<<

1 4

2 Defining the current situation - 3

4 Paul R Hunter, Helen Risebro

5 INTRODUCTION 6 The first step in any economic appraisal or evaluation is to understand the 7 underlying problem being addressed (see Chapter 1). Clearly such analysis of 8 drinking-water interventions will have a strong public health element. This 9 chapter introduces the reader to the role that epidemiology has in determining 10 the burden of in a community that may be attributable to lack of access 11 to safe drinking-water and/or adequate sanitation. In order to determine how 12 important such an issue may be one has to ask three questions: 13 What is the burden of disease in the target group? Burden of disease is a 14 function of the and severity of the disease. 15 What proportion of the burden of disease is due to deficiencies in access to 16 drinking-water to be remedied by the intervention? 17 Are there any spin-off livelihood responses that would result from the outcomes 18 of the intervention?

1 19 This chapter primarily concerns itself with answering the first two questions. 20 Specific data challenges related to livelihood analysis will be raised in the next 21 chapter. 22 23 The rest of this chapter is primarily concerned with assisting decision makers 24 in gathering evidence to enable them to make an informed decision whether or 25 not there is a public health need for an intervention, as a prerequisite to 26 undertaking a full economic assessment. The chapter introduces the reader to 27 some of the methods of epidemiology to better understand epidemiological 28 papers and reports. The chapter then goes on to describe how existing analyses 29 may be used to estimate disease burden

30 MEASURES OF DISEASE OCCURRENCE 31 The two predominant measures of disease occurrence are and 32 incidence.

33 Prevalence 34 Prevalence measures the amount of disease in a population at a given time 35 point and can be expressed as a percentage or shown as cases/population: 36 37 Number of existing cases in defined population at a given point in time 38 Number of people in the defined population at the same point in time 39 40 The point prevalence is a single assessment of a fixed point in time whereas 41 the period prevalence is the percentage of a population that are cases at any time 42 within a stated period. Period prevalence is preferred in infectious disease 43 epidemiology as it can be used when there are repeated or continual assessments 44 of the same individuals over a period of time (such as, multiple episodes of 45 diarrhoea). 46 Longitudinal prevalence can be calculated using the following formula (Morris, 47 1996): 48 49 Number of days with diarrhoea 50 Number of days under observation

51 Incidence 52 Incidence measures the number of new cases of disease in a population over a 53 specific time period. When the population is constant the incidence risk is 54 measured as:

2 55 56 Number of people who develop disease over a defined period of time 57 Number of disease-free people in that population at the start of the time period 58 59 When the population is not constant, for example, through deaths, migration, 60 births, or through additional participant recruitment, the incidence rate should 61 be calculated: 62 63 Number of new events in a defined population over a defined period of time 64 Total person-time at risk during the defined period of time 65 66 When studying illness that last a short time (days or a few weeks), such as 67 acute diarrhoea, then incidence would usually be the most appropriate measure. 68 For more protracted , such as the health effects of arsenic poisoning, 69 prevalence would be the more appropriate measure.

70 ESTIMATING DISEASE OCCURRENCE 71 There are different approaches to estimating disease occurrence in a population. 72 The choice of approach will depend on many different factors such as the 73 amount of resources available and the accuracy of result required. Whatever 74 approach is used, one of the most important starting points is to develop a case 75 definition.

76 Case definition 77 The case definition is essential for both the epidemiological studies and any 78 subsequent cost-benefit analyses. The case definition will enable the researcher 79 to know whether or not a particular health event should be included in the 80 analysis and will enable the cost-benefit analyst to determine the cost of the 81 disease outcome. A case definition may be based on symptoms such as the 82 presence of diarrhoea or clinical features of arsenic poisoning or the results of 83 laboratory investigations such as whether or not a stool sample is positive for 84 Cryptosporidium. For example, WHO defines diarrhoea as three or more loose 85 or fluid stools (which take the shape of a container) in a 24 hour period (WHO 86 1993). Case definitions may also include age ranges, geographical location or 87 dates of onset. 88 Whatever case definition is used, it should be clear and standardised to 89 minimise disease misclassification bias. Standardising case definitions becomes 90 especially important when using more than one field 91 researcher/interviewer/clinician or more than one community; this is because 92 definitions of diarrhoea can be culture- and person-specific. For example, a

3 93 study conducted in a rural municipality in Nicaragua in Central America 94 identified a classification encompassing nine different types of diarrhoea 95 (Davey-Smith et al. 1993). The classifications used in Nicaragua were 96 influenced by the place/person consulted for treatment. The source of any 97 existing health care utilisation data should therefore always be carefully 98 considered.

99 Primary surveys 100 Where prior information is not available or suspected to be unreliable it may be 101 most appropriate to collect data directly from the population concerned. Such 102 primary data is especially valuable for estimating the burden of disease for 103 illnesses that are unlikely to cause people to visit their local health care 104 provider. As such they are particularly valuable for self-reported diarrhoea. 105 They can also be especially valuable in poor or remote communities, with 106 limited access to health care. In these circumstances even people with severe 107 and chronic disease may not come to the attention of the health services. On the 108 other hand such population surveys are poor at identifying uncommon illnesses. 109 These surveys usually involve a questionnaire and this may solely be concerned 110 with determining whether the respondent reports various symptoms to enable a 111 diagnosis on whether or not they satisfy the case definition. Alternately, some 112 form of physical examination, or even laboratory or radiological examination 113 may be included. For example, stool samples may be collected in a study of 114 gastrointestinal disease. Examination of the teeth and radiological examination 115 of the skeleton may be necessary for exposure to fluoride at toxic levels. 116 117 There are two forms of the population survey; the cross-sectional survey and 118 the . 119 Cross-sectional studies are a relatively quick way of getting an estimate of 120 disease incidence or prevalence in a community. Cross-sectional studies look at 121 the disease status of all or a sample of a population at a particular moment in 122 time. In general each individual would be contacted only once. For diseases 123 with seasonal variation in their incidence the results of the survey would clearly 124 depend on what time of year the study was carried out. Conducting repeat 125 studies or lengthening the duration of the data collection period may ameliorate 126 the results. Cross-sectional studies can be conducted using various ways of 127 contacting the participants. The choice of approach will depend on resources 128 available, costs and existing communications. Researchers may contact people 129 by visiting them in their homes, or by post or telephone. Response rates are 130 generally poorest for postal surveys (typically around 50%), slightly better for 131 telephone surveys and best for direct visits to the home. Clearly, if examination 132 needs to be undertaken, the face to face contact is essential.

