^^ili^l^^^^.^ -O- ': ,
-------
EPA 560/5-83-003
June 1983
AIRBORNE ASBESTOS LEVELS IN SCHOOLS
by
\
Paul C. Constant, Jr.
Fred J. Bergman
Gaylord R. Atkinson
Donna R. Rose
Midwest Research Institute
EPA Contract No. 68-01-5915
and
Donna L. Watts
Everett E. Logue
Ty D. Hartwell
Research Triangle Institute
EPA Contract No. 68-01-5848
and
Bertram P. Price
Julius S. Ogden
Battelle Columbus Laboratories
Dr. Frederick W. Kutz, Project Officer, Contract No. 68-01-5915
Dr. Joseph J. Breen, Task Manager, Field Studies Branch
Mr. Joseph S. Carra, Project Officer, Contract No. 68-01-5848
Ms. Cindy R. Stroup, Task Manager, Design and Development Branch
U.S. Environmental Protection Agency
Office of Pesticides and Toxic Substances
Office of Toxic Substances
Exposure Evaluation Division
401 M Street, S.W.
Washington, D.C. 20460
-------
DISCLAIMER
This document has been reviewed and approved for publication
by the Office of Toxic Substances, Office of Pesticides and Toxic
Substances, U.S. Environmental Protection Agency, according to the
Agency's peer review system. The use of trade names or commercial
products does not constitute Agency endorsement or recommendation
for use.
-------
PREFACE
This report presents the results of the collaborative effort of two
primary contractors for the Exposure Evaluation Division (EED) of the Office
of Toxic Substances: Midwest Research Institute (MRI) under EPA Contract No.
68-01-5915 and Research Triangle Institute (RTI) under EPA Contract No. 68-
01-5948. MRI's overall responsibility was program management and sampling
and analysis. RTI's overall responsibility was statistical design and data
analysis. Support in field sampling, electron microscopy, and statistical
analysis was provided under subcontract by Battelle Columbus Laboratories
(BCL); and in quality assurance analyses of field samples under subcontract
by Illinois Institute of Technology (IITRI) and Colorado School of Mines
Research Institute (CSMRI).
The work reported on herein is of a task under the EPA program entitled
"Sampling and Analysis of Selected Toxic Substances." This program is under
the leadership of Dr. John E. Going and was under the overall management of
Dr. James L. Spigarelli, Director of the Analytical Chemistry Department of
Midwest Research Institute. Mr. Steven R. Williams was RTI's program manager.
The MRI work was conducted under the technical management of Mr. Paul C.
Constant, Jr. He was assisted by Mr. Fred J. Bergman, who supervised field
sampling activities. Ms. Teresa Costello, Ms. Kim Cowherd, Ms. Marilyn Gabriel,
Mr. Mike Kalinoski, Ms. Lily Leong, Mrs. Donna R. Rose, and Ms. Carolyn Winter
were MRI's field personnel. Mr. Bergman and Mrs. Rose were field crew leaders.
Mr. Gaylord R. Atkinson was responsible for the analysis of the bulk samples.
He was assisted by Mrs. Rose. Dr. Dennis Takade was the quality assurance
coordinator and also performed the sample and traceability audit. Mr. John
Hosenfeld assisted Dr. Takade. The BCL subcontract with MRI was managed by
Dr. Charles W. Townley. The late Mr. Eric Schmidt was responsible for BCL's
field sampling. His field samplers were Mr. Curtis Bridges, Mr. Gregory L.
Headington, Mr. Salo E. Miller, and Ms. Amanda Bush Osburn. Mr. Julius S.
Ogden was responsible for the analysis of the air samples. He was assisted
by Mrs. Sandra J. Anderson, Mrs. Carolyn F. Dye, Mrs. Irene E. Green,
Mr. Doyle F. Kohler, and Mr. Carl W. Melton. Dr. Bertram P. Price provided
the statistical assistance at BCL.
The RTI work was conducted under the supervision of Dr. Tyler D. Hartwell,
who also was responsible for the statistical analysis. He was assisted with
the statistical analyses by Dr. Donna L. Watts and Dr. Everett E. Logue.
Dr. Watts was responsible for the survey design and selection of sites. She
was assisted by Ms. Denise Melroy. The statistical data file audit was per-
formed by Ms. Debbie Whitehurst.
The rater team that inspected the sampling sites to obtain consensus
ratings of specific aspects of the asbestos-containing material and the site
included Dr. Joseph J. Breen and Ms. Cindy R. Stroup of EED, Dr. Everett Logue
of RTI, and two persons from the school district surveyed.
111
-------
CONTENTS
Preface iii
List of Figures vii
List of Tables ix
Acknowledgments xv
Executive Summary xvii
1. Introduction 1
2. Conclusions 3
3. Overall Quality Assurance Program 5
Organization 5
Operation 5
Protocols 6
Quality assurance data 6
Audits 7
Site-specific ratings 7
4. Sample Design and Selection 9
Study population 10
Sample design 11
5. Field Survey 15
Air sampling 15
Sampling system 16
Field operations 16
Sample handling 18
Bulk sampling 18
Sample collection 19
Sample handling 19
Chain of custody 19
Site-specific ratings 19
6. Sample Analysis 21
Air samples 21
Method of analysis 21
Discussion 31
Quality assurance 32
Bulk samples 39
Method of analysis 39
Discussion 40
Quality assurance 54
7. Statistical Analysis 57
Analysis methods 58
Airborne chrysotile concentration at asbestos-containing
material sites, control sites, an3 outdoor ambient sites. 59
Exposure levels 59
Schoolwide elevated levels 62
Levels of the assessment factors 62
Algorithm score 72
Bulk chrysotile levels 75
Other major components 75
Releasability 78
Other covariables 80
Long-term variability 85
-------
CONTENTS (continued)
Regression analyses 87
Alternative classification analyses 99
Comparison of raters 112
References 117
Appendixes
A. Design considerations and sampling scheme 121
B. Protocol for air sampling 147
C. Protocol for creating and maintaining chain of custody 151
D. Protocol for the sampling and analysis of insulation material
suspected of containing asbestos 155
E. Analytical protocol for air samples 159
F. Friable material site-specific rating forms 163
G. Classification analyses based upon the 50th percentile
(84 ng/m3) of the airborne chrysotile distribution 167
VI
-------
74
LIST OF FIGURES
Number Page
1 Air sampling system 17
2 TEM analytical precision fo airborne asbestos concentration:
coefficient of variation versus mean (fibers/cc) 37
3 Releasability ratings and the range of the ratings of bulk
samples from each of the 48 sites sampled 53
4 Boxplot of airborne chrysotile concentrations (ng/m3) for
asbestos-containing material sites, indoor control sites,
and ambient sites (population estimates) 61
5 Airborne chrysotile concentration (ng/m3) versus the consensus
score rating for the 48 asbestos-containing material sites . .
6 Airborne chrysotile concentration (ng/m3) versus average bulk
chrysotile percentage at 48 asbestos-containing material
sites 76
7 Airborne chrysotile concentration (ng/m3) by other major
components of bulk samples at 48 asbestos-containing
material sites 77
8 Airborne chrysotile concentration (ng/m3) by average bulk
chrysotile percentage and other major component at
48 asbestos-containing friable material sites 79
9 Airborne chrysotile concentration (ng/m3) versus releasability
categories at 48 asbestos-containing material sites 81
10 Plot of airborne chrysotile concentration (ng/m3) versus new
score (releasability substituted for friability) for
asbestos-containing material friable sites 83
11 Log (ng/m3 airborne chrysotile + 1.0) versus the predicted
value from Model II (original algorithm factors, with
releasability replacing friability) for asbestos-containing
friable material sites 94
12 Airborne chrysotile concentration (ng/m3) versus the predicted
value from Model II (original algorithm factors, with
releasability replacing friability) for asbestos-containing
friable material sites 95
13 Log (ng/m3 airborne chrysotile + 1.0) versus the predicted
value from Model III (releasability, cleaning, and water
damage) for asbestos-containing friable material sites .... 96
VII
-------
LIST OF FIGURES (continued)
A-l Hypothesized relationship between air level and algorithm score. 126
A-2 Original asbestos exposure assessment decision tree 127
A-3 Hypothetical distribution of airborne asbestos concentration
levels 129
D-l Procedure for analysis of asbestos materials 157
Vlll
-------
LIST OF TABLES
Number Page
1 Distribution of Eligible Sampling Areas in the School District
with Respect to Asbestos Content, Friability, Condition,
Exposure, and Accessibility 12
2 Results of Analysis of Airborne Asbestos Fibers Collected by
Filtration and Analyzed by Transmission Electron Microscopy:
Chrysotile Concentration in ng/m3 22
3 Results of Analysis of Airborne Asbestos Fibers Collected by
Filtration and Analyzed by Transmission Electron Microscopy:
Amphibole Concentration in ng/m3 24
4 Fundamental Data Obtained from the Analysis of the Airborne
Asbestos Fiber Collected by Filtration and Analyzed by
Transmission Electron Microscopy 26
5 The Number of Chrysotile Bundles and Clusters Observed on the
Filter but Not Used in the Mass Calculations 33
6 Quality Assurance Results of Transmission Electron Microscopy
Analysis 35
7 Quality Assurance Results of Duplicate Sample Analysis for
Chrysotile 36
8 Quality Assurance Results of Field Blank Analysis for
Chrysotile 38
9 Quality Assurance Results of Laboratory Blank Analysis for
Chrysotile 39
10 Results of Polarized Light Microscopic Analysis of Bulk Samples
for Volume of Chrysotile, Size of Fibers, Coating on Fibers,
Releasability Rating of Material, and Nonasbestos Components . 41
11 Variance Components for PLM Measurement of Bulk Samples 55
12 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites, Control Sites, and
Ambient Sites 60
13 Schools Where Airborne Chrysotile Concentration at Indoor
Control Sites Appears To Be Elevated Above Concentration at
Outdoor Ambient Site 63
14 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Condition of Material . . 64
IX
-------
LIST OF TABLES (continued)
Number Pa££
15 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Accessibility of
Material 65
16 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Airstream Status 66
17 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Proportion of
Material Exposed 67
18 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Water Damage of Material. 68
19 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Activity 69
20 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Friability of Material. . 70
21 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Chrysotile Content of
Material 71
22 Summary of Relationships Between Airborne Chrysotile
Concentration (ng/m3) and Original Algorithm Factors for
Asbestos-Containing Friable Material Sites 71
23 Exposure Assessment Algorithm Scores for the Asbestos-
Containing Material Sites in the School District 73
24 Bulk Chrysotile Content at the Asbestos-Containing Material
Sites 75
25 Bulk Chrysotile Content at the Asbestos-Containing Material
Sites by Other Major Components 78
26 Distribution of the Asbestos-Containing Material Sites
with Respect to Friability and Releasability (Population
Estimates) 82
27 Candidate Covariables 84
28 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
Containing Friable Material Sites by Cleaning Category .... 85
29 Variability of Airborne Chrysotile Concentration (ng/m3) over
Time at Three Asbestos-Containing Friable Material Sites and
the Corresponding Control and Ambient Sites 86
-------
LIST OF TABLES (continued)
Number Page
30 Summary of Unweighted Regression Models: Relationship of
Surrogate Measurements to Airborne Chrysotile Concentration. . 89
31 Unweighted Regression Models: Relationship of Surrogate
Measurements to Airborne Chrysotile Concentration 91
32 Weighted Survey Regression: Relationship of Surrogate
Measurements to Airborne Chrysotile Concentration 92
33 Summary of Unweighted Regression Models: Relationship of
Surrogate Measurements (Excluding Cleaning Practices) to
Airborne Chrysotile Concentration 97
34 Three-Variable Regression Models: Relationship of Surrogate
Measurements (Excluding Cleaning Practices) to Airborne
Chrysotile Concentration 98
35 Predicting Low/High Airborne Chrysotile Concentration with a
Discriminant Function Based upon Seven Original Algorithm
Factors 100
36 Summary of Discriminant Analyses 101
37 Predicting Low/High Airborne Chrysotile Concentration with
the Algorithm Score Dichotomy 103
38 Predicting Low/High Airborne Chrysotile Concentration with
a Releasability Dichotomy 104
39 Predicting Low/High Airborne Chrysotile Concentration with
Three Regression-Based Dichotomies 105
40 Predicting Low/High Airborne Chrysotile Concentration with
Other Regression-Based Dichotomies 106
41 Predicting Low/High Airborne Chrysotile Concentration with
the Original Decision Tree Based Upon Proportion of Material
Exposed, Material Condition, Accessibility, and Friability . . 107
42 Predicting Low/High Airborne Chrysotile Concentration with a
Decision Tree Based Upon Releasability and Water Damage. . . . 108
43 Predicting Low/High Airborne Chrysotile Concentration with a
Decision Tree Based Upon Releasability and Airstream Status. . 109
44 Predicting Low/High Airborne Chrysotile Concentration with a
Decision Tree Based Upon Releasability and Bulk Sample
Chrysotile Content 110
XI
-------
LIST OF TABLES (continued)
Number gage
45 Predicting Low/High Airborne Chrysotile Concentration with a
Decision Tree Based Upon Releasability, Water Damage (or
Material Condition), and Airstream Status Ill
46 Classification Analysis Summary (Airborne Chrysotile
Concentration 75th Percentile Dichotomy) 113
47 Rater Consistency:Disagreement with Consensus Frequencies. . . . 115
48 Analysis of Variance for Disagreement Proportions 116
A-l Algorithm Factors and Their Weighted Scores 124
A-2 Site Description for Decision Tree Algorithm 128
A-3 First-Stage Strata, Constructed from Asbestos Content and
Friability Categories 130
A-4 Allocation of Sample Schools Among First-Stage Strata 131
A-5 Probabilities of Selection for Sample Schools 133
A-6 Number of Eligible Asbestos-Containing Material Sites on
Second-Stage Frame 134
A-7 Description of Friability/Condition/Exposure/Accessibility
Categories 135
A-8 Allocation of Second-Stage Sample of Asbestos-Containing
Material Sites Among Schools and Friability/Condition/
Exposure/Accessibility Categories 137
A-9 Asbestos-Containing Material Sites and Selection Probabilities
in Second Stage of Selection 138
A-10 Summary of Control Site Selection 142
A-11 Selection of Long-Term Sampling Sites: Stratum 1 143
A-12 Selection of Long-Term Sampling Sites: Stratum 2 143
A-13 Selection of Long-Term Sampling Sites: Stratum 3 144
G-l Predicting Low/High Airborne Chrysotile Concentration with
Three Regression-Based Dichotomies 168
G-2 Predicting Low/High Airborne Chrysotile Concentration with the
Original Decision Tree Based upon Proportion of Material
Exposed, Material Condition, Accessibility, and Friability . . 169
xii
-------
LIST OF TABLES (continued)
Number
G-3 Predicting Low/High Airborne Chrysotile Concentration with a
Decision Tree Based upon Releasability and Water Damage. . . . 170
G-4 Predicting Low/High Airborne Chrysotile Concentration with a
Decision Tree Based upon Releasability and Airstream Status. . 171
G-5 Predicting Low/High Airborne Chrysotile Concentration with a
Decision Tree Based upon Releasability and Bulk Sample
Chrysotile Content 172
G-6 Predicting Low/High Airborne Chrysotile Concentration with a
Decision Tree Based upon Releasability, Water Damage (or
Material Condition), and Airstream Status 173
XI11
-------
ACKNOWLEDGMENTS
The helpful technical guidance and dedicated administrative support of
Dr. Joseph J. Breen, Ms. Cindy R. Stroup, Mr. Larry E. Longanecker,
Mr. Joseph S. Carra, Dr. Frederick W. Kutz, and Mr. Martin P. Halper of EPA
are gratefully acknowledged. The close cooperation and unwavering support of
the school district's asbestos committee and administrative staff, and the
principals, teachers, and custodians of the schools surveyed are also appre-
ciated. To all parties who contributed to a very complex cooperative effort
successfully accomplished under stringent time and cost constraints, we are
grateful.
xv
-------
EXECUTIVE SUMMARY
The U.S. Environmental Protection Agency's Office of Toxic Substances
has an ongoing asbestos-in-schools program. As part of this program, a study
was conducted in a large, urban U.S. school district to: (1) document expo-
sures to airborne asbestos resulting from friable, asbestos-containing mate-
rials in schools and (2) validate existing exposure assessment tools (includ-
ing the algorithm) and determine relationships between asbestos air levels
and individual or combined assessment factors. The study covered five major
areas of work: the development of a survey design, the design of a quality
assurance plan, the execution of a field survey, the chemical analysis of
field samples, and the statistical analysis and interpretation of data.
The survey plan was based on a sample design that allows valid inferences
to be made for the field study area concerning airborne asbestos levels and
the relationships between these levels and the assessment tools or individual
factors and combinations. The study sites were randomly selected from all
the student activity areas of schools in the school district that contained
asbestos materials. Data concerning auxiliary variables such as room dimen-
sions, floor covering (carpeted, tile, wood), cleaning practices, air condi-
tioning, and room temperature were collected in addition to air and bulk
samples.
A detailed statistical analysis of the data collected was undertaken.
The main intent of this analysis was first to document the asbestos air levels,
bulk asbestos levels, assessment factors, and algorithm scores in the school
district and then to examine various relationships. In particular, it was of
interest to determine how well airborne chrysotile levels could be predicted
by the assessment factors, singly and in combination.
The quality assurance plan covered all aspects of the study: the objec-
tive of the plan, personnel and their responsibilities, and field and labora-
tory compliance requirements such as standard operating procedures, protocols,
monitoring, documentation, and communications. Protocols were prepared and
followed on air sampling, sampling and analysis of insulation material sus-
pected of containing asbestos, chemical analysis of air samples, and chain of
custody of samples. External quality assurance analyses were performed on
both air and bulk samples. The results of complete audits performed on all
data were satisfactory.
A total of 48 asbestos-containing sites in 25 schools were sampled for
airborne asbestos. The air was monitored for five consecutive days at each
school while students were present. At each school, an outdoor ambient sam-
ple, an indoor control sample from an area without asbestos-containing mate-
rial, and up to four indoor samples from student activity rooms that contained
asbestos material were collected simultaneously.
xvi i
-------
The total of 116 air samples and 9 blank filter samples were collected.
They were analyzed by transmission electron microscopy (TEM) techniques for
asbestos fiber concentration. The mean airborne asbestos levels for the three
types of sites studied are:
Site
Mean Standard error
(ng/m3) of the mean
Asbestos-containing material
Indoor control
Outdoor ambient
179
53
6
42
20
4
The statistical test of difference in air level means among the three types
of sites is significant at the level < 0.01.
To investigate the variability of airborne asbestos with time, 3 of the
48 sites were selected for long-term study (three consecutive school weeks).
There is evidence of substantial variation over time in airborne asbestos
levels in the same room. For the current data, this variation was approxi-
mately 100% (i.e., the coefficient of variation over 3 weeks = standard
deviation/mean = 100%).
Three bulk samples were collected from each of the 48 asbestos-containing
sites. In addition, one double sample (taken side-by-side) was collected at
each site to provide a duplicate for quality assurance. A total of 192 bulk
samples were collected for fiber analysis by polarized light microscopy (PLM)
techniques. The asbestos (chrysotile) content ranged from 5 to 63% with a
mean of 16%.
In addition to standard PLM analysis, the bulk samples were examined for
physical characteristics that could be of importance. There were three major
matrix components (perlite, vermiculite, and glass wool), three dominant fiber
sizes (fine, medium, and coarse), and two fiber coat categories (coated and
not coated). Based on this analysis, a fiber releasability rating system was
developed. Bulk samples were rated according to how readily the asbestos
fibers would be released from the bulk material. The releasability ratings
at each site were compared to the airborne asbestos level at that site. The
results of the comparisons demonstrated a relationship between the bulk sample
releasability rating and the airborne asbestos levels; i.e., a high releas-
ability rating corresponded to elevated airborne asbestos levels.
Each of the 48 sites selected for the study was rated by a five-member
panel on the following assessment factors: material condition, accessibility,
whether it is part of an air moving system, how much of the material is ex-
posed, presence of water damage, activity level, and degree of friability.
The panel of raters first independently rated each site. Then, a consensus
xvi 11
-------
rating was determined for each factor. A comparison between the consensus
and the individual independent ratings found good agreement. The consensus
ratings were reported and used in the statistical analysis of the relationship
between factors and measured concentration levels of airborne asbestos.
The factors material condition, accessibility, activity, and friability
were not related to airborne asbestos levels for these data. The factor com-
binations available in the school district were limited. For example, almost
all activity areas were greater than 10% exposed; no sites were found where
the condition was rated "severely damaged"; and no sites were found where
accessibility was rated "not accessible."
The principal conclusions related to the first study objective are:
(1) airborne asbestos levels inside school buildings with asbestos are signif-
icantly higher than outdoor ambient levels due to the release of asbestos fi-
bers from asbestos-containing materials inside those buildings; and (2) within
a school building, asbestos fibers are transported from rooms having asbestos-
containing materials to rooms without these materials.
The principal conclusions related to the second study objective are:
(1) the existing algorithm is not a valid predictor of exposure to airborne
asbestos levels; (2) the amount of asbestos in the bulk material is not a
valid predictor of exposure to airborne asbestos levels; (3) the releasability
rating system developed in this study is related to levels of airborne asbes-
tos. (Additional studies are underway to attempt to validate this system.)
xix
-------
SECTION 1
INTRODUCTION
The widespread use of asbestos over the years has raised the issue of
increased cancer and chronic respiratory disease risk in various segments of
the population. Pulmonary cancer, mesothelioma, and fibrosis of the lung are
known to be associated with exposure to asbestos in certain work places, such
as where asbestos is mined and milled or where asbestos materials and products
are manufactured or used (NCI 1978, Peto et al., 1982, Zivy 1982). Currently
there is considerable concern that asbestos-containing materials that were
used extensively in schools from 1945 to 1973 for fire-retarding purposes and
acoustical insulation are deteriorating or being disturbed, with the release
of asbestos fibers into the air of the buildings. The resultant exposure of
the students, teachers, and school staff to the airborne asbestos may result
in asbestos-related diseases. A rule proposed by the U.S. Environmental
Protection Agency (EPA) requiring the identification of friable asbestos-
containing materials in schools and the notification of those exposed to the
materials was published in the Federal Register (45 FR 61966) in September
1980. The final rule was published May 27, 1982, in the Federal Register (47
FR 23360) and was effective June 28, 1982.
The Exposure Evaluation Division (EED) of EPA's Office of Toxic Substances
(OTS) has been providing a variety of technical support to the OTS asbestos-
in-schools program since its inception. This support includes the development
of sampling and analytical protocols; guidance for quality assurance programs
appropriate for implementation at state and local levels; a PLM analytical
proficiency program for laboratories analyzing bulk samples; international
collaborative work with Laboratoire d1Etude des Particules Inhalees, Paris,
on the collection and analysis of bulk samples and their relation to airborne
asbestos levels; preparation of an photomicrographic asbestos particle atlas;
preparation of a design study for a field survey to collect data on airborne
asbestos levels and the physical characteristics of the bulk materials; and
the investigations of exposure rating systems designed to identify buildings
that have significant airborne asbestos exposures.
Two procedures that have been proposed for characterizing airborne expo-
sures are the algorithm and the decision tree, both of which are discussed in
detail in Appendix A. The usefulness of both these procedures in assessing
exposure has been an ongoing concern to EPA. Over the past 2 years, there
have been a number of investigations directed at validating assessment tools
(USEPA 1979a, USEPA 198la). The experience gained has resulted in a field
study design that proposed air sampling to validate these procedures (Price
et al. 1980). A design report addressing the statistical aspects of the study
design including expectations for precision in the final results was prepared
as an aid in the planning process (USEPA 198lc).
1
-------
That design was the basis for the field study presented in this report.
The field study had two objectives:
1. To document exposures to airborne asbestos resulting from friable
asbestos-containing materials in schools.
2. To validate existing exposure assessment tools (including the
algorithm) and determine relationships between asbestos air levels
and individual or combined assessment factors.
These objectives were addressed by surveying the school buildings of a
large urban school district. The principal conclusions of the study are
given in Section 2. The overall quality assurance program that was used is
described in Section 3. A discussion of the sample design which was the basis
for the survey is given in Section 4. A discussion of the field survey, which
covers all aspects of the air-sampling and the bulk sampling activities, is
given in Section 5. Discussions of the transmission electron microscopy (TEM)
techniques and the polarized light microscopy (PLM) techniques that were used
to analyze the air and bulk samples respectively are given in Section 6. The
results from these analyses are summarized in Section 6 also. A discussion
of the statistical analyses that were used and their results are given in
Section 7- A list of pertinent references is given at the end of Section 7-
Appendixes A through G present the study background, the sampling protocol
for air sampling, the protocol for creating and maintaining chain of custody
of the samples, the protocol for the sampling and the analysis of insulation
material suspected of containing asbestos, the analytical protocol for air
samples, copies of the data sheets that were used in rating sites on each
algorithm factor, and additional statistical analyses.
-------
SECTION 2
CONCLUSIONS
The principal conclusions from this study are given below. The first
conclusion meets the first objective, which is to document potential exposures
to airborne asbestos from asbestos-containing material. The second conclusion
further supports this objective. The third, fourth, and fifth conclusions
meet the second objective, which is to validate existing exposure assessment
tools and determine relationships between asbestos air levels and individual
or combined exposure factors.
1. Airborne asbestos levels in buildings with asbestos-containing mate-
rials are greater than outdoor ambient levels due to the release of asbestos
fibers from those materials. On the average, the higher levels were found in
rooms with asbestos-containing materials, relatively lower levels were found
in rooms without asbestos-containing materials (indoor controls), and the
lowest levels were found in ambient sites on the schools' roofs. Tests of
significance found the average levels in the three types of sites to be sig-
nificantly different (.01 level).
2. Within a school building, asbestos fibers are transported from rooms
having asbestos-containing materials to rooms without these materials. There
is evidence of schoolwide elevated airborne asbestos levels in more than half
of the schools sampled. This was demonstrated by airborne asbestos levels in
indoor control rooms that were significantly higher than ambient levels.
3. The existing algorithm is not a valid predictor of exposure to air-
borne asbestos levels. When the algorithm scores for the 48 sites were com-
pared with the airborne asbestos level, the Pearson correlation coefficient
was found to be -0.17 (P-value = 0.25).
4. The amount of asbestos in the bulk material is not a valid predictor
of exposure to airborne asbestos levels. When the average percentages of the
asbestos in the bulk materials in the 48 sites were compared with airborne
asbestos levels, the Pearson correlation coefficient was found to be -0.06
(P-value = 0.71).
