Tracking Pilot Project - 1

1
Rapid Assessment and Characterization of
Environmental Risks (RACER)

• Academic Partners for Excellence in
  Environmental Public Health Tracking (APEX)
• Webinar
• April 25, 2008

2
Investigators

• Gary M. Marsh, Ph.D.
• Professor of Biostatistics
• Director, Center for Occupational
• Biostatistics and Epidemiology
• Michael Cunningham, M.S.
• Research Specialist IV

3
Premise

• Two primary objectives of the EPHTN
• to identify populations at risk and respond to
  outbreaks, clusters, and emerging threats
• to establish the relationship between
  environmental risks and disease
• From Americas Environmental Health Gap Why
  the Country Needs a
• Nationwide Health Tracking Network,
  Pew Environmental Health
• Commission and Johns Hopkins
  University Dept. of Health Policy and
• Management, September 2000.

4
Rationale

• Relatively little research has focused on extent
  to which inherent limitations of environmental
  investigations impact on our ability to identify
  and characterize environmental risks
• Few software tools available that focus on
  epidemiological / statistical aspects of
  environmental investigations

5
Goals

• Goal 1 - to help fill the research gap by
  evaluating the sensitivity of some commonly used
  statistical methods for identifying and
  characterizing risks
• Goal 2 - To provide a web-based software tool
  that will enable public health officials and
  others to evaluate feasibility of conducting
  health tracking activities or targeted
  investigations
• Rapid Assessment and Characterization of
• Environmental Risks (RACER)

6
Commonly Encountered Environmental Exposure
Scenarios

• Scenario 1
• Source of environmental exposure is identified
• Proportion of population exposed and
  exposure-level health risks are known or
  estimable
• Scenario 2
• Suspected source of environmental exposure
• A number of health outcomes are observed in a
  population over time
• Scenario 3
• Suspected source of environmental exposure
• Exposure level-specific populations and
  associated risks are neither known nor estimable

7
Components of Rudimentary Empirical Model

• A community population of specified size resides
  around point source of environmental exposure
• Specify background rate of any disease of
  interest
• Specify exposure-response relationship for
  environmental risk
• Divide population into subpopulations by distance
  from exposure source and specify proportion of
  subpopulations exposed

8
Components of Rudimentary Empirical Model
(contd)

• Use relative measure of effect to compare
  observed events in exposed subpopulations to
  number expected had background rates prevailed
• User can specify representative range of values
  for basic determinants of statistical
  characteristics
• Use empirical models to examine sensitivity of
  measures such as statistical power and MDRR for
  set of realistic environmental exposure scenarios

9
Reference and Potentially Exposed Subpopulations
Exposure Point Source at Center
10
Airborne Exposure to Point Source with Proportion
of Subpopulations Exposed Downwind of Source
11
Groundwater Exposure to Point Source with
Proportion of Subpopulations Exposed Downstream
of Source
12
Scenario 1 Known Information

• Reference population
• Background rates used to calculate the expected
  cases
• Potentially exposed population
• Population size with given age distribution
• Proportion exposed in each sub-region
• Environmental exposure
• Exposure level-specific relative risks

13
Scenario 1 Estimated by RACER

• Statistical power to observe a specified, true
  elevation in the background rate over given range
  of observation times
• Minimum time required to achieve a given power
  assuming constant exposure risk and population
  over time

14
Scenario 1 Hypothetical Example

• Residents of Town A have been exposed to ionizing
  radiation
• Assume a 75 increased incidence in thyroid
  cancer at the levels of radiation that were
  measured in the community
• The exposure boundary encompasses the entire
  community and the radiation levels are constant
  throughout the exposed region
• The size of the population living within the
  exposed area is known
• Assume all residents are exposed

15
Scenario 1 Hypothetical Example
16
Scenario 1 Hypothetical Example

• Question
• If the residents of Town A have been exposed to
  these levels of ionizing radiation, how can this
  exposure be characterized in terms of the ability
  to detect a true 1.75-fold increase in thyroid
  cancer incidence?

