global burden of disease - pakistan presentation
Upload: institute-for-health-metrics-and-evaluation-university-of-washington
Post on 24-Dec-2014
1.120 views
DESCRIPTION
Outline of the Global Burden of Diseases, Injuries, and Risk Factors StudyTRANSCRIPT
UNIVERSITY OF WASHINGTON
The Global Burden of Diseases, Injuries, and Risk Factors Study
2
Presentation Outline
• Goal and key attributes
• Project structure and partners
• Mortality
• Causes of Death
• Systematic reviews
• Analysis of disease-specific data for calculating YLD
• Disability Weights measurement
• Future work
3
GBD Goal
• To produce new, robust, and reliable estimates of burden for all major diseases, injuries, and risks that are widely disseminated, understood, and easily used by policymakers, researchers, funders, and practitioners.
4
Key Attributes
• Producing specific DALY, YLL, and YLD estimates for over 300+ diseases/injuries and 40+ risk factors by age and sex for 21 regions for the years 1990, 2005, and 2010.
• Providing a consistent time trend (methods for current ‘00, ’02, ‘04 estimates are not comparable to ‘90).
• Providing first comprehensive revision of Disability weights since 1996. Many Burden estimates done after the original study had used ad hoc DW based on Dutch study.
• Providing improved analytical tools to facilitate Burden estimates and policy use.
5
Presentation Outline
• Goal and key attributes
• Project structure and partners
• Mortality
• Causes of Death
• Systematic reviews
• Analysis of disease-specific data for calculating YLD
• Disability Weights measurement
• Future work
Organizational Structure
Core TeamExternal
Advisory Board
COD Sub-Team Rafael Lozano
CRA Sub-TeamMajid Ezzati
DW Sub-TeamJosh Salomon
Mortality Sub-TeamChris Murray and
Alan Lopez
YLD Sub-TeamRafael Lozano
and Colin Mathers
Cluster A CVD, COPD, Cancer
Majid EzzatiHarvard University
Cluster B Child/Maternal
Bob BlackJohns Hopkins
University
Cluster C Injuries and Mental
HealthTheo Vos
University of Queensland
Cluster D Communicable
DiseasesNeff Walker
Johns Hopkins University
Cluster ENoncommunicable
DiseasesCatherine MichaudHarvard University
• Vision, decision-making, and leadership are handled by the core team, a group of 12 key individuals from the collaborating institutions.
• Specific analytical tasks are grouped into (1) Causes of Death, (2) Comparative Risk Assessment, (3) Disability Weights, and (4) Mortality Estimation. Each of these “subteams” are led by 1 or 2 members of the core team to guide each category’s specific scientific progress and analysis.
• Management of diseases, injuries and risks are organized into clusters and are led by 1 member of the core team who oversees the cluster’s expert groups.
• Expert groups are comprised of knowledgeable specialists of a disease, injury, or risk.