4 133 A cohort study follows a group of individuals over a period of time. During this 134 period, researchers monitor for the appearance of the disease outcome of 135 interest. Usually the initial contact includes an expanded baseline questionnaire. 136 There are several ways of recording the presence or absence of illness. Probably 137 the simplest method is the repeat visit of the researcher who may contact the 138 participant by phone, post or visit every one or more weeks and ask whether or 139 not there has been any illness since the previous visit. Increasingly email or 140 phone text messages are also being used. Sometimes people are asked to keep a 141 daily diary of symptoms which is then collected by the researcher. An example 142 of a pictorial diary used for recording frequency and consistency of stool is 143 shown in figure 1 (Wright et al. 2006). 144 Some of the advantages and disadvantages of cross-sectional and cohort studies 145 are listed in table 1. In general, cross-sectional surveys are quicker and less 146 costly but they can suffer from recall bias in that people may over-report very 147 recent or current diarrhoea (Boerma et al. 1991). Cohort studies are generally 148 more expensive and take longer, but they are better at detecting time 149 relationships to possible risk factors. Also cohort studies seem to suffer from a 150 fall-off in enthusiasm for reporting (respondent fatigue) which could lead to an 151 under-estimate of actual disease burden (Strickland et al. 1996; Verbrugge 152 1980). 153 154

5 155 Figure 1. Example of a symptom diary (Wright et al., 2006). 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175

176 Sampling 177 In the studies described so far it is unlikely that the analysts will have sufficient 178 resources to survey the entire population. What they will have to do is sample a 179 proportion of the actual population. Sampling is that part of statistical practice 180 concerned with the selection of individual observations intended to yield some 181 knowledge about a population of concern, especially for the purposes of 182 statistical inference. One of the problems with sampling is that it can lead to 183 bias. For example, if only people from the wealthier communities are sampled, 184 disease incidence may be underestimated. In order to avoid systematic sampling 185 errors, it is advisable to select a random sample where each sampling unit 186 (individual, household, family etc) should have an equal probability of selection. 187 As shown in table 2, there are various approaches to random sampling. 188

6 189 Table 1: Advantages and disadvantages of the cross-sectional and cohort design 190 for estimating disease occurrence (from Bowling and Ebrahim 2005). Study Advantages Disadvantages Design • Quick and relatively easy as no • Problem with direction of follow-up is required causality (reverse causality) • Can be used to investigate • In aetiological studies, survival many exposures and outcomes bias a potential problem as the • Useful for measuring the true sample is based on prevalent burden of disease in a rather than incident cases population • Not efficient for rare diseases • Recall bias • Not suitable for diseases of short duration

Cross-sectional • Prevalence estimates are potentially biased by low response rates • Migration in and out of population influences prevalence • Direct measurement of • Can be expensive and time- incidence of disease in exposed consuming and unexposed groups • Not useful for rare diseases • Time relationships between • Historical cohorts are very exposure and disease known dependent on quality of records • Reduced bias in exposure • Losses of follow-up can bias measurement findings () • Multiple outcomes can be • Outcome assessment can be studied influenced by knowledge of Cohort • Allows direct calculation of exposure (information bias) • Effects of rare exposures can be evaluated by careful selection of cohorts • Natural history of disease can be evaluated 191 192 The choice of sampling method will depend upon a number of factors, 193 including time, resources and study design. A combination of sampling 194 strategies may be employed. For example, in a study conducted in the Republic 195 of Congo, Mock et al. (1993) selected an equal number of villages from four 196 provinces (stratified cluster sampling). Subsequently, a systematic sample of 197 women with children <30 months was selected from each village. In Zimbabwe,

7 198 Waterkeyn and Cairncross (2005) took a representative systematic sample of 25 199 health clubs from each of two districts. Three hosts (the 10th, 20th and 30th 200 member) and 4 neighbours of each belonging to the health club were 201 selected from each health club register (cluster sample), resulting in 15 members 202 from each health club. 203 The sample chosen for analysis should be representative of the population 204 under study and large enough to make statistical inference. Where prevalence of 205 a condition is low, a larger sample is required. If a sample is too small, there 206 will be a lack of precision in results. Specific statistical methods exist for 207 calculating required sample size; an array of text books and journal articles 208 cover this in more detail (see: Woodward, 2005; Kirkwood and Sterne, 2006). If 209 possible, it is advisable to contact a statistician before selecting a sample and 210 designing the survey or investigation.

211 Utilizing existing local health data 212 A less costly and usually more rapid way of determining disease occurrence is 213 to use existing data collected by local health services. Ethical issues still usually 214 apply to collecting such data, especially when it is possible to identify which 215 individual relates to each report. The big problem with these datasets is that they 216 are based only on people who are ill enough or have sufficient resources to 217 present themselves to the health services. Surveillance may not even capture all 218 people who present to health care. For example if surveillance is based on 219 laboratory positive results then patients who do not have samples taken will not 220 be included in the analysis. This is the so-called reporting pyramid (Wheeler et 221 al. 1999; O’Brien 2007).

8 222 Table 2. Approaches to sample selection Sampling Description Method Simple random Each individual or unit is numbered and a sample of the sampling required size is selected using a random numbers table or a computer generated list of random numbers.

Systematic A system is in place to select a smaller sample from a random sampling larger population. Sampling may begin at a random point but all individuals or units thereafter follow sequence. If the list is ordered in a particular way, this can lead to serious bias. If the list is ordered in a random way, this is called Quasi-random sampling. Stratified random The inventory of people is divided into specific subgroups sampling (strata) of, for example, age and sex categories, and a random sample is drawn from each. This sampling method may be practiced when a specific subgroup is over- represented and an even distribution of people in strata is required. Cluster random The population is divided into sub-populations (clusters) sampling and a random sample of these clusters is selected (this differs from stratified sampling where all subgroups are selected). For example, to minimise travel costs and time, it may be appropriate to randomly select a number of villages from within a rural area. Multistage A random sample of subgroups is drawn from the target random sampling population; further random samples can be drawn from these subgroups. Sampling for A list of numbers in the study area is obtained or generated telephone (non-residential numbers should be excluded from the interviews sample) and a random is sample is subsequently selected either manually or by means of random digit dialling. A less resource intensive method is the use of computer generated lists 223 224 Examples of the types of health care data that may be available include: 225 • National morbidity and mortality data 226 • Consultation data 227 o Hospital episode and admissions data 228 o Emergency department visits (e.g. Heffernan et al. 2004) 229 o General practice or other health post/clinic consultation data 230 (Boussard et al. 1996)

9 231 o Telephone helplines (Rodman et al. 1998; Cooper et al. 232 2003)). 233 • Specialist communicable/infectious disease surveillance centres: 234 o International (e.g. WHO, Enter-net) 235 o National (e.g. CDSC in the UK, CDC in the US, KTL in 236 Finland, RIVM in the Netherlands etc.) 237 o Regional or local centres 238 • Proxy indicators of EID and EID outbreaks in the community: 239 o Anti-diarrhoeal pharmaceutical sales or prescriptions (Sacks et 240 al., 1986; Beaudeau et al. 1999; Edge et al. 2004) 241 o School/work absenteeism records.

242 Extrapolating from previous studies 243 It may well be the case that researchers have already conducted studies in the 244 community of interest. Alternately, it is likely that others have conducted studies 245 that have estimated disease occurrence rates in similar communities to the one 246 of interest. If these studies have been published in the peer-reviewed literature 247 they will be listed on one of the on-line searchable databases such as PubMed, 248 MEDLINE and ISI Web of Science. PubMed is available free of charge. 249 When using existing literature it is important to critically assess the quality of 250 the study and the extent to which can be extrapolated at least to the community 251 of interest. Recommendations on how to assess the quality of such studies are 252 published elsewhere (Blettner et al. 2001; Khan et al. 2001; Rushton, 2000; 253 Downs & Black, 1998). 254 A particularly valuable source of information on diarrhoeal disease incidence 255 in developing countries is provided by the Measure DHS (Demographic and 256 Health Surveys) Project http://www.measuredhs.com/. This is a USAID funded 257 project which has been collecting and analysing data from developing nations 258 since 1984. For many countries there may be reports already available that give 259 estimates of diarrhoeal incidence, albeit only in children under five years old.

