Transcript Introduction and Oveview - World Health Organization

Protecting our Health from
Climate Change:
a Training Course for Public
Health Professionals
Chapter 6: Analyzing the Health
Effects of Weather, Climate and
Climate Change
STRATUS CONSULTING
Some definitions
 Weather: the day-to-day atmospheric conditions in
a specific place at a specific time
 Climate: the average state of the atmosphere and
the underlying land or water in a specific region
over a specific time scale
 Climate change: variation in either the mean state
of the climate or in its variability, persisting for an
extended period (typically decades or longer)
STRATUS CONSULTING
Types of analysis
OBSERVATIONAL
(1) Episodes or event analysis: heat wave, flood, drought…
(2) Time-series analysis: mortality vs temperature,
precipitation
(3) Seasonality: diarrhoea, aero-allergens
(4) Changes in geographical distribution:
temperature/precipitation vs VBDs
MODEL-BASED
(1) Health burdens: risk assessments
(2) Decisions analysis of health impact of policy options
Decades
Climate change
Years
Climate
Seasonality
Weather
Months
Days
HISTORICAL
EVIDENCE
(recent past)
Conventional epidemiology,
observation
FUTURE
IMPACTS
(mid century
onwards)
Models, synthesis,
‘triangulation’
Daily changes: two approaches

Episode analysis
- transparent
- risk defined by comparison to local baseline

Regression analysis of all days of year (time-series)
- uses full data set
- requires fuller data and analysis of confounders
- can be combined with episode analysis
No. of deaths/day
Principles of episode analysis
Smooth function of date
Triangle: attributable
deaths
Smooth function of date
with control for influenza
Period of heat
Influenza ‘epidemic’
Date
60
40
20
0
0
100
200
Mean daily temperature (degrees Celsius)
80
300
DEATHS, LONDON, 2003
01jan2003
01apr2003
01jul2003
Date
01oct2003
01jan2004
01jan2003
0
0
200
01apr2003
20
40
60
300
01jul2003
Date
80
16 August
1 August
Mean daily temperature (degrees Celsius)
100
DEATHS, LONDON, 2003
01oct2003
01jan2004
100
150
200
250
300
DEATHS, LONDON, 2001-2003
0
10
20
Mean temperature / Celsius
30
Interpretation

Common sense, transparent

Relevant to PH warning systems
But

How to define episode?
- relative or absolute threshold
- duration
- composite variables

Uses only selected part of data

Most sophisticated analysis requires same methods
as for regression of all days of year
Time-series regression

Short-term temporal associations

Daily/weekly

Suitable for episodes or effects of local fluctuations
in meteorological parameters

U- or V-shape of temperature-response function

Different lags
Parameterization: hockey-stick models
f(ti, ß) = ßctc,i + ßhth,i
Relative
risk
tc,i = max[tc - ti,0] ‘cold’
th,i = max[th - ti,0] ‘heat’
Cold slope
Heat slope
Tc
Th
Temperature
Minimum mortality
range
The model
(log) rate = ß0 +
ß1(high temp.)
ß2(low temp.)
ß3(pollution)
+
+
ß2=cold slope
+
ß4(influenza)
+
ß5(day, PH)
+
ß6(season)
+
ß7(trend)
ß1=heat slope
measured
confounders
unmeasured
confounders
Lags

Heat impacts short: 0-2 days
Cold impacts long: 0-21 days

Vary by cause-of-death
- CVD: prompt
- respiratory: slow

Should include terms for all relevant lags
Increase in mortality/degrees Celsius
below cold threshold
Lags for cold-related mortality, London
Cardiovascular death
Respiratory death
2
2
1.5
1.5
1
1
0
5
10
15
0
Time lag (days)
5
10
15
LAG: 0-1 DAYS
(HEAT)
LAG: 2-13 DAYS
(COLD)
Relative
risk
Relative
risk
Temperature
Threshold for
heat effect
Temperature
Threshold for
cold effect
Mortality displacement: schema
MORTALITY
A
B
HEAT
strong correlation
Period of
averaging
weaker
absent
Constrained distributed lag model:
“harvesting” interpretation
RR increment per degree
Constrained distributed lag model for hightemp
(Quartic polynomial)
‘Prompt’ adverse effects
1.01
1
Deficit 3 weeks later - harvesting?
.99
0
5
10
15
Lag
20
25
30
Controlling for season
TEMPERATURE
MORTALITY
SEASON
X
X
Infectious disease
Diet
Human behaviours
UNRECORDED FACTORS
Methods of seasonal control