Organizational structure
8
Johns Hopkins University
Harvard University
University of Queensland
Institute for Health Metrics and Evaluation
World Health Organization
44 expert groups, with
over 800 members worldwide
Collaborating Partners
Expert Groups
CancersCardiovascular DiseasesChronic Respiratory DiseasesClimate ChangesIndoor Air PollutionMetabolic RisksNutritional RisksOutdoor Air PollutionPhysical InactivitySocioeconomic FactorsTobacco
ARI Meningitis SepsisChild NutritionCongenital and NeonatalDiarrheaMalariaMaternal ConditionsSelected Vaccine Preventable DiseasesWater, Sanitation, Hygiene
Alcohol UseCollective ViolenceIllicit Drug UseIntimate Partner and Sexual ViolenceLead ExposureMental DisordersMusculoskeletalNeurological DisordersOccupational RisksOther InjuriesRoad Traffic Accidents
HepatitisHIV/AIDSParasitic &Vector DiseasesSTIsTuberculosisUnsafe Sex
DentalDiabetesGastrointestinalGenitourinary DiseasesHearing LossHemoglobinopathiesSkin DiseasesVision Loss
Cluster C Injuries and Mental Health
Theo VosUniversity of Queensland
Cluster A CVD, COPD, Cancer
Majid EzzatiHarvard University
Cluster B Child/Maternal
Bob BlackJohns Hopkins University
Cluster ENoncommunicable Diseases
Catherine MichaudHarvard University
Cluster D Communicable Diseases
Neff WalkerJohns Hopkins University
44
10
Systematic reviews of
prevalence/ incidence
Peer Review
2nd Revision internal
consistency analysis by
disease
Internal consistency analysis by
disease
Causes of death
consistency analysis
Revised internal
consistency by disease
Systematic reviews of disabling sequelae
Peer ReviewRevision of disabling sequelae
Functional health status
Disability weights
Start
First Consultative Review Meeting
Identify diseases and disabling sequelae
Data input from1. Vital registration
deaths by age,sex, and cause
2. Disease registries3. Surveys- DHS, WHS
Data input fromAll cause mortality
by age and sex from vital registration data, survey data, and other sources
YLD
GBD Operational Components
Causes of death
YLL
Shapes Key Colors Key
Inputs Process Outputs Data input fromHealth state
measurement in representative samples
by sequelae
Expert Groups
Core Team
Peer Review
Expert Groups & Core Team
Expert Group Meetings
Estimation of
valuation function
11
Start
First Consultative Review Meeting
Identify diseases and disabling sequelae
GBD Operational Components – Comparative Risk Assessment
Shapes Key Colors Key
Inputs Process Outputs
Expert Groups
Core Team
Peer Review
Expert Groups & Core Team
Expert Group Meetings
Revised exposure
data
Revision ofdisease end-
points andhazard
magnitude
Peer Review
Peer Reviewof
hazardous effects
Exposure metric &
assessment of systematic
bias
Systematic reviews of exposure data and
hazardous effects
Data input from1. Health
examination surveys2. Risk factor indicator
registries (FAO, WHO, etc)
PAF(Individual& multiple
risks)
Addictive substances
• Tobacco use
• Alcohol use
• Illicit drug use
Environmental
• Unsafe water, sanitation, and hygiene
• Urban ambient air pollution
• Household air pollution from solid fuel use
• Lead exposure
• Passive smoking / Environmental tobacco smoke
• Food contamination
• Road and vehicle safety
Violence related
• Sexual violence
• Intimate partner violence
• Collective violence
• Possession of firearms
Undernutrition (child and maternal)
• Folic acid deficiency
• Anaemia and/or iron deficiency
• Small-for-gestational age
• Growth retardation
• Suboptimal breasfeeding
• Vitamin A deficiency
• Zinc deficiency
Reproductive and sexual
• Unsafe sex
• Unwanted pregnancies
Risks related to medical practice
Genetic
Systemic
• Global climate change
• Socioeconomic factors
Other selected risks to health
• Osteoporosis
12
Risk Factors Occupational
• Risks for injuries
• Carcinogens
• Airborne particulates
• Ergonomic stressors
• Noise
• Pesticides
• Other
Metabolic, nutritional and lifestyle
• High blood pressure
• High cholesterol
• High blood glucose
• Dietary fats
• High BMI
• Low intake of fruit and vegetable
• Physical inactivity
• Other nutritional
13
21 GBD Regions
14
Age Groups for Results
• <1 month
• 1 – 11 months
• 1 – 4 years
• 5 – 9 years
• 10 – 14 years
• 15 – 19 years
• 20 – 24 years
• 25 – 34 years
• 35 – 44 years
• 45 – 54 years
• 55 – 64 years
• 65 – 74 years
• 75 – 84 years
• 85+ years
15
Presentation Outline
• Goal and key attributes
• Project structure and partners
• Mortality
• Causes of Death
• Systematic reviews
• Analysis of disease-specific data for calculating YLD
• Disability weights measurement
• Future work
16
• Generating regional estimates for age and sex
• All cause mortality based on demographic sources recording the event of death.