260 Using existing global estimates 261 Probably the most common illness linked to inadequate water and sanitation is 262 diarrhoeal disease. There have been various attempts to estimate diarrhoeal 263 disease burden globally (Lopez et al. 2001; Bern 2004; Prüss and Havelaar 264 2001). Diarrhoeal disease is one of the most significant contributors to 265 preventable disease burden. It is estimated that diarrhoeal diseases are the forth 266 most common cause of death and second most important contributor to disease 267 burden globally (Prüss and Havelaar 2001). There are an estimated 3 million 268 deaths and almost 100 million DALYs lost (explained in Chapter 9) from 269 diarrhoeal disease per year. The burden of disease due to diarrhoeal illness falls 270 most heavily on the youngest and especially children under five years old.

10 271 Diarrhoeal disease also predominantly affects the poorest countries and the 272 poorest communities within those countries. Diarrhoea does not even feature in 273 the top 10 causes of disease burden in developed countries (Prüss and Havelaar 274 2001). Appendix 1 gives a table with estimated mortality and morbidity from 275 diarrhoeal disease.

276 ESTIMATING DISEASE ATTRIBUTABLE TO A SPECIFIC 277 ENVIRONMENTAL RISK 278 Once studies as described above have shown that a particular disease is an 279 important public health problem in the community, the next step is to identify 280 the important environmental risk factors that are driving the occurrence of that 281 disease. There are three approaches to estimating the contribution of a risk 282 factor; some form of quantitative risk assessment, one or more epidemiological 283 investigations, or the use of pre-existing global estimates. Furthermore, the DHS 284 data sources referred to above can be used to estimate disease burden 285 attributable to poor water and sanitation.

286 Risk assessment 287 Quantitative Microbial Risk Assessment (QMRA) has become a popular tool in 288 recent years especially in North America. QMRA uses existing data about the 289 and distribution of pathogenic microorganisms or indicator bacteria 290 to estimate risk to human health. The four stages of a QMRA are: 291 1. Hazard Identification 292 2. Exposure Assessment 293 3. Dose-Response Analysis 294 4. Risk Characterisation 295 An accessible review of QMRA methodology is given elsewhere (Hunter et 296 al. 2003). Howard et al. (2006) give a good example of its use in developing 297 country small rural settings. 298 QMRA has advantages over epidemiological studies where the disease 299 under investigation is uncommon; where the costs of an epidemiological study 300 would be too great or where serious time constraints apply. On the other hand 301 the input variables and especially the exposure values may not be known with 302 any degree of accuracy and so there may be large uncertainty around the results. 303 It is highly unlikely that the analyst will be able to generate dose-response 304 data de novo, as there will be plenty of studies already reported that have this 305 information. For some chemical agents such as arsenic there are reasonably well 306 defined relationships between concentration of arsenic in drinking water and 307 risk of disease (WHO, 2001). The principle problem with arsenic is that there 308 are different disease outcomes, such as skin lesions and cardiovascular disease 309 each of which have different dose-response curves. For microbiological dose

11 310 response curves these have had to be extrapolated, often from minimal data and 311 so carry with them significant uncertainty. Another issue with microbiological 312 dose-response curves is that they are usually -specific and so require 313 data on or estimations of the concentrations of multiple likely to be 314 present in the drinking-water options under investigation. 315 Contrary to dose-response data, exposure data will often be specific to the 316 community under investigation. However, national or local authorities such as 317 health agencies, administrative bodies or water and sewerage utilities often 318 undertake routine analyses of water quality in their jurisdictions. Even if routine 319 datasets are not available there may well be previous studies that have collected 320 water samples for analysis in the community of interest or in other similar 321 communities. 322 Primary data collection would include collection of data on basic microbial or 323 chemical analyses such as E. coli counts or arsenic concentrations. Primary data 324 collection for risk assessment would also include basic sanitary surveys. 325 It is unlikely that sampling for specific pathogens such as Cryptosporidium or 326 noroviruses would be available. The analyst is referred to the publications by 327 (Hunter et al. 2003) and Howard et al. (2006) for further information and 328 particularly for information on how to conduct the risk characterisation.

329 Epidemiological approaches 330 There are several epidemiological methods available to estimate the 331 contributions of specific risk factors or pathways to disease burden. 332 These include ecological studies, case-control studies, cohort studies and 333 prospective studies. In regard to waterborne disease, the advantages and 334 disadvantages of these different methods have been discussed in detail 335 elsewhere (Hunter et al. 2002). In addition to the book by Hunter et al. (2002) 336 there are many introductory text books on epidemiology and consequently this 337 discussion will not concentrate on describing such methods. 338 One of the easiest epidemiological approaches is the . In 339 ecological studies, the unit of observation and analysis is at the group, 340 population or community level rather than at the level of the individual. 341 Frequently the data used for observation and analysis is derived from existing 342 data sources (secondary data). The Geographical Information Systems (GIS) 343 described in box 1 can be very useful in compiling and analysing this type of 344 data. One of the simplest ecological approaches is to estimate disease incidence 345 in areas with high and low exposure to the potential risk factor such as poor 346 quality water. However, such studies may be susceptible to significant 347 confounding. For example, relatively wealthy populations often have access to 348 high quality water. In the case of a poor community with no sanitation, answers 349 to the question how much of the excess illness compared to a wealthy

12 350 community with well managed sanitation is due to sanitation and how much to 351 other factors linked to wealth will be full of confounding issues. 352 353 Box 1

A Geographic Information System (GIS) is a computer based graphics software program which enables the user to capture, store, manipulate, analyse and display spatially referenced data from a number of sources. Separate data sources based on geographically referenced information can be connected via relational databases and used to display any number of data attributes in the form of maps. The integrated information can be used in local and regional resource and environmental planning initiatives as well as spatial studies of infectious disease. For example, in Germany GIS has been applied to spatial patterns of diarrhoeal illness in relation to groundwater and surface water supplies (Dangendorf et al., 2002). In Nigeria, GIS was used to evaluate the health impact (diarrhoeal illnesses) of 39 separate water sources displayed in terms of layers, such as, hydrology, geology, and environmental pollution (Njemanze et al., 1999). Examples of GIS software include EDINA Digimap (for example, through which the UK Ordnance Survey Data Collection can be accessed) and ESRI ArcGIS 9.1 (http://www.esri.com). The University of Edinburgh hosts a site with a comprehensive index of GIS resources: http://www.geo.ed.ac.uk/home/giswww.html. It should be noted that when any type of data is gathered at diverse geographical areas it may be difficult to compile at the small area level without complex analysis and this can lead to problems of interpretation. 354 355 Cross-sectional studies have been discussed above for the collection of 356 disease incidence data. Such studies can also be useful for collecting data on 357 potential risk factors. 358 In case-control studies people with an illness are interviewed about past 359 exposure to possible risk factors. The same questions are also asked of controls 360 (people without the disease). Assumptions about the importance of particular 361 risk factors are then made, based on statistical analyses of the proportion of 362 cases and controls reporting exposure. Case control studies are valuable for 363 investigating multiple risk factors. However, they are usually around a single 364 disease and it can be difficult to use the output from these studies for estimating 365 disease burden attributable to a single transmission pathway 366 By comparison, cohort studies follow people with different levels of exposure 367 (say those with and without access to improved sanitation) and observe how 368 much disease develops in each group. The big advantage is that many different 369 disease outcomes can be observed. However, cohort studies are more costly than 370 case-control studies as people are followed over time.

