Embed Size (px)
Citation preview
Burden of DiseaseEstimation
Jae-Hyun ParkSungkyunkwan University
School of medicine2015.11
1
Contents
1. Measuring Health status2. Summary Measure of Public Health3. Estimation of DALY4. New approach and IHME5. Results of GBD 20106. Burden of Disease in Korea
2
Learning objectives
• Understanding Concept of Burden of disease and DALY
• Understanding New approaches of DALY estimation (GBD 2010)
• Understanding the results of Korea and making use of it
Keywords: Burden of Disease, Global Burden of Disease, DALY
3
Reliable health data and statistics are the foundation of health policies, strategies, and
evaluation and monitoring…….
Evidence is also the foundation for sound health information for the general public.
Margaret Chan 2007
If you are going to work, work on something important
William Foege, 2006
4
Defining Health
• “A state of complete physical, mental, and social well-being and not merely the absence of disease or infirmity”
WHO Charter, 1948
5
Concept of Physical, Mental, Social
PHYSICAL
MENTAL
SOCIAL
0
1
1
10 = DEATH1 = FULL LIFE
Metaphor of Life
(1,1,1)
6
Different Perspectives of Measuring Health
PATIENTS
PAYERSPROVIDERS
MANAGERS
“A Case of Fever That Requires Admission in a Ward”
What is my Quality of Life?
What is the Temperature? “How Much Did it Cost?
“How Long Did the Patient Stay? (LoS)”
7
Indexes that measure health states
• Mortality– Crude mortality rate– Mortality proportion due to specific diseases– Infant mortality rate
• Morbidity– Incidence– Prevalence– Case fatality
• Quality of life measure– QALY
8
The Epidemiologic Transition
• Underlying reasons for the demographic transition– Change in disease pattern
• Reduction in malnutrition and communicable diseases
9
“ONE MEASURE TO RULE THEM ALL”
10
QALY(Quality adjusted life year)
11
SMPH(Summary Measure of Public Health)
An integrative measure of population health that combines both mortality and morbidity data
to represent overall population health
In a single number(IOM, 1998).
12
SMPH & DALY
SMPHDALY
Quantityof Health
Mortality
(Morbidity)
Demography
Life expectancyYears lived
Qualityof Health
HRQL
(Morbidity)
Preference weighting(Disability Weight D)
SMPHDALY
Quantityof Health
Mortality
(Morbidity)
Demography
Life expectancyYears lived
Qualityof Health
HRQL
(Morbidity)
Preference weighting(Disability Weight D)
13
SMPH(Summary Measure of Public Health)
Partial Measure: Morbidity Partial Measure: Mortality
Population morbidity, disability,health-related quality of life
Average life expectancy or years lived
Summary Measures of Population Health
DALY (Disability-adjusted life years)HeaLY (Health-related life years)DFLE (Disability-free life expectancy)DALE (Disability-adjusted life expectancy)
14
Health Expectancy= A + f(B)= full health + f(B)
Health Gaps= C + g(B)= mortality gap + g(B)
Health Expectancies VS Health Gaps
15
Composite Measures of Population Health
• Health Expectancy=A+f(B)– Disability-free Life Expectancy
(DFLE)– Health Adjusted Life Expectancy
(HALE).