5. The releasability rating system that was developed in this work is
related to airborne asbestos levels in the school district examined. When
the releasability ratings of the 48 sites were compared with the airborne
asbestos levels, the Spearman correlation coefficient was found to be 0.44
(P-value = < 0.01). (Releasability is based on microscopic characterization
of the bulk material and is discussed on page 78.)
-------
SECTION 3
OVERALL QUALITY ASSURANCE PROGRAM
Quality assurance was an important consideration in the execution of
this study. It covered the organization of the team and the operation of all
aspects of the work. The major components of the quality assurance program
are summarized below.
I. ORGANIZATION
This was a collaborative study for EPA's Office of Toxic Substances (OTS)
that was performed under two prime contracts. Midwest Research Institute (MRI),
under EPA Contract No. 68-01-5915, was a prime contractor who worked for the
Field Studies Branch of the Exposure Evaluation Division of OTS. Research
Triangle Institute (RTI), under EPA Contract No. 68-01-5848, was a prime con-
tractor who worked for the Design and Development Branch of the same division.
There were eight major areas of activities: management, survey design,
statistical sampling, quality assurance, field survey, sample analysis, statis-
tical analysis, and final report preparation. The division of responsibility
was: MRI—management, quality assurance, field survey, sample analysis, and
final report preparation; RTI--survey design, statistical sampling, quality
assurance, statistical analysis, and report preparation. Battelle Columbus
Laboratories (BCL), under subcontract, had the responsibility for the analysis
of the air samples that were collected. BCL assisted MRI with the field sur-
vey and assisted RTI with survey design, statistical analysis, and data in-
terpretation. Other MRI subcontractors were the Illinois Institute of Tech-
nology Research Institute (IITRI), who performed quality assurance analysis
of air samples, and the Colorado School of Mines Research Institute, who per-
formed quality assurance analysis of bulk samples. MRI, RTI, and BCL collab-
orated in the preparation of this report.
II. OPERATION
The study was initiated with a planning meeting with EPA, MRI, RTI, and
BCL personnel. The mission of the study, which is given in Section 1, was
well understood by all parties. The plan of operations was established and
specific assignments and responsibilities were made. This four-party team
worked together closely throughout the study.
In the execution of the plan all field and laboratory compliance require-
ments and a thorough documentation system were followed. Good communications
via telephone conversations, meetings, written communiques, and reports were
maintained among all parties. Plans were modified as required for effective-
ness of operation.
-------
Protocols were prepared to ensure uniformity in sampling and chemical
analysis. Quality assurance data were collected in order to monitor the data
acquisition process. A chain of custody was established and quality assurance
audits were conducted to ensure that all procedures were followed. The qual-
ity assurance program is briefly summarized in the paragraphs that follow.
Evaluation of quality assurance data on sampling and chemical analysis are
found in the sections treating those topics, Sections 5 and 6, respectively.
Details of the protocols are found in the Appendices B through E.
A. Protocols
Existing protocols were adapted when possible. The EPA bulk sampling
and quality assurance protocol (USEPA 1980b) was followed and is summarized
in Appendix D. The protocol for analyzing bulk samples for asbestos content
specifies the use of PLM and closely follows the steps given in the Asbestos
Particle Atlas that was prepared by Walter C. McCrone (1980) and the EPA PLM
analytical protocol (USEPA 1982b, 1980 draft version). The protocol for TEM
analysis of filters (Appendix E) generally follows an EPA method (USEPA 1978).
The protocol specifies sample preparation, fiber counting, and units conver-
sion to nanograms per cubic meter.
A chain of custody (Appendix C) was developed both for the handling of
air samples taken on filters and for bulk samples. The chain of custody was
implemented to ensure that data being reported are in fact data from those
samples that were taken in the field. It also allowed the renumbering of
samples prior to chemical analysis to prevent analyst bias from affecting the
results.
A protocol for collecting air samples was also developed and is described
in Appendix B. It provides guidance for placing the sampler, specifies the
type of sampling equipment to be used, and specifies the sampling operating
characteristics including flow rate, sampling time, and other basic parameters.
The protocol for collecting bulk samples (Appendix D) specifies how to define
a sampling area and how to select appropriate locations for sampling.
B. Quality Assurance Data
For TEM analysis of filters, data were obtained on laboratory blanks and
field blanks. A portion of the filters were selected at random for duplicate
analysis, and other filters were divided for analysis at an independent lab-
oratory.
The detection limit for the TEM analysis is one fiber observed while
scanning grid openings. The resulting detectable quantity, based on the pro-
tocol and the volume of air sampled, is 0.002 ng/m3. The quantification limit
depends on the number of fibers observed during the TEM analysis, which was 1
to 941. Thus, the number of significant figures for the TEM results ranges
from 1 to 3.
For the PLM analysis of bulk samples, side-by-side samples were divided
into two groups at random. One group was used to assess intralaboratory
variation; the other group was split between the primary and a secondary lab-
oratory to investigate interlaboratory variation.
-------
Further information on TEM and PLM quality assurance analyses can be
found in Section 6.
C. Audits
Traceability audits were conducted (1) to determine and establish sample
and data traceability and (2) to determine if sampling and analysis protocols
were followed. The audit results indicated that all bulk samples were trace-
able, and sampling and analysis were performed according to designated proto-
cols. All air samples except one were traceable. This one air sample was
one of a group that was put on hold pending recycling of the sites for samples
of longer durations of sampling. The repeat sample was used; therefore, the
one not traceable was not needed. Inconsistencies in fiber counts were dis-
covered. Records were corrected, and air-level calculations were repeated to
eliminate the inconsistencies. This resulted in over 20 revisions in air
levels. Sampling and analysis were completed according to protocol.
An audit was also conducted on the data file that was prepared for sta-
tistical analysis. The objectives of the audit were to verify all data values
and to verify that the data were correctly identified with the school and the
sampling area within the school. The results indicated one minor error in
the averaging of the percent asbestos for bulk samples from a site. The error
was corrected, and it was determined that the change was of no material impor-
tance to the results of the statistical analysis.
D. Site-Specific Ratings
The sites where air sampling and bulk sampling took place were inspected
by a panel of five trained raters. Their primary purpose was to obtain con-
sensus ratings of the assessment factors (Appendix A). The consensus ratings
were obtained at each site by polling the five raters after all raters had re-
corded their independent assessment of the factors on a data sheet. A simple
majority rule was invoked when there was disagreement among the five indepen-
dent assessments. The results of the statistical analyses of this information
are given in Section 7.
-------
SECTION 4
SAMPLE DESIGN AND SELECTION
This section outlines the sample design and selection for this study of
airborne asbestos levels. A more detailed description of the design (USEPA
1981c) and its development is provided in Appendix A.
The study population, or population of inferential interest, was defined
in terms of all student activity areas in the study school district. The
study population consisted of student activity areas that had asbestos-
containing material. The sample design was developed to satisfy the informa-
tional requirements of the study for this population. The first objective of
the study was to document potential exposure to airborne asbestos in schools
in which asbestos-containing materials were present. This objective required
estimating airborne asbestos levels at these sites and comparing these esti-
mates with estimates for the following types of sites: (1) student activity
areas without asbestos-containing material, or control sites, and (2) outdoor
ambient sites. To ensure estimates of the desired precision, the sample de-
sign controlled the distribution of the sample with respect to these types of
sites.
The second objective of the study was to examine the relationship between
asbestos air levels and individual or combined assessment factors (including
the algorithm). This second objective required estimating and comparing air-
borne asbestos levels of the sites with different assessment factor level
combinations. The sample design controlled the distribution of the sample
with respect to certain assessment factor level combinations.
These two study objectives were mutually conflicting as to the prescribed
allocation of sites among the factor level combinations; that is, the alloca-
tion that was to give the most precise estimates for the first objective was
not the same as the allocation that was to give the most precise estimates
for the second objective. This was because the factor level combinations of
interest were not present in equal proportions in the study population. To
satisfy the study objectives, the sample design employed a compromise
allocation.
The sample design, outlined below, was a two-stage design with stratifi-
cation imposed on each stage. First-stage sampling units were schools having
asbestos-containing material in student activity areas. Second-stage sampling
units were student activity areas, or sites (e.g., classrooms, auditoriums,
and gymnasiums). A total of 48 asbestos-containing material sites were se-
lected in 25 sample schools. Additionally, one control site (a site without
asbestos-containing material) was selected from each of the 25 schools in the
first-stage sample.
-------
At each of the 48 asbestos-containing material sites, the following ac-
tivities were to be conducted: (1) air sampling during five consecutive
school days while students were in attendance, (2) bulk sampling of asbestos-
containing material, (3) scoring assessment factors, and (4) collecting other
relevant site information (e.g., cleaning procedures used and weather condi-
tions during air sampling). The field activities are described in detail in
Section 5 of this report. With the exception of bulk sampling, the above
activities were also performed at the 25 control sites. Additionally, air
sampling was conducted outside each of the 25 schools. At three sites ran-
domly selected from the 48 asbestos-containing material sites, air sampling
was to be conducted during three consecutive periods of five school days; con-
currently, air sampling was conducted at the control sites and at outdoor am-
bient sites associated with these three sites.
The sample design employed for this study was a statistically valid de-
sign that permitted estimation for the study population free from selection
bias. (Bias refers to the average difference between estimated values and
the true value of a given parameter for the population of interest.) Using a
probability sample design with demonstrable inferential ability for the study
area was one aspect of the overall effort to ensure data quality for this
study. A purposive selection of schools and sites within schools was not used
because it would not permit statistically valid inferences to be made for the
study population; the results would apply only to those sites where data were
collected. Also, there would be little assurance that the results were not
influenced by researcher preconceptions or objectives.
Because this study was of a geographically restricted area, as opposed
to a national study, the inferential ability of the information generated by
the study was concomitantly restricted. It was still thought important, how-
ever, that a probability sample design be employed to allow conclusions to be
drawn at the level of the study population. The study population was a real-
world collection of sites that might well be typical of many other settings.
An examination of the population frequencies of assessment factor combinations
suggested that this school district is similar in that regard to other school
systems.
I. STUDY POPULATION
The study population consisted of all the school district's student ac-
tivity areas, or eligible sites, in schools that had asbestos-containing mate-
rial in any student activity area. Eligible sites within a school included
classrooms, corridors, gymnasiums, locker rooms, cafeterias, kitchens, librar-
ies, and auditoriums. The term "classroom" also refers to special purpose
rooms such as music rooms and laboratories. The following school areas were
not considered to be eligible sites: administrative offices, teachers'
lounges, custodial rooms, storage rooms, mechanical rooms, and restrooms.
This definition of which sites are eligible for the study was based on an in-
terest in sites with higher activity levels and an interest in exposure of
students to airborne asbestos. Restrooms were not considered eligible sites
because they were not thought to be operationally feasible.
10
-------
A major reason for conducting the study in this particular school dis-
trict was that, through its own asbestos program, the district had collected
information that could be used to advantage in the selection of sites. School
personnel had recently inspected all the district's schools for asbestos-
containing material. In schools where material suspected of containing asbes-
tos was found, the total area was partitioned into sampling areas according
to the USEPA procedure (1980b, 1980c). Bulk samples were taken and data on
the assessment factors were collected.
The school personnel's inspection produced 134 eligible sampling areas.
At the time the sample was selected, laboratory analyses of bulk samples had
been reported for 92 of these sampling areas. Table 1 shows the distribution
of eligible sampling areas with respect to asbestos content, friability, con-
dition, exposure, and accessibility based on the ratings by school district
personnel. According to these ratings, the district's eligible sampling areas
exhibited only 10 of the 32 possible combinations of the factor levels shown
in Table 1. The fact that many of the factor level combinations appeared not
to be present in the study population limited the comparisons that could be
made to satisfy the second objective of examining the relationships of the
assessment factors to airborne asbestos levels. However, factor level combin-
ations that occurred very infrequently in real-world settings were not of
major interest to this study.
II. SAMPLE DESIGN
The sample design employed was a two-stage design with stratification
imposed on both stages. The first-stage frame consisted of all schools having
student areas in which material known to or suspected of containing asbestos
was present. The first-stage frame was stratified into nine classes formed
by combining three categories of asbestos content—low, high, and unknown--
with three categories of friability—low, moderate, and high. Schools were
classified according to the asbestos content and friability of the sampling
areas. The first-stage sample of 25 schools was allocated among the nine
strata approximately in proportion to size but with greater emphasis on known
asbestos content and high asbestos content. Size measures based on school
enrollment were used. The required number of schools was selected from each
stratum with probability proportional to size and without replacement.
The second-stage frame consisted of all eligible sites in the first-stage
sample of 25 schools. The second-stage frame was stratified according to
presence or absence of asbestos-containing material. Forty-eight asbestos-
containing material sites and 25 control sites (sites with no asbestos-
containing material) were selected. To the extent possible and given the
following restrictions, the second-stage sample of 48 asbestos-containing
material sites was allocated among schools and assessment factor categories
proportional to the number of sites: (1) at least one site must be selected
from each of the 25 schools in the first-stage sample, and (2) at least one
site must be selected from each nonempty, factor combination category. Eli-
gible asbestos-containing material sites in each school (or school and factor
combination category) were listed in order according to location. The re-
quired number of sites was then selected by random systematic sampling.
11
-------
pj
Table 1. Distribution of Eligible Sampling Areas in School District
with Respect to Asbestos Content, Friability, Condition,
Exposure, and Accessibility
Asbestos
c
content
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
High
High
High
High
High
High
High
High
High
High
High
High
High
High
High
High
Friability
Low
Low
Low
Low
Low
Low
Low
Low
High
High
High
High
High
High
High
High
Low
Low
Low
Low
Low
Low
Low
Low
High
High
High
High
High
High
High
High
g
Condition
Good
Good
Good
Good
Bad
Bad
Bad
Bad
Good
Good
Good
Good
Bad
Bad
Bad
Bad
Good
Good
Good
Good
Bad
Bad
Bad
Bad
Good
Good
Good
Good
Bad
Bad
Bad
Bad
No. of eligible'
Exposure Accessibility sampling areas
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Lo
High
Low
High
0
0
11
0
0
0
0
0
0
0
49
5
0
0
6
1
0
0
2
0
0
0
0
0
0
0
13
3
0
0
1
1
(continued)
12
-------
Table 1 (continued)
Asbestos
content Friability Condition Exposure Accessibility*
No. of eligible'
sampling areas
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Low
Low
Low
Low
Low
Low
Low
Low
High
High
High
High
High
High
High
High
Good
Good
Good
Good
Bad
Bad
Bad
Bad
Good
Good
Good
Good
Bad
Bad
Bad
Bad
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
0
0
1
0
0
0
0
0
0
0
34
2
0
0
4
1
Total
134
A sampling area was eligible for study if it contained at least one
eligible site with material suspected of containing asbestos. Eligible
sites within a school included classrooms, corridors, gymnasiums, locker
rooms, cafeterias, kitchens, libraries, and auditoriums. The following
were not considered eligible sites: administrative officers, teachers'
lounges, custodial rooms, storage rooms, mechanical rooms, and restrooms.
Based on sampling area ratings by school district personnel. Asbestos
content was taken from the school district's asbestos program laboratory
results, reported by June 26, 1981.
Asbestos content: Low = > 1% and < 20%, High = ^ 20%
c
d
e
f
8
Friability: Low = 1, High =2,3
Condition: Good =0,2, Bad = 5
Exposure: Low = 1, High = 4
Accessibility: Low =0,1, High = 4
13
-------
At each of the 25 schools in the first-stage sample, one control (indoor
ambient) site was also selected from the list of sites without asbestos-
containing material. All such sites within a sample school had equal probabil-
ity of being selected. Any student activity area without asbestos-containing
material was eligible to be a control site; there was no requirement that con-
trol sites be a certain distance from asbestos-containing material sites.
Additionally, the sample schools varied widely as to the proportion of the
school area having asbestos-containing material.
Alternate sites were named for each of the selected sites so that field
personnel could substitute an alternate site whenever a selected site was
found to be ineligible (contrary to floor plan information) or whenever there
was nonresponse at a selected site. Nonresponse could occur, for example, if
a teacher or school official would not permit air sampling at the selected
site or if air sampling was not possible at the site because of the lack of
an electrical outlet, repeated vandalism, etc. An alternate site could not
be substituted for a selected site simply for convenience. The procedures
for substitution were carefully followed by field personnel.
14
-------
SECTION 5
FIELD SURVEY
The survey included two types of sampling: air sampling, which was per-
formed by MRI and BCL, and bulk sampling, which was performed by MRI. The
air sampling activity started on May 4, 1981, and ended on June 2, 1981. The
bulk sampling started on May 28, 1981, and ended on June 4, 1981. The air
sampling period was fixed because the sampling had to be done while students
were in school. RTI selected the sites to be surveyed. The statistical basis
for the field survey plan was discussed in Section 4. The protocols that were
followed for the air sampling activity and bulk sampling can be found in Appen-
dices B and D. These protocols are adaptations of those used by Laboratoire
d1Etude des Particules Inhalees, Paris (USEPA 1980f).
I. AIR SAMPLING
Forty-eight indoor sites in which asbestos-containing material was pres-
ent were selected for air sampling in the 25 study schools. To investigate
the variability of airborne asbestos with time, three long-term sites were
randomly selected from the 48 sites.
The survey plan called for the collection at each school of one outdoor
ambient air sample,* one indoor control air sample from an area constructed
with materials not containing asbestos, and up to four samples from rooms that
contained asbestos. All samples at a school were to be collected simultane-
ously. Long-term air sampling at three schools consisted of repeating the
sampling procedure two additional times at the outdoor ambient site, the in-
door control site, and one indoor sample site.
The air sampling phase consisted of collecting a sample at each site for
five consecutive school days during school hours while students were present.
The sampling rate was to be approximately 5 liters/min, for a total volume of
air to be sampled of approximately 10 m3.
The term "sample" used in the discussion of the field survey and fiber
analysis activities is not to be confused with the term "sample" that
is used in the statistical sense. In the statistical discussions,
"sample" refers to the subset of units selected. In this section,
"sample" refers to the air or bulk material collected for chemical
analysis.
15
-------
A. Sampling System
The air sampling system used is shown in Figure 1. A programmable timer
was set to start the system at the beginning of the class day and to stop the
system at the end of the class day. The sampled day ran from 8:30 AM to 3:30
PM in elementary schools and junior high schools and from 8:00 AM to 3:00 PM
in senior high schools, for a total of 7 hr/day. At sites where the system
was started on days other than Monday, the timers were set to skip Saturday
and Sunday, resume on Monday, and stop after five consecutive school days had
been sampled.
Observation of the sampling systems during the first several days of sam-
pling revealed that in setting the on-off stops on the timer, diaphragm pumps
started and stopped from 15 min early to 15 min late. If the total volume
were calculated from the timer setting, the cumulative error could be as much
as 15%. It was decided to add running time meters to the sampling system to
eliminate this potential source of error. Electric clocks, which were readily
available, were used (running time meters could not be located). This permit-
ted determining the actual time sampled provided sampling was not interrupted.
B. Field Operations
An air sampling scheme staggered over time was used to survey the 25
schools in accordance with the sampling protocol presented in Appendix B.
During field operations some of the samples were lost and some were collected
for an inadequate or unknown length of time. These deficient samples resulted
from filters being vandalized, equipment theft, equipment failures, power in-
terruptions, and field crew errors.
In an effort to obtain satisfactory samples for as many sites as possible,
repeat sampling, or recycling of the deficient sample sites, was pursued.
Because of the limited time available before the end of the school year, it
was apparent that all deficient sample sites could not be recycled. Whenever
possible, recycling was done so that all sites at a school were sampled simul-
taneously. A list of sites that did not meet the program criteria was pre-
pared, and as equipment became available those sites showing the greatest de-
viations were recycled first. Priority was given to sites where samples had
not been obtained initially. Next priority was given to sites where sampling
had been less than the required 5 days, with sites ranked according to the
time sampled—shortest periods to longest periods. At those sites where a
5-day sample could not be collected before the end of the school year, two
samples (filters) were collected. This was accomplished by operating the sam-
pling system until the last day of school. The filter was then recovered and
a new filter was installed. The sampling system was then operated with the
second filter installed until five days of sampling were completed. A hold-
for-analysis label was placed on the recycled samples. At the completion of
field sampling, all samples were reviewed and those most nearly meeting the
program criteria were released for analysis. The remaining filters were put
on hold.
Each field team member was given a logbook for recording data. Most
of the type of data collected is given in the sampling protocol document
16
-------
Orifice
Detail
\
Brass Disk 0.209" Dia.
1/16" Thick Center
Drilled '68 & Soft
Soldered in Place (0_2090,.)
1/2" Deep
8 Turns of
1/4 "Copper
Tubing Wound
4" Diameter
3 Foot 1/4" x 3/16"
Rubber Vacuum Tubing
Swage lolc B-2-MHC-4T
Hose Connector to Male
Pipe 1/8" Male Pipe to
1/4" I.D. Tubing
Thomas Industries Inc.
Pump Model 107CA18
Gel man Filter Holders
Model 4202 47mm Open
Faced Magnetic
Clamp, Medium
Utility 3-Finger
Jaw Vinylized
36" Long Rod
Tube Fitting. Male Elbow 90°
1/8" Male Pipe Threaded to
1/4" Tube
Swage lolc B-200-2-J
Indicator
WW Grainger
6X136
s
7 Day Programmable
Timer
Grainger 2E214
Power Cord
Figure 1. Air sampling system.
17
-------
(Appendix B). Additional items recorded included the relative humidity and
temperature at each site at the beginning and end of the sampling period,
whether the air conditioner was running during the sampling period, whether a
window was open, whether there was a rug or other floor covering at the sample
site, and the frequency and method of cleaning the classroom and hallway where
the sampler was located. To document the sampling times, the electric clock
readings were recorded, and to identify the equipment used at a site, property
tag numbers were recorded. Black and white photographs were taken of sampling
sites. Climatological data for the area and period in which air sampling took
place were obtained from the National Oceanic and Atmospheric Administration.
Because of the number of problems that developed in keeping the sites
operational for the five-day period, a walk-through procedure was instituted
after the first week. This procedure consisted, as time was available, of
walking through each active school and observing each system. As problems
with a system developed, corrective action was taken, including re-plugging
in the power cords, resetting timers, exchanging or replacing missing or mal-
functioning equipment, and reconnecting the hoses to the sampling heads. If
filters were damaged early in the sampling period, new filters were installed
and sampling was continued. At one school where several sampling systems had
been tampered with, all sites at this school were recycled. During the walk-
through operation and near the middle of the sampling period (between two and
three days after sampling was started), and whenever possible, the temperature,
relative humidity, and airflow in the sampling system were taken at each site.
C. Sample Handling
The air samples were handled according to the protocol given in Appendix
C. To facilitate sample tracking, the filter holder labeling number was modi-
fied from that given in the protocol. Each modified number contained the
letter "B" (for Battelle Columbus Laboratories) or "M" (for Midwest Research
Institute) to designate the organization collecting the sample, followed by
the operator's initials and the sample number. The sample numbers were as-
signed sequentially by each operator to the filters he or she recovered. At
this time, the operator entered the sample number in the logbook for the col-
lection site and completed a chain-of-custody form.
II. BULK SAMPLING
Four bulk samples were collected from each of the 48 indoor asbestos-
containing sites. Samples were collected from three randomly selected points
at each site. A double sample taken side by side was collected at one of the
three points to provide a duplicate for quality assurance.
At the time the statistical selection of sampling points was made, all
room dimensions were not available to RTI. Some of the sampling points were
identified, therefore, as a fraction of the room length and width. The field
sampling team located a sampling point by measuring the room and converting
the fractional value to a unit of measure. If the sampling point could not be
reached because of its location (above a light fixture or other obstruction),
RTI was contacted for an alternate sampling point.
18
-------
A. Sample Collection
Bulk samples were collected by cutting away a section of the asbestos-
containing material. A section of material 1 to 8 cm3, depending on the con-
dition and thickness of the material, was collected. The collected samples
were placed directly into labeled, snap-covered plastic bottles for shipment
to MRI. At the same time, chain-of-custody forms were prepared and the sample
number and site description were entered in the logbook.
B. Sample Handling
The bulk samples were transported to MRI and released to a quality con-
trol representative, who logged them in and assigned them permanent numbers
on a random basis. The quality control representative then identified and
removed the duplicates and, from this set of duplicates, chose, on a random
basis, a number of them for external quality assurance analysis. The remain-
ing duplicates were put with the remaining samples, and all of these were
given to MRI analysts to be analyzed. Further details of the bulk sampling
procedure can be found in the sampling protocol presented in Appendix D.
III. CHAIN OF CUSTODY
The protocol used for creating and maintaining the chain of custody of
bulk and air samples is given in Appendix C. At the end of each week, MRI
gave its chain-of-custody forms and the filters it had collected to BCL field
personnel, who transported the air samples and the chain-of-custody forms to
Battelle Columbus Laboratories where the samples were analyzed. At the com-
pletion of the bulk sampling, all the bulk samples and the chain-of-custody
forms were transported by MRI field personnel to MRI. At MRI the QA samples
were removed and sent by air to the Colorado School of Mines for analysis.
The remaining samples were analyzed at MRI.
IV. SITE-SPECIFIC RATINGS
After the air sampling and the bulk sampling components of the data col-
lection effort had been completed, a third site-specific component of the data
collection effort was undertaken. Five raters walked through each room or
area where both air and bulk samples were collected and recorded their struc-
tured assessments of physical characteristics of the site on specially de-
signed data collection forms (Appendix F). In addition, one of the raters
interviewed available school personnel about typical cleaning practices rele-
vant to the sites in each school.
The purpose of these site inspections was to obtain consensus ratings of
the asbestos-containing material's general condition, the extent of water dam-
age, the quantity of material exposed, the degree of accessibility, the degree
of friability, the level of human activity present, and the presence of an
air plenum. The seven consensus ratings (Table A-l of Appendix A) at each
site were combined with information from the laboratory analysis of bulk sam-
ples of the asbestos-containing material to yield an algorithm score. Appen-
dix A contains a general description of the asbestos exposure assessment al-
gorithm used to calculate the score.
19
-------
Consensus ratings of the seven subjective algorithm factors were obtained
at each site by polling the five raters after they had recorded their indepen-
dent assessment of the factors on a data sheet. A simple majority rule was
invoked when there were disagreements among the five independent assessments.
