly small Policy makers need to recognize the of climate variability differs greatly proportion of total health burden, between pre-monsoon and monsoon full scale of the health threat that is compared with water-borne diseases posed by rapid but poorly planned seasons. The estimated impact of an such as diarrhea or malnutrition. On incremental change in climate variables urbanization in Bangladesh. average, about 1 and 0.9 percent of (temperature, humidity, and rainfall) deaths are reported from malaria and Cost-effectiveness analysis (CEA) is relatively small, but the event of dengue fever, respectively, compared should be fully implemented extreme precipitation has a large with 10.7 percent from diarrhea by both government units and and statistically significant impact on (even higher when taking account of donor agencies to inform the disease incidence. malnutrition). Priorities in the short allocation and prioritization Lack of access to sanitation facilities, term should therefore be placed on of public resources. Despite the particularly in urban areas, is improving the collection of data on significant investment in health over identified as a key confounding VBDs from both public and private the past two decades in Bangladesh, factor determining the incidence of health facilities to identify changes in evidence on the cost-effectiveness of seasonal patterns and the geographic different interventions is lacking. The childhood illnesses (fever and ARI). distribution of VBDs to improve CEA of different health adaptation The health burden associated with monitoring and surveillance. While interventions in Bangladesh, including these three childhood illnesses is Bangladesh historically has had a investment in an early warning system projected to be about 14 million strong vector control program, data and risk reduction measures, projects disability-adjusted life years (DALY), collection focusing on the impact and to improve access to safe drinking accounting for about 3.4 percent gross cost of a variety of VBD programs in water and sanitation services, and domestic product (GDP) by 2050. Policy Bangladesh is critical for identifying nutrition-focused programs, is carried simulations also suggest that excess cost-effective interventions, ranging out using a variety of data sources. health burden attributable to climate from community-based awareness The CEA results suggest that nutrition change can be mitigated completely initiatives and projects to improve programs are more cost-effective than through targeted investment in the access to rapid diagnostic devices and the other interventions, although the traditional areas of development— essential drugs. findings should be interpreted with namely, improving access to basic caution due to weakness in the data. environmental services (water, Rapid urbanization and a growing sanitation, and electricity), female urban slum population are Given that a significant amount education, and child nutrition. quickly changing the population of financial resources, will be dynamics in Bangladesh, which committed to the area of climate Strong seasonal patterns are has implications for climate- change adaptation and resilience identified between climate induced health risks. However, the in Bangladesh, CEA should be used variability and vector-borne initial spatial targeting assessment routinely in allocating resources to diseases (VBDs), but future efforts provides strong evidence of the programs and projects. The results of should focus on filling the evidence poor geographic targeting of public CEA should be disseminated widely Executive Summary | 13 among stakeholders as a means to reliable data on the cost and impact improve transparency as well as to of different projects and programs develop an evidence-based decision- and to implement CEA with rigor and making process for selecting of consistency, evidence-based policy projects or programs and identifying making will remain simply an empty new areas of policy interventions. promise. Without concerted efforts to collect 14 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? CHAPTER 1 HEALTH AND CLIMATE CHANGE PATHWAYS ANDTHE BANGLADESH CONTEXT Chapter 1 | Health and Climate Change Pathways andthe Bangladesh Context | 15 I t has been recognized for centuries that weather fluctuations and BOX 1.1  Weather and Climate Variability versus Climate Change seasonal-to-interannual climate variability affect health outcomes. The Weather and climate variability refers to the day-to-day change in meteoro- geographic distribution of and logical parameters including temperature, precipitation, humidity, and winds. seasonal variations in many infectious Extreme weather events are significant deviations of meteorological variables, diseases are clear evidence of the links such as floods caused by excessive rainfall, droughts, storm surges, and heat between climate and health. However, waves (extreme temperature). to understand the links between climate and health, it is important to Climate variability refers to short-term changes in the average meteorological distinguish between weather and conditions over a time scale, such as a month, a season, or a year. In contrast, climate variability and climate change climate change refers to changes in average metrological conditions and (see box 1.1). seasonal patterns over a much longer time horizon, often over 50 or 100 years. Recent studies show that climate change is projected to shift the mean value of Most climate change studies so far temperature and precipitation, increase climate variability and the frequency have focused on the effect of rising and intensity of extreme weather events, and alter seasonal patterns (for temperatures on the intensity of example, delay the onset of the monsoon season or lengthen the hot season). El Nino-Southern Oscillation (ENSO) events (IPCC 2001.) The ENSO cycles influence interannual variability in temperature and rainfall and the likelihood of extreme weather events such as floods, storms, and droughts, which, in turn, have implications for local conditions. Therefore, the impact of global warming on local temperatures, precipitation, humidity, and seasonal patterns is likely to vary substantially across regions. Most global climate models project that climate change will increase the frequency and severity of extreme weather events in the coming decades and consequently lead to more serious and adverse health conse- mortality and morbidity from extreme hygienic conditions, and, in the event quences globally (Confalonieri et weather events such as floods and of natural disasters, cause outbreaks al. 2007; McMichael, Woodruff, and storms. of disease among displaced popu- Hales 2006; WHO 2009; IPCC 2001). lations in overcrowded shelters. However, climate variability is not the Climate variability and extreme During heat waves, excess mortality is only factor that will determine health weather events also affect health greatest among the elderly and those outcomes. Many other factors, such indirectly through their effect on the with preexisting conditions. There as health system capacity, socio-eco- nomic conditions, and demographic replication and spread of microbes is comprehensive literature (Hales, characteristics, can have a larger effect and vectors. For example, epidemi- Edwards, and Kovatz 2003; Patz et al. on health, either independently or by ological evidence shows that vector- 2010; Burke et al. 2001) on the links modifying climate effects (WHO and borne diseases such as dengue fever, between climatic and weather condi- PAHO 2010). malaria, and yellow fever are associ- tions and various infectious diseases ated with warm and humid weather and health outcomes. conditions; influenza becomes more Pathways from Climate prevalent during cooler seasons. In However, the impact of climate and to Health the tropics, diarrheal diseases typically weather variability on population The complex pathways through which peak during and after the monsoon health can be modified significantly climate conditions affect population season, and both flood and droughts by non-climatic factors. These include health are shown in figure 1.1. Climate are found to be linked to an increased improving access of the population conditions affect health outcomes risk of diarrhea. to basic environmental infrastruc- both directly and indirectly. The direct ture, such as safe drinking water and impacts on health include outbreaks Extreme weather events, such as sanitation facilities, and improving and spread of infectious diseases, droughts, floods, and cyclones, can the capacity of institutions to adapt, thermal stress-related mortality contaminate drinking water, cause such as better climate forecasting, from extreme high temperature, and water shortages that result in poor disaster management, and health 16 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? FIGURE 1.1  Pathways from Climate Change to Health Adaptive Modulating capacity influences Health effects Temperature-related Mitigate illness and death capacity Extreme Micorbial weather-related Regional contamination health effects weather pathways changes Water and food- Mitigatation Transmission borne diseases Driving measures Heatwaves dynamics Forces Vector-borne and Extreme weather Agro- rodent-borne Population diseases Dynamics Greenhouse ecosystems, gases (GHG) CLIMATE Temperature hydrology emissions CHANGE Effects of food and Unsustainable Precipitation water shortages economic Socioeconomics, development demographics Mental, nutritional, infectious and other health effects Health-specific Natural adaption causes measures Research needs Evaluation of adaption Source: WHO 2003. service provision. Further, locational the world, ranked sixth on the 2011 80 percent of annual precipitation characteristics such as population United Nations national disaster risk occurring during this period. Its density and the geographic charac- index. Climate variability and extreme geographic location and the hydro- teristics that determine local vulnera- weather events, such as inland logic impact of the GBM basin render bility to climate-related risk also play and coastal floods, droughts, trop- it particularly vulnerable to a range of important roles. Over the long run, ical cyclones, and storm surges, are climate risks, including various types many unforeseeable social and envi- projected to become more frequent of flooding (inland monsoon flooding, ronmental changes, human adap- and severe as the average temperature tropical cyclones, and storm surges) tation in response to climate-related rises with climate change (World Bank and periodic droughts. Once every risks, and advancement in vaccine 2010). three to five years, up to two-thirds and drugs can all modify the impact of Bangladesh is inundated by floods, of climate change on health and the Bangladesh is located at the tail end resulting in loss of life, outbreaks of probability of disease outbreaks. of a delta formed by three major disease, and damage to infrastructure, rivers—the Ganges, the Brahmaputra, housing, agriculture, and livelihoods The Bangladesh Context and the Meghna—and two-thirds of (World Bank 2010). the country is less than 5 meters above Climate conditions and climate sea level.1 The southwest summer Future changes in temperature and change projections monsoon is the major hydrologic precipitation in Bangladesh have Bangladesh is one of the most driver in the Ganges-Brahmaputra- been projected based on 16 global climate-vulnerable countries in Meghna (GBM) basin, with more than circulation models for three emissions Bangladesh is a subtropical country located between latitudes of 22°N and 27°N, with distinct seasonal climatic patterns. The winter season is 1  from November to March, and the hot and humid summer season is from April to October. The incidence of rainstorms is highest during the hot season from March to May, and about 70 percent of total annual rainfall occurs during the monsoon season from May to September. Rainfall varies significantly across regions, with annual average rainfall reaching about 1,500 millimeters at Rajshahi, 2,150 millimeters at central Dhaka, rising to 2,900 millimeters in the southeast at Chittagong, and 4,200 millimeters in the northeast at Sylhet (U.K. Meteorological Office). Chapter 1 | Health and Climate Change Pathways andthe Bangladesh Context | 17 scenarios (Yu et al. 2010). Temperature is predicted to rise during all months Table 1.1  Estimated Total Deaths, by Cause, 2007–08 and seasons relative to the historical WHO data, 2008 DHS data, 2007 data for 1980–99, with the median (ages 0–14) (ages 0–5) warming prediction of 1.55°C (degrees Celsius) by the 2050s. Annual and Cause of death Number % % wet season precipitation is projected (thousands) to increase, with a median increase All causes 221.7 100.0 100.0 in precipitation of 4 percent by the Infectious and parasitic diseases 57.5 25.9 15.1 2050s. The extent of land flooding Diarrheal diseases 23.8 10.7 2.0 is also projected to increase, given the existing flood protection infra- Malaria 2.3 1.0 — structure, with an average increase in Dengue 2.1 0.9 — flooded area of 3 percent in 2030 and Respiratory infections 38.3 17.3 — 13 percent in 2050. The proportion of flooded area is projected to be highest Lower respiratory infections 38.2 17.2 — during July through September, Upper respiratory infections 0.1 0.1 — peaking in August. Sea-level rise is Otitis media 0.0 0.0 — projected to be one of the most critical Pneumonia — — 22.0 climate risks for Bangladesh due to its long coastline and high population Perinatal conditions 91.0 41.0 25.8 density. Prematurity and low birth weight 28.6 12.9 — Birth asphyxia and birth trauma 30.3 13.7 — Major health issues Neonatal infections and other 32.1 14.5 — The 2008 Bangladesh Climate Change conditions Strategy and Action Plan prepared Nutritional deficiencies 4.5 2.0 0.4 by the government of Bangladesh highlighted the need to implement Injuries and drowning 12.2 5.5 9.2 surveillance systems for existing and new disease risks and to ensure that Source: WHO data for 2008 and 2007 DHS. Note: — = not available. health systems are prepared to meet future demands. According to the strategy document, Table 1.1 summarizes the leading factors, including differences in the causes of death among children (ages data sources, population covered, ages 0–14) based on the World Health covered, and definitions of diseases, Climate change is likely to increase Organization (WHO) data and verbal illnesses, and infection.2 the incidence of water-borne and autopsies from the 2007 Bangladesh air-borne diseases. Bacteria, para- Demographic and Health Survey Table 1.2 presents the trends in key sites, and disease vectors breed faster (DHS) for children ages 0–5. The large indicators of child health for the in warmer and wetter conditions and where there is poor drainage and discrepancies in the causes of death decade of the 2000s covering three sanitation. In view of this, it will (proportion of all deaths) from the two childhood illnesses and two measures be important to implement public data sources are likely due to several of child malnutrition. health measures (immunization, improved drainage, sanitation, and hygiene) to reduce the spread of these Table 1.2 Child Health: Incidence of Illnesses and Malnutrition, 2000–11 diseases and to improve access to health services for those communities % of children likely to be worst affected by climate Year ARI Diarrhea Wasting Stunting Underweight change. Unless these steps are taken, the health of many of the poorest and 2000 18.3 6.1 10.3 44.7 47.7 most vulnerable people will deterio- 2004 20.8 7.5 15.3 51.2 43.2 rate. Acute illness is known to be one of the main triggers driving people 2007 13.1 9.8 17.4 43.4 41.1 into extreme poverty and destitution 2011 5.8 4.6 16.2 41.2 36.2 in Bangladesh. Sources: DHS, 1999–2000, 2004, 2007, and 2011. The WHO estimates rely on vital statistics that are collected from primary health facilities, while the estimated proportions in the DHS data come 2  from verbal autopsies, relying primarily on the recall of the child’s caretaker in the households selected in the 2007 DHS. Only a few illnesses and diseases are defined in the verbal autopsy, whereas the vital statistics have much wider coverage, including illnesses, diseases, and conditions. 18 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Childhood illnesses. Diarrhea, acute prevalent in the coming decades in deaths from kala-azar were reported, respiratory infection (ARI), and fever3 Bangladesh. This is due to fact that although it is considered an important are major causes of under-five deaths, climate change is projected to affect vector-borne disease in Bangladesh. accounting for about a quarter of total the patterns of seasonality—in partic- Despite the recent attention paid to childhood deaths in Bangladesh. ular, prolonging the monsoon season vector-borne diseases in the public A range of climate variables, living and raising temperatures, which will health discussions of climate change, conditions, and socio-economic affect the abundance and spread of water-borne diseases, in particular factors can have an impact, directly many disease vectors. The number of diarrhea, are a more important health and indirectly, on the incidence and malaria cases reported increased from concern with regard to the total geographic distribution of these child- 1,556 in 1971 to 51,773 in 2011. Table number of deaths in the population hood illnesses. Repeated common 1.3 summarizes the major vector and (see table 1.3). childhood infections can also result water-borne diseases for the total in child malnutrition because they population in Bangladesh. reduce the child’s ability to absorb Mortality and emerging health issues nutrients during or even beyond the associated with natural disasters. In Dengue fever was an unfamiliar Bangladesh, natural disasters such as period of sickness. disease in Bangladesh until its flooding, cyclones, and storm surges outbreak in the summer of 2000 in have been a major source of health Child malnutrition. The rates of child three major cities—Dhaka, Chittagong, threats to the population, in terms malnutrition in Bangladesh are among the highest in the world. About half and Khulna—with the highest case- of lives lost, injuries, and disease of preschool-age children, equivalent load recorded at 6,132 in 2002. During outbreaks. Bangladesh has a long to more than 9.5 million children, are the period of 2000–11, a total of 23,518 history of natural disasters. According stunted, and more than 36 percent are cases and 239 deaths from dengue to data collected by the European classified as underweight.4 Recurring fever were reported in Bangladesh. No Detailed Mortality Database, between natural disasters exacerbate malnu- trition by wiping out crops, homes, safe water sources, and livelihoods. Table 1.3 Major Infectious Diseases: Number of Cases and Deaths, Recent evidence from a large number by Cause, 2000–11 of countries highlights the two-way   Malaria Dengue Kala-azar Diarrhea causal relationship between malnu- trition and the frequency of child- Cases Year Cases Deaths Cases Deaths Cases Deaths hood illnesses.5 Children with poor (thousands) nutritional status face far greater risk 2000 54,223 478 5,551 93 7,640 1,556 475 of mortality and severe illness due to 2001 54,216 490 2,430 44 4,283 1,866 521 common childhood infections, such 2002 62,269 588 6,132 58 8,110 2,599 1,022 as pneumonia, diarrhea, malaria, human immunodeficiency virus (HIV), 2003 54,654 577 486 10 6,113 2,287 1,282 acquired immunodeficiency syndrome 2004 58,894 535 3,934 13 5,920 2,246 1,170 (AIDS), and measles, suggesting that 2005 48,121 501 1,048 4 6,892 2,152 929 the health burden associated with childhood illnesses and malnutri- 2006 32,857 307 2,200 11 9,379 1,962 239 tion can be even higher with higher 2007 59,857 228 466 0 4,932 2,335 537 climate risk. 2008 84,690 154 1,153 0 4,824 2,295 393 2009 63,873 47 474 0 4,301 2,619 360 Vector-borne diseases. Dengue fever, malaria, and kala-azar, although 2010 55,873 37 409 0 2,810 2,427 345 less significant as a share of total 2011 51,773 36 1,362 6 2,534 2,268 70 health burden (accounting for about 2 percent of total deaths each Source: Ministry of Health. year), are projected to become more Note: There were no reported deaths caused by kala-azar. Fever is a symptom of many childhood illnesses. It was not recorded in the DHS during the two-week survey period, and it is not classified by the 3  WHO as a cause of death. 4 Three of the most commonly used measures of child nutritional status are underweight, stunting, and wasting. Stunting (height for age) refers to shortness, reflecting growth achieved pre- and postnatal, that is, long-term and cumulative effects of inadequate nutrition and poor health status. Wasting (weight for height) is a measure of an acute or short-term effect of an occurrence of illness, such as a bout of diarrhea that causes loss of body fluids and the consequent reduction in calorie intake. Underweight (weight for age) is a good indicator for children under 24 months, but it does not take height into account. 