Moving averages
Fourier series (trigonometric terms):
Fn(x) = a0 + (a1cos(x) + b1sin(x))+ …
+ (ancos(nx) + bnsin(nx))
where a0, b0, a1, b1,… are coefficients of
Fn(x)
Smoothing splines
Stratification by date
Other…
Inter-annual variation:
10
2.0
9
1.5
La Nina
1.0years
SOI (Southern Oscillation Index)
7
6
0.5
5
0.0
4
-0.5
3
SOI
8
1
-1.0
El Nino years
-1.5
0
-2.0
Hales and Woodward, 1999
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
2
1970
number of epidemics
example of dengue
epidemics in the South Pacific 1970-1998
Seasonality
Cases of
diarrhoeal
disease
Current distribution
Distribution under
global warming?
Date of year
Summary of time-series





Provide evidence on short-term associations of weather
and health
Robust design
Repeated finding of direct heat + cold effects
Some uncertainties over PH significance
Uncertainties in extrapolation to future
(No historical analogue of climate change)
Changes in geographical distribution of
disease
(1) BIOLOGICAL MODELS
 Use of (laboratory derived) biological evidence
(2) STATISTICAL MODELS
 Analyses of disease prevalence or vector abundance
in relation to geographical factors
BITING FREQUENCY OF MOSQUITO
Bites per day
0.3
0.2
0.1
15
20
25
30
35
40
Temperature /deg Celsius
Incubation period (days)
INCUBATION PERIOD OF PARASITE
60
40
20
15
20
25
30
35
40
Temperature /deg Celsius
PROBABILITY OF MOSQUITO SURVIVAL
P(S) per day
0.75
TRANSMISSION POTENTIAL
1
0.5
0.8
0.6
0.25
15
20
25
30
35
Temperature /deg Celsius
40
0.4
0.2
0
15
20
25
30
35
Temperature /deg Celsius
40
Estimated population at risk of dengue fever:
(A) 1990, (B) 2085
Source. Hales S et al. Lancet (online) 6 August 2002.
http://image.thelancet.com/extras/01art11175web.pdf
Future burdens: risk assessment
GHG emissions
scenarios defined by IPCC
GCM model:
Generates series of maps of
predicted future distribution of
climate variables
Health impact model Generates
comparative estimates of the regional
impact of each climate scenario on
specific health outcomes
Level
1
2
3
1
Conversion to GBD
‘currency’ to allow
summation of the effects of
different health impacts
Age group (years)
0-4
5-14
15-29
1.0
1.0
1.0
1.2
1.2
1.2
1.7
1.7
1.7
1.0
1.0
1.0
30-44
1.0
1.2
1.7
1.0
45-59
1.0
1.2
1.7
1.0
60-69
1.0
1.2
1.7
1.0
70+
1.0
1.2
1.7
1.0
2
3
1
2
3
1
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
1.2
1.7
1.0
2
3
1
2
3
1.2
1.7
1.0
1.2
1.7
1.2
1.7
1.0
1.2
1.7
1.2
1.7
1.0
1.2
1.7
1.2
1.7
1.0
1.2
1.7
1.2
1.7
1.0
1.2
1.7
1.2
1.7
1.0
1.2
1.7
1.2
1.7
1.0
1.2
1.7
Relative mortality (% of daily average)
Heat-related mortality (Delhi)
Temperature
distribution
140
120
100
80
0
10
20
30
Daily mean temperature /degrees Celsius
40
Uncertainties
• EXTRAPOLATION
(going beyond the data)
• VARIATION
(..in weather-health relationship -- largely unquantified)
• ADAPTATION
(we learn to live with a warmer world)
• MODIFICATION
(more things will change than just the climate)
Changing vulnerability
• Changes in population
- Demographic structure (age)
- Prevalence of weather-sensitive disease
• Environmental modifiers
• Adaptive responses
- Physiological habituation (acclimatization)
- Behavioural change
- Structural adaptation
- PH interventions
Conclusions

Most methods of ‘climate’ attribution based on
analysis of weather-health associations: episode
analysis, time-series, seasonality, inter-annual
variations

Relevance to climate change limited by
uncertainties over multiple effect-modifiers –
changes in vulnerability of population & health

Modelling intrinsic to assessment of likely future
burdens and the effect of adaptation options, but
entails many uncertainties