13 371 An intervention may constitute new sanitation facilities, improved access to 372 water, a change in the water treatment process, or the introduction of 373 educational or behavioural programs. Interventions can be introduced in a 374 variety of study designs including cross-sectional, cohort and randomised 375 controlled trial (RCT). Intervention studies can be conducted under natural 376 conditions (i.e. accidental trials, such as, outbreaks), under uncontrolled 377 conditions (i.e. public measures, such as, the introduction of a new water 378 treatment plant), or under controlled conditions (i.e. clinical trials or field 379 studies) (Payment and Hunter 2003). 380 A particular form of intervention study is the Randomised Controlled Trial. 381 The RCT is one of the most robust epidemiological study designs and permits 382 simultaneous comparison of outcomes in a group of individuals. Study 383 participants are randomly assigned to one or more intervention groups expected 384 to influence disease status, or the control group which receives either the status 385 quo or a placebo (sham) intervention. A comprehensive review of randomised 386 and quasi-randomised controlled trials assessing diarrhoeal disease according to 387 water quality intervention can be found in Clasen et al. (2006).

388 Existing global and regional estimates 389 In the absence of local data, or data from similar situations, use can be made of 390 global estimates by determining what water and sanitation scenarios 391 communities fit into. This will then give rough estimates of the disease risk 392 attributable to inadequate water, sanitation and hygiene. There have been a 393 number of attempts to estimate the global burden of disease that may be 394 attributable to unimproved drinking water, sanitation and hygiene (Prüss et al. 395 2002). The authors estimated that globally these factors are responsible for 396 4.0% of all deaths and 5.7% of disease burden. The authors then went on to 397 define different scenarios of water and sanitation provision and estimated deaths 398 and attributable disease burden for each (Prüss et al. 2002). These scenarios are 399 presented in table 3. 400 401 Table 3: Exposure scenariosa Scenario Description Environmental Minimum and faecal-oral realistic relative pathogen load risk of diarrhoeal disease relative to scenario 1 VI Population not served with improved water Very high 6.1 / 11.0 supply and no improved sanitation in countries which are not extensively cover by those services (less than 98% coverage), and where water supply is not likely to be routinely controlled Vb* Population having access to improved water Very high 4.9 / 8.7

14 supply but not served with improved sanitation in countries which are not extensively covered by those services (less than 98% coverage), and where water supply is not likely to be routinely controlled Va* Population having access to improved High 3.8 / 6.9 sanitation but no improved water supply in countries where less than 98% of the population is served by water supply and sanitation services, and where water supply is likely not to be routinely controlled. IV Population having access to improved water High 3.8 / 6.9 supply and improved sanitation in countries where less than 98% of the population is served by water supply and sanitation services and where water supply is likely not to be routinely controlled. III Piped water in-house and improved High 2.5 / 4.5 sanitation services in countries where less than 98% of the population is served by water supply and sanitation services, and where water supply is likely not to be routinely controlled. II Population having access to piped water in- Medium to 2.5 / 2.5 house where more than 98% of the low population is served by those services; generally corresponds to regulated water supply and full sanitation coverage, with partial treatment of sewage and is typical in developed countries. I Ideal situation, corresponding to the Very low 1 / 1 absence of transmission of diarrhoeal disease through WSH 402 a Based on Prüss et al. 2002; Prüss-Üstün et al. 2004 403 * Transitions between exposure levels Va and Vb do not generally occur 404 405 Scenario I represents the minimum theoretical risk, namely no disease transmission 406 through unsafe WSH; scenario II is the situation typically encountered in developed 407 countries. These scenarios have very low to medium loads of faecal-oral pathogens in 408 the environment, characterized by more than 98% coverage of improved water supply 409 and sanitation. Scenarios III-VI are based in a high faecal-oral pathogen environment, 410 typical for developing countries with less advanced water and sanitation provision. 411 Scenario III represents piped water in-house and improved sanitation but does not occur 412 widely.

15 413 NON-COMMUNICABLE DISEASE 414 In contrast to disease of microbiological origin, burden of disease estimates due 415 to chemical contamination at a global level are less well developed. A wide 416 range of chemical and radiological contaminants of drinking water have been 417 implicated in human disease (Hunter 1997). Although toxicological assessments 418 do not have to deal with the problems of prior immunity, there are considerable 419 difficulties in assessing disease burden. There is still some considerable debate 420 over whether or not many chemicals in water actually cause disease such as 421 aluminium and dementia (Hunter 1997; Flaten 2001). Even where the 422 association is accepted the balance of acute disease and latent or chronic disease 423 may be far from clear. Recent considerations of burdens of disease due to 424 chemical contaminants of water have included nitrate (Fewtrell 2004), arsenic 425 (Fewtrell et al. 2005) and fluoride (Fewtrell et al. 2006). 426 Musculoskeletal disease makes a significant contribution to global disease 427 burden that causes substantial disability in both the developed and developing 428 world (Brooks 2006). How much of that burden of disease is due to carrying 429 water is unclear. There is uncertainty over the total disease burden and the 430 proportion due to carrying water. It is likely that the major health effect of 431 carrying water is low back pain and global estimates of disease burden due to 432 back pain are lacking (Brooks 2006). Also the authors are not aware of any good 433 studies directly linking the carrying of water and back pain. When such 434 information is lacking then the assessor may have to extrapolate from studies of 435 similar exposures. For example, the finding that children that carry backpacks to 436 school that are too heavy may suffer from increased musculoskeletal symptoms 437 (Moore et al. 2007). In the absence of any such study the assessor may have to 438 fall back on a process of soliciting expert opinion. 439 Non-communicable disease poses a number of challenges for conducting an 440 economic analysis. These include: 441 • As already discussed, existing epidemiological evidence is often 442 relatively poor compared to microbiological data. 443 • More so than for infectious diseases the actual disease burden can vary, 444 substantially, even between very similar communities situated relatively 445 close to each other due to marked differences in contaminant 446 concentrations. (Rahman et al. 2005) 447 • For some contaminants, drinking-water (either through consumption or 448 through cooking) is the only or predominant source of exposure (e.g. 449 arsenic, Fewtrell et al. 2005). For others, water is one on several possible 450 routes of exposure (e.g. lead, Romieu et al. 1994). 451 In these cases it may be difficult to determine the disease burden attributable 452 to water exposure. Reliance on national or regional estimates is also likely to be 453 problematic. These problems with assessing non-communicable disease burden

16 454 are probably responsible for their general exclusion from earlier economic 455 analyses. However, given that many of these illnesses would be chronic and 456 may be common in effected communities their contribution to disease burden is 457 likely to be substantial. It is likely that the exclusion of non-communicable 458 disease from analysis would, in certain communities, heavily understate the 459 benefits of water supply improvement. 460 This chapter has considered different approaches to estimating disease burden 461 within a community and then estimating what proportion of that disease may be 462 attributed to a specific environmental risk factor. Some of the approaches would 463 take considerable resources to implement, though usually the cost-benefit 464 analyst should be able to find previous studies that would sufficient for his/her 465 purposes. Global and regional estimates, though probably lacking precision for 466 any single country may still be sufficient for most purposes, unless there is 467 evidence of chronic disease of a non-communicable nature.