5 Based on a UNICEF report (UNICEF 2013) a child who is severely underweight (weight for height) is 9.5 times more likely to die of diarrhea than a child who is not, and the risk of death of a stunted child is 4.6 times higher. Moreover, malnutrition is also a consequence of infections and repeated episodes of illness. Chapter 1 | Health and Climate Change Pathways andthe Bangladesh Context | 19 1980 and 2010, Bangladesh experi- overcrowding in temporary shelters also been recorded among popu- enced 234 natural disasters, causing with inadequate drinking water lations living in the coastal regions more than US$17 billion in total and poor sanitation. The displaced (Khan et al. 2011). damage. The total number of people population is particularly exposed to killed as a result of natural disasters the risks of infectious diseases when The following two chapters analyze between 1980 and 2010 was about humans and livestock share shelters. the statistical links between climate 191,836; on average, 6,188 people are variability and health. Chapter 2 killed each year. In the coastal regions, the contami- quantifies the impact of climate nation of drinking water by salinity variability on the incidence of three Many outbreaks of water-borne as a result of storm surges caused by childhood illnesses using nationally diseases have been recorded during cyclones, rising sea levels, cyclone representative health surveys that and after major floods. While and storm surges, and upstream collect childhood diseases among complete records of disease cases withdrawal of freshwater has been children under the age of five. Chapter in the aftermath of major natural identified as the main cause of 3 focuses on the statistical correla- disasters are not available, officials emerging health problems. The rising tion between climate conditions and from risk management and disaster incidence of several health conditions vector-borne diseases using aggre- management units have reported and diseases, including hypertension, gate disease records collected by the evidence of outbreaks of various premature delivery due to pre-ec- Ministry of Health covering the entire infectious diseases as a result of lampsia, ARI, and skin diseases has population. CHAPTER 2 QUANTIFYING THE HEALTH IMPACT OF CLIMATE VARIABILITY ON CHILDHOOD ILLNESSES Chapter 2 | Quantifying the Health Impact of Climate Variability on Childhood Illnesses | 21 T his chapter aims to quantify the policy makers to plan health adapta- two nationally representative health impact of climatic variability tions in a medium- to long-term time surveys using geographic information on the incidence of childhood frame. system (GIS) information. illnesses—diarrhea, acute respiratory infection (ARI), and fever—while The following analysis is conducted Weather data controlling non-climatic confounding at the national level using informa- The Bangladesh Meteorological factors. While climate conditions such tion from the geographically refer- as weather fluctuations and seasonal Department has been collecting enced national health surveys and and interannual climate variability monthly weather data since 1970 in local weather stations located across influence many infectious diseases, 35 weather stations located across the country and spanning more disease incidence can be modified the country. The five weather vari- than 30 years. This study is based by many non-climatic factors. These ables (rainfall, average temperature, on uniquely rich data sources— include drinking water sources, maximum temperature, minimum combining health, a wide range of sanitation facilities, sewerage system socio-environmental information, temperature, and relative humidity) infrastructure, the capacity of public and local climate conditions—and have been collected daily,8 but only health services (such as disease large in spatial and temporal scale. monthly data are made publicly avail- surveillance, control, and treatment It attempts to address limitations in able. Figure 2.1 summarizes monthly systems), human adaptation responses current research in the area of health average weather variables constructed to health risks and climate variability, and climate links by controlling from the 35 weather stations from 1970 and population migration. In the confounding factors and taking to 2010 for six regions. context of Bangladesh, these non-cli- account of seasonality effects.6 matic factors are particularly important The monthly climate variables in urban areas, where population provide important climate features density is high and the slum popula- Key Data Sources for Bangladesh. While there is little tion is rapidly increasing as a result of Quantifying the impact of climate spatial variation in temperature massive rural-urban migration over variability on disease incidence or (maximum, minimum, and average) the past decade. health in general remains a formi- and humidity, monthly rainfall varies dable empirical challenge because of substantially across regions, in partic- Quantifying the health impact of data requirements and the complex ular during the monsoon season both climate and non-climatic factors pathways from climate conditions to of May to September. The average provides valuable information for the health. The analyses are extremely monthly temperature starts to rise formulation of public policies in two data intensive, requiring local infor- in April, reaches the peak (around areas. First, knowledge of the rela- mation on a sufficiently large spatial 30°C) in May, and starts to fall in tive importance of different health and temporal scale (taking account September. The maximum tempera- determinants can inform the design of seasonality) and covering health, ture during the hot season can vary of measures to maximize the health local climate conditions, the features between 30°C and 35°C. The humidity benefits of interventions. Second, the of ecosystem and topography, as well level starts to rise in April, reaching statistical links between climate vari- as a wide range of socio-economic the peak in June and continuing ability and disease incidence provide conditions.7 The complete list of data through September. critical parameters for projecting future sources in this study is summarized health burdens attributable to climate in appendixes A, B, and C. In the Both incremental changes in average change. While these projections are following analysis, the extensive climate variables and climate vari- subject to large uncertainties due to weather station data are matched ability—in particular, extreme weather the complex pathways from climate to the primary sampling units or events—affect disease outcomes. In conditions to health, they provide clusters (villages in rural areas and the following analysis of the statistical some benchmark estimates that help neighborhoods in urban areas) of the links between climate conditions and 6 A few studies have been carried out to analyze the statistical links between climate variables and infectious diseases in Bangladesh, focusing on diarrhea, cholera, kala-azar, and malaria. These are district case studies, covering the drought-prone district of Rajshahi, the flood-prone district of Manikganj, the salinity-affected district of Satkhira, and the malaria-epidemic region of Chittagong Hill Tracts (Matsuda et al. 2008; Pascual and Dobson 2005; Islam and Uyeda 2007; Hashizume et al. 2007) examine the association between climate variability and diarrhea cases in Dhaka City. While these studies provide useful information on the statistical association between climate variables and disease incidence, they are of local scale and do not control important confounding factors, such as household drinking water, sanitation facilities, and socio-economic conditions, that are important determinants of disease incidence (with the exception of Hashizume et al. 2007). Del Ninno and Lundberg (2005) use an indirect measure of extreme weather events—the household flood index—to estimate the impact of floods on the nutrition of under-five children in Bangladesh. Many leading researchers in the field of public health and climate change (Burke et al. 2001; Patz et al. 2010; McMichael, Woodruff, and Hales 7  2006; Hales, Edwards, and Kovats 2003; Gage et al. 2008; Rogers and Randolph 2006) have called for concerted efforts to improve the collection of data on health, local socio-economic conditions, and environmental infrastructure by integrating GIS and remote-sensing data on a wide variety of environmental parameters that are known to affect local disease risks (for example, forest coverage and soil conditions). 8 The completeness of these observations for the five variables varies across the 35 weather stations. 22 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? FIGURE 2.1  Average Monthly Weather Variables, by Region and Month, 1970–2011 Note: The monthly averages are constructed using monthly data from 1970 to 2011 collected from 35 weather stations located across the six regions. the incidence of disease, both the level (cold) month is defined as when the locations of primary sampling units and variability of climate variables temperature is higher (lower) than or clusters (villages in rural areas and are matched to the GIS-referenced 1 (or 2) standard deviation from the neighborhoods in urban areas). national health surveys. long-term mean at a specific month and location. The 2004 and 2007 DHS, the fifth Excessive precipitation (floods) or and sixth waves of health surveys extreme low rainfall (droughts) is The Bangladesh national health that have been conducted since conventionally defined as when rain- surveys 1993, are chosen for analyzing the fall at a given location and month is links between health and climate. bigger (smaller) than 1 (or 2) stan- The Bangladesh nationally represen- The choice of these two surveys is dard deviation from the long-term tative Demographic and Health Survey driven by the survey months and the average. The location and monthly (DHS) is used in the analysis. The availability of GIS information in the long-term average climate conditions survey’s sample framework covers the surveys. The survey months are from are established using the monthly entire population residing in private January to May in the 2004 DHS and weather station data from 1970 to dwelling units in the country based from March to August in the 2007 DHS, 2010. The extreme temperature is on the enumeration areas created in presenting the opportunity to analyze similarly defined—the extreme hot the 2001 census.9 Map 2.1 displays the the effect of climate variability for Bangladesh is divided into six administrative divisions; each division is divided into zilas, and each zila is divided into upazilas. Each urban area in 9  the upazila is divided into wards, and each ward is divided into mahallas; each rural area in the upazila is divided into union parishads, and each union parishad is divided into mouzas. The urban areas are stratified into three groups: (1) standard metropolitan areas, (2) municipalities, and (3) other urban areas. Chapter 2 | Quantifying the Health Impact of Climate Variability on Childhood Illnesses | 23 Map 2.1  DHS Survey Locations, 2004 and 2007 pre-monsoon and monsoon seasons information collected in the DHS with key household-level variables are separately. Table 2.2 presents the the weather station data at a relatively summarized in tables 2.2 and 2.3. It is sample distribution for the 2004 and high resolution. notable that there is little variation 2007 DHS by survey month and by in water or the availability of elec- region. In addition to detailed information on tricity over the time period. There is, childhood diseases, the DHS collects however, some variation in sanitation The 2004 and 2007 DHS collected GIS information on household-level access and education, but one would expect information for each primary sampling to a wide range of basic services, such much more improvement over the unit. The GIS data provide the critical as drinking water sources, sanita- period. link for spatially matching disease tion facilities, and electricity, as well incidence as well as socio-economic as socioeconomic conditions. The Table 2.2 Household Distribution in 2004 and 2007 DHS, by Region and Survey Month Survey year and month Barisal Chittagong Dhaka Khulna Rajshahi Sylhet Total 2004 DHS January 304 490 546 — — 521 1,861 February 314 27 1 258 261 367 1,228 March 134 352 155 374 321 56 1,392 April — 498 403 203 314 — 1,418 May — 136 426 31 416 — 1,009 Total 752 1,503 1,531 866 1,312 944 6,008 2007 DHS March 207 133 — — — 139 479 April 455 421 34 — — 591 1,501 May 129 586 139 104 102 388 1,448 June — 135 379 257 355 — 1,126 July — — 356 353 497 — 1,206 August — — 377 — 13 — 390 Total 791 1,275 1,285 714 967 1,118 6,150 Note: — = no households surveyed. 24 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Table 2.3  Summary of Key Variables in 2004 and 2007 DHS Variable  2004 DHS 2007 DHS Sample size 6,908 6,150 Water sources (%) Piped water = 1 7.32 6.03 Tube well = 2 84.52 80.36 All other = 9 8.16 13.61 Sanitation Septic tank = 11 11 24.36 Slab latrine = 21 15 0.00 Pit latrine = 22 33 16.73 Open latrine = 23 27.51 31.33 All other = 99 13.52 27.58 Has electricity 41.06 42.33 Has TV 24.57 28.39 Religion Islam = 1 91.26 91.19 Hindu = 2 8.26 8.06 Other = 99 0.48 0.62 Location Urban = 1 30.01 34.26 Rural = 2 69.99 65.74 Sex head of household Male 93.01 90.86 Mother’s education No education = 0 36.78 27.25 Primary = 1 31.09 31.33 Secondary = 2 26.45 34.05 Higher = 3 5.67 7.37 Quantifying the Health community-level covariates, that is, precipitation event * toilet types). confounding factors. Therefore, the Another example is identifying the Impact of Climate Variability estimated impact of climate variability synergy effect of integrated health The analytical model can be interpreted as a causal effect interventions—whether the combi- The links between the incidence of as opposed to a statistical correla- nation of improving both drinking childhood illnesses and climate vari- tion between climate and disease water sources and sanitary facilities ability are analyzed here using the incidence. (for example, access to water * access multilevel (hierarchical) modeling to sanitation) is more effective in approach. The statistical presenta- Second, the multilevel model can be reducing disease incidence than the tion of this model is summarized in extended to test various health adap- sum of two separate interventions box 2.1. This modeling approach has tation policy variables. For example, (improving drinking water sources and several analytical advantages over what is the differential effect of sanitation facilities). the existing biological and statistical extreme rainfall on disease incidence modeling approaches that are widely between households with sanitary Seasonality of disease incidence used in epidemiology to study the toilets and those with open latrines, In the DHS, cases of diarrhea, ARI, links between climate and health. holding all other factors constant? and fever are collected for children First, it allows the incorporation of This interactive effect can be tested under the age of five. The informa- both climate variables and a range by incorporating in the model an tion is recorded based on the moth- of individual, household, and interactive term (for example, extreme er’s response to the question asking Chapter 2 | Quantifying the Health Impact of Climate Variability on Childhood Illnesses | 25 whether her children had experi- enced episodes of any of the three BOX 2.1  The Model illnesses during the period two weeks preceding the survey date. Diarrhea is The probability of event (Y)— a child who lives in household h located in not defined medically, and answers village i experiences an episode of illness at survey month t — is assumed to to this question are left to the moth- follow the logistic distribution: er’s perception and diagnostics. This means that reporting biases may arise 1 (B2.1.1) Phi = E (Y = 1 | Xi) = . as mothers with more education are 1 + e?? + ??Xi more likely to recognize the symptoms associated with diarrhea than mothers Phi The odds ratio defined as is commonly used in the medical field and with less or no education. A child is 1 – Phi recorded to have experienced ARI if epidemiology. The logistic regression model, which is the log transformation the mother reported any one of three of equation B2.1.1, has a linear relationship between the odds ratio and Xi, symptoms, including short but rapid the covariates. The specification of the logistic regression model is as follows: breathing, difficulty breathing, or labored respiration.10 Phi (B2.1.2) Log ( ) = a + b1 Climatei + b2 SEhi + b3 Villagei + b4 Vi + uhi. 1 – Phi The DHS does not collect information on malaria or dengue fever, which Where, Climatei is the vector of climate variables (both level and variability) are major vector-borne diseases in in villagei, SEhi is the vector of individual- and household-level variables Bangladesh. However, fever can be (for instance, age, gender, birth order, drinking water source, type of sanita- a major manifestation of malaria, tion facilities, and other key socio-economic characteristics such as mother’s which is more prevalent after the education level or religion). Villagei is a vector of village-level variables, such end of the rainy season, as well as as the proportion of households with access to safe water and sanitation, a symptom of a wide range of acute access to health facilities, presence of a village drug store or village medical infections that can occur through all personnel, and presence of a health outreach program. Vi is the unobservable seasons in Bangladesh.11 It is esti- village-specific random effect, which is likely correlated with Villagei, and mated that the proportion of malaria uhi is the residual following i.i.d. distribution and is not correlated with the among all fever cases in Bangladesh, included explanatory variables. on average, is about 12 percent (ICDDR,B 2007). The model specification incorporates both the level of precipitation and extreme rainfall events. Excessive rainfall can have different health conse- Table 2.4 summarizes the incidence of quences depending on the season and topography as well as household illness by month and season estimated and community environmental infrastructure (for example, drinking water from the 2004 and 2007 DHS. During sources, sanitation facilities, waste discharge, and sewerage systems). the pre-monsoon season, ARI is more prevalent among young children, with an incidence of 14 percent compared to 10 percent during the monsoon season. The incidence of fever is higher in the monsoon season than during the pre-monsoon season. Based on the two health surveys, the incidence 10 The combined symptoms that are associated with ARI are similar to pneumonia or bronchiolitis. In the diagnosis guidance of the World Health Organization (WHO), the symptoms of ARI include convulsions, shrunken eyes, high respiratory counts, noisy breathing, and high body temperature. Fever can be a common symptom of many 11  non-life-threatening illnesses, including both vector- and water-borne diseases such as ARI and diarrhea, and of life- threatening illnesses, including cerebral malaria, meningitis, septicemia, and typhoid. Its causes can be associated with local endemic and epidemic diseases that may or may not be seasonal. 26 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? in humidity increases the incidence Table 2.4 Disease Incidence, by Survey Month and Season, 2004 and 2007 by 4 and 6 percent, respectively. No statistically significant effect is found Survey year and month Fever Diarrhea ARI for diarrhea during the pre-monsoon 2004 DHS season, although Hashizume et al. January 27.18 7.50 17.39 (2007) confirm a positive association February 35.2 6.64 15.30 between the number of cases and temperature, using weekly diarrheal March 39.36 6.69 20.63 cases collected in Dhaka hospitals from Cold season: January–February 29.89 7.21 16.68 January 1996 to December 2002. April 44.98 7.26 19.53 The effect of temperature and May 41.29 6.22 15.06 humidity levels during the monsoon Pre-monsoon: March–May 41.97 6.76 18.61 season of May to August (2007 DHS Total 37.19 6.94 17.85 survey) is, however, markedly different from that of the pre-monsoon season. Higher temperature is associated with 2007 DHS a lower incidence of ARI, but higher March 43.60 6.02 17.10 humidity has no significant effect. April 34.22 10.00 16.20 The differential impact of climate conditions on disease incidence across May 31.43 9.91 11.59 seasons may be due to the distinct Pre-monsoon: March–May 34.06 9.43 13.98 seasonal patterns of climate variables June 33.22 7.48 9.65 in Bangladesh. July 41.19 9.38 12.14 During the pre-monsoon season August 39.84 12.24 11.35 when the average temperature in Monsoon: June–August 37.68 9.07 10.98 Bangladesh is in the range of 15–28°C Total 36.07 9.23 12.32 before peaking at 30°C in May, an increase in temperature is likely to foster the development and spread of of diarrheal illness appears constant The impact of climate on disease a wide range of pathogens and para- through all seasons. However, using incidence sites, causing more illnesses such as the hospital data for both children The regression results using the multi- fever and ARI among young children. and adults, Teshima et al. (2007) level model confirm that climate vari- The temperature reaches above 30°C find two peak seasons for diarrhea ability is strongly linked with the inci- during the monsoon season, possibly in Bangladesh: the pre-monsoon dence of childhood illnesses and that approaching the limit of physiological (March and April) and the end of the the impact of climate varies by season. tolerance for pathogens, as an increase monsoon season. Table 2.5 summarizes the estimated in temperature can destroy many types impact of climate variables on the of pathogens, reducing the disease incidence. The nonlinear relationship A direct comparison of disease inci- odds ratio of fever, diarrhea, and ARI between temperature and the survival dence by month and season using among children under the age of five, of a wide range of pathogens has also the 2004 and 2007 DHS is complicated controlling for confounding factors at been documented in epidemiological by the fact that the primary sampling multiple levels: child, household, and studies for other countries (Patz et al. unit locations changed between 2004 community. The full regression results 2010). and 2007 (map 2.1). For example, from the complete model specification the average disease incidence for the are presented in tables D.1, D.2, and The effect of extreme heat (tempera- same survey months (March–May) is D.3 in appendix D. ture) on health is also investigated markedly different in the 2004 and in the analysis. An extreme hot Effect of temperature and humidity. (cold) event is defined as when the 2007 surveys. This indicates that the During the pre-monsoon season temperature is higher (lower) than 1 location effect is particularly important (January–April), results from the 2004 (or 2) standard deviations from the in estimating disease incidence. DHS show that high temperature and long-term mean at a specific month Therefore, for disease monitoring humidity levels increase the inci- and location. However, unlike precip- purposes, the national disease surveil- dence of fever and ARI. A 1° increase itation, extreme heat events were lance system should be considered a in average temperature increases the very rare in the months and locations more reliable source of data than the odds ratio of fever by about 21 percent surveyed in 2003 and 2004; therefore, national health surveys. and ARI by 14 percent, respectively, it is not possible to use the model to controlling for household socio-envi- test the effect of extreme heat events ronmental factors; a one-unit increase on disease incidence. Chapter 2 | Quantifying the Health Impact of Climate Variability on Childhood Illnesses | 27 Table 2.5 Impact of Climate Variables on Disease Incidence, 2004 and 2007 DHS   Pre-monsoon season Monsoon season Climate variable Fever Diarrhea ARI Fever Diarrhea ARI Temperature at survey month 1.21*** 0.87 1.14* 0.92 0.88 0.51* Relative humidity at survey month 1.04* 1.04 1.06* 1.01 1.01 0.97 Rainfall at survey month Rainfall level (log form) 0.81* 1.39 0.76 1.32 0.95 0.78 Extreme rainfall event (dummy variable; base case = normal rainfall) Heavy rainfall event 1.82* 1.23 1.47 0.67* 0.74 1.03 Low rainfall event 0.96 1.27 0.89 1.44 0.95 0.96 Rainfall-one month lag Rainfall level (log form) 0.84*** 1.04 0.87 1.27* 1.25 1.02 Extreme rainfall event (dummy variable; base case = normal rainfall) Heavy rainfall event 1.18 1.09 1.28 0.72 0.56* 1.08 Low rainfall event No No No 0.96 0.85 0.89 Non-climatic covariates (included) Child age and gender Yes Yes Yes Yes Yes Yes Household socio-economic factors (water sources, Yes Yes Yes Yes Yes Yes sanitation, mother’s education, religion) Community factors (clinic, health program) Yes Yes Yes Yes Yes Yes Note: The estimates of all non-climatic covariates are presented in appendix B. *** p<0.001; * p<0.05. Effect of precipitation. The effect of sources12 as well as the hygienic envi- population (McMichael, Woodruff, precipitation on disease incidence ronment overall. and Hales 2006), or improving water is more complex, and a distinction quality through the dilution of should be made between excessive Compared to the incremental change pathogens in an environment where rainfall (floods) events and an incre- in precipitation, excessive rainfall water quality is poor.13 mental increase in precipitation. events have a much larger effect on The results from the DHS data show the incidence of fever. Relative to The nonlinear health impact of that an incremental increase in the normal rainfall conditions, excessive precipitation may be attributable level of precipitation during the rainfall events increase the odds ratio to seasonal patterns in precipita- pre-monsoon season, both at the of fever by 82 percent during the tion in Bangladesh. As shown in survey month and with a one-month pre-monsoon season, but reduce it figure 2.2, the average monthly lag, decreases the incidence of fever, by 33 percent during the monsoon rainfall normally peaks in June and but has no effect on diarrhea or ARI, season, controlling for all other vari- continues through July and August. everything else being the same (the ables. What explains the nonlinear full results include socio-economic relationship between rainfall and Therefore, a heavy rainfall event variables and household access to fever incidence across seasons? Some during the monsoon season may basic services and are reported in the studies show that, in tropical and generate a positive health benefit by appendix D, tables D.1, D.2, and D.3). subtropical regions, high precipi- reducing the vector population and This finding can possibly be explained tation may cause the outbreak of diluting the contamination of patho- by the fact that higher rainfall during diarrhea, whereas excessive rainfall gens in groundwater sources (as dry seasons makes groundwater more may reduce vector-borne diseases reflected in the estimated negative plentiful, improving the quantity of by flushing larvae from their habitat impact of heavy rainfall events on drinking water available from safe in pooled water, reducing the vector the odds ratio of fever, 33 percent). Using DHS data from 25 countries in Africa, Bandyopadhyay, Wang, and Kanji (2011) find that high rainfall during the survey period in the dry 12  season reduces the prevalence of diarrheal disease by around 3 percentage points. For example, 51 percent of water-borne disease outbreaks were preceded by precipitation events above the 80th percentile of the monthly average for 13  the particular weather station using a database of all reported outbreaks of water-borne disease in the United States from 1948 to 1994. 28 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Caveats on data limitations. While carry out more local-level validation called for future analytical efforts national health surveys are valu- studies to explore the best options for focusing on empirical studies that use able for analyzing the links between collecting rainfall data for research in a wide range of analytical methods climate and health, the lack of inte- the area of climate change and health.14 and data to assess the consistency of grated climate information collected Islam and Uyeda (2007) validate precip- the climate-disease relationship in at the survey locations limits their itation data by comparing rainfall data different societal contexts and across a usefulness. The reliance on matching collected from satellite-based Tropical variety of temporal and spatial scales. 