Uncertainty and data availability for the global burden of disease estimates 2000-2002

Colin D. Mathers

Evidence and Information for Policy Working Paper

World Health Organization Revised November 2005

Executive summary
During the years 1998 to 2004, the World Health Organization (WHO) undertook a progressive reassessment of the GBD for the years 2000 to 2002, with consecutive revisions and updates published annually in WHO's World Health Reports. These updates drew on a wide range of data sources to develop internally consistent estimates of incidence, health state prevalence, severity and duration, and mortality for over 130 major causes, for 17 sub-regions of the world. WHO program participation in the development and finalization of these estimates ensured that estimates reflected all information and knowledge available to WHO at the time, and to the extent possible. This working paper provides an overview of data availability, an examination of levels of uncertainty for GBD estimates for 2002, and a summary of the levels of evidence for country-level estimates of burden of disease. While it is difficult to quantify the exact numbers of data sources used for the YLD estimates for GBD 2000-2002, an attempt has been made to provide an approximate count by region. Of the more than 8,000 datasets estimated to have been used for the GBD 2000-2002 estimation of YLD, nearly 6,600 relate to Group I causes, and only 18 datasets to Group III causes (injuries). Furthermore, one quarter of the datasets relate to populations in sub-Saharan Africa, and around one-fifth to populations in high income countries. Not counting again studies which also contributed to the estimation of cause-specific mortality rates, an additional 1,370 datasets were used for the estimation of YLL. In total, the GBD 2000-2002 has drawn on in excess of 10,000 datasets or studies. The GBD 2000-2002 analyses have been used to prepare estimates of mortality and burden of disease for WHO Member States for the year 2002. Mortality estimates were based on analysis of latest available national information on levels of mortality and cause distributions as at mid-2004. YLD estimates were based on the GBD analyses of incidence, prevalence, duration and severity of conditions for the relevant epidemiological subregion, together with national and subnational level information available to WHO. Country-level estimates have been classified broadly into three classes according to the levels of evidence available. Uncertainty in the GBD prior estimates for countries must be taken into account when making crossnational comparisons, and needs to be carefully communicated and interpreted by epidemiologists and policy makers alike. Estimates of mortality in countries where there is no functioning vital registration system for causes of death will always be substantially more uncertain than those derived from systems where all deaths are registered and medically certified. The same may be said for the quantification of disability due to various conditions, where the gap in data availability between rich and poor countries is likely to be even more extreme than for mortality. This working paper summarizes provisional attempts to quantify the levels of uncertainty relating to all-cause and cause-specific estimates of mortality, and provides estimates of uncertainty ranges for total deaths by sex for WHO Member States in 2002. Relative uncertainty ranges varied from less than 1 per cent in countries with good death registration systems up to plus or minus 25 per cent in countries with limited information on mortality.

1. Introduction
In 1993 the Harvard School of Public Health in collaboration with WHO and the World Bank assessed the Global Burden of Disease (GBD). As well as generating comprehensive and consistent set of estimates of mortality and morbidity by age, sex and region for the world for the first time (Murray and Lopez 1996a; Murray and Lopez 1996b; Murray and Lopez 1996c; World Bank 1993), the GBD study also introduced a new metric the disability adjusted life year (DALY) to quantify the burden of disease. During the years 1998 to 2004, the World Health Organization (WHO) undertook a progressive reassessment of the GBD for the years 2000 to 2002, with consecutive revisions and updates published annually in WHO's World Health Reports. These updates drew on a wide range of data sources to develop internally consistent estimates of incidence, health state prevalence, severity and duration, and mortality for over 130 major causes, for 17 sub-regions of the world (Mathers et al. 2003a). WHO program participation in the development and finalization of these estimates ensured that estimates reflected all information and knowledge available to WHO at the time, and to the extent possible. The original GBD study analysed and synthesized a large volume of data on population health to produce comprehensive and comparable information on the causes of loss of health globally, regionally and particularly for low and middle income countries, where there are considerable limitations in data availability and comparability. The GBD study made estimates even for causes of burden where there was limited data and considerable uncertainty, to ensure that causes with limited information were not implicitly considered to have zero burden and hence ignored by health policy makers (Murray, Mathers, and Salomon J.A. 2003). The basic philosophy guiding the GBD approach is that there is likely to be information content in almost all sources of health data, provided they are carefully screened for plausibility and completeness; and that internally consistent estimates of the global descriptive epidemiology of major conditions are possible with appropriate tools, investigator commitment and expert opinion. This philosophy remains central to the GBD updates for 2000 to 2002, which have expanded the framework of the original GBD study in order 1. to quantify the burden of premature mortality and disability by age, sex, and region for 135 major causes or groups of causes; 2. to develop internally consistent estimates of the incidence, prevalence, duration, and casefatality for over 500 sequelae resulting from the above causes; 3. to analyze the contribution to this burden of major physiological, behavioural, and social risk factors by age, sex and region. Methods and data sources are discussed in detail in other working papers (Mathers et al. 2003a; Salomon et al. 2001); this working paper provides an overview of data availability and an examination of levels of uncertainty for GBD estimates for 2002.

2. GBD version changes

GBD Version 1 estimates for 2000 Version 1 of the GBD estimates for the year 2000 were published in the World Health Report 2001. The data sources and methods for Version 1 estimates were documented in a discussion paper available on the WHO website (Murray et al. 2001b). Subsequent versions of the GBD estimates have been progressively made available in the World Health Reports 2002 to 2004 and are also available on the WHO website at www.who.int/evidence/bod. Apart from the incorporation of new epidemiological data for specific causes, later versions incorporated improvements in mortality data availability, and improvements in methods used for life table and cause of death analysis. These are summarized in a more detail below. For these reasons, the later Versions 4 and 5 GBD estimates for 2002 are not directly comparable with earlier versions, with non-minor changes in estimates for some causes and age groups. GBD Version 2 estimates for 2000 Version 2 estimates formed the basis for the comparative risk assessments for 26 major risk factors and the analyses of the cost-effectiveness of interventions for these risks which were the main topic of the World Health Report 2002 (World Health Organization 2002). Full details of the CRA analyses for these risk factors were published in book form in 2004 (Ezzati et al. 2004). A summary of the CRA methods is also available on the WHO website as a supplementary file to the World Health Report 2002. GBD Version 3 estimates for 2002 The World Health Report 2003 reported burden of disease estimates for 2002 based on Version 3 revisions of the GBD. The data sources and methods used for these Version 3 revisions were documented in Discussion Paper 54 (Mathers et al. 2003a). These estimates contained substantial revisions in some areas: Following the Scientific Peer Review (Anand et al. 2003), significant improvements were made in both data and methods used to calculate life expectancies for WHO Member States. There was a substantial effort to obtain recent death registration data, surveys and censuses, particularly from developing countries. This r esulted in significant changes and improvements in estimates of child and adult mortality levels for some countries. The CODMOD model for estimation of the proportions of deaths across major cause groups was revised to take into account the much larger historical dataset on cause distributions of deaths. For countries without useable cause of death registration data, there were substantial revisions in the regional cause-of-death distributions used to estimate deaths for detailed cause groups within major cause groups. Version 3 of the GBD 2000 included new reviews of the epidemiological data and new or revised disease models for many causes. In particular, the age distribution of HIV/AIDS deaths was substantially revised to use the detailed age-specific e stimates prepared by UNAIDS (Walker et al. 2003). The total war deaths of 310 million reported for 2000 in the World Health Report 2001 and in the World Report on Violence were s ubstantially revised downwards to around 170
5

million in 2002. This revision resulted from a combination of correction of several errors (resulting in significant numbers of war deaths in several countries where there should have been few) and of a reduction in estimated numbers of deaths for a number of the larger conflicts, reflecting downwards revisions in several of the data sources. As part of the analysis effort to produce the version 3 estimates, consistent back revisions of estimates for the year 2000 were also carried out. Version 3 estimates for the year 2001 were then interpolated between the estimates for the years 2000 and 2002. These version 3 estimates for 2001 were used to prepare the global burden of disease estimates and tabulations used i the n Disease Control Priorities Project (Jamison et al. 2006; Lopez et al. 2006). GBD Version 4 estimates for 2002 The GBD estimates for 2002 were updated and republished in the World Health Report 2004. These Version 4 estimates were updated only to incorporate new information received from Member States as a result of the consultations in 2003, but too late for inclusion in the World Health Report 2003. Life tables for eight countries were revised from those prepared for the World Health Report 2003 to include new vital registration information received too late for inclusion in that report. Additionally, mortality and DALY estimates were revised to reflect new estimates from WHO programs and partners for mortality due to six causes, based on improved data and recent evidence: Latest provisional estimates of HIV/AIDS mortality available for countries as at the beginning of February 2004 were incorporated. TB prevalence and mortality for 2002 were also revised, as published in WHO's Global Tuberculosis Control report for 2004. Estimates of measles mortality were revised to take into account new information on the effects of supplemental immunization campaigns in reducing measles mortality. Estimates of deaths due to pertussis, polio and tetanus for 2002 were also revised to take into account new information on notifications and immunization coverage. GBD Version 5 estimates for 2002 The 2004 Report on the global AIDS epidemic (UNAIDS 2004) contained substantially revised estimates of HIV/AIDS mortality for many countries, taking into account new and different sources of data, such as national household surveys, as well as improved information on emerging epidemics in Eastern Europe, Asia and the Americas. For some countries, these estimates differed substantially from the draft estimates used in Version 4. As well as updating HIV/AIDS mortality to reflect the 2004 Report estimates, further updates were also carried out for malaria, schistosomiasis and intestinal helminths. Country-specific estimates of malaria mortality and incidence were updated to reflect recent work carried out in collaboration with other WHO programs and external expert groups to refine and revise these country-specific estimates of malaria mortality (Korenromp et al. 2003; Rowe et al. 2005).