Appendix F contains examples of the three forms used to structure the collec-
tion of the site-specific data. The cover sheet was used to record informa-
tion about the auxiliary variables, one sheet was used to record the algorithm
data, and the third sheet collected assessment factor information. Each of
the five raters independently completed both of the latter forms before a con-
sensus vote on the algorithm was taken at each site. The cover sheet for each
site was filled out by one of the EPA researchers; the other EPA researcher
recorded the consensus data.
Data concerning several auxiliary variables were collected during each
site inspection along with the data used to obtain the consensus ratings.
The auxiliary variable data included measurements of room length and width
and an estimate of room height; a notation of where the suspicious material
was located; a characterization of the room floor (carpeted, tile, wood); and
information about the frequency of sweeping, wet mopping, dry mopping, and
vacuuming. This auxiliary information was used in some of the statistical
analyses that are discussed in Section 7-
During the 3-hour training session for the five raters, assessment fac-
tors were defined, the algorithm description was presented in detail, and
slides of typical asbestos-containing materials were viewed. The discussion
of the algorithm was led by two individuals employed by the U.S. EPA who had
previous experience using the algorithm in several situations. The EPA per-
spective was supplemented by that of two individuals employed by the school
district who had previous experience inspecting school buildings for asbestos-
containing materials. Another perspective was supplied by a researcher under
contract to EPA who had previous experience analyzing algorithm data. Thus,
all of the raters used in the study were familiar with the asbestos-in-schools
problem, though they had had different levels of experience using the algo-
rithm in the field.
A caveat which is pertinent to the optimal classification analysis (see
Section 7) is that the air plenum data collected in this study should be
viewed as airstream data. No air plenums were present in any of the 48
asbestos-containing material sites. Rather, the asbestos-containing material
was located in an airstream from an air conditioning/heating unit at 23 sites.
The direction of the airstream was ascertained by noting whether or not the
ceiling was blackened in a characteristic pattern.
20
-------
SECTION 6
SAMPLE ANALYSIS
Two types of chemical analyses were performed: transmission electron
microscopy (TEM) for air samples and polarized light microscopy (PLM) for bulk
samples. A total of 125 filters were analyzed by Battelle Columbus Laborator-
ies (BCL) using TEM techniques. Of those, 116 were air samples collected at
the 25 schools: 54 were indoor samples (from classrooms or other student ac-
tivity areas); 31 were indoor control samples (from areas with presumably no
asbestos material); 31 were outdoor ambient samples; and 9 were field blanks
(filters). Quality assurance analysis of air samples was performed by the
IIT Research Institute (IITRI). Midwest Research Institute (MRI) analyzed
192 bulk samples using PLM techniques. The quality assurance analysis of bulk
samples was done by the Colorado School of Mines Research Institute (CSMRI).
I. AIR SAMPLES
Sample preparation and microscopic examination of the air samples were
carried out according to the Analytical Protocol for Air Samples (Appendix E)
established at the beginning of the program. This protocol is based on the
U.S. EPA Provisional Methodology Manual (USEPA 1978). Tables 2 and 3 summar-
ize the analytical results of chrysotile and amphibole, respectively, that
were obtained from the 126 samples* collected by the field sampling team. In
approximately one-half the number of schools, the air level of asbestos fibers
at the control site was greater than at least one of the air levels at the
asbestos-containing material sites. This phenomenon is discussed in Section
7.II.B. Fibers counted, fibers on filter, mass on filter, and volume of air
sampled for the individual samples are given in Table 4.
A. Method of Analysis
The samples were logged in and randomly numbered so that the analyst
would not know sample sites or relative locations of the samples. Four ana-
lysts performed the analyses on two transmission electron microscopes. A
senior analyst was always available for consultation in the event of a ques-
tion about the identification of a fiber or particle. The microscopic exam-
ination of the prepared grids was carried out at a magnification of 20,OOOX.
Each grid opening to be counted was selected randomly and then systematically
scanned to cover the full opening. The fibers observed were identified as
chrysotile, amphibole, or other.
One field blank was lost; therefore, only 125 samples are given in the table
21
-------
Table 2. Results of Analysis of Airborne Asbestos Fibers Collected by Filtration and Analyzed
by Transmission Electron Microscopy: Chrysotile Concentration in ng/m3
School
no.
1
2
3
4
5
6
7
8
9
10
11
12a
13a
Ua
15
16
17
Site
Week
1
1
2
3
1
2
3
1
1
1
1
1
1
1
2
3
1
1
2
1
2
3
1
2
3
1
1
1
Outdoor ambient
0.4
4.10
1.30
8.88
2.7
17.6
< 0.002
0.02
1.44
0.36
0.1
0.3
0.50
0.55
10.8
1.1
0.9
0.6
NA
0.83
NA
NA
0.77
NA
NA
3.5
2.5
40.6
Indoor control
69.3
3
2.6
< 0.002
166
14.6
45.5
1.4
16.9
43.9
5.01
362
0.1
72.9
50.3
73.4
51.0
NA
0.50
NA
NA
< 0.002
49.8
NA
NA
21.8
69.3
4.4
1
28.4
11.6
NA
NA
NA
NA
11.0
9.66
23.6
37
68.0
0.1
1
10
0.76
1.5
11.2
0.04
NA
18.7
NA
NA
NA
NA
77.6
69.3
425
149
2
8.95
73
NA
NA
38.2
NA
NA
NA
NA
82.8
92.7
NA
NA
NA
NA
NA
NA
< 0.002
NA
55.8
NA
NA
150
NA
NA
NA
NA
NA
3
108
422
NA
NA
105
27.5
6.17
NA
NA
43.0
93.3
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
89.0
89
NA
NA
NA
NA
NA
4
NA*
32.3
61.2
6.8
332
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
(continued)
-------
Table 2 (continued)
School
no.
18
19
20
21
22
23
24
25a
Site
Week
1
1
1
1
1
1
1
1
2
Outdoor ambient
0.57
0.90
0.35
0.63
1.1
2.84
0.38
1.3
NA
Indoor control
46.8
0.5
288
19
0.2
2.2
25.2
116
NA
1
7.93
644
420
153
312
196
77.8
8.58
NA
2
NA
NA
566
484
69.6
NA
11
NA
4.3
3
NA
NA
NA
NA
NA
NA
NA
245
NA
4
NA
NA
NA
NA
NA
NA
NA
NA
NA
a Some samples were collected for too short a period or an unknown length of time due to filters being
vandalized, equipment stolen, equipment failures, power interruptions, and field crew error. Repeat
sampling or recycling of questionable sites was done on a priority basis. In some cases, recycling
had to be done during another week. Therefore, the control and site samples were not taken during
the same week.
Note: The concentrations of nine field blanks were 0.30, 1.00, 0.00, 1.00, 0.60, 0.00, 0.00, 0.60,
and 0.00 ng/m3.
* NA = not applicable (no sample taken).
-------
Table 3. Results of Analysis of Airborne Asbestos Fibers Collected by Filtration and Analyzed
by Transmission Electron Microscopy: Amphibole Concentration in ng/m3
KJ
-P-
School
no.
1
2
3
4
5
6
7
8
9
10
11
12a
13a
I4a
15
16
17
Site
Week
1
1
2
3
1
2
3
1
1
1
1
1
1
1
2
3
1
I
2
1
2
3
1
2
3
1
1
1
Outdoor ambient
< 0.002
< 0.002
< 0.002
< 0.002
3
2
< 0.002
< 0.002
2
0.5
< 0.002
< 0.002
< 0.002
< 0.002
< 0.002
< 0.002
< 0.002
0.6
NA
< 0.002
NA
NA
0.11
NA
NA
0.6
< 0.002
< 0.002
Indoor control
< 0.002
2
0.1
< 0.002
24
4
< 0.002
< 0.002
3
< 0.002
< 0.002
79
< 0.002
42
5
< 0.002
< 0.002
NA
< 0.002
NA
NA
< 0.002
< 0.002
NA
NA
3
0.002
0.2
1
< 0.002
< 0.002
NA
NA
NA
NA
< 0.002
< 0.002
< 0.002
< 0.002
< 0.002
< 0.002
< 0.002
1
< 0.002
< 0.002
< 0.002
< 0.002
NA
< 0.002
NA
NA
NA
NA
< 0.002
< 0.002
52
3
2
< 0.002
< 0.002
NA
NA
< 0.002
NA
NA
NA
NA
< 0.002
< 0.002
NA
NA
NA
NA
NA
NA
< 0.002
NA
< 0.002
NA
NA
< 0.002
NA
NA
NA
NA
NA
3
< 0.002
93
NA
NA
4
< 0.002
< 0.002
NA
NA
< 0.002
< 0.002
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
< 0.002
< 0.002
NA
NA
.NA
NA
NA
4
NA*
< 0.002
13
< 0.002
6
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
(continued)
-------
Table 3 (continued)
NJ
t_n
School
no.
18
19
20
21
22
23
24
25a
Site
Week
1
1
1
1
1
1
1
1
2
Outdoor ambient
< 0.002
2
< 0.002
< 0.002
< 0.002
< 0.002
4
< 0.002
NA
Indoor control
9
0.5
< 0.002
83
< 0.002
< 0.002
1
< 0.002
NA
1
5
< 0.002
< 0.002
< 0.002
< 0.002
< 0.002
13
< 0.002
NA
2
NA
NA
< 0.002
5
< 0.002
NA
2
NA
< 0.002
3
NA
NA
NA
NA
NA
NA
NA
< 0.002
NA
4
NA
NA
NA
NA
NA
NA
NA
NA
NA
a Some samples were collected for too short a period or an unknown length of time due to filters being
vandalized, equipment stolen, equipment failures, power interruptions, and field crew error. Repeat
sampling or recycling of questionable sites was done on a priority basis. In some cases, recycling
had to be done during another week. Therefore, the control and site samples were not taken during
the same week.
Note: The concentration of each of the nine field blanks was 0.00.
* NA = not applicable (no sample taken).
-------
Table 4. Fundamental Data Obtained from the Analysis of the Airborne Asbestos Fiber Collected by
Filtration and Analyzed by Transmission Electron Microscopy
Chrysotile
Sample
no.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
School3 Siteb
20
20
20
20
2
2
2
2
2
2
11
11
11
c
18
18
18
c
14
7
7
2
2
2
8
8
8
1
1C
OA
2
4(1)
1(1)
IC(1)
2(1)
3(1)
OA(1)
1
OA
1C
c
OA
1C
1
c
1
3
OA
IC(2)
4(2)
OA(2)
OA
1
1C
Estimate
No. of of number
fibers of fibers
counted on filter
941
405
16
724
122
117
71
164
419
72
125
9
107
1
24
158
111
3
120
238
1
81
201
35
4
3
632
1.18 x
5.08 x
1.00 x
9.09 x
3.83 x
2.46 x
4.50 x
1.03 x
2.63 x
4.50 x
1.96 x
1.00 x
6.72 x
6.00 x
7.50 x
4.96 x
1.16 x
2.00 x
7.54 x
2.99 x
6.00 x
5.10 x
1.26 x
2.20 x
2.00 x
2.00 x
7.94 x
9
10*
6
10*
107
10'
10«
108
106
107
10
10^
107
10
10^
107
IO7
10?
7
10*
10
io6
106
105
wi
5
10*
Mass on
filter (g)
4.02 x
3.03 x
3.50 x
5.31 x
3.32 x
1.20 x
2.90 x
7.51 x
4.04 x
3.80 x
1.17 x
9.00 x
5.11 x
6.00 x
5.40 x
4.58 x
7.73 x
1.00 x
2.69 x
1.01 x
1.00 x
2.80 x
6.38 x
1.20 x
3.00 x
1.00 x
3.76 x
1Q-6
10-9
10~6
10~7
10~R
10~-7
10"-6
10 I
10~7
10"n
10"?
ID' 7
10
10
io"7
10-9
10~7
10
10
10~?
10-8
10_q
10
10"^
io'6
Araphibole
Estimate
No. of of number
fibers of fibers
counted on filter
0
0
0
0
0
0
3
0
3
0
0
0
0
0
0
6
2
0
0
0
0
1
8
0
0
0
26
0
0
0
0
0
0
2.00 x 10^
° 6
2.00 x 10
0
0
0
0
0
0 ,
2.00 x 10|?
2.00 x W
0
0
0
0 ,
6.00 x IO;
5.00 x 10
0
0
0
3.30 x 10
Air volume
Mass on sampled
filter (g) (m3)
0
0
0
0
0
0 -8
2.00 x 10
° -7
9.00 x 10
0
0
0
0
0
0 _
8.00 x 10"J
5.00 x 10
0
0
0
0
i.oo x 10 ::
1.00 x 10
0
0
0
7.80 x 10
9.58
10.55
9.98
9.38
10.27
10.35
9.75
10.29
9.56
9.29
10.40
10.71
10.01
c
9.46
9.77
9.75
c
3.47
10.14
10.45
10.75
10.44
9.66
9.83
10.77
10.37
(continued)
-------
Table 4 (continued)
N)
Chrysotile
Sample
no.
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
School3 Siteb
c
14
14
14
14
25
25
25
25
2
2
2
4
4
4
19
19
19
10
24
15
15
15
12
12
12
3
3
c
1C
2
3
OA
1
1C
3
OA
K3)
4(3)
OA(3)
1
1C
OA
OA
1C
1
IC(1)
OA
OA
1
1C
OA
1
2
1(3)
OA(3)
Estimate
No. of of number
fibers of fibers
counted on filter
0
114
99
184
76
114
331
343
13
0
99
114
113
64
1
22
8
789
204
16
59
144
123
8
1
0
122
0
0
7.16 x
1.20 x
2.31 x
1.40 x
1.43 x
2.08 x
4.31 x
8.20 x
0
7.80 x
1.79 x
1.42 x
4.00 x
6.00 x
1.40 x
1.00 x
9.91 x
1.28 x
5.00 x
3.70 x
9.04 x
2.58 x
5.00 x
6.00 x
0
2.56 x
0
7
10
10*
6
l°c
10
10*
105
f.
107
107
io{
1°A
10*
10
108
10ft
X
10c
10
107
10
10'
S
104
10
10
Mass on
filter (g)
0
4.52 x
7.80 x
1.88 x
6.40 x
8.30 x
1.14 x
2.32 x
5.20 x
-
6.90 x
9.03 x
9.14 x
1.50 x
2.00 x
9.10 x
5.00 x
6.32 x
8.30 x
4.10 x
3.30 x
6.47 x
2.04 x
6.00 x
3.00 x
-
9.80 x
-
io"7
10
10
10~R
10 ~ ,
— Q
~ Q
io-9
-R
10-8
10-8
10 °
10
10
l°"o
">:2
10 ,
— /
1° n
»>:
JO-7
-7
10 '
— u
10 *
10
-ft
10 8
Amphibole
Estimate
No. of of number
fibers of fibers
counted on filter
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
7
1
0
7
1
2
0
2
1
0
0
0
0
0
0
0
0 ,
2.00 x 10
0
0
0
0
0
0
0
0
0
0
4.00 x 10^
1.00 x 10
0 c.
f\
4.00 x 10°
3.00 x 10
1.00 x 10
° 5
4.00 x 10,
Li
6.00 x 10
0
0
0
0
Air volume
Mass on sampled
filter (g) (m3)
0
0
0
0
9.00 x 10
0
0
0
0
0
0
0
0
0
0 R
2.00 x 10 "J
5.00 x 10
0 7
— 7
5.20 x 10 '
5.00 x 10~
5.00 x 10
0
3.00 x 10 „
— u
5.00 x 10
0
0
0
0
c
9.08
9.77
9.18
8.36
9.67
9.88
9.48
4.11
9.52
10.23
10.18
9.46
10.95
9.23
10.04
10.61
9.82
11.39
10.78
9.42
9.33
9.35
8.84
8.25
8.80
8.94
9.80
(continued)
-------
Table 4 (continued)
fO
oo
Chrysotile
Sample
no.
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
a b
School Site
3
3
1
1
1
12
6
6
6
6
6
10
3
3
3
24
21
21
10
10
10
17
17
17
c
5
5
c
3(3)
IC(3)
1C
2
OA
1C
1C
1
2
3
OA
OA(1)
2(1)
IC(1)
3(1)
(2)
OA
1
IC(2)
1(2)
OA(2)
1C
1
OA
c
1
1C
c
Estimate
No. of of number
fibers of fibers
counted on filter
102
103
106
117
3
13
128
95
132
121
11
17
163
575
472
85
10
148
180
26
101
54
651
102
3
125
149
1
1.28 x
6.47 x
4.45 x
1.84 x
2.00 x
8.20 x
8.04 x
6.00 x
1.66 x
7.60 x
6.90 x
1.10 x
5.12 x
3.61 x
2.96 x
1.10 x
6.30 x
1.86 x
1.13 x
1.60 x
1.59 x
3.40 x
4.09 x
6.41 x
2.00 x
3.93 x
4.68 x
1.00 x
10
10_7
107
wl
10*
IOT"
107
10
io8
10s
10?
107
10g
10_
10,
7
105
JOg
106
107
10^
108
!
-------
Table 4 (continued)
N>
VO
Chrysotile
Sample
no .
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
School3 Siteb
23
23
23
c
16
16
16
c
9
9
9
22
22
22
22
c
7
7
7
13
10
3
3
24
24
21
21
d
3
OA
1
1C
c
OA
1
1C
c
1C
1
OA
1
2
OA
1C
c
1C
1
2
3
KD
4(1)
OA(i)
1C
1
1C
2
d
IC(2)
Estimate
No. of of number
fibers of fibers
counted on filter
181
805
78
0
22
659
246
0
4
18
15
338
270
22
3
2
101
210
274
352
99
509
16
104
113
119
398
d
137
1.89 x 10g
5.06 x 10,
4.90 x 10
0 ,
2.80 x 10g
8.30 x 10*
1.54 x 10
0
3.00 x 10^
1.10 x 10*
9.40 x 10o
4 25 x 10
~ • &~-J A J-V/rt
1.70 x lOg
1.40 x 10
2.00 x 10^
K
1.00 x 10^
6.34 x 100
i oo in"
1.32 X 10,.
1.72 x 10fi
2.21 x 10*
1.20 x 10 '
6.39 x 10*
2.00 x 10?,
~ n-> _. i A '
6 . t.1 X 1U_
7.10 x 10'
1.80 x 10 '
5.00 x 10
d _
7
4.30 x 10
Mass on
filter (g)
2.31 x
1.83 x
1.70 x
.
2.30 x
3.56 x
5.65 x
_
1.00 x
9.80 x
4.70 x
3.10 x
6.46 x
1.10 x
2.00 x
1.00 x
4.61 x
6.55 x
6.72 x
5.38 x
1.20 x
3.34 x
2.60 x
2.81 x
8.22 x
1.75 x
4.43 x
d
1.45 x
10-8
10
10
o
io'8
— r\
10 °
lO'7
10-9
10
10~7
10~«
10-9
10-8
10 °
- /
-7
<7
10-6
10-8
10-7
10-7
10 -,
io"'
io"6
— 7
10 '
Amphibole
Estimate
No. of of number
fibers of fibers
counted on filter
0
0
0
0
0
9
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
2
4
1
1
o
1
d
1
0
0
0
0
0
1.00 x
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1.00 x
3.00 x
5.00 x
3.00 x
6,00 x
5.00 x
1.00 x
d
3.00 x
7
w'
106
10*
105
105
10.
106
10°
^
io5
Air volume
Mass on sampled
filter (g) (m3)
0
0
0
0
0 7
— /
4.00 x 10
0
0
0
0
0
0
0
0
0
0
0
0
0
° -8
1.00 x 10 °
6.00 x io"r
3.00 x 10~*
1.00 x 10~^
1,00 x 10"'
8.00 x 10"'
4.00 x 10
d R
— o
4.00 x 1C
8.15
9.34
7.69
c
8.92
8.37
8.16
c
10.54
9.69
9.36
9.92
9.28
9.61
10.47
c
9.21
9.63
7.25
6.05
11.27
10.06
9.73
11.14
10.57
9.24
9.14
d
9.96
(continued)
-------
Table 4 (continued)
U)
o
Chrysotile
Estimate
Sample
no.
113
114
115
116
117
118
119
120
121
122
123
124
125
126
a The
b The
in
School
3
3
13
13
13
5
10
10
10
1
1
25
c
13
school
site at
a h
3 Siteb
3(2)
OA(2)
OA
2
1
OA
IC(3)
1(3)
OA(3)
3
1
2
c
1C
where the
No. of
fibers
counted
112
105
35
272
100
14
165
53
16
374
133
62
0
0
of number
of fibers
on filter
7.
1.
2.
1.
6.
1.
1.
3.
1.
2.
2.
3.
sample was
the school where the
which the sample
was taken.
03 x 10
65 x 10
20 x 10
71 x 10
28 x 10
80 x 10
04 x 10
33 x 10
00 x 10
35 x 10
09 x 10
90 x 10
0
0
taken.
sample
7
8
7
6
8
q
j
e
O
7
f.
\J
No. of
Mass on fibers
filter
2.98 x
1.92 x
8.50 x
5.35 x
1.97 x
1.60 x
7.25 x
1.40 x
8.90 x
1.03 x
1.36 x
1.70 x
-
0
See Tables
was
See Tables
taken.
2 and 3
(g) Counted
10^ 0
10" 3
10_7 °
10 0
10 0
10"; 3
10:
lO.o 0
10 , 0
io"k o
10 ft 0
10 0
0
0
2 and 3.
The number in
1C = indoor
Amphibole
Estimate
of number
of fibers
on filter
0
5.00 x 10
0
0
0
4.00 x 10
0
0
0
0
0
0
0
0
parentheses
control and
Air volume
Mass on
filter (g)
0 R
2.00 x 10
0
0
0
2.00 x 10
0
0
0
0
0
0
0
0
identifies the
sampled
Cm3)
10.80
10.87
10.14
9.59
10.55
11.02
9.86
8.84
8.16
9.49
4.79
3.83
c
3.12
week
OA = outdoor ambient.
c Filter blanks. No air sampled.
d Sample No.
111 was a
blank filter that was
lost.
Therefore, there were no data.
-------
The length and width of the chrysotile and amphibole fibers were re-
corded. The fiber length was measured using the number of concentric circles
on the viewing screen that the fiber crossed (each circle segment was 0.25 pm
at 20,OOOX). The width was measured in millimeters on the viewing screen.
The fiber was aligned with the millimeter scale on the side of the viewing
screen (1 mm = 0.05 pm at 20,OOOX). The volume of the fiber was then computed
assuming the fiber to be a right circular cylinder. The mass of the fiber
was calculated using a density of 2.6 g/cm3 for the chrysotile and 3.0 g/cm3
for the amphibole. Appropriate filter area factors and dilution factors were
used to extrapolate from the fibers actually counted and measured to the total
number of fibers per filter and total nanograms of asbestos per filter.
The detection limit for this type analysis is one fiber observed while
10 grid openings are scanned. The protocol calls for the counting of 100 fi-
bers or 10 grid openings but never any partial grid openings. One fiber ob-
served in 10 grid openings would correspond to 4 x 103 fibers per filter when
the extrapolation is made to total filter area. If the one fiber were of av-
erage dimensions (1 pm long x 0.05 |Jm in diameter), the mass would be 2 x
10 li g per filter. Since most of the air volumes per sample were approxi-
mately 10 m3, the minimum detectable quantities would be 2 x 10 12 g/m3 or
0.002 ng/m3.
The quantification limit for these analyses depends upon the number of
fibers observed during the TEM analysis. The number of fibers counted during
TEM analysis ranges from a minimum of 1 to a maximum of 941. Thus, the number
of significant figures for the results will range from 1 to 3.
The method of selecting the filter fraction to be ashed was generally
the same for each sample. The large amount of debris collected on many of
the filters made the low temperature ashing procedure a necessity. After ash-
ing, the residue containing the asbestos fibers was resuspended in 100 ml of
water using the ultrasonic bath to ensure that the fibers were removed from
the ashing tube walls. The resuspended sample was then divided into 10-ml,
20-ml, and 70-ml aliquots, and each aliquot was filtered onto a Nuclepore fil-
ter. The three aliquots gave the analyst some flexibility in finding a suit-
able fiber loading for TEM examination.
B. Discussion
Initial examination of some of the filters showed debris on the prepared
filter grid. The debris often necessitated using the 20-ml aliquot rather
than the 70-ml aliquot. The debris apparently was composed of dust and paint
chips collected at the sampling site on the sample filter. Any organic fibers
or other organic debris was removed during the low temperature ashing proce-
dure.
Fiber bundles and fiber clusters in the samples that contained large
quantities of asbestos required special attention. Bundle is defined as a
group of fibers bound together that makes the determination of its constitu-
ents difficult. Often it was possible to identify one end of a fiber, but it
was not always possible to identify positively the constituents of the bundle.
Cluster is defined as several overlapping and cross-linked individual fibers.
31
-------
Fibers in a cluster that could be seen as individual fibers were counted as
individual fibers, but when the individual fibers could not be distinguished,
they were considered a cluster and recorded as such, but not counted. Many
times the cluster formed around a paint chip. Another problem was in deter-
mining whether these bundles and clusters were evenly distributed throughout
the filter or were isolated events. (The chrysotile fibers were observed to
form bundles and clusters, but the amphibole fibers did not.)
The way in which bundles and clusters are handled can greatly affect the
quantity of asbestos calculated for each filter. The analysis does not in-
clude the bundles and clusters in the calculation primarily because the ana-
lyst could not be sure of uniform distribution or rely on the volume calcula-
tions associated with the bundles and clusters. Table 5 lists the number of
bundles and clusters for each sample. There were 82 samples that had some
bundles or clusters. Fifteen of these were outdoor ambient samples and 23
were indoor control samples. The remaining 47 were asbestos-containing mate-
rial site samples. The 15 outdoor ambient samples averaged 2.7 bundles/
clusters per sample. If the total 31 outdoor ambient samples (16 had no
bundles/clusters) taken are considered, the average is 1.3. The 23 indoor
control samples averaged 7.8 bundles/clusters per sample. If the total 31
indoor control samples (8 had no bundles/clusters) taken are considered, the
average is 5.7. The 47 site samples averaged 8.3 bundles/clusters per sample.
If the total 54 site samples (7 had no bundles/clusters) taken are considered,
the average is 7.2
Bundles and clusters have some mass; therefore, the calculations repre-
sent a minimum value for the quantity of asbestos for these samples. The sam-
ples with higher asbestos concentrations tended to have more bundles and clus-
ters. The bundles and clusters were observed on the TEM-prepared filter and
must have been deposited as such on the filter during air sampling. The ul-
trasonification procedure that followed the low temperature ashing tended to
break up the fiber bundles and clusters. The primary purpose of Bonification
was to ensure the removal of fibers from the glass test tube in which the ash-
ing took place. All samples were subjected to the same low temperature ashing
and Bonification procedure, done according to the protocol; therefore, the
effect is assumed to be the same for each sample.