35 weather stations to the survey Rainfall Measuring Mission (TRMM) The analysis presented here takes an location can result in large imprecision and from the rain gauge network from important first step in this direction. in the measurement of local climate 1998 to 2002 in Bangladesh. They conditions and, consequently, in the conclude that TRMM is useful for esti- In the context of Bangladesh, future estimated impact of climate variability mating the average values of rainfall studies should seek to use disease data on the incidence of diseases. in Bangladesh, but that satellite data collected from the national surveillance overestimated the rainfall during the system and other sources of climate Rainfall is widely regarded to be pre-monsoon period and in dry regions data, including both weather stations the most challenging meteorolog- and underestimated it during the and remote-sensing precipitation data, ical parameter to measure due to its monsoon period and in wet regions. to supplement studies using national spatial and temporal variability. Along health surveys. In particular, these the spatial dimension, rainfall can The review of existing literature studies should investigate the impact vary over short distances and interact on health and climate change—in of extreme precipitation on infectious with local climate and topography. particular, the links between climate diseases by season. Research efforts The availability of low-cost modern variability and infectious disease, should focus particularly on flood- rain gauges and advances in remote- malnutrition, and other climate-sen- prone areas where population density sensing precipitation products that sitive diseases—shows that this is high, as credible evidence is needed estimate global and high-resolution important area is still in its infancy. to assist policy makers in designing precipitation open up the possibility to Leading researchers and experts have health adaptation programs. In particular, the spatial resolution of satellite-based products may have become less a concern, with several precipitation products (including 14  RFE, ARC, and CMORPH) achieving high spatial resolution of 0.1 x 0.1 (degree), covering an area of about 11 square kilometers). Such a level of resolution is likely adequate for measuring village-level precipitation in many rural areas. Chapter 2 | Quantifying the Health Impact of Climate Variability on Childhood Illnesses | 29 Impact of household environmental conditions on health Table 2.6 Impact of Water and Sanitation Facilities on Disease Incidence, Odds Ratio, 2004 and 2007 The estimation of the health impact of household environmental services Survey and variable Number of Fever Diarrhea ARI focuses on drinking water sources and households sanitation facilities, as well as their Pre-monsoon season, 2004 DHS interactive effect with an extreme Urban precipitation event (defined as the dummy variable in the model). The Access to water sources drinking water sources and sanitation Relative to piped water 473 facilities are markedly different in urban Tube well and other 1,598 1.072 0.525 1.165 and rural households; therefore, a sepa- rate analysis is carried out for urban and Sanitation facility rural areas (the full estimation results Relative to septic sink 555 are presented in appendix D, tables D.1, Slab latrine 435 1.34 0.807 1.933** D.2, and D.3). The estimated impact of Pit latrine 465 1.241 0.76 1.924* household environmental conditions on health is summarized in table 2.6. Open latrine 483 1.202 1.15 1.859* No toilet 133 1.251 0.771 2.943** Poor sanitation facilities are identi- Rural fied as a key determinant of disease incidence in urban but not in rural Access to water sources areas. During the pre-monsoon Relative to tube well 4,574 season in urban areas, a child living Surface and other 254 1.058 0.862 1.435 in a household with unsanitary toilet facilities (slab, pit, or open latrine) is Sanitation facility two times as likely as a child living in a Relative to septic or slab 824 household with access to a flush toilet Pit latrine 1,790 1.011 0.882 0.774 (septic sink) to experience an episode Open latrine 1,415 1.159 1.179 0.902 of ARI and nearly three times as likely if living in a household with no toilet, No toilet 800 1.09 1.241 0.761 holding all other factors constant. Monsoon season, 2007 DHS Urban During the monsoon season, unsanitary toilet facilities significantly increases the Access to water sources incidence of fever. Given that fever is a Relative to piped water 350 symptom of many infectious diseases, Tube well 1,535 1.268 0.968 this finding suggests that households Surface and other 222 1.853 1.384 with poor sanitary facilities are likely to be located in areas with unsanitary living Sanitation facility environments, which are conducive to Relative to flush or septic sink 649 the development and spread of a wide Flush to pit 345 1.834*** 1.42 range of pathogens and disease vectors. Pit latrine 311 1.802** 1.562 The results are consistent with find- Open latrine 425 1.284 0.943 ings of the World Bank Water and No toilet 377 1.654*   1.477 Sanitation Program (WSP 2011), which are based on the 2007 DHS. That study Rural found that inadequate sanitation is Access to water sources responsible for economic losses of Relative to tube water 3,435 about US$4.22 billion, equivalent Surface and other 608 0.832 0.937 1.124 to 6.3 percent of the country’s gross domestic product (GDP) each year. Sanitation facility Relative to flush or septic sink 504 Contrary to findings in other studies, Pit latrine 718 0.787 0.782 0.643* no statistically significant effect of drinking water sources on disease Open latrine 1,502 0.8 0.675 0.767 incidence is established in these two No toilet 1,319 0.788 1.004 0.676 national surveys. The estimated impact of drinking water sources shows that a *** p<0.001; ** p<0.01; * p<0.05. 30 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? child living in a household using surface zz Second, as expected, household incidence estimated in the previous water is more likely to contract diseases, environmental conditions are identi- section provides critical parameters such as fever and ARI, although the fied as an important determinant of for projecting future health burden estimates are not statistically signifi- the incidence of childhood diseases, in the context of climate change. It cant.15 The lack of variation in drinking although the interactive effect of is important to acknowledge that water sources is likely the underlying extreme precipitation with drinking the health impact of climate change reason for the lack of statistical signif- water sources or sanitation facilities and variability remains highly uncer- icance. In 2004 about 77 percent of is not statistically significant. The tain, and extrapolating future health urban households and 72 percent of estimation results show that unsan- impacts of climate variability based rural households had a tube well. In itary toilet facilities are a key factor on estimated statistical relationships 2007, about 95 percent of urban house- determining the incidence of fever at the local level can be highly impre- holds had piped water, and 85 percent and ARI, but only in urban areas. cise (National Research Council 2001). of rural households had a tube well. zz Third, on average the incidence However, projections are useful for of childhood illnesses estimated understanding the potential economic The impact of water contamination is from the national surveys varies by and social implications of disease investigated by incorporating lagged survey month and is strongly influ- burden. In this section, future disease precipitation variables in the model. enced by survey location. Therefore, burden is projected under several In general, surface water is quickly using disease surveillance data to assumptions (see box 2.2). contaminated in the event of heavy estimate disease incidence at disag- rainfall in areas where sewerage gregate levels may be more reliable The climate projection is taken from systems and sanitation services are than using national health surveys. the global circulation models adopted poorly constructed, but it can take a in World Bank (2009) which studies zz Finally, the findings provide useful much longer time to observe the impact climate adaptation in Bangladesh. The information on the links between of groundwater contamination after vast majority of climate change predic- health- and climate-related condi- heavy rainfall. This suggests that the tions relevant to Bangladesh have tions, but they should be interpreted time period for observing the occur- with caution. The pathways from been made using regional climate rence of disease outbreaks varies by climate conditions to health are models. These models project that season and depends on the quality of complex, and non-climatic factors the rate of warming in South Asia will local water supply and sewerage infra- and human adaptation responses be significantly faster than that seen structure. The regression results with to climate variability are often in the twentieth century and more lagged climate variables, however, do important underlying determinants rapid than the global mean rate of not show a statistically significant effect of disease dynamics and population warming. During December, January, of extreme rainfall on the incidence of health. These factors are difficult and February, warming is expected to the three childhood illnesses. to account for in some cases due to be at its greatest and to be associated data deficiency and limitations in with a decrease in precipitation, while Summary of Results the existing analytical approaches. the consensus of regional models is While these results are broadly consis- The most effective means to improve that summer rainfall will increase. tent with evidence from the existing our understanding of the links between epidemiological studies that focus on Extreme weather events are projected health and climate change may include the impact of climate change on disease (a) improving the collection and use to become more frequent in South transmission and incidence, new find- of ecological and environmental data Asia, including heat waves and exces- ings emerge from this analysis: using remote-sensing technologies and sive rainfall. Tropical cyclone inten- meteorological conditions from satellites sity is also expected to rise by 10–20 zz First, the impact of climate vari- at higher levels of resolution and (b) percent and sea surface temperature is ability on disease incidence varies building a base of data for sufficiently expected to increase by 2–4°C. Glacial sharply between pre-monsoon and long periods of time and across regions. and sea-ice melt and the expansion monsoon seasons. The estimated The analysis in this section takes a first of the oceans due to higher mean impact of an incremental change step in this direction in the context of temperature suggest that sea-level of a climate variable (temperature, Bangladesh. rise is certain (the minimum rise of humidity, and rainfall) is relatively about 40 centimeters by the end of the small, but the impact of extreme Projecting Future Health century is projected based on the most precipitation on disease incidence is conservative climate change estimates; large and statistically significant. The Burden Map 2.2 presents the projected spatial impact of extreme precipitation on The statistical relationship between distribution of disease incidence, disease incidence varies by season. climate variation and disease temperature, and floods by 2050. The interactive effect of extreme rainfall and household environmental services (drinking water sources and sanitation facilities) is also investigated. 15  The regression results show no statistically significant interaction effect on health. This finding is at odds with widely held perceptions that the impact of extreme precipitation on disease incidence depends on the household’s source of drinking water and type of sanitation facilities. Chapter 2 | Quantifying the Health Impact of Climate Variability on Childhood Illnesses | 31 BOX 2.2  Key Assumptions for Projecting Disease Burden by 2050 First, the projection of future temperature and precipitation is based on 16 global circulation models from the fourth report of the Intergovernmental Panel on Climate Change (Yu et al. 2010), which predicted temperature rises during all months and seasons. The median warming prediction for Bangladesh across the models by the 2050s is 1.55°C. The projected monthly precipitation assumes that both annual and monsoon season precipitation increases, but that precip- itation during the post-monsoon dry season does not rise. A median increase in precipitation of 4 percent by the 2050s compared with the baseline of 1980 for Bangladesh is assumed. Second, the level of public investment in water and sanitation in the coming four decades is assumed to be sufficient to improve access to safer drinking water and sanitation services taking account of population growth. It is assumed that by 2050 all households currently without safe water sources will be connected to piped water and that households currently using open latrines will be provided with sanitation facilities that connect to a septic tank. Map 2.2  Spatial Distributions of Disease Incidence, Temperature, and Floods by 2050 32 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Table 2.7 summarizes the projection results by 2050. Under the scenario of Table 2.7 Projected Health Burden by 2050 a rise in average temperature of 2°C   Estimation Projection and increased frequency of excessive from DHS rainfall events across all regions (from data 30 to 40 percent) in Bangladesh, the incidence of ARI and fever is projected Indicator 2004 2007 2050 to increase, but the incidence of Climate variable diarrhea is projected to remain Average temperature (survey months) 23.5 27.8 29.0 unchanged. The future health burden Probability of flooding (%) 22.6 30.6 40.0 of these three childhood illnesses is projected to be about 14 million Disease incidence  disability-adjusted life years (DALY), ARI accounting for about 3.4 percent of Incidence (%) 18.7 12.3 23.0 GDP by 2050, based on the United Nations population projections for Cases (thousands) 9,009 5,734 14,220 Bangladesh for 2050.16 Diarrhea Incidence (%) 7.0 9.2 7.3 This projection should also be Cases (thousands) 3,376 4,296 4,529 regarded as the lower-bound estimate for at least two reasons. First, it covers Fever only three childhood illnesses. Many Incidence (%) 39.2 36.1 46.3 studies in Bangladesh show that other Cases (thousands) 18,916 16,787 28,605 diseases in adults, including hyper- tension due to water salinity caused Population ages 0–14 (thousands) 48,222 46,541 61,833 by coastal flooding, malnutrition due to both increased food insecu- Note: The average temperature refers to the survey months. Flooding is defined as monthly rainfall above 1 standard deviation for a particular location and month. rity and high disease incidence (the two-way causation between nutrition and childhood illnesses, and mental disorders, have already become more place. As discussed in chapter 1, the conditions on child health is estimated prevalent in recent years and could be impact of climate and weather vari- using the Bangladesh 2011 DHS (see attributed to climate change). ability on population health can be appendix D, table D.4). The results significantly modified by non-climatic are consistent with evidence compiled Second, in the coming decades, the factors, such as access to safe drinking from many low-income countries by population—in particular, in urban water and sanitation facilities, access to the World Health Organization (WHO and coastal areas—is likely to be more basic health care services, health infor- 2003, 2009; WHO and PAHO 2010), exposed to a variety of health risks mation, and better hygiene. Promoting indicating that children’s health (both of much bigger magnitude as the development can be the best option disease incidence and nutritional frequency of extreme weather events for addressing climate change–related status) is influenced by a range of rises. health issues. This begs the question of household-level factors, such as living how much the projected excess health conditions, access to safe water, sani- How Much Climate Impact burden (as a result of climate change, tary toilet facilities, access to electricity, Can Be Mitigated through holding all factors constant) can be and maternal and child-care practices averted through investment in key (such as hygiene practices). Development? areas that are known to have a signifi- The projection in the previous section cant impact on health. Table 2.8 presents the results of a simu- shows that temperature warming lation exercise illustrating how much and more frequent extreme weather This section illustrates the impact of of the excess health burden, in terms events, including both inland and development investment on mitigating of the incidence of childhood illnesses coastal flooding, will increase the the excess health burden associated and premature deaths, that is attrib- health burden under the assump- with climate change in the context of utable to the effects of climate change tion of maintaining the status quo, child health measured by both disease (warming and increased frequency of meaning that no additional targeted incidence and nutritional status. The extreme weather events) can be averted health adaptation interventions take impact of household environmental by investing in basic infrastructure The calculation of DALY for children under the age of five is the sum of morbidity and mortality from three illnesses. Morbidity is estimated from 16  the projected disease cases for 2050, assuming that the average length per episode is four days, a DALY disability severity weight of 0.11 and a DALY age weight of 0.31, taken from WHO guidelines (Murray and Lopez 1997). Mortality is calculated based on total disease cases for 2050, which are estimated based on the conditional probability of mortality for each illness and discounted years of life lost per child, based on WHO guidelines. Chapter 2 | Quantifying the Health Impact of Climate Variability on Childhood Illnesses | 33 this chapter—that the impact of Table 2.8 Simulation of the Health Impact of Climate Change (Warming climate variability on the inci- and Flooding) by 2030 under Two Scenarios: Current versus dence of childhood illnesses varies Higher Levels of Investment sharply between pre-monsoon and monsoon seasons—suggest   Scenario 2: Additional that policy makers should focus on investment selectively targeting the implemen- to promote tation of disease control programs, development both temporally (focusing on peak Impact Scenario 1: Infrastructure Education Nutrition All three seasons) and spatially to reach the Current level combined most vulnerable locations in terms of investment of climate and health risks. Incidence of illness zz The identification of vulnerable 0.432–0.474 0.395 0.396 0.400 0.371 locations should be based on a (base line: 0.42 in 2011) comprehensive spatial database, which covers climate-sensitive Averted deaths in children diseases (incidence and outbreaks) 464 −224 −215 −160 −605 and their seasonal patterns, ages 0–14 (thousands) patterns of extreme climatic events (such as heavy rainfall or prolonged Note: The incidence of childhood illness is projected based on regression estimation using periods of drought), and local envi- the 2011 DHS, assuming that (a) average temperature increases by 1.5°C and (b) the proba- ronmental infrastructure (drinking bility of flooding increases from the current level of 0.225 to 0.35 by 2030 across the whole water sources, sanitation, sewerage country. The population between the ages of 0 and 14 was 51 million in 2011(based on WHO system, and waste management). figures) and is projected to be 66 million in 2020, and 61.8 million in 2030, based on U.S. The spatial database should be projections. Overall development investment achieves (a) universal access to safe drinking at a disaggregate level, such as water (that is, tube well in rural areas and piped water in urban areas), sanitary toilet unions in rural areas or slum and facilities, and household access to electricity; (b) secondary female education attainment; non-slum areas in urban areas. and (c) elimination of child stunting in 2030. zz The disease data collected from the national surveillance system and the (drinking water sources, sanitation, and (MDG) targets set for 2015 (see box climate data collected from various electricity access), female education, E.1 in appendix E). In this scenario, sources, including both weather and improving children’s nutrition. about 605,000 child deaths would be station and remote-sensing precip- averted. This is more than the total itation equipment, should supple- Table 2.8 presents the simulated deaths that are projected to occur as a ment the national health surveys. impact on the incidence of child- result of climate change, which indi- Research efforts should focus hood illnesses and premature deaths cates that targeting social investment in particularly on flood-prone areas averted under two scenarios by 2030. areas that are particularly important to with high population density to The first one is the case where total child health could mitigate completely provide location- and context-spe- the excess child deaths attributable cific information for improving the investment is only sufficient to keep to climate change. The message is implementation of disease manage- up with population growth, which clear: while climate change is indeed ment programs in these locations. means that the current levels of access to basic services, female education, an important issue and can affect health both directly and indirectly, zz While the adverse health conse- and child nutrition status remain the quences of climate change have same in 2030. The projected child focusing resources on traditional areas of development, such as improving gained recognition in recent years, deaths attributable to climate change it is important not to lose sight would be around 464,000 in 2030. access to safe drinking water, sanita- tion, electricity, female education, and of the importance of socio-envi- child nutrition, is the best option for ronmental conditions for health. The second scenario assumes that A simple message is that focusing mitigating the adverse health impacts additional investments are made resources on traditional areas of of increased climate risks. to achieve the following targets: (1) development, such as improving universal access to basic environmental access to safe drinking water, sani- services; (2) secondary female educa- Key Messages tation facilities, electricity, female tion attainment; and (3) elimination zz Given that resource limitations education, and child nutrition of child stunting by 2030. These targets (for example, medical personnel and improving resilience and risk are set in reference to both the current and drug availability) have always management capacity, can be the level estimated from 2011 DHS and been one of the key constraints best health adaptation option in the Millennium Development Goal in Bangladesh, findings from the face of increased climate risk. CHAPTER 3 VECTOR-BORNE DISEASES: HOTSPOTS AND CLIMATE LINKAGES Chapter 3 | Vector-Borne Diseases: Hotspots and Climate Linkages | 35 E pidemiological studies across The Bangladesh government, in the This chapter provides some evidence the world show that climate 2008 Climate Change Strategy and on the links between climate condi- variability has a direct influence Action Plan, highlighted the impor- tions and these three VBDs using on the replication and movement of tance of addressing the emerging several data sources that are available microbes and vectors of many vector- public health risks associated with in Bangladesh. Section 3.1 summarizes borne diseases (VBDs). The temporal and three VBDs: malaria, dengue fever, the epidemiology of VBDs. Section 3.2 spatial changes in temperature, precip- and kala-azar (see box 3.1). However, reviews existing data sources on key itation, and humidity affect the biology studies that focus on the links disease data in Bangladesh. Section 3.3 and ecology of vectors and intermediate between climate conditions and the analyzes the temporal and spatial hosts and, consequently, the risk of incidence of VBDs in Bangladesh are distribution of the three VBDs using vector-borne disease transmission extremely limited, with the exception the most recently available data in among the exposed population. Socio- of malaria. This scarcity of empirical high-prevalence regions. Section 3.4 economic and environmental condi- studies in the area of VBDs is largely presents a statistical analysis of the tions—in particular, rapid urbanization a result of poor data sharing and correlation between climate variables and population migration—are also underuse of VBD information and and the incidence of VBDs. identified as important determinants of weather station data that are readily the spread and outbreak of VBDs. available in Bangladesh. BOX 3.1  Major VBDs in Bangladesh Malaria is one the most common vector-borne diseases in tropical regions, and pathogens are protozoan that are transmitted to humans by the Anopheles mosquito. Anopheles mosquitos tend to prefer a temperature range from 24°C to 27°C. If the overall temperature rises, their habitat may be reduced, and the breeding period may be shifted and prolonged, leading to a possible change in malaria distribution in Bangladesh. Dengue fever is transmitted by Aedes aegypti, a type of mosquito that feeds primarily on humans and thrives especially in urban environments where still water and plant containers are abundant. The intensity of dengue vectors is deter- mined mainly by the availability of breeding sites such as water containers. Based on studies in some areas of Bangkok, the abundance of vectors is largely independent of rainfall. Kala-azar is caused by the female sand fly Phlebotomus argentipes on the Indian subcontinent, including Bangladesh. The disease is lethal if left untreated and affects approximately 500,000 new patients annually worldwide, with 60 percent of new cases on the Indian subcontinent. Climate adaptation measures (for example, building more embank- ments in response to sea-level rise) may be conducive to the spread of visceral leishmaniasis vectors, increasing the number of cases of kala-azar. Kala-azar cases in Bangladesh are found to cluster near flood control embankments. 