3.

Availability of data for GBD estimates

The GBD 2000-2002 incorporated a range of new data sources to develop internally consistent estimates of incidence, health state prevalence, severity and duration, and mortality for 135 major causes, by sex and by eight age groups. Data on mortality levels and causes of death
6

According to data provided by 112 Member States to WHO, only about one third of the estimated 56 million deaths occurring annually are recorded in death registration systems (Mathers et al. 2005). If the sample registration systems of India and China are considered to provide information on their whole populations, then information is available for around 72 percent of the global population. In recent years, considerable attention has also been placed on obtaining data on child and maternal mortality through instruments such as the Demographic and Health Surveys (DHS) and UNICEF's Multiple Indicator Cluster Surveys (MICS). The sources of information on levels of child and adult all-cause mortality used to construct life tables for 192 WHO Member States for the GBD 2002 are given in the Annex Tables to Discussion Paper 54 (Mathers et al. 2003a). These sources of information are summarized in Table 1 by region and by type of data. The regions used in this table are defined in Annex Table A-1. In the last decade, computerization of death registration data at country level and electronic transmission to WHO have considerably improved the timeliness of information received. The number of countries submitting their underlying causes of death data to WHO using 10th revision of the International Classification of Diseases (ICD-10) has increased from 4 in 1995 to 75 in 2003. There are still around 50 countries reporting data using the ICD-9 (9th revision) and only one country using the 8th revision (Mathers et al. 2005).
Table 1: Availability of data for the estimation of all cause mortality rates by age and sex, by type of data and region.
East Asia and Pacific Europe & Central Asia Latin America & Caribbea n Middle East & North Africa

Type of data
Number of countries with all cause mortality data: Death registration data for 2001 Complete Incomplete (a) Death registration data for years prior to 2001 (b) Complete Incomplete (a) Data for levels of child and adult mortality Data for levels of child mortality only Number of data collections Country-years of death registration data used 2001 available 2001 not available (c) Other sources of information on child and adut mortality Total data sets used

High incom e

South Asia

SubSaharan Africa

Total

1 1

11 6

25 2

2 5

1 0

0 1

1 0

41 15

1 3 12 4

0 9 1 0

5 2 3 0

3 13 6 3

0 1 9 4

0 1 4 2

0 2 2 42

9 31 37 55

4 89

17 132

28 114

7 279

1 45

1 23

1 29

59 711

70 163

22 171

16 158

122 408

67 113

48 72

190 220

535 1305

(a) Completeness of death registration data assessed using standard demographic methods (see Lopez et al. 2002) (b) Includes countries where death registration data for years prior to 2001 were used to project levels of child and adult mortality to construct a life table based on a country standard derived from the death registration data. (c) Also includes countries where death registration data were used to poject levels of child and adult mortality as inputs to the WHO logit life table system using a global standard.

Table 2: Availability of data for the estimation of causes of death by age and sex, by type of data and region.
East Asia and Pacific 1 5 Europe & Central Asia 11 16 Latin America & Caribbea n 5 25 Middle East & North Africa 1 3

Type of data
Number of countries Death registration data (a) (coverage of 85% or more) Death registration data (coverage <85%) adjusted using cause of death models Sample registration and surveillance (b) No data- use of cause of death models and regional model pattern of causes of death Epidemiological estimates for mortality due to specific causes used where applicable

High incom e 29 5

South Asia 0 1

SubSaharan Africa 1 3

Total 48 58

2 14

0 0

0 3

0 2

1 10

1 6

1 42

5 77

(c)

(d)

(e)

(c)

(c)

(c)

(c)

(c)

Per cent of population Death registration data (a) (coverage of 85% or more) Death registration data (coverage <85%) adjusted using cause of death models Sample registration and surveillance (b) No data- use of cause of death models and regional model pattern of causes of death Total 0.0 52.7 94.4 13.0 0.1 0.0 0.2 19.5

5.6 73.3

47.3 0.0

5.3 0.0

84.2 0.0

50.0 1.7

1.4 74.5

8.6 1.4

17.1 39.1

21.1 100.0

0.0 100.0

0.3 100.0

2.9 100.0

48.1 100.0

24.2 100.0

89.8 100.0

24.3 100.0

(a) The threshold of coverage of 85% used for causes of death differs from that used for registration of deaths (95%) since the biases from underreporting of the fact of death are more serious for assessing levels of all cause mortality than for assessing the distribution of causes. (b) Inciudes countries with death registration or surveillance systems relying heavily on verbal autopsy methods for ascertaining causes of death. (c) AIDS, tuberculosis, measles, pertussis, poliomyelitis, tetanus, acute lower respiratory infections, Chagas disease, maternal conditions, perinatal conditions, cancers, drug use disorders, rheumathoid arthritis and war (d) AIDS, drug use disorders and war (e) Drug use disorders and war

Death registration data containing useable information on cause of death distributions were available for 107 countries, the majority of these in the high income group, Latin America and the Caribbean and Europe and Central Asia (see Table 2). For more detailed information on these data sources, refer to the Annex Tables of Discussion Paper 54 (Mathers et al. 2003a). Where the latest available year was earlier than 2002, death registration data from 1980 up to the latest available year were analyzed as a basis for projecting recent trends for specific causes, and these trend estimates were used to project the cause distribution from the latest available year. When estimating cause of death

distributions for very small countries, an average of the three last years of data were used to minimize stochastic variation. Specific epidemiological estimates for some causes were also taken into account in analysing causes of death for countries (Mathers et al. 2003a). Table 3 summarizes the numbers of studies (population-based epidemiological studies, disease registers and notifications systems) that contributed to the estimation of mortality due to 21 specific causes of death, including HIV/AIDS, malaria and tuberculosis. In excess of 2,700 datasets contributed to the estimates for these 21 causes of death, with almost one third of these relating to sub-Saharan Africa. Estimates of deaths by cause, age and sex were carried out for 226 countries and territories drawing on a total of 770 country-years of death registration data, as well as 535 additional sources of information on levels of child and adult mortality, and in excess of 2,700 data sets providing information on specific causes of death in regions not well covered by death registration systems.

Table 3: Numbers of data sets (a) contributing to epidemiologically based estimates of deaths due to specific causes, by region and cause
East Asia and Pacific 24 14 Europe & Central Asia 27 26 0 Latin America & Caribbea n 34 27 15 Middle East & North Africa 16 13 8

Cause
Tuberculosis (b) HIV/AIDS (b) Diarrheal diseases Childhood-cluster diseases Pertussis Poliomyelitis Diphtheria Measles Tetanus Meningitis Hepatitis B and C Malaria Tropical-cluster diseases Trypanosomiasis* Chagas disease Schistosomiasis Acute lower infections respiratory

High income 31 29 0

South Asia 8 5 21

SubSahara n Africa 39 37 24

Total 179 150 73

14 22 12 22 48 23 40 9

33 27 25 18 23 18 27 0

124 37 8 22 34 43 113 0

64 32 2 32 27 30 47 2

14 15 14 12 40 12 43 1

8 8 8 8 32 4 18 7

45 47 46 47 79 27 67 142

302 192 115 127 289 157 355 161

0 0 6 2 6 14 7 14 11 3 296

0 0 0 0 0 32 0 12 11 1 280

0 0 1 2 1 10 0 40 43 7 545

0 31 3 18 9 27 7 12 18 1 438

0 0 8 0 5 11 11 10 10 0 243

0 0 0 9 4 13 19 3 6 0 181

36 0 37 18 20 49 12 14 15 6 807

36 31 55 49 45 156 56 105 114 18 2,765

Maternal conditions (all causes) Unsafe abortion Perinatal causes Malignant neoplasms Drug use disorders War Total

10

(a) includes population-based epidemiological studies, disease registers, surveillance and notifications systems Where possible, regional and global totals refer to numbers of separate studies, or country-years of reported data from surveillance or notifications systems. (b) Totals refer to numbers of countries for which data were available, not to total data sets or country-years.

11

Although these epidemiological studies allow estimation of deaths for certain causes in populations without death registration data, they do not cover many important causes of death in these populations, such as cardiovascular disease or injuries. In order to address these information gaps, models for estimating broad cause-of-death patterns were used as the starting point for indirect methods of estimating attributable mortality for a comprehensive list of detailed causes. The causeof-death model used in the GBD 1990 was substantially revised and enhanced for estimating deaths by broad cause group in regions with limited information on mortality. The statistical model was enhanced by the adaptation of models for compositional data that were previously developed in other areas, and a substantially larger dataset of 1,613 country-years of observations used for analysis. Income per capita was added to the model as an explanatory variable, in addition to level of all cause mortality (Salomon and Murray 2002). Data on incidence, prevalence, duration and severity of conditions Estimating the years lived with a disability (YLD) is the most complex component of burden of disease analysis, since it requires systematic assessments of the available evidence on incidence, prevalence, duration and severity of a wide range of conditions. Various methods have been developed to reconcile often fragmented and partial estimates available from different studies. A specific software tool, DisMod, has been developed to assist in the development of internally consistent estimates (Barendregt et al. 2003). A wide range of data sources were used for the analysis of incidence, prevalence and YLD for the GBD 2000-2002. These included: (1) Disease registers Disease registers record new cases of disease based on reports by physicians and/or laboratories. Registers are common for infectious diseases (e.g. tuberculosis), cancer, congenital anomalies, a number of relatively rare diseases (e.g. cystic fibrosis or thallassaemia), and sometimes for conditions such as diabetes, schizophrenia and epilepsy. For some Group I conditions, WHO programs maintain up-to-date databases based on diseases registers, population surveys and epidemiological studies. These have been used where available. (2) Population surveys Interview surveys such as the National Health Interview Survey in the USA can provide selfreported information on disabilities, impairments and diseases. However, self-report data is generally not comparable across countries (Murray et al. 2002; Sadana et al. 2002). It is also often difficult to attribute impairment to the underlying causes, and, there are often considerable differences between lay self-reporting of disease causes and the actual underlying disease causes in terms of defined GBD disease categories. In general, the results of health examination surveys have contributed more to YLD calculations than self-reported interview surveys. The CIDI and DIS questionnaires used in mental health surveys are examples of standard questionnaires based on self-report that have undergone validity testing and have been used in assessing YLD for mental disorders for the GBD 2000-2002. (3) Epidemiological studies Some of the most useful sources of information for the GBD 2000-2002 were populationbased epidemiological studies. Particularly, longitudinal studies of the natural history of a disease have provided a wealth of information on the incidence, average duration, levels of severity, remission and case fatality. Such studies are rare because they are very costly to
12