• Health Gap (Healthy Life Lost)=C+g(B)– Healthy Life Years (HeaLY)– Disability Adjusted Life
Year(DALY)
A
B
C
AGE%
Sur
vivi
ng
Health expectancy 1:Disability-free life expectancy (DFLE)
0
10
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50 60 70 80 90 100
Age (years)
Survivors (%)
Weight = 1
0
0
0
17
Health expectancy 2:Disability-adjusted life expectancy (DALE)
0
10
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50 60 70 80 90 100
Age (years)
Survivors (%)
Weight = 1
w2
w3
w4
DALE0 = HE1.0 + w2*HE2.0 + w3*HE3.0 + w4*HE4.0 (wh : weight for state h) 18
Disability Adjusted Life Expectancy at birth, Males 1999
(years)<3535 - 39.940 - 44.945 - 54.955 - 64.9>= 65No Data
19
Health gapDALY (Disability-adjusted life years)
DALY = YLL(years of life lost) + YLD(years lived with disability)
disability
deathextent of disability (De)
birth age at onset (Ao)
age at death (Af)
duration of disability (Dt)
expectation of life at age of onset (E(Ao)) 20
Concept of DALY
disability
death
birth onset death Expectation of life
Cirrhosis
deathIntermittent disability
onset death Expectation of life
birth
Epilepsy
birth Expectation of life
disability
Polio
onset 21
YEARS of LIFE LOST (YLL)
Natural Death at Age 86 years
He Dies at the Age of 50 Years Due to Road Traffic Injury
This person has therefore lost 86-50 = 36 YEARS of HEALTHY LIFE
22
Concept: YLL for Populations
Number of Deaths in the Population for An Age Group
Multiplied By
Life Expectancy At Birth for that Age (~ 86 – Age in Years)
23
Years of Life Lost (YLL)
• YLL = N x Lx
YLL=Years of life lost to premature mortality
– N=Number of deaths in the population– Lx =Standard life expectancy at age of death– X=Age of Death
• Example:– 10 deaths at 50 = 10 x Lx=10 x 34=340 YLL
24
YEARS Lived with Disability (YLD)
Natural Death at Age 86 years
He suffers RTI at the Age of 50 Years and LIVES with 50% Disability
This person has therefore lost (86-50) * 0.5 = 18 YEARS of HEALTHY LIFE
25
Years Lived with Disability (YLD)
• YLD = I x DW x d
– YLD=Years of life lived with disability
– I = Number of incident cases in the population– DW = Disability Weight
• Scale 0 (perfect health) to 1 (death)– d = Duration of disability (years)
• 10 cases of mental retardation due to lead at birth:– 10 x 0.36 x 80 years = 288 YLD
26
Value Choices for the DALY
• Time discounting: 3%– Falling mortality– Increasing costs
• Age weighting– non uniform weights– less weight to years lived at
younger and older ages
• Disability weights– Largely based on GBD 1990
study with some revisions.– For local prioritization, may
adjust to suit cultural preferences
AGE
% S
urvi
ving
Global Burden of Disease StudyMurray and Lopez, 1996
• Quantified Health effects for 107 diseases and injuries in 8 regions in 1990
• Comprehensive and consistent estimates of morbidity and mortality by age, sex, and region
• Introduced the DALY– YLL from premature
death and years lived in less than full health
Chris Murray (L) and Alan Lopez (right) considered to be the “grand fathers” of the Global Burden of Disease Approach
28
Global Burden of Disease Goals
• Measure loss of health due to comprehensive set of disease injury and risk factor causes in a comparable way
• Decouple epidemiological assessment from advocacy
• Inject non-fatal health outcomes into health policy debate
• Use a common metric for burden of disease assessment using summary measure for population health and cost-effectiveness analysis
WHO Global Burden of Disease 2004 Report
29
GBD Philosophy• Quantities of interest are total events or states at population
levels
• Best available data used to make estimates
• Corrections for major known biases to improve cross-population compatibility
• Comprehensive set of disease and injury causes– nothing is left out in principle
• No blanks in the tables, only wider uncertainty intervals
• Internal consistency used as a tool to improve validityWHO Global Burden of Disease 2004 Report
30
GBD 2004 Update (2008)
• YLL update by age, sex, and cause for 192 states
• YLD estimates for 52 causes
• UNAIDS, UNICEF, RBM, IARC, WHO surveillance
• Addition of “refractory errors”
• Revision of “angina pectoris”and CVA estimates
31
Global Cause of Death by Category
• Group I– Communicable plus
maternal, perinatal and nutritional conditions
• Group II– Non-communicable
conditions (eg, heart disease, stroke, cancer)
• Group III– Injuries including motor
vehicle accidents, homicide, and suicide
Group I
Group II
GroupIII
58.8 million deaths, 2004
Murray and Chen, 1995
Leading Causes of GBD, 20042030
WHO Global Burden of Disease 2004 Report33
Global Burden of Disease: New Approach
• Now Seven Organizations:– IHME– Imperial College– University of Queensland– Johns Hopkins– University of Tokyo– WHO
• Data from 187 countries, 291 conditions, 1160 sequela, 220 disease states
• Synthesis of Epidemiological Data
34
• Institute for Health Metrics and Evaluation, University of Washington
• Providing independent, rigorous, and scientificmeasurement and evaluations
• “Our goal is to improve the health of the world’s populations by providing the best information on population health”
• Core funding by the Bill & Melinda Gates Foundation and the State of Washington
• Created in 2007• 80 researchers, 60 staff
35
<#>
IHME improved IT infrastructure
<#>
GBD 2010 Previous methodPrevalence * DW
“True” systematic reviews and synthesis of all available data
Consistency check between disease parameters
Adjustments for comorbidity
Uncertainty quantified
DWs: paired comparisons; population surveys
Incidence * duration * DW
Choice of single data set for a given population/time
Consistency check between disease parameters
Comorbidity ignored
No uncertainty
DWs: panel of health experts; person trade off
New approach
38
38
Prevalence
Disability weight surveys
DWs
Severity distribution
YLDsSystematic review
DisMod-MR
Covariates:‒ Study characteristics
• Definition• Study type• Representative?