36 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? The Epidemiology of Vector- Several studies have shown that However, the impact of climate vari- rainfall is positively associated with ability on the incidence of many Borne Diseases dengue incidence. Rainfall could vector-borne and water-borne Malaria, dengue fever, and kala-azar facilitate the spread and transmis- diseases can be significantly modified are all transmitted between human sion of vector-borne diseases by by local environmental conditions hosts by vectors.17 The effect of climate expanding breeding sites. Heavy rain- and human adaptation responses. variability on the incidence and case- fall contributes to inland water bodies For example, in a tropical region such load of VBDs is mediated through its and flooding almost every year in as Bangladesh, drought can lead to impact on each component of the Bangladesh, creating a suitable envi- an increase in dengue fever because transmission cycle of different patho- ronment for the vector. more people may store water in open gens. The transmission components containers in areas where access to include pathogen (viral, bacterial, Humidity is another critical climatic piped water is limited, thus increasing and parasites), vector (mosquito), factor that can influence the survival the number of breeding sites for non-biological vehicle (water, soil), and transmission of VBD pathogens mosquitos. Another striking example non-human reservoir (mice, rodent), and vectors. Most studies that include of the importance of non-climatic and human host. the humidity variable show a consis- factors on VBDs is the sharp difference tent pattern of rising humidity being in dengue fever cases along the United Many studies show that climate vari- positively associated with disease States and Mexico border, with about ables such as temperature, precip- incidence, suggesting that a high level 5,033 cases reported on the Mexico itation, and humidity affect the life of humidity is conducive to the repli- side, compared with only 100 cases on cycle of many disease pathogens cation and survival of many pathogens the United States side between 1980 and vectors and, consequently, the and viruses. Humidity is found to be and 1996; given the identical climatic outbreak and incidence of disease.18 one of the most critical determinants conditions, the difference is attribut- Changes in temperature are therefore of dengue fever (Hales, Edwards, and able almost entirely to socio-environ- likely to have the biggest impact on Kovatz 2003). mental factors. the transmission of vector-borne diseases. Transmission occurs at two While these climate variables are ranges of temperature: 14–18°C at the Key Data Sources and Issues correlated with a high incidence of lower end and 35–40°C at the upper VBDs, the seasonal patterns of infec- for VBDs end.19 An increase in temperature in tious diseases are one major pathway Three main institutions in Bangladesh the lower range has a significant and through which climate change can are directly involved in the collection nonlinear impact on the extrinsic have a potentially drastic effect on the of data on major water- and vector- incubation period and consequently dynamics and outbreaks of disease borne diseases. These include (1) the on the transmission of disease (Watts (Pascual and Dobson 2005). For Directorate General of Health Services et al. 1987). A rise in temperature example, long-term climate change (DGHS) under the Ministry of Health above 34°C generally has a negative affects seasonal patterns, altering the and Family Welfare (MoHFW); (2) the impact on the survival of vectors and timing of the onset of the monsoon Institute of Epidemiology, Disease parasites (Rueda et al. 1990), reducing season or lengthening the hot season, Control, and Research (IEDCR); and transmission risks. However, at around all of which can affect the key compo- (3) the International Center for Disease 30–32°C, vector capacity can increase nents of the transmission cycle. A and Diarrhoeal Research, Bangladesh substantially as result of a reduction longer monsoon season may affect the (ICDDR,B). in the extrinsic incubation period.20 reproduction rate of pathogens and Reeves et al. (1994) find that extremely increase the coverage of still water The DGHS is responsible for the high temperatures can actually to facilitate the geographic spread of Management Information System (MIS) increase mosquito mortality, which mosquitos and, consequently, the of the health system, which records a decreases the transmission of arboviral spread of VBD risk. wide range of diseases collected from disease (related to dengue fever). health facilities across the country. In epidemiology, infectious diseases are broadly classified as anthroponoses if the natural reservoir of the pathogen is human and zoonoses 17  if it is animal. Direct transmission is defined as when the pathogen is transmitted directly between human hosts or between animals through physical contact or droplet exposure, while indirect transmission is defined as when the pathogen is transmitted between human hosts by either a physical vehicle (water) or a biological vector (mosquito). 18 For example, Hales, Edwards, and Kovats (2003) show that warm ambient temperature and high humidity increase the probability of transmission by shortening the time for mosquitos to become infectious, resulting in higher vector-borne incidence. 19 Jetten and Focks (2006) find that warming temperature increases the incidence of dengue fever up to a threshold and that, when passing the threshold, rising temperature reduces the incidence of many vector-borne diseases. For regions where cases of dengue are already present, the impact of an increase in average temperature is substantial—the aggregate epidemic risk is increased by an average of 31–47 percent for a 1°C increase in mean temperature (Patz et al. 1996). 20  The relationship between temperature and the survival of a wide range of pathogens is complex and nonlinear (Patz et al. 2010). Extreme temperatures (too high and too low) are destructive to the survival of pathogens because their physiological tolerance has a limit. But a gradual increase in temperature in the spectrum of moderate temperatures is likely to increase the survival and spread of pathogens and viruses. Chapter 3 | Vector-Borne Diseases: Hotspots and Climate Linkages | 37 The IEDCR also collects information on water- and vector-borne diseases Table 3.1 Key Data Sources for VBDs through disease surveillance and Indicator Dengue Malaria Kala-azar disease outbreak investigations. ICDDR,B has also developed a surveil- Source Disease Control Room National Malaria Malaria and Parasitic lance system for diarrheal disease of Directorate General Control Program Disease Control Unit and vector-borne infections (malaria) of Health Services (NMCP) of DGHS (M&PDC) of DGHS (DGHS) over the past decade. In addition, the incidence of water- and vector- Period and Monthly from 2000 Monthly from 2007 to Yearly from 1994 to borne diseases among children under frequency to 2011 2011; yearly 2000 to 2007; monthly from the age of five is collected through present 2008 to 2011 nationally representative Demographic Level District District Yearly data at district and Health Surveys (DHSs) at regular level; monthly data intervals of every three years by at upazila (subdistrict) level the National Institute of Population Research and Training (NIPORT). Comments Cases were confirmed Cases were confirmed Cases were confirmed Table 3.1 summarizes the data sources by different by different govern- by different govern- for these three VBDs. government and ment hospitals ment hospitals private hospitals of (mostly upazila health (mostly upazila health Bangladesh complex) and nongov- complex) As part of the capacity assess- ernment organizations ment, focus group interviews were conducted to identify major data issues and ascertain the quality The MoHFW has taken some initiatives For dengue fever, a total of 23,518 cases of data from different agencies in to improve the health MIS to ensure and 239 deaths were reported during charge of collecting health data in the delivery of timely and reliable the period of 2000–11. The highest Bangladesh. Poor data quality and disease information. These include caseload was recorded in 2002 (6,132), inadequate institutional capacity, regular and timely publication of followed by 2000 (3,964), and the ranging from the primary data sources health bulletins, modernization of the lowest caseload was in 2010 (409). In (the health facility level) to the final data collection and storage system in the case of kala-azar, a total of 104,313 centralized data system (health MIS) a central department, and creation cases were recorded from the 36 were identified as the key problems by of an assessment report using Health districts during 1994–2011. A declining all agencies. Metric Network assessment tools. The trend of cases is seen from 2004 to key findings from the focus group 2011. The number of malaria cases The inaccuracy of disease data result interviews suggest that efforts to remained stable between 2000 and largely from lack of computer facilities 2005 but spiked in 2008. However, improve the quality of health data and designated staff for data collec- due to inconsistent data collection should target lower-tier facilities, tion and database management at the across locations and over time, caution facility level, in particular hospitals. such as subdistrict-level hospitals and should be taken in reaching any The responsibility for collecting and clinics where the regulator’s collection conclusions regarding the trends of recording disease data rests on doctors of disease data is taking place, and VBDs using these data sources. and nurses who are already over- include private health facilities. Data loaded with health service delivery. should be verified using multiple Figure 3.1, panel b, presents the The information collected on VBDs is data sources to check consistency and seasonal patterns of three VBDs often not based on laboratory tests reliability before they are published in using the average monthly number due to lack of diagnostic devices. the health MIS. of disease cases constructed for the Also, the data are not collected using period 2000–11. A clear seasonality a standardized health format; for The Temporal and Spatial pattern is evident for dengue fever this reason, the data reported to the and malaria, but not for kala-azar. health MIS is likely to be incomplete Distribution of VBDs The dengue fever season starts in and inconsistent across regions and Temporal and seasonal trends July, peaks in August (accounting for over time. 37.5 percent of total annual cases on Figure 3.1, panel a, presents the trends average), and declines in October Another important data deficiency is of three VBDs from 2000 to 2011 and (20.5 percent). In the Dhaka area the omission of disease cases reported their seasonal patterns based on where more than 98 percent of cases from private health practitioners monthly VBD cases collected from are reported, no cases were found where a large proportion of VBD the government disease surveillance from January through April during patients are treated. It is estimated system managed by the DGHS. The 2000–11. An outbreak of dengue fever that more than half of total annual caseloads for all three VBDs vary errat- and dengue hemorrhagic occurred in cases of disease not recorded in the ically from year to year, with no clear Dhaka in 2000, with more than 5,551 health MIS because cases are treated in trends of increase or decrease observed hospitalized cases and 93 reported private sector health facilities . for this period. deaths between July and December. 38 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? FIGURE 3.1  Trends and Monthly Disease Patterns of VBDs, 2000–11 While malaria cases are recorded depending on local climate conditions, that proximity to hospitals deter- throughout the year, the malaria season geographic shift of seasonal patterns, mines whether cases of dengue are starts in June (accounting for about 13.2 local environmental conditions, and diagnosed. percent of annual total cases), peaks in population movement. Map 3.1 pres- July (15.1 percent), and declines from ents the spatial distribution of dengue In Bangladesh, 45 out of 64 districts September onward (accounting for fever, kala-azar, and malaria using the are recorded to have kala-azar about 12.2 percent in September). The most recently available data of 2011. cases. Mymensingh is the most seasonal pattern of kala-azar disease endemic region, accounting for about is less obvious, with the caseload rising The cases of dengue fever were 48.5 percent of total cases, followed by gradually from March and declining reported primarily in two cities, Dhaka Pabna (11.4 percent) and Tangail (9.4 from October onward. and Khulna. Using geographic infor- percent) (Rahman et al. 2008). Five mation system (GIS) technology, Ali et upazilas in the Mymensingh District Disease hotspots al. (2005) showed that dengue clusters have the highest number of kala-azar In general, the infectious disease are less identifiable in areas farther cases. At the aggregate level, about hotspots are likely to change over time, away from major hospitals, suggesting 5,000 new cases of kala-azar are Map 3.1  Geographic Distribution of Vector-Borne Diseases in 2011 Chapter 3 | Vector-Borne Diseases: Hotspots and Climate Linkages | 39 reported every year, although officially in the correlation analysis to allow for some prediction of the caseload of recorded cases are likely an under- the long incubation period of many of dengue fever in August and thereby estimation due to lack of access to the vectors.21 help in preparedness. For example, diagnostic devices and an inadequate preventive and case control measures public health surveillance system Dengue fever should be put in place in high- in many parts of the high-endemic risk locations such as Dhaka when areas. The analysis focuses on Dhaka District, where most (about 92 percent) of the abnormal climate (for example, higher total number of dengue cases were rainfall and hotter weather than long- The map of malaria shows that the term means) is observed one or two majority of malaria cases occurred in recorded during the period 2000–11. The results presented in table 3.2 show months preceding the peak month of 13 districts close to the areas bordering India and Myanmar, accounting for a strong correlation between lagged August. 98 percent of the reported cases of climate variables (rainfall, tempera- malaria identified in Bangladesh. ture, and humidity) and dengue Malaria These 13 districts are difficult to reach cases. Monthly rainfall (both current The statistical analysis is based on due to the hilly terrain and lack of rainfall and rainfall lagged up to monthly data collected by ICDDR,B surveillance and information systems. three months) is positively correlated from three hilly districts (Khagrachhari, Therefore, substantial underreporting with dengue cases, with the highest Rangamati, and Bandarban) and one of malaria cases in this region by the correlation observed in two-month coastal district (Cox’s Bazar) from 2007 Ministry of Health is highly plau- lagged rainfall (0.73). The average to 2011. Table 3.3 presents the results, sible. Bandarban, Khagrachari, and temperature (lags of up to three which show that high rainfall is posi- Rangamati, collectively known as months) is also significantly correlated tively correlated with malaria cases in the Chittagong Hill Tracts districts, with dengue cases, as is humidity recorded the highest incidence of level. These results provide important Rangamati and Khagrachhari, while malaria in Bangladesh. Rangamati evidence on the links between in Bandarban the positive association District has the highest number of climate variability and dengue fever in is strongest with one-month lagged cases of malaria throughout the year, Bangladesh.22 rainfall. High humidity is also posi- followed by the districts of Bandarban tively correlated with malaria cases. In and Khagrachari. Moulvibazar District, The findings of stronger correlation all the four districts, high temperature situated in the northeast part of between lagged climate variables increases the number of malaria cases the country, is also classified as a (lags of up to two to three months) with a one- to two-month lagged high-malaria-risk area, with malaria and the cases of dengue fever are effect, and the correlation between cases rising significantly from June particularly useful. They suggest that climate variables and malaria cases to August in 2011 to reach the level of closer monitoring of climate conditions becomes stronger as the location of Chittagong District. (rainfall, temperature, and humidity) the district is farther away from the in June and July (one or two months coastal area, with the correlation in Statistical Correlation preceding August, the expected peak Cox’s Bazar (an inland district) being between Climate Conditions month for dengue fever) can provide the weakest. and VBDs In this section, the statistical correla- Table 3.2 Spearman Rank Correlation of Dengue Cases and Climatic Factors tion between climate variables and VBD cases is analyzed using the Factor Lag 0 Lag 1 Lag 2 Lag 3 monthly disease cases collected Total rainfall 0.39 0.64 0.73 0.54 from the health surveillance system Average minimum temperature 0.43 0.66 0.67 0.55 managed by DGHP and validated by Average maximum temperature 0.05 * 0.26 0.44 0.58 ICDDR,B. The Spearman rank correla- tion coefficient is used to assess the Average humidity 0.60 0.74 0.65 0.33 strength of correlation. Both current and lagged climate variables are used * Statistically not significant. Highest correlations are marked as bold. For example, research shows that the dengue vector, Aedes mosquito, takes 7 to 45 days to become an adult from an egg. 21  22 Other studies have found similar results in the other settings. In Rio de Janeiro, Brazil, dengue incidence was significantly correlated with a 0-month time lag for minimum and maximum temperature and a 1-month time lag for accumulated rainfall. Minimum relative humidity and total rainfall at a 2-month time lag and minimum and maximum temperature at a one-month time lag were correlated with dengue incidence in Guangzhou, China. In Guadeloupe, French West Indies, dengue incidence was most positively correlated with three climatic factors: relative humidity with a 7-week lag, minimum temperature at a 5-week lag, and mean temperature at an 11-week lag. In Barbados, the strongest correlation was found between dengue incidence with vapor pressure at a 6-week lag, precipitation at a 7-week lag, minimum temperature at a 12-week lag, and maximum temperature at a 16-week lag. In Kaohsuing City in Taiwan, China, amount of rainfall, relative humidity, and minimum and maximum temperature were positively correlated at a 2-month lag with dengue incidence for 1995–2000 and at a 3-month lag for 2001–08. For 1992–2001, dengue incidence in Jakarta, Indonesia, was correlated with a 2-month lag of rainfall and 1-month lag of temperature. 