• The sum of all cause specific deaths for any age-sex group must equal, and not exceed, the overall mortality envelope for that age-sex group.
Mortality
17
Mortality Estimation
18
Mortality estimation: Synthesis
Gaussian Process Regression:
•Synthesizes discordant time series of mortality estimates into a best estimate of smooth trend
• Assigns probabilities to different functions according to how likely they are to be the true function
• Uses prior beliefs, the data and the uncertainty in the data to inform those probabilities
19
Child Mortality in Nicaragua: Example of Using VR and Survey and Census Data
20
Child Death Numbers: East Sub-Saharan Africa
21
Child Death Numbers: Global
22
GPR for Adult mortality: Nicaragua
23
GPR for Adults: Zimbabwe
24
Using 5q0 and 45q15 to a Complete Lifetable
WHO uses modified logit lifetable system (Murray et al 2003) to generate complete lifetables.
HIV mortality is modeled separately from demographic sources and added on after demographic estimation.
We wanted to improve the performance of model life table systems and avoid modeling strategies that are not empirically based.
Over 18 months, developed a semi-parametric approach to the modified logit lifetable system that improves performance and captures the empirical impact of HIV.
25
Key Attributes of Mortmatch
Based on 5q0, 45q15 and HIV sero-prevalence searches database of (nearly 8000 lifetables) for nearest matches using Mahalanobis distance.
Matched lifetables used to establish standard life table.
Modified logit transformation including bend factors used to estimate full survivorship curve based on matched standard.
26
Validation Results-1
01
2
0 1 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95100
excludes outside values
Entry parameters from HMD: Males only
relative error in lx: MORTMatch relative error in lx: modmatch
27
Presentation Outline
• Goal and key attributes
• Project structure and partners
• Mortality
• Causes of Death
• Systematic reviews
• Analysis of disease-specific data for calculating YLD
• Disability Weights measurement
• Future work
Sources: Gathering COD Data
Types of sources
o Verbal Autopsies
o Household Surveys
o Hospital Records
o Sentinel Registration
o Demographic Surveillance Systems
o Sample Registration Systems
o Vital Registration with Certification of Cause of Death
Data providers
• WHO mortality database (Geneva)
• PAHO, EMRO, WPRO mortality databases
• National Ministries of Health
• Networks: INDEPTH, Matlab, India, etc.
• Researchers
• Literature Review
Almost all information related with causes of death is useful.
29
> 5,000 country-years observed
• VR data don’t get more than 100 countries per year• VA and maternal deaths added important value
Data for more of 180 countries and territories
31
Almost 1 billion deaths from 1950 to 2008, only 34% of total expected
Sub_Saharan_Africa_West
Oceania
Latin_America_Andean
Asia_South
Sub_Saharan_Africa_Southern
Caribbean
Asia_Central
Latin_America_Southern
North_Africa_Middle_East
Asia_East
Australasia
Latin_America_Tropical
Asia_Southeast
Asia_Pacific_High_Income
Europe_Central
Europe_Eastern
Latin_America_Central
North_America_High_Income
Europe_Western
1,000 10,000 100,000 1,000,000 10,000,000 100,000,000
39,085
463,434
5,754,853
5,785,132
5,933,272
6,163,084
11,626,166
15,112,752
16,329,319
18,095,408
18,738,326
36,770,945
41,879,227
60,659,801
66,331,482
76,332,493
77,616,744
124,404,521
285,614,616
%Developed Countries 55Eastern & Central Europe 18Latin-America and Carb 16Asia & Oceania 7.5 Middle Eats & North Afr 1.9S.S.A. 0.7