‒ Country characteristics.• GDP• Access to health servic
es• Conflict
‒ Adjustment data points‒ Pooling info‒ Predicting “gaps” ‒ Consistency between pa
rameters
Analytical steps
39
GBD Data and Model Flow Chart
40
GBD – it’s big data
• 187 countries• 1990, 2005 and 2010• 291 causes / 1160 specific out
comes• 66 risk factors• 20 age groups• Male/female/total• 5 key metrics: deaths, YLLs, pr
evalence, YLDs, DALYs
• Surveys
• Censuses
• Vital registration
• Disease registries
• Hospital records
• Surveillance systems
• Mortuaries / burial sites
• Police records
• Literature reviews
41
42
43
44
The Global Health Data Exchange (GHDx.org)
45
Challenges of YLD estimation
Data sources
Uncertainty
• No single source of data for YLDs from all conditions
• Inconsistency and gaps in information
• Uncertainty from data itself, lack of data, disability weights
Process specifications
• Complex disease epidemiology
• Severity distributions of health states
• Comorbidity
46
Data adjustments
Data issue Adjustment
Inconsistent case definition
Measurement instrument bias
Non-representative population bias
Incompleteness
Selection bias
Outlier studies
Correct for at-risk population
Downweight
Adjust upwards
Crosswalk
47
Methods
• DisMod-MR
• Natural history models
• Geospatial models
• Back-calculation models
• Registration completeness models
48
SusceptibleS(t)
CasesC(t)
Death from general mortalities, M(t)
Case specific deathD(t)
i : incidence rate
r : remissionrate
m
m
f
Dismod modeling
49
DisMod-MR• Bayesian Disease Modeling Meta-Regression tool
• Negative binomial statistical model
• Performs crosswalks to adjust for methodological variation
• Incorporates assumptions to inform the model
• Borrows strength using covariates and super-region, region, and country random effects to inform regions/countries with little or no data
• Forces consistency among disease parameters
50
Is negative-binomial distribution the best choice?
DisMod-MR
51
DisMod-MRBayesian meta-regression
52
52
Three estimation strategies with DisMod-MR
Direct estimation of disease sequelae
Maternal sepsis
Disability envelopes for etiological attribution
Otitis media Congenital Meningitis Other causes
Hearing loss
Disability envelopes for disease sequelae Diabetes mellitus
Diabetic neuropathy
Diabetic foot ulcer
Diabetic amputation
Uncomplicated diabetes
Diabetic retinopathy
53
DisMod-MR output
• Epidemiological parameters estimated for:– 187 countries– Years 1990, 2005, 2010– Single-year age groups– Both sexes
• Estimates repeated 1,000 times to define uncertainty
Need to build in reality of comorbidity
54
Comorbidity adjustment
1 Simulate comorbidity distribution
• Use prevalence and disability weights across hypothetical 20,000 people in each demographic group
2 Calculate combined disability weights (CDW)𝐷𝐷𝐷𝐷𝑛𝑛𝑛𝑛𝑛𝑛,𝑖𝑖 = 1 − 1 − 𝐷𝐷𝐷𝐷1,𝑖𝑖 ∗ 1 − 𝐷𝐷𝐷𝐷2,𝑖𝑖 ∗ ⋯∗ 1 − 𝐷𝐷𝐷𝐷𝑛𝑛,𝑖𝑖
where n = number of health states observed for individual i
3 Reaggregate by disease sequela
• Apportion CDWs to each of the contributing sequelae in proportion to the DW of a sequela on its own
4 Quantify uncertainty
• Repeat 1,000 times to estimate uncertainty
Comorbidity-adjusted YLDs with uncertainty55
Intervention 2:Prevent 1y of deafn
ess for
2000individuals
Intervention 1:Extend life by 1y in
1000healthy individuals
Disability weights in the 1996 GBD revision
Person trade-off: which would you choose?
