Transcript No Slide Title

Environment, Society, Climate and Health:
Analysis, Understanding and Prediction
PART 2
Mark L. Wilson
Department of Epidemiology
and
Global Health Program
School of Public Health
The University of Michigan
Colloquium on Climate and Health
NCAR
Boulder, Colorado
23 July, 2004
Some examples of studies from our group
• Importance of environment (BROADLY defined)
• Role of spatial pattern of people, people-environment
- spatial autocorrelation as a problem
- spatial pattern as a source of insight
• Attempt to integrate individual- and population-level
• Increasing use of time series, time-space analyses
• Longer-term: integrate these analyses and underlying
methods with more "upstream" causes
• Summarize NAS report findings
Spatially-Extensive Examples
•Remote sensing of environment
•Large scale active surveillance
•Population Census – human / animal
Cutaneous Leishmaniasis - Turkey
(Collaboration with Aksoy et al.)
(Collaboration with Aksoy et al.)
(Collaboration with Aksoy et al.)
(Collaboration with Aksoy et al.)
(Collaboration with Aksoy et al.)
(Collaboration with Aksoy et al.)
(Collaboration with Aksoy et al.)
(Collaboration with Aksoy et al.)
(Collaboration with Aksoy et al.)
Dengue Fever and Water Sources - Peru
Schneider et al. 2004 in press (collaboration with Morrison, et al.)
N
Average Wing Length by City Block
Averag e W ing L en gth
1.86 67 - 2 .404 5
2.40 45 - 2 .575
2.57 5 - 2.6 496
2.64 96 - 2 .8
2.8 - 3 .1 16 7
Zone s
BG
IQ
MC
MY
PT
PU
SA
TA
Schneider et al. 2004 in press (collaboration with Morrison, et al.)
Spatial Patterns of Malaria Risk - Kenya
Macdonald et al. in prep (collaboration with Hawley, Hightower, et al.)
Macdonald et al. in prep (collaboration with Hawley, Hightower, et al.)
Hot spot and cold spot clusters for
Anopheles gambiae
Macdonald et al. in prep (collaboration with Hawley, Hightower, et al.)
Some Conclusions
•
•
•
•
Clusters of apparently higher risk
No obvious link to mosquito breeding sites
Associations with crude measures of SES weak
Pattern of higher risk suggests possible role of
regional environmental factors
• Generates new hypotheses for more focused
studies
Other Malaria Studies
• Malawi - analysis of role of ITNs in reducing childhood
anemia and mortality (Don Mathanga)
- measures SES, knowledge, access and use
- ITNs highly effective, also efficacious
- ORs for income, educ., housing all signif.
• Kenya - urban malaria and patterns of environmental and
SES inequality (Jose Siri)
- cases/controls, questionnaire KAP, household
environment, RS and ground-based environmental data
- strong spatial clustering of cases, environment vars.
- KAP and SES data being analyzed
Temporally-Extensive Data
- Examples
•Long-term samples of environment
•Systematic surveillance of cases
•Population Census – human / animal
250
Viral Meningitis in Michigan
200
Number of Cases
150
• Collaboration with State Epidemiologists
• County-specific case data from 1993-2001
• Cases adjusted to county population & area
100
50
0
1993
1994
1995
1996
1997
Year
1998
1999
2000
2001
Viral Meningitis - Michigan, 1993-2001
• Time series of cases
• Autocorrelation function
ACF=0.43 at 3-year lag
Viral Meningitis in Michigan
42
counties,
July Oct.,
2001
1
2
3
5 counties
+ Detroit,
July Oct., 1998
2
counties,
Aug Oct., 2001
Figure 5. The Three Most Likely Overall Spatio-Temporal Clusters
 The three clusters were each significant, with p-value = 0.01
 In the most likely cluster (#1), children <10 years old were 42% of all cases,
while for all 8,803 cases, this age group constituted 34%. X2 test for specified
proportions: X2=36.5, d.f.=1,
p-value <0.0001
Greene et al. In press
Influenza and Environment
• Investigation of relationships among epidemic
onset, duration, magnitude, predominant
circulating strain(s) - and climate signals
• Are climate-influenza relationships regionspecific?
• Are relationships consistent across years?