C. Quality Assurance
The transmission electron microscopy analysis was carried out by four
persons trained in microscopy. A senior analyst was always available for con-
sultation if there was a question about identification. The sample prepara-
tion was also carried out by persons who had previous experience in preparing
samples for microscopic examination. Both the preparation and analysis were
carried out according to the protocol contained in Appendix E.
Although this protocol is accepted and used by expert microscopists,
there are factors that contribute to the possibility of having relatively
large variabilities in results. These factors include agglomeration (bundles
and clusters) during the ashing operation that is not included in the fiber
count, since the number of fibers cannot be ascertained; the possible transfer
(loss or gain) of fibers when a number of filters are processed at the same
32
-------
Table 5. The Number of Chrysotile Bundles and Clusters Observed on the Filter
but Not Used in the Mass Calculations
Sample
no.
1
2a
4
5
6
8
9.
10b
11
13b
is"
16a
17
19
20b
21b
22a
23
27a
29a
30
31.
32b
33
34a
35
38
39b
40
4la
43b
45
46a
47b
49
50a
54
56
57a
58a
59
62a
Chrysotile
bundles/clusters
54
40
40
14
3
3
6
1
2
12
2
2
3
4
3
3
2
5
22
5
1
12
4
1
15
8
4
4
4
3
1
8
8
2
1
3
8
1
1
4
1
3
Sample
no.
63
68
71,
72b
73
74a
76b
77a
78
79b
81
82a
84b
85
86a
88b
89
90a
93
95
96
97b
100a
101
102
103
104
105
106b
107a
108
109a
110
112a
113.
114b
116
117
119a
120
122
123
Chrysotile
bundles /clusters
6
9
7
1
1
10
2
1
23
1
2
2
6
2
6
5
35
4
2
5
7
4
5
14
22
14
4
7
3
9
5
7
5
2
4
1
13
6
13
1
6
3
Sample from an indoor control site. There were eight other indoor control
samples (Nos. 7, 37, 44, 61, 69, 92, 98 and 126), but there were no bundles
or clusters on these.
Sample from an outdoor control site. There were 16 other outdoor ambient
samples (Nos. 3, 12, 24, 25, 36, 42, 48, 51, 55, 60, 66, 67, 94, 115, 118
and 121), but there were no bundles or clusters on these.
33
-------
time, or the loss when a single filter is processed; the effectiveness of the
dispersion of fibers during the sonication process; the production of a non-
uniform deposit of the fibers during the filtration operation; and the rela-
tive error in calculating the mass concentration of fibers. Further improve-
ment of the protocol is recognized as being needed but was beyond the scope
of this task.
The quality assurance aspect of the analytical part of this program is
summarized in the following paragraphs. As a quality assurance measure, sam-
ples were randomly selected for analysis by IITRI. Table 6 lists the samples
selected and the results obtained at IITRI and BCL.
The filters were divided at BCL, and one-half of each filter was hand-
carried to IITRI; the deposited side was kept up at all times. Results for
some of the samples showed large differences even though the same analytical
protocol was used by both laboratories. Since it did not seem likely that
there could be enough inhomogeneity of the filter deposit to produce these
large differences, each laboratory reexamined the prepared grids where differ-
ences were observed. They concluded that the microscopic examinations had
been correct and that the variation must have been due either to inhomogeneity
in the filter deposit or to laboratory contamination during preparation of the
filters for microscopic examination. BCL repeated the analysis of samples 95
and 110 starting with a new section of the original air sample filter. The
results of this second analysis, shown in Table 6, indicated that there was a
high concentration of chrysotile in these samples. During IITRI's reexamina-
tion of sample 98 by a qualitative comparison of a second dilution, no amphi-
bole fibers were detected. IITRI believed that the first dilution examined
was contaminated; therefore, the results of this sample should be discarded,
or used with great caution.
Table 7 shows the results of duplicate and replicate analyses for chryso-
tile conducted at BCL. The duplicate analyses were conducted by a second ana-
lyst using the same grid preparation as was used in the original analyses. In
a few cases, different aliquot preparations were analyzed. The replicate
analyses were performed using two independent preparations. Figure 2 shows
the relative variation* of the analysis pairs plotted against the average
value of the pairs. The symbol "Z" denotes the duplicates and "A" denotes
the replicates. Except for very low concentrations, the percentage relative
variation does not exceed 40% and is independent of the measured asbestos con-
centration level. The levels of variation experienced in this study are con-
sistent with other reports on the use of TEM for asbestos analysis (USEPA
1980f). The variation in the airborne asbestos concentration levels across
sites sampled in this study far exceeds 40%. The indication is that the ana-
lytical variability is within an acceptable range and the data are appropriate
for their intended use.
Relative variation is also known as the coefficient of variation computed
as standard deviation divided by mean or s/x.
34
-------
Table 6. Quality Assurance Results of Transmission Electron Microscopy Analysis
Asbestos concentration (ng/m3)
Sample no.
5
10
22
25
50
88
90
92
95
98
110
IITRI
Chrysotile
6.50
1.40
0.250
0.770
22.6
0.230
45.7
0.090
15.5
1.30
4.40
Amphibole
0.030
1.60
0
0
113
32.0
0
0
0
l,740a
0
BCL
Chrysotile
32.3
4.10
2.64
0.306
21.8
2.54
69.3
0.103
312
0.214
484
Amphibole
0
0
0.103
0
2.68
0
0
0
0
0
4.86
BCL-repeat
Chrysotile Amphibole
514 0
275 0
In a qualitative examination of a second dilution, no amphibole fibers were observed. IITRI believes
the first dilution was contaminated; therefore, this result should be discarded or be used with great
caution.
-------
Table 7. Quality Assurance Results of Duplicate Sample
Analysis for Chrysotile
Reported Duplicate
Sample no. ng/m3 ng/m3
Duplicates
17 7.93 13.5
20 99.3 70.2
32 0.770 1.50
39 8.88 4.66
47 0.379 0.200
67 0.554 0.720
72 0.634 1.20
76 10.8 6.28
79 40.6 56.6
94 0.504 1.50
96 69.6 80.1
97 1.14 0.710
104 10.4 16.0
Replicates
95 312 514
110 484 275
Samples analyzed by two different TEM operators (same
grid preparation but different areas).
36
-------
I.UU
0.80
c
o
o
•g °-60
>
lt_
0
c
.2
| 0.40
M-
8
u
0.20
0
z
-
Z - Duplicate
A - Replicate
e\
2 - Two of the duplicates
(Z's) at the same point
-
Z
Z
z
z
Z A
Z
Z ZZ A
Z Z
Z
-7 1 1 1 II
100
200
300
400
500
Mean
Figure 2. TEM analytical precision for airborne asbestos concentration:
coefficient of variation versus mean (fibers/cc).
37
-------
There was one field filter blank from each school for a total of 25.
Nine of these were selected at random and analyzed. Table 8 presents BCL's
results of the analysis of the nine field filter blanks. The filter blanks
were taken directly from the filter box, placed in a petri filter holder, and
carried to BCL along with the exposed samples. The analyst did not know which
samples were field blanks; these samples were prepared and analyzed like all
other samples. The low asbestos counts indicate that the laboratory or han-
dling contamination was no greater than 1.3 ng per filter.
Table 8. Quality Assurance Results of
Field Blank Analysis for Chrysotile
Sample no. ng/filter
14
18
28
80
83
87
91
99
125
0.060
1.0
0
0.80
0.12
0
0
0.13
0
a There was no amphibole detected.
Table 9 shows the results obtained from analysis of laboratory blanks.
The laboratory blank was either a blank filter in an ashing tube or an empty
tube placed beside each sample tube. Each sample was ashed in a test tube
(the test tubes were never reused), and each sample test tube had a blank test
tube placed beside it in the low temperature ashing chamber. The laboratory
blank provided a continuous check on possible contamination during the prepa-
ration procedure. These results show that a few fibers were picked up by the
blank during preparation, but this small amount had no effect on the quanti-
ties observed on the actual samples.
38
-------
II. BULK SAMPLES
Table 9. Quality Assurance Results
of Laboratory Blank Analysis
for Chrysotile
Blank no. ng/filter
E
E
E
E
E
E
E
E
E
E
E
E
E
E
E
E
E
E
E
E
E
1
2
3
4
5
6
7
8
9
10
11
39
42
51
56
61
64
73
83
89
100
1
1
1
0.3
0.6
0
0.4
0.3
0.08
0.2
0.09
0.4
0.1
0.1
0.3
0.1
0.08
0
0.2
0.1
0
Blank filters added to the
analysis scheme for a check on
laboratory contamination.
Three bulk samples and one duplicate were collected at each of the 48
air sampling sites for a total of 192 bulk samples. Twenty-four of the dupli-
cates were randomly selected for quality assurance analysis by Colorado School
of Mines Research Institute (CSMRI). The remaining duplicates, as well as
the balance of the 192 samples, were analyzed at Midwest Research Institute
(MRI).
A. Method of Analysis
The samples were analyzed by PLM according to the protocol given in Ap-
pendix D. For the analysis MRI used a stereo zoom microscope capable of 8X
to 40X magnification and equipped with a built-in illuminator and external
illuminator for oblique illumination, and a polarizing microscope capable of
100X magnification and equipped with an external illuminator and dispersion
staining objective.
39
-------
Each bulk sample was emptied onto a clean weighing paper, and the entire
sample was examined as a whole through the stereomicroscope for layering, ho-
mogeneity, and the presence of fibrous material. Identification of macro-size
nonfibrous components was usually possible at this point.
Four or more subsamples of the bulk sample were selected using the stereo-
microscope. They were then mounted onto a clean microscope slide in the appro-
priate index of refraction liquids for examination through the polarizing mi-
croscope.
The PLM procedure consisted of observing the characteristics of the sub-
sample components with transmitted polarized light, crossed polars, slightly
uncrossed polars, crossed polars plus the first-order red compensator, and
the central stop dispersion staining objective. The observations obtained
using the various techniques were used to identify the fibrous and some of
the nonfibrous components on the basis of morphology, signs of elongation,
and refractive index/dispersion staining colors.
Quantitation of the asbestos was achieved by stereomicroscopic observa-
tion of the entire bulk sample through the stereomicroscope and PLM examina-
tion of the subsamples. The volume percentages of the various components were
estimated in relationship to the whole sample.
B. Discussion
The results of all the analyses of bulk samples, including the quality
assurance analyses, are presented in Table 10. These results are grouped by
location and by site. Chrysotile asbestos was found in all the bulk samples,
and traces of amosite were found in three samples (539D, 641D, 583). The ma-
jority of samples contained 1 to 50% asbestos (by volume), which is one of
the three ranges the algorithm considers: less than 1%, 1 to 50%, and greater
than 50%.
A number of the samples presented analytical difficulties. Binders pres-
ent in these samples hindered the visibility of the fibers when using the po-
larizing microscope. In an attempt to eliminate this hindrance, the samples
were treated with hexametaphosphate to remove the binders. This did not work
well, possibly because of the small size of the fibers.
Five types of decorative insulation materials were seen in the study of
samples: a perlite-based material; a mixed perlite-vermiculite-based material
containing significant amounts of cellulose fiber; a glass-wool-based material;
and a double-layered insulation consisting of a vermiculite-based layer and
perlite-based layer. The perlite layer was the exposed surface at the sam-
pling sites.
A significant number of the bulk samples were double-layered, both layers
containing chrysotile asbestos. MRI examined and reported the two layers of
these materials separately. The outer layer data were used in the statistical
analysis of results.
40
-------
Table 10. Results of Polarized Light Microscopic Analysis of Bulk Samples for
Volume of Chrysotile, Size of Fibers, Coating on Fibers,
Releasability Rating of Material, and Nonasbestos Components
Bulk
School sample
no. Site no.
1 1 585^
585*
510
510*
510 QA
634 D
533 D
1 2 691^
691
613 f
534 D*
534 D*
678 D
678 D
1 3 504*1
504*
517*
517f
674 DJ:
674 D
674 QA
658 D8
Analysis
lab
M
M
M
M
M
M
C
M
M
M
M
M
M
M
M
M
M
M
M
M
M
C
Chrysotile
volume %
20
22
20
20
10
8
34
25
15
33
28
25
5
5
30
20
30
28
10
15
10
10
Fiber size
(by %)
F
50
25
M
50
75
L
0
0
Nonasbestos components
Coat R.RC volume %
6 Perlite 60, binder 19
Vermiculite 60, binder
3 Perlite/binder 75-80
15
Vermiculite/binder 75-80
20
40
70
75
70
90
90
100
50
30
70
50
30
10
20
10
10
0
50
30
10
10
0
15
5
0
0
0
0
40
Vermiculite 60
7 Perlite/binder 85
Perlite 52, gypsum 8
6 Perlite 75
Vermiculite 85, glass 2
7 Perlite/binder 65
5 Perlite/plaster 65
Vermiculite/binder 65,
cellulose 10
5 Perlite 95
Vermiculite 90
8 Perlite/binder 70
F Vermiculite 60, binder
8 Perlite/binder 70
Vermiculite 60-70,
binder 10-15
5 Perlite/binder 90
Vermiculite 60, binder
Vermiculite 60,
perlite 25, binder 5
Perlite 50, vermiculite
20, gypsum 18
(continued)
20
20
-------
Table 10 (continued)
Bulk
School sample
no. Site no.
2 1 672
6llf
611*
611 QA
655 D
646 D*
646 D
2 2 575
548
622 D
594 D
2 3 527
680
557 D
•f
568 D*
568 D
2 4 645
649
532 D
502 D
Analysis
lab
M
M
M
M
M
M
M
M
M
M
C
M
M
M
M
M
M
M
M
C
Chrysotile
volume %
5
8
15
5
8
10
18
35
23
25
15
12
10
1
5
5
10
1
10
2
Fiber size
(by %)
F
100
100
20
20
100
95
29
60
40
90
100
90
5
M
0
0
70
70
0
5
70
40
50
10
0
10
95
L Coat
0
0
10
10
0
0
1
0 F
10
0
0
0
0
R.R6
4
3
6
3
8
7
6
8
6
9
7
3
8
7
Nonasbestos components
volume %
Perlite/binder 95
Perlite/binder 90
Vermiculite 50, binder 25
Vermiculite 75,
perlite 10, binder 10
Vermiculite 85 , binder 5
Perlite/binder 90
Vermiculite 55-60, binder 20
Perlite/binder 65
Perlite/binder 75
Vermiculite/binder 75
Vermiculite 65, binder 10,
gypsum 5 , carbonate 5
Perlite/binder 85-90
Perlite 70, binder 20
Perlite/binder 20,
Vermiculite 60,
cellulose 20
Perlite 75, binder 15-20
Vermiculite 50, perlite
20, cellulose 20
Vermiculite 40, cellulose
perlite 15, binder 10
Vermiculite 50, cellulose
perlite 15, binder 5-10
Vermiculite 30, cellulose
perlite 20, binder 15
Vermiculite 33, perlite 55
(continued)
-------
Table 10 (continued)
School
no .
3
3
3
3
4
5
Bulk
sample
Site no.
1 636
523
508
540
540
2 620
623
643
629
3 673
656
615
525
4 552
633
698
581
1 521
647
697
562
1 576
640
640
592
692
D
D
QA
D
D
D
D
D
D
D
D
QA
D
D
Analysis
lab
M
M
C
M
M
M
M
M
C
M
M
M
M
M
M
M
C
M
M
M
M
M
M
M
M
M
n
Fiber size
Chrysotile (by %)
volume %
6
12
8
8
10
5
8
5
12
5
4
3
8
6
8
6
5
5
3
5
15
60
60
50
20
AM
TV/
F
100
99
100
60
95
95
100
100
100
100
100
100
100
100
100
99
100
100
0
0
20
0
0
M
0
1
0
40
5
5
0
0
0
0
0
0
0
0
0
1
0
0
10
10
40
20
20
L Coat
0
0
0
0
0 F
0 F
0
0
0
0
0
0
0
0
0 F
0
0
0
90
90
40
80
80
R.Re
4
6
5
4
4
4
5
5
4
4
4
6
6
6
4
5
3
4
4
3
o
^.
Nonasbestos components
volume %
Perlite 60,
Perlite 68,
Perlite 37,
Perlite 82,
Perlite 70,
Perlite 80,
Perlite 90,
Perlite 70,
Perlite 54,
Perlite 70,
Perlite 70,
Perlite 80,
Perlite 90
Perlite 80,
Perlite 80,
Perlite 80,
Perlite 48,
Perlite 70,
Perlite 70,
Perlite 70,
Perlite 70,
binder 32
binder 20
carbonate
binder 10
carbonate
binder 15
binder 2
binder 25
carbonate
binder 25
binder 26
binder 17
binder 14
binder 10
binder 5
carbonate
binder 25
binder 27
binder 23
binder 15
53
20
31
44
Glass wool/binder 40
Glass wool
Glass wool
Glass wool
u-Lasib wool
35, binder
50
75, binder
60
5
5
(continued)
-------
Table 10 (continued)
Bulk
School sample
no. Site no.
6 1 682
604
577 D
537 D
6 2 588
665
664 D
570 D
6 3 598
650
650 QA
639 D
551 D
7 1 580
632
681 D
545 D
7 2 599
696
669 D
542 D
7 3 612
612 QA
Ana lysis
lab
M
M
M
M
M
M
M
C
M
M
M
M
M
M
M
M
C
M
M
M
C
M
M
Chrysotile
volume %
10
10
10
25
20
15
10
40
15
15
30
25
22
5
5
8
2
12
12
8
7
8
7
Fiber
(by
F
45
90
10
10
10
5
5
25
20
50
10
25
100
100
90
100
100
100
90
60
size
%)
M
50
5
70
30
60
20
45
50
50
40
40
50
0
0
10
0
0
0
10
40
,1
L Coat
5
5
20
60
30
75
50
25 F
30
10
50
25
0
0
0
0
0
0
0 F,S
0
R.Re
8
8
8
6
9
7
8
9
8
8
8
4
5
3
4
5
6
3
Nonasbestos components
volume %
Perlite/binder 90
Perlite/binder 90
Perlite/binder 90
Perlite/binder 70
Perlite/binder 80
Perlite/binder 85
Perlite/binder 90
Perlite 45, cellulose 10,
binder 3
Perlite/binder 80-85
Perlite/binder 85
Perlite 60, binder 10
Perlite/binder 75
Perlite/binder 78
Perlite 70, binder 25
Perlite 75, binder 20
Perlite 65, binder 27
Perlite 59, carbonate 31,
gypsum 3
Perlite 70, binder 18
Perlite 70, binder 18
Perlite 70, binder 20
Perlite 43, carbonate 45
Vermiculite 82, binder 10
Vermiculite 85, binder/
clay/gypsum 8
(continued)
-------
Table 10 (continued)
Bulk
School sample
no. Site no.a
660
677 D
657 D
8 1 541
648
554 D
662 D
9 1 621
690
603 D
694 D
10 1 619
564
689 D
511 D
11 1 624
700
702 D
590 D
590 QA
12 1 701
561
670 D
Analysis
lab&
M
M
C
M
M
M
M
M
M
M
M
M
M
M
C
M
M
M
M
M
M
M
M
Chrysotile
volume %
14
6
4
70
60
66
50
60
50
45
40
18
20
20
15
20
15
12
30
15
10
15
5
Fiber size
(by %)
F
100
100
10
0
0
0
0
0
0
0
10
5
20
30
10
35
10
10
60
15
60
M
0
0
20
10
25
10
5
20
10
10
50
15
50
65
90
60
60
70
35
84
40
L Coat
0 S
0
70
90
75
90
95
80
90
90
40
80
30 S
5 S
0 S
5 S
30 F,S
20
5 S
1 S
0 S
R.R6
6
5
2
5
2
4
5
3
2
3
4
4
3
4
7
5
3
5
1
4
Nonasbestos components
volume %
Vermiculite 76, binder 10
Perlite 89, binder 5
Perlite 52, carbonate 45
Glass wool 30
Glass wool 30, plaster/
gypsum 10
Glass wool 33
Glass wool 50
Glass wool 35,
surface coat 5
Glass wool 50
Glass wool 50, binder 5
Glass wool 55, binder 5
Vermiculite 62, binder 20
Vermiculite 70, binder 10
Vermiculite 75, perlite 5
Vermiculite 74, clay 5
Vermiculite 65, binder 10
Vermiculite 65, binder 20
Vermiculite 75, binder 10
Vermiculite 60, binder 7
Vermiculite 60, carbonate
clay 15
Vermiculite 80, binder 10
Vermiculite 60, binder 25
Vermiculite 85, binder 5-1
(continued)
-------
Table 10 (continued)
School
no.
Bulk
sample
Site no.
Analysis
lab
Chrysotile
volume %
Fiber size
(by %)
F M L
Nonasbestos components
Coat R.R6 volume %
12
13
13
13
593 D
688
538
506 D
573 D
555
518
509 D
536 D
f
679*
679*
515*
515 f
f
609 D*
•F
609 D£
556 D*
•f
556 D
556 QA
62gf
•F
628*
571*
-F
571
571 QA
M
M
M
M
M
M
M
C
M
M
M
M
M
M
M
M
M
M
M
M
M
M
20
10
10
5
5
10
8
15
10
7
10
18
15
12
8
Vermiculite 50, other
glass 20, gypsum 5,
carbonate 5
10
25
17
20
15
10
8
2
80
25
95
20
100
100
100
20
75
5
80
0
0
0
0
0
0
0
0
0
0
S
F,S
S
F,S
5
4
2
4
3
5
4
Vermiculite 80, binder 10
Vermiculite 50, binder 25
Vermiculite 60, binder 20
Vermiculite 50, binder 30
Perlite 85
Perlite 90
Perlite 90
Perlite 27, carbonate 67,
100
100
100
100
50 40
100
100
clay 3
Perlite/gypsum 85
Vermiculite 50, binder 35
Perlite/binder 90-95
Vermiculite 50, binder 45
Perlite/binder 89
Vermiculite 40, binder 50
Perlite 85
Vermiculite 65, binder
10
Perlite 10,
binder 13
25
vermiculite 70
Perlite/binder 90
Vermiculite 50, binder 30
Perlite 85
Vermiculite 60, binder 25
Perlite 20, vermiculite
60, carbonate 10
(continued)
-------
Table 10 (continued)
Bulk
School sample
no. Site no.
512 D*
512 D
671 D
14 1 618
637
637 QA
661 D
582 D
14 2 530
528
528 QA
574 D
567 D
567 QA
14 3 503
627
627 QA
687 D
614 D
Analysis
labb
M
M
C
M
M
M
M
C
M
M
M
M
M
M
M
M
M
M
M
Chrysotile
volume %
15
8
2
10
14
10
10
10
18
10
12
15
20
10
20
5
7
10
15
£
Fiber size
(by %)
F
100
5
25
25
10
10
30
10
5
20
80
15
20
75
30
M
0
90
75
75
80
90
50
80
75
70
15
80
70
25
60
L Coat
0
5 S
0 S
0 S
10 S
0
20
10 S
20 F
10
5 F,S
5
10
0
10
R.Re
6
3
2
3
7
4
3
6
5
3
4
5
Nonasbestos components
volume %
Perlite 60, binder 25
Vermiculite 67, binder 25
Perlite 53, carbonate 32,
vermiculite 5, gypsum 4
Vermiculite/binder 90
Vermiculite 60, binder 20,
coating 5
Vermiculite 80, binder 5,
opaques 5
Vermiculite/binder 90
Vermiculite 75, binder 10
Vermiculite 70, binder 10
Vermiculite 60, binder 30
Vermiculite 75, binder 10
Vermiculite 60, binder 23
Vermiculite 60, binder 15
Vermiculite 80, binder 10
Vermiculite 60, binder 10,
cellulose 5, surface
coat 5
Vermiculite 60, binder 25
Vermiculite 80, binder 8,
cellulose 5
Vermiculite 70, binder 10
Vermiculite 55, binder 25
(continued)
-------
Table 10 (continued)
00
Bulk
School sample
no. Site no.
15 1 616
565
565 QA
535 D
642 D
16 1 519
543
675 D
586 D
17 1 601
526
526 QA
569 D
550 D
18 1 635
635
666
566 D
572 D
19 1 560
560 QA
607
547 D
563 D
Analysis
lab
M
M
M
M
C
M
M
M
M
M
M
M
M
C
M
M
M
M
M
M
M
M
M
M
Chrysotile
volume %
45
40
30
35
30
20
21
20
25
8
15
10
15
10
10
20
10
8
15
40
15
12
25
33
Fiber size
(by %)
F
0
0
10
0
10
5
10
5
100
95
60
90
95
100
100
95
40
60
90
25
15
M
10
10
20
5
65
90
30
90
0
5
40
10
5
0
0
5
50
40
5
75
65
L Coat
90
90
70
95
25
5
60
5 S
0 S
0 F
0
0 F
0
0
0
0
10
0
5
0
20
R.Re
2
3
4
4
5
6
7
4
4
3
5
4
2
4
8
7
7
8
Nonasbestos components
volume %
Glass wool 50
Glass wool 55
Glass wool 30, carbonate 1
Glass wool 65
Glass wool 35, carbonate 3
Vermiculite 75 , binder 5
Vermiculite 59, binder 20
Vermiculite 70, binder 10
Vermiculite 50, binder 25
Vermiculite 60, binder 20
Vermiculite 45 , binder 40
Vermiculite 80, binder 10
Vermiculite 50, binder 30
Vermiculite 45, carbonate
binder 10, gypsum 8
Perlite/binder 90
Vermiculite 55, binder 25
Perlite/binder 90
Perlite 70, binder 20
Perlite/binder 80,
cellulose 5
Perlite 60
Perlite 60, carbonate 25
Perlite 73, binder 10
Perlite 75
Perlite 66
(continued)
-------
Table 10 (continued)
Bulk
School sample
no. Site no.