32
VA Literature Search Process
Search “Verbal Autopsy”
• 971 studies from Google Scholar• 320 studies from PubMed• 687 studies from Google-country search• 1978 total articles culled
Check for duplicates• 473 duplicate studies between Google Scholar and
PubMed
Screen studies using criteria
• 818 total studies screened from Google Scholar• 234 total studies screened from Google• 5 reviewers for studies from PubMed• 3 reviewers for studies from Google Scholar• 2 reviewers for studies from Google search
Extract data from selected studies • 172 studies from which data was extracted
33
VA Literature Screening Criteria
• Four criteria:
1. Population based study
2. Using verbal autopsy method
3. Open to any age group
4. Open to any set of causes
34
All Child Maternal 0
10
20
30
40
50
60
70
80
Types of Studies in the Cause of Death Verbal Autopsy Database
Study Category
Nu
mb
er o
f S
tud
ies
35
1.3 million deaths from VA studies by GBD region and source
Europe Western
Europe Eastern
Caribbean
Asia Central
Oceania
LA Central
LA Southern
SSA Central
Asia East
SSA Southern
NA Miiddle East
Asia Southast
SSA West
SSA East
Asia South
10 100 1,000 10,000 100,000 1,000,000
Turkey
Andra P
Cherg
India DHS
INDEPTH
Tanzania Nat VA
Matlab
India SRS
India SCD
Sist Rew
1,000 10,000 100,000 1,000,000
36
1.3 million deaths from VA studies
37
Assess •Evaluate basic attributes: ICD format, coverage, age groupings, administrative regions, etc.•Check for consistency: totals/subtotals, clumped data, double counting
Correct •Correct restriction violations, redistribute unknown age and unknown sex deaths•Calculate remainder codes to reconcile totals and subtotals
Mapping •Based in list of 39 causes for adults and 26 for children under five•Comparable among countries and across ICD revision (ICD 1st to ICD 10th)•Also map to GBD Cause list 317 causes (280 are CoD)
Disaggregate •Disaggregate tabulated codes•Break up large age groups
Redistribute •Identifying Garbage Codes and Targets•Redistribute garbage codes
Cleaning: Preparing VR Data for Analysis
38
Name of List and Number of Causes
* Causes of death
317(290)*
GBD 2005 Cause List (ICD 10 4 digit)
3
24
39
CodMod I
CodMod II
CodMod B
• ICD 1,2,3,4,5,6,7,8,9,10
• ICD 9 BTL, ICD 10 Tab A
• China (ICD 9 and 10)
• Russia (ICD 9 and 10)
39
Mapping GBD Cause List with ICD Revisions and Other Tabulated List
BTL1 2 3 4 5 Tab B 6,7 Tab A 89 tab
9 VA 10Tab
10
GBD 2005 Cause List (317 )
GBD 1990 Cause List (100)
CODMOD level 2 (24)
CODMOD level B(39)
ICD and other formats1900 2000
CODMOD II
3
24
CODMOD I
Intentional injuries C24
Unintentional injuriesC23
Congenital anomaliesB22
Genitourinary diseases---Skin diseases----Musculoskeletal diseasesB21
Digestive diseases---Oral conditionsB20
Respiratory diseasesB19
Cardiovascular and circulatory diseasesB18
Mental and behavioral disorders--- Neurological conditions-- Sense organ diseasesB17
Endocrine, nutritional, blood and immune disorders B16
Diabetes mellitusB15
Malignant neoplasm and B. Other neoplasmB14
Small poxA13
Nutritional deficienciesA12
Perinatal and infant causesA11
Maternal conditionsA10
Respiratory infectionsA9
Meningitis and encephalitis and Hepatitis and Other infectious diseasesA8
Parasitic and vector diseasesA7
MalariaA6
Selected Vaccine Preventable Childhood DiseasesA5
Intestinal infectious diseasesA4
STDs excluding HIVA3
HIV/AIDSA2
TuberculosisA1
This level can be presented with 24 causes and subgroups and 3 big groups
CODMOD II
CODMOD B
3
24
39
CODMOD I
CODMOD II
CODMOD B
24.3War and civil conflict and Legally sanctioned deaths