40 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Table 3.3 Spearman Correlation between Malaria Cases and Climate Conditions District and climatic factors Lag 0 Lag 1 Lag 2 Lag 3 Khagrachhari District Total rainfall 0.73 0.63 0.36 0.04* Average minimum temperature 0.7 0.66 0.48 0.16* Average maximum temperature 0.15* 0.38 0.57 0.57 Average humidity 0.65 0.38 0.01* –0.41* Rangamati District Total rainfall 0.7 0.66 0.49 0.08* Average minimum temperature 0.71 0.72 0.53 0.19* Average maximum temperature 0.20* 0.5 0.6 0.53 Average humidity 0.7 0.39 0.02* –0.37* Bandarban District Total rainfall 0.59 0.66 0.46 0.16* Average minimum temperature 0.66 0.8 0.72 0.43 Average maximum temperature 0.34 0.57 0.62 0.48 Average humidity 0.76 0.67 0.45 0.15* Cox’s Bazar District Total rainfall 0.39 0.44 0.24* 0.02* Average minimum temperature 0.29 0.43 0.5 0.35 Average maximum temperature –0.08* 0.21* 0.57 0.68 Average humidity 0.44 0.38 0.22 –0.07* *Statistically not significant. Highest correlations are marked as bold. Table 3.4 Spearman Rank Correlations of Kala-azar Cases and Climatic Factors Factor Lag 0 Lag 1 Lag 2 Lag 3 Total rainfall 0.26 0.25 0.05 –0.10 Average minimum temperature 0.34* 0.17 –0.15 –0.20 Average maximum temperature 0.34* 0.14 0.01 –0.11 Average humidity 0.14 0.11 –0.04 –0.14 *Statistically significant. Highest correlations are marked as bold. Kala-azar Another noteworthy finding is that the correlated with kala-azar cases, but correlation changes from positive to weaker than temperature and rainfall Table 3.4 presents the Spearman negative after two- or three-month variables. correlation analysis for kala-azar. The lags, although the lagged correla- correlation between kala-azar cases tion is statistically insignificant. For While we did not find a significant and climate variables is much weaker example, the correlation between the correlation between climate vari- than that for dengue fever and malaria. current monthly average temperature ables and kala-azar, some research The Spearman rank correlation coeffi- and kala-azar cases is 0.34, but it has been conducted in this area in changes to –0.15 for temperature with Bangladesh. Using data collected cients are statistically significant only for a two-month lag and to –0.2 with a in the highly endemic district of temperature (minimum and maximum). three-month lag. The results show a Mymensingh from 2000 to 2004, Alam Unlike the results for dengue fever and strong positive correlation between (2005) investigated the links between malaria, no lagged climate variables are temperature and number of cases. the abundance of the kala-azar correlated with kala-azar cases. Average humidity is also positively vector and environmental factors, Chapter 3 | Vector-Borne Diseases: Hotspots and Climate Linkages | 41 focusing on soil conditions (tempera- VBD control programs should target zz Addressing existing and emerging ture, moisture, pH level, and organic resources (diagnostic equipment and VBDs in Bangladesh, future studies composition) and housing structure. medical staff) not only in high-risk should focus on filling in the The results show that organic compo- locations but also in seasons with high information gap on the cost-ef- sition is significantly correlated with caseloads. fectiveness of various VBD control the abundance of vector species and interventions. Given the large that houses with mud walls are twice Interdisciplinary approach should uncertainties about the impact as likely to have vector species than be promoted for research on VBDs of climate variability, social and houses with walls made of tin. and climate links. For example, environmental conditions, as well the epidemiological research that is as urbanization and population In another study, using active surveil- based on laboratory or field exper- migration on the replication and lance data from 2006 to 2009 collected iments and mathematical modeling movement of microbes and vectors, jointly by ICDDR,B and the University of the dynamics of infectious disease it is plausible that vector-con- of Tsukuba in Trishal upazila of the epidemics can be combined with the trol programs in the context of Mymensingh District, establishes a dose-response between the expo- Bangladesh may be less cost-ef- statistically significant correlation sure of pathogens and the disease fective than case management between the kala-azar caseload and incidence to assess the health impact interventions. However, lack of household location as well as a gender of future climatic risk factors. This rigorous cost-effectiveness analysis effect on the incidence of kala-azar approach has been used by the U.S. in this area limits the possibility disease (in general, males have a Environmental Protection Agency to of providing concrete answers to higher risk of contracting kala-azar questions of this nature. develop controls for key water-borne than females). disease agents and by the World zz A first step in this direction is to Health Organization, which conducted Summary of Results collect data on the impact and cost a global assessment of the health of a variety of VBD programs in The findings presented here shed impact of climate change (Campbell- Bangladesh, covering vector control light on the seasonal patterns of Lendrum, Corvalan, and Pruss-Ustun programs, community-based three VBDs and their sensitivity to 2010). awareness initiatives, and projects climate variables. The results confirm designed to improve access to rapid a strong correlation between short- Key Messages diagnostic devices and essential term climate variability and VBD drugs. zz While the incidents of VBDs are cases in high-prevalence regions of projected to increase with climate Bangladesh. In the case of dengue zz To provide information for change in Bangladesh, as high- fever and malaria, all climate variables improving the targeting of resources lighted in the government’s Climate (temperature, rainfall, and humidity), to address VBDs, efforts should Change Strategy and Action Plan, both current and lagged, have a focus on three key areas: (1) VBDs pose a less critical public strong correlation with disease case- improving data collection on VBDs load. For kala-azar, only temperature health threat than water-borne from both public and private clinics is significantly correlated with the diseases such as diarrhea or the to identify correctly the changes in number of cases. Soil conditions and high prevalence of malnutrition. seasonal patterns and geographic housing structure were also found to This is due to the fact that a wide distribution of VBDs; (2) inte- be correlated with the abundance of range of non-climatic factors grating epidemiological data with the kala-azar vector in the high-inci- (ecological, environmental, and environmental and climate data dence district of Mymensingh. socio-economic factors) are known and promoting interdisciplinary to influence the incidence of VBDs research to improve the robustness However, the correlation analysis and that the health burden associ- of the evidence; and (3) conducting does not control for local socio-eco- ated with VBDs has remained rela- spatial assessment of local health nomic and environmental condi- tively small (1 and 0.9 percent of facilities (availability of trained tions and therefore has limitations deaths from malaria and dengue, medical staff to diagnose and treat for projecting future VBD burden or respectively, compared with 10.7 VBD cases and essential drugs and mapping the geographic distribution percent from diarrhea, which is diagnostics equipment) against the of VBDs in Bangladesh.23 The findings even higher when taking account of projected risk of VBD outbreaks in suggest that the implementation of malnutrition). high-prevalence locations. Although socio-economic factors may influence individual cases, the data collection methods used preclude us from understanding their 23  influence on specific cases. CHAPTER 4 POPULATION DYNAMICS AND SPATIAL TARGETING: IMPLICATIONS FOR CLIMATE CHANGE AND HEALTH Chapter 4 | Population Dynamics and Spatial Targeting: Implications for Climate Change and Health | 43 O ne of the key features of dynamics have evolved over the past (ranging from 12 million in Dhaka population dynamics in decade in Bangladesh. Second, it to 3.4 million in Bogra). The least Bangladesh is the rapid but presents the measurement of adap- populated districts were Bandarban, poorly planned urbanization process tation capacity using complementary Rangamati, Khagrachhari, Meherpur, that has been taking place over the data sources. Finally, it conducts Jhalkati, Narail, and Borguna (popu- past few decades, with the share of a spatial mapping of geographic lation between 390,000 in Bandarban urban population rising from about targeting of public financing of to 918,000 in Borguna). 19 percent in the 1980s to about 30 health adaptation capacity, including percent in 2012. The large popula- basic environmental infrastructure The population growth rate also tion movement and the increased and health facilities and disease varies substantially by region. Over population density in urban areas—in prevalence. the period of 2001 to 2011, the average particular, the rapid expansion of district population growth rate was urban slums, driven by a combination Key Features of Population about 1.2 percent. However, some of poverty and recurring natural disas- districts grew much faster (Gazipur, ters—are likely to alter the patterns Dynamics 6.7 percent; Dhaka, 4.2 percent; and spread of disease, exposing the According to the United Nations popu- Narayanganj, 3.6 percent; Sylhet, 3.4 population to health risks. lation projections, the Bangladesh percent), while a few districts shrank population will reach 218 million (Barisal, −0.13 percent; Jhalkati, −0.17 The results from the previous two by 2050, from around 147 million in percent; Khulna, −0.25 percent; and chapters show strong geographic 2012, with urban areas estimated to Bagerhat, −0.47 percent). The under- and seasonal patterns of infectious absorb the additional 70 million of lying reasons for the negative popu- diseases in Bangladesh. While these total population growth. While the lation growth in these districts are not spatial and temporal characteristics growth of rural population is projected fully established, although extreme of the transmission of vectors and to halt by 2025,24 the urban popu- weather events, natural disasters, and diseases can be influenced by climate lation will increase from the current poverty are all likely to have contrib- variability and extreme weather 34 million to more than 100 million uted to the large outflows of popula- events, local adaptation—in particular, by 2050, predominantly as a result of tion. The complete list of population living conditions and the treatment rural-urban migration (Streatfield and growth by district is summarized in and control of disease—play a critical Zunaid 2008). table D.5 in appendix D. role in mitigating the adverse health consequences of climate risk. For More alarming, the fastest growth is The district-level information, example, malaria was once common taking place in urban slums, at around in many parts of Europe, but it is no however, presents only a partial 7 percent a year, twice the average longer found in these regions as a picture, concealing variations of growth rate of the urban population. result of improved living conditions population density between rural and Based on the most recent data, the (water supply infrastructure, sewerage proportion of slum population in urban areas and, more important, system, and screening of homes in urban areas is about 35.2 percent on within urban areas. The rapid but addition to aggressive vector-control average, but in Dhaka, more than poorly planned urbanization in programs). one-third (37.4 percent) of the total Bangladesh has been associated with population (9,136,182) is currently many emerging health problems, Therefore, assessing the geographic living in slum areas, compared with particularly in urban slums where targeting of measures to fight the about 19.5 percent in Khulna City the population growth rate is the prevalence of disease, at the disaggre- Corporation. highest. However, the health status gate level, provides critical information of the urban slum population is for informing the design and imple- The rapid population growth and imprecisely estimated due to sampling mentation of health adaptation inter- rural-urban migration have shaped biases in current nationally repre- ventions. Such an assessment would the spatial distribution of population sentative household surveys such as help the Government of Bangladesh density. While Bangladesh is the third the Demographic and Health Surveys and its development partners to most densely populated country in (DHSs) and household expenditure improve their ability to plan, prioritize, the world, substantial variations exist surveys. These surveys are important and target policy interventions and across regions. Map 4.1 presents the sources of information on the spatial investments to reach the population current population density in 2011 distribution of health status and that is most vulnerable to infectious and changes between 2001 and 2011 social-economic conditions, but the diseases in locations where the health for 66 districts using the 2001 and sampling design has not adequately adaptation capacity is more limited. 2011 population censuses. In 2011, the captured the new features of popula- most populated districts were Dhaka, tion dynamics—in particular, the rapid This chapter focuses on three issues. Chittagong, Comilla, Mymensingh, increase in urban-slum population First, it looks at how population Tangail, Sylhet, Gazipor, and Bogra (see box 4.1). Another approximately 30 million will be added to the current 108 million people living in rural areas by 2050. 24  44 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Map 4.1  Population Density and Growth BOX 4.1  Issue of Sampling Bias National household surveys such as the DHS omit the A comparison of health indicators using the 2006 UHS and urban slum population. The health indicators for urban the 2007 DHS reveals the extent of sampling bias as a result areas across regions, therefore, do not capture a signifi- of inadequate coverage of the slum population in national cant proportion of the urban slum population. The recent health surveys. Table B4.1.1 presents the incidence of diar- joint initiatives by the government and the U.S. Agency rhea and stunting for Dhaka areas using the two surveys. for International Development (USAID) have attempted to The comparison shows large discrepancies in these two address this issue. The major outcomes of this joint effort are health indicators. Diarrhea incidence was about 11.3 percent the three waves of Urban Health Surveys (UHSs) conducted in Dhaka using the DHS data, while it was much lower in 2006, 2011, and 2013, covering slum and non-slum areas according to the 2006 UHS (4.3 percent in non-slum areas, of the major city corporations, including Dhaka, Chittagong, 8.2 percent in small- and medium-size slum areas, and Khulna, Rajshahi, Barisal, and Sylhet. 6.8 percent in large slum areas). Table B4.1.1 Potential Bias of Health Indicators of Urban Population: An Example of Dhaka 2006 UHS Lag 1 Lag 2 2007 DHS, Data source Large slum Medium, small slum Non-slum all areas Number of observations 721 664 546 960 Health indicators Diarrhea incidence (%) 6.8 8.2 4.3 11.3 Stunting rate (%) 59.4 51.2 39.6 28.2 Note: The survey months are March–July for the 2006 UHS and April–August for the 2007 DHS for urban Dhaka area. Child malnutrition status in urban areas is significantly using the 2006 UHS (39 percent in non-slum areas in underestimated using the DHS data: the average stunting Dhaka, 51 percent medium- and small-size slums, and rate among children under five in Dhaka is about 28 59 percent in large slums). percent using the 2007 DHS, compared with the estimate Chapter 4 | Population Dynamics and Spatial Targeting: Implications for Climate Change and Health | 45 The average population density in Measuring Health Adaptation The 2011 HFS collected data on the an urban area is about 23,378 per availability of medical inputs (equip- Capacity ment, drugs, and personnel), the square kilometer (ranging from 7,152 The local capacity to mitigate health provision of various types of health in Barisal City Corporation to 29,857 in risks associated with population services, and information on consumer Dhaka Metropolitan Area). In Dhaka migration and high density as well or client satisfaction obtained from slums, population density is around as climate change depends on both both public and private facilities 205,415 per square kilometer, which health and non-health sectors. covering all tiers of administra- is about 8 times the average for Therefore, a comprehensive measure tive units (from community clinics urban areas and almost 300 times the of local health adaptation capacity to district hospitals). One of the average for rural areas (755 per square should cover local preventive and important aspects of health sector treatment health services (disease capacity is the delivery of quality kilometer) (Streatfield and Zunaid health services. This capacity, in turn, surveillance systems and availability 2008). depends on the availability of equip- of medical staff and essential drugs), local basic environmental infrastruc- ment, skills, and training of medical Changing population dynamics and ture, and a risk and disaster manage- personnel and the overall level of spatial distribution have important ment system, to name a few. Both management at the facility level. implications for public health and public and private sector capacity to Information on these aspects of health the spread of infectious diseases. address current and future health care are often difficult to collect in national health surveys. The risk of disease endemics can be risks should be incorporated in the heightened significantly in locations measurement of capacity. In the following analysis, four core where public investment in basic health sector–related indicators, In reality, however, information that environmental services such as safe covering equipment availability, essen- covers comprehensive measures of tial drug availability, staffing quality, drinking water, sanitation facilities, adaptation capacity at a disaggregate and sewerage systems are lacking as a and patient satisfaction, are constructed level are often incomplete or not avail- from the 2011 HFS to measure the result of poor urban planning. able. This limits the scope of adap- quality of health service delivery. A challenge for policy makers is to tation capacity assessments that are Capacity in the non-health sector is determine how to prioritize and allo- urgently needed to inform decisions on measured by indicators of environ- cate public resources to support the public resource allocation across both mental infrastructure constructed from development of health adaptation sectors and locations. In this section, the 2011 census, covering the propor- the district-level measures of health tion of households with access to safe capacity in locations where the local adaptation capacity are constructed drinking water, access to sanitary toilet population is most vulnerable to using complementary data sources, facilities, and access to electricity. The health risk and climate shocks and including the 2011 Bangladesh health spatial distribution of overall health local health and basic environmental facility survey (HFS) and the 2011 adaptation capacity in 2011 is presented infrastructure is particularly lacking. census. in map 4.2. Map 4.2  Spatial Distribution of Overall Health Adaptation Capacity 46 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? According to the 2011 census, the Assessing the Efficiency of allocated to disease-prevalent districts that have inadequate access locations. to sanitation facilities (defined as less Geographic Targeting than 30 percent of a population with Spatial assessment of adaptation Diarrhea is the most common access to sanitary facilities) include capacity involves identifying disease infectious disease in Bangladesh, Bandarban, Thakurgraon, Gaibandha, hotspots and analyzing spatial accounting for about 10 percent of Sunamganj, Nilphamari, Nawabganj, correlation between disease incidence total deaths every year. The national Naogaon, and Rangama. The districts and level of local adaptation capacity. disease surveillance system provides that have inadequate access to safe Malaria is concentrated in three relatively reliable records of diar- drinking water (defined as less than districts (Bandarban, Rangamati, and rhea cases across the country. The 80 percent of the population with Khagrachhari), which accounted for top 12 districts with high incidence access to safe drinking water) include about 92 percent of total cases during of diarrhea cases in 2011, ranked Bangamati, Bandarban, Bagerhat, 2007–11. The cases of dengue fever in descending order, are Kushtia, Khagrachhari, and Sylhet. were reported primarily in Dhaka, and Kishoreganj, Munshiganj, Manikganj, the cases of kala-azar were reported Bandarban, Randarban, Rajshahi, These district-level measures of health mainly in five upazilas of Mymensingh Khagrachhari, Magura, Hennaidah, adaptation capacity, constructed District. Satkhira, and Shariyatpur. Table 4.1 from the 2011 HFS and the 2011 census presents the spatial distribution of together with disease data, provide From a public health point of view, the incidence of diarrheal disease critical information for assessing how information on the availability of against the ranking of health and well public resources are spatially diagnostic equipment and essential environmental services for the top targeted. That is, do districts with drugs in high-endemic regions is 12 districts of high-diarrhea prev- high disease incidence have adequate essential for assessing whether public alence. The ranking provides some adaptation capacity measured by recourses are well targeted to address evidence of poor targeting, with the availability of health services the risks of VBDs. Unfortunately, districts of high diarrhea prevalence and the level of basic environmental no such information is collected in ranked low in access to safe water infrastructure? The following section Bangladesh. However, focus group and sanitation as well as in essential assesses the efficiency of spatial interviews found that in many loca- drug availability. targeting using these district-level tions at high risk of VBDs, lack of diag- indicators. nostic devices and poorly equipped The efficiency of geographic targeting surveillance systems have been a key is further assessed using the Pearson issue in monitoring VBDs, suggesting correlation coefficient between the that inadequate resources are being incidence of diarrhea and a range of Table 4.1  Spatial Analysis: Disease Incidence versus Local Capacity, 2011 Zila Diarrhea Improved Safe drinking Health equipment Essential drug incidence (%) sanitation water availability availability (1 best; 65 worst) (1 best; 65 worst) (1 best; 25 worst) (1 best; 25 worst) Kushtia 4.96 38 14 5 19 Kishoreganj 3.83 52 41 10 21 Munshiganj 3.80 6 27 7 20 Manikganj 3.63 20 19 — — Bandarban 3.57 64 63 — — Rajshahi 3.48 42 28 — — Khagrachhari 3.46 56 61 11 6 Magura 3.38 13 6 — — Jhenaidah 3.37 41 10 1 23 Satkhira 3.36 39 57 11 16 Shariyatpur 3.34 5 33 9 1 Khulna 3.30 12 56 9 6 Rajbari 3.11 10 15 — — Rangamati 3.10 57 64 11 6 Note: — = no data. Chapter 4 | Population Dynamics and Spatial Targeting: Implications for Climate Change and Health | 47 adaptation capacity indicators for 66 safe water access is −0.3, equipment infectious disease in Bangladesh, districts in 2011.25 Adaptation capacity availability (−0.1), essential drug similar results may be found for other is defined as previous investments availability (−0.08), staffing index health indicators, suggesting that in water supply and sanitation and (−0.04), and patient satisfaction overall public resources for health in public health service delivery. The index (−0.03). adaptation are poorly targeted in results of the correlation analysis, as Bangladesh. presented in table 4.2 and figure 4.1, These findings provide strong evidence show that, with the exception of of poor geographic targeting of public The finding of poor geographic access to electricity, all capacity indi- resources in the area of health adap- targeting of public resources in the cators are negatively correlated with tation interventions; locations with area of public health is consistent the incidence of diarrheal disease. larger disease burden, in general, have with findings from other studies in For example, the Pearson correlation weaker adaptation capacity. Given Bangladesh. Two important spatial between diarrhea incidence and that diarrhea is the most common assessments have been carried out in Table 4.2 Pearson Correlation Coefficient between Adaptation Capacity and Disease Incidence: 2011 Indicator Pearson Median correlation Basic infrastructure capacity % of households with access to safe drinking water (tap and tube well are defined as safe −0.30 96.30 drinking water) % of households with access to sanitary toilet facility (water-sealed and non-water-sealed −0.04 61.45 toilets are defined as sanitary) % of households with access to electricity 0.17 48.65 Basic health service provision capacity Equipment availability index (%) −0.10 66.67 Essential drug availability index −0.08 50.00 Staffing index −0.04 61.48 Patient satisfactory index −0.03 75.20 Incidence of diarrhea (%) 1.39 Sources: Basic infrastructure indicators are from 2011 census, and health service indexes are from 2011 HFS. FIGURE 4.1  Correlation between Health Facility and Disease Incidence: 2010 District-Level Data The choice of diarrhea reflects the wide geographic coverage of cases, as all the vector-borne diseases under study are concentrated in limited 25  geographic locations. 