undertake. As they are often conducted in a particular region or town, judgment is needed to extrapolate results to the whole population. (4) Health facility data In the majority of cases, routine data on consultations by diagnosis was not found to be very helpful in estimating YLD. Facility based data unless the coverage of the health system is near complete will always be based on biased samples of the conditions present in the community. Likewise, hospital deaths are unlikely to be useful due to the same problems of selection bias. Examples of conditions that were estimated from hospital data with national or quasi-national population coverage include: perinatal and maternal conditions, meningitis, stroke, myocardial infarction, some sequelae identifiable via data on surgical interventions and injuries. The epidemiological reviews underlying the GBD 2000-2002 estimates of YLD have been progressively documented and published in draft form on the WHO website (www.who.int/evidence/bod) and in peer reviewed publication. Discussion Paper 54 provides an overview of data sources and methods for various specific causes together with references to more detailed documentation (Mathers et al. 2003a). While it is difficult to quantify the exact numbers of data sources used for the YLD estimates for GBD 2000-2002, an attempt has been made to provide an approximate count by region (Table 4). This table counts numbers of data sources (registers, notifications, health facility and other official data sets, and epidemiological studies for each of the causes included in the GBD. For some causes, only counts were available of the numbers of countries in each region for which country-specific data were used. In some cases, it was not feasible to do an exact recount of studies by region, and an approximate regional break down was estimated from prior counts according to the 17 epidemiological subregions used in the GBD (Mathers et al. 2003b). Additionally, it was not always possible to be consistent in the counting of studies carried out across multiple countries or multiple years. Finally, it must be emphasized that there is huge variability in the information content across studies or data sets and that small epidemiological studies count equally in Table 4 with national hospital inpatient data on injuries for an entire population-year. Thus the counts in Table 4 should be treated as reasonably indicative of the empirical bases underlying the GBD 2000-2002 without overinterpreting differences between causes or regions. That said, it is striking that of the more than 8,000 datasets estimated to have been used for the GBD 2000-2002 estimation of YLD, nearly 6,600 relate to Group I causes, and only 18 datasets to Group III causes (injuries). Furthermore, one quarter of the datasets relate to populations in subSaharan Africa, and around one-fifth to populations in High income countries. While this predominance of data relating to Group I conditions, and to sub-Saharan Africa, is not entirely surprising, the paucity of data for some of the leading noncommunicable diseases is more surprising. For example, for several of the leading causes of burden among mental disorders, one or no useable population-based studies were found for some regions. For ischaemic heart disease, very few studies of the incidence or prevalence of angina pectoris or acute myocardial infarction were found outside high income countries. Assuming that for causes in Table 4 where the counts relate to countries rather than datasets, there are on average two datasets per country, then overall, approximately 8,700 datasets contributed to the estimation of YLD. Not counting again studies which also contributed to the estimation of causespecific mortality rates, an additional 1,370 datasets were used for the estimation of YLL. In total, the GBD 2000-2002 has drawn on in excess of 10,000 datasets or studies.

13

Table 4: Numbers of country data sources (a) contributing to the estimation of YLD, by region and cause.
East Asia and Pacific Europe & Central Asi Latin America & Caribbea n Middle East & North Africa

GBD Cause Category

I. Communicable, maternal, perinatal and nutritional conditions Tuberculosis (b) Sexually transmitted diseases excluding HIV (c) HIV/AIDS (b) Diarrheal diseases Childhood-cluster diseases (d) Pertussis . . Poliomyelitis Diphtheria Measles . Tetanus

High incom e

South Asia

SubSaharan Africa

Total (a)

24 143 14 155

27 318 26 0

31 297 29 0

34 148 27 27

16 45 13 55

8 99 5 29

39 406 37 91

179 1,456 150 357

14 22 12 22 48 23 4 36 9

33 27 25 18 23 18 4 23 0

124 37 8 22 34 43 28 85 0

64 32 2 32 27 30 6 41 2

14 15 14 12 40 12 6 37 1

8 8 8 8 32 4 10 8 7

45 47 46 47 79 27 11 56 98

302 192 115 127 289 157 69 286 117

Meningitis Hepatitis B Hepatitis C (e) Malaria Tropical-cluster diseases Trypanosomiasis (b) Chagas disease Schistosomiasis Leishmaniasis (f) Lymphatic filariasis (b) Onchocerciasis Leprosy (b) Dengue (g)

0 0 6 3 29 0 32 91 10 11 29 15 4

0 0 0 7 0 0 10 0 1 0 0 0 0

0 0 1 4 2 0 3 15 4 0 6 5 9

0 31 3 15 8 6 8 170 0 4 23 15 0

0 0 8 13 5 0 14 0 0 5 13 12 2

0 0 0 4 5 0 8 4 3 4 10 30 2

36 0 37 20 40 26 45 2 0 19 53 18 7

36 31 55 66 89 32 120 282 18 43 134 95 24

Japanese encephalitis (b) Trachoma Intestinal nematode infections Acute lower infections Otitis media Maternal conditions Maternal haemorrhage Maternal sepsis Hypertensive disorders of pregnancy Obstructed labour Abortion Perinatal conditions Low birth weight (i) Birth asphyxia & trauma respiratory

3 2 1 2 32

0 0 0 0 10

9 11 2 2 27

2 3 1 2 11

0 0 0 0 13

1 1 2 1 49

13 14 12 14 156

28 31 18 21 32

28 7

27 0

33 0

33 7

15 11

7 19

41 12

184 56

14

15

Table 4 (continued): Numbers of country data sources (a) contributing to the estimation of YLD, by region and cause.
East Asia and Pacific Europe & Central Asi Latin America & Caribbea n Middle East & North Africa

GBD Cause Category

Nutritional deficiencies Protein-energy malnutrition (j) Iodine deficiency Vitamin A deficiency Iron-deficiency anemia II. Noncommunicable diseases Malignant neoplasms Incidence Survival Diabetes mellitus- Type I (l) Diabetes mellitus- Type II Neuro-psychiatric conditions Unipolar depressive dis. Bipolar affective disorder Schizophrenia Epilepsy (k) Alcohol use disorders Alzheimer/other dementias Parkinson disease (k) Multiple sclerosis Drug use disorders Post-traumatic stress dis. Obsessive-compulsive dis. Panic disorder Insomnia (primary) Migraine Mental retardation attrib. to lead exposure Sense organ diseases Vision disorders (b) Hearing loss, adult onset Cardiovascular diseases Rheumatic heart disease Ischemic heart disease Cerebrovascular disease Other forms of heart disease Respiratory diseases COPD Asthma Musculoskeletal diseases Rheumatoid arthritis

High incom e

South Asia

SubSaharan Africa

Total (a)

61 17 10 14

28 13 2 1

15 20 8 0

116 13 12 1

37 17 4 15

30 12 4 5

132 44 29 33

419 136 67 69

11 3 22 6

8 4 12 4

25 15 41 8

11 1 17 5

10 0 5 8

2 1 1 3

14 0 2 6

81 24 100 40

5 2 4 1 24 10 2 4 11 1 2 2 2 6 10

5 1 3 1 43 3 1 24 11 0 0 0 2 2 12

27 14 25 7 56 87 7 116 43 6 14 22 9 11 23

6 1 3 6 39 3 1 3 18 1 3 3 5 5 21

3 1 1 1 13 0 0 5 10 0 0 1 1 2 4

4 2 3 4 5 4 1 1 6 0 1 0 1 0 14

6 0 6 8 34 3 1 1 15 0 0 2 1 1 9

56 21 45 28 214 110 13 154 114 6 20 30 21 43 93

11 5 0 15 3 4 0

3 0 0 0 11 8 0

9 12 0 7 58 28 5

4 1 0 12 0 1 0

5 1 0 9 2 5 0

4 5 0 15 4 0 0

19 1 0 26 1 6 0

55 25 0 84 79 52 5

24 17

10 14

32 74

10 20

4 12

16 6

8 7

104 149

29

16

Osteoarthritis Congenital malformations Oral conditions (f) III. Injuries Total (m)

1 3 22 3 1,155

1 42 24 1 914

9 5 27 7 1,735

1 29 32 1 1,239

0 6 15 0 590

2 9 7 0 522

1 6 35 6 1,955

15 100 162 18 8,096

(a) includes population-based epidemiological studies, disease registers, surveillance and notifications systems , but excludes death registration data (see Table 1 and 2). Where possible, regional and global totals refer to numbers of separate studies, or country-years of reported data from surveillance or notifications systems. Global total may include global review studies not counted in subtotals for regions. For some causes, regional subtotals were approximately estimated from subtotals for WHO regions and subregions. (b) Totals refer to numbers of countries for which data were available, not to total data sets or country-years. (c) Regional subtotals estimated from current distribution of studies in the WHO STI surveillance database (d) Regional subtotals estimated from subtotals of numbers of studies by WHO region, rather than by re-accessing original databases. (e) Count of country-years of data, data available for 133 countries. (f) Approximate estimate from current WHO database, original extraction from surveillance data sources not available. (g) Country years of surveillance reports (approximate minimum estimate for Latin America and Caribbean). (h) Global total for number of countries with quantitative data, approximate regional subtotals pending detailed count. (i) Approximate estimate based on finally published literature review. (j) Regional distribution of the 419 national studies used assumed similar to that of the current 442 national studies in the WHO malnutrition database. (k) Approximate minimum estimate. Several global reviews used and studies not separately counted. (l) Total of 100 population-based registries in 50 countries. (m) Actual numbers of studies used exceed the minimums shown here, based on summed table entries for specific causes irrespective of whether counts of data sets or countries.