468 REFERENCES 469 Beaudeau P, Payment P, Bourderont D, Mansotte F, Boudhabay O, Laubiès B and 470 Verdière J (1999) A time series study of anti-diarrheal drug sales and tap-water 471 quality. Int. J. Environ. Health Res; 9(4):293-312. 472 Bern C (2004) Diarrhoeal diseases. In: Murray CJL, Lopez AD, Mathers CD (2004). 473 Global Burden of Disease and Injury Series Volume IV: The Global Epidemiology 474 of Infectious Diseases. WHO, Geneva. 475 Blettner M, Heuer C, Razum O (2001) Critical reading of epidemiological papers: a 476 guide. Eur J Pub Health;11:97-101. 477 Boerma JT, Black RE, Sommerfelt AE, Rutstein SO, Bicego GT (1991) Accuracy and 478 completeness of mothers' recall of diarrhoea occurrence in pre-school children in 479 demographic and health surveys. Int J Epidemiol; 20(4):1073-80. 480 Boussard E, Flahault A, Vibert JF & Valleron A J (1996) Sentiweb: French 481 communicable disease surveillance on the world wide web. BMJ; 313(7069):1381- 482 1382. 483 Bowling A, Ebrahim S (2005) Handbook of Health Research Methods: Investigation, 484 Measurement and Analysis. Open University Press, Maidenhead. 485 Brooks PM 2006) The burden of musculoskeletal disease - a global perspective. Clinical 486 Rheumatology 25: 778-781. 487 Clasen T, Roberts I, Rabie T, Schmidr W, Cairncross S (2006) Interventions to improve 488 water quality for preventing diarrhoea. Cochrane Database of Systematic Reviews; 489 Issue 3, Art. No.: CD004794. DOI: 10.1002/14651858.CD004794.pub2. 490 Cooper DL, Smith GE, O'Brien SJ, Hollyoak VA & Baker M (2003) What can Analysis 491 of Calls to NHS Direct Tell us about the Epidemiology of Gastrointestinal 492 in the Community? Journal of ; 46 (2): 101-105. 493 Dangendorf F, Herbst S, Reintjes R, Kistemann T (2002) Spatial patterns of diarrhoeal 494 illnesses with regard to water supply structures--a GIS analysis. Int J Hyg Environ 495 Health; 205(3):183-91. 496 Davey Smith G, Gorter A, Hoppenbrouwer J, Sweep RMP, Gonzalez C, Morales P, 497 Pauw J, Sandiford P (1993) The cultural construction of childhood diarrhoea in rural

17 498 Nicaragua: relevance for epidemiology and health promotion. Soc Sci Med; 499 36:1613-1624. 500 Downs SH, Black N (1998) The feasibility of creating a checklist for the assessment of 501 the methodological quality both of randomized and non-randomised studies of 502 health care interventions. J Epidemiol Community Health; 52:377-84. 503 Edge VL, Pollari F, Lim G, Aramini J, Sockett P, Martin SW, Wilson J & Ellis A (2004) 504 Syndromic surveillance of gastrointestinal illness using pharmacy over-the-counter 505 sales. A retrospective study of waterborne outbreaks in Saskatchewan and Ontario. 506 Can. J. Public Health; 95(6):446-450. 507 Fewtrell L (2004) Drinking-Water Nitrate, Methemoglobinemia, and Global Burden of 508 Disease: A Discussion. Environmental Health Perspectives 112, 1371–1374. 509 Fewtrell, L., Fuge, R. and Kay D. (2005) An estimation of the global burden of disease 510 due to skin lesions caused by arsenic in drinking water. Journal of Water and 511 Health 3, 101-107. 512 Fewtrell, L., Smith, S., Kay D. and Bartram, J (2006) An attempt to estimate the global 513 burden of disease due to fluoride in drinking water. Journal of Water and Health 514 4(4), 533-542. 515 Flaten TP (2001) Aluminium as a risk factor in Alzheimer's disease, with emphasis on 516 drinking water. Brain Research Bulletin 55, 187-196. 517 Heffernan R, Mostashari F, Das D, Karpati A, Kulldorff M & Weiss D (2004) Syndromic 518 Surveillance in Public Health Practice, New York City. Emerging Infectious 519 Diseases; 10(5):858-864. 520 Howard G, Pedley S, Tibatemwa S (2004) Quantitative Microbial Risk Assessment to 521 estimate health risks attributable to water supply: Can the technique be applied in 522 developing countries with limited data? Journal of Water and Health; 04.1:49-65. 523 Hunter, P.R. (1997) Waterborne disease: Epidemiology and Ecology. Wiley, Chichester, 524 UK. 525 Hunter PR, Payment P, Ashbolt N & Bartram J (2003) Chapter 3: Assessment of Risk. 526 Assessing Microbial Safety of Drinking Water: Improving Approaches and 527 Methods. OECD/WHO guidance document. In: Ronchi E & Bartram J (eds.), pp. 528 79-109. ISBN: 9264099468. OECD/WHO, Paris. 529 Hunter PR, Waite M, Ronchi E (eds.) (2002) Drinking Water and Infectious Disease: 530 Establishing the Link. CRC Press, Bocata Raton. 531 Khan KS, ter Reit G, Popay J, Nixon J, Kleijnen J (2001) Stage II, Phase 5: Study 532 Quality Assessment. In: CRD Report 4: Undertaking systematic reviews of research 533 on effectiveness. Available at (accessed 15/03/07): 534 http://www.york.ac.uk/inst/crd/pdf/crd4_ph5.pdf 535 Kirkwood BR, Sterne JA (2006) Essential Medical Statistics (2nd Edition). Blackwell 536 Science Ltd, Oxford. 537 Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJL (2001). Global and regional 538 burden of disease and risk factors, 2001: systematic analysis of population health 539 data. Lancet, 367 (9524), 1747-1757. 540 Mock NB, Sellers TA, Abdoh AA, Franklin RR (1993). Socioeconomic, environmental, 541 demographic and behavioral factors associated with occurrence of diarrhea in young 542 children in the Republic of Congo. Soc Sci Med; 36(6):807-16. 543 Moore, M.J., White, G.L. and Moore, D.L. (2007) Association of relative backpack 544 weight with reported pain, pain sites, medical utilization, and lost school time in 545 children and adolescents. Journal of School Health 77, 232 – 239.

18 546 Morris SS, Cousens SN, Kirkwood BR, Arthur P, Ross DA (1996) Is prevalence of 547 diarrhoea a better predictor of subsequent mortality and weight gain than diarrhoea 548 incidence? American Journal of Epidemiology; 144(6):582-8. 549 Njemanze PC, Anozie J, Ihenacho JO, Russell MJ, Uwaeziozi AB (1999) Application of 550 risk analysis and geographic information system technologies to the prevention of 551 diarrheal diseases in Nigeria. Am J Trop Med Hyg; 61(3):356-60. 552 O'Brien SJ & Halder SLS (2007) GI Epidemiology: infection epidemiology and acute 553 gastrointestinal infections. Alimentary Pharmacology & Therapeutics; 25 (6):669- 554 674. 555 Payment P, Hunter PR (2003) Intervention Studies. In: Hunter PR, Waite M, Ronchi E 556 (2003) Drinking Water and Infectious Disease: Establishing the Links. CRC Press & 557 IWA Publishing, London. 558 Prüss, A. and Havelaar, A. (2001) The Global Burden of Disease study and applications 559 in water, sanitation and hygiene. In: Fewtrell L, Bartram J (eds) Water Quality: 560 Guidelines, Standards and Health. Risk assessment and management for water- 561 related infectious disease. IWA Publishing, London 2001, pp 43 – 59. 562 Prüss A, Kay D, Fewtrell L, Bartram J (2002) Estimating the burden of disease from 563 water, sanitation and hygiene at a global level. Environmental Health Perspectives; 564 110(5):538-542. 565 Rahman MM, Sengupta MK, Ahamed S, Chowdhury UK, Lodh D, Hossain MA, Das B, 566 Saha KC, Kaies I, Barua AK, Chakraborti D. (2005) Status of groundwater arsenic 567 contamination and human suffering in a Gram Panchayet (cluster of villages) in 568 Murshidabad, one of the nine arsenic affected districts in West Bengal, India. 569 Journal of Water and Health 3: 283-96. 570 Rodman J, Frost F & Jabukowski W (1998) Using nurse hotline calls for disease 571 surveillance. Emerg. Infect. Dis; 4(2):329-32. 572 Romieu I, Palazuelos E, Hernandez Avila M, Rios C, Munoz I, Jimenez C, Cahero G. 573 (1994) Sources of lead exposure in Mexico City. Environmental Health 574 Perspectives 102: 384-9. 575 Rushton L (2000) Reporting of occupational and environmental research: use and misuse 576 of statistical and epidemiological methods. Occup Environ Med; 57:1-9. 577 Sacks JJ, Lieb S, Baldy LM, Berta S, Patton CM, White MC, Bigler WJ & Witte JJ 578 (1986) Epidemic Campylobacteriosis associated with a community water supply. 579 Am. J. Public Health; 76(4): 424-428. 580 Strickland MJ, Crawford JMac, Shen L, Willkins JR III (2006) Time-dependent record 581 keeping fatigue among youth completing health diaries of unintentional injuries. 582 Journal of Safety Research; 37: 487–492. 583 Verbrugge LM (1980) Health Diaries. Medical Care; 18(1):73-95. 584 Waterkeyn & Cairncross (2005) Creating demand for sanitation and hygiene through 585 Community Health Clubs: a cost-effective intervention in two districts in 586 Zimbabwe. Soc Sci Med; 61(9):1958-70. 587 Wheeler JG, Sethi D, Cowden JM, Wall PG, Rodrigues LC, Tompkins DS, Hudson MJ, 588 Roderick PJ (1999) Study of infectious intestinal disease in England: rates in the 589 community, presenting to general practice, and reported to national surveillance. 590 The Infectious Intestinal Disease Study Executive. BMJ; 318(7190):1046-50. 591 Woodward M (2005) Epidemiology: Study design and data analysis. Second Edition. 592 Chapman & Hall/CRC, London. 593 WHO (1993) The management and prevention of diarrhoea: practical guidance. 3rd 594 edition. World Health Organisation, Geneva.

