Accounting for Death in the Tri-national Impact Assessment UN / ECE Symposium, London 18th / 19th February 2001

1
Accounting for Death in the Tri-national Impact
AssessmentUN / ECE Symposium, London 18th /
19th February 2001
Nino Künzli, MD PhDAssistant Professor (PD)
Institute of Social and Preventive
MedicineUniversity Basel Nino.Kuenzli_at_unibas.ch
HEIFeb01.ppt
2
Focus

• Background of the Trinational Study
• General Method
• Attributable Death
• Interpretation and uncertainties
• Conclusion

HEIFeb01.ppt
3
(No Transcript)
4
Question of the Authorities of Austria, France,
Switzerland What are the health costs of
traffic-related air pollution ?
UNECE1.ppt
5
Requires interdisciplinary approaches
Air pollution concentrations Air
Hygiene Population exposure distribution GIS Expos
ure - Response function Epidemiology Derivation
of attributable cases Epi Risk
Assess. Costing Economy
UNECE.ppt
6
Uncertainties

• multiple sources of uncertainties
• majority not statistical uncertainties
• conceptual and data uncertainties
• gaps at interdisciplinary interface

• At least approache
• with risk function based 95 CI

UNECE.ppt
7
Definition of Air Pollution Exposure

• Availability of effect functions
• Availability of exposure data
• no multiple counting (correlated pollutants same
  sources)

• PM10

at least
UNECE.ppt
8
Selected Attributable Health outcomes
Attributable Death Long-term
effect Hospital admissions (cardio-resp.) Shor
t-term effect Chronic bronchitis incidence
(adults) Long-term effect Bronchitis episodes
(children) Short-term effect Restricted
activity days Short-term
effect Asthma attacks (children adults)
Short-term effect
UNECE.ppt
9
Impact Assessment Model
Death
Exposure-response function
Attributable death
7.5 ug/m3
at least
PM10
PM10 reference
10
Exposure-Response Functionfor Mortality

• Weighted average of
• ACS-Study (Pope et al., 1994)
• Harvard 6-City Study (Dockery et al, 1993)
• 4.3 (95 CI 2.6-6.1) increase per 10
  ?g/m3 PM10
• 360 (200-520) attributable death
• per 10 ?g/m3 PM10 and 1 Mio.
  inhabitants

UNECE.ppt
11
UNECE.ppt
12
Frequency distribution of total PM10 population
exposure
UNECE.ppt
13
Air Pollution Attributable Death in Austria,
France and Switzerland
Attributable Death (adults ³ 30 yrs.) Total
cases 40600 (24600-56'900) or
6 (3.6-8.4) of total deaths Traffic
related 21800 (13300- 30600)
UNECE.ppt
14
Monetarization of attributable death in the
Trinational Study
source Willingness-To-Pay
estimates value 900000 EUR per
statistical fatality total 36.5 Mio. EUR
See Sommer et al, in OECD Report Ancillary
Benefits and Costs of Greenhouse Gas Mitigation,
New York 2000
UNECE.ppt
15
Air pollution related health costs 50000
Mio. EUR per year traffic related 27000 Mio
EUR per year
UNECE.ppt
16
Key Outcome Uncertainties
1) Why cohort study estimates? 2) Attributable
death and life time lost 3) Are US-Studies
relevant for Europe? 4) Are effects source
specific? 5) Are attributable death
preventable
UNECE.ppt
17
1) Why Cohort study based Attributable Death ?
OR Why not time-series based estimates, which
are 5-10 times smaller ?
UNECE.ppt
18
Frailty and Death
Frailty (susceptibility for death)
Death
Time
Birth
Age at death
19
Frailty and Exposure
Frailty (susceptibiltiy for death)
Death
Time
Birth
Age at death
Start smoking
20
Contribution of Air Pollution
Frailty (susceptibility for death)
SMOG EPISODE
Death
Cumulative exposure
Chronic Bronchitis
Time
Birth
Age at death
21
Focus of the Time-series analysis
Frailty (susceptibilty for death)
Death
Time
Birth
22
Time-series counting advancement of death
Death counts
Pollution
Time
Period of exposure
Time to event
Period of events
SHORT PERIODS
SHORT PERIODS
SHORT PERIODS
SHORT PERIODS
SHORT PERIODS
UNECE.ppt
23
Time-series

• 1) Time-series captures
• short-term effects only
• not all short-term effects
• 2) Life time lost not known

• 3) Long-term effect on life shortening not
  measured

gt gt incomplete impact assessment
24
Focus of the Cohort Study
Frailty (susceptibilty for death)
Time
Birth
25
Cohort studies

• capture cumulative total effects on time to death

gt more complet counting of effects

• BUT
• no distinction of acute and long-term
• only US studies
• YoLL not published /QALY not assessed

26
Long-term and short-term cases(Künzli et al, Am
J Epidemiol 2001 in press, adapted from COMEAP,
1998)
All Air Pollution cases
long-term effect
short-term effect
Mixed effects
All Death
27
2) Number of deathversus years of life lost

• Both
• Life expectancy in a population
• annual number of deaths
• depend on
• age specific death rates

28

• BUT
• published are only number of deaths !