4.

GBD prior estimates for WHO Member States

The GBD 2000-2002 analyses have been used to prepare estimates of mortality and burden of disease for WHO Member States for the year 2002. These are available in summarized form by country and cause on the WHO website at www.who.int/evidence/bod (select links to estimates, then to 2002 revised, then to country estimates). Mortality estimates were based on analysis of latest available national information on levels of mortality and cause distributions as at mid-2004. YLD estimates were based on the GBD analyses of incidence, prevalence, duration and severity of conditions for the relevant epidemiological subregion, together with national and subnational level information available to WHO (Mathers et al. 2003a). The GBD used the population estimates for WHO Member States prepared by the UN Population Division (2002 revision). Initial WHO estimates and technical explanations were sent to Member States for comment in 2003. Comments or data provided in response were discussed with them and incorporated where possible. The prior estimates for countries should, however, still be interpreted as the best estimates of WHO, based on the evidence available to it in mid-2004, rather than the official estimates of

17

Member States. They have been computed using standard categories and methods to ensure crossnational comparability and may not be the same as official national estimateis produced using alternate, potentially equally rigorous methods. Where feasible, country-specific YLD estimates were made for a number of causes. For the estimates for 2002, these included childhood vaccine preventable diseases, malnutrition, HIV/AIDS, cancers, and diabetes (see Table 5). For other causes, regional YLD estimates were used, together with country-specific and regional cause of death information, to develop country-specific estimates of YLD by age and sex. The five methods used are described below and summarized in Table 5. Country-specific Data For the causes listed in the first row of Table 5, YLD estimates were made directly for Member States using available data on the incidence or prevalence of each condition. For the Group I conditions, death registration data together with databases of country-level studies developed by WHO programmes and UNAIDS were used to estimate country-level incidences and prevalences. Methods used to estimate the prevalence of malignant neoplasms at the country level, based on national incidence, mortality and survival data, were described by Mathers et al (Mathers et al. 2002). Diabetes prevalence was based on updated country-level prevalence estimates prepared by the WHO Management of Non-communicable Diseases and Mental Health Programme (NMH) according to the revised WHO definition of diabetes cases (Wild et al. 2004). Variations in the prevalence of unipolar depressive disorders in some European countries, Australia, New Zealand, and Japan were estimated directly from relevant population studies for the GBD (Ustun et al. 2004). For other countries i the A (lowest mortality) regions, country-specific prevalences Pc for males n and females aged 15-59 have been estimated using a regression model on suicide rates as follows: P c = PR + 0.0919(S C - SR ) where PR is the regional depression prevalence, SC and SR are the country and regional suicide rates (ages 15-59 both sexes combined). For other regions, it was assumed that the variation of depression prevalence with suicide rate was half that of A countries, and the range of variation was restricted from a m inimum of one half the regional average to a maximum of twice the regional average. YLD/YLL Ratios For specific disease and injury causes where mortality was responsible for a significant proportion of the total burden (incidence YLD/YLL ratio less than 5), regional estimates of incidence YLD/YLL ratios by age and sex together with country-level estimates of YLL were used to estimate countrylevel YLD. This process ensured that country-specific knowledge on the epidemiology of the disease (as reflected in the country-level mortality estimates of that disease) was used to adjust the regionallevel patterns of disability due to that cause. Group cause YLD/YLL Ratios For certain causes, regional age-sex specific incidence/mortality ratio for the overall cause group were applied to country-specific mortality for each cause in the group to estimate incidence and YLD (see Table 5). This avoided problems resulting from statistical fluctuations in countries with small numbers of deaths for certain cause-sex-age-groups.

18

Regional Prevalence YLD Rates For specific disease and injury causes where mortality is not responsible for a significant proportion of the total burden (incidence YLD/YLL ratio is 5 or higher), or where there is insufficient evidence to predict variations in YLD rates from variations in mortality rates, regional estimates of YLD rates per 1 000 population by age and sex were used together with country-level population distribution estimates to estimate YLD for each country.

19

Table 5. Methods used to estimate country-level YLD for 2002

Method Imputation of country-specific Causes incidence rates from regional incidence rates Country-specific prevalence or incidence data used Tuberculosis, HIV/AIDS, pertussis, diphtheria, measles, tetanus, schistosomiasis, lymphatic filariasis, onchocerciasis, trachoma, abortion, protein-energy malnutrition, iodine deficiency, diabetes mellitus, unipolar depressive disorders*, alcohol use disorders, drug use disorders, sense organ disorders excluding others, asthma Hepatitis B, hepatitis C, malaria, lower respiratory infections, malignant neoplasms, rheumatic heart disease, hypertensive heart disease, inflammatory heart disease, other cardiovascular diseases, peptic ulcer disease, cirrhosis of the liver, nephritis and nephrosis

Country-specific data

Incidence/ mortality ratio short duration

Regional age-sex specific incidence/mortality ratio applied to country-specific mortality to estimate incidence. Same approach used for prevalence

Incidence/ mortality ratio long duration

Regional age-sex specific Meningitis, endocrine disorders, ischaemic heart disease, incidence/mortality ratio applied to cerebrovascular disease, COPD, other respiratory diseases, country-specific mortality to other digestive diseases, other genitourinary system diseases estimate incidence. Resultant country/regional sex specific allages YLD[0,0] ratio applied to regional prevalence YLD total for that sex to estimate countryspecific prevalences by age and sex Regional age-sex specific incidence/mortality ratio applied to country-specific mortality of the cause group to estimate incidence Maternal haemorrhage and sepsis, hypertensive disoders of pregnancy, perinatal causes, congenital malformations excluding cleft lip and palate and Down syndrome, , road traffic accidents, poisonings, falls, fires, drownings, other unintentional injuries, self-inflicted injuries, violence, other intentional injuries Sexually transmitted diseases excluding HIV, diarrhoeal diseases, poliomyelitis, trypanosomiasis, Chagas disease, leishmaniasis, leprosy, dengue, Japanese encephalitis, intestinal nematode infections, other infectious diseases, upper respiratory infections, otitis media, obstructed labor, vitamin A deficiency, iron-deficiency anaemia, other nutritional disorders, other neoplasms, bipolar affective disorder, schizophrenia, epilepsy, Alzheimer and other dementias, multiple sclerosis, post-traumatic stress disorder, obsessive-compulsive disorder, panic disorder, insomnia (primary), migraine, mild mental retardation attributable to lead exposure, other neuropsychiatric disorders, other sense organ disorders, appendicitis, benign prostatic hypertrophy, skin diseases, musculoskeletal diseases, cleft lip and palate, Down syndrome, oral conditions

Group cause incidence/ mortality ratio

Regional rates

Regional age-sex specific incidence and prevalence rates applied to countries

Average of ratio and rate method results

Regional age-sex specific Other maternal conditions, Parkinson disease, war incidence and prevalence rates and incidence and prevalence rates from the ratio methods are averaged and applied to countries

* Some country data plus regression model based on suicide rates.

Levels of evidence for country-level information on mortality and burden of disease Country-level estimates of deaths and DALYs by cause (for all ages and both sexes combined) have been made available on the WHO website at www.who.int/evidence/bod with summary information on levels of evidence as follows.

20

The data sources and methods used for the estimation of total deaths (all causes) for each Member State are summarized into four levels of evidence: Level 1a Complete death registration data available for years 2001 and/or 2002. Level 1b Complete death registration data available for an earlier time period. All cause mortality projected to 2002. Level 2a Death registration data available for years 2001 and/or 2002. Completeness for latest year estimated using standard demographic methods for child deaths under age 5 and for deaths at ages 5 and over. Estimated completeness used to adjust death registration data. Level 2b Death registration data available for available for an earlier time period. Completeness for latest year estimated using standard demographic methods for child deaths under age 5 and for deaths at ages 5 and over. Estimated completeness used to adjust death registration data and then all cause mortality was projected to 2002. Level 3a Country information for years 2001 and/or 2002 available on levels of child mortality (between ages 0 and 5) and adult mortality (between ages 15 and 60). WHO modified logit life table system used with global standard to estimate all cause mortality. Level 3b Country information available for an earlier time period on levels of child mortality (between ages 0 and 5) and adult mortality (between ages 15 and 60). Child and adult mortality levels were projected to 2002 and the WHO modified logit life table system used with global standard to estimate all cause mortality. Level 4a Country information for years 2001 and/or 2002 available on level of child mortality (between ages 0 and 5) only. Levels of adult mortality excluding HIV/AIDS and war deaths predicted from child mortality. HIV/AIDS and war deaths added separately and WHO modified logit life table system used with global standard to estimate all cause mortality. Level 4b Country information prior to 2001 available on level of child mortality (between ages 0 and 5) only. Levels of adult mortality excluding HIV/AIDS and war deaths predicted from projected 2002 HIV/AIDS-free child mortality. HIV/AIDS and war deaths added separately and WHO modified logit life table system used with global standard to estimate all cause mortality. The data sources and methods used for the estimation of deaths by cause for each Member State are also summarized into four levels of evidence. Additional information for estimating country-level deaths due to certain specific causes were obtained from studies, WHO technical Programmes and UNAids for the following conditions: AIDS, tuberculosis, measles, pertussis, poliomyelitis, tetanus, acute lower respiratory infections, Chagas, maternal conditions, perinatal conditions, cancers, drug use disorders, rhumathoid arthritis and war. Level 1a Complete death registration data with cause of death coded using ICD-9 or ICD-10 available for years 2001 and/or 2002. Less than 10% of deaths coded to ICD codes for symptoms, signs, and ill-defined conditions, injuries where the intent is not determined; cardiovascular "garbage" codes, and cancer deaths coded to categories for secondary or unspecified sites. Level 1b Complete death registration data available . Complete death registration data with cause of death coded using ICD-9 or ICD-10 available for for an earlier time
21