19 595 WHO (2001) Arsenic and arsenic compounds. Environmental Health Criteria 224 596 (Second edition). Geneva, World Health Organization. Available at (accessed 597 15/03/07): http://whqlibdoc.who.int/ehc/WHO_EHC_224.pdf 598 Wright JA, Gundry SW, Conroy R, Wood D, Du Preez M, Ferro-Luzzi A, Genthe B, 599 Kirimi M, Moyo S, Mutisi C, Ndamba J, Potgieter N (2006) Defining episodes of 600 Diarrhoea: Results from a Three-country Study in Sub-Saharan Africa. J Health 601 Popul Nutr; 24(1):8-16. 602

20 Appendix 1. Estimated mortality and morbidity from diarrhoeal disease.

Number Estimated Age- Estimated Estimated rural with Number total Estimated standardized total DALYs Age- Total unimproved rural deaths deaths per death rates DALYs per standardized Rural water with no ('000), by 100,000 per 100,000 ('000), by 100,000 DALYs per Population population supply tap Country cause population by cause cause population 100,000 total ('000) ('000) ('000) ('000) Afghanistan 41.16 179.51 135.69 1343.11 5857.41 3157.16 28574 21716 14984 21716 Albania 0.01 0.17 0.23 1.15 36.72 36.33 3112 1743 105 924 Algeria 8.14 26.05 24.26 295.41 944.82 803.00 32358 13267 2653 5572 Andorra 0.00 0.89 0.41 0.02 26.47 31.42 67 6 0 Angola 48.77 369.90 233.97 1608.10 12197.30 5832.30 15490 9914 5948 9814 Antigua and Barbuda 0.00 0.60 0.57 0.10 131.56 124.71 81 50 6 11 Argentina 0.43 1.13 1.03 49.46 130.23 126.38 38372 3837 767 2110 Armenia 0.06 1.90 3.51 3.52 114.48 167.91 3026 1089 218 370 Australia 0.03 0.17 0.11 5.18 26.53 29.88 19942 1595 0 Austria 0.01 0.06 0.02 1.88 23.22 29.27 8171 2778 0 0 Azerbaijan 0.92 11.08 14.03 36.54 440.47 461.75 8355 4178 1713 3384 Bahamas 0.00 1.23 1.49 0.40 130.09 119.78 319 32 4 6 Bahrain 0.01 1.47 5.64 0.61 85.77 88.56 716 72 Bangladesh 68.19 47.41 41.25 2297.81 1597.83 1095.60 139215 104411 29235 104411 Barbados 0.00 1.33 0.95 0.26 97.43 110.04 269 129 0 Belarus 0.02 0.24 0.41 3.09 31.10 45.95 9811 2845 0 2134 Belgium 0.09 0.88 0.36 2.73 26.50 30.79 10400 312 Belize 0.03 11.41 8.70 1.27 505.77 368.41 264 137 25 51 Benin 7.34 111.95 84.24 241.10 3676.15 2016.08 8177 4497 1934 4407

21 Bhutan 1.32 60.39 43.47 43.99 2008.46 1220.03 2116 1926 770 Bolivia 4.17 48.21 38.34 142.28 1645.74 1105.12 9009 3243 1038 1816 Bosnia and Herzegovina 0.01 0.17 0.25 1.29 31.30 40.95 3909 2150 86 494 Botswana 0.36 20.51 21.64 13.30 751.31 522.96 1769 849 85 611 Brazil 17.33 9.83 9.89 735.40 417.23 393.97 183913 29426 12653 24424 Brunei Darussalam 0.00 0.45 0.38 0.32 91.79 79.13 366 84 Bulgaria 0.03 0.33 0.69 2.65 33.30 54.47 7780 2334 70 654 Burkina Faso 21.79 172.64 122.04 717.31 5681.93 2801.47 12822 10514 4836 10514 Burundi 9.29 140.76 95.40 305.80 4632.19 2497.33 7282 6554 1507 6488 Cambodia 11.84 85.74 71.54 386.82 2801.10 1726.07 13798 11176 7265 10953 Cameroon 14.38 91.41 66.46 480.20 3052.94 1864.37 16038 7698 4311 7544 Canada 0.26 0.83 0.46 8.68 27.77 31.77 31958 6072 61 3765 Cape Verde 0.10 22.46 18.13 3.77 830.90 586.62 495 213 57 204 Central African Republic 4.69 122.75 87.61 151.19 3958.76 2310.77 3986 2272 886 2272 Chad 11.38 136.37 96.10 375.06 4493.03 2256.30 9448 7086 4039 6944 Chile 0.25 1.62 1.71 19.10 122.36 117.80 16124 2096 880 1300 China 108.40 8.32 10.06 5055.36 388.19 441.57 1307989 784793 258982 337461 Colombia 2.66 6.10 5.59 133.79 307.37 259.73 44915 10330 2996 5062 Comoros 0.39 51.98 43.31 13.26 1776.08 1047.90 777 497 90 477 Congo 1.57 43.20 39.24 52.06 1433.11 800.35 3883 1786 1304 1715 Cook Islands 0.00 5.45 4.63 0.05 249.06 195.69 18 5 1 Costa Rica 0.12 2.92 3.63 5.79 141.44 130.00 4253 1659 133 315 Côte d'Ivoire 16.67 101.84 87.84 539.23 3295.03 2065.51 4540 1861 0 651 Croatia 0.01 0.20 0.17 1.16 26.14 32.12 11245 2699 594 1376 Cuba 0.23 2.03 1.66 12.61 111.88 123.41 826 256 0 0 Cyprus 0.00 0.53 0.72 0.62 78.41 92.41 10229 2660 0 239 Czech Republic 0.01 0.06 0.04 2.36 23.08 29.33 17872 9830 2556 9338 Democratic People's Republic of Korea 3.21 14.23 16.47 129.33 573.74 602.45 22384 8730 0 2532 Democratic Republic of the Congo 112.21 219.15 148.16 3669.63 7167.11 3557.20 55853 37980 26966 37600 Denmark 0.07 1.39 0.61 1.58 29.61 32.38 5414 812 0 0 Djibouti 0.84 120.66 93.21 27.87 4023.51 2252.51 779 125 51 118