20 1 693
549
589
584
584
20 2 505
659
625
641
21 1 522
630
605
539
21 2 653
513
544
644
22 1 602
668
683
516
22 2 583
600
608
676
D
D
QA
D
D
D
D
D
D
D
D
D
D
Analysis
lat>
M
M
M
M
M
M
M
M
M
M
M
M
C
M
M
M
M
M
M
M
M
M
M
M
M
p
Fiber size
Chrysotile (by %)
volume %
16
16
12
30
20
14
12
20
12
15
18
16
24
14
14
15
22
25
20
40
40
35
22
30
15
F
99
99
70
5
50
75
50
10
30
85
80
80
15
20
95
25
25
60
70
50
10
60
20
40
M
1
1
25
75
50
25
30
60
60
15
20
15
75
80
5
25
65
40
25
49
85
30
75
60
L Coat
0
0
5 F
20
0
0 F
20
30
10
0
0
5
10
0
0
50
10 S
0 S
5 S
1
5 S
10 F,S
5 S
0 F,S
R.R6
7
6
7
7
9
8
9
7
8
8
6
9
8
7
8
3
4
3
4
5
5
Nonasbestos components
volume %
Perlite 60,
Perlite 70,
Perlite 60,
binder 24
binder 14
binder 28
Perlite/binder 70
Perlite 65,
Perlite 80,
carbonate
binder 6
15
Perlite 90-95
Perlite/binder 78
Perlite 80,
Perlite 65,
Perlite 70,
Perlite 70,
Perlite 40,
Perlite 70,
Perlite 70,
Perlite 70,
Perlite 70,
Vermiculite
Vermiculite
Vermiculite
Vermiculite
Vermiculite
Vermiculite
cellulose
Vermiculite
Vermiculite
binder 8
binder 20
binder 11
binder 12
carbonate
binder 16
binder 14
cellulose
binder 3
50, binder
70, binder
50, binder
50, binder
45 , binder
60, binder
5
60, binder
75, binder
28
15
20
5-10
10
10
20
15,
10
10
(continued)
-------
Table 10 (continued)
t_n
O
Bulk
School sample
no. Site no.
23 1 638
553
684 D
529 D
24 1 610^
610*
559*
559 ,
+
651 D*
651 D1
591 D
24 2 546
597
685 D
514 D
25 1 507
558
652 D
520 D
25 2 654
595
524 D
578 D
578 QA
Analysis
lab
M
M
M
C
M
M
M
M
M
M
C
M
M
M
C
M
M
M
C
M
M
M
M
M
Chrysotile
volume %
8
16
12
19
6
9
10
8
5
11
8
18
18
15
20
5
8
7
2
12
10
10
5
5
Fiber size
(by %)
F
99
45
90
100
100
100
25
5
75
100
90
100
100
100
100
99
60
M
1
50
9
0
0
0
50
25
20
0
10
0
0
0
0
1
40
L Coat
0
5 F
1
0
0
0
25 F
75 S
5
0 F,S
0 S
0 F
0 F,S
0 F,S
0 F
0 F,S
0
R.R6
6
7
6
4
5
4
7
6
7
2
7
3
5
2
6
Nonasbestos components
volume %
Perlite 70, binder 28
Perlite 60, binder 14
Perlite 70, binder 18
Perlite 53, carbonate 20,
gypsum 4
Perlite 5, vermiculite 80
Perlite 5-10, vermiculite 70
Perlite 20, vermiculite 75
Perlite 20, vermiculite 42
carbonate 25
Vermiculite 70, binder 12
Vermiculite 72, binder 10
Vermiculite 65, binder 20
Vermiculite 70, binder 20,
surface coat 5
Vermiculite 60, binder 40
Vermiculite 75, binder 10
Vermiculite 49, carbonate
cellulose 5
Vermiculite 70, binder 18
Vermiculite 70, binder 20
Vermiculite 40, binder 50
Vermiculite 75, binder 20
Vermiculite 80, carbonate
binder 12
(continued)
-------
Table 10 (continued)
School
no .
Bulk
sample
Site no.
Analysis
lab
Chrysotile
volume %
c
Fiber size
(by %)
F M L
Nonasbestos components
Coat R.Re volume %
25
531
587
587 QA
695 D
606 D
M
M
M
M
15
5
2
15
100 0 0 F,S
100 0 0 F,S
100 0 0 F,S
S
3 Vermiculite 50, binder 30,
cellulose 5
4 Vermiculite 50, binder 40,
cellulose 5
Vermiculite 85, carbonate
5 Vermiculite 50, binder 25,
surface coat 10
Vermiculite 55, carbonate
binder 6, gypsum 5
a QA indicates duplicate analysis by different MRI analyst. D indicates duplicate samples. A double
sample was taken side by side at one of the three bulk sampling points at a site to provide a
duplicate for quality assurance.
b The laboratory that analyzed the sample: M for Midwest Research Institute (MRI) and
C for Colorado School of Mines Research Institute (CSMRI).
c Fiber size: Fine - single fiber of length and width difficult to pick out and mount.
Medium - single fiber or small bundles suitable to mount as they are for PLM.
Large - fibers and bundles of sufficient size that they would have to be subdivided
or separated for PLM.
d Coating: S indicates that there was coating on the suface; F indicates that there was coating on
the fibers themselves.
e Releasability rating.
f Double-layered sample, perlite surface exposed.
g CSMRI reported average of both layers.
-------
CSMRI noted layering and prepared their samples for analysis to include
a proportional amount of both layers. The subsample was then homogenized be-
fore slides were prepared for the PLM work. In most cases there was little
difference in the asbestos content of the layers of any given sample.
Characteristics other than the asbestos content that were observed from
the analysis of the bulk samples included the fiber or fiber bundle size, the
degree that the asbestos fibers were bound to the matrix or were coated with
a binder, and the toughness or brittleness of the insulation matrix. These
characteristics are not evident by sample examination without magnification,
but differences between samples are readily seen under the low power magnifi-
cation of a stereomicroscope. The fiber size and coating information are
given, respectively, in columns 6 and 7 of Table 10.
Examination of the results indicated that there were three major matrix
components (perlite, vermiculite, and glass wool), three dominant fiber sizes
(fine, medium, and coarse), two fiber coat categories (coated and not coated),
and four major ranges of asbestos content (0-10%, 10-20%, 20-30%, and greater
than 30% by volume).
The question arose as to whether this information was related to air
levels of asbestos. Tentative subjective ratings of these data according to
the apparent availability of releasable fibers from the bulk material indi-
cated a possible relationship between the ratings and the air levels. There-
fore, each bulk sample was reexamined with a low power stereomicroscope and
rated on an arbitrary scale of 1 through 9 for the apparently available free
asbestos fibers, or the asbestos releasability factor. The results of this
subjective rating, which was based on asbestos content, fiber size, brittle-
ness of the matrix, and apparent freedom of the individual asbestos fibers,
are shown in Figure 3.
The numbers on the ordinate identify the schools and the sites at the
schools where the bulk samples were taken. Of the 48 school sites, 23 sites
had three bulk samples rated, 23 sites had four bulk samples rated, and 2
sites had two bulk samples rated. The abscissa gives range of releasability.
The horizontal lines represent the range of releasability (1 to 5) of the
samples taken from a site. The numbers appearing above the horizontal lines
represent the number of bulk samples taken from a site that were rated for
releasability. The numbers in brackets to the right of the horizontal lines
are the releasability ratings, on a scale of 1 to 9, of each bulk sample taken
from a site. The numbers in parentheses within the brackets are the duplicate
samples. The number in parentheses to the right of the brackets is the aver-
age of the values within the brackets. The average was calculated by first
averaging the duplicates and then using that value with the remaining values
in the brackets to calculate the average number in parentheses to .the right
of the brackets.
The information in Figure 3 indicates that there is a significant varia-
tion in releasability rating values among those samples taken from the same
site. The largest spread in releasability ratings for a set of samples from
a site is 5; the lowest is zero; and the average of all 48 sites is 2.3. The
reason for this variability can be that the material on the ceiling at a site
52
-------
-2
a.
.*
3
ca
c
0
~o
o
25-3
25-2
25-1
24-2
24-1
23-1
22-2
22-1
21-2
21-1
20-2
20-1
19-1
18-1
17-1
16-1
15-1
14-3
14-2
14-1
13-3
13-2
13-1
12-2
12-1
11-1
10-1
9-1
8-1
7-3
7-2
7-1
6-3
6-2
6-1
5-1
4-1
3-4
3-3
3-2
3-1
2-4
2-3
2-2
2-1
1-3
1-2
1-1
- — RSI 3 rt 1(41
-4-[6,7](6.5) O')-^
—4— f(4).4,5](4.3)
— S [(6), 6, 7] (6. 3)
2[(5),5](5)
_ ..4, _ _ ,_ (T7 8) 8 9 1 ( 8 ?)
[~(AI g 81 (7 3)
..4 ,. r?7 0) ft oi(g 3)
[~(7 71 6 71 (6 71
4 f(7 8) 7 8 1(7 5)
4 r/2 41 4 5"! (41
— L_[(3),4,4](3.7)
4 . F(6 71 4 5~H5 21
r ' •* M T "iTfX 71
4 rp A) A ?](<; 3)
3 r/3\ T 2~\ (f 71
J F(6) 4 5~i(5)
3 r/^v 5 S^-> / j\' J
4 ,-... , l"..\...' - f(2 '41' 4 51 (4)
4 -_.. ffj 51 4 7~U51
l_V-i,3J,t ,/ J V-j;
-J [(3),4,4](3.7)
_4_[(2,3),3,5](3.5)
4 T(? 4) ? ^]('' "»)
_ J,_ _ ..!"( A) 4 5l(?1
...u,3 , . . , V( 3) 4 *>! (4)
_J [(8,8),8,9](8.3)
3 r/g\ 7 o"l (g\
,..4 „ ., rii R1 fi ft 1(7 7)
4 ^(? •?} A 4]p ^)
3 r(A) 4LA"i(s •?)
_i — [(4,4),5,5](4.7)
3 [(4), 4, 4 ](4)
- . .. . . !~(^} 4 A1(s)
R^l 7 8l (71
4 _. • - Q 3 6) 3 4] (3 8)
3 u> "" ' ,~V[J7) 3 g]{A
4 ri-> 9) 6 8l (^ 3)
4 r/ c c\ 6 7~i / 6\
HO, j;, o,- j ^y
J ..... ... ri T1 1 6l (5 31
, , ^(O.a.fijis.-j
3)
1234
Releasability Rating Range
HOTES
1. The numbers runni.ig vertically at :he left identify Me
schools and the sites at the schools where the bulk samples were taken.
Of the 4S school sites, 23 sites had three bulk samples rated, 23 sites
had four bulk samples rated, an-J 2 sites had two bulk samples rated.
2. The horizontal lines represent the range of reieasabi1ity
(1 to 5) of the samples taker from a site.
3. The numbers appearing above the horizontal lines represent
the nunOer of bulk samples tanen from a site that were rated 'or re|6asa-
btlUy.
4. The numbers in brackets to the right of the horizontal lines
are the releasability ratings, on a scale of 1 to 3, of each bulk samole
taken from a site. The numbers in parentheses within the brackets =re the
duplicate samples. The number in parentheses to the right of the orackets
Is the average of the values within the brackets. The average «as calcu-
lated by first averaging the duplicates and then jsing that /aloe with the
remaining values in the brackets to calculate the average numoer in ,i<:renthes<
to the right of the brackets.
Figure 3. Releasability ratings and the range of the ratings of
bulk samples from each of the 48 sites sampled.
53
-------
is not homogeneous, that the subjectivity bounds are that great, or both. It
may well be that more experience in this type of rating could narrow the vari-
ability bounds in the rating. The spread in the asbestos content for sets of
bulk samples is considerably less than the spread in the releasability ratings.
This releasability work was statistically analyzed, and the results are given
in Section 7.
C. Quality Assurance
All samples were handled according to the sample handling procedure given
in the protocol for Sampling and Analysis of Insulation Material Suspected of
Containing Asbestos, which can be found in Appendix D. Neither the relation-
ship of the samples with each other nor their individual source was known to
the analyst. The Protocol for Creating and Maintaining Chain of Custody,
which can be found in Appendix C, was followed by MRI.
The side-by-side duplicate samples were divided blindly and at random
into two groups of 24 each. One set of samples was sent to CSMRI for analy-
sis. In addition, 10% of all samples retained at MRI were randomly selected
and given an independent examination by a second MRI analyst.
The CSMRI analyst prepared four to five representative subsamples from
each of the 24 quality assurance bulk samples. Each of these macrosize sub-
samples was thoroughly mixed using tweezers. From these subsamples, three
slides were prepared as oil dispersion mounts of portions of the material for
examination through a polarizing microscope with a 60-W tungsten light source,
set up for 44X and 125X magnification. Components were identified on the
basis of morphology, refractive index, and other optical properties. Quanti-
tation was achieved by observing the microsize subsamples with the microscope.
The analytical procedures used by CSMRI were basically the same as those-
used by MRI, but there were some differences. MRI included a low power stereo-
microscopic examination for determining sample layering and gross uniformity.
Also, MRI used a polarizing light microscope equipped with a dispersion stain-
ing objective as an aid in refractive index determination. PLM examinations
by MRI were made at the single magnification of 100X.
The results of the quality assurance analyses in terms of percentage rel-
ative variation (standard deviation divided by mean, multiplied by 100) are
as follows. For the MRI laboratory replications, the relative variation was
36%. This value may be interpreted as an estimate of analytical error associ-
ated with PLM. For the side-by-side duplicate samples, the relative variation
for samples analyzed at MRI was 52%; for the samples that were divided between
MRI and CSMRI, the relative variation was 57%. The relative variation observed
at MRI (52%) is an estimate of variation consisting of three components--side-
by-side sampling variation, intralaboratory variation, and analytical error.
The relative variation observed between MRI and CSMRI (57%) is an estimate of
variation based on four components--side-by-side sampling variation, intra-
laboratory variation, interlaboratory variation, and analytical error.
54
-------
These error levels are consistent with error estimates reported in the
literature for PLM (USEPA 1982b). The small difference in analysis of side-
by-side samples between MRI and CSMRI (52% versus 57%) indicates that the PLM
protocol was well controlled.
An alternate view of the quality of the bulk sample data is found in a di-
rect analysis of the variance components for these measurements. An analysis
of variance components was conducted using VMCPNLS, a general purpose program
for a completely nested, multistage design (Shah 1979). The variance compo-
nents are summarized by percentage contribution in Table 11. The variation
across schools and sites, which are the primary variables being studied, far
exceeds the variation due to sampling and laboratory analysis. These results
indicate that the sampling and analysis protocols were in control at a level
sufficient to produce data that are consistent with their intended use in the
study.
Table 11. Variance Components for PLM Measurement of Bulk Samples
Component
Percentage of total variation'
Schools
Sites
Location
Side-by-side
Laboratory replication
53
31
7
7
2
a Rounded to the nearest percent.
55
-------
SECTION 7
STATISTICAL ANALYSIS
A major emphasis of this study was the statistical analysis of the data
collected at the sites in the schools. The main purpose of this analysis was
to document the air levels, bulk levels, assessment factors, and algorithm
scores, and then to examine the relationships between these levels and the
factors and scores. In particular, it was of interest to determine how well
airborne asbestos levels could be predicted by the various variables one at a
time and in combination. Since amphibole levels were generally scarce, only
chrysotile data were analyzed.
As in any data analysis, a great many exploratory analyses were carried
out with the data, a subset of which produced useful results. Accordingly,
this section describes principal results and does not present the many pre-
liminary analyses that were undertaken. The following results are reported
in this section.
(1) Average airborne chrysotile levels by general location (ambient,
indoor control, asbestos-containing material site);
(2) Average airborne chrysotile levels by individual assessment factors;
(3) Average airborne chrysotile levels by a dichotomized algorithm
score;
(4) Airborne chrysotile values by bulk chrysotile percentage and bulk
sample components;
(5) Average airborne chrysotile levels by bulk sample releasability;
(6) Average airborne chrysotile levels by cleaning category; and
(7) Variability of airborne chrysotile values over time.
The analysis methods used are discussed in more detail in the following
subsection. The regression analyses are focused upon the relationship between
airborne chrysotile values and a set of possible predictors or independent
variables. The best-predicting regression equations have been identified by
examining the empirical performance of various candidate models. Alternative
classification analyses are also discussed. These analyses are focused upon
predicting whether or not the airborne chrysotile level at a specified site
is higher or lower than some specified reference level. The classification
analyses include unweighted discriminant analyses, two analyses based upon an
57
-------
algorithm score dichotomy or a releasability dichotomy, a series of analyses
based upon regression models, and a series of decision tree analyses based
upon the assessment factors and/or releasability.
I. ANALYSIS METHODS
The summary statistics presented in this section quite often include
three measures of central tendency, namely, the arithmetic mean, the median,
and the geometric mean. The arithmetic mean is given because it is the most
familiar measure of central tendency; however, it can be unduly inflated when
the distribution of the data is highly skewed. In the present case, the dis-
tribution of air levels does appear to be skewed; therefore, the geometric
mean and median are also presented. These latter two statistics are consid-
ered better measures of central tendency for skewed data.
Probability values (p-values) from statistical tests of hypotheses con-
cerning means and correlations are presented in tables and figures. Formally,
the p-value associated with means indicates the probability of erroneously
rejecting the statistical hypothesis that the means of interest are equal,
whereas the p-value associated with a correlation indicates the probability
of erroneously rejecting the statistical hypothesis that the correlation of
interest is zero. Practically, the p-value associated with means indicates
whether or not the size of the difference between means can be attributed to
random variation, whereas the p-value associated with a correlation indicates
whether or not the size of the correlation can be attributed to random varia-
tion. The smaller p-values indicate a larger statistical difference between
means, or between a correlation and zero, while the larger p-values indicate
a smaller statistical difference between means, or between a correlation and
zero.
In most cases, population estimates (weighted estimates) are presented
rather than unweighted estimates. This reflects the fact that the data col-
lected in this project were collected under a probability sampling framework;
therefore, by taking this into account (i.e., by correctly weighting the data),
estimates for the entire urban school district may be obtained even though
all schools in the district were not sampled. For example, when the popula-
tion mean of air levels for asbestos-containing sites is presented, this is
an estimate of the mean for all asbestos-containing student activity sites in
the entire urban school district. In those cases where unweighted estimates
are given (i.e., estimates that ignore the sampling weights), this reflects
the fact that preliminary statistics are being examined before the final
weighted estimates are presented.
In the analyses presented in this section, the factor scores used are
the consensus scores of all five raters. This was done to avoid an excessive
number of analyses and because analysis of rater differences did not indicate
large rater-to-rater disagreement.
58
-------
II. AIRBORNE CHRYSOTILE CONCENTRATION AT ASBESTOS-CONTAINING MATERIAL SITES,
CONTROL SITES, AND OUTDOOR AMBIENT SITES
A. Exposure Levels
The first objective of this study was to document probable exposure to
airborne asbestos resulting from asbestos-containing materials in schools. In
general, higher levels of airborne chrysotile were found in student activity
areas with asbestos-containing materials, and relatively lower levels of air-
borne chrysotile were found in student activity areas without asbestos-
containing materials (indoor controls in schools with asbestos-containing
material sites). The lowest levels were found at ambient sites on the roofs
of the schools.
Table 12 gives the estimated air levels of chrysotile for the asbestos-
containing material sites, the control sites, and the ambient sites. Reading
from Table 12, the average air level of chrysotile in asbestos-containing ma-
terial sites is 179.46 ng/m3 with a standard error of the mean of 41.99 ng/m3.
The geometric mean is 80.45 ng/m3 with a standard error of 23.62 ng/m3. The
minimum value observed is 0 ng/m3, the maximum is 644 ng/m3, and the median
is 92.70 ng/m3. These population estimates are based on a sample of 48
asbestos-containing material sites. It is estimated that there are 2,698
asbestos-containing material sites in the school district.
The test of difference in air level means among asbestos-containing mate-
rial sites, control sites, and ambient sites is significant at the level < .01
For geometric means, this test is significant at the level < .01. Addition-
ally, differences in air level means and geometric means were tested between
pairs of site types. The levels of significance are as follows: asbestos-
containing material versus ambient, mean significant at < .01 (geometric mean
significant at < .01); control versus ambient, .03 (.03); and asbestos-
containing material versus control, .01 (.02). These test results indicate
that generally exposure to airborne asbestos is higher in rooms containing
asbestos materials than in rooms without. The results also indicate that
airborne asbestos levels inside buildings with asbestos materials are higher
than outdoor ambient levels.
Figure 4 contains a boxplot of airborne chrysotile levels based upon the
48 asbestos-containing material sites, 19 of the indoor control sites, and 25
ambient sites. (Two control sites were omitted because friable material was
found, and four were omitted because they were not student activity areas.)
The boxplot emphasizes the airborne chrysotile gradient observed in the pres-
ent study; for example, the 25th percentile of the airborne chrysotile distri-
bution at the asbestos-containing material sites is only slightly larger than
the 50th percentile (median) observed at the indoor control sites. A similar
shifting of the airborne chrysotile distributions is evident when the data
from the control sites are compared to the data from the ambient sites.
59
-------
Table 12. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites, Control Sites, and
Ambient Sites
Asbestos-containing
Population estimates
Mean (p < .01)a
Standard error of the
mean
friable material
sites
179.46
41.99
Control
sites
53.09
19.64
Ambient
sites
6.10
4.34
Geometric mean,
(P < .01)°
Standard error of the
geometric mean
Minimum
Median
Maximum
Number of sample sites
Estimated number of
population sites
80.45
23.62
0.00
92.70
644.00
48
2,698
13.15
8.35
0.00
21.80
362.00
19C
2,077
2.16
1.07
0.02
0.90
40.60
25
109
Test of difference (in airborne chrysotile means) among asbestos-containing
friable material sites, control sites, and ambient sites: level of signi-
ficance < .01.
Test of difference (in airborne chrysotile geometric means) among asbestos-
containing friable material sites, control sites, and ambient sites: level
of significance < .01.
Of the twenty-five control sites, two were omitted because asbestos-
containing friable material was found, and four were omitted because they
were not student activity areas.
60
-------
300
250
^ 200
150
100
50
Asbestos-Containing
Friable Material
Site Value Above
300 ng/m3:
644
566
484
425
422
420
332
312
75th percentile (245)
Mean (179)
50th percentile (93)
25th percentile (33)
75th percentile (66)
Mean (53)
50th percentile-(22)
Mean (6)
75th percentile (4)
' 50th percentile (1 )
Asbestos-Containing
Friable Material Sites
Control Sites
Ambient Sites
Figure 4. Boxplot of airborne chrysotile concentrations (ng/m3) for
asbestos-containing material sites, indoor control sites, and outdoor
ambient sites (population estimates).
61
-------
B. Schoolwide Elevated Levels
As discussed in the preceding subsection, the average airborne chrysotile
level is significantly higher for control sites, as well as for asbestos-
containing material sites, than for ambient sites. This result demonstrates
that exposure to airborne asbestos probably occurs beyond those sites which
actually contain the asbestos materials.
In 15 of the 25 sample schools, the air level of chrysotile in the con-
trol site appears to be elevated above the ambient site's level. Air sampling
results for these 15 schools are listed in Table 13. In nine of the schools,
the control site's air level of chrysotile exceeds the air level of chrysotile
in at least one of the school's asbestos-containing material sites.
When the analytical focus shifts from the school district level of aggre-
gation to the individual school level, a problem develops. Specifically, this
study was designed to yield reasonably precise estimates for the school dis-
trict as a whole. Hence, the sample sizes within schools are too small to
yield precise estimates at that level. Indeed, the precision or the variance
of an estimate cannot be obtained when there is only one observation of a kind
at a school (e.g., a single indoor or outdoor control site). Thus, due cau-
tion should be exercised when comparing individual airborne chrysotile values
(asbestos-containing material site/inside control/outside control) at the
school level.
The transport of airborne asbestos from its source to other school areas
is not fully understood. In this study, any student activity area without
asbestos-containing material was eligible to be an indoor control site; there
was no requirement that indoor control sites be a certain distance from
asbestos-containing material sites. Any further consideration might include
airflow and ventilation, traffic patterns, etc. In the second school (School
3) on the list in Table 13, all student activity areas have asbestos-
containing material. An office was selected as the control site for this
school, and the hallway outside the office and classrooms near the office had
asbestos-containing material. In light of this, it is reasonable to expect
an elevated air level of chrysotile in this control site. A different situa-
tion is present in the sixth school on the list (School 10). The only
asbestos-containing material in the school is in the asbestos-containing ma-
terial site sampled and small areas close to it. The control site is located
on the floor above, where there are no sites with asbestos-containing material.
For this school, a reasonable explanation of the elevated air level of chryso-
tile observed in the control site is not immediately obvious.
C. Levels of the Assessment Factors
Tables 14 through 21 describe the air levels of chrysotile at asbestos-
containing material sites, by factor levels, for each of the following assess-
ment factors: condition of material, accessibility, airstream status, expo-
sure, water damage, activity level, friability, and chrysotile content. Table
22 summarizes results of testing differences in air level geometric means among
levels of each factor. Differences were found to be significant for chrysotile
content. None of the other differences were found to be significant.
62
-------
Table 13. Schools Where Airborne Chrysotile Concentration at Indoor
Control Site Appears to be Elevated Above Concentration
at Outdoor Ambient Site
School no.
1
3
5
6
8
10
11
14
15
16
18
20
21
24
25
Ambient
0
2
17
0
1
0
0
0
10
1
0
0
3
2
0
0
0
0
1
.40
.70
.60
.00
.44
.36
.30
.55
.80
.10
.90
.77
.50
.50
.57
.35
.63
.38
.30
Airborne
site
(wi)a
(W2)
(W3)
(Wl)
(W2)
(W3)
chrysotile
Control
69.
166.
14.
45.
16.
43.
362.
72.
50.
73.
51.
49.
21.
69.
46.
288.
19.
25.
116.
30
00
60
50
90
90
00
90
30
40
00
80
80
30
80
00
00
20
00
concentration (ng/m
site
(Wl)
(W2)
(W3)
(Wl)
(W2)
(W3)
3)
Asbestos -containing
friable material sites
28
8
108
105
27
6
11
38
332
23
37
82
43
0
10
0
1
11
77
150
80
69
425
7
420
566
153
484
77
11
8
4
245
.40
.95
.00
.00
.50
.17
.00
.20
.00
.60
.00
.80
.00
.10
.00
.76
.50
.20
.60
.00
.00
.30
.00
.93
.00
.00
.00
.00
.80
.00
.58
.30
.00
(Wl)
(W2)
(W3)
(Wl)
(W2)
(W3)
Wl = sampling week 1; W2 = sampling week 2; and W3 = sampling week 3
63
-------
Table 14. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Condition of Material
Condition
Population estimates No damage Moderate damage Severe damage
Mean (p = .24)* 206.28 121.76 NA"
Standard error of the 56.90 37.15 NA
mean
Geometric mean 103.58 45.99 NA
(p = -15)a
Standard error of the 36.60 18.80 NA
geometric mean
Number of sample sites 32 16 0
a Level of significance for test of difference between means.