• Expert panel used ‘person trade-off’ to assign values to 22 indicator conditions
56
1
Disability weights
2 3 4 5 6 7
Class 1:• Vitiligo on face
Class 4:• Below-knee amputation• Deafness
Class 7:• Active psychosis• Quadriplegia
Disability weights in the 1996 GBD revision
• Expert panel used ‘person trade-off’ to assign values to 22 indicator conditions
• These 22 conditions used as operational definitions of 7 disability classes
57
1
Disability weights
2 3 4 5 6 7
Rheumatoid arthritis cases
Average disability weight=0.2*0.07 + 0.4*0.18 + 0.4*0.30=0.21
Disability weights in the 1996 GBD revision
• Expert panel used ‘person trade-off’ to assign values to 22 indicator conditions
• These 22 conditions used as operational definitions of 7 disability classes
• Remaining conditions allocated across classes to compute average weights
58
59
Disability weights measurement study goals
• Derive weights for all 220 health states capturing nonfatal outcomes from 291 disease and injury causes in GBD 2010
• Address criticisms of previous approaches by:– Focusing on valuations from community respondents…– … in a diverse range of settings– … using suitable measurement methods
• Specific research aims– Develop valid and reliable data collection tools for population-bas
ed surveys– Empirical examination of variation in weights
59
60
Study components• Population-based household surveys
• Face-to-face interviews in Tanzania, Bangladesh, Indonesia, Peru
• Telephone interview in random sample of US households
• Focus on paired comparisons for 108 health-states
• Key objectives include comparative analysis across diverse settings and benchmarking Internet survey against community samples
• Open-access Internet surveys• Available in English, Spanish and Mandarin
• Key objectives are to fill in gaps with remaining sequelae and to anchor scale for paired comparison responses
Bangladesh
Tanzania
60
Web survey included 16,328 respondents from 167 countries
Web survey
1 - 910 - 4950 - 99100 - 499500+
61
62
Measurement methods: paired comparisons
• Primary mode of eliciting responses is paired comparison• Respondents hear (or read) two descriptions of hypothetical people, e
ach with a randomly selected condition
• Respondents indicate which person is healthier
• Paired comparison questions chosen for relative ease of comprehension, administration and analysis
• Literacy and numeracy not essential
• Health comparisons not tied to external “calibrators” such as risk
• Appealing intuitive basis and established strategies for analysis
62
Paired comparison exampleThe first person has vision problems that make it difficult to see and recognize faces of family or friends across a room.The second person has severe back and leg pain, which causes difficulty dressing, sitting, standing, walking, and lifting things. The person sleeps poorly and feels worried. Imagine that both people will have these problems for the rest of their lives. Who would you say is healthier overall, the first person or the second person?