22 Dominica 0.00 1.78 1.64 0.11 146.00 137.45 79 22 2 9 Dominican Republic 1.25 14.50 13.20 49.23 571.36 473.51 8768 3507 316 1333 Ecuador 1.97 15.39 13.48 77.46 604.67 493.75 13040 4955 545 2725 Egypt 13.13 18.63 16.62 464.12 658.26 503.96 72642 42132 1264 10954 El Salvador 0.96 14.92 12.16 37.93 591.31 446.31 6762 2705 811 1677 Equatorial Guinea 0.46 94.88 64.10 14.87 3089.71 1686.11 492 251 146 251 Eritrea 2.58 64.75 59.15 87.03 2180.72 1216.58 4232 3386 1456 3386 Estonia 0.00 0.04 0.08 0.34 25.75 35.69 1335 401 4 108 Ethiopia 63.16 91.58 66.07 2099.77 3044.86 1634.76 75600 63504 56519 63504 Fiji 0.06 6.65 10.72 1.67 200.65 181.35 841 404 198 375 Finland 0.03 0.59 0.29 1.35 26.03 30.78 5235 2042 0 163 France 0.55 0.93 0.46 18.10 30.24 33.91 60257 14462 0 0 Gabon 0.41 31.70 23.30 14.72 1126.65 732.57 1362 204 108 188 Gambia 0.94 67.72 54.65 31.11 2240.50 1350.47 1478 1094 252 1061 Georgia 0.04 0.68 0.65 1.95 37.75 47.14 4518 2169 716 1561 Germany 0.27 0.33 0.17 20.38 24.73 30.99 82645 9917 0 298 Ghana 9.69 47.32 41.25 329.34 1608.77 1062.71 21664 11699 4211 11231 Greece 0.00 0.00 0.00 2.50 22.80 29.18 11098 4328 Grenada 0.00 5.09 5.07 0.15 184.50 174.39 102 60 4 15 Guatemala 3.39 28.14 20.79 122.89 1021.04 620.09 12295 6516 521 2281 Guinea 9.48 113.42 90.44 312.12 3733.96 2054.15 9202 5889 3828 5830 Guinea-Bissau 1.63 112.24 82.35 52.91 3652.10 1824.93 1540 1001 511 1001 Guyana 0.31 40.12 43.76 8.82 1155.11 1020.94 750 465 79 256 Haiti 5.56 67.63 51.40 185.46 2256.80 1529.70 8407 5212 2293 5056 Honduras 1.67 24.69 19.16 61.58 908.17 597.36 7048 3806 723 1446 Hungary 0.02 0.19 0.15 2.70 27.26 35.20 10124 3442 69 310 Iceland 0.00 0.39 0.21 0.08 27.55 29.45 292 20 0 0 India 456.36 43.48 38.89 15253.84 1453.37 1152.88 1087124 782729 133064 720111 Indonesia 35.50 16.35 16.68 1264.05 582.16 522.81 220077 116641 36159 109642 Iran (Islamic Republic of) 8.68 12.75 14.79 321.72 472.63 473.40 68803 22705 3633 Iraq 14.06 57.35 47.22 469.50 1915.53 1156.59 28057 9259 4629 6203 Ireland 0.01 0.14 0.08 1.05 26.85 29.24 4080 1632 65 Israel 0.06 0.88 0.66 2.25 35.62 33.07 6601 528 0 11

23 Italy 0.04 0.07 0.03 13.07 22.74 29.34 58033 19151 766 Jamaica 0.13 4.91 4.60 7.04 268.10 240.42 2639 1267 152 684 Japan 1.30 1.02 0.52 36.65 28.75 35.11 127923 43494 0 3914 Jordan 0.69 13.02 10.86 26.47 496.70 344.02 5561 1168 105 222 Kazakhstan 0.35 2.29 2.58 15.62 101.00 114.85 14839 6529 1763 4766 Kenya 24.62 78.06 62.21 817.60 2592.22 1661.56 33467 19746 10663 17376 Kiribati 0.03 33.26 28.90 0.97 1120.97 827.41 97 49 23 39 Kuwait 0.01 0.31 0.41 2.07 84.91 78.83 2606 104 Kyrgyzstan 0.66 13.04 11.62 25.94 511.97 440.93 5204 3435 1168 2507 Lao People's Democratic Republic 5.44 98.48 73.21 175.89 3181.17 1913.67 5792 4576 2608 4301 Latvia 0.01 0.35 0.36 0.68 29.07 41.12 2318 788 32 323 Lebanon 0.27 7.56 7.95 11.17 310.65 295.46 3540 425 0 64 Lesotho 1.51 83.87 67.65 49.24 2735.46 1864.97 1798 1474 354 1356 Liberia 4.87 150.38 97.18 159.04 4909.56 2532.50 3241 1718 825 1718 Libyan Arab Jamahiriya 0.44 8.17 7.57 18.18 333.86 281.98 5740 746 Lithuania 0.00 0.12 0.07 0.91 26.28 32.92 3443 1136 500 Luxembourg 0.00 0.84 0.54 0.14 32.16 35.32 459 37 0 1 Madagascar 18.90 111.75 76.11 621.55 3674.25 1993.93 18113 13222 8595 12958 Malawi 19.51 164.32 105.07 631.65 5320.89 2696.91 12608 10465 3349 10255 Malaysia 0.31 1.29 1.31 27.93 116.53 98.23 24894 8962 358 1165 Maldives 0.05 16.57 14.96 1.90 615.41 379.79 321 228 55 228 Mali 22.74 180.19 126.88 750.48 5945.43 2962.38 13124 8793 5628 8617 Malta 0.00 0.00 0.00 0.10 24.80 29.14 400 32 0 1 Marshall Islands 0.01 21.02 18.20 0.39 738.85 548.15 60 20 1 Mauritania 3.61 128.48 86.82 119.66 4262.68 2330.64 2980 1103 617 959 Mauritius 0.01 1.13 1.53 1.44 118.75 120.66 1233 690 0 0 Mexico 5.57 5.46 6.08 253.51 248.63 214.56 105699 25368 3298 7103 Micronesia (Federated States of) 0.04 33.84 28.29 1.23 1139.92 788.02 110 77 5 Monaco 0.00 1.35 0.82 0.01 37.36 42.77 35 0 Mongolia 0.86 33.77 32.75 30.56 1194.13 1047.87 2614 1124 787 1113 Morocco 6.59 21.92 20.84 231.99 771.46 644.88 31020 13028 5732 10814