* NA = not applicable.
700
600
£? 500
O)
c
T 400
0
VI
X
6 300
0)
0
| 200
100
0
LEGEND: A = 1 Obs, B = 2 Obs, etc.
A
—
A
A
B A
A
A
A
- A
A A
A
B A
E A
B B
C B
C D
9 c
1 1 1
Total
179.46
41.99
80.45
23.62
48
No Damage
Moderate Damage
Condition
Severe Damage
64
-------
Table 15. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Accessibility of
Material
Accessibility
Not
Population estimates accessible
Mean (p = .99)a NA"
Standard error of the NA
mean
Geometric mean NA
(P = -49)a
Standard error of the NA
geometric mean
Number of sample sites 0
Rarely
accessible Accessible Total
179.34 180.92 179.46
45.41 124.87 41.99
87.23 28.37 80.45
25.59 44.06 23.62
44 4 48
a Level of significance for test of difference between means.
* NA = not applicable.
LEGEND: A = 1 Obs, B = 2 Obs, etc.
700
600
«* 500
u>
C
T 400
o
X
L_
u 300
0)
C
o
_Q
» 200
100
0
A
-
A
A
B A
—
A
A
-
A
_ A
B
A
C
F
D
E
F A
_ H B
1
Not Accessible
Rarely Accessible
Accessibility
Accessible
65
-------
Table 16. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Air Stream Status
Population estimates
Mean (p = .08)a
Standard error of the
mean
Geometric mean
(p = -47)a
Standard error of the
geometric mean
Number of sample sites
Air Stream
Not in air stream
115.89
25.65
65.69
14.00
25
Status
In air stream
239.86
65.45
97.49
48.26
23
Total
179.46
41.99
80.45
23.62
48
a Level of significance for test of difference between means.
LEGEND: A = 1 Obs, B = 2 Obs, etc.
PO
\
c
0)
o
X
_c
U
c
o
-Q
3
700
600
500
400
300
200
100
0
—
—
—
B
-
A
A
~ C
C
B
C
_ E
A
A
A
A
A
A
A
B
C
B
B
B
E
1
Not in Air Stream |n Air Stream
Air Stream Status
66
-------
Table 17 Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Proportion of
Material Exposed
Population estimates
Mean
Standard error of the
mean
Geometric mean
Standard error of the
Not
exposed
*
NA
NA
NA
NA
Exposure
10% or less
exposed
NA
NA
NA
NA
Greater than
10% exposed
179.46
41.99
80.45
23.62
Total
179.46
41.99
80.45
23.62
geometric mean
Number of sample sites
48
48
* NA = not applicable.
67
-------
Table 18. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Water Damage of
Material
Water damage
Population estimates
Mean (p = .70)a
Standard error of the
mean
Geometric mean
(P = -28)a
Standard error of the
geometric mean
Number of sample sites
None
170.78
65.77
57.56
27.52
30
Minor Moderate or major Total
178.49
36.37
120.51
25.52
13
289.73
126.89
164.67
109.34
5
179.46
41.99
80.45
23.62
48
a Level of significance for test of difference between means.
LEGEND: A = 1 Obs, B = 2 Obs, etc.
/uu
600
°°E 500
\
CO
c
>—x
TT 400
0
>.
-C
^ 300
c
o
< 200
100
0
A
—
A
A
B A
—
A
A
—
A
— A
B
A
B A
C B A
A B A
D A
F A
— H
1
B
1 1
Minor
Water Damage
Moderate or Major
68
-------
Table 19. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Activity
Population estimates
Mean (p = .91)a
Standard error of the
mean
Geometric mean
(p = -76)a
Standard error of the
geometric mean
Number of sample sites
None or low
172.24
42.58
83.77
33.91
25
Activity
Moderate
200.36
89.32
94.58
43.01
18
High
142.67
92.07
39.85
40.44
5
Total
179.46
41.99
80.45
23.62
48
a Level of significance for test of difference between means.
LEGEND: A = 1 Obs, B = 2 Obs, etc.
700
600
"g 500
O)
c
•*-'
J> 400
VI
X
r
U 300
0)
c
o
_D
^ 200
100
0
—
A
_ _
B
—
A
A
— A
A
A
_
E
A
C
C
— E
None or Low
A
A
A
A
C
A
B
B
C
C
Moderate
Activity
A
A
A
B
High
69
-------
Table 20. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Friability of Material
Population estimates
Mean (p = .28)a
Standard error of the
mean
Geometric mean
(p = -57)a
Standard error of the
geometric mean
Number of sample sites
Low
199.86
77.50
102.54
50.74
12
Friability
Moderate
187.12
55.55
85.49
31.14
29
High
99.00
32.65
35.23
31.52
7
Total
179.46
41.99
80.45
23.62
48
a Level of significance for test of difference between means.
LEGEND: A = 1 Obs, B = 2 Obs, etc.
700
600
£* 500
E
£
^ 400
0
X
b.
5 300
-------
Table 21. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Chrysotile Content
of Material
Population estimates
Mean (p < .01)b
Standard error of the
mean
Geometric mean,
(p < -01)b
Standard error of the
geometric mean
Number of sample sites
Chrysotile Content
Low
(£ 1%)
NA'V
NA
NA
NA
0
Moderate
(> 1%, ^ 50%)
184.72
42.74
89.92
24.55
46
High
(> 50%)
0.96
0.07
0.93
0.10
2
Total
179.46
41.99
80.45
23.62
48
a As categorized in the original algorithm.
b Level of significance for test of difference between means.
* NA = not applicable.
Table 22. Summary of Relationships Between Airborne Chrysotile Concentration
(ng/m3) and Original Algorithm Factors for Asbestos-containing
Friable Material Sites
Algorithm factors
Sign of regression
relationship
Results of test
Condition
Accessibility
Air stream status +
c
Exposure NA
Water damage +
Activity
Friability
Chrysotile content
b
NS
NS
NS
NA
NS
NS
NS ,
fl
Sig.
a Based on geometric means (Tables 14-21).
b NS = not significant at .10 level.
c NA = not applicable.
d Sig. = significant at .10 level.
71
-------
Though not significant, as water damage increases (from none to minor to mod-
erate or major) the average air level of chrysotile increases. For the fac-
tors of activity, friability, and chrysotile content, though the differences
among factor levels were not found to be significant, the directions of the
relationships were not consistent with those expected. For example, among
activity levels (Table 19), the average air level of chrysotile is lowest for
high activity. In interpreting these results, it should be noted that only
one factor at a time is being considered; i.e., there are no adjustments for
extraneous factors.
These adjustments may be necessary because of the possible confounding
effects of extraneous factors upon the relationship of interest. For example,
when the relationship between friability and airborne chrysotile levels is
being characterized, the possible effect of the bulk chrysotile percentage,
and water damage could be taken into account by including friability, bulk
chrysotile percentage, and water damage in an appropriate statistical model.
Thus the bivariate analyses should be viewed as preliminary until the appro-
priate multivariate analyses are completed.
Table 20 describes air levels of chrysotile by the factor friability,
which has three levels — low, moderate, and high. For sites with a friability
rating of low, the mean air level of chrysotile is 199.86 ng/m3 with a stan-
dard error of 77.50 ng/m3. The geometric mean is 102.54 ng/m3 with a standard
error of 50.74 ng/m3. There are 12 sample sites with a low friability rating,
29 sites moderate, and 7 sites high. Differences in geometric means among
friability levels are very significant (p = .57). However, geometric means
decrease (from 102.54 ng/m3 to 85.49 ng/m3 to 35.23 ng/m3) as friability in-
creases. This is the opposite of the relationship originally hypothesized--
that levels of chrysotile increase as friability increases.
In Table 14, note that there are no sample sites with a condition rating
of severe damage. It follows that no conclusions can be drawn from this study
about the effect of severe damage on air level of chrysotile. Also, none of
the sample sites has an accessibility rating of not accessible (Table 15).
All 48 asbestos-containing material sites in the sample have the same exposure
rating, greater than 10% exposed, so no comparisons of air levels for differ-
ent exposure categories can be made (Table 17).
D. Algorithm Score
Table 23 gives the distribution of the exposure assessment algorithm
scores (formed from consensus factor ratings) estimated for the asbestos-
containing material sites in the school district. The mean score is 29 with
a standard error of 3.2. The median score is 28. The information in Table
23 is useful for judging whether or not a given score value is high or low
with respect to the score distribution in the school district.
Figure 5 is a plot of the air level of chrysotile versus the algorithm
score for the 48 asbestos-containing material sites in the sample. The
Pearson correlation coefficient (-0.17) is not significantly different from
zero (p = .25). This indicates that there is no linear relationship between
the algorithm score and the air levels of chrysotile observed during the study.
72
-------
Table 23. Exposure Assessment Algorithm Scores for the Asbestos-
Containing Material Sites in the School District
Population estimates
Exposure
assessment
algorithm
score
Mean
Standard error of the mean
Minimum
Median
Maximum
Quantiles
10
20
30
40
50
60
70
80
90
Number of sample sites
Estimated number of population sites
29
3
10
28
108
12
16
24
24
28
28
30
36
40
48
2,698
73
-------
700 r
600-
500-
•> 400-
8
11
c
o
300-
200-
100-
0-
t
A
A A
A
X
,
A
A A
A A A X
A X
A A A A
A B X
A B X
1 f 1 1 1 1
Pearson Correlation Coefficient = -0. 17
P-value = 0.25
AltoritBB icore category
Population ettuates la* (i 28) Kiih f It} Tottl
Mean (p « 0.12)' 216.33 109.77 171. .«
Standard error of Che 59.31 :t73 .!.»
A xan
SmeKtric mean 100.49 52 52 iO. .i
(p = 0.24)'
Standard error of toe 37. SS V9. SO :}.QZ
k geonecrlc oeaa
Hutber of itmflt ncea 29 19 •>
A
a Level of significance for teat of difference between neana.
,
A
A
A
k A A
k A A
A
B A
k ? |A A | | A | | | | 1 1 *_J
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 HO
Algorithm Score (Consensus Rating)
Figure 5. Airborne chrysotile concentration (ng/m3) versus the
consensus score rating for the 48 asbestos-containing material sites
74
-------
(A strong positive linear relationship would be exhibited by all points on
the plot lying close to a straight line from lower left to top right).
The table in Figure 5 shows the air levels of chrysotile (population es-
timates) by algorithm score category. The low/high score division is made at
the weighted median score of 28. The difference in geometric means between
score categories is not significant. The difference in means has a p-value
of .12, and the mean air level is lower for the high algorithm score category.
E. Bulk Chrysotile Levels
Bulk chrysotile levels are summarized in Table 24. Figure 6 displays a
plot of the air level of chrysotile versus the percentage of chrysotile in
asbestos-containing material for the 48 sample sites. The Pearson correlation
coefficient (-0.06) is not significantly different from zero, indicating that
there is no linear relationship between the bulk percentage of chrysotile and
the air level of chrysotile. The table in Figure 6 shows the air levels of
chrysotile (population estimates) by chrysotile content category. Chrysotile
content < 20% was classified as low, ^ 20% as high. It should be noted that
the original algorithm categorized at ^ 50% and > 50%; however, for the cur-
rent data, only two sites have bulk levels greater than 50% (Table 21.) The
differences in means and geometric means between asbestos content categories
are not significant.
Table 24. Bulk Chrysotile Content at the
Asbestos-Containing Material Sites
Population estimates
Chrysotile content (%)
Mean
Standard error of the mean
Minimum
Median
Maximum
Number of sample sites
16
2
5
13
63
48
F. Other Major Components
During laboratory analysis, the other major components in each bulk sam-
ple were observed--perlite, vermiculite, or glass wool. Figure 7 displays a
plot of the air level of chrysotile versus the other major components in the
site's asbestos-containing material, and the table in Figure 7 summarizes
these data. The mean and the geometric mean air levels of chrysotile are
significantly different among the other major components. The average air
level of chrysotile is highest for perlite and lowest for glass wool.
75
-------
E
O)
,c
*•**
_o
'•§
SL
k.
_C
u
0)
£
t
;£
*JC
700
600
500
400
300
200
100
0
—
-
A
A
A A
-
A
_
A
— A
A A
A
A A
A AA A /
A A
AA A A
A A A AA A
— B AA A AA
1 1 1 1
LEGEND: A = 1 Obs, B = 2 Obs, etc.
A
Pearson Correlation Coefficient = -0.06
P-value = 0.71
Chrviotile conte&t
A
Hean (p » .37)' 151.88 254.76
Standard error of the 30.58 111.67
mean
A Geometric Man 77 26 91.76
(p - .S2)4
Staadard error of the 21.91 05.36
geoawtric gean
Nunber of aanple sites 35 13
a Level of significance for teat of difference between neana.
A
V A
A A
A
A
A A
1 1 1 1 1 1 1
Tocil
[J9..6
-1.99
SO. IS
23.62
U
A
|
10 15 20 25 30 35 40 45
Bulk Sample Chrysotile Content (%)
50
55
60
Figure 6. Airborne chrysotile concentration (ng/m3) versus average
bulk chrysotile percentage at 48 asbestos-containing material sites.
76
-------
~
1
O)
c
-2f
I
X
-C
(J
u
I
<
700
600
500
•J\J\J
400
300
O AA
200
100
0
A
A
A
S
A
_
A
C
D
B
0
0
c
Perlite
LEGEND: A = 1 Obs, B = 2 Obs, etc.
Spearman Correlation Coefficient = -0.32
P-volue = 0.03
Otlltr nalot coWDMnt
Population TiTiMTat Porlit* Vcmiculita Gla«» wool
1«*« (p < 01)b 215.51 132.53 11.29
Ius4ud *mt at U* 63.47 31.42 7.19
«u
A SMMUic MU^ 116.92 56.9? 3.9S
(» < .01)
StMdinl >rcn »i th« 37.7] 30.15 3.66
A
tak«r of >«fU ilt«« 2> 16 4
.. t Tkla cAtagerr iAclud«> tve tltn that **cb bad bulk aupUi of
A b L«v«i of ii|BiflcABC« for t««t of diff«r«ac« b«tw»«a MAIU.
B
A A
A
B *
1 1 _
Vermiculite Gloss Wool
Total
179.46
41.09
80.45
23.62
41
different
Bulk Sample Other Major Component
Figure 1- Airborne chrysotile concentration (ng/m3) by other major
components of bulk samples at 48 asbestos-containing material sites
77
-------
Table 25 describes the percentage of chrysotile in asbestos-containing
material by other major component. For sites with perlite and vermiculite,
the average percentages are similar; however, the average percentage of as-
bestos for glass wool sites is much higher. This can be seen in Figure 8,
which is a plot of air level of chrysotile versus chrysotile percentage that
identifies sites according to other major components. The four glass wool
sites are the sites with the highest asbestos percentage, but they have rela-
tively low air levels of chrysotile. The friability rating given these glass
wool sites was high; three sites had a rating of 3, the other a rating of 2.
They had a low releasability rating, and air levels were low. Caution should
be exercised in reaching conclusions about glass wool since this information
is based on observations at only four sites.
Table 25. Bulk Chrysotile Content at the Asbestos-Containing
Material Sites by Other Major Components
Other major component (%)
o
Population estimates Perlite Vermiculite Glass wool Total (%)
Mean
Standard error
Median
Sample size
14
2
13
28 sites
15
2
13
16 sites
50
1
51
4 sites
16
2
13
48 sites
a This category includes two sites that each had bulk samples of different
compositions (a perlite/vermiculite composite).
G. Releasability
The detailed microscopic examination of major nonasbestos components of
the bulk samples led to the development of a new factor, referred to as re-
leasability. This factor indicates how readily the bulk material might re-
lease the asbestos fibers into the air. During microscopic examination, the
material was assigned a releasability rating between 1 (low) and 9 (high).
These ratings were based upon the material's asbestos content, fiber size,
brittleness of the matrix, and the apparent freedom of the individual asbestos
fibers (see Section 6 for more detailed discussion of the releasability rat-
ing). The releasability rating for an asbestos-containing material site was
computed by averaging (weighted) releasability ratings of bulk samples from
that site. Releasability ratings were grouped into three categories: 1-4,
low; 5-6, moderate; and 7-9, high.
78
-------
700
600
«E 500
o>
c
7 400
5
IM
X
u 300
200
LEGEND: Symbol is value of other
major component.
P = perlite
V = vermiculite
G = glass wool
100
p
pv p
p
iVP
p p
P PV
p
p p
VP
V
p
v
w p
w
©
©
I®
10 15 20 25 30 35 40
Bulk Sample Chrysotile Content (%}
45
50
55
60
65
Figure 8. Airborne chrysotile concentration (ng/m3) by average bulk
chrysotile percentage and other major component at 48 asbestos-containing
friable material sites.
79
-------
Figure 9 displays the air levels of chrysotile by releasability category,
and the table in Figure 9 summarizes these data. Both the mean and geometric
mean air levels of chrysotile are significantly different among releasability
categories. As releasability increases (low to moderate to high), average
airborne chrysotile also increases.
Table 26 shows the distribution of asbestos-containing material sites
with respect to releasability, as observed by microscopic examination of bulk
samples, and friability, as observed by inspection of the sites. A tendency
towards an inverse relationship between releasability and friability can be
seen in Table 26. All but one of the high friability sites are low in releasa-
bility, and the majority of the low friability sites are high in releasability.
A new algorithm score was computed by substituting releasability for fri-
ability in the original score computations. Figure 10 is a plot of the air
level of chrysotile versus this new score for the 48 sample sites. The corre-
lation coefficient is .45, positive, and significantly different from zero.
(Recall that the correlation coefficient using the original score is not sig-
nificantly different from zero, -0.17.)
H. Other Covariables
Table 27 lists other variables about which information was collected at
the asbestos-containing material sites. Room height and room volume are in-
cluded in the regression analyses which follow. The other factors listed in
Table 27 are not included in the analyses because little variability was ex-
hibited by these factors across the 48 sites in the sample. When there is
little variability, a meaningful assessment of the effect of different levels
of the factor on airborne asbestos is not possible.
Information was collected on cleaning practices employed at the sites in
the sample. Cleaning practices were divided into four categories—wet mopping
only, vacuuming only, dry mopping with an oil-saturated mop, and sweeping.
There is not much variability in cleaning practices; 38 of the 48 sites were
dry mopped with oil. Table 28 describes the air levels of chrysotile by
cleaning category. The air level means and geometric means are significantly
different among cleaning categories. The average air level of chrysotile is
higher for sites that were swept. Again, caution should be exercised in reach-
ing conclusions since this information is based on only four sites that were
swept. Cleaning differs from other factors considered, such as releasability
or condition of the friable material, because it can be changed more easily.
If sweeping does indeed contribute to an elevated level of airborne asbestos,
then changing cleaning procedures seems to be a reasonable approach to reduce
potential asbestos exposure rather than just using the cleaning factor in
assessing the extent of potential exposure.
80
-------
Population »ciutM
dun (p < .01)*
Standard error of tha
•aaa
GtoaMtric Man
(p < .01)'
Standard error of the
l»OMtric Man
Hi»b«r of ia«pU litaa
ton
93.96
24.92
31.79
18.03
19
Ralaaiabilit?
Moderate
123.17
32.29
74.19
25.56
17
High
380.35
82.11
280.46
84.44
12
Toul
179. 4«
41.99
30.45
23.62
•3
for t«»t of difftc«nc« b«tva«n sauna.
700
600
<"£ 500
^
I
ju 400
0
w
u 300
t>
c
3
3 200
100
0
LEGEND: A =1 Obs, B = 2 Obs, etc.
Spearman Correlation Coefficient 0.44
P-value <0.01
_
A
—
A
A
—
A
A
A
A
i- B
C A
B B
A B
C C
- G C
I 1
Low Moderate
Releasability
A
A
A
B
A
A
B
B
A
1
High
Figure 9. Airborne chrysotile concentration (ng/m3) versus releasability
categories at 48 asbestos-containing material sites.
81
-------
Table 26. Distribution of the Asbestos-Containing Material Sites
with Respect to Friability and Releasability
(Population Estimates)
Friability
Releasability Low Moderate High Total
Low
Percentage of sites 5 16 12 33
Standard error (%) 46 78
Number of sample sites 2 11 6 19
Moderate
Percentage of sites 5 34 < 1 40
Standard error (%) 5 9 < 1 9
Number of sample sites 2 14 1 17
High
Percentage of sites 16 11 0 27
Standard error (%) 77 0 10
Number of sample sites 84 0 12
Total
Percentage of sites 26 62 12 100
Standard error (%) 10 12 70
Number of sample sites 12 29 7 48
82
-------
700
600
500
= 400
^300
c
< 200
100
Legend: A - 1 obs. , B = 2obs., etc.
Pearson Correlation Coefficient = 0.45
P-value <0.01
A B
A A A A A
A A A A
A A A B
A A B A A
ABA A BAA A
i I I T I | T I | I
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72 76
New Algorithm Score (Releasability Replacing Friability)
Figure 10. Plot of airborne chrysotile concentration (ng/m3) versus new score
(releasability substituted for friability) for asbestos-containing material
friable sites.
83
-------
Table 27- Candidate Covariables
Missing data 0<_j ^_.c.c .^
Variable
Temperature (N = 36)
Humidity (N = 36)
Sampling height (N = 44)
Room height (N = 48)
Room volume (N = 48)
Na
12
12
4
0
0
%
(25.0)
(25.0)
( 8.3)
NA*
NA
Mean
24.8°C
64.2%
1.4 m
3.5 m
792 m3
deviation
1.6°C
7.4%
0.4 m
1.47 m
1,271 m3
variation
0.06
0.12
0.29
0.42
1.60
Air conditioning status 4 (8.3)
(N = 44)
Friable material location 0 NA
(N = 48)
mean
90% operating
98% on ceilings
Floor carpeting (N = 48)
Tile floor (N = 48)
Wood floor (N = 48)
Suspended ceilings (N = 48)
a Number of sample sites.
V* Pr*£»-F-F-ir*-j£ant~ nf •X7a*"Ta1"-ir*r»
0
0
0
0
std.
NA
NA
NA
NA
deviation
94% uncarpeted
92% had tile
100% did not
100% did not
floor
have
have
* NA = not applicable.
84
-------
Table 28. Airborne Chrysotile Concentration (ng/m3) for Asbestos-
containing Friable Material Sites by Cleaning Category
Population estimates
Cleaning category'
Total
Mean (p < .01)
Standard error of the
mean
Geometric mean,
(p = .01)D
Standard error of the
geometric mean
Number of sample sites
19.28 39.83 159.96 368.41 179.46
4.36 13.52 31.71 163.64 41.99
16.64 32.78
6.00 10.81
76.48 182.09 80.45
23.24 151.97 23.62
38
48
a Cleaning category: 1 = wet mopped only; 2 = vacuumed only; 3 = dry mopped
with oil-saturated mop; 4 = swept (includes two sites that were swept and
also dry mopped with oil-saturated mop).
b Level of significance for test of difference between means.
I. Long-Term Variability
To examine the variability of air levels in chrysotile over time, air
sampling was conducted at 3 of the 48 asbestos-containing material sites for
three consecutive weeks, instead of just 1 week. Air sampling was also con-
ducted for 3 weeks at the control and ambient sites corresponding to these
asbestos-containing material sites. Substantial variability over time was
observed. Caution should be employed in interpreting the data because they
are based upon a small number of sites. The long-term sampling data are
listed in Table 29.
For the long-term site selected from Group 1 (large student activity
areas), the following air levels of chrysotile are observed: week 1, 10.00
ng/m3; week 2, 0.76 ng/m3; and week 3, 1.50 ng/m3. The average over 3 weeks
at this site is 4.09 ng/m3, and the standard deviation is 5.13 ng/m3, result-
ing in a coefficient of variation equal to 1.25. This variability over time
at one site can be compared to the variability in air levels of chrysotile
among the different sites in Group 1; the coefficients of variation, 1.25 and
1.17, are very similar. The coefficient of variation over all asbestos-
containing material sites is 1.06. Comparisons like this can be made using
the other two long-term sites as well as the long-term controls and ambients.
It follows that variability over time at the same site appears to be com-
parable to variability among different sites within the various groups. This
85
-------
Table 29. Variability of Airborne Chrysotile Concentration (ng/m3) over Time at Three Asbestos-containing
Friable Material Sites and the Corresponding Control and Ambient Sites
OO
School 10
Swimming Pool Room
(Group la)
School 3
Classroom 204
(Group 2b)
Site Control Ambient Site Control Ambient
Airborne Chrysotile (ng/m3)
Week 1
Week 2
Week 3
Mean
Standard deviation
Coefficient of variation
10.00 72.90 0.55 105
0.76 50.30 10.80 27
1.50 73.40 1.10 6
A. 09 65.53 4.15 46
5.13 13.19 5.77 52
1.25 0.20 1.39 1
.00 166.00 2.70
.50 14.60 17.60
.17 45.50 0.00
.22 75.37 6.77
.01 80.00 9.48
.13 1.06 1.40
School 2
Classroom 221
(Group 3C)
Site Control
32.30 3.00
61.20 2.60
6.80 0.00
33.43 1.87
27.22 1.63
0.81 0.87
Ambient
4.10
1.30
8.88
4.76
3.83
0.80
Variability of Airborne Chrysotile Concentration (ng/m3)
for
Population estimates
Mean
Standard error of the mean
Standard deviation ,
Coefficient of variation
Median
Number of sample sites
a Group 1: large student
b Group 2: classrooms and
the Asbestos-containing Friable
Group
I3 2b
14.80 209.32
8.16 45.11
17.31 130.22
1.17 0.62
8.58 196.00
9 10
activity areas.
corridors with high friability,
Material Sites by Group
3C
186.78
53.80
204.29
1.09
92.70
29
or a combination of high
Total
179.46
41.99
190.86
1.06
92.70
48
asbestos content
and
moderate friability (according to the school district rating).
c Group 3: classrooms and corridors with low friability, a combination of low asbestos content and
moderate friability, or a combination of unknown asbestos content and moderate friability
(according to the school district rating).
d Coefficient of variation = standard deviation/mean.
-------
result is limited by the small number of sites on which it is based as well
as by the time period of only 3 weeks.