• Respondents in household surveys each answered 15 of these questions, with pairs of sequelae drawn at random from the universe of 108 × 108 possible pairs
63
64
Analyzing paired comparison dataBangladesh Tanzania
• Simple ordering of outcomes based on “winning” proportions, as in various websites using paired comparisons to rank large pools of competitors
http://kittenwar.com/
64
65
Analyzing paired comparison dataBangladesh Tanzania
Won 78% of 511 battles Won 77% of 539 battles
Winningest
Lost 80% of 846 battles Lost 80% of 1872 battles
Losingest
• Simple ordering of outcomes based on “winning” proportions, as in various websites using paired comparisons to rank large pools of competitors
http://kittenwar.com/
65
-4 -2 0 2 4 6 8
depression blindnessarthritis
90%
66
-4 -2 0 2 4 6 8
depression blindnessarthritis
67
Results: new disability weights
68
The universe of Cause of Death dataIHME attempted to identify all available data on causes of death for 187 countries from 1980 to 2010• IHME used 9 different sour
ces of CoD data • IHME collected data on aro
und 600 million deaths in the last 30 years
• Data available varies by disease:
– More on maternal, cancer, injuries
– Less on NTD, diarrhea and LRI pathogens
Type Site years Countries
Vital Registration 2,798 130
Verbal Autopsy 486 66
Cancer registries 2,715 93
Police Reports 1,129 122
Surveys/Census 1,564 82
Maternal Mortality Surveillance
83 8
Deaths in health Facilities 21 9
Burial and Mortuary 32 11
Country−years of vital registration, 1980−2010
69
Assessment and enhancement of data qualityand comparability
1. Assessment of completeness
2. Causes of death mapping3. Redistribution of misclassi
fied causes of death4. Age and age-sex splitting5. Smoothing for stochastic
variation due to small numbers
6. Outlier detection
Percent garbage from ICD vital registration
70
Modeling causes of death
1. Causes of death ensemble modeling, CODEm (133 causes), including all major causes except HIV. CODEm selects models and ensembles of models based on out-of-sample performance.
2. Negative binomial (12 causes).3. Fixed proportion models (27 causes).4. Disaggregation by pathogens or sub-causes (36 cause
s).5. Natural history models (8 causes).6. Mortality shock regressions (2 causes).
71
72
73
Change in Percent of DALYs Due to NCDs 1990-2010
74
Disability Transition: Share of DALYS Due to YLLs and YLDs by Region, 1990 and 2010
1990 2010
75
Fourth Pattern: the Majority of Burden in sub-Saharan Africa is Still from MDGs 4, 5 and 6
76
Uncertainty Varies by Cause
77
Leading Causes of DALYs by Country, 2010
78
BOD due to risk factor
WHO, The world health report 2002 79
Methods
1. Calculate the proportion of deaths or disease burden holding other independent factors unchanged.
2. Counterfactual analysis: What if risk exposure was at a different level?
3. 67 risk factors and clusters of risk factors.
4. 20 age groups, both sexes, 187 countries, and for 1990, 2005, and 2010.
80
GBD 2010 – risks quantifiedUnimproved water and sanitationUnimproved waterUnimproved sanitationAir pollutionAmbient particulate matter pollutionHousehold air pollution from solid fuelsAmbient ozone pollutionOther environmental risksResidential radonLead exposureChild and maternal undernutritionSuboptimal breastfeeding
Non-exclusive breastfeedingDiscontinued breastfeeding
Childhood underweightIron deficiencyVitamin A deficiencyZinc deficiency
Tobacco smoking and secondhand smokeTobacco smokingSecond-hand smokeAlcohol and other drugsAlcohol useDrug use (opioids, cannabis, amphetamines)Physiological risks for chronic diseasesHigh fasting plasma glucoseHigh total cholesterolHigh systolic blood pressure High body mass index Low bone mineral densitySexual abuse and violenceChildhood sexual abuseIntimate partner violence
81
Dietary risk factors and physical inactivityDiet low in fruitsDiet low in vegetablesDiet low in whole grainsDiet low in nuts/seedsDiet low in milkDiet high in unprocessed red meatDiet high in processed meatSugar-sweetened beveragesDiet low in fibreDiet low in calcium Diet low in seafood omega-3Diet low in polyunsaturated fatty acid (PUFA)Diet high in trans fatty acidsDiet high in sodiumPhysical inactivity and low physical activity
Occupational exposuresOccupational exposure to asbestosOccupational exposure to arsenicOccupational exposure to benzeneOccupational exposure to berylliumOccupational exposure to cadmiumOccupational exposure to chromiumOccupational exposure to dieselOccupational exposure to formaldehydeOccupational exposure to nickelOccupational exposure to PAHsOccupational exposure to secondhand smokeOccupational exposure to silicaOccupational exposure to sulfuric acidOccupational exposure to asthmagensOccupational exposure to particulates and gasesOccupational noiseOccupational risk factors for injuryOccupational low back pain
GBD 2010 – risks quantified (2)
