Anna Hansell Presentation

Report
Environment and Health
Atlas for England and
Wales
Anna L Hansell
Assistant director of Small Area Health Statistics Unit
MRC-HPA Centre for Environment and Health
School of Public Health, Imperial College London
Frontiers in Spatial Epidemiology Symposium
Overview of the Environment and Health Atlas for
England and Wales
What is the aim of the EHA?
• To provide information about geographical variation of disease that
may be related to environmental pollutants
• To provide information about geographical variation of selected
pollutants
• To form a basis for development of hypotheses, further research and
public health action.
How can we access it?
• Print version
• On-line interactive version
The print version will be available both as hard copy to purchase and
as downloadable pdf chapters.
Example: Exposure and Health Outcome Maps
The Environment and Health Atlas
17 authors
12 data
providers
22 reviewers
7 SAHSU team
members
24 Sense
About Science
attendees
4 audiences
Researchers
Public health
Policy makers
Public
Environmental Exposures and Health Outcomes
Environmental Exposures
Health Outcomes
Birth Outcomes
- Air Pollution
• NO2, PM10
- Radon
- Metals
• cadmium, lead
- Agricultural Pesticides
- Sunshine Duration
- Chlorination disinfection
by-products
⁻
⁻
⁻
⁻
⁻
⁻
⁻
⁻
⁻ Mesothelioma
Lung cancer
Breast cancer
Prostate cancer
Malignant melanoma
Bladder cancer
Leukaemia
Brain cancer
Liver cancer
- Still Births
⁻ Coronary heart
disease mortality
⁻ COPD mortality
⁻ Kidney disease
mortality
- Low birth Weight
Exposures
• Air pollution (NO2, PM10): Land
Use Regression for 2001
annual mean concentrations
from the national air quality
network and predictor
variables related to traffic,
population, land use and
topography
• Radon
• Sunshine duration
• Metals in soil (Cd, Pb)
• Agricultural pesticides
• Chlorination disinfection
byproducts
Statistical methods
Poisson framework with BYM model used for all analyses
• Allows to overcome the excess variability due to small
numbers (rare disease/small areas)
• Takes into account the spatial dependence in risks using
– Global smoothing (unstructured heterogeneity)
– Local smoothing (spatial heterogeneity)
• a parameter in one area is influenced by the average value of
its neighbours
• variability quantified by a conditional variance depending on the
number of neighbours
yi ~ Poisson(λi)
Log λi = α + Ui + Vi
U ~ CAR( W, σ2u )
Vi ~ Normal( 0, σ2v )
Smoothed Relative Risk of male lung cancer
incidence adjusted for age and deprivation
• Ward-level maps 1985-2009
• Male/female separately
• Chloropleth maps
• A diverging or bi-polar
scheme is used as risks are
below or above average.
• Health outcomes use same
nine point scale. Each
increment represents 12%
increase in the log RR
• Colour scheme can be read
by colour blind individuals
Smoothed and posterior probability maps
Smoothed Relative Risk of male lung cancer
incidence adjusted for age and deprivation
Posterior probabilities of male lung cancer
incidence adjusted for age and deprivation
Posterior probabilities may be interpreted as the strength of (statistical) evidence
of an excess/reduced risk in each area
Contextual maps
–
–
–
–
–
Topography
Administrative boundaries
Population density (1981, 1991 & 2001)
Urban/rural distribution
SES (Carstairs)
Health outcome chapters – context
• Each chapter contains a brief overview of the disease
including background, risk factors and time trends to provide
context for the maps.
Age-standardised lung cancer incidence and mortality in England
and Wales 1985-2009
Lung cancer incidence rates in males and females in
2002
Health outcome chapters – text
• Text presents information about the outcome and important
risk factors based on authoritative reviews and meta-analyses
• Key points are displayed in text boxes.
Summary text box
 Lung cancer is one of the commonest cancers and is strongly associated
with cigarette smoking.
 In England and Wales there were approximately 35,000 (19,800, male and
15,600 female) new cases of lung cancer in 2009, the last year of data for
the maps.
 The maps show highest risks for lung cancer in conurbations and
industrial areas of England and Wales. This is likely to reflect past smoking
patterns and occupational exposures (e.g. to asbestos) with a smaller
potential contribution from air pollution exposures.
•
Health outcome chapters – statistics
Region
East Midlands
East of England
London
North East
North West
South East
South West
Wales
West Midlands
Yorkshire & the Humber
Rate per
100,000 (age
adjusted)
85.08
76.04
94.10
120.20
103.94
75.79
71.08
92.89
89.57
98.85
Statistical summary: Male
lung cancer incidence.