period. Less than 10% of deaths coded to ICD codes for symptoms, signs, and illdefined conditions, injuries where the intent is not determined; and cardiovascular and cancer "garbage" codes. Level 2a Death registration data available for years 2001 and/or 2002. Adjustments to cause of death distribution required for incomplete registration and/or for use of non-ICD9 or ICD-10 coding and/or for the more than 10% of deaths coded to ill-defined conditions, and cardiovascular, cancer and injury "garbage" codes. Level 2b Death registration data available for latest year earlier than 2001. Adjustments to cause of death distribution required for incomplete registration and/or for use of non-ICD-9 or ICD-10 coding and/or for the more than 10% of deaths coded to illdefined conditions, and cardiovascula, cancer and injury "garbage" codes. Level 3 Level 4 Country information on causes of death available based on verbal autopsy methods. Country information on causes of death not available for most causes. Cause of death modelling used to estimate broad distribution of causes of death for Groups I, II and III by age and sex for the country level of all cause mortality and per capita income. Cause of death patterns within the three major cause groups based on death registration data from other countries in the region. Further country-level information and data on specific causes listed above was also used. Country data on causes of death (Levels 1 to 3) used to adjust regional YLD distributions for causes with significant case fatality. Partial country-specific information on incidence or prevalence of non-fatal causes available. Level 4 information on causes of death and regional estimates of YLD used for most causes. Partial country-specific information on incidence or prevalence of some nonfatal causes available.

Two levels of evidence are distinguished for YLD estimates as follows: Level 3

Level 4.

In the summary tables available on the WHO website, columns containing estimates of deaths or DALYs for WHO Member States in 2002 are colour-coded to summarize the levels of evidence and uncertainty. xThree colour codes are used to denote the following levels:
Total mortality Level 1or Level 2 Causes of death Level 1 or Level 2

Colour BLUE

YLD Level 3 Death registration data, complete or incomplete, containing useable information on causes of death is available for the country, and used to adjust regional YLD distributions for causes with significant case fatality. Partial country-specific information on incidence or prevalence of non-fatal causes available. Other forms of information on child and adult mortality or causes of death (eg. verbal autopsy) available. Country-specific information on mortality for specific causes available.Partial country-specific information on incidence or prevalence of nonfatal causes available. Country information on level of adult mortality not available and it was predicted from child mortality level OR cause of death information for most causes not available, and cause pattern predicted using cause-of-death models. Partial country-specific information on incidence or prevalence of non-fatal causes available.

YELLOW

Total mortality or causes of death have level 3 evidence, neither is level 4.

RED

Level 4 OR Level 4

Level 4

22

Estimated 95% uncertainty ranges for the expected number of total deaths for 2002 are also given for each country in the tables relating to deaths and provide some indication of the level of uncertainty for estimates in that country (see following section). Cause-specific estimates will generally have even greater relative uncertainty ranges, as will DALY estimates, with the possible exception of some causes such as HIV/AIDS where separate country-specific analyses have been used.

23

5.

Uncertainty analysis

The GBD 2000-2002 estimated mortality and burden of disease for a comprehensive set of disease and injury causes and for all regions of the world, including regions with limited, incomplete and uncertain data. It is thus important to attempt to provide some analysis and guidance on levels of uncertainty in estimates, in order to allow the user of the information to assess whether the information uncertainty range is compatible with the purpose at hand (Murray, Mathers, and Salomon J.A. 2003). This is difficult to do, since apart from the large number and disparate nature of the data sources used, there is often limited information or knowledge of the quality and potential biases in the data. This section gives an overview of initial efforts to quantify the uncertainty associated with the estimation of deaths by cause, and with epidemiologic estimates of incidence and prevalence for the GBD. Uncertainty may arise from incomplete information (e.g., when we base estimates for a population on observations from a sample), from potential biases in information (e.g., how representative for a whole population are estimates from a study of a subgroup, how validly does the survey instrument address the quantity of interest), from disagreements between information sources (e.g., when we have several studies giving different estimates for the same quantity), from model uncertainty (e.g., the variables or functional form specified in a regression model), or it may be inherent to the data generation process itself (e.g., we may only infer risks from event counts in a population, which means that we can never know the risks themselves with certainty. Most measurement involves not only random (stochastic) error but also systematic error, arising from biases in the measurement instrument (e.g., unrepresentativeness of a sampling frame for a survey), or from inaccuracies in the assumptions used to infer the actual quantity from the available data (e.g., estimating prevalence of a disease for a country from studies of representative subpopulations). Much of the uncertainty in estimates of deaths or DALYs is associated with the assessment of systematic errors in primary data. While most countries have some information about prevalence, incidence and mortality from some diseases and injuries, and some information on population exposure to risk factors, it is generally fragmented, partial, incomparable and diagnostically uncertain. The general approach of the GBD to describing and estimating uncertainty in quantities of interest is to express them as probability distributions using a Bayesian interpretation of probability as expressing uncertainty of an observed or hypothetical event given a set of assumptions about the world (Morgan and Henrion 1990). Probability distributions can thus be used to express uncertainty about epidemiologic quantities such as the prevalence of depression in a particular population, or the underlying risk of mortality due to a specific cause in a specific population. Analytic methods for dealing with uncertainty have been facilitated enormously by the revolution in computer technology. Using such technology, the general approach to propagating uncertainty in estimates relies on numerical simulation method (King, Tomz, and Wittenberg 2000; Salomon et al. 2001; Vose 2000). Uncertainty estimates for all cause mortality and life expectancies Methods for estimating life tables for each of 192 WHO Member States are described elsewhere (Lopez et al. 2002). For those countries with vital registration data projected using time series regression models on the parameters of the logit life table system, uncertainty around the regression coefficients was accounted for by taking 1000 draws of the parameters using the regression
24

estimates and variance covariance matrix of the estimators. For each of the draws, a new life table was calculated. For countries that did not have time series data on mortality by age and sex, point estimates and ranges around 5q0 and 45q15 for males and females were developed on a country-by-country basis as elsewhere (Lopez et al. 2002). For countries without useable information on levels of adult mortality, 45q15 was estimated, along with uncertainty ranges, based on regression models of 45q15 versus 5q0 as observed in a set of almost 2000 life tables judged to be of good quality(Murray et al. 2001a). Estimated levels of child and adult mortality were then applied to a modified logit life table model, using a global standard, to estimate the full life table, and HIV/AIDS deaths and war deaths added to total mortality rates where necessary. Estimates of uncertainty ranges for HIV/AIDS and war deaths were also incorporated into the life table uncertainty analysis where necessary. The final estimated uncertainty ranges for 5q0 and 45q15 for 192 WHO Member States are plotted in Figure 1 for males and females. Using Monte Carlo simulation methods, 1000 random life tables were generated by drawing samples from normal distributions around these inputs with variances defined in reference to the defined ranges of uncertainty for 5q0 and 45q15. In countries where uncertainty around 5q0 and 45q15 was considerable due to a paucity of survey or surveillance information, the samples were drawn from wide distributions but then constrained within prior specified maximum and minimum possible values for 5q0 and 45q15. For each country, the results of this analysis were 1000 different simulated life tables which were then used to describe ranges around key indicators such as life expectancy at birth and age and sex-specific mortality rates (Table 6). For high income countries where relatively complete death registration data was available, the uncertainty ranges for LE at birth were typically around 0.1 to 0.2 years. For regions such as Latin America and the Caribbean, where death registration data was available for most countries but was often not complete, the uncertainty ranges were larger, typically around 0.5 years. For regions with partial data on child mortality only, where adult mortality was predicted from child mortality, the uncertainty ranges were much larger, and for sub-Saharan Africa were typically around 5 to 10 years. This translated to considerable heterogeneity in uncertainty ranges for l for estimates of all cause and cause-specific mortality levels across WHO Member States. Uncertainty estimates for country-level estimates of mortality Provisional attempts were made to estimate indicative uncertainty ranges for the estimates of death by cause for the GBD in 2002. These uncertainty ranges t ok into account uncertainty in the o expected number of total deaths for 2002 (life table uncertainty), uncertainty in the estimated proportions of broad cause groups I, II and III (where relevant for countries without vital registration data or with incomplete coverage), uncertainty in the diagnosis of underlying cause, uncertainty arising from the miscoding of underlying cause, and fundamental Poisson uncertainty in the estimated death rate arising from the observation of a finite number of deaths in a fixed time interval. This analysis was carried out country by country. Uncertainty in the underlying cause attribution of total deaths was estimated in terms of the relative uncertainty of the proportion of deaths due to each specific cause. The estimates of cause-specific relative uncertainty were based on advice from nosologists and experts in the area of cross-country mortality analysis on the general levels of uncertainty in attribution of specific causes within the major Groups I, II and III, together with detailed advice on particular causes with known higher levels of attribution uncertainty under the International Classification of Diseases (ICD). Information on the latter causes derived from cross-country, cross-time and cross-ICD-revision comparative analyses, together with information from a variety of country-specific coding quality studies involving recoding or dual coding of deaths and comparisons with the original attributed causes.
25