III. REGRESSION ANALYSES
The previous analyses focused upon the relationship between airborne
chrysotile and a number of possible predictors of airborne levels, but only
two variables were considered in each separate analysis — the dependent varia-
ble (airborne chrysotile) and one predictor (e.g., material condition, releas-
ability, or the algorithm score). The analyses described in this subsection,
on the other hand, focus on the relationship between airborne chrysotile and
a set of possible predictors or independent variables. The reason for these
multivariable analyses is the recognition that the candidate predictors are
theoretically and empirically related to one another, and that predictions
can generally be improved if multiple variables are considered simultaneously.
The continuous nature of airborne chrysotile measurements suggested that
multiple linear regression could be used to construct a series of models that
might be reasonably predictive of the chrysotile level in sites with specified
characteristics; thus, standard multiple regression procedures were generally
employed. However, two preliminary decisions had to be made to define the
boundaries of the analytical problem. Specifically, the pool of candidate
predictors had to be defined, and transformations of the dependent variable
had to be considered.
The candidate predictors placed in the independent variable pool included
material condition, accessibility, air stream status, water damage, activity,
friability, average bulk sample chrysotile percentage, releasability, room
volume, room height, and typical cleaning practices. The "exposed surface"
variable was excluded from the set of candidate predictors because all 48
asbestos-containing material sites were 100% exposed. Releasability and the
cleaning practices variable were included because previous bivariate analyses
suggested that these factors are somewhat predictive of airborne chrysotile
levels (see Figure 9 and Table 28). Room volume and height were included be-
cause of special interest in airborne chrysotile levels in rooms with large
volumes and/or high ceilings. Specifically, there was the issue of the impact
of room volume or ceiling height upon the relationship between the algorithm
factors and airborne chrysotile levels. The average bulk sample (chrysotile)
percentage variable was included instead of the coded (grouped) bulk sample
variable because there may be some loss of useful information when the esti-
mated percentage of chrysotile in the friable material is replaced by a code
representing a crude range of percentages. Variables such as temperature,
humidity, air sampling height, and type of flooring were not included in the
independent variable pool because they had little variation across the 48
asbestos-containing friable material sites (see Table 27).
Transformations of the dependent variable had to be considered because
the data suggested that the airborne chrysotile distribution in the study area
was positively skewed. A transformation of the general form log (airborne
chrysotile level + k), where k is a constant, was an obvious choice because
of the positive skew, but the choice for k was more arbitrary. A value of
87
-------
1.0 ng/m3 for k was assumed because the available data suggest that the back-
ground level of airborne chrysotile in the United States is less than 10 ng/m3
(Nicholson et al. 1978).
The next issue after specifying the independent variable pool and the
dependent variable transformation was selecting a procedure for identifying a
simple predictive regression model. A number of stepwise model building pro-
cedures are available, but many focus exclusively upon the "statistical sig-
nificance" of predictor contributions to the regression model. Thus, statis-
tical considerations may produce a model which does not fit well with substan-
tive or theoretical considerations. To minimize this possibility, a two-stage
model selection procedure was used. During the first stage, R2 statistics
for all possible models of a given size were examined. (The R2 statistic
ranges between 0.00 and 1.00 and indicates the proportion of the observed air-
borne chrysotile variation which is explained by the independent variables.)
In this way, a number of better predicting models were identified. During
the second stage, the algebraic signs and the p-values for model coefficients
were examined. Emphasis focused upon identifying predictive models with the
expected coefficient sign (positive and negative) and small p-values. Thus,
the identification of a "best" model was based upon statistical and substan-
tive considerations.
Table 30 displays a summary of unweighted regression models. The un-
weighted models are based upon relationships present in the data from the 48
selected sites. Weighted models should be used when the analytical focus is
shifted from the 48 selected sites to the target population of sites in the
study area. Inspection of the R2 statistics associated with the one-variable
models suggests that releasability is the best single predictor of airborne
chrysotile. Specifically, releasability accounts for approximately 22% of
the airborne chrysotile variation. Inspection of the two-variable models and
the correponding R2 statistics reveals that the addition of the cleaning vari-
able (or the room volume variable) explains an additional 4% of the airborne
chrysotile variation. Inspection of the three-variable models suggests that
the addition of water damage to the model with releasability and the cleaning
variable accounts for another 5% of the airborne chrysotile variation. How-
ever, the addition of a fourth variable to the model increases the R2 statis-
tic by less than 5%. This leveling off of the R2 statistic indicates that
most of the statistically useful information contained in the independent vari-
able pool has been accounted for by the variables releasability, cleaning,
and water damage (or room volume). These variables account for only one-third
of the observed airborne chrysotile variation.
Consideration of other models underlines the importance of releasability
from a predictive perspective. Specifically, a model based upon seven of the
original algorithm variables (exposed surfaces excluded) yields an R2 statis-
tic (R2 = .19) less than that yielded by releasability alone (R2 = .22).
Moreover, the substitution of releasability for friability increases the R2
associated with the algorithm-based model from .19 to .31. This latter model
is no less predictive than the optimal three-variable model identified above
(R2 = .31).
88
-------
Table 30. Summary of Unweighted Regression Models: Relationship of
Surrogate Measurements to Airborne Chrysotile Concentration
Dependent variable = Log (ng/m3 airborne chrysotile + 1.0)
Independent variable R2
(1) One-variable models
Activity < 0.01
Condition < 0.01
Air stream status < 0.01
Accessibility 0.03
Water damage 0.04
Friability 0.06
Chrysotile content (%) 0.06
Cleaning 0.08
Room height 0.11
Room volume 0.13
Releasability 0.22
(2) Two-variable models with highest R2s
Releasability, water damage 0.24
Releasability, cleaning 0.26
Releasability, chrysotile content (%) 0.26
Releasability, room volume 0.27
(3) Three-variable models with highest R2s
Releasability, water damage, chrysotile content (%) 0.29
Releasability, cleaning, chrysotile content (%) 0.29
Releasability, room volume, water damage 0.30
Releasability, cleaning, water damage 0.31
(4) Four-variable models with highest R2s
Releasability, cleaning, water damage, air stream status 0.32
Releasability, cleaning, water damage, room height 0.33
Releasability, cleaning, water damage, room volume 0.33
Releasability, cleaning, water damage, chrysotile content (%) 0.35
(5) Original algorithm factors (% chrysotile content)
R2 = 0.19
(6) Original algorithm factors (% chrysotile content), with releasability
replacing friability
R2 = 0.31
Proportion of airborne chrysotile variation explained by the independent
variables.
89
-------
Table 31 displays the coefficient signs and p-values associated with six
unweighted regression models. Model I is based upon the original algorithm
factors and was mentioned above. Model II is similar except that releasabil-
ity was substituted for friability. Model III is the optimal three-variable
model, while Models IV and V are two of the more predictive four-variable
models. Model VI is a variant of Model V.
An important aspect of Models I and II in Table 31 is that four of the
seven predictors in each model have negative coefficients, and only one of
the coefficients in Model II is significantly different from zero. Present
theory (and speculation) concerning the determinants of airborne chrysotile
predicts that each of these factors should be positively related to the de-
pendent variable. Thus, Model III is much more satisfactory because all the
predictors have positive coefficients and all of the coefficients are signifi-
cantly different from zero (p = .10 level).
Table 32 displays the coefficients and p-values associated with six
weighted regression models. The weighted models take into account the proba-
bility sample design used in site selection and are applicable to the target
population of sites in the study area. Because the 48 sample sites were not
selected with equal probabilities, sampling weights were used to adjust in-
formation to facilitate valid statistical inferences from the 48 sites to the
study area. (Recall that the study area consists of all school district
student activity areas with material known to or suspected of containing
asbestos.) The models presented in Table 32 are similar to the models pre-
sented in Table 31 except that the estimate of each model coefficient and its
standard error have been slightly altered by the weighted calculations.
Model-specific comparisons between Tables 31 and 32 reveal that in each case
the weighted regression model has a slightly larger R2 statistic. This phe-
nomenon highlights the fact that the models fit better in the actual study
area than in a hypothetical study area with a distribution of sites directly
proportional to the distribution of sample sites.
The smaller R2 statistic associated with Model I in Table 32 argues
against the use of this set of predictors based upon the original algorithm
variables. Thus, attention should focus upon one of the remaining models
which include releasability. Model II should also be rejected because the
majority of the coefficient estimates are not significantly different from
zero and several have negative signs. Model III is a strong candidate for
the most reasonable model because all three coefficients are positive. The
remaining models in Table 32 have slightly larger R2 statistics, but each has
a negative coefficient with a large p-value.
90
-------
Table 31. Unweighted Regression Models: Relationship of Surrogate
Measurements to Airborne Chrysotile Concentration
Dependent variable = Log (ng/m3 airborne chrysotile + 1.0)
Probability
Independent variable Sign of coefficient value
Model I. Original algorithm factors (R2 = 0.193)
Condition - 0.37
Accessibility - 0.46
Air stream status + 0.75
Water damage + 0.12
Activity + 0.53
Friability - 0.27
Chrysotile content (%) - 0.12
Model II. Original algorithm factors, with releasability replacing friability
(R* = 0.3lT
Condition - 0.76
Accessibility - 0.43
Air stream status - 0.99
Water damage + 0.20
Activity + 0.73
Releasability + < 0.01
Chrysotile content (%) - 0.12
Model III. Releasability, cleaning, and water damage (R2 = 0.31)
Releasability + < 0.01
Cleaning + 0.05
Water damage + 0.08
Model IV. Releasability, cleaning, water damage, and air stream status
(R* = 0.32T
Releasability + < 0.01
Cleaning + 0.04
Water damage + 0.06
Air stream status - 0.43
Model V. Releasability, cleaning, water damage, and chrysotile content (%)
(R* = 0.35)
Releasability + < 0.01
Cleaning + 0.06
Water damage + 0.06
Chrysotile content (%) - 0.10
Model VI. Releasability, cleaning, water damage, and chrysotile content
category (Category 1 is < 20%. Category 2 is ^20%) (R* =0.31)
Releasability + < 0.01
Cleaning + 0.05
Water damage + 0.08
Chrysotile content category - 0.71
a Proportion of airborne chrysotile variation explained by the independent
variables.
91
-------
Table 32. Weighted Survey Regression: Relationship of Surrogate
Measurements to Airborne Chrysotile Concentration
Dependent variable = Log (ng/m3 airborne chrysotile +1.0)
Probability
Independent variable Sign of coefficient value
Model I. Original algorithm factors (R2 = 0.22 )
Condition - < 0.01
Accessibility - 0.88
Air stream status + 0.94
Water damage + 0.10
Activity + 0.52
Friability - 0.85
Chrysotile content (%) - 0.22
Model II. Original algorithm factors, with releasability replacing friability
~~^
Condition - 0.07
Accessibility - 0.79
Air stream status + 0.97
Water damage + 0.24
Activity + 0.69
Releasability + 0.01
Chrysotile content (%) - 0.15
Model III. Releasability, cleaning, and water damage (R2 = 0.32)
Releasability + < 0.01
Cleaning + 0.13
Water damage + 0.28
Model IV. Releasability, cleaning, water damage, and air stream status
(Rz = 0.33)
Releasability + < 0.01
Cleaning + 0.12
Water damage + 0.32
Air stream status - 0.67
Model V. Releasability, cleaning, water damage, and chrysotile content (%)
(R* = 0.367
Releasability + < 0.01
Cleaning + 0.21
Water damage + 0.31
Chrysotile content (%) - 0.15
Model VI. Releasability, cleaning, water damage, and chrysotile content
category (Category 1 is < 20%, Category 2 is £ 20%) (R* =0.33)
Releasability + < 0.01
Cleaning + 0.16
Water damage + 0.30
Chrysotile content category - 0.67
a Proportion of airborne chrysotile variation explained by the independent
variables .
92
-------
Figures 11 through 13 can be used to obtain a more intuitive appreciation
of the predictive character of regression models with R2 statistics less than
.500. Figure 11 displays a plot of the observed airborne chrysotile values
(on a log scale) versus the predicted values obtained from an unweighted re-
gression model based upon releasability and six of the original algorithm fac-
tors (see Model II, Table 31). (Figure 5 displays a plot of airborne chryso-
tile values versus algorithm score values.) If the model were perfectly pre-
dictive, the R2 statistic would be 1.00 and all of the points in Figure 11
would fall along a straight line. The scatter of the points about this
straight line represents airborne chrysotile variation which is not statis-
tically explained by the seven independent variables. Figure 12 is identical
to Figure 11, except that the observed and the predicted values have not been
transformed. Figure 13 displays a similar plot of (logged) observed versus
predicted airborne chrysotile values except that the regression model (see
Model III, Table 31) contains only three independent variables (releasability,
cleaning, and water damage).
Regardless of which of the better predicting models is selected as most
reasonable, it is clear from the R2 statistics in Table 32 and the plots in
Figures 11 and 13 that only moderately accurate predictors of airborne chryso-
tile have been identified. A highly satisfactory set of predictors should
have an R2 statistic of at least .80, and the model coefficients should have
algebraic signs consistent with physical theory.
In summary, the data suggested that the four-variable models were not
meaningfully superior to the three-variable model based upon releasability,
cleaning, and water damage. However, the inclusion of a "cleaning" variable
in an exposure assessment algorithm may create additional measurement problems
because school personnel have to be queried to obtain the relevant information.
There was some interest, therefore, in optimal models based upon an indepen-
dent variable pool which did not contain the cleaning variable. Table 33 pre-
sents a series of optimal unweighted regression models based upon a variable
pool without cleaning. The best two-, three-, and four-variable models are
displayed.
Inspection of Table 33 reveals that the optimal three-variable model from
this pool includes releasability, room volume, and water damage. Thus, the
exclusion of the cleaning variable allows room volume into the three-variable
model, while the inclusion of a fourth variable does not appreciably increase
the R2 statistic. Table 34 presents additional characteristics of the optimal
three-variable model obtained when cleaning is excluded. Specifically, it is
of interest that releasability and water damage have positive coefficients,
while the room volume coefficient is negative. After adjusting for the ef-
fects of releasability and water damage, larger sites (rooms) appear to have
lower airborne chrysotile concentrations. A slightly negative room volume/
airborne chrysotile relationship was first observed in the bivariate analyses.
Overall, two optimal three-variable models were identified and judged to
be reasonable taking statistical and substantive issues into consideration.
The first model is based upon releasability, water damage, and cleaning, and
the second is based upon releasability, water damage, and room volume.
93
-------
3. 6
u
0
JL
A A A
A A
A A
A _A_
_L
Legend: A = 1 obs., B = 2obs., etc.
R2 = 0.31
A
A A
A A
A
A A
A
A A
AA
J_
1.0 1.4 1.8 2.2 2.6 3.0 3.4 3.8 4.2 4.6 5.0
Model II Predicted Value of Log (ng/m3 Airborne Chrysotile + 1.0)
5.4 5.8
Figure 11. Log (ng/m^ airborne chrysotile + 1.0) versus the predicted value from
Model II (original algorithm factors, with releasability replacing friability)
for asbestos-containing friable material sites.
-------
VO
Ul
700;
600
*£ 500
>
:= 400
^ 300
j
< 200
100
A A A
A A
A
A A
ABA A
A
A
A
A
A
A A
AA A
A A
A
A
A A
A
A
Legend: A = 1 obs., B = 2obs., etc.
R2 = 0.31
20 40 60 80 100 120 140 160 180 200 220
Model II Predicted Value of Airborne Chrysotile (ng/m3)
240 260 280 300
Figure 12. Airborne chrysotile concentration (ng/m^) versus the predicted value from
Model II (original algorithm factors, with releasability replacing friability) for
asbestos-containing friable material sites.
-------
c?
*-~
+
0)
7~
s
J-
u
c
4
**•
n
i
c
I
8
7
6
5
4
3
2
1
0
-
-
A
A
A
A
_
A
A
A
A
A
A A
-
A
-1 1 1 1 1 C
1.0 1.4 1.8 2.2 2.6
Legend: A = 1 obs. , B = 2obs. , etc.
R2 =0.31
A A
A
A A A
A A
A
A A
A A A
B A A A
A A
A
A A
A A
A
A
B A
A A
1 1 1 1 1 1 1 1
3.0 3.4 3.8 4.2 4.6 5.0 5.4 5.8
Model Ml Predicted Value of Log (ng/m3 Airborne Chrysotile + 1.0)
Figure 13. Log (ng/nP airborne chrysotile + 1.0) versus the predicted value from
Model III (releasability, cleaning, and water damage) for asbestos-containing
friable material sites.
-------
Table 33. Summary of Unweighted Regression Models: Relationship of
Surrogate Measurements (Excluding Cleaning Practices) to
Airborne Chrysotile Concentration
Dependent variable = Log (ng/m3 airborne chrysotile + 1.0)
Independent variable
R2
(1) Two-variable models with highest R2s
Releasability, room height 0.24
Releasability, water damage 0.24
Releasability, chrysotile content (%) 0.26
Releasability, room volume 0.27
(3) Three-variable models with highest R2s
Releasability, room volume, chrysotile content (%) 0.28
Releasability, room volume, activity 0.29
Releasability, water damage, chrysotile content (%) 0.29
Releasability, room volume, water damage 0.30
(4) Four-variable models with highest R2s
Releasability, room volume, water damage, friability 0.31
Releasability, room volume, water damage, accessibility 0.31
Releasability, room volume, water damage, activity 0.31
Releasability, room volume, water damage, chrysotile content (%) 0.32
Proportion of airborne chrysotile variation explained by the independent
variables.
97
-------
Table 34. Three-Variable Regression Models: Relationship of Surrogate
Measurements (Excluding Cleaning Practices) to Airborne
Chrysotile Concentration
Dependent variable = Log (ng/m3 airborne chrysotile + 1.0)
Probability
Independent variable Sign of coefficient value
Optimal unweighted three-variable model (R2 = 0.30 )
Releasability + < 0.01
Room volume - 0.06
Water damage + 0.16
Corresponding weighted three-variable model (R2 = 0.35)
Releasability + 0.02
Room volume - 0.03
Water damage + 0.46
a Proportion of airborne chrysotile variation explained by the independent
variables.
98
-------
The first model is only slightly more predictive of airborne chrysotile
levels, but the second may be more useful in abatement programs if accurate
data on cleaning practices are not easily obtainable. The reliability and
generalizability of releasability measurements should be documented in addi-
tional studies, and other predictors of airborne chrysotile levels should be
identified. Both of the optimal three-variable models explain less than half
of the observed airborne chrysotile variation.
IV. ALTERNATIVE CLASSIFICATION ANALYSES
The true airborne chrysotile level at a site with specified characteris-
tics can be estimated if a satisfactory regression model has been developed
and the pertinent independent variables heve been accurately measured. How-
ever, an estimate of the true airborne chrysotile level at a specified site
may be of less interest than a prediction of whether the airborne chrysotile
level at a specified site is higher or lower than some specified reference
level. That is, when abatement decisions are made, a specific abatement pro-
cedure is either followed, or it is not, regardless of the exact airborne as-
bestos level at the site. In essence, the prediction or identification of
sites with high or low airborne chrysotile levels is a classification problem,
and several types of classification analyses were completed. Different def-
initions of high or low airborne chrysotile levels were used in the classifi-
cation analyses.
Table 35 summarizes the results from a preliminary two-group discriminant
analysis. The dependent variable was an airborne chrysotile dichotomy created
by using the unweighted median (68.6 ng/m3) as a cut point. The independent
variables were the seven original algorithm factors (exposed surfaces ex-
cluded) . According to Table 35 the linear discriminant function correctly
classified 60% of the observations. However, the performance of all two-group
discriminant models (and the other classification analyses) should be compared
with performance of an unbiased "coin" which could correctly classify 50% of
the available observations.
Table 36 presents the results from a series of discriminant analyses
similar to the one displayed in Table 35. Comparisons between the analyses
in Table 36 emphasize the effect of different airborne chrysotile cut points
upon two statistics of interest—the percentage of all observations correctly
classified, and the percentage of observations above the cut point correctly
classified. The latter statistic indicates the sensitivity of the classifica-
tion. Table 36 demonstrates that the sensitivity of the classification is
decreased when the cut point is increased. Table 36 also suggests that total
number of classification errors can be reduced by selecting a cut point higher
or lower than 68.6 ng/m3, but the data do not suggest whether 11.1 ng/m3 (25th
percentile) or 149.5 ng/m3 (75th percentile) is preferable. Of course, sub-
stantive issues such as the ambient level of airborne chrysotile could be ad-
dressed when the optimal cut point is designated.
Comparisons between Table 36 analyses also emphasize the effect of dif-
ferent models upon the statistics of interest. Specifically, the substitution
of releasability ratings for friability ratings decreases the impact of a
99
-------
Table 35. Predicting Low/High Airborne Chrysotile Concentration with
a Discriminant Function3 Based upon Seven Original Algorithm Factors
Distribution of the 48 Asbestos-containing Friable Material Sites:
Observed Airborne Chrysotile Concentration Versus Predicted
Observed airborne Predicted by discriminant function
chrysotile Low High Total
concentration (ng/m3) (^ 68.6) (> 68.6)
Low (
High
Total
^ 68.6°)
(> 68.6)
16
11.
27
8
13
21
24
24
48
% Correct = 29/48 = 60.4%
Sensitivity = 13/24 = 54.2%
a Discriminant analysis with pooled covariance matrix and prior probabilities
proportional to sample sizes.
b The factor exposure was not included because of zero variability. The
original algorithm categorization of bulk sample chrysotile content was
used.
c 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable material
sites.
100
-------
Table 36. Summary of Discriminant Analyses
Discriminant model characteristics
Airborne chrysotile
concentration Independent variables
cut point
Statistics
% Correct
Sensitivity (%)
11.1 ng/m3 (25th %)a 7 algorithm factorsb
68.6 ng/m3 (50th %) 7 algorithm factors
149.5 ng/m3 (75th %) 7 algorithm factors
79.2
60.4
79.2
100.0
54.2
25.0
11.1 ng/m3 (25th %) Releasability rank
68.6 ng/m3 (50th %)
149.5 ng/m3 (75th %)
and 6 algorithm factors
Releasability rank
and 6 algorithm factors
Releasability rank
and 6 algorithm factors
79.2
64.6
81.2
100.0
62.5
41.7
a Percentile of the observed airborne chrysotile concentration distribution
for the 48 asbestos-containing friable material sites.
b The factor exposure was not included because of zero variability. The
original algorithm categorization of bulk sample chrysotile content was
used.
c The factor friability was replaced by the releasability rank (possible
values 1-9) determined from bulk sample inspection.
101
-------
changing cut point upon the sensitivity of the classification. When the
original seven algorithm factors are included in the discriminant model, the
sensitivity of the classification decreases to 54.27» and 25% when the two
higher cut points are used. When releasability is substituted for friability,
the sensitivity of the classification decreases to 62.5% and 41.7%, respec-
tively. Substitution of the releasability ratings also decreases the total
number of classification errors when the higher cut points are used.
Table 37 summarizes an attempt at predicting high and low airborne chrys-
otile sites with an algorithm score dichotomy. The algorithm score dichotomy
was based upon the median consensus score (28), while the airborne chrysotile
dichotomy was based upon the 50th percentile (68.6 ng/m3) and the 75th percen-
tile (149.5 ng/m3). Inspection of the percentages of observations correctly
classified in Table 37 suggests that all the previous discriminant analyses
outperformed the simple algorithm score dichotomy.
Table 38 summarizes an attempt at predicting high/low airborne chrysotile
sites with a releasability dichotomy. High releasability was defined as a
rating greater than 6, while the airborne chrysotile dichotomy was based upon
either the 50th or the 75th percentile. Inspection of the percentages of ob-
servations correctly classified in Table 38 suggests that the releasability
dichotomy also outperformed the algorithm score dichotomy. Yet two of the
discriminant models in Table 36 performed slightly better.
Tables 39 and 40 (and G-l in Appendix G) summarize various attempts at
predicting high/low airborne chrysotile sites with several regression-based
dichotomies. The regression models used in these analyses correspond to some
of the unweighted models presented in the previous section. The motivation
for these analyses was to demonstrate how regression models can be used to
classify sites with respect to a designated cut point.
Tables 41 through 45 (and G-2 through G-6) summarize various attempts at
predicting high/low airborne chrysotile sites with decision trees based upon
some of the original algorithm factors and releasability. Parallel analyses
using either the 50th (Tables G-2 to G-6) or the 75th (Tables 41 to 45) air-
borne chrysotile percentile to identify high/low airborne chrysotile sites
were completed. Table 41 (and G-2) is concerned with the original decision
tree based upon exposed surfaces, material condition, accessibility, and fri-
ability, while Table 42 (and G-3) is concerned with a tree based upon releas-
ability and water damage. Table 43 (and G-4) examines a tree based upon re-
leasability and airstream status, while Table 44 (and G-5) examines a tree
based upon releasability and the bulk chrysotile percentage. Table 45 (and
G-6) summarizes the performance of a more complex tree based upon releasabil-
ity, water damage (or material condition), and airstream status.
102
-------
Table 37. Predicting Low/High Airborne Chrysotile Concentration with
the Algorithm Score Dichotomy
Distribution of the 48 Asbestos-containing Friable Material Sites:
Observed Airborne Chrysotile Concentration Versus Predicted
Observed airborne
chrysotile
concentration (ng/m3)
Predicted by score dichotomy
Low , High
(Score ^ 28D)
(Score > 28)
% Correct = 21/48 = 43.8%
Sensitivity = 8/24 = 33.3%
% Correct = 23/48 = 47.9%
Sensitivity = 3/12 = 25.0%
Total
Low (^
High (>
Total
68. 6C)
68.6)
13
16
29
11
_8
19
24
24
48
Low (^
High (>
Total
149.5)
149.5)
20
_9
29
16
J3
19
36
12
48
a Consensus score based upon the original eight factors.
b 28 = 50th percentile (population estimate) of the algorithm scores for
asbestos-containing friable material sites.
c 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
d 149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
103
-------
Table 38. Predicting Low/High Airborne Chrysotile Concentration with
a Releasability Dichotomy
Distribution of the 48 Asbestos-containing Friable Material Sites:
Observed Airborne Chrysotile Concentration Versus Predicted
Observed airborne
chrysotile
concentration (ng/m3)
Low (£ 68. 6a)
High (> 68.6)
Total
Predicted by
Low
(Ranks 1-6)
21
15
36
releasability rank
High
(Ranks 7-9)
3
_9
12
Total
24
24
48
% Correct = 30/48 = 62.5%
Sensitivity = 9/24 = 37.5%
Low (5
High (>
Total
149. 5b)
149.5)
30
_6
36
6
_6
12
36
12
48
% Correct = 36/48 = 75.0%
Sensitivity = 6/12 = 50.0%
a 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
b 149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
104
-------
Table 39. Predicting Low/High Airborne Chrysotile Concentration
with Three Regression-bated Dichotomies
Distribution of the 48 Asbestos-containing Friable Material Sites:
Observed Airborne Chrysotile Concentration Versus Predicted
Observed airborne
Chrysotile
Predicted by seven algorithm factors8
(Regression model I, Table 31)
Total
concentration (ng/m3)
Low (S 149. 5b)
High (> 149.5)
Total
Low
(S 149.5)
36
10
46
High
(> 149.5)
0
2
2
36
12
48
% Correct = 38/48 = 79.2%
Sensitivity = 2/12 = 16.7%
Observed airborne
Chrysotile
concentration (ng/m3)
Predicted by six algorithmdfactorsc
and releasability
(Regression model II, Table 31)
Total
Low (S 149.5)
High (> 149.5)
Total
Low
(£ 149.5)
35
9
44
High
(> 149.5)
1
3
4
36.
