EHS 655 Lecture 8 Exposure grouping metrics and

EHS 655 Lecture 8: Exposure grouping, metrics, and a bit more on data cleaning

What we’ll cover today o A bit more on grouping exposures o Exposure metrics o A bit more on data cleaning o Stata 2

How can we quantify grouping effectiveness? o Look at variance within and between groups o One summary measure is referred to as contrast (ability to create groups with different exposures) o Contrast 3

Contrast o Grouping affects standard error of exposure-response slope very adversely if low contrast between groups n o Ratio of within-group to between-group variance >1 leads to low contrast So, grouping schemes must be optimized n A priori grouping by job may not address factors that actually affect exposure 4

Grouping – contrast Niewenheijsen 1997 5

Job exposure matrix o Large-scale grouping strategy o Essentially a giant table with 3 axes n n n o Agent Exposure Time Often used in large population-based studies n Few or no measurements on study subjects 6

Simple job exposure matrix example http: //depts. washington. edu/fmrwrkr/needs 1. 1_v 1/Image 81. gif 7

More complex job exposure matrix example http: //oem. bmj. com/content/62/4/272/F 3. large. jpg 8

Grouping by exposure AIHA, Strategy for Assessing and Managing Occupational Exposures, 3 rd ed, 2006 9

EXPOSURE METRICS o How we quantify/define exposure WR Ott, AC Steinemann, LA Wallace, 2007 10

Examples of exposure metrics Nieuwenhuijsen, Lowson, Venables, Newman Taylor, 199511

Examples of exposure metrics (applied to noise) o Average exposure Where T is total duration t is an interval within duration N is total number of intervals LA is A-weighted noise level Note: this is the level in the Work. LEQd. BA variable Chapter 3, AIHA Noise Manual, 2003 12

Average vs. Time-Weighted Average (TWA) o TWA normalizes measurements to a standard exposure duration (typically 8 hours) Where LA 8 hn = 8 -hour equivalent exposure T is duration of exposure (in hours) Note: you have duration of exposure in minutes Chapter 3, AIHA Noise Manual, 2003 13

Why convert average to TWA? o o Allows for direct comparison of exposures measured over different periods For example: n n Both of these exposures have the same LEQ average However, TWA for right-hand exposure would be 2 Xas high, since monitoring duration is 2 X as long 14

Peakiness metrics o Ratio of peak or maximum to average? Peak Average Moretto A, Handbook Clin Neurol, 2015 15

Example of exposure metrics (applied to noise) o Ratio exposure (peakiness) o LMax/LEQ ratio = log 10 10(LMAX/10) 10(LEQ/10) o Note exponentiation, i. e. 10(LMAX/10) = 10^(LMAX/10) o Provides summary of highest exposure during measurement to average exposure during measurement ( ) 16

Example relationships between exposure metrics Nieuwenhuijsen, Lowson, Venables, Newman Taylor, 1995 17

Relationship of annual exposure metrics by person Question: do we want these to be highly or poorly correlated? Seixas, Neitzel, Sheppard, Goldman, 2004 18

Metrics – agent of interest Friesen et al 2007 19

Metrics – average and variability Seixas, Neitzel, Stover, Sheppard, Feeney, Mills, Kujawa, 2012 20

Combination of exposure groups and exposure metrics Seixas, Neitzel, Sheppard, Goldman, 2004 21

Metrics and exposure limits AIHA, Strategy for Assessing and Managing Occupational Exposures, 3 rd ed, 2006 22

A BIT MORE ON DATA CLEANING o Example of how to treat missing data Davies, Teschke, Kennedy, Hodgson, Demers, 2008 23

Example: data cleaning Seias, Neitzel, Sheppard, Goldman, 2004 24

On to Stata o New data this week in dataset v 4 25

On to Stata o What exposure metrics can we look at in our data? o Average and TWA n n o Maximum n o How to compute TWA? By individual? By group? Peakiness metric? n n How to compute peakiness? Maximum / average? 26

On to Stata o TWA = LEQ + log 10(Runtime/480) o Peakiness = 10 x log 10(10^(LMax/10) / 10^(LEQ/10)) 27

On to Stata o Bivariate analysis examples n n n n tabulate varname 1 varname 2 tab 2 varname 1 varname 2 varname 3 tabstat varname, stat(mean sd count) bysort varname 1: tabstat varname 2, stat(mean sd count) table varname 1, contents(mean varnamex sd varnamex) by(varname 2) twoway scatter varname 1 varname 2 Graph matrix varname 1 varname 2 varname 3, half Graph box varname 1, over(varname 2) 28