48 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Bangladesh in recent years, including Key Messages national surveys should be treated poverty mapping at the subdistrict with caution. Given the inade- zz The results from this chapter, quate sampling of the urban slum level using the income/expenditure combined with evidence from population, spatial analysis based measure of poverty (World Bank other spatial assessment studies on existing national surveys may 2013) and the spatial assessment of (World Bank 2010; UNICEF 2013) provide misleading information, in multiple indicators covering educa- provide strong evidence of the particular, for the spatial targeting tion and health outcomes, access poor geographic targeting of public to health services, and access to of programs or projects to improve resources in health and basic access of the population to basic safe water and sanitation facilities environmental infrastructure in (UNICEF 2013). services and disease control. Bangladesh. While these findings are not linked directly to climate zz The issue of inadequate water and Combining spatial information from change, they suggest that locations both assessments, the 2009 UNICEF sanitation in urban areas is urgent with high disease risk are particu- and needs to be addressed. As the report highlights the strong spatial larly lacking in access to health and correlation between lack of access total population of Bangladesh basic environmental services that is projected to increase by 64.6 to basic social services and income/ are essential in reducing health million to about 217.9 million expenditure poverty. The report risk. However, the findings are people between 2010 and 2030, recommends the policy of “blanket only applicable at the aggregate with three-fourths of that growth geographic targeting” of basic level (districts) because data are occurring in urban areas, the health social services to the most deprived lacking at a more disaggregate level implications of these population geographic areas of Bangladesh as in rural areas and in urban slum dynamics and the rapid population a means of achieving MDGs related and non-slum areas. These find- growth in urban slums should be to equity as well as accelerating the ings point to the need for better rigorously assessed. Policy makers overall reduction in poverty. targeting of resources in the future, need to recognize the full scale of given the increasing risk of climate the health threat posed by rapid The World Bank (2010) also shows that change and disease incidence. but poorly planned urbanization public resources in the health sector in Bangladesh. The threat will are not spent effectively and health zz The sampling design of nation- continue to worsen if climate facilities are not adequately financed ally representative surveys such change is overlaid on these existing and staffed. In particular, the current as DHS and expenditure surveys capacity constraints. It is therefore resource allocations are not effectively has not been updated to capture important for the Government of reaching the poor. Inequalities in adequately the fast-growing urban Bangladesh to act quickly on the access to health services remain a key slum population, one of the most issue of inadequate safe water and issue and contribute to disparities in important features of popula- sanitation services, particularly in health outcomes between the rich tion dynamics in Bangladesh. slums where the urban population and poor.26 Indicators constructed from these has been growing the fastest. The recent Multiple Indicator Cluster Survey conducted in 2009 shows huge geographic disparities in health outcomes in Bangladesh, with the 26  under-five mortality rate, one of the important measures of population health, ranging from 102 deaths per 1,000 births in the district of Sherpur to only 43 per 1,000 in Pabna. Chapter 4 | Population Dynamics and Spatial Targeting: Implications for Climate Change and Health | 49 CHAPTER 5 COST-EFFECTIVENESS ANALYSIS OF HEALTH ADAPTATION INTERVENTIONS Chapter 5 | Cost-Effectiveness Analysis of Health Adaptation Interventions | 51 O ver the past few decades, interventions. This is particularly true outcome that is central to the program. Bangladesh has invested in the area of health due to the scarcity For example, many nutrition-focused extensively in health and of impact evaluation studies and cost programs generate multiple benefits, education, basic infrastructure, and data. For example, no conclusive impact including child health, school atten- risk and disaster reduction measures, studies have examined the two large- dance, and future earnings. In CEA, such as early warning systems and scale national nutrition projects—the only the nutrition benefit is quanti- construction of embankment and Bangladesh Integrated Nutrition Project fied, but in CBA all benefits (nutrition, shelters. These policy interventions (BINP) and the National Nutrition Project schooling, and future earnings) are have strengthened the country’s (NNP)—despite the significant amount of estimated. In CBA, the net benefit is overall adaptation capacity and resil- resources committed to the two projects calculated as total benefits minus total ience to climate risks, contributing (see Hossain et al. 2002; White 2006).27 costs, discounted back to the present to the improvement in development value in the base year of the program outcomes, especially health outcomes This chapter illustrates the use of when the costs and impact last over from the reduction in premature CEA for evaluating health adaptation many periods. CBA is most useful deaths. The fundamental policy chal- interventions using a variety of data when policy analysts are studying a lenge, for both decision makers in the sources available in Bangladesh. single program or policy to determine government and development part- The first section explains why CEA whether its total benefits to society ners is how to maximize the cost-ef- is preferred to CBA in the context of exceed the costs or when comparing fectiveness of resources in a fiscally health adaptation options. The second alternative programs in order to constrained environment. illustrates how CEA can be applied to identify the one that is achieving the rank investment in three key areas greatest net benefit to society. Decision Cost-effectiveness analysis (CEA) and central to averting premature deaths, makers in line ministries, such as cost-benefit analysis (CBA) are useful including investment in an early health and education, often consider tools for identifying adaptation inter- warning system and risk management this question. ventions (programs and projects), measures, in improving access to safe guiding decisions about where public drinking water, sanitation services, As a practical tool for guiding decision resources should be prioritized to and electricity, and in nutrition-fo- makers on where to target limited maximize development outcomes. cused interventions. resources for achieving a specific In reality, however, these tools are outcome cost-effectively, CEA is gener- rarely used, which is evident at all ally preferred to CBA in two situations. levels of policy making. For example, Why Cost-Effectiveness First, CEA is most useful when policy according to some estimates, bilateral Analysis Rather Than Cost- makers have identified a policy target and multilateral development partners Benefit Analysis? (for example, reducing the incidence currently allocate 98 percent of their of childhood disease) and want to disaster management funds for relief A large body of literature has been determine which alternative programs and reconstruction and only 2 percent developed to guide practitioners or projects will achieve the greatest for investment in an early warning on how to conduct CEA and CBA in outcome for a given budget. In system and risk reduction measures different contexts (Levin and McEwan contrast, the questions posed in CBA (Mechler 2005). This is clearly not a 2001; U.S. Department of Health and are broader and may not be of imme- cost-effective solution. The review of Human Services 1996; U.K. Government diate concern to policy makers in line four decades of World Bank project 2011). CEA calculates the costs of a ministries, whose primary objectives appraisal documents by the World program in relation to its targeted are to improve specific outcomes of the Bank’s Independent Evaluation Group outcomes (often a single outcome such population, such as health, education, (IEG 2010) concludes that the applica- as children’s school enrollment rate or or employment. tion of CBA in project appraisal often malnutrition rate). In CEA, the ratio of lacks consistency and objectivity, varies quantifiable outcome or impact of a Second, CEA is a better option in greatly in quality and rigor, and often program or project to a given cost asso- cases where the central outcomes fails to use the existing data sources ciated with its implementation (or the or targets are either intangible or and results from impact evaluation inverse ratio measuring the amount otherwise difficult to monetize. For studies that are widely available in of cost required to achieve a given example, placing a monetary value the published literature for individual impact, referred to as the cost-effec- on health, including mobility and countries where projects are taking tiveness ratio) is calculated to compare mortality, is particularly controversial place. cost-effectiveness across alternative and subject to large uncertainties. programs. The commonly adopted measure of In Bangladesh, despite extensive population health—disability-ad- public investment over the past CBA takes this process one step further justed life years (DALY)—combines several decades, little is known about by comparing costs with total program in one measure the time lived with the cost-effectiveness of different benefits rather than just the single disability and the time lost due to The BINP was carried out between 1996 and 2002, with funding of US$59.8 million, and the NNP was implemented between 2002 and 2011, with 27  funding of US$124.42 million. 52 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? premature mortality. The calculation to development, policy interventions is, the counterfactual), controlling of DALY imposes several subjective in traditional areas, such as access for all confounding factors. Many assumptions, including societal to basic services (health care, safe complementary methods, such as preferences for giving different drinking water, and sanitary toilet randomization experiments and weights to the health of people of facilities), female education, and regression-based methods, have different ages and to disabilities. child nutrition, will continue to be been developed to estimate the The monetary value per DALY is important in improving develop- benefit of various policy interven- far from agreed on among policy ment outcomes, including health. tions with the choice of methods, analysts. In this context, CEA is less In addition, investment in early depending on data availability (for restrictive, allowing users to apply warning systems and risk manage- a review of methods, see Gertler their own judgment to quantify the ment will enhance the population’s et al. 2010). health benefits or to avoid placing a resilience to climate shocks, gener- monetary value simply by using the ating substantial benefits measured In the following CEA analysis, the number of averted premature deaths by a reduction in economic losses, number of deaths avoided is used as as a measure of program impact. For mortality, and disease outbreaks from the measure of health benefit from the a good review on the choice of CEA natural disasters. three interventions. The methods used and CBA and the applications of CEA, to quantify the health benefits are see Dhaliwal et al. (2012) . This section presents a simplified summarized in table 5.1. CEA to compare three interventions In the context of health adaption with the central objective of reducing The estimated health benefit from interventions in Bangladesh, all premature deaths: (1) achievement the first option—achieving 100 three key areas of policy interven- of universal household access to percent household access to safe tions (investment in early warning safe drinking water, sanitation, and water, sanitation, and electricity—is system and risk reduction measures, electricity; (2) elimination of stunting about 133,000 child deaths averted improved access to safe drinking through targeted nutrition and health in 2011 using the World Health water and sanitation services, and programs; and (3) investment in Organization (WHO) estimate of nutrition-focused programs) have one early warning and risk management central objective in common: averting 25 percent of deaths caused by child- capacity.28 hood illnesses and 2011 population premature deaths. Therefore, CEA is preferred to CBA. data. The health benefit from the Benefit estimation second option—nutrition programs The major difficulty with CEA or CBA that eliminate stunting—is taken An Application of Cost- is that it is challenging to identify as from the policy simulation carried Effectiveness Analysis well as place monetary values on all out by USAID and FANTA (2012), which in Health Adaptation benefits and costs associated with shows that about 160,000 child a program or project. The benefit of deaths could be averted through the Interventions policy interventions, in the context reduction of stunting in 2011. It is often said that development is of health adaptation, should be the best form of adaptation. This measured as the change in health The benefit of investing in early means, while the risk of climate outcomes by comparing the scenarios warning and risk reduction systems change adds additional challenges with and without interventions (that is measured as the number of deaths Table 5.1  Summary of Methods for Quantifying Benefits Intervention Method Outcome indicator Reaching 100% household access The marginal impact of access to water, sanitation, and electricity on the Child deaths averted to safer water, sanitation, and incidence of childhood illness is estimated using regression method. (under 14 years of age) electricity in 2011 Child deaths averted are estimated using simulation based on the esti- mated marginal impact and the WHO estimate of 25% of child deaths from illnesses. Eliminating stunting by 2011 Based on estimates from existing studies; about 35–55 percent of deaths Child deaths averted among children under five are estimated to be caused by malnutrition (under 15 years of age) (Pelletier et al. 1995; Black et al. 2008; Hossain and Bhuyan 2009) Investing in early warning and Control and treatment method Deaths averted (all ages) risk reduction systems These policy targets include (1) universal access to safe drinking water (that is, tube well in rural areas and piped water in urban areas), sanitary 28  toilet facilities, and electricity; (2) female attainment of a secondary education; and (3) elimination of child stunting. They are set in reference to the current level of access to basic services, female education, and child malnutrition estimated from the 2011 DHS and the Millennium Development Goal targets set for 2015 (see appendix 5A, box 5.1). Chapter 5 | Cost-Effectiveness Analysis of Health Adaptation Interventions | 53 and sanitation and child nutrition Table 5.2  Major Cyclones, 1970–2009 programs, the cost data on these Landfall date Landfall Wind speed Storm surge Human projects are hard to come by. Data location (kilometers (meters) casualties on operation and maintenance per hour) (number) costs, on institutional overhead November 12, 1970 Bhola, Meghna, 224 10 300,000 costs, and on the costs of commu- Estuary nity-driven health programs, are April 29, 1991 Noakhali 235 7.6 133,882 especially scarce. However, the Chittagong recent study by USAID and FANTA November 15, 2007 Sundarban, 250 6–8 2,388 (2012) provides a good example for Borguna estimating the costs of implementing November 15, 2007 Burma 240 6–8 138,366 a comprehensive national program for nutrition in Bangladesh for the Sources: EMDAT. period of 2011–21.29 avoided using mortality data from past Cost estimation In the following, the cost estimation of major natural disasters in Bangladesh, Quantifying the cost of a program a national nutrition program is taken focusing on cyclones. Table 5.2 (or policy intervention), covering from the FANTA study. It estimates that summarizes the major cyclones both investment costs (for example, the cost of providing effective nutrition crossing the Bangladesh coast from construction of water pipes) and services covering the entire territory 1970 to 2009. recurrent costs (like maintenance of Bangladesh is in the range US$130 and management) can also be chal- million–$170 million per year (which The 1970 and 1991 cyclones were lenging. First, program costs depend is relatively small compared with the similar in severity, as measured by on a wide range of local factors, in annual health sector budget). wind speed and storm surge, and particular, local implementation so were the 2007 cyclones that hit The cost estimation of the disaster capacity; therefore, cost data are Bangladesh and Burma. Therefore, context specific and not easily trans- risk mitigation investment shows the reduced death tolls from these ferable if no cost data are collected. that, since 1971, the Government cyclones can be used to measure the Second, in most cases, incremental of Bangladesh, with support from avoided deaths as a result of invest- cost, rather than unit cost, should international development agen- ments in an early warning system be used to measure program costs, cies, has invested more than US$10 and other disaster risk management which is not easily constructed. billion to make the country more measures or the benefit of these For example, the incremental cost resilient to climate change, covering interventions. The reduced deaths of expanding household access to many areas such as risk and natural are about 170,000 by comparing the piped, safe drinking water can be disaster management, transportation, 1970 and 1991 cyclones and 136,000 quite low in areas where the system and agriculture. This is equivalent by comparing the 2007 cyclones in of underground water supply has to an annualized cost in the range Bangladesh and Burma. The health been built compared with locations US$120 million–US$200 million. benefit—averted mortality—can be without a source of water. The same However, it is not possible to disag- attributed largely to investment in an applies to costing the investment in gregate the total cost of investment to early warning system and other risk early warning and risk management obtain the cost of investment in risk mitigation measures, although overall systems. A review of the literature and natural disaster management for development (such as better housing, shows that very few program impact the CEA or the annual incremental cyclone shelters, and roads) also studies in developing countries cost of investment in risk mitigation played an important role, suggesting report comprehensive cost estimates for the CEA. that these numbers are likely the (Dhaliwal et al. 2012). upper-bound estimation of benefits The cost data for the three interven- from investing in an early warning In Bangladesh, despite exten- tions of interest here are summarized systems. sive investments in water supply in box 5.2. The FANTA estimation exercise covers five procedures: (1) identifying an appropriate structure for the program, (2) selecting the necessary 29  interventions and activities, (3) determining a management structure and a method of service provision, (4) identifying the inputs and obtaining the unit costs of each activity and input, and (5) estimating the program costs for duration of the project. 54 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Cost-effectiveness analysis: BOX 5.2  Cost of Adaption Options An illustration The benefit-cost ratio estimated using zz Early warning system. More than US$10 billion has been invested FANTA’s benefit and cost estimation in reducing disasters and risks in Bangladesh over the past 50 years. is presented in the last column of These investments include infrastructure (polders, cyclone shelters, and table 5.3. The results show that nutri- cyclone-resistant housing) and early warning systems. These investments tion programs are highly cost-effective have significantly reduced the damages and losses from extreme weather in preventing premature deaths in events, in particular, the number of deaths and injuries avoided. The Bangladesh (more than 600 deaths average annual cost is assumed to be in the range US$120 million–US$200 of children under 14 averted per US$1 million. million), compared with the other two options. Carrera et al. (2012) carried zz Nutrition. The FANTA study estimates that the cost of providing effec- out a CEA using data from 14 coun- tive nutrition services covering all of Bangladesh are in the range US$130 tries on the impact of national health million–US$170 million per year. This is relatively small compared with the programs on the number of under- annual health sector budget, and it is achievable. five deaths averted. Their results show that about 81 deaths in children zz Safe water and sanitation investment. A water sanitation project report under five can be averted by investing for Bangladesh (World Bank 2005) estimates that the cost of constructing US$1 million in national health a water point and connecting 500 people to a location is about US$25 per programs such as early and exclusive household, for sanitation, it is US$45 per household. breastfeeding, complementary infant feeding, and improved water, sanita- tion, and hygiene practices. However, these estimates should be interpreted with caution due to uncertainties about some of the model parameters and baseline data. The result from the CEA should also be treated with caution due to data weakness, and the results should be used mainly for illustration purposes. Drawing policy conclusions regarding the cost-effectiveness of different interventions requires further efforts to collect data on costs, more rigorous impact evaluation studies using a variety of data sources and validation of data, and impact studies using complementary approaches. Over the past few years, a signifi- cant amount of resources have been channeled to improving malnutrition in Bangladesh. Unfortunately, little evidence has been generated from properly designed impact evaluation studies in this area (see box 5.3). Existing studies show a limited impact of nutrition-based programs. It is plausible that nutrition programs can have a much larger impact in reducing malnutrition and disease prevalence when combined with sanitation projects and a hygiene education campaign targeted at mothers. This is obvious based on the well-established medical evidence of the two-way causal effect of disease incidence and malnutrition among young children Chapter 5 | Cost-Effectiveness Analysis of Health Adaptation Interventions | 55 Table 5.3  Cost-Benefit Analysis: Comparison of Three Adaptation Options Benefit-cost ratio Policy option  Annualized cost Annualized Number of DALY per US$1 (US$ million) benefit deaths averted (number of per US$1 million deaths averted) 100% access of safer water and sanitation 300–500 133,000 (under 14) 260-443 0.04-0.06 Elimination of stunting 130–170 160,000 (under 14) 627-820 0.07-0.09 Early warning and risk management systems 120–200 14, 000 (all ages) 70-116 0.004-0.01 Note: The total number of households is 31,863,396 based on the 2011 census, which is used to estimate the cost of achieving 100% access to safe water and sanitation. The number of deaths avoided from natural disasters as a result of the early warning and risk management systems is calculated under the assumption of a return period of 10 years (that is, a probability of 10 percent for every year). The annualized cost of eliminating stunting is adjusted by a 50% increase from FANTA cost estimates. The DALY is estimated using a disability weight of 0.02 and life expectancy of 75. and the important health determi- rigorous evaluation methods based on zz The results of CEA should be widely nants of access to safe water, sanitary appropriately collected data including disseminated among stakeholders toilet facilities, and electricity. Impact both baseline and follow-up surveys. as a means to improve transparency evaluation studies on whether inte- as well as develop an evidence- grated programs in health adaptation based decision-making process for interventions—such as nutrition, Key Messages selecting projects and programs water and sanitation, and hygiene zz Given that considerable amount and identifying new areas of policy education targeted to the caretakers of of public resources have been interventions. children—can effectively reduce child allocated to improving climate malnutrition as well as disease inci- change adaptation and resilience in zz However, without concerted efforts dence would provide valuable infor- Bangladesh, CEA should be used in both to collect reliable data on cost mation needed to guide the design policy making to guide the alloca- and impact of different projects and implementation of future health tion and prioritization of resources. or programs and to implement adaptation interventions. This recommendation should apply CEA with rigor and consistency, both to key government units and evidence-based policy making will The analysis presented here illus- to donor agencies. remain simply an empty promise. trates how various complementary approaches, including regres- sion-based impact analysis and policy BOX 5.3  Two Major Nutrition Projects simulation exercises, can be used to conduct cost-benefit analysis to assess different health adaptation The Bangladesh Integrated Nutrition Project, with funding of a US$59.8 interventions. While the cost-benefit million, was implemented between 1996 and 2002, covering a population of 15.6 million in 59 thanas (a locality with a population of approximately ratio presented in this chapter should 200,000–450,000 people). The impact evaluation study (Hossain, Duffield, be taken with caution due to the and Taylor 2005) shows no evidence that the project achieved its objectives weakness of data quantifying both of reducing child malnutrition, measured by the prevalence of underweight costs and benefits, the CEA exercise children under the age of two. identifies an important information deficiency in guiding policy makers The Bangladesh National Nutrition Project, with funding in the amount of seeking to target limited resources to US$124.42 million (which is about 7.5 percent of annual aid received by the maximize health outcomes. country), began in 2000, with the objective of supporting the government’s 15-year vision to extend community nutrition services to the entire country. The key data gaps include lack of cost Project impact studies found no conclusive evidence regarding the impact of data on projects and programs related the project (White and Masset 2006). to investment in expanding house- hold access to safe drinking water, Based on child nutritional status estimated from the 2000 and 2011 DHS, child sanitation, and electricity, separately nutrition improved