Figure 1. Adult mortality versus child mortality for 191 WHO Member States, 2002

1000 900 Adult Mortality (45q15 per 1000) 800 700 600 500 400 300 200 100 0 0 50 100 150 200 250 300 Child Mortality (5q0 per 1000 live births) 350

Males

400

1000 900 Adult Mortality (45q15 per 1000) 800 700 600 500 400 300 200 100 0 0 50 100 150 200 250 300 Child Mortality (5q0 per 1000 live births) 350

Females

400

26

Table 6. Estimated 95% uncertainty ranges for total deaths in 2002, WHO Member States
Estimated total deaths (000) Lower uncertaintyUpper uncertainty bound bound 348 19 158 1 255 1 277 24 124 69 60 2 2 1030 2 138 101 1 69 15 49 31 34 1180 1 104 213 104 123 188 220 2 60 109 81 8712 231 4 33 0 19 49 75 7 101 217 116 797 57 7 1 696 28 188 1 435 1 287 34 129 72 70 2 3 1184 2 148 105 2 123 30 104 38 45 1262 1 112 337 174 219 330 225 3 99 192 87 9528 257 8 62 0 21 52 78 8 105 386 314 1294 57 13 1 Relative uncertainty (a) 36 20 9 7 29 8 2 19 2 2 7 1 11 7 1 4 2 8 31 36 38 10 14 3 2 4 25 29 30 30 1 24 26 28 4 4 5 40 32 1 5 4 2 12 2 33 49 25 0 34 1

Member State Afghanistan Albania Algeria Andorra Angola Antigua and Barbuda Argentina Armenia Australia Austria Azerbaijan Bahamas Bahrain Bangladesh Barbados Belarus Belgium Belize Benin Bhutan Bolivia Bosnia and Herzegovina Botswana Brazil Brunei Darussalam Bulgaria Burkina Faso Burundi Cambodia Cameroon Canada Cape Verde Central African Republic Chad Chile China Colombia Comoros Congo Cook Islands Costa Rica Croatia Cuba Cyprus Czech Republic Cte d'Ivoire Democratic People's Republic of Korea Democratic Republic of the Congo Denmark Djibouti Dominica

Total 485 22 173 1 307 1 281 26 127 70 64 2 2 1107 2 144 103 1 87 21 73 35 41 1225 1 107 250 120 160 236 222 2 76 149 84 9135 244 5 44 0 18 50 77 7 103 259 204 986 57 8 1

27

Dominican Republic Ecuador Egypt El Salvador Equatorial Guinea Eritrea

57 77 495 41 8 41

53 72 485 37 5 27

60 81 505 43 9 55

6 5 2 8 26 34

Table 6 (continued). Estimated 95% uncertainty ranges for total deaths in 2002, WHO Member States
Estimated total deaths (000) Lower uncertaintyUpper uncertainty bound bound 18 818 5 48 495 10 11 54 801 138 111 1 74 86 21 6 91 34 120 2 9979 1541 356 191 30 36 553 18 973 23 168 287 1 4 42 59 33 22 36 53 18 41 3 146 220 19 1464 6 49 517 21 22 70 830 289 116 1 88 151 34 8 158 51 125 2 10678 1711 412 237 32 38 588 21 1018 26 197 521 1 5 48 76 33 26 60 94 30 42 3 265 330 Relative uncertainty (a) 0 30 7 1 2 35 34 13 2 36 2 6 8 29 25 14 30 20 2 0 3 5 7 11 4 3 3 7 2 7 8 29 1 11 7 13 1 7 26 29 26 2 0 30 21

Member State Estonia Ethiopia Fiji Finland France Gabon Gambia Georgia Germany Ghana Greece Grenada Guatemala Guinea Guinea-Bissau Guyana Haiti Honduras Hungary Iceland India Indonesia Iran (Islamic Republic of) Iraq Ireland Israel Italy Jamaica Japan Jordan Kazakhstan Kenya Kiribati Kuwait Kyrgyzstan Lao Peoples Democratic Republic Latvia Lebanon Lesotho Liberia Libyan Arab Jamahiriya Lithuania Luxembourg Madagascar Malawi

Total 18 1060 5 49 499 15 16 61 815 209 114 1 82 114 27 7 112 42 122 2 10378 1626 384 213 31 35 571 20 973 23 184 407 1 5 45 67 33 24 46 69 23 41 3 201 258

28

Malaysia Maldives Mali Malta Marshall Islands Mauritania Mauritius Mexico Micronesia (Federated States of) Monaco Mongolia

119 2 243 3 0 40 8 470 1 0 19

115 2 191 3 0 31 8 442 1 0 19

122 2 315 3 1 55 8 489 1 0 20

3 6 26 4 8 30 2 5 13 3 2

Table 6 (continued). Estimated 95% uncertainty ranges for total deaths in 2002, WHO Member States
Estimated total deaths (000) Lower uncertaintyUpper uncertainty bound bound 139 293 344 22 0 222 136 27 24 177 1456 0 44 6 1247 0 13 41 25 159 421 344 92 1 266 47 247 1973 116 0 1 1 1 0 1 73 75 118 1 102 169 476 757 38 0 245 142 28 28 350 2528 0 46 11 1526 0 15 52 28 182 477 360 97 2 284 49 271 2757 190 0 1 1 1 0 2 128 145 124 1 202 Relative uncertainty (a) 10 24 40 28 24 5 2 3 9 35 27 2 2 29 10 1 6 11 6 7 6 2 2 5 3 3 5 16 28 1 1 1 6 2 35 28 35 3 1 38

Member State Morocco Mozambique Myanmar Namibia Nauru Nepal Netherlands New Zealand Nicaragua Niger Nigeria Niue Norway Oman Pakistan Palau Panama Papua New Guinea Paraguay Peru Philippines Poland Portugal Qatar Republic of Korea Republic of Moldova Romania Russian Federation Rwanda Saint Kitts and Nevis Saint Lucia Saint Vincent and the Grenadines Samoa San Marino Sao Tome and Principe Saudi Arabia Senegal Serbia and Montenegro Seychelles Sierra Leone

Total 154 385 520 28 0 233 139 27 26 245 2006 0 45 8 1386 0 14 47 27 170 448 352 94 1 275 48 259 2406 131 0 1 1 1 0 1 97 103 121 1 132

29

Singapore Slovakia Slovenia Solomon Islands Somalia South Africa Spain Sri Lanka Sudan Suriname Swaziland Sweden Switzerland Syrian Arab Republic Tajikistan Thailand

18 50 18 3 175 680 356 146 346 3 26 91 61 71 54 419

17 49 18 3 151 513 343 127 244 3 22 89 59 67 48 387

19 51 18 3 249 816 368 163 479 4 32 93 63 75 60 453

6 3 1 13 28 22 4 12 34 8 19 2 3 6 10 8

Table 6 (continued). Estimated 95% uncertainty ranges for total deaths in 2002, WHO Member States
Estimated total deaths (000) Lower uncertaintyUpper uncertainty bound bound 18 4 50 1 10 53 407 41 0 311 760 9 577 587 2372 30 165 1 109 487 125 190 270 20 10 85 1 11 59 464 43 0 470 809 10 623 661 2472 31 179 1 120 543 233 282 335 Relative uncertainty (a) 6 43 28 4 4 6 7 3 1 20 3 2 4 6 2 1 4 12 5 5 31 20 12

Member State The former Yugoslav Republic of Macedonia Timor-Leste Togo Tonga Trinidad and Tobago Tunisia Turkey Turkmenistan Tuvalu Uganda Ukraine United Arab Emirates United Kingdom United Republic of Tanzania United States of America Uruguay Uzbekistan Vanuatu Venezuela (Bolivarian Republic of) Viet Nam Yemen Zambia Zimbabwe

Total 19 7 62 1 12 56 437 42 0 388 783 9 599 596 2421 31 172 1 115 516 171 232 284

Based on this advice, for cause distributions derived from vital registration data coded using ICD10, it was generally estimated that diagnostic uncertainty and coding uncertainty together resulted in approximate relative 95% uncertainty ranges of 3% for Group I causes (communicable diseases, maternal and perinatal conditions and nutritional deficiencies), 7% for Group II causes (noncommunicable diseases), and 2% for Group III causes (injuries). Uncertainty ranges will be larger for some specific causes known to have greater levels of diagnostic or coding error, and for certain Member States using older ICD or other cause coding systems or verbal autopsy methods, or where
30