82
Calculating risk factor burden
1. Select risk-outcome pairs;
2. Estimate exposure distributions to each risk factor in the population;
3. Estimate cause effect sizes: relative risk per unit of exposure for each risk-outcome pair;
4. Choose a counterfactual exposure distribution: theoretical minimum risk exposure distribution (TMRED); and
5. Compute attributable burden, including uncertainty.
83
Three essential elements
84
Population attributable fraction, PAF
• Two categories:
– 𝑃𝑃𝑃𝑃𝑃𝑃 = 𝑃𝑃(𝑅𝑅𝑅𝑅−1 )𝑃𝑃 𝑅𝑅𝑅𝑅−1 +1
• Multiple categories:
– 𝑃𝑃𝑃𝑃𝑃𝑃 = ∑𝑖𝑖=1𝑛𝑛 𝑃𝑃𝑖𝑖(𝑅𝑅𝑅𝑅𝑖𝑖− 1)
∑𝑖𝑖=1𝑛𝑛 𝑃𝑃𝑖𝑖 𝑅𝑅𝑅𝑅𝑖𝑖−1 +1
– Where RRi is the relative risk for exposure category i and Pi is the fraction of the population in exposure category i.
• Continuous quantity
– 𝑃𝑃𝑃𝑃𝑃𝑃 = ∫𝑥𝑥=0𝑚𝑚 𝑅𝑅𝑅𝑅 𝑥𝑥 𝑃𝑃 𝑥𝑥 𝑑𝑑𝑥𝑥− ∫𝑥𝑥=0
𝑚𝑚 𝑅𝑅𝑅𝑅 𝑥𝑥 𝑃𝑃′ 𝑥𝑥 𝑑𝑑𝑥𝑥
∫𝑥𝑥=0𝑚𝑚 𝑅𝑅𝑅𝑅 𝑥𝑥 𝑃𝑃 𝑥𝑥 𝑑𝑑𝑥𝑥
– Where RR(x) is the relative risk at exposure level x, P(x) is the (observed or estimated) population distribution of exposure, P’(x) is the counterfactual distribution of exposure, i.e., the TMRED, and m the maximum exposure level.
Stephen S Lim. et al. The Lancet, Volume 380, Issue 9859, Pages 2224 - 2260, 15 December 2012
85
Exposure: ambient PM pollution
86
PM2.5 (µg per m3)
Exposure: ambient PM pollution (2)
• Satellite-based measures of aerosol optical depth (AOD)• TM5 chemical transport models• Calibrated against ground-based PM2.5 sensors
87
Burden of disease attributable to 20 leading risk factors in 2010, expressed as a percentage of global disability-adjusted life years, both sexes
88
89
GBD PROFILE: South KoreaPercent decline in age-specific mortality rate by sex from
1990-2010 in South Korea
90
Ranks for top 25 causes of YLLs 1990-2010, South Korea
91
South Korea YLDs by cause and age 2010
92
Burden of disease attributable to 15 leading risk factors in 2010, expressed as a percentage of Korea DALYs
93
94
Global Burden of Diseases, Injuries, and Risk Factors Study 2013
• Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and chronic diseases and injuries in 188 countries, 1990–2013: a systematic analysis for the Global Burden of Disease Study 2013 (Published: June 8, 2015)
• Global, regional, and national age–sex specific all-cause and cause-specific mortality for 240 causes of death, 1990–2013 (Published: December 17, 2014)
• Global, regional, and national incidence and mortality for HIV, tuberculosis, and malaria during 1990–2013 (Published: July 22, 2014)
• Global, regional, and national levels and causes of maternal mortality during 1990–2013 (Published: May 2, 2014)
• Global, regional, and national levels of neonatal, infant, and under-5 mortality during 1990–2013 (Published: May 2, 2014)
95
Thank you
96
