1. EHS 655 Lecture 7: Exposure grouping strategies

2. Exposure analysis in a nutshell

3. What we’ll cover today Attenuation bias
Sampling issues and temporal trends
Exposure groups
Stata commands

4. ATTENUATION BIAS
Heederik, Attfield, 2000

5. Repeated measurements
Loomis, Kromhout, 2004

6. Ways to reduce attenuation bias
Increase number of exposure measurements per person
Increase between-subject variability so inter-individual exposure range is larger

7. SAMPLING ISSUES Several different sampling strategies
Haphazard/convenience
Worst case
Representative
Random
How exposures were collected may determine what we can infer from them
Also need to consider temporal trends
Question: what are the strengths and weaknesses of each of these approaches?

8. Sampling issues - temporal trends
Heederik, Attfield, 2000

9. Sampling issues – temporal trends
Davies, Teschke, Kennedy, Hodgson, Demers, 2008

10. Individual-level assessment
Exposure varies greatly over time and space
Measure individuals’ exposures
Use repeated measurements on individual to calculate individual average (only their data)
Often considered gold standard approach
Ignacio and Bullock, AIHA, 2006

11. Question
If individual-level approach is gold standard, come up with at least two reasons we might want to create exposure groups?

12. Reasons we might not want individual-level assessments
Even with repeated measurements, measured average level is at best approximation of true exposure
N of repeated measurements per individual usually small (scarcity of data)
Higher within-person variability and/or smaller inter-person variability = more attenuation of exposure-response relationship
Logistical/financial challenges
Lack of direct access to individuals

13. EXPOSURE GROUPS
To address scarce individual data, we commonly pool data across individuals
Create group estimates
Increased amount of data in groups reduces random variability associated with estimate
Apply group estimates to individuals who likely have similar exposures
May not be truly applicable to all individuals in group

14. Group-level assessment
Create subgroups based on common features of exposure
Ideally, all workers within each group measured
Must consider between-group, within-group, within-individual variability
Tielemans, Kupper, Kromhout, Heederik, Houba, 1998

15. Advantages of grouping
Often more effective than individual-level assessment
Especially when temporal variability large
Logistically less demanding than individual approach
Should result in almost unbiased estimated coefficients of exposure-response relationship
Attenuation very small with grouped exposure data

16. Grouping and attenuation
Seixas and Sheppard, 1996

17. Exposure groups Originally “Homogeneous exposure groups” (HEGs)
Current trend is “Similar exposure groups” (SEGs)
Groups frequently defined by common structures
E.g., job title, work area, activity, behavior, agent, street, etc
Individuals in different groups assumed to have different exposures
Individuals within groups also have different exposures
Differences in activities, behaviors, protective equipment

18. Exposure groups
Statistical measures of central tendency often applied to groups
Mean, median, mode
We treat every individual in the group as though they are exposed at the this measure of central tendency
For categorical exposures, we assume groups are exposed (1) or unexposed (0)
Assumes zero variability in groups

19. What do we need to think about when grouping people?
Specificity
Summarizes variance within group
Large variance = highly specific
Precision
Summarizes variance between groups
Large variance = highly precise

20. Exposure grouping goals
Goal 1: create groups which retain true individual differences (specificity)
i.e., within-group variability small compared to between-group variability
Goal 2: create groups that are as large as possible (precision)
Leads to exposure estimates that are more precise than individual worker means

21. Tradeoff between specificity and precision
Individual-based strategy
Precise (small standard error) but biased estimates of exposure-response relationship
Group-based strategy
Unbiased (high validity) but imprecise (large standard error) estimates of exposure-response relationship
Validity, not precision, most important

22. Specificity vs precision (from the reading)
Werner, Attfield, 2000

23. Specificity vs precision
Within-group variation
Between-group variation
Heederik, Attfield, 2000

24. Grouping strategies A priori A posteriori
Grouping takes place before measurement is made
A posteriori
Grouping takes place after measurement is made

25. Ways to create exposure groups
Group by
Activity type
Process
Agent
Exposure pathway
Location
Time
Etc.

26. Example: grouping by agent
Note variation in group size, exposure range, repeated measurements

27. Example: grouping by job title
Lewne, Plato, Bellander, Alderling, Gustavsson, 2010

28. Example - grouping outcomes
Possibilities
Within group variability > between group variability
Between group variability > within group variability
Question: which of these grouping approaches is preferred?
Rappaport, Kromhout, Symanski, 1993

29. Group size and attenuation bias
Reducing attenuation bias
Increasing number of subjects per group can be as effective as increasing number of measurements per subject

30. Question
What do you think of this as a grouping strategy, and why?

31. Resources
EPA ExpoBox

32. On to Stata More basic data manipulation commands
Create dummy variables for groups
example: tabulate varname, gen(newvarname)
Let’s make or identify some groups and compare them

33. On to Stata Bivariate analysis examples tabulate varname1 varname2
tab2 varname1 varname2 varname3
tabstat varname, stat(mean sd count)
bysort varname1: tabstat varname2, stat(mean sd count)
table varname1, contents(mean varnamex sd varnamex) by(varname2)
twoway scatter varname1 varname2
Graph matrix varname1 varname2 varname3, half
Graph box varname1, over(varname2)