17
Scenario 1 Known Information

• Population distribution of Town A in 2007, by age
  group (only interested in the five listed age
  groups)

Age Group Potentially Exposed Population
21-34 1110
35-49 1445
50-59 2374
60-69 2219
70 852
Total 8000
18
Scenario 1 Known Information

• 100 of the town residents live in the potential
  exposure region
• 100 of those living in the potential exposure
  region are assumed to be exposed
• Exposure to the ionizing radiation at the
  observed levels is known to inflate the baseline
  thyroid cancer incidence by a factor of 1.75

19
Scenario 1 Reference Rates

• Baseline (background) thyroid cancer incidence
  rates in general population (SEER)

Age Group Baseline incidence rate (per 100,000)
21-34 5.6
35-49 5.6
50-59 6.8
60-69 7.5
70 8.4
20
Scenario 1 Additional Assumptions

• Increased exposure risk is constant across age
  groups and through time
• No latency
• No adjustment for in- or out-migration or
  mortality

21
Scenario 1 RACER on the Web
22
Scenario 1 Start Menu
23
Scenario 1 Input Screen 1
24
Scenario 1 Input Screen 2
25
Scenario 1 Input Screen 3
26
Scenario 1 Input Screen 4
27
Scenario 1 RACER Calculates Expected Cases
28
Scenario 1 Input Screen 6
29
Scenario 1 Summary Options
30
Scenario 1 Output Option 1
31
Scenario 1 Output Option 2 Table
32
Scenario 1 Output Option 2 Plot
33
Scenario 1 Output Option 3 Table
34
Scenario 1 Output Option 3 Plot
35
Scenario 1 Hypothetical ExampleKey Results

• After the first year of exposure, there is only
  7 power to detect a true excess risk of 1.75 in
  this population.
• After 10 years, the inflation of the background
  risk due to the exposure should result in enough
  cases to achieve 50 power.
• The power to observe a true exposure elevated
  risk of 1.75 reaches 80 after 30 years.

36
Scenario 2 Known Information

• Scenario 2
• Suspected source of environmental exposure
• A number of health outcomes are observed in a
  population over time
• Reference population
• Background rate used to calculate the expected
  cases in exposed population
• Exposed population
• Observed cases (health outcome of interest)
• Population size with given age distribution

37
Scenario 2 Estimated by RACER

• The ratio of observed cases to expected had
  background rates prevailed on exposed population
  (SMR, SIR, etc.)
• The number of additional cases and observation
  time required to achieve desired power

38
Scenario 3 Known Information

• Scenario 3
• Suspected source of environmental exposure
• Exposure level-specific populations and
  associated risks are neither known nor estimable
• Exposure elevated risks or observed number of
  cases (health outcomes) in the population of
  interest are unknown
• Elevated risk due to environmental exposure is
  suspected

39
Scenario 3 Estimated by RACER

• Given a fixed proportion of individuals exposed
  in a region, plot relationship of power and the
  observed/expected ratio over time for a range of
  exposure elevated relative risks
• Given a fixed exposure elevated relative risk,
  plot relationship of power and the percentage of
  exposed individuals over time

40

• Scenario 3 Hypothetical Example
• Showing 2-D Contour Plots
• Percent of Population Exposed is Fixed,
• Observation Time and Exposure-Elevated
• Relative Risks (RRee) Vary

41
Scenario 3 Hypothetical Example
Observed/Expected Power at Fixed Exposed
42

• Scenario 3 Hypothetical Example
• Showing 3-D Contour Plots
• Exposure-Elevated Relative Risk (RRee) is Fixed,
• Percent of Population Exposed and
• Observation Time Vary

43
Power as a Function of Exposed (ew) and Time
(years)Comparison of 4 Plots
RRee 1.2
RRee 1.4
RRee 1.6
RRee 1.8
44
Possible Future RACER Enhancements

• Account for geographic population mobility
• Extend standardization to other factors (e.g.,
  race, sex)
• Time dependent exposure-elevated relative risks
• Account for latency/incubation period of health
  endpoints
• Case-crossover power and sample size module

45
Acknowledgements

The investigators wish to thank the following
groups for their support and guidance Centers
for Disease Control and Prevention Environmental
Public Health Tracking Program Grant
5U19EH000103-02 (Academic Partners for
Excellence in Environmental Public Health
Tracking) University of Pittsburgh Academic
Consortium for Excellence in Environmental Public
Health Tracking (UPACE-EPHT) (a
collaboration with Drexel University)
46
Questions?