• Indirect estimates of years of life lost
• life table approache
• assumptions about age structure of cases

29
Mortality in the Trinational Study

• 1) Direct estimates of attributable death
• ( gt age 30 non-violent death)
• 2) Years of life lost
• Indirect estimates, used for costing
• assumptions
• the attributable death are due to cardiopulmonary
  causes
• same age distribution as all cardiopulmonary
  death
• thus typical age of an air pollution death
  75-80 yrs.

30
Average reduction in population life expectancy

• For a population with a 10 ?g/m3 higher PM10
• Sommer et al. (Trinational Study) 0.5 yr
• other estimates 6 mo. to 2 yrs.
• (Brunekreef Pope Leksell Hurley Miller Rabl
  etc.)

31
3) We transfered the US cohort results to Europa
!

• Consistency for other outcomes between US and
  European countries
• Coherence between different outcomes

UNECE.ppt
32
Mortality (adjusted Relative Risk) and long-term
mean pollution (PM2.5)Harvard Six-City Cohort
Study, Dockery et al, NEJM 1993 329 (24)1753-9
Relative Risk
mean PM2.5
UNECE.ppt
33
Annual mean PM10 and expiratory capacity (FVC) in
the 8 SAPALDIA areas Ackermann-Liebrich et al,
AJRCCM 1997 155 (1)122-129
FVC deviation from predicted
PM10 annual mean (?g / m3)
UNECE.ppt
34
Other Evidence?
Europe French PAARC suggestive abstract Dutch
cohort suggestive abstract SAPALDIA ?? Needs
other 10 yrs. ECRHS ?? ?
USA Abbey et al 1999 interaction with
sex? Krewski et al 2000 interaction with SES?
UNECE.ppt
35
4) Are attributable death source specific?

• Assumption in tri-national study
• PM10 effects are independent of PM source
• -gt one single effect slope

UNECE.ppt
36
Percent Change (and 95th CI) in Daily Death per
10 ?g/m3 PM2.5 mass an source specific
elements1979-1988, Harvard Six Cities Laden et
al, Env Health Perspect 2000 108 (10) 941-947
Motor (Pb)
PM2.5 mass
Coal (Se)
Crustal (Si)
37
Percent Change (and 95th CI) in Daily Death per
10 ?g/m3 PM2.5 mass an source specific
elements1979-1988, Harvard Six Cities Laden et
al, Env Health Perspect 2000 108 (10) 941-947
Motor (Pb)
PM2.5 mass
Coal (Se)
Crustal (Si)
38
Percent Change (and 95th CI) in Daily Death per
10 ?g/m3 PM2.5 mass an source specific
elements1979-1988, Harvard Six Cities Laden et
al, Env Health Perspect 2000 108 (10) 941-947
Motor (Pb)
PM2.5 mass
Coal (Se)
Crustal (Si)
39
Percent Change (and 95th CI) in Daily Death per
10 ?g/m3 PM2.5 mass an source specific
elements1979-1988, Harvard Six Cities Laden et
al, Env Health Perspect 2000 108 (10) 941-947
Motor (Pb)
PM2.5 mass
Coal (Se)
Crustal (Si)
40
5) Are attributable death also preventable
death?
UNECE.ppt
41
Unknown time domaines

• Duration of long-term exposure
• Relevant time window of exposure
• Time between exposure and effect
• Time between intervention and benefit?

UNECE.ppt
42
Time-to-Benefit (death, long-term) (Sommer et
al., 1999)
Attributable Costs
10
20
Time (yrs)
Clean air intervention
UNECE.ppt
43
Competing risks
Removal of one long-term health risk factor will
modify the importance of the remaining
risks attributable death gt preventable
death
UNECE.ppt
44
TSP decrease and reduction in bronchitis
prevalence (Germany, 1993 versus 1995, children,
age 5-14 yrs)
70
Hettstedt
60
Bitterfeld
bronchitis prevalence ()
Zerbst
50
40
Year
30
1993
30
40
50
60
70
1995
3
annual mean of TSP
m
g/m
(Heinrich et al, AJRCCM 2000 1611930-36)
COP6DenHaag.ppt
45
Acute bronchitis in children and annual mean
PM10 The SCARPOL Study
25
23
Genf
21
Lugano
Anières
19
Biel
17
Zürich
Bern
15
Prevalence of dry cough at night ()
Langnau
13
Payerne
11
Montana
9
Rheintal
7
5
5
10
15
20
25
30
35
Annual mean PM10 mg/m3
3
Jahresmittel PM10(
m
g/m
)
C. Braun-Fahrländer AJRCCM 1997
UNECE.ppt
46
Conclusion

• Air pollution accounts for 4-8 of death
• Attributable death from cohorts adequate
  approache
• BUT gaps in evidence and uncertainties
• Best effect estimate?
• Years of life time lost?
• European long-term studies?
• Source attribution ?
• How many are preventable death? When?

UNECE.ppt
47
Thanks
Nino.Kuenzli_at_unibas.ch Institute of Social and
Preventive Medicine University Basel
48
Susceptibility and Time-to-event
Probability of event

• High susceptibility
• Early response

• Low susceptibility
• Late response

Time
Exposure