moderately: with the prevalence of underweight and for urban and rural areas, investment stunting at the national level reduced from 47 and 36 percent, respectively, in in an early warning system, and risk 2000 to 44 and 41 percent, respectively, in 2011. But how much the reductions management measures. Also lacking at the national level can be attributable to the NNP remains unknown. are impact evaluation studies that use CHAPTER 6 CONCLUSIONS AND WAY FORWARD Chapter 6 | Conclusions and Way Forward | 57 Bangladesh is one of many countries weather events are projected to be to implement CEA with rigor and already facing enormous challenges due significant by 2050, but well-tar- consistency, evidence-based policy to extreme events such as droughts and geted development investments making will remain simply an empty land and coastal flooding. Added to can mitigate all of the excess health promise. these challenges are demographic and burden attributable to climate socio-economic factors, such as rapid change. A simple message is that population growth and fast urbaniza- focusing resources in traditional Way Forward tion, poverty, poor health conditions, areas of development, such as There is an urgent need in Bangladesh to water scarcity, and inadequate sanitary improving access of the population incorporate health concerns into the deci- conditions. Climate change is, therefore, to safe drinking, sanitation, elec- sions and actions of sectors, as they plan an additional stressor that is expected tricity, female education, and child to mitigate and adapt to climate change, to increase the burden of diseases, nutrition, as well as strengthening so that these decisions and actions notably through increased morbidity the capacity to manage risks can be enhance health. Measures to mitigate the and mortality. Bangladesh is particularly the best health adaptation option in adverse health impacts of climate change vulnerable to outbreaks of infectious, the face of increased climate risk. need to be considered beyond the health water-borne, and vector-borne diseases. sector: for example, integrating health zz Second, policy makers need to concerns in water and sanitation, urban These risks and diseases are not new, recognize the full scale of the health planning, early warning, and a disaster and the health sector is already tackling threat as a result of rapid but poorly management systems. Resources should them. However, the capacity to cope planned urbanization and act be spatially targeted to reach the most with potentially increasing levels of quickly on the issue of inadequate vulnerable locations that are likely to be risks and diseases arising from climate safe water and sanitation services, change is still limited in Bangladesh. particularly in urban slum areas with at high climate and health risk to ensure Furthermore, there is insufficient the fastest population growth. The cost-effectiveness. capacity for assessment, research, and evidence of poor geographic targeting communication on climate-sensitive of past public investment in health While the study attempts to use all health risks, as well as insufficient and environmental infrastructure possible data sources, future efforts capacity to design and implement miti- highlights the need to develop a should focus on developing the insti- gation and adaptation programs. comprehensive spatial database—at tutional capacity within Bangladesh to a disaggregate level such as in rural develop and manage a database that Recognizing the importance of climate areas and slum and non-slum loca- integrates GIS, health surveys, and a change in the broader health context, tions in urban areas—for monitoring disease surveillance system with local Bangladesh became one of the first and improving spatial targeting. climate information and environmental countries to establish a climate change facilities at the national level. Such cell in the Ministry of Health. The Climate zz Third, despite several decades of capacity is needed to improve research in Change and Health Promotion Unit experience in programs and proj- the area of health and climate change. (CCHPU) has been established under the ects designed to improve health The research activities should be carried Ministry of Health and Family Welfare. in Bangladesh, evidence on the out as part of the government’s efforts CCHPU’s objectives are (1) to coordinate cost-effectiveness of different inter- to develop climate adaptation policies to all health promotional activities of ventions is lacking. Considerable ensure evidence-based policy making. intra- and interministerial initiatives; resources will be committed to (2) to increase awareness of the health climate change adaptation and To meet these objectives, the CCHPU consequences of climate change; (3) to resilience in the future, and needs considerable strengthening, strengthen the capacity of health systems cost-effectiveness analysis should particularly in the capacity to collect to provide protection from climate-re- be conducted routinely when allo- and analyze data, to conduct commu- lated risks through e-health and cating resources to programs and nity outreach, to implement and eval- telemedicine; (4) to ensure that health projects, both by government units uate pilot projects to reduce climate concerns are addressed in decisions and by development partners. health risks, and to mainstream health to reduce risks from climate change in concerns across various sectors. A other key sectors; (5) to conduct research zz The results of CEA should be widely stronger CCHPU could play a catalytic and evaluate and monitor programs disseminated among stakeholders related to health promotion and climate as a means to improving trans- role in combating the health impacts of change; and (6) to coordinate emergency parency as well as developing climate change and protecting human medical services and school health evidence-based decision making health from current and projected promotion to reduce health hazards for selecting projects and programs risks due to climate change through during disasters and emergencies. and identifying new areas of policy its analytical, advisory, and outreach interventions. The political economy services and in collaboration with nature of the decision-making other research organizations like the Key Messages process in Bangladesh should be International Center for Disease and Three key messages emerge from this fully recognized, but this is a subject Diarrhoeal Research, Bangladesh. study: beyond the scope of this study. At the same time, climate change Without concerted efforts to collect concerns need to be mainstreamed in zz First, the health impacts of increased reliable data on cost and impact of the Health Nutrition and Population climate variability and extreme different projects or programs and Sector Program for Bangladesh. APPENDIX A  DATA SOURCE FOR WATER- AND VECTOR- BORNE DISEASES Three main institutions in Bangladesh diarrheal disease and vector-borne The IEDCR collects information on are directly involved in collecting infection surveillance. In addition, water- and vector-borne diseases health information including data for the National Institute of Population through its disease surveillance water- and vector-borne diseases: (1) Research and Training (NIPORT) obtains system. Table A.2 shows the systems Directorate General of Health Services limited data on water- and vector- that are collecting information related (DGHS) under the Ministry of Health borne diseases through nationally to water- and vector-borne disease and Family Welfare (MoHFW), (2) the representative Demographic and through the IEDCR surveillance Institute of Epidemiology, Disease Health Surveys (DHSs) at regular system, the Diarrheal Disease and Control, and Research (IEDCR), and (3) intervals. Enteric Infection surveillance system, the International Center for Disease and the DHS.31 Priority communi- and Diarrhea Research in Bangladesh The MIS collects information from cable disease surveillance (PCDS) and (ICDDR,B). The DGHS is responsible the regular information systems of sentinel surveillance (SS) in selected for the Management Information different health programs such as areas obtain information on various System (MIS) of the health system, malaria, tuberculosis, Essential Service diseases, including water- and and the IEDCR collects information Delivery (ESD), Communicable Disease vector-borne diseases like diarrheal on water- and vector-borne diseases Control (CDC). Hospitals provide the diseases (acute watery diarrhea and by conducting disease surveillance, CDC with information related to water- bloody dysentery), malaria, and investigating outbreaks, and training and vector-borne diseases.30 Table A.1 kala-azar. researchers. The ICDDR,B focuses on shows the status of data for water- and vector-borne diseases. Table A.1  Health MIS: Diseases and Coverage Key information Coverage (year, region) Name of organization and website Pneumonia, no pneumonia (cough and 2002: 420 upazila in 50 districts and Integrated Management of Childhood Illnesses cold), diarrhea, fever (malaria), fever communities in 15 upazilas (monthly) (IMCI), (ESD), DGHS (www.dghs.gov.bd) (not malaria) among under-five children Malaria burden, P. falciparum, P. vivax, death 13 districts of eastern and northern parts Communicable Disease Control, DGHS of the country (www.dghs.gov.bd) Kala-azar (visceral ieishmaniasis), cases, 1999: 100 upazilas; 2000: 26 districts Communicable Disease Control, DGHS deaths (www.dghs.gov.bd) Dengue, cases, deaths 2000: reported from facilities Communicable Disease Control, DGHS and individual doctors (www.dghs.gov.bd) Diarrhea (demographic, etiologic, clinical, 2000: reported from facilities, whole Communicable Disease Control, DGHS and therapeutic aspects of patients) country (www.dghs.gov.bd) and ICDDR,B (www.icddrb.org) 30 Computers have been provided to all national and regional tertiary hospitals, 64 district health managers, and many of the 464 subdistrict hospitals. These computers are connected through the Internet. Hospital-based service data are still collected in formats compiled locally, with limited possibility of disaggregation. 31  Upazila health and family planning officers and civil surgeons are responsible for conducting surveillance locally. Appendix A  Data Source for Water- and Vector-Borne Diseases | 59 Table A.2  Data Sources of Three Surveillance Systems System and name of organization Key information Coverage (years, regions) IEDCR Priority communicable disease surveil- Diarrheal diseases (acute watery diarrhea and 2007 onward; data collected weekly from at lance (PCDS) bloody dysentery), malaria, kala-azar least one upazila in each of 64 districts Institutional disease surveillance Diarrheal diseases (acute watery diarrhea and Monthly disease profiles from the medical bloody dysentery), malaria, kala-azar college hospitals and specialized institutions Sentinel surveillance (SS) Diarrheal diseases (acute watery diarrhea and Information obtained from eight selected bloody dysentery), malaria, kala-azar upazilas of eight selected districts; discon- tinued in 2009 and merged with PCDS Diarrheal Disease and Enteric Infection surveillance system ICDDR,B (www.icddrb.org) Socio-economic, demographic, housing, and Dhaka hospital from 1979; Matab hospital environmental conditions, feeding practices, from 1999; extensive microbiological assess- use of drugs and fluid therapy at home, ments of fecal samples (microscopy, culture, clinical, anthropometric measurements, and and enzyme-linked immunosorbent assay) treatments received at facilities Demographic and Health Survey National Institute of Population Research Diarrhea, fever, ARI, socio-demographic DHS 1993–94, 1996–97, 1999–2000, 2004, and Training (NIPORT), characteristics 2007, 2011 (national and divisional coverage) (www.niport.gov.bd) and ICF-MACRO (www.measuredhs.com) APPENDIX B CLIMATE AND NATURAL DISASTER DATA Three key departments are respon- surface and upper-air observatories, information, and issue warning sible for collecting climate- and radar and satellite stations, agro-me- messages. disaster-related data: the Bangladesh teorological observatories, geo-mag- Meteorological Department (BMD), netic and seismological observatories, The DMIN links 64 districts and under the Ministry of Defense; the and a meteorological telecommunica- 482 subdistrict offices (hardware and Bangladesh Water Development Board tion system. software support, Internet, training (BWDB); and the Disaster Management in information and communication Information Network (DMIN) portal The BWDB collects ground and surface technology) through its DMIN portal within the Disaster Management water information and monitors with the DMB, the BMD, the FFWC, Bureau (DMB). floods and drought situations. The and the Cyclone Preparedness Program Flood Forecasting and Warning Centre (CPP). It plays an important role in risk The BMD collects real-time meteoro- (FFWC) of the BWDB was established reduction by providing information on logical information from 35 observato- in 1972 and became a Division of the risks, mapping risk reduction activities, ries. It also provides routine weather Processing and Flood Forecasting Circle maintaining information databases forecasts for the public, farmers, mari- of BWDB Hydrology, which, in turn, is on disaster management capacity, ners, and aviators and issues warnings under the control of the chief engineer and operating a national informa- for severe weather phenomena such for hydrology. The FFWC is working to tion portal. Table B.1 summarizes the as tropical cyclones, tornados, and develop flood forecast and inundation data collection from the three key floods. It maintains a network of models, disseminate flood forecast departments. Table B.1  Weather and Disaster Management Data System Key information Coverage (years, regions) Data source Rainfall surface data Daily up to 2002; three-hourly from 2003 35 weather stations of BMD (www.bmd.gov.bd) Maximum and minimum temperature Three-hourly Dew point temperature, relative humidity, present and Daily past weather, cloud and height Date and time of magnitude and intensity of At the time of occurrence earthquake Voice data High-frequency wireless network, 67 stations Water data from BWDB: Flood Forecasting and Warning Centre (www.ffwc.gov.bd) Telemetry system satellite imagery 14 stations; GMS, NOAA-12 and NOAA-14 Monitoring of cloud and depression movements; Catchment area = 82,000 square kilometers; estimation of precipitation from cloud temperature total length of modeled rivers = 7,270 kilometers; analysis; cyclone monitoring; flood forecast modeling: number of catchments = 216; number of fore- one-dimensional fully hydrodynamic model (MIKE 11 cast stations = 30; flood maps generated from HD) incorporating all major rivers and floodplains model results via GIS link to model (MIKE 11 GIS) Real-time data management: GIS-based map showing Generation of flood status at upazila level water level and rainfall status (flood watch); display of forecast water levels and discharges; automatic genera- tion of flood forecast bulletins Climate change database: natural and geographic; phys- Comprehensive Disaster ical and infrastructure; hydrometrological data (tempera- Management Programme ture, rainfall, sunshine, humidity, evaporation, wind (CDMP), www.cdmp.org.bd speed, aerosol, water level, and sea surface temperature) APPENDIX C NATIONAL HEALTH AND FACILITY SURVEY Health Facility Survey corporations and a sample of district units belonged to one of two types of municipalities. community: slum and non-slum areas As shown in table C.1, the nationally of the city corporations. These units representative random sample of The 2006 UHS had three main served as the basis for the basic statis- public sector health facilities at the objectives: (1) to obtain a profile of tical domains of the study: district, upazila, and union levels health problems and health care– has been selected for the 2011 Health seeking behavior in urban areas of zz Dhaka Metropolitan Area large slum Facility Survey (HFS). Within each Bangladesh; (2) to identify vulnerable areas (by population) tier, especially for maternal and groups and examine their health child health centers, subcategories profile and health care–seeking zz Dhaka Metropolitan Area small and are district hospitals, maternal and behavior; and (3) to examine the medium slum areas (by population) child welfare centers, upazila health individual, household, and neighbor- complexes, union health and family hood factors associated with health zz Dhaka Metropolitan Area non-slum areas welfare centers, union subcenters and outcomes and behaviors in urban rural dispensaries, and community areas. The 2006 UHS was designed zz Chittagong City Corporation slum clinics. to expand the base of knowledge areas regarding population health and Urban Health Survey health-related behavior in urban zz Chittagong City Corporation areas of Bangladesh, with a particular non-slum areas The Government of Bangladesh and emphasis on understanding vulnera- the U.S. Agency for International bility and environmental risk in urban zz Slum areas of the remaining city Development jointly conducted the settings. corporations (Khulna, Rajshahi, 2006 Urban Health Survey (2006 Barisal, and Sylhet) UHS) designed to obtain a broad The 2006 UHS was based on a multi- health profile of the urban popu- stage sampling scheme under which zz Non-slum areas of the remaining lation of Bangladesh. The 2006 UHS the primary sampling units were city corporations (Khulna, Rajshahi, is a rich, micro-level survey of the crafted explicitly to reflect a mean- Barisal, and Sylhet) health of communities, households, ingful notion of urban community or and individuals throughout the city neighborhood. The primary sampling zz District municipalities. Table C.1  Number of Health Facilities Selected for the 2011 Health Facility Survey Facility type Sample size Total a District hospital 40 59 Maternal and child health center 50 97 Upazila health complex 80 414 Union health and family welfare center 532 3,806 Community clinic 758 9,722 Private facilities at district level 43 2,501 a   As reported by the Ministry of Health. APPENDIX D  ANALYTICAL RESULTS Table D.1  Logit Regression Results Using Full Sample (Odds Ratio) (Random-Effect Estimation) Pre-monsoon season (2004 DHS) Monsoon season (2007 DHS) Variable Fever Diarrhea ARI Variable Fever Diarrhea ARI L_rain_svy | 0.805* 1.392 0.764 L_rain_svy | 1.323 0.953 0.787 rain_heavy1sd_D | 1.815* 1.226 1.473 rain_heavy1sd_D | 0.676* 0.742 1.032 rain_light1sd_D | 0.967 1.278 0.896 rain_light1sd_D | 1.445 0.956 0.963 L_rain_svy_L | 0.837*** 1.035 0.879 L_rain_svy_L | 1.267* 1.246 1.023 rain_heavy1sd~L | 1.18 1.09 1.286 rain_heavy1sd~L | 0.724 0.561* 1.084 rain_light1sd~L | rain_light1sd~L | 0.963 0.856 0.895 avetemp_svymth | 1.212*** 0.878 1.135* avetemp_svymth | 0.928 0.888 0.514* hum_svymth | 1.041* 1.047 1.058* hum_svymth | 1.014 1.016 0.976 _Iregion_2 | 2.645 0.739 1.669 _Iregion_2 | 0.541 0.722 0.987 _Iregion_3 | 2.536 0.701 2.307 _Iregion_3 | 0.565 0.752 0.748 _Iregion_4 | 1.329 3.568 2.322 _Iregion_4 | 0.426** 0.627 0.675 _Iregion_5 | 3.353* 0.877 2.089 _Iregion_5 | 0.510* 0.453 0.691 _Iregion_6 | 3.424* 2.422 2.702 _Iregion_6 | 0.188*** 0.425 0.534 _Iarea_2 | 0.731* 0.72 0.741 _Iarea_2 | 0.922 0.852 1.102 _Iwealth_du_2 | 1.003 0.724 1.019 _Iwealth_du_2 | 1.08 0.878 0.917 _Iwealth_du_3 | 0.666** 0.510* 0.682* _Iwealth_du_3 | 1.096 1.062 0.867 _Iwealth_du_4 | 0.600** 0.424* 0.454*** _Iwealth_du_4 | 1.005 0.946 0.704 _Iwealth_du_5 | 0.547** 0.359* 0.493* _Iwealth_du_5 | 0.95 0.587 0.436** _Iwater_2 | 0.771 0.230*** 0.756 _Iwater_2 | 1.105 0.490* 1.213 _Iwater_9 | 0.905 0.311* 0.939 _Iwater_9 | 0.912 0.422* 0.924 _Itoilet_21 | 1.398 1.107 1.71 _Itoilet_22 | 1.093 0.754 1.003 _Itoilet_22 | 1.219 0.974 1.589 _Itoilet_23 | 1.016 0.607* 0.947 _Itoilet_23 | 1.302 1.345 1.777* _Itoilet_99 | 0.997 0.905 0.899 _Itoilet_99 | 1.412 1.48 1.828* _Iwom_edu_1 | 1.134 1.145 1.015 _Ielect_1 | 1.076 1.4 1.334 _Iwom_edu_2 | 1.141 0.962 0.877 _ITV_1 | 1.316* 0.662 0.914 _Iwom_edu_3 | 0.901 0.367** 0.664 _Ireligion_2 | 0.721 1.415 0.574* _Ibord_new_2 | 0.927 0.892 1.011 _Ireligion_99 | 0.703 1.293 0.549 _Ibord_new_3 | 1.021 0.856 0.981 _Isex_head_2 | 1.014 0.728 0.866 _Ibord_new_99 | 1.065 1.021 0.97 _Iwom_edu_1 | 1.233* 1.273 1.394* age_chd | 0.883 1.164 0.824 _Iwom_edu_2 | 1.291* 1.117 1.429* age_chdsq | 0.991 0.921* 0.991 _Iwom_edu_3 | 1.094 1.141 1.092 _cons | 0.22 2.472 1.71e+09* _Ibord_new_2 | 0.998 1.162 1.07 _Ibord_new_3 | 1.122 0.892 1.345 _Ibord_new_99 | 1.158 1.004 1.12 age_chd | 1.045 1.039 0.781* age_chdsq | 0.946* 0.926 1 _cons | 0.001*** 0.085 0.000** _cons | 0.044*** 0.288* 0.166*** _cons | 0.132*** 0.224*** 0.110*** *** p<0.001; ** p<0.01; * p<0.05. Appendix D  Analytical Results | 63 Table D.2  Logit Regression Results Using Urban Households (Odds Ratio) (Random-Effect Estimation)   Pre-monsoon season (2004 UHS) Monsoon season (2007 DHS) Variable Fever Diarrhea ARI Variable Fever ARI rain_heavy1sd_D | 1.292 0.965 1.838 L_rain_svy | 1.225 1.915* rain_light1sd_D | 1.224 0.897 1.069 rain_heavy1sd_D | 0.747 0.354** avetemp_svymth | 1.086*** 1.008 0.996 rain_light1sd_D | 1.503 1.599 hum_svymth | 1.023 1.015 0.986 avetemp_svymth | 0.842* 0.752** _Iwater_urb_2 | 1.072 0.525 1.165 hum_svymth | 1.001 1.038 _Itoilet_ur_21 | 1.34 0.807 1.933** _Iwater_urb_2 | 1.274 0.98 _Itoilet_ur_22 | 1.241 0.76 1.924* _Iwater_urb_3 | 2.362 1.152 _Itoilet_ur_23 | 1.202 1.15 1.859* _Itoilet_ur_2 | 1.835*** 1.4 _Itoilet_ur_99 | 1.251 0.771 2.943** _Itoilet_ur_22 | 1.875** 1.479 _Iregion_2 | 0.803 0.269** 1.169 _Itoilet_ur_23 | 1.278 0.889 _Iregion_3 | 1.026 0.428* 0.747 _Itoilet_ur_99 | 1.684* 1.479 _Iregion_4 | 0.741 0.391* 0.763 _Iregion_2 | 0.699 0.526 _Iregion_5 | 1.178 0.477* 0.711 _Iregion_3 | 1.203 0.452 _Iregion_6 | 1.676 0.516 0.885 _Iregion_4 | 0.981 0.337* _Iwealth_du_2 | 1.082 1.361 0.975 _Iregion_5 | 1.16 0.298* _Iwealth_du_3 | 0.929 0.829 0.846 _Iregion_6 | 0.686 0.281 _Iwealth_du_4 | 1.013 1.097 0.796 _Iwealth_du_2 | 0.773 0.65 _Iwealth_du_5 | 1.089 0.552 0.833 _Iwealth_du_3 | 0.751 0.592 _Ielect_1 | 0.778 0.917 1.029 _Iwealth_du_4 | 0.945 0.988 _ITV_1 | 1.01 0.631 0.967 _Iwealth_du_5 | 0.976 0.587 _Ireligion_2 | 0.674 1.425 0.861 _Ielect_1 | 1.684* 1.479 _Ireligion_99 | 2.038 0 5.147 _Ielect_7 | 0.741 1.149 _Isex_head_2 | 1.017 1.357 0.814 _ITV_1 | 0.251 0.595 _Iwom_edu_1 | 1.112 1.033 1.058 _ITV_9 | 0.999 1.00E+00 _Iwom_edu_2 | 1.084 0.895 1.036 _Ireligion_2 | 2.14E+09 0 _Iwom_edu_3 | 0.975 0.986 0.898 _Ireligion_99 | 0.675 0.789 _Ibord_new_2 | 0.89 1.379 0.71 _Isex_head_2 | 1.426 0 _Ibord_new_3 | 1.153 0.597 1.37 _Iwom_edu_1 | 1.256 0.966 _Ibord_new_99 | 1.149 0.878 1.047 _Iwom_edu_2 | 1.517* 0.661 age_chd | 0.951 0.83 0.701* _Iwom_edu_3 | 1.311 0.552* age_chdsq | 0.964 0.994 1.023 _Ibord_new_2 | 1.148 0.478* _cons | 0.018** 0.175 0.803 _Ibord_new_3 | 1.017 1.114 _Ibord_new_99 | 1.023 1.108 age_chd | 1.222 1.01 age_chdsq | 1.053 1.14 _cons | 0.953 0.926           16.597 2.439 lnsig2u | lnsig2u | _cons | 0.084*** 0 0.062** _cons | 0.134*** 0 *** p<0.001; ** p<0.01; *p<0.05. 