Potential Mechanisms
• Climate could influence:
– onset of transmission
– cessation of transmission
– patterns of contagion
– apparent inter-epidemic virus disappearance
– regional synchrony of transmission
– virus extra-host survival
– human immunity
– disease expression
– human-to-human contact patterns
– non-human host abundance and behavior
Specific Potential Mechanisms
• Temperature
– virus survival
– defense mechanisms of URT
– crowding
• Humidity
– Assays show infectivity of influenza virus declined rapidly
under conditions of 40% humidity1
– Conditions of low indoor humidity during winter could
promote virus survival and ↑ transmission
1Saito
et al. (2003) Options for the Control of Influenza V, poster.
Environmental Data: Multivariate El Niño
Southern Oscillation Index (MEI)
• Sea-level pressure
• Zonal and
meridional
components of the
surface wind
• Sea surface
temperature
• Surface air
temperature
• Total cloudiness
fraction of the sky
•MEI and ENSO temporal patterns similar
http://www.cdc.noaa.gov/~kew/MEI/mei.html
Influenza Data from France
Surveillance: Influenza-Like-Illness (ILI) all France
(500 physicians, 88 provinces) 1984 - present
2000
1750
ILI / 100,000
1500
1250
1000
750
500
250
0
nov-84
nov-86
nov-88
nov-90
nov-92
nov-94
nov-96
nov-98
CALENDAR DATE
Viboud, et al. (2004) European Journal of Epidemiology (in press)
Influenza - Climate Variability Temporal Pattern
MEI
Dominant
Strain
Temporal pattern of ILI epidemic magnitude (red) categorized as below or above average (1984-2000) and
annual excess mortality (black) (1984-1997). Symbol size is proportional to the value it represents. Monthly
Multivariate ENSO Index (MEI) shown as blue curve, left y-axis. Influenza virus variants predominantly
circulating in France are indicated for each winter.
Viboud, et al. (2004) European Journal of Epidemiology (in press)
Summary of Effect in France
5
2500
ILLNESS
(MILLIONS)
4
2000
3
1500
2
1000
1
500
0
0
WARM
COLD
ENSO CONDITIONS
EXCESS
DEATHS
WARM
COLD
ENSO CONDITIONS
1979-2000: Influenza-related morbidity and mortality
greater during cold ENSO conditions
Viboud, et al. (2004) European Journal of Epidemiology (in press)
Observed Flu Seasonality - U.S.
• NCHS pneumonia and influenza / respiratory and circulatory
mortality
–
Age, race, sex, county, metropolitan statistical area, underlying cause, and up
to 8 axis conditions listed, with monthly resolution
• National Hospital Discharge Database
• Circulating Strains/Vaccine/Match, 1978 – 2003
Preliminary Analysis in USA
• For all regions, influenza negatively correlated with MEI
• (consistent with Viboud et al. 2004)
– Effect varies by region, with weakest correlation in New England, strongest in Pacific,
and intermediate values in between
– Regional trend interesting; repeat using properly grouped seasons from NCHS data and
examining lagged climatic drivers
• Viral pneumonia also neg. correlated w/ MEI, but E-W trend not seen
• MEI not correlated with a chronic respiratory disease, asthma
Greene et al. In preparation
Time-Space Extensive Data - Examples
Raccoon Rabies Expansion - Connecticut
•
•
•
•
All cases of raccoon rabies, 1991-1996
Georeferenced to location where found
Spatio-temporal analysis of spread
Trend surface velocity analysis for
vectors indicating direction and rate of
spread
• Simulation modeling of importance of
rivers
Raccoon Rabies Expansion - Connecticut
First case in each township indicated by darker
color
Raccoon Rabies Expansion - Connecticut
Best fit trend surface vector field showing
direction and velocity of spread of the infection
Lucey et al. 2002
Influenza Time-Space Spread - France
Week 1
Week 3
Week 5
Week 7
Summary
•
Disease and environmental patterns typically vary
temporally, spatially, spatio-temporally
•
Environmental factors affect most diseases, but
especially so for infectious diseases
•
Analysis of space and time patterns can help clarify
confounding, identify new associations, develop new
hypotheses… and determine lack of independence
•
Challenge: better integrate these analyses and
underlying methods with studies of more "upstream"
causes
What are the health implications
for these unprecedented climatic
events?
Under the Weather: Climate, Ecosystems and Infectious Disease
Committee Members
Johns Hopkins University
DONALD BURKE (Chair)
Indiana University
ANN CARMICHAEL
U.S. Department of Agriculture
DANA FOCKS
University of Southern Mississippi
DARELL GRIMES
University of California, Berkeley
JOHN HARTE
University of Alberta
SUBHASH LELE
Maastricht University, Netherlands
PIM MARTENS
University of Washington
JOHNATHAN MAYER
National Center for Atmospheric Res.