cause of death models were used to estimate death distributions across Groups I, II and III. In the latter case, an additional relative uncertainty for the estimation of Group I, II and III proportions would arise from the prediction uncertainty ranges associated with the CODMOD regression model. Additional uncertainty arises from the redistribution of general, cancer, cardiovascular, and injury illdefined cause codes and incomplete coverage of vital registration data. The analysis of uncertainty in cause-of-death (COD) estimates at country level thus combined quantitative country-specific information on uncertainty levels in all-cause mortality and, in some cases also in major cause group distributions, together with quantified average relative uncertainty ranges for specific cause atttributions based on expert advice and adjusted for specific causes and also for country-specific information on data sources, type of cause information available and indicators of data quality. For most causes, uncertainty ranges were greater than those of the all cause mortality estimates, since there is additional uncertainty associated with cause attribution, as described above. For example, the relative uncertainty ranges for ischaemic heart disease ranged from around 12% for high-income countries to (-24%, +34%) for Sub-Saharan Africa. While the uncertainty range for high-income countries may seem surprisingly large, it reflects not only uncertainty in overall mortality levels, but also uncertainty in the attribution of underlying cause, and in the attribution of causes coded to garbage codes or to the ICD chapter for 'symptoms, signs and ill-defined conditions'. The proportion of deaths coded to these two groups of causes is surprisingly large for some high-income countries (Mathers et al. 2005). Uncertainty arising from epidemiological estimates Uncertainty in YLD estimates is mainly determined by the uncertainty in (a) epidemiologic estimates for prevalence and/or incidence of disability associated with specific causes or cause groups, (b) disability weights arising from uncertainty in health state valuations and, in some cases, also in the disability severity distribution associated with a condition. The quantities of interest for the Global Burden of Disease Study are the underlying rates of incidence, remission and mortality for defined causes for whole populations for a specified time period, and the assessment of these often requires synthesis of multiple study data, or adjustment for biases in terms of population, age groups, or time period. A major source of uncertainty for the GBD estimates is the uncertainty associated with extrapolating from one or more subgroups to a regional population. For example, how representative of the incidence and prevalence patterns of dementia in sub-Saharan Africa are two or three population-representative studies for rural or urban populations in specific regions of specific countries? The uncertainty associated with extrapolating from a set of studies in sub-populations to the regional population is related to systematic biases, and is much more difficult to quantify than uncertainty associated with stochastic variation due to sample size or measurement error. For a subset of the GBD causes, analysts carrying out reviews and analyses for the estimation of YLD also undertook an estimation of levels of uncertainty in regional prevalences. These uncertainty levels took into account not only typical levels of measurement error in the input data sets but also semi-qualitative judegements on the uncertainty arising from the lack of representativeness of the available data for each region. The resulting uncertainty ranges varied considerably across causes, ranging from relatively certain estimates for some causes such as polio for which there are intensive surveillance systems in place, to highly uncertain estimates for other causes, such as osteoarthritis where for some regions, not a single useable dataset was found, and for others, the latest available
31

data was decades old. Table 4 summarizing numbers of data sources used for YLD estimates by cause and region provides one indication of the relative uncertainty associated with YLD estimates for different causes. For some causes, such as stroke and ischaemic heart disease, YLD estimates were essentially derived from estimates of cause-specific mortality, via models of regional variations in case fatality rates. In such cases, YLD uncertainty will be significantly higher than the uncertainty ranges associated with cause-specific mortality estimates, as there is also considerable uncertainty in case fatality rates for most low- and middle-income countries, and in models u to infer non-fatal sed disease burden from mortality. In general, we can conclude that YLD uncertainty will generally be greater than YLL uncertainty, and also varies across causes according to both the typical uncertainty associated with measurement of incidence or prevalence according to GBD case definitions, and also with the number and representativeness of available studies. For a subset of 16 major causes of YLD for which indicative uncertainty ranges were estimated by analysts carrying out revisions and updating of YLD estimates, the typical uncertainty ranges for regional prevalence estimates ranged from 10% to 90%, with a median value of 41%. Uncertainty ranges were generally higher for low- and middle-income countries than for high income countries.

6.

Discussion and conclusions

Uncertainty in the GBD prior estimates for countries must be taken into account when making crossnational comparisons, and needs to be carefully communicated and interpreted by epidemiologists and policy makers alike. Estimates of mortality in countries where there is no functioning vital registration system for causes of death will always be substantially more uncertain than those derived from systems where all deaths are registered and medically certified. The same may be said for the quantification of disability due to various conditions, where the gap in data availability between rich and poor countries is likely to be even more extreme than for mortality. The GBD project has attempted to maximize the use of available population-based data, and, even for regions and causes where data are sparse, to use the available evidence and the best available methods to make inferences. In excess of 770 country-years of death registration data and over 3,000 additional sources of information on levels of child and adult mortality and on specific causes of death were used to estimate global and regional patterns of mortality. Together with the more than 8,700 data sources used for the estimation of YLD, the GBD 2000-2002 has incorporated information from over 10,000 datasets relating to population health and mortality. This represents the largest synthesis of global information on population health carried out to date. Despite the perceptions of some critics that the Global Burden of Disease study is inadequately empirically based for some regions, and particularly in Africa (Cooper et al. 1998), it is notable that fully one third of the more than 10,000 data sources used by the GBD 2001 relate to sub-Saharan African populations, albeit with the serious limitations in the information available on mortality noted above. The fact that estimates are possible does not obviate the need to put higher priority on addressing the serious lack of information on levels of adult mortality and causes of death in some regions of the world, particularly Sub-Saharan Africa. Low- and middle-income countries can benefit from the advantages of death registration without fully implementing a system of complete population coverage and medical certification (Rao, Bradshaw, and Mathers 2004; Setel et al. 2005; Yang et al. 2005). There is also a massive lack of good population-based epidemiological data for developing regions, particularly for non-communicable diseases. Despite the fact that ischemic heart disease and stroke

32

are among the leading causes of burden of disease in most regions of the world, there are very few recent and reliable sources of information on the prevalence and severity distribution of chronic cardiovascular conditions and long-term disability following stroke. Similarly, there are very few population-based studies, even in high income countries, of the prevalence of chronic lung disease or musculoskeletal conditions. Cross-national surveys such as the WHO's World Health Survey, conducted during 2002 and 2003 in 73 countries, will fill some gaps in information for some chronic diseases and mental disorders (Ustun et al. 2003). However, there remain significant issues with the comparability of prevalence data derived from self-report questionnaires on symptoms for mental disorders, angina, and other chronic diseases, which will need to be addressed. With increasing pressure on resources for health in all countries, priority setting in the health sector will depend more and more on comprehensive, comparative information about the impact of diseases, injuries and risk factors on population health. To improve the usefulness of the Global Burden of Disease results for setting and monitoring global health priorities, a much more concerted effort is needed to obtain, uncover and critically assess data sets on the health of populations in all countries. This must be a key focus of future efforts to assess the burden of disease.

Acknowledgements
In producing this short working paper, I have drawn on documentation prepared by the many people involved in the Global Burden of Disease studies, and acknowledged this as far as possible through citations. I would particularly acknowledge the assistance of Doris Ma Fat and Mie Inoue, who compiled comprehensive summaries of mortality data sources (Mathers et al. 2003) which we have drawn on here, and on the assistance of Chalapati Rao, who carried out analyses of latest mortality data for a number of countries such as India, China, Iran and Turkey. I would also acknowledge considerable advise and assistance in the analysis of uncertainty from Joshua Salomon, Steven Begg and Niels Tomijima. Many people have contributed to the estimation of global burden of disease both inside and outside WHO. I wish to particularly acknowledge the contributions of staff in various WHO programs, and expert groups outside WHO, who have provided advice and collaborated in the reviews of epidemiological data and in the estimation of burden of disease. Current and former staff of the Evidence and Information for Policy cluster who worked directly on the GBD analyses include Omar Ahmad, Jose Ayuso, Prerna Banati, Stephen Begg, Christina Bernard, Cynthia Boschi-Pinto, Marisol Concha, Carmen Dolea, Majid Ezzati, Brodie Ferguson, Mie Inoue, Kim Moesgaard Iburg, Jeremy Lauer, Matilde Leonardi, Steve Lim, Alan Lopez, Doris Ma Fat, Rafael Lozano, Chris Murray, Sue Piccolo, Chalapati Rao, Tanuja Rastogi, Eduardo Sabat, Joshua Salomon, Toshi Satoh, Kenji Shibuya, Claudia Stein, Lana Tomaskovic, Niels Tomijima, Thomas Truelsen, Bedirhan Ustn, Marie -Claude von Rulach, Sarah Wild and Hongyi Xu.

References

Anand, S, W Ammar, T Evans, T Hasegawa, K Kissimova-Skarbek, A Langer, and et.al. 2003. "Report on the Scientific Peer Review Group on Health Systems Performance Assessment". In Health systems performance assessment: debates, methods and empricism, edited by Murray, C. J. L and D. Evans (Geneva: World Health Organisation).

33

Barendregt, J., G. J. van Oortmarssen, T. Vos, and Murray C.J.L. 2003. "A generic model for the assessment of disease epidemiology: the computational basis of DisMod II". Population Health Metrics 1:4. Cooper, R. S., B. Osotimehin, J. S. Kaufman, and T. Forrester. 1998. "Disease burden in subSaharan Africa: what should we conclude in the absence of data?". The Lancet 351, no. 9097:208-210. Ezzati, M, A Lopez, A Rodgers, and C. J. L Murray. 2004. Comparative Quantification of Health Risks: Global and Regional Burden of Disesae Attributable to Several Major Risk Factors. Geneva: WHO. Jamison, D. T., J. G. Breman, A. R. Measham, G. Alleyne, M. Claeson, D. Evans, P. Jha, A. Mills, and P. Musgrove. 2006. Disease control priorities in developing countries, second edition. Oxford: Oxford University Press. King, G., M. Tomz, and J. Wittenberg. 2000. "Making the most of statistical analyses: improving interpretation and presentation". American Journal of Political Science 44, no. 2:341355. Korenromp, E., B. G. Williams, E. Gouws, C. Dye, and R. W. Snow. 2003. "Measurement of trends in childhood malaria mortality in Africa: an assessment of progress toward targets based on verbal autopsy". Lancet Infectious Diseases 3, no. 6:349-358. Lopez, A. D., O. Ahmad, M. Guillot, B. Ferguson, J Salomon, Murray C.J.L, and K. H. Hill. 2002. World Mortality in 2000: life tables for 191 countries. Geneva: World Health Organization. Lopez, A. D., C. D. Mathers, M. Ezzati, Murray C.J.L., and D. T. Jamison. 2006. Global burden of disease and risk factors., Disease Control Priorities Project. Oxford: Oxford University Press. Mathers, C. D., C. Bernard, K. M. Iburg, M. Inoue, D. Ma Fat, K Shibuya, C. Stein, N. Tomijima, and H. Xu. 2003a. Global Burden of Disease in 2002: data sources, methods and results. GPE Discussion Paper No. 54. Geneva, World Health Organization. Mathers, C. D., A. D. Lopez, C. Stein, D. Ma Fat, C. Rao, M. Inoue, K Shibuya, N. Tomijima, C. Bernard, and H. Xu. 2003b. Deaths and disease burden by cause: global burden of disease estimates for 2001 by World Bank country groups. Washington., The World Health Organization (WHO), the World Bank, and the Fogarty International Center, U.S. National Institutes of Health (NIH). DCPP Working Papers Series No. 18, Second Project on Disease Control Priorities in Developing C ountries (DCPP). Retrieved 25 July, 2005 from http://www.fic.nih.gov/dcpp/wps.html. Mathers, C. D., D. Ma Fat, M. Inoue, C. Rao, and Lopez A.D. 2005. "Counting the dead and what they died from: an assessment of the global status of cause of death data". Bulletin of the World Health Organization, no. March.

