A Model to Systematically Employ Professional Judgment in the Bayesian Decision
Analysis for a Semiconductor Industry Exposure Assessment
Supplement
TABLE SI. Table of values used in the structured QLRA to determine Health Effect rating
for each chemical agent.
LD-50 (mg/kg)
4 HR. LC-50 (ppm)
Practically non-toxic; nonirritating; no chronic effects
5000-15,000
10,000-100,000
1
Slightly toxic; slight to mild
irritation; minor reversible
injury
500-<5000
1000-<10,000
2
Moderately toxic; moderate
irritation; temporary reversible
injury
50-<500
100-<1000
3
Highly toxic; severely irritating;
corrosive; irreversible injury
without prompt medical
attention
1-<50
10-<100
4
Extremely toxic; lifethreatening, permanent
damage may result from single
or repeated exposure
<1
<10
Health Effect
Rating*
0
*=
Toxicity Description
For chemical agents with data indicating potential human carcinogenicity, a health effect rating of 3 or 4 was
assigned. These chemical agents include IARC Group 1 (carcinogenic to humans) and 2A (probably
carcinogenic to humans); NTP "known carcinogens"; ACGIH category A1 (confirmed human carcinogen)
and A2 (suspected human carcinogen); and chemicals for which the data support sufficient evidence of
carcinogenicity in humans (or limited evidence of carcinogenicity in humans, combined with sufficient
evidence in experimental animals). Known reproductive chemical agents were assigned a 4 rating and
suspect reproductive agents were assigned a 3 rating.
TABLE SII. Table of values used in the structured QLRA to determine Dispersion rating
for each chemical agent.
Dispersion
Rating*
1
Vapor Pressure
(@20 – 25⁰ C)
Particulate Description
<25 mmHg
Bound material
2
25-<100 mmHg
Large non-airborne
3
100-250 mmHg
Potentially airborne/inhalable
4
>250 mmHg
Fine light airborne/inhalable
TABLE SIII. Table of values used in the structured QLRA to determine Implemented
Exposure Controls rating for each chemical agent.
Control
Rating*
0
1
2
3
4
Description of Implemented Exposure Controls
Closed system; minimal potential for release to work environment
Semi-closed system; release potential at identified points; effective engineering controls
in place at identified points
Open system; effective engineering controls in place to contain/remove airborne
contaminants
Open system; some degree of engineering controls; effective use of administrative and
PPE controls
Open system; ineffective controls
TABLE SIV. Table of values used in the structured QLRA to determine Frequency and
Duration rating for each chemical agent based on related task.
*
Duration
<15 mins
15 mins – <1 hr
1 hr - <2 hrs
2 hrs – 4 hrs
>4 hrs
<1/month
1
1
1
1
1
Frequency
<daily/>monthly
1
2
2
3
4
daily
1
2
3
4
5
TABLE SV. Table of values used in the structured QLRA to determine Overall Exposure
Potential rating for each chemical agent.
Product of Dispersion, Controls and
Frequency/Duration Ratings
<10
10 – 20
21 – 30
31- 40
>40
Overall Exposure Potential Rating
0
1
2
3
4
Example BDA calculation
The following example illustrates how BDA was implemented. The chemical agent of interest is
isopropyl alcohol. The work task and SEG considered is normal operations in Line B Photo
process area. The OEL is 200 ppm isopropyl alcohol as an 8-h TWA.
For this work task and SEG, historical exposure monitoring data are available. Historical data
summary statistics:
Median Sample
(ppm)
size
0.52
18
Min
(ppm)
<0.042
Max
(ppm)
0.82
95%
UCL
(ppm)
6.92
% censored
38.9
Sub method
MLE
Category
(95%
UCL)
1
Given that historical data are available, a data-informed prior is developed by BDA. The datainformed prior is developed using BDA with an uninformed prior and the historical data. The
posterior distribution is then used as the data-informed prior. The BDA decision chart is:
The same information is displayed in tabular form:
Prior
type
Flat
0-De
minimis
(<1%
OEL)
0.463
1-Highly
Controlled
(1-10%
OEL)
0.529
2-Well
Controlled
(10-50%
OEL)
0.006
3Controlled
(50-100%
OEL)
0.001
4-Poorly
Controlled
(>100%
OEL)
0.001
Category
(BDA)
1
The prior distribution is updated with newly collected exposure monitoring data. The summary
of these data is:
Median Sample
(ppm)
size
0.017
3
Min
(ppm)
<0.02
Max
(ppm)
<0.05
95%
UCL
(ppm)
0.52
% censored
100
Sub method
LOD/2
Category
(95%
UCL)
0
Note that because 100% of results are censored, the results are replaced by one-half the limit of
detection (LOD) for use in BDA. These data are used to define the likelihood distribution in
BDA. The BDA software summary of the likelihood distribution is presented as a chart:
The same information is displayed in tabular form:
Prior
type
Datainformed
0-De
minimis
(<1%
OEL)
1-Highly
Controlled
(1-10%
OEL)
2-Well
Controlled
(10-50%
OEL)
3Controlled
(50-100%
OEL)
4-Poorly
Controlled
(>100%
OEL)
Category
(BDA)
0.97
0.029
0.001
0
0
0
The BDA software then calculates the posterior distribution. The posterior distribution is
presented as a chart:
The same information is displayed in tabular form:
Prior
type
0-De
minimis
(<1%
OEL)
1-Highly
Controlled
(1-10%
OEL)
2-Well
Controlled
(10-50%
OEL)
3Controlled
(50-100%
OEL)
4-Poorly
Controlled
(>100%
OEL)
Category
(BDA)
Datainformed
0.967
0.033
0
0
0
0
The highest probability is in category 0, de minimis. Category 0 is the event that the 95th
percentile of the lognormal distribution of exposure is < 1% of the OEL. The probability that the
95th percentile of the lognormal distribution of exposure is in category 0 is 0.967. Using the
decision rule that the category with the highest probability is the most-likely category, this
analysis indicates that most-likely exposure category is 0, de minimis.