LINDA MEARNS
University of Colorado / NOAA
ROGER PULWARTY
Emory University
LESLIE REAL
Intl. Research Inst. for Climate Prediction
CHET ROPELEWSKI
University of South Florida
JOAN ROSE
University of Texas Medical Branch
ROBERT SHOPE
NASA Goddard Space Flight Center
JOANNE SIMPSON
University of Michigan
MARK WILSON
LAURIE GELLER
SUSAN ROBERTS
JONATHAN DAVIS
NRC Staff
Board on Atm. Sciences and Climate
Ocean Studies Board
Institute of Medicine
NAS Committee Tasks
1) provide a critical review of the linkages between climate
variability and emergence/transmission of infectious
disease agents, and to explore feasibility of using this
information to develop a fuller understanding of the possible
impacts of long-term climate change.
2) develop an agenda for future research activities that could
further clarify these linkages.
3) examine the potential for establishing disease early-warning
systems based on climate forecasts and for developing
effective societal responses to such warnings.
Sponsors: USGCRP, CDC, NOAA, NASA, NSF, EPA, DOI, EPRI
KEY FINDINGS 1:
Climate-Disease Linkages
Weather fluctuations and seasonal-tointerannual climate variability influence
many infectious diseases
• Characteristic geographic distributions and seasonal variations of many
infectious diseases (IDs) are prima facie evidence of linkages with weather
and climate.
• Studies have shown that temperature, precipitation, humidity affect life
cycles of many pathogens, vectors (directly and indirectly); this, in turn,
may influence timing, intensity of outbreaks.
• However, ID incidence also affected by other factors (e.g. sanitation, public
health services, population density, land use changes, travel patterns).
• The importance of climate relative to these and other variables must be
evaluated in the context of each situation.
KEY FINDINGS 2:
Climate-Disease Linkages
Observational and modeling studies must
be interpreted cautiously
• Numerous studies showing associations between climatic variations and
ID incidence can not fully account for complex web of causation underling
disease dynamics; most are not reliable indicators of future changes.
• Various models simulating effects of climatic changes on incidence of
diseases (e.g. malaria, dengue, cholera) are useful heuristic tools for
testing hypotheses and undertaking sensitivity analyses; they are not
intended to serve as predictive tools; often exclude physical/biological
feedbacks and human adaptation.
• Caution needed in using these models to create scenarios of future disease
incidence, providing early warnings, and developing policy decisions.
2050 projection (from Martens et al., 1999)
2050 projection (from Rogers and Randolph, 2000)
KEY FINDINGS 3:
Climate-Disease Linkages
The potential disease impacts of global
climate change remain highly uncertain
• Changes in regional climate patterns caused by long-term global
warming could affect potential geographic range of many diseases.
• However, if climate of some regions becomes more suitable for
transmission of particular disease agents, human behavioral adaptations
and public health interventions could serve to mitigate many adverse
impacts.
• Basic public health protections (adequate housing, sanitation),and new
interventions (vaccines, drugs), may limit future distribution & impact
of some infectious diseases, regardless of climate-associated changes.
• These protections, however, depend on maintaining strong public health
programs, and assuring vaccine and drug access in poorer countries.
KEY FINDINGS 4:
Climate-Disease Linkages
Climate change may affect the evolution
and emergence of infectious diseases
• Potential impacts of climate change on the evolution and emergence of
infectious disease agents are an additional highly uncertain risk.
• Ecosystem instabilities from climate change and concurrent stresses (e.g.
land use changes, species dislocation, increasing global travel) could
influence genetics of pathogenic microbes through mutation and
horizontal gene transfer.
• New interactions among hosts and disease agents could occur, fostering
emergence of new infectious disease threats.
ANTHROPONOSES
Direct Transmission
Indirect Transmission
Humans
Humans
Vector
Humans
Vector
Humans
ZOONOSES
Animals
Animals
Vector
Animals
Humans
Vector
Animals
Humans
KEY FINDINGS 5:
Climate-Disease Linkages
Potential pitfalls exist in extrapolating
climate and disease relationships among
spatial and temporal scales
• Relationships between climate and infectious disease are often highly
dependent upon local-scale parameters.
• Difficult or impossible to extrapolate these relationships meaningfully to
broader spatial scales.
• Temporal climate variability (seasonal, interannual) may not represent a
useful analog for long-term impacts of climate change.
• Ecological responses on such time scales (e.g. El Niño event) may be
significantly different from the ecological responses and social
adaptations expected under long-term climate change.