24.2 Interpersonal violence
24.1 Self-inflicted injuries
23.8 Accidental exposure to other and unspecified factors 23.7 Accidental poisoning by and exposure to noxious substances (acute or chronic)
23.6 Exposure to smoke, fire and flaes, contact with heat and hot substances
23.5 Accidental drowning and submersion
23.3 Falls
23.1Transport Injures
20.2 Other digestive diseases
20.1 Cirrhosis of the liver
18.5 Other circulatory diseases
18.4 Cerebrovascular disease
18.2 Ischaemic heart disease
14.9 Other malignant and benign neoplasm
14.6 Cervix and Corpus uteri cancer
14.5 Breast cancer
14.4 Larynx , Trachea, bronchus and lung cancers
14.2 Stomach cancer
14.1 Esophagus cancer
8.3 Other infectious diseases
8.1 Meningitis and encephalitis
From the 24 causes we are dividing Malignant Neoplasm, CVD and Injuries
42
Evolution of Garbage Codes in GBD Studies
• 1990: ill defined; heart failure and atherosclerosis; cancer without defined site and injuries ill defined
• 2000: same codes of 1990 with better methods of redistribution
• 2005: Completely different approach, based on new concepts and methodso More garbage codes and more targetso Sequences for redistributiono Methods of redistribution
Distribution of Garbage Codes by Type and Region
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
SSA Asia LA Europe C&E ALL Europe W Caribbean N.America Australasia
SpecialsImmediateSequelaeIntermediateI&D UNSCancerIll Def
% o
f G
C
• ~20% total deaths from VR are GCs
• 10 causes accumulate 75%
• Intermediate causes are the most important Garbage Codes
Causes ICD 10 %
Ill-defined R00-R99 26.0
Heart Failure I50 18.0
Renal Failure N18 6.4
Atherosclerosis I70 6.0Malignant neoplasm without specification of site C80 4.8
Septicaemia A41 4.2
Essential (primary) hypertension I10 3.0
Exposure to Unspecified factor X59 2.7
Pulmonary embolism I26 2.2
Respiratory Failure J96 2.0
Percent of deaths with garbage codesSelect Countries of the Americas, circa 2005
45
Causes of Death Modeling Strategy Challenges• Dependent variable: age-specific rates or age-specific cause-
specific mortality fractions.
• Model each cause as a function of critical covariates available for most countries/sites: GDP, education, tobacco consumption, HIV sero-prevalence, TFR, DTP coverage, SBA, water and sanitation, war, disasters ….
• Covariates only explain 30-40% of the variance depending on cause.
• Sparse data for some developing regions
• Compositional bias, data in each time period reflects a changing set of countries/sites
• Small numbers – VA studies and small countries have huge sampling and non-sampling variation
46
Causes of Death Modeling Strategy – 3 Steps• Step 1 – run outlier resistant models using basic covariates
including negative binomial regression, quantile (L1) regression.
• Evaluate residuals – drop outliers using Box-plot methods, assess correlations over space and time in residuals using heatmaps.
• Use local regression methods (two-dimensional Loess) to model residuals. Space dimension relatedness is based on the observed correlation structure in the heatmaps.
47
48
Using VR data for 2005
49
50
51
Validation of Models
• Many variants possible at each stage. How to choose most valid predictive models and how to pool results across a range of models.
• Three tests of predictive validity:
1) Exclude 20% of country-years at random and predict for them out of sample
2) Exclude last 10 years of sequence for all countries and predict them out of sample
3) Exclude 20% of countries and predict them entirely out of sample.