34

Mathers, C. D., K. Shibuya, C. Boschi-Pinto, A. D. Lopez, and C. J. Murray. 2002. "Global and regional estimates of cancer mortality and incidence by site: I. Application of regional cancer survival model to estimate cancer mortality distribution by site". BMC Cancer 2, no. 1:36. Morgan, M. G. and M. Henrion. 1990. Uncertainty: a guide to dealing with uncertainty in quantitative risk and policy analysis. Cambridge: Cambridge University Press. Murray, C. J. L, B. Ferguson, A. D. Lopez, M. Guillot, J. A. Salomon, and O. Ahmad. 2001a. The modified logit lifetable system: principles, empirical validation, and application. Health Systems Performance Assessment Discussion Paper. Global Programme on Evidence for Health Policy. Geneva, WHO. Murray, C. J. L and A. D. Lopez. 1996a. "Evidence-based health policy -- lessons from the Global Burden of Disease Study". Science 274, no. 5288:740-743. Murray, C. J. L and A. D. Lopez. 1996b. Global health statistics. Vol. 2, Global Burden of Disease and Injury Series. Cambridge: Harvard University Press. Murray, C. J. L and A. D. Lopez. 1996c. The global burden of disease: a comprehensive assessment of mortality and disability from diseases, injuries and risk factors in 1990 and projected to 2020. Vol. 1, Global Burden of Disease and Injury Series. Cambridge: Harvard University Press. Murray, C. J. L, A. D. Lopez, C. D. Mathers, and C. Stein. 2001b. The Global Burden of Disease 2000 project: aims, methods and data sources. GPE Discussion Paper No. 36. Geneva, WHO. Murray, C. J. L, C. D. Mathers, and Salomon J.A. 2003. "Towards evidence-based public health". In Health systems performance assessment: debates, methods and empiricism, edited by Murray, C. J. L and D. Evans (Geneva: World Health Organization). Murray, C. J. L, A. Tandon, J. A. Salomon, and C. D. Mathers. 2002. "New approaches to enhance cross-population comparability of survey results". In Summary measures of population health: concepts, ethics, measurement and applications, edited by Murray, C. J. L, J. A. Salomon, C. D. Mathers, and A. D. Lopez (Geneva: World Health Organization). Rao, C., D. Bradshaw, and C. D. Mathers. 2004. "Improving death registration and statistics in developing countries: lessons from sub-Saharan Africa". Southern African Journal of Demography 9, no. 2. Rowe, A. K., R. W. Steketee, S. Y. Rowe, R. W. Snow, E. Korenromp, J. A. Schellenberg, C. Stein, B. Nahlen, J Bryce, and R. J. Black. 2005. "The burden of malaria mortality among African children in the year 2000". Submitted for publication. Sadana, R., C. D. Mathers, A. D. Lopez, C. J. L Murray, and K. Moesgaard-Iburg. 2002. "Comparative analyses of more than 50 household surveys on health status". In Summary measures of population health: concepts, ethics, measurement and applications, edited

35

by Murray, C. J. L, J. A. Salomon, C. D. Mathers, and A. D. Lopez (Geneva: World Health Organization). Retrieved 25 July, 2005 from www.who.int/pub/smph/en/index.html. Salomon, J. A., C. D. Mathers, C. J. L Murray, and B. Ferguson. 2001. Methods for life expectancy and healthy life expectancy uncertainty analysis. GPE Discussion Paper No. 10. Geneva, WHO. Available on the worldwide web at http://www.who.int/evidence. Salomon, J. A. and C. J. L. Murray. 2002. "The epidemiologic transition revisited: compositional models for causes of death by age and sex". Population and Development Review 28, no. 2:205-228. Setel, P. W., O. Sankoh, V. A. Velkoff, C. Rao, C. D. Mathers, Y. Gonghuan, Y. Hemed, P. Jha, and Lopez A.D. 2005. "Sample registration of vital events with verbal autopsy: innovative approaches to measuring and monitoring vital statistics". Bulletin of the World Health Organization. UNAIDS. 2004. 2004 report on the global AIDS epidemic. Geneva, UNAIDS. Retrieved 25 July, 2005 from http://www.unaids.org/epidemic_update/report/index.html. Ustun, T. B., J. L Ayuso-Mateos, S. Chatterji, C. D. Mathers, and C. J. L. Murray. 2004. "Global burden of depressive disorders in the year 2000". British Journal of Psychiatry 184:386392. Ustun, T. B., S. Chatterji, A. Mechbal, Murray C.J.L, and WHS Collaborating Groups. 2003. "The World Health Surveys". In Health systems performance assessment: debates, methods and empiricism, edited by Murray, C. J. L and D. Evans (Geneva: World Health Organisation). Vose, D. 2000. Risk analysis: a quantitative guide. New York, Wiley. Walker, N., K. A. Stanecki, T. Brown, J. Stover, S. Lazzari, J. M. Garcia-Calleja, B. Schwartlnder, and P. D Ghys. 2003. "Methods and procedures for estimating HIV/AIDS and its impact: the UNAIDS/WHO estimates for the end of 2001". AIDS 17, no. 15:22152225. Wild, S., G. Roglic, A. Green, R. Sicree, and H. King. 2004. "Global prevalence of diabetes: estimates for the year 2000 and projections for 2030". Diabetes Care 27, no. 5:10471053. World Bank. 1993. World Development Report 1993. Investing in health. New York: Oxford University Press for the World Bank. World Health Organization. 2002. World Health Report 2002. Reducing Risks, Promoting Healthy Life. Geneva: WHO. Retrieved 25 July, 2005 from www.who.int/whr. Yang, G. H., J. Hu, K. Q. Rao, J. Ma, C. Rao, and Lopez A.D. 2005. "Mortality registration and surveillance in China: History, current situation and challenges". Population Health Metrics 3:3.

36

37

Annex Table A-1: Regional reporting categories for GBD 2002 data sources.
Region East Asia and Pacific Countries included American Samoa, Cambodia, China, DPR Korea, Fiji, Indonesia, Kiribati, Lao People's Democratic Republic, Malaysia, Marshall Islands, Micronesia (Federated States Of), Mongolia, Myanmar, Palau, Papua New Guinea, Philippines, Samoa, Solomon Islands, Thailand, Timor-Leste, Tonga, Vanuatu, Viet Nam

Europe and Central Albania, Armenia, Azerbaijan, Belarus, Bosnia And Herzegovina, Bulgaria, Croatia, Czech Republic, Asia Estonia, Georgia, Hungary, Isle of Man, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Poland, Republic of Moldova, Romania, Russian Federation, Serbia and Montenegro, Slovakia, Tajikistan, The Former Yugoslav Republic Of Macedonia, Turkey, Turkmenistan, Ukraine, Uzbekistan High Income Andorra, Aruba, Australia, Austria, Bahamas, Bahrain, Belgium, Bermuda, Brunei Darussalam, Canada, Cayman Islands, Channel Islands, Cyprus, Denmark, Faeroe Islands, Finland, France, French Polynesia, Germany, Greece, Greenland, Guam, Iceland, Ireland, Israel, Italy, Japan, Kuwait, Liechtenstein, Luxembourg, Monaco, Netherlands, Netherlands Antilles, New Caledonia, New Zealand, Nothern Mariana Islands, Norway, Portugal, Qatar, Republic of Korea, San Marino, Singapore, Slovenia, Spain, Sweden, Switzerland, United Arab Emirates, United Kingdom, United States of America, United States Virgin Islands

Latin America and Antigua and Barbuda, Argentina, Barbados, Belize, Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, Caribbean Dominica, Dominican Republic, Ecuador, El Salvador, Grenada, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Puerto Rico, Saint Kitts And Nevis, Saint Lucia, Saint Vincent And The Grenadines, Suriname, Trinidad And Tobago, Uruguay, Venezuela Middle East and North Africa South Asia Sub-Saharan Africa Algeria, Djibouti, Egypt, Iran (Islamic Republic Of), Iraq, Jordan, Lebanon, Libyan Arab Jamahiriya, Malta, Morocco, Occupied Palestinian Territory, Oman, Saudi Arabia, Syrian Arab Republic, Tunisia, Yemen Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, Sri Lanka Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Cape Verde, Central African Republic, Chad, Comoros, Congo, Cte dIvoire, Democratic Republic of the Congo, Equatorial Guinea, Eritrea, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Namibia, Niger, Nigeria, Rwanda, Sao Tome and Principe, Senegal, Seychelles, Sierra Leone, Somalia, South Africa, Sudan, Swaziland, Togo, Uganda, United Republic of Tanzania, Zambia, Zimbabwe Anguilla, British Virgin Islands, Cook Islands, Falkland Islands, French Guiana, Gibraltar, Guadeloupe, Holy See, Martinique, Montserrat, Nauru, Niue, Pitcairn, Runion, Saint Helena, Saint Pierre et Miquelon, Tokelau, Turks and Caicos Islands, Tuvalu, Wallis and Futuna Islands, Western Sahara

Other

38