Observed and expected
numbers, Standardised
Incidence Ratios (SIRs) and
smoothed Relative Risk (RRs)
by census ward 1985-2009
Rate per 100,000
95% Confidence
(age and
Intervals
deprivation
adjusted)
(84.29 to 85.88)
86.02
(75.38 to 76.70)
82.17
(93.39 to 94.82)
84.63
(118.99 to 121.42)
100.94
(103.24 to 104.64)
98.41
(75.25 to 76.34)
86.35
(70.45 to 71.72)
79.85
(91.93 to 93.87)
89.47
(88.84 to 90.30)
85.73
(98.06 to 99.65)
92.13
95% Confidence
Intervals
(85.21 to 86.83)
(81.37 to 82.98)
(83.86 to 85.41)
(99.66 to 102.24)
(97.74 to 99.09)
(85.56 to 87.15)
(78.89 to 80.82)
(88.47 to 90.48)
(85.02 to 86.46)
(91.36 to 92.91)
Male lung cancer agestandardised incidence rates per
100,000 people by region of
England and Wales, 1985-2009
Development of interactive Atlas
Interactive Atlas
Public Engagement
• Working with Sense about Science, a charitable
organisation that aim to improve understanding of science
and evidence - workshops and meetings with representative
target audience
1. Workshop 1: Interpretation of the maps
2. Workshop 2: Chapter content
3. Workshop 3: Online interactive atlas
• Presentations and discussions with the MRC-HPA Centre
Community Advisory Board
Consultation
Why
haven’t
you
done x?
You should
have done (the
whole atlas like)
this
People will overlay
the health and
exposure maps
I’ve no
idea what
this means
Giving people
information on
exposure implies it is
a health risk and is
irresponsible
You will alarm people
who live in an area at
higher risk
Difficult decisions!
Feedback from consultation – 1
Feedback on disease maps:
- One colour ramp
instead of a
divergent colour
ramp
- The same scale to
be used across all
health outcome
maps to allow them
to be comparable
- Defined categories
instead of a
continuous scale
- Provide the highest
and lowest number
of cases
Feedback on interpretation of the
maps:
- Revise text so that
the chapter acts as
an easy reference
guide for interpreting
the maps
- Provide a complete
worked example of
how to interpret the
mpas
- Remove statistical
methods and move to
an appendix
Feedback from consultation – 2
• Traffic light colours: Red=danger
• Displaying uncertainty on maps
• Interpretation
• Interpretation when not
shown on other
environmental exposure
maps
Simplification of a usually complicated subject
Balancing scientific rigour against getting a message across
Public Health Message
• Will the maps tell me if my
area is bad?
Scientific discussion
• Why haven’t you presented
unadjusted maps?
• Will I be able to tell if the
exposure in my area is
giving me cancer?
• Is drinking tap water a risk factor
for bladder cancer in men?
• Everyone becomes an
expert if they think they
understand it!
• You should use different
statistical methods
• It’s of statistical interest only
It’s probably right when nobody is happy!
Conclusions
• Using sophisticated (statistical, cartography, literature review)
techniques doesn’t mean the target audience will understand!
• Output may need to be simplified to be accessible and
meaningful
• Risks oversimplification
• Presenting information to a range of different audiences is
difficult
• Consultation results in a more useful output but
• Takes a lot of time
• Throws up the unexpected
• Needs a thick skin!
SAHSU Atlas Team
Anna Hansell
Lea Fortunato
Ellen McRobie
Clare Pearson
Linda Beale
Daniela Fecht
Helga Laszlo
Oliver Robinson
Lars Jarup
Rebecca Ghosh
Peter Hambly
Kees de Hoogh
Paul Elliott
Federico Fabbri
Kevin Garwood
Exposures
• Air Pollution (NO2, PM10) – Land Use Regression model
(100m x100m grid) using data from National Air Quality
Archive’s
• Radon Potential- Modelled by the HPA and the British
Geological Survey (BGS)
• Sunshine Duration - Meteorological Office, through
MIDAS Land Surface Station Data
• Metals (Cadmium, lead) - Data collected and analysed by
the Centre for Ecology and Hydrology as part of the 2000
Countryside Survey
• Agricultural Pesticides - Data from The Pesticides Usage
Survey, conducted by The Food and Environment Research
Agency
• Chlorination Disinfection By-products – Data from 10
Water companies

similar documents