• Long-term climate change may influence regional climate variability
patterns, hence limiting the predictive power of current observations.
climate
mean temperature,
precipitation, humidity,
extreme weather events
ecology
vegetation, soil moisture,
species competition
transmission biology
microbe replication/movement,
vector reproduction/movement,
microbe/vector evolution
disease outcome
Risk, rate of transmission
Spread to new areas
social factors
sanitation, vector control,
travel/migration,
behavior/economy,
population/demographics
KEY FINDINGS 6:
Climate-Disease Linkages
Recent technological advances should
improve modeling of infectious disease
epidemiology
• New techniques in several disparate scientific disciplines may encourage
different approaches to infectious disease models.
• Advances include sequencing of microbial genes, satellite-based remote
sensing of ecological conditions, Geographic Information System (GIS),
new analytical techniques, increased computational power.
• Such technologies should improve analyses of microbe evolution and
distribution, and of relationships to different ecological niches.
• This may dramatically improve abilities to quantify disease impacts
from climatic and ecological changes.
KEY FINDINGS 7:
Disease "Early Warning" Potential
Future epidemic control strategies should
complement "surveillance and response"
with "prediction and prevention"
• Current epidemic control strategies depend largely on surveillance for
new outbreaks followed by a rapid response to control the epidemic.
• Climate forecasts and environmental observations could help identify
areas at risk of epidemics, thus aiding efforts to limit or prevent.
• Operational disease early warning systems not yet feasible due to limited
understanding of climate/disease relationships and climate forecasting.
• Establishing goal of developing early warning capacity will foster the
needed analytical, observational, and computational developments.
KEY FINDINGS 8:
Disease "Early Warning" Potential
Effectiveness of early warning systems
will depend upon context of their use.
• Where risk mitigation is simple and low-cost, early warning may be
feasible given only general understanding of climate/disease associations.
• If mitigation actions are significant, precise and accurate prediction may
be necessary, requiring more thorough mechanistic understanding of
underlying climate/disease relationships.
• Value of climate forecasts depends on disease agent and locale (e.g.
reliable ENSO-related disease warnings restricted to regions with clear,
consistent ENSO-related climate anomalies).
• Investment in sophisticated warning systems not effective use of
resources where capacity for meaningful response is lacking, or if
population not highly vulnerable to hazards being forecasted.
KEY FINDINGS 9:
Disease "Early Warning" Potential
Disease early warning systems cannot
be based solely on climate forecasts
• Need for other appropriate indicators (e.g. meteorological, ecological,
epidemiological surveillance) that complement climate forecasts.
• Such combined information may permit a “watch” to be issued for
regions, and a “warning” if surveillance data confirms projections.
• Vulnerability and risk analyses, feasible response plans, and strategies
for effective public communication needed as part of system.
• Climate-based early warning for other applications (e.g. agricultural
planning, famine prevention) may provide many useful lessons.
prediction
surveillance
epidemic
early
cases
climate
forecasts
environmental
observations
sentinel
animals
Time
climate
forecasts
ongoing epidemiological
surveillance and
environmental observations
disease
watch/warning
risk analysis,
vulnerability
assessment
response
strategy
public
communication
evaluation,
feedback
KEY FINDINGS 10:
Disease "Early Warning" Potential
Development of early warning systems
should involve active participation of
the system’s end users
• Input from stakeholders (e.g. public health officials, local policymakers)
needed to help ensure that forecast information is provided in a useful
manner and that effective response measures are developed.
• Probabilistic nature of climate forecasts must be clearly explained to
communities using these forecasts, allowing development of response
plans with realistic expectations of possible outcomes range.
RESEARCH RECOMMENDATIONS
• Research on climate and infectious disease linkages must be
strengthened
• Further development of transmission models needed to assess
risks posed by climatic and ecological changes
• Epidemiological surveillance programs should be
strengthened
• Observational, experimental, and modeling activities must be
coordinated
• Research on climate and infectious disease linkages inherently
requires interdisciplinary collaboration
Unpredictability of climate-disease
linkages suggests reducing human
vulnerability is most prudent public
health strategy
• Understanding of climate linkages to ecosystems and health not solid,
making early warning systems not yet feasible.
• Some unpredictability will always be present.
• Thus, strengthening of public health infrastructure (e.g. vector control,
water treatment systems, vaccination programs) should be high priority.
• Reducing overall vulnerability of populations at risk is the most prudent
strategy for improving health.
Thank you….
Questions?