64 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Table D.3  Logit Regression Results Using Rural Households (Odds Ratio) (Random-Effect Estimation)   Pre-monsoon season (2004 DHS) Monsoon season (2007 DHS ) Variable Fever Diarrhea ARI Variable Fever Diarrhea ARI L_rain_svy | 1.052 1.141 1.017 rain_heavy1sd_D | 1.494** 1.539 1.232 rain_heavy1sd_D | 0.769 0.641 1.323 rain_light1sd_D | 1.028 0.843 1.049 rain_light1sd_D | 1.079 1.151 1.398 avetemp_svymth | 1.059*** 0.97 1.021 avetemp_svymth | 0.806*** 0.916 0.835** hum_svymth | 1.008 1.027 1.023 hum_svymth | 1.018 1.026 1.011 _Iwater_rur_2 | 1.058 0.862 1.435 _Iwater_rur_2 | 0.828 0.925 1.139 _Itoilet_ru_22 | 1.011 0.882 0.774 _Itoilet_ru_22 | 0.796 0.793 0.629* _Itoilet_ru_23 | 1.159 1.179 0.902 _Itoilet_ru_23 | 0.801 0.688 0.768 _Itoilet_ru_99 | 1.09 1.241 0.761 _Itoilet_ru_99 | 0.78 1.024 0.679 _Iregion_2 | 0.956 1.173 0.738 _Iregion_2 | 0.741 0.962 1.24 _Iregion_3 | 0.657** 0.773 0.570** _Iregion_3 | 0.906 0.922 0.691 _Iregion_4 | 0.672* 1.561 0.76 _Iregion_4 | 0.722 0.9 0.649 _Iregion_5 | 0.872 0.991 0.552** _Iregion_5 | 0.876 0.655 0.768 _Iregion_6 | 0.948 1.02 0.679 _Iregion_6 | 0.490* 0.826 0.912 _Iwealth_du_2 | 0.984 0.777 1.054 _Iwealth_du_2 | 1.122 0.921 0.933 _Iwealth_du_3 | 0.811* 0.79 0.852 _Iwealth_du_3 | 1.096 1.133 0.733 _Iwealth_du_4 | 0.734* 0.909 0.625** _Iwealth_du_4 | 1.043 1.007 0.557** _Iwealth_du_5 | 0.692* 0.713 0.531** _Iwealth_du_5 | 0.925 0.636 0.406** _Ielect_1 | 1.183 1.368 1.182 _Ielect_1 | 0.866 1.486* 0.801 _ITV_1 | 1.076 0.470** 0.94 _Ielect_7 | 1.017 0.603 0.702 _Ireligion_2 | 0.703* 1.006 0.703 _ITV_1 | 1.11 0.629* 1.207 _Ireligion_99 | 0.532 1.889 0.401 _Ireligion_2 | 0.638** 0.621 0.841 _Isex_head_2 | 0.942 0.708 0.934 _Ireligion_96 | 0 0 0 _Iwom_edu_1 | 1.058 1.226 1.094 _Ireligion_99 | 0.457 1.459 0 _Iwom_edu_2 | 1.073 1.021 1.009 _Isex_head_2 | 1.14 0.936 1.035 _Iwom_edu_3 | 0.823 0.679 0.522* _Iwom_edu_1 | 0.981 1.139 1.095 _Ibord_new_2 | 1.09 1.389 0.968 _Iwom_edu_2 | 1.063 1.133 0.984 _Ibord_new_3 | 0.952 1.399 0.864 _Iwom_edu_3 | 0.839 0.866 0.648 _Ibord_new_99 | 0.973 1.316 0.827 _Ibord_new_2 | 1.033 0.967 1.023 age_chd | 1.036 1.164 0.788* _Ibord_new_3 | 1.136 0.86 1.093 age_chdsq | 0.950** 0.905** 1.007 _Ibord_new_99 | 1.141 1.087 1.071 _cons | 0.133* 0.020** 0.080* age_chd | 0.827* 1.229 0.765* age_chdsq | 1.002 0.917* 1.006         _cons | 112.136** 0.112 26.867 _cons | 0.113*** 0.156*** 0.241*** _cons | 0.153*** 0.217*** 0.088*** *** p<0.001; ** p<0.01; *p<0.05. Appendix D  Analytical Results | 65 Table D.4  Estimation of Marginal Impact on the Incidence of Childhood Illness and Malnutrition (Odds Ratio), 2011 DHS   Rural sample  Urban sample Variable Illnesses WFH<2sd HFA<2sd Variable Illnesses WFH<2sd HFA<2sd Monsoon season 1.1 1.2 0.9 Monsoon season 1.07 1.13 0.70* Water source (reference: tube well) Water source (refer- ence: piped water) Piped water 1.13 0.61 0.55 Tube well 1.34* 0.63* 1.09 No water 1.35 1.47 1 No water 1.25 0.94 1.58 Sanitation (reference: improved Sanitation (reference: latrine) flush latrine) Pit latrine 1.16 1.16 1.40** Improved latrine 1.09 1 1.09 Open latrine 1.08 1.08 1.67*** Pit latrine 0.97 1.31 1.42* No latrine 1.06 1.15 2.00*** Open latrine 1.34 1.29 1.4 Flush 1.17 1.12 0.89 No latrine 0.81 0.65 1.42 Has electricity 0.95 0.89 0.83* Has electricity 0.77 0.78 0.58** Mother’s education Mother’s education (reference: secondary and above) (reference: secondary and above) Primary education 1.20* 1.14 1.41*** Primary education 1.19 1.3 1.57*** No education 0.97 1.19 1.67*** No education 1.11 1.26 2.07*** Stunted 1.20** 1.40*** Stunted 1.19 1.39* Sex_head = female 1.24* 0.97 1.04 Sex_head = 2 1.31 0.91 0.93 Child age 0.80*** 1.06* 1.18*** Child age 0.87*** 1.07 1.15*** Constant 0.93 0.10*** 0.22*** Constant 0.76 0.14*** 0.28*** Number of observations 5,200 5,200 5,200 Number of 2,135 2,135 2,135 observations Note: WFH = weight for height; HFA = height for age; sd = standard deviation. *** p<0.001; ** p<0.01; * p<0.05. Four types of health adaptation The magnitude of the net impact of zz Electricity. Access to householde- interventions are found to have a these four policy variables is summa- lectricity reduces the incidence of statistically significant health impact: rized as follows: stunting.32 A child living in a house- household sanitation facilities, access hold with electricity is 20 percent to electricity, nutritional status, and zz Sanitation. Upgrading household’s less likely to be stunted in rural mother’s education. The findings sanitation facilities has a large areas and 40 percent less likely in Bangladesh are broadly consis- impact on reducing the incidence in urban areas, everything else tent with evidence compiled from of chronic malnutrition (stunting). being the same. This finding may many low-income countries by the In urban areas, holding all other be linked to food storage. Unsafe World Health Organization (WHO 2011) conditions constant, a single food storage due to a shortage of showing consistently that children’s intervention—upgrading to a flush electricity has been identified as a health status (disease incidence and latrine or a septic sink from a pit serious health risk in many parts nutrition status) is influenced by a latrine—reduces the incidence of of Bangladesh due to the long range of household factors, such stunting by 42 percent. The health period of heat and humidity in the as living conditions with access to impact is even bigger in rural summer season. safe water and sanitary toilet facil- areas: upgrading to an improved ities, the availability of diverse and latrine from a pit, an open, or no zz Nutrition. Children’s nutritional nutrient-rich food, and maternal latrine, on average, reduces the status is an important determinant and child-care practices (hygiene incidence of stunting by 40, 67, and of childhood illnesses. Children practices). 100 percent, respectively. who are stunted are 20 percent more likely to suffer from illness The health benefits of household access to electricity are much less discussed in policy, although some important evidence exists on the impact of 32  access to electricity on reducing child mortality based on cross-country analysis of DHS data (Wang 2003). 66 | Climate Change and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? (diarrhea, ARI, or fever) and 40 zz Mother’s education. As expected, than a child living in a household percent more likely to be under- improving mothers’ education has with a mother with a secondary weight, holding other factors a large impact on reducing inci- education, all factors being the constant. The estimated impact dence of disease and prevalence same. Similarly, a child living with of stunting is similar among the of malnutrition. The results show a mother with primary education urban and rural populations. This that a child living in a household only is about 20 percent more finding indicates that significant with an uneducated mother is likely to get sick than a child cared health benefits can be generated twice as likely to be stunted in for by a mother with secondary by investing in nutrition-focused urban areas and 30 percent more education. programs. likely to be stunted in rural areas Table D.5  District-Level Change in Population, 2001–2010 Zila Population Growth Zila Population Growth 2001 2011 rate (%) 2001 2011 rate (%) Gazipur 2,031,891 3,403,912 6.75 Nator 1,521,336 1,706,673 1.22 Dhaka 8,511,228 12,043,977 4.15 Netrokona 1,988,188 2,229,642 1.21 Narayanganj 2,173,948 2,948,217 3.56 Jhenaidah 1,579,490 1,771,304 1.21 Sylhet 2,555,566 3,434,188 3.44 Chuadanga 1,007,130 1,129,015 1.21 Bandarban 298,120 388,335 3.03 Kushtia 1,740,155 1,946,838 1.19 Cox’s Bazar 1,773,709 2,289,990 2.91 Jessore 2,471,554 276,4547 1.19 Sunamganj 2,013,738 2,467,968 2.26 Munshiganj 1,293,972 1,445,660 1.17 Noakhali 2,577,244 3,108,083 2.06 Magura 824,311 918,419 1.14 Moulvibazar 1,612,374 1,919,062 1.90 Gaibandha 2,138,181 2,379,255 1.13 Habiganj 1,757,665 2,089,001 1.89 Meherpur 591,436 655,392 1.08 Brahmanbaria 2,398,254 2,840,498 1.84 Rajbari 951,906 1,049,778 1.03 Panchagar 836,196 987,644 1.81 Tangail 3,290,696 3,605,083 0.96 Narsingdi 1,895,984 2,224,944 1.74 Faridpur 1,756,470 1,912,969 0.89 Rangamati 508,182 595,979 1.73 Jamalpur 2,107,209 2,292,674 0.88 Comilla 4,595,557 5,387,288 1.72 Naogaon 2,391,355 2,600,157 0.87 Khagrachhari 525,664 613,917 1.68 Manikganj 1,285,080 1,392,867 0.84 Nilphamari 1,571,690 1,834,231 1.67 Jaipurhat 846,696 913,768 0.79 Laksmipur 1,489,901 1,729,188 1.61 Shariyatpur 1,082,300 1,155,824 0.68 Pabna 2,176,270 2,523,179 1.59 Satkhira 1,864,704 1,985,959 0.65 Feni 1,240,384 1,437,371 1.59 Chandpur 2,271,229 2,416,018 0.64 Chapai Nawabganj 1,425,322 1,647,521 1.56 Sherpur 1,279,542 1,358,325 0.62 Kurigram 1,792,073 2,069,273 1.55 Borguna 848,554 892,781 0.52 Chittagong 6,612,140 7,616,352 1.52 Patuakhali 1,460,781 1,535,854 0.51 Sirajganj 2,693,814 3,097,489 1.50 Bhola 1,703,117 1,776,795 0.43 Thakurgaon 1,214,376 1,390,042 1.45 Narail 698,447 721,668 0.33 Mymensingh 4,489,726 5,110,272 1.38 Madaripur 1,146,349 1,165,952 0.17 Rajshahi 2,286,874 2,595,197 1.35 Gopalganj 1,165,273 1,172,415 0.06 Rangpur 2,542,441 2,881,086 1.33 Pirojpur 1,111,068 111,3257 0.02 Lalmonirhat 1,109,343 1,256,099 1.32 Barisal 2,355,967 2,324,310 −0.13 Dinajpur 2,642,850 2,990,128 1.31 Jhalkati 694,231 682,669 −0.17 Bogra 3,013,056 3,400,874 1.29 Khulna 2,378,971 2,318,527 −0.25 Kishoreganj 2,594,954 2,911,907 1.22 Bagerhat 1,549,031 1,476,090 −0.47 Appendix D  Analytical Results | 67 Table D.6  Spatial Analysis: Disease Incidence versus Local Capacity, 2011 Zila Diarrhea Improved Safe Health Essential Zila Diarrhea Improved Safe Health Essential incidence sanitation drinking equip- drug incidence sanitation drinking equip- drug (%) (1 best; water ment avail- (%) (1 best; water ment avail- 65 worst) (1 best; avail- ability 65 worst) (1 best; avail- ability 65 worst) ability (1 best; 65 worst) ability (1 best; (1 best; 25 worst) (1 best; 25 worst) 25 worst) 25 worst) Kushtia 4.96 38 14 5 19 Comilla 1.37 15 43 11 7 Kishoreganj 3.83 52 41 10 21 Thakurgaon 1.31 63 3 — — Munshiganj 3.80 6 27 7 20 Jamalpur 1.30 45 21 6 18 Manikganj 3.63 20 19 — — Chapai 1.23 59 31 — — Nawabganj Bandarban 3.57 64 63 — — Sunamganj 1.21 61 54 11 15 Rajshahi 3.48 42 28 — — Cox’s Bazar 1.21 44 52 4 17 Khagrachhari 3.46 56 61 11 6 Feni 1.20 17 36 — — Magura 3.38 13 6 — — Sirajganj 1.08 34 24 5 3 Jhenaidah 3.37 41 10 1 23 Jhalkati 1.07 2 46 9 20 Satkhira 3.36 39 57 11 16 Naogaon 1.06 58 48 1 20 Shariyatpur 3.34 5 33 9 1 Tangail 1.01 31 29 5 1 Khulna 3.30 12 56 9 6 Jaipurhat 0.98 46 23 11 20 Rajbari 3.11 10 15 — — Narail 0.86 22 8 — — Rangamati 3.10 57 64 11 6 Bogra 0.83 29 9 3 10 Bagerhat 2.87 9 62 9 5 Brahmanbaria 0.81 25 34 11 12 Gopalganj 2.80 3 30 — — Jessore 0.80 35 2 4 16 Sherpur 2.52 47 42 — — Nator 0.74 27 38 11 1 Chandpur 2.51 28 45 11 19 Nilphamari 0.74 60 22 — — Sylhet 2.39 32 60 9 13 Chittagong 0.47 14 44 10 14 Netrokona 2.39 55 51 1 1 Faridpur 0.45 8 16 5 2 Dinajpur 2.37 48 4 11 22 Bhola 0.44 36 32 — — Habiganj 2.36 49 53 12 19 Panchagar 0.41 30 37 13 11 Chuadanga 2.24 51 7 — — Pirojpur 0.41 16 59 — — Laksmipur 2.17 18 49 9 1 Kurigram 0.39 40 17 2 8 Madaripur 2.12 24 25 — — Borguna 0.36 19 55 9 24 Narsingdi 2.04 33 11 — — Pabna 0.32 26 39 5 13 Moulvibazar 1.89 43 58 11 4 Lalmonirhat 0.31 37 18 1 4 Meherpur 1.60 50 26 — — Patuakhali 0.26 21 12 — — Dhaka 1.60 1 1 1 26 Barisal 0.13 4 40 1 25 Mymensingh 1.57 53 47 8 1 Gaibandha 0.12 62 35 5 9 Narayanganj 1.42 11 20 — — Noakhali 0.12 23 50 — — Gazipur 1.40 7 13 12 20 Rangpur 0.11 54 5 1 1 Sources: Diarrhea incidence is from the health Management Information Service, health facility capacity measures are from the 2011 national HFS, water and sanitation access are from the 2011 census. Note: — = no data. APPENDIX E POLICY SIMULATION OF HEALTH IMPACT The health benefit is estimated under Policy interventions in improving which uses different data sources and the assumption of achieving the nutrition are also shown to generate modeling methods, found that over following targets by 2020: (1) universal large health benefits. While the exact the 2011–12 period about 160,000 access to safe drinking water (that is, relationship between malnutrition child deaths could be averted by tube well in rural areas and piped and mortality is not well estab- reducing the prevalence of stunting. water in urban areas), sanitary toilet lished, the policy simulation suggests The health benefit from the three facilities, and electricity; (2) female that eliminating child malnutrition policy interventions combined is esti- secondary education attainment; (stunting) could avert more than mated to reduce more than one-third and (3) elimination of child stunting. 93,000 deaths, representing about of stunting cases and to avert more These policy targets are set in reference one-third of total deaths among chil- than 70 percent of child deaths to both the current level of access to dren 0–14 years of age.34 The policy annually in Bangladesh. basic services, female education, and simulation by USAID and FANTA (2012), child malnutrition estimated from the 2011 Demographic and Household Survey (DHS) and the Millennium Table E.1  Policy Simulation Targets Development Goal targets set for 2015 (see table E.1). Indicator Current level, MDG target, Simulation 2011 2015 target, 2020 Table E.2 presents the estimated Rural area health benefits from achieving the policy targets set for 2020 and 2030, Access of tube well (%) 85.5 96.5 100 respectively, taking account of popu- Improved latrine (%) 9.5 55.5 100 lation growth. The health benefit is Has electricity (%) 43.8 100 measured in terms of the number of Women with secondary and above 45.8 100 reduced cases of illness and malnu- education (%) trition and the number of deaths averted among children 0–14 years of Stunting (height for age z score < 2 36.8 0 age. The simulation results show that standard deviations) (%) improving household living condi- Population (thousands) 112,510 129,849 tions (access to safe drinking water, Urban area better sanitation facilities, and access Access of piped water (%) 38 100 100 to electricity) is estimated to reduce about one-quarter of stunting cases Latrine flush to septic (%) 42 85.5 100 and avert more than 46 percent of Has electricity (%) 82.5 100 deaths among children 0–14 years Women with secondary and above 59.7 100 of age in 2020.33 Achieving female education (%) secondary education attainment alone stunting (height for age z score < 2 28.7 0 is projected to reduce about 15 percent standard deviations) (%) of stunting cases and one-third of child deaths. Population (thousands) 26,743 80,399 Source: Current levels are based on the 2011 DHS; 2020 population projection is based on U.S. projection. Using the 2007 DHS data, WPS (2011) shows that inadequate sanitation is responsible for economic losses of about US$4.22 billion, equivalent to 33  6.3 percent of the country’s gross domestic product each year based on estimates from the 2007 DHS data, with premature mortality and other health-related diseases and disability accounting for about 84.3 percent of total cost. The World Health Organization (WHO) has estimated that almost 10 percent of the global burden of disease could be prevented through water, sanitation, and hygiene interventions. 34 In Bangladesh, the number of children under five classified as stunted and wasted was 5,958,000, and 2,251,000, respectively, in 2011 (UNICEF 2013). The projected number of stunting and wasting by 2020, under the scenario of “business as usual”—meaning that no health adaptation measures are taken—will rise to 7,690,000 and 2,905,000, respectively (attributable to population growth). Appendix E  Policy Simulation of Health Impact | 69 Table E.2 Estimating the Health Benefit of Targeted Policy Interventions Target Infrastructure Women’s Nutrition All combined education Reduced stunting, ages 0–5 (thousands) 1,908 1,156 none 2,563 % stunted in 2020 24.8 15.0 none 33.3 % stunted in 2030 20.6 12.5 none 27.6 Averted death, ages 0–14 (thousands) 133 96 93 212 % deaths averted in 2020 46.5 33.6 32.4 74.0 % deaths averted in 2030 49.8 35.9 34.6 79.2 Note: The table makes the following assumptions: (1) the proportion of urban population increases from 28 percent in 2011 to 35 percent in 2020 and to 45 percent in 2030, (2) the population ages 0–14 years old is 51 million in 2011 (WHO 2009), 66 million in 2020, and 61.8 million in 2030, (3) the total deaths of people 0–14 years of age is estimated at 221 million in 2011 (WHO and PAHO 2010), 286 million in 2020, and 267 million in 2030, and (4) the population 0–5 years of age is estimated at 19 million in 2011 (UNICEF 2013), 24.5 million in 2020, and 22.9 million in 2030. REFERENCES Alam, M. S. 2005. “Analysis of Kala-Azar Situation in Del Ninno, C., and M. Lundberg. 2005. “Treading Water: Mymensingh from 2000 to 2004.” Master’s diss., ICDDR,B. The Long-Term Impact of the 1998 Flood on Nutrition in Bangladesh.” Economics and Human Biology 3 (1): 67–96. Ali, A., A. Amani, A. Diedhiou, and T. Lebel. 2005. “Rainfall Estimation in the Sahel. Part II: Evaluation of Dhaliwal, I., D. E. Duflo, R. Glennerster, and C. Tulloch. Rain Gauge Networks in the CILSS Countries and Objective 2012. “Comparative Cost-Effectiveness Analysis to Inform Intercomparison of Rainfall Products.” America Meteorology Policy in Developing Countries: A General Framework Society 44 (11): 1707–22. with Applications for Education.” J-PAL Working Paper, Massachusetts Institute of Technology, Abdul Latif Jameel Poverty Action Lab (J-PAL), Boston. Bandyopadhyay, S., L. M. Wang, and S. Kanji. 2011. “The Impact of Rainfall and Temperature Variation on Children’s Gage, K. L., T. R. Burkot, R. Eisen, and E. Hayes. 2008. Diarrhea Incidence in Sub-Saharan Africa.” Applied “Climate and Vector-Borne Diseases.” American Journal of Geography 33 (April): 63–72. Preventive Medicine 35 (5): 436–50. Black, R. E., L. H. Allen, Z. A. Bhutta, L. E. Caulfield, M. Onis, Gertler, P. J., S. Martinez, P. Premand, L. B. Rawlings, and M. Ezzati, C. Mathers, and J. Rivera. 2008. “Maternal and C. M. J. Vermeersch. 2010. Impact Evaluation in Practice. Child Undernutrition: Global and Regional Exposures and Washington, DC: World Bank. Health Consequences.” Lancet 371 (9608): 243–60. Hales, S., S. J. Edwards, and R. S. Kovats. 2003. “Impacts on Burke, D. et al., eds. 2001. Under the Weather: Climate, Health of Climate Extremes.” In Climate Change and Human Ecosystems, and Infectious Disease. Washington, DC: Health, ch. 6. Geneva: World Health Organization. National Academy Press. Hashizume, M., B. Armstrong, S. Hajat, Y. Wagatsuma, Abu Campbell-Lendrum, D. H., C. F. Corvalan, and A Pruss- S. G. Faruque, T. Hayashi, and D. A. Sack. 2007. “Association Ustun. 2010. “How Much Disease Could Climate Change between Climate Variability and Hospital Visits for Cause?” In Climate Change and Human Health, ch. 7. Noncholera Diarrhea in Bangladesh: Effects and Vulnerable Geneva: World Health Organization. Groups.” International Journal of Epidemiology 36 (5): 1030–37. Carrera, C., A. Azrack, G. Begkoyian, J. Pfaffmann, E. Ribaira, T. O’Connell, and P. Doughty. 2012. “Equity in Child Survival, Hossain, S. M., A. Duffield, and A. Taylor. 2005. “An Health, and Nutrition 2: The Comparative Cost-Effectiveness Evaluation of the Impact of a US$60 Million Nutrition of an Equity-Focused Approach to Child Survival, Health, Programme in Bangladesh.” Health Policy Plan 20 (1, and Nutrition: A Modelling Approach.” Lancet 380 (9850): January): 35–40. 1341–51. ICDDR,B (Department of International Centre for Diarrhoeal Confalonieri, U., B. Menne, R. Akhtar, K. L. Ebi, M. Disease Research, Bangladesh). 2007. “Climate Change Hauengue, R. S. Kovats, B. Revich, and A. Woodward. and Bangladesh.” ICDDR,B News and Events, October 22. 2007. “Human Health.” In Climate Change 2007: Impacts, http://203.190.254.12/news/index.jsp?idDetails=130&typeOf- Adaptation and Vulnerability, edited by M. L., O. F. News=Newspercent20andpercent20Events. Canziani, J. P. Palutikof, P. J. van der Linden, and C. E. Hanson. Contribution of Working Group II to the fourth IEG (Independent Evaluation Group). 2010. “Cost-Benefit assessment report of the Intergovernmental Panel on Climate Analysis in World Bank Projects.” World Bank, Washington, Change Parry Cambridge University Press, Cambridge, United DC. Kingdom. REFERENCES | 71 IPCC (Intergovernmental Panel on Climate Change). 2001. Pascual, M., and A. Dobson. 2005. “Seasonal Patterns of Regional Impacts of Climate Change. Report of the Third Infectious Diseases.” PLoS Med 2 (1): e5. IPCC. Geneva: IPCC. Patz, J. A., P. R. Epstein, T. A. Burke, and J. M. Balbus. 1996. Islam, N., and H. Uyeda. 2007. “Use of TRMM in Determining “Global Climate Change and Emerging Infectious Diseases.” the Climatic Characteristics of Rainfall over Bangladesh.” Journal of the American Mosquito Control Association 275 Remote Sensing of Environment 108: 264–76. (3): 217–23. Jetten, T., and D. Focks. 2006. “Potential Changes in Patz, J. A., A. K. Githeko, J. P. McCarty, S. Hussein, U. the Distribution of Dengue Transmission under Climate Confalonieri, and N. De Wet. 2010. “Climate Change and Warming.” Wageningen Agricultural University, Graduate Infectious Diseases.” In Climate Change and Human Health, School for Production Ecology, Department of Entomology, ch. 6. Genva: World Health Organization. The Netherlands. Reeves, W. C., J. L. Hardy, W. K. Reisen, and M. M. Milby. Khan, A. E., A. Ireson, S. Kovats, S. K. Mojumder, A. Khusru, 1994. “Potential Effect of Global Warming on Mosquito- A. Rahman, and P. Vineis. 2011. “Drinking Water Salinity Borne Arboviruses.” Journal of Medical Entomology 31: and Maternal Health in Coastal Bangladesh: Implications of 323–32. Climate Change.” Environmental Health Perspectives 119 (9): 1328–32. Pelletier, D. L., E. A. Frongillo Jr., D. G. Schroeder, and J.-P. Habicht. 1995. “The Effects of Malnutrition on Child Mortality Levin, H. M., and P. J. McEwan. 2001. Cost-Effectiveness in Developing Countries,” WHO Bulletin OMS73, World Health Analysis, 2d ed. Thousand Oaks, CA: Sage Publications. Organization, Geneva. Matsuda, F., S. Ishimura, Yukiko Wagatsuma, T. Higashi, Rahman, R., M. Bangali, H. Kabir, F. B. Naher, and S. Taiichi Hayashi, Abu S. G. Faruque, David A. Sack, and M. Mahboob. 2008. “Kala-azar Situation in Bangladesh.” Nishibuchi. 2008. “Prediction of Epidemic Cholera Due to National Guideline and Training Module for Kala-azar Vibrio Cholerae O1 in Children Younger Than 10 Years Using Elimination in Bangladesh. 1st ed. CDC, DGHS, Ministry of Climate Data in Bangladesh.” Epidemiology and Infection Health and Family Welfare. 136 (1): 73–79. Rogers, D. J., and S. E. Randolph. 2006. “Climate Change McMichael, A., R. Woodruff, and S. Hales. 2006. “Climate and Vector-Borne Diseases.” Advances in Parasitology 62: Change and Human Health: Present and Future Risks.” 345–81. Lancet 367 (9513): 859–69. Rueda, L. M., K. J. Patel, R. C. Axtell, and R. E. Stinner. MoEF (Ministry of Environment and Forest). 2008. 1990. “Temperature-Dependent Development and Survival “Protecting Health from Climate Change.” Climate Change Rates of Culex Quinquefasciatus and Aedes Aegypti (Diptera: Cell, February. http://www.climatechangecellbd.org/publi- Culicidae).” Journal of Medical Entomology 27 (5): 892–98. cations/18protectinghealthfromcc.pdf. Streatfield, P. K., and A. K. Zunaid. 2008. “Population Murray, C. J., and A. D. Lopez. 1997. “Global Mortality, Challenges for Bangladesh in the Coming Decades.” Journal Disability, and the Contribution of Risk Factors: Global of Health, Population, and Nutrition 26 (3): 261–72. http:// Burden of Disease Study.” 349 (9063): 1436–42. www.icddrb.org/images/jhpn263-Population_Challenges_ for_Bangladesh.pdf. National Research Council. 2001. Under the Weather, Climate, Ecosystems, and Infectious Disease. Washington, Teshima A., M. Yamada, T. Hayashi1, Y. Wagatsuma, and DC: National Academy Press. T. Terao. 2007. “Climate Impact on Seasonal Patterns of 72 | Climate Change REFERENCES | 72 and Health Impacts: How Vulnerable Is Bangladesh and What Needs to Be Done? Diarrheal Diseases in Tropical Areas.” Working Paper. World Bank. 2005. “Water Sanitation Project Report.” World ICDDR,B. Bank, Washington, DC. U.K. Government. 2011. The Green Book: Appraisal and ———. 2009. “Economics of Adaptation to Climate Change: Evaluation in Central Government. London. Bangladesh.” World Bank, Washington, DC. UNICEF (United Nations Children’s Fund). 2013. Improving ———. 2010. “Bangladesh Public Expenditure and Child Nutrition: The Achievable Imperative for Global Institutional Review: Towards a Better Quality of Public Progress. Geneva: UNICEF. Expenditure.” World Bank, Washington, DC. USAID and FANTA (Food and Nutrition Technical Assistance). ———. 2013. Bangladesh Poverty Assessment: Assessing 2012. Investing in Nutrition Now: A Smart Start for Our a Decade of Progress in Reducing Poverty, 2000–2010. Children, Our Future; Estimates of Benefits and Costs of a Washington, DC: World Bank. Comprehensive Program for Nutrition in Bangladesh, 2011– 2021. Washington, DC: FANTA. WSP (Water and Sanitation Program). 2011. “The Economic Impacts of Inadequate Sanitation in Bangladesh.” World U.S. Department of Health and Human Services. 1996. “Cost- Bank, Water and Sanitation Program, Washington, DC. Effectiveness in Health and Medicine: Report to the U.S. Public Health Service by the Panel on Cost-Effectiveness in Yu, W. H., M. Alam, A. Hassan, A. S. Khan, A. C. Ruane, Health and Medicine.” Washington, DC: DHHS. C. Rosenzweig, D. C. Major, and J. Thurlow. 2010. Climate Change Risks and Food Security in Bangladesh. Washington, Wang, L. 2003. “Determinants of Child Mortality in Less DC: World Bank. Developed Countries: Empirical Findings from Demographic and Health Surveys.” Health Policy 65 (3) 56: 277–99. Watts, D. M., et al. 1987. “Effect of Temperature on Vector Efficiency of Aedes Aegyoti for Dengue 2 Virus.” American Journal of Tropical Medicine and Hygiene 36: 143–52. White, H., and E. Masset. 2006. “The Bangladesh Integrated Nutrition Program: Findings from an Impact Evaluation.” Journal of International Development 19: 627–52. WHO (World Health Organization). 2003. “Summary.” In Climate Change and Human Health: Risks and Responses. Geneva: WHO. ———. 2009. Protecting Health from Climate Change: Connecting Science, Policy, and People. Geneva: World Health Organization. WHO and PAHO (Pan American Health Organization). 2010. Protecting Health from Climate Change: Vulnerability and Adaptation Assessment. Geneva: World Health Organization.