52
Presentation Outline
• Goal and key attributes
• Project structure and partners
• Mortality
• Causes of Death
• Systematic reviews
• Analysis of disease-specific data for calculating YLD
• Disability weights measurement
• Future work
Systematic Reviews
• Objective:o To evaluate and interpret all available research evidence relevant
to a particular condition
• To date we have: o Recruited over 800 experts worldwide
o Worked with experts and Core team to revise the cause list
o Begun processing epidemiological reviews from experts
• Next steps:o Upcoming expert group meeting May 2010
o Complete systematic epidemiological reviews
o Peer review
53
54
Cardiomyopathy Epi Review ProcessInitial search of primary databases (n=28,382)
Duplicates removed
Eligibility screening (n=301)Articles evaluated by 2 or more
abstractors and excluded by criteria on closer review (n=174)
Data abstractionRegional experts review
articles and collect data on standardized collection sheets.
Adjustments for bias and missing data is reviewed.
(n=pending)Incidence
Mortality
Case fatality
Prevalence
Inclusion criteria :1. diagnostic methods2. ICD coding3. epidemiological factors,4. population-based demographics
55
Cardiomyopathy Epi Review Data
56
Presentation Outline
• Goal and key attributes
• Project structure and partners
• Mortality
• Causes of Death
• Systematic reviews
• Analysis of disease-specific data for calculating YLD
• Disability weights measurement
• Future work
57
Analysis of Disease-Specific Data for YLD
YLD = Disability Weight x Incidence x Duration
The GBD links losses of health to disease and injury causes through the concepts of cases and sequelae.
For incident cases of a given disease or injury in the population, there will be a distribution of current and future health states in the population, and the GBD maps this distribution of health states to a small set of discrete entities for which epidemiological estimates and YLD calculations are made.
Case definitions are based upon expert group guidance
58
Cardiomyopathy Disease Model
59
• DisMod III uses a compartmental model of disease progression to infer consistent epidemiological parameters from sparse and noisy data.
Generic Model of Disease
DisMod III
StatesS: healthy (susceptible)C: diseased (condition of interest) D: dead from the diseaseM: dead from all other causes
Transition ratesi: incidencer: remissionƒ: case fatalitym: all other mortality
60
DisMod III Analysis: CardiomyopathyCardiomyopathy for males in Asia Pacific High Income region in 2005
61
DisMod III Analysis: CardiomyopathyCardiomyopathy for males in Asia Pacific High Income region in 2005
62
Cardiomyopathy Prevalence
63
Presentation Outline
• Goal and key attributes
• Project structure and partners
• Mortality
• Causes of Death
• Systematic reviews
• Analysis of disease-specific data for calculating YLD
• Disability Weights measurement
• Future work
64
• Objectives:o Derive disability weights for ~250 sequelae, which capture the
major health consequences of all of the causes in the GBD Study
o Address criticisms of previous approaches:─ Focus on valuations from community respondents in a “Disability Weights
Measurement Survey”
─ Use of techniques that are well-matched to the intended measurement construct
o Provide transparent, standardized and replicable approach that will easily accommodate additions or amendments
Disability Weights Measurement
65
Disability Weights Measurement
• Disability weights provide the bridge between mortality and non-fatal outcomes in disability adjusted life years (DALYs)
• Disability weights quantify overall health levels associated with different states, on a continuum between perfect health (which has a value of 0) and death (which has a value of 1)o Construct reflects decrements from perfect health, distinct from broader
notions of well-being or social value
o Must be measured on meaningful cardinal scale
66
Disability Weights Measurement
• Survey componentso Community surveys in 6 sites (Tanzania, Indonesia, Bangladesh, Peru,
South Africa, United States), focusing on random paired comparison and time trade-off questions for 108 sequelae
o Open access Web-based surveys including all sequelae, and paired comparison, time trade-off and population equivalence questions
o Community surveys are using computer-assisted personal interview approach with laptops
Household interview in Pemba, TZ 10/23/2009
67
Presentation Outline
• Goal and key attributes
• Project structure and partners
• Mortality
• Causes of Death
• Systematic reviews
• Analysis of disease-specific data for calculating YLD
• Disability weights measurement
• Future work