24 Mozambique 30.13 162.54 121.84 979.93 5286.27 2978.92 19424 12237 9055 11992 Myanmar 24.93 51.02 47.60 816.83 1672.03 1376.36 50004 35003 8051 34303 Namibia 0.85 43.19 32.40 29.19 1488.53 903.48 2009 1346 256 902 Nauru 0.00 7.83 6.80 0.04 324.52 245.32 13 0 Nepal 16.73 67.97 54.04 551.57 2241.36 1405.76 26591 22602 2486 20116 Netherlands 0.03 0.20 0.10 4.13 25.68 29.64 16226 5517 0 0 New Zealand 0.01 0.27 0.22 1.15 29.94 32.69 3989 558 Nicaragua 1.40 26.25 20.51 51.22 960.01 604.87 5376 2258 835 1648 Niger 25.32 219.31 160.02 832.39 7210.52 3381.54 13499 10394 6652 10394 Nigeria 134.46 111.21 84.72 4381.77 3623.95 2035.29 128709 66929 46181 64921 Niue 0.00 0.00 0.00 0.00 79.39 67.03 1 0 0 0 Norway 0.09 1.98 0.77 1.44 31.83 33.27 4598 920 0 0 Oman 0.17 6.23 5.55 7.96 287.51 197.99 2534 557 Pakistan 118.40 78.98 57.66 3904.13 2604.30 1570.41 154794 102164 11238 86839 Palau 0.00 4.98 4.33 0.05 235.82 180.45 20 6 0 5 Panama 0.21 6.71 5.89 10.02 327.05 275.27 3175 1365 287 382 Papua New Guinea 2.40 42.94 40.91 80.28 1437.04 909.15 5772 5022 3415 4821 Paraguay 0.86 15.00 11.82 34.53 601.64 415.00 6017 2527 809 1895 Peru 4.40 16.45 14.19 170.76 637.95 511.09 27562 7166 2508 4371 Philippines 12.23 15.56 16.77 411.04 523.08 422.48 81617 31014 5583 23881 Poland 0.03 0.09 0.08 10.37 26.84 33.97 38559 14652 586 Portugal 0.01 0.12 0.10 2.63 26.14 31.70 10441 4698 Qatar 0.01 1.00 1.38 0.63 105.22 99.35 777 62 0 Republic of Korea 0.26 0.54 0.71 62.75 132.29 130.75 47645 9053 2625 5522 Republic of Moldova 0.03 0.76 1.14 2.20 51.57 70.52 4218 2278 273 2073 Romania 0.15 0.66 0.95 10.41 46.51 67.91 21790 9806 8237 8531 Russian Federation 0.87 0.60 0.76 55.26 38.35 54.77 143899 38853 4662 18649 Rwanda 13.27 160.39 121.00 435.51 5264.60 2781.61 8882 7106 2203 7035 Saint Kitts and Nevis 0.00 8.65 9.03 0.11 263.73 262.62 42 29 0 8 Saint Lucia 0.01 4.31 5.15 0.30 202.09 196.63 159 110 2 27 Saint Vincent and the Grenadines 0.00 0.74 0.78 0.16 135.03 122.32 118 48 3 13 Samoa 0.02 9.48 7.39 0.68 382.94 261.73 184 144 19 69

25 San Marino 0.00 0.00 0.00 0.01 22.44 29.19 28 3 Sao Tome and Principe 0.06 36.02 25.74 1.98 1265.66 778.23 153 95 26 77 Saudi Arabia 1.68 7.16 6.10 74.49 316.70 211.06 23950 2874 Senegal 7.35 74.56 64.52 247.92 2515.77 1516.13 11386 5693 2277 4725 Serbia and Montenegro 0.15 1.40 2.01 7.54 71.53 97.04 10510 5045 706 1816 Seychelles 0.00 2.00 2.24 0.10 118.61 114.74 80 40 10 10 Sierra Leone 12.90 270.82 170.53 417.93 8771.91 4599.53 5336 3202 1729 3170 Singapore 0.02 0.49 0.54 2.61 62.41 71.00 4273 0 Slovakia 0.01 0.11 0.11 1.48 27.40 34.21 5401 2268 23 159 Slovenia 0.01 0.26 0.15 0.47 23.78 29.66 1967 964 Solomon Islands 0.22 47.68 42.76 7.31 1577.19 947.46 466 387 135 383 Somalia 15.56 164.18 112.61 509.59 5375.36 2565.43 7964 5177 3779 5177 South Africa 13.59 30.37 29.07 478.58 1069.23 898.38 47208 20299 5481 13804 Spain 0.28 0.68 0.32 10.60 25.86 31.23 42646 9809 0 98 Sri Lanka 0.68 3.59 4.18 32.59 172.34 185.02 20570 16250 4225 15600 Sudan 19.45 59.16 46.57 655.02 1992.25 1259.44 35523 21314 7673 18543 Suriname 0.04 9.43 8.57 1.78 412.25 345.98 446 103 28 53 Swaziland 0.61 57.36 44.72 20.72 1937.46 1183.83 1034 786 361 676 Sweden 0.05 0.60 0.21 2.22 24.99 30.20 9008 1531 0 0 Switzerland 0.03 0.37 0.16 1.75 24.40 30.03 7240 2317 0 23 Syrian Arab Republic 2.05 11.81 10.63 78.34 450.70 324.35 18582 9291 1208 2601 Tajikistan 2.03 32.80 28.99 68.33 1103.02 871.72 6430 4887 2541 3909 Thailand 4.64 7.47 8.42 184.97 297.42 307.40 63694 43312 0 36382 The former Yugoslav Republic of Macedonia 0.03 1.71 2.21 1.62 79.29 93.32 2030 812 Timor-Leste 0.14 19.08 18.58 4.55 616.21 518.70 887 816 359 726 Togo 2.96 61.72 50.39 96.75 2015.21 1189.41 5988 3832 2453 3832 Tonga 0.01 8.78 6.82 0.36 355.74 258.82 102 68 0 16 Trinidad and Tobago 0.02 1.28 1.43 1.44 110.67 116.37 1301 312 37 103 Tunisia 0.66 6.75 7.58 27.49 282.59 294.04 9995 3598 648 2231 Turkey 6.94 9.87 10.86 235.89 335.47 296.89 72220 23833 1668 4052

26 Turkmenistan 1.10 22.88 20.97 38.91 811.64 708.23 4766 2574 1184 1827 Tuvalu 0.00 28.66 24.37 0.10 971.75 744.16 10 4 0 Uganda 30.22 120.86 86.24 988.55 3953.64 1907.56 27821 24482 10772 24482 Ukraine 0.16 0.32 0.51 16.16 33.04 48.14 46989 15506 1396 8063 United Arab Emirates 0.02 0.75 1.09 2.78 94.65 95.13 4284 643 0 193 United Kingdom 0.64 1.08 0.52 17.55 29.72 34.21 59479 6543 0 131 United Republic of Tanzania 31.89 87.91 74.31 1058.77 2918.63 1652.12 37627 24081 12281 23359 United States of America 1.49 0.51 0.30 84.23 28.94 30.94 295410 59082 0 0 Uruguay 0.09 2.80 1.96 4.49 132.48 134.58 3439 241 0 39 Uzbekistan 0.54 2.12 2.01 37.74 146.80 126.64 26209 16774 4193 12580 Vanuatu 0.03 13.77 11.02 1.06 513.71 336.92 207 159 77 115 Venezuela (Bolivarian Republic of) 1.58 6.28 6.99 71.44 283.19 239.56 26282 3154 946 1230 Viet Nam 10.75 13.39 13.82 395.50 492.66 461.44 83123 61511 12302 57820 Yemen 19.06 98.69 73.18 623.05 3225.67 1634.92 20329 15043 5265 13539 Zambia 15.50 144.88 110.58 501.27 4685.65 2522.23 11479 7347 4408 7200 Zimbabwe 6.18 48.17 41.99 204.68 1594.65 1039.85 12936 8408 2354 7988

27