12
48
% Correct = 38/48 = 79.2%
Sensitivity = 3/12 = 25.0%
Observed airborne
Chrysotile
concentration (ng/m3)
Predicted by releasability , room
volume, and water damage
(Unweighted regression model, Table 34)
Low High
(S 149.5) (> 149.5)
Total
Low (S 149.5)
High (> 149.5)
Total
35
9
44
1
3
4
36
12
48
% Correct = 38/48 = 79.2%
Sensitivity = 3/12 = 25.0%
Of the eight algorithm factors, exposure was not included because of zero
variability. Bulk sample Chrysotile content was included as the actual
percentage instead of the original algorithm categorization.
149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
Of the eight algorithm factors, exposure was not included because of zero
variability, and friability was replaced by releasability category. Bulk
sample Chrysotile content was included as the actual percentage instead of
the original algorithm categorization.
Releasability category.
105
-------
Table 40. Predicting Low/High Airborne Chrysotile Concentration with
Other Regression-based Dichotomies
Percentage of the 48 Asbestos-containing Friable Material
Sites Correctly Classified
T , , . ,n Airborne chrysotile concentration cut point
Independent variables * *V
in the regression model 68.6 ng/m3 149.5 ng/m3
Releasability (nine ranks) 60.4% 77.1%
and six algorithm
factors
Releasability (nine ranks), 60.4% 77.1%
room volume, and
water damage
a 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
b 149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
c Of the eight algorithm factors, exposure was not included because of zero
variability, and friability was replaced by releasability rank.
106
-------
Table 41. Predicting Low/High Airborne Chrysotile Concentration with
the Original Decision Tree Based upon Proportion of Material
Exposed, Material Condition, Accessibility, and Friability
(0,0)
0,1 b
(33,11)
0,1
(0,0)
0,1
Exposure
4 1 (36,12)
Condition
0,2 1 (36,12)
Accessibility
3 i (3,1)
Friability
(°'0) 1 II- 1
(3J) > Hi,h
2,3 ' h'yh
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
chrysotile
concentration (ng/m3)
Low (^ 149. 5C)
High (> 149.5)
Total
Predicted by tree
Low High
33 3
11 1
44 4
Total
36
12
48
% Correct = 34/48 = 70.8%
Sensitivity = 1/12 = 8.3%
a Number of sites with low airborne chrysotile concentration (^ 149.5 ng/m3),
number of sites with high airborne chrysotile concentration (> 149.5 ng/m3)
b Algorithm factor codes.
c 149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
107
-------
Table 42. Predicting Low/High Airborne Chrysotile Concentration with
a Decision Tree Based upon Releasability and Water Damage
(17,2)a
lb
Releasability
(6,6)
3
High
1
(13,4)
(8,2)
0
Water Damage
(5,2)
1,2C
fc Hinli
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
Chrysotile
concentration (ng/m3)
Predicted by tree
Low
High
Total
Low (<
High (>
Total
149. 5d)
149.5)
25
_4
29
11
8
19
36
12
48
% Correct = 33/48 = 68.8%
Sensitivity = 8/12 = 66.7%
a Number of sites with low airborne chrysotile concentration (S 149.5 ng/m3),
number of sites with high airborne chrysotile concentration (> 149.5 ng/m3).
b Releasability codes: 1 = ranks 1-4, 2 = ranks 5-6, 3 = ranks 7-9.
c Water damage codes.
d 149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
108
-------
Table 43. Predicting Low/High Airborne Chrysotile Concentration with
a Decision Tree Based upon Releasability and Air Stream Status
(17,2)a
lb
Releasability
(6,6)
3
w. I_J • L
— +• High
(13,4)
(7,0)
0
Air Stream Status
(6,4)
lc
k. LJT 1
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
Chrysotile
concentration (ng/m3)
Predicted by tree
Low
High
Total
Low (£ I49.5d)
High (> 149.5)
Total
24
_2
26
12
10
22
36
12
48
% Correct = 34/48 = 70.8%
Sensitivity = 10/12 = 83.3%
a Number of sites with low airborne chrysotile concentration (^ 149.5 ng/m3),
number of sites with high airborne chrysotile concentration (> 149.5 ng/m3)
Releasability codes: 1 = ranks 1-4, 2 = ranks 5-6, 3 = ranks 7-9.
Air stream status codes.
149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
109
-------
Table 44. Predicting Low/High Airborne Chrysotile Concentration with
a Decision Tree Based upon Releasability and Bulk Sample
Chrysotile Content
(17,2)a
lb
Releasability
(6,6)
3
1
(13,4)
High
(11,3)
< 20%
Bulk Sample Chrysotile Content
(2,1)
>20%
High
Distribution of the 48 Asbesots-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
chrysotile
concentration (ng/m3)
Predicted by tree
Low
High
% Correct = 35/48 = 72.9%
Sensitivity = 7/12 = 58.3%
Total
Low (£
High (>
Total
149. 5C)
149.5)
28
_5
33
8
_7
15
36
12
48
Number of sites with low airborne chrysotile concentration (^ 149.5 ng/m3),
number of sites with high airborne chrysotile concentration (> 149.5 ng/m3)
Releasability codes: 1 = ranks 1-4, 2 = ranks 5-6, 3 = ranks 7-9.
149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
110
-------
Table 45. Predicting Low/High Airborne Chrysotile Concentration with
a Decision Tree Based upon Releasability, Water Damage (or
Material Condition), and Air Stream Status
High
(17,2)°
1b
Releasability
(6,6)
3
2 v (13,4)
(9,3)
(MC = 0)
W
(A,
(3,0)
Water Damage
or
(Material
Condition )
D=1'2 (41)
4C=2,5) ^4>''
Air Sfronm Sfnfn« U»U ^
0 1
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
Chrysotile
concentration (ng/m3)
Low
Predicted by tree
High
Total
Low (^ 149. 5d)
High (> 149.5)
Total
29
_5
34
7
J_
14
36
12
48
% Correct = 36/48 = 75.0%
Sensitivity = 7/12 = 58.3%
a Number of sites with low airborne chrysotile concentration (^ 149.5 ng/m3),
number of sites with high airborne chrysotile concentration (> 149.5 ng/m3)
b Releasability codes: 1 = ranks 1-4, 2 = ranks 5-6, 3 = ranks 7-9.
c Water damage and (material condition) codes.
d 149.5 ng/m3 = 75th percentile for the 48 asbestos-containing friable
material sites.
Ill
-------
Two points should be considered before the most "successful" classifica-
tion analyses can be identified from the various discriminant, univariate,
regression, and decision tree approaches presented in Tables 35 through 45
and in Appendix G. First, if sites with several hundred nanograms per cubic
meter of airborne chrysotile are of primary concern, then attention could
focus upon those analyses which used the 75th airborne chrysotile percentile
to identify high/low airborne chrysotile sites because all of the observed
"high" sites had airborne levels in excess of 149.5 ng/m3- Second, when clas-
sification tables are inspected, equal attention could be paid to "the overall
percentage of observations correctly classified" and to the "sensitivity" of
the analysis. In the present context, sensitivity can be defined as the per-
centage of observed (true) high airborne chrysotile sites predicted by the
variables in the model or the tree. More specifically, if most sites are ob-
served low, then a classification procedure which predicts all sites to be
low will have a high overall percentage of sites correctly classified. How-
ever, its sensitivity to detecting high level sites will be zero. Table 46
displays a summary of the classification analyses based upon the 75th airborne
chrysotile percentile.
Inspection of Table 46 reveals that all of the analyses based upon the
higher cutpoint (75th percentile) yielded a percentage of total observations
correctly classified in excess of 68.0 except for the univariate score dichot-
omy which yielded a (low) percentage of 47.9. This relatively poor perfor-
mance of the algorithm score is consistent with earlier findings. Table 46
also highlights the large range of sensitivity statistics yielded by the vari-
ous analyses. The very lowest sensitivity (8.3%) is associated with the orig-
inal version of the decision tree based upon exposed surfaces, material condi-
tion, accessibility, and friability. The highest sensitivity (83.3%) is asso-
ciated with the decision tree based upon releasability and airstream status.
In the light of the information summarized in Table 46, the optimal classifi-
cation analysis appears to be a decision tree based upon releasability and
airstream status. Yet more research needs to be done to reduce the number of
false positives yielded by this tree (see Table 43). In addition, it should
be noted that a discriminant model (based upon the 75th percentile cut point,
releasability, and airstream status with equal prior probabilities) can be
designed to yield the same classification results as the optimal tree. More-
over, the sensitivity of this discriminant model can be increased to 92% (if
the prior probabilities of classification are .25 and .75, but the overall
percentage correct drops to 48.0). The advantage of a decision tree versus a
discriminant function is that the former is easily understood by nonstatis-
ticians.
V. COMPARISON OF RATERS
Each of the 48 student areas selected for the study was rated by five
researchers from the project team on each assessment factor—condition, acces-
sibility, part of air moving system, exposure, water damage, activity, and
friability (see Site-Specific Ratings in Section 5). The raters acted inde-
pendently at each student area to provide data that reflect the measurement
precision that is associated with the factors. In addition, after the rating
was completed at each site, the five raters reached a consensus rating for
each factor. The consensus ratings were reported and used in the statistical
112
-------
Table 46. Classification Analysis Summary (Airborne Chrysotile Concentration
75th Percentile Dichotomy)
Analytical
approach
Discriminant
analysis
Univariate
dichotomy
Regression
analysis
Decision
tree
Results
Table
36
36
37
38
39
39
39
40
40
41
42
43
44
45
Independent variables Sensitivity3
(%)
7 algorithm factors
Releasability rank and
6 algorithm factors
Algorithm score (L/H
dichotomy)
Releasability (L/H
dichotomy)
7 algorithm factors
6 algorithm factors and
releasability category
Releasability category, room
volume, and water damage
Releasability rank and
6 algorithm factors
Releasability rank, room
volume, and water damage
Proportion of material
exposed, material condi-
tion, accessibility,
and friability
Releasability category and
water damage
Releasability category and
air stream status
Releasability category and
bulk sample chrysotile
content
Releasibility category, water
damage (or material condi-
tion) , and air stream status
25.0
41.7
25.0
50.0
16.7
25.0
25.0
25.0
16.7
8.3
66.7
83.3
58.3
58.3
% Correctly,
classified
79.2
81.2
47.9
75.0
79.2
79.2
79.2
77.1
77.1
70.8
68.8
70.8
72.9
75.0
a Percentage of the 12 observed "high" sites correctly classified by the
independent variables.
b Percentage of the 48 asbestos-containing friable material sites correctly
classified by the independent variables.
113
-------
analysis of the relationship between factors and measured airborne asbestos
concentration levels.
Any application of the factors or combination of factors as a method of
ordering the severity of asbestos exposure among schools or student areas
within schools will probably involve only one rater. The use of a consensus
in this study is an operational method to reduce the actions of five raters
to that of one rater. The additional rating data give information on the in-
ternal consistency of the factors. Lack of variation among raters is an indi-
cation that the factors are measuring well-defined characteristics. Wide var-
iation among raters suggests that the factors are measuring poorly defined
characteristics. An assessment of consistency was performed by comparing in-
dividual ratings with the consensus rating.
There are a variety of methods to summarize the rater data to reflect
the degree of consistency in the factors. For the purposes of this analysis,
the number and proportion of disagreements between individual ratings and con-
sensus ratings were used. Table 47 displays the basic data. The rows corre-
spond to the 48 asbestos-containing friable material sites within schools.
Column 1 is an identifying code. Columns 2 through 8 correspond to the expo-
sure assessment algorithm factors. The values appearing in each column are
counts of the number of rater disagreements with the consensus rating. For
example, in row 1, column 4, the 1 indicates that one individual rating dis-
agreed with the consensus on the factor air movement at student area 5-1. In
row 22 (labeled 55-1), the 3 in column 6 indicates that three individual rat-
ings disagreed with the consensus rating on water damage at student area 55-1.
The last two rows of the table are summaries. The row labeled "count"
contains a row count of the number of student areas from among the 48 where
there was any type of disagreement; the row labeled "total" contains the total
number of disagreements tabulated over all student areas. Columns 9 and 10
are similar summaries taken across factors for each student area. For example,
at student area 5-2 there was disagreement on two factors (activity and fria-
bility) with a total of three disagreements (two on activity and one on fria-
bility) .
In total, there are 1,680 opportunities for disagreement (5 raters, 7
factors, 48 student areas). There were 166 disagreements observed. Assuming
that the raters treated each factor at each student area as an independent
rating, an estimate of the rate of disagreement is .099 (9.9%). Using a 95%
confidence interval as an approximation to assess the variability in the esti-
mate indicates that it is unlikely that the disagreement rate exceeds .113
(11.3%).
From the last row in Table 47 there is an indication that the factors
activity and friability may be less reliable than the other five factors.
The percentages of disagreement for these two factors are 17.5% and 14.2%,
respectively. The maximum disagreement percentage taken over the remaining
five factors is 10.4%. An analysis of variance computation to compare dis-
agreement proportions among factors indicated significant differences (see
Table 48). This result suggests that activity and friability are harder to
measure than the other factors and merits further consideration. Indeed, this
result highlights the relative subjectivity of the factor rating procedures
used in the application of the algorithm.
114
-------
Table 47. Rater Consistency:Disagreement with Consensus Frequencies
School
site
1-1
1-2
1-3
2-1
2-2
2-3
2-4
3-1
3-2
3-3
3-4
4-1
5-1
6-1
6-2
6-3
7-1
7-2
7-3
8-1
9-1
10-1
11-1
12-1
12-2
13-1
13-2
13-3
14-1
14-2
14-3
15-1
16-1
17-1
18-1
19-1
20-1
20-2
21-1
21-2
22-1
22-2
23-1
24-1
24-2
25-1
25-2
25-3
COUNT
TOTAL
Cond.
1
2
1
1
1
1
1
1
1
2
2
2
2
2
1
1
1
1
1
19
25
Access.
1
2
1
1
1
1
2
2
2
1
1
1
1
1
1
15
19
Air
1
2
1
1
1
1
1
2
1
2
1
1
1
1
1
1
1
3
1
19
24
Factor
Exp. WDam.
1
2
3
2
1
2
1
1
2
1
1
1
1
1
3
0 14
0 22
Summary
Activ.
2
2
1
2
1
1
1
2
2
1
2
1
1
2
2
1
1
1
3
1
2
2
1
2
1
1
1
1
1
29
42
Fri.
1
1
1
1
1
1
1
2
1
2
1
1
2
2
1
1
1
2
3
1
1
1
1
3
1
1
25
34
Count
1
2
1
1
2
1
5
2
2
1
2
2
4
2
1
1
4
5
3
2
4
5
6
3
2
2
4
2
3
5
6
3
1
2
2
3
4
4
4
0
2
1
3
3
1
0
2
0
121
Total
1
3
2
2
2
2
5
3
2
1
2
2
5
3
1
1
6
6
5
2
6
8
11
4
3
2
5
5
4
8
8
3
2
4
3
4
4
4
4
0
2
1
7
5
1
0
2
0
166
115
-------
Q
Table 48. Analysis of Variance for Disagreement Proportions
Source
Asbestos- containing
material sites
Factors
Error
TOTAL
Sum of
squares
1.813
0.310
4.753
6.876
df
47
5
235
287
Mean
square
0.039
0.062
0.020
F Ratio
1.95
3.10b
a The calculations were performed on ratios obtained as the number of
disagreements divided by the number of raters (five in each case).
The factor "exposure" was excluded since all student areas were rated
as being exposed.
b p = .05.
116
-------
REFERENCES
Cochran WG. 1977. Sampling techniques, 3rd ed. New York: John Wiley and
Sons.
Hansen MH, Hurwitz WN, Madow WG. 1953. Sample survey methods and theory,
vol. I. New York: John Wiley and Sons.
Holt MM. 1979. SURREGR: Standard errors of regression coefficients from
sample survey data. Research Triangle Park, NC: Research Triangle Institute.
Kish L. 1965. Survey sampling. New York: John Wiley and Sons.
McCrone WC. 1980. The asbestos particle atlas. Ann Arbor, MI: Ann Arbor
Science, 122 pp.
McCrone WC, McCrone LB, Delly JG. 1978. Polarized light microscopy.
Ann Arbor, MI: Ann Arbor Science.
NCI. 1978. Natl Cancer Inst. Asbestos: an information resource. Bethesda,
MD: U.S. Dept. Health, Education, and Welfare, Public Health Service,
National Institutes of Health. DHEW Pub. (NIH) 78-1681.
Nicholson WJ. 1978. Mount Sinai Sch. of Medicine, City University of
New York. Control of sprayed asbestos surfaces in school buildings.
Washington, DC: Natl. Inst. of Environmental Health Sciences. Contract
N01-ES-7-2113.
Nicholson WJ, Rohl AN, Sawyer RN, et al. 1979. Mount Sinai Sch. of Medicine,
City University of New York. Control of sprayed asbestos surfaces in school
buildings: a feasibility study. Washington, DC: Natl. Inst. of Environ-
mental Health Sciences. Contract N01-ES-7-2113.
Patton JL, Melton CW, Schmidt EW, et al. 1980. Battelle Columbus Labs. As-
bestos in schools. Draft final report. Washington, DC: Office of Pesticides
and Toxic Substances, U.S. Environmental Protection Agency. Contract 68-01-
3858.
Peto J, Seidman H, Selikoff IJ. 1982. Mesothelioma mortality in asbestos
workers: implication for models of carcinogenesis and risk assessment.
Br. J. Cancer 45:124.
Price B, Melton C, Schmidt E, Townley C. 1980. Battelle Columbus Labs.
Airborne asbestos levels in schools: a design study. Report. Washington,
DC: Office of Pesticides and Toxic Substances, U.S. Environmental Protection
Agency. Contract 68-01-3858.
117
-------
Rajhans GS, Sullivan J. 1981. Asbestos sampling and analysis. In: Identi-
fication and characterization of asbestos. Ann Arbor, MI: Ann Arbor Science.
Selikoff IJ, Hammond EC, eds. 1979. Health hazards of asbestos exposure.
Annals of the New York Academy of Sciences, Vol. 330. New York: New York
Academy of Sciences.
Shah BV. 1979. VMCPNLS: Program to compute variance components. Research
Triangle Park, NC: Research Triangle Institute.
Taylor DH, Bloom JS. 1980. Hexametaphosphate pretreatment of insulation
samples for microscopical identification of fibrous constituents. Microscope
28:47-49.
USEPA. 1975. U.S. Environmental Protection Agency. Asbestos fiber atlas.
Environmental Protection Technology Series. Research Triangle Park, NC:
Office of Research and Development, USEPA. EPA 650/2-75-036. PB-244 766.
USEPA. 1978. U.S. Environmental Protection Agency. Electron microscope
measurement of airborne asbestos concentrations—a provisional methodology
manual. Environmental Protection Technology Series. Research Triangle Park,
NC: Environmental Sciences Research Laboratory, USEPA. EPA 600/1-77-178.
USEPA. 1979a. U.S. Environmental Protection Agency. Asbestos-containing
materials in school buildings: a guidance document, 2 Parts. Washington, DC:
Office of Toxic Substances, USEPA. Report C00090.
USEPA. 1979b. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Asbestos-containing materials in school buildings. Advance notice
of proposed rulemaking. (44 FR 54676).
USEPA. 1980a. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Asbestos analytical programs bulk sample analysis: New York City
and Maryland. Washington DC: USEPA. EPA 560/13-80-21. PB81-14672 2.
USEPA. 1980b. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Asbestos-containing materials in school buildings: guidance for
asbestos analytical programs. Washington, DC: USEPA. EPA 560/13-80-017A.
PB81-24358 6.
USEPA. 1980c. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Asbestos-containing materials in school buildings: guidance for
asbestos analytical programs. Statistical background document. Washington
DC: USEPA. EPA 560/5-80-17B.
USEPA. 1980d. U.S. Environmental Protection Agency. Office of Pesticides
and Toxic Substances. Friable asbestos-containing materials in schools;
proposed identification and notification. (45 FR 61966).
118
-------
USEPA. 1980e. U.S. Environmental Protection Agency. Office of Pesticides
and Toxic Substances. Identifying potential asbestos exposures in schools:
the Montgomery County experience. Washington, DC: USEPA. EPA 560/13-80-039.
PB81-24350 3.
USEPA. 1980f. U.S. Environmental Protection Agency. Office of Pesticides
and Toxic Substances. Measurement of asbestos air pollution inside buildings
sprayed with asbestos: Paris. Washington, DC: USEPA. EPA 560/13-80-026.
PB81 14700 1.
USEPA. 1980g. U.S. Environmental Protection Agency. Office of Pesticides
and Toxic Substances. Support document: asbestos containing materials in
schools; economic impact analysis of identification & notification; proposed
rule, Section 6, Toxic Substances Control Act. Washington, DC: USEPA.
EPA 560/12-80-004. PB81-14164 0.
USEPA. 1980h. U.S. Environmental Protection Agency. Office of Pesticides
and Toxic Substances. Support document: commercial and industrial uses of
asbestos; economic analysis of reporting forms; proposed rule - Section 8(a),
Toxic Substances Control Act. Washington, DC: USEPA. EPA 560/4-81-001.
PB81-15616 8.
USEPA. 1981a. U.S. Environmental Protection Agency. Office of Pesticides
and Toxic Substances. Asbestos in schools. Washington, DC: USEPA.
EPA 560/5-81-002. PB81-22582 3.
USEPA. 1981b. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Asbestos-containing materials in school buildings: bulk sample
analysis quality assurance program, Round 2. Washington, DC: USEPA.
EPA 560/5-81-001. PB81-22584 9.
USEPA. 1981c. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Airborne asbestos levels in schools: a design study. Washington,
DC: USEPA. EPA 560/5-81=006.
USEPA. 1982a. U.S. Environmental Protection Agency. Asbestos; friable
asbestos-containing materials in schools; identification and notification.
(47 FR 23360).
USEPA. 1982b. U.S. Environmental Protection Agency. Bulk sample analysis
for asbestos content: evaluation of tentative method. Research Triangle Park,
NC: Environmental Materials Science Laboratory, USEPA. EPA 600/4-82-021.
Zivy, P. 1982. Pulmonary and pleural radiology of 6,063 workers exposed to
asbestos in an industrial environment. World Symposium on Asbestos, Montreal,
Quebec, Canada, May 24-27, 1982. Sponsors: Governments of Canada and Quebec
and the Commission of European Communities.
119
-------
APPENDIX A
DESIGN CONSIDERATIONS AND SAMPLING SCHEME
121
-------
SAMPLE DESIGN AND SELECTION
After consideration and development of the informational requirements
needed to satisfy the objectives of the study, the first step in designing a
survey is to determine the population and subpopulations of interest, the pa-
rameters to be estimated, and constraints with respect to estimation preci-
sion, cost, and time. Given the statement of the informational requirements,
operational study population definitions can be developed and appropriate sam-
pling frames constructed. The sample design and survey operating procedures
can then be developed to meet the estimation precision and cost requirements.
Given the probability structure specified by the sample design, the appropri-
ate statistical analyses can be prescribed.
The design of the survey for this study of airborne asbestos levels in
schools involved many operationally complex issues, e.g., air sampling instru-
mentation and methodology, logistics regarding both equipment and field staff,
and laboratory analysis methodology. This appendix gives the considerations
in development of the survey design (USEPA 1981). This appendix also de-
scribes the sample design and the results of sample selection.
I. DESIGN CONSIDERATIONS
A. Objective One
The first objective of the study was to collect air sampling data in
order to document potential exposure to airborne asbestos from asbestos-
containing materials in schools. A number of air sampling studies have been
conducted to establish the level of airborne asbestos in public buildings and
in specific areas of buildings where asbestos-containing materials are present
and to compare these levels to ambient concentrations. Nicholson (1978) re-
ported on ambient asbestos samples collected in 48 U.S. cities in 1969 and
1970. The majority of ambient samples contained less than 20 ng/m3 of asbes-
tos. Other studies reported by Nicholson that measured airborne asbestos both
inside and outside buildings showed indoor averages ranging from 2.5 to 200
ng/m3 and outdoor averages from 0 to 48 ng/m3. In Paris airborne asbestos
concentrations inside public buildings and outdoors were studied from 1974
through 1976 (USEPA 1980f). The results showed ambient concentrations of
chrysotile asbestos, with maximum concentration observed to be 10 ng/m3. The
distribution of airborne asbestos concentrations inside buildings where
asbestos-containing materials were identified ranged from below 10 ng/m3 to
approximately 800 ng/m3.
The results of these studies provide evidence that airborne asbestos
levels are elevated when asbestos-containing materials are present. None of
these previous studies, however, had a probability sample design to estimate
airborne levels in a defined study population.
Plans for the current study specified that airborne asbestos concentra-
tion levels were to be determined by air sampling to collect fibers on a fil-
ter followed by transmission electron microscopy analysis to estimate concen-
trations in nanograms per cubic meter (ng/m3). The indoor concentrations at
122
-------
sites where asbestos-containing materials were present were to be compared
with those at indoor sites where no asbestos was present and with ambient
Levels. At each school where air sampling was conducted, an indoor control
sample and an ambient sample were also to be collected.
From a statistical perspective the ideal way to accomplish the first ob-
jective is to do a national study of schools, and sites within schools se-
lected according to a probability sample design. The study results would pro-
vide unbiased national estimates along with stated precision of the airborne
asbestos concentration levels and the concentration differential between sus-
pect sites and controls. However, it was determined that a national study
was neither feasible nor necessary at this time. Difficulties associated with
soliciting cooperation of school systems, establishing background information
to support the implementation of the sampling design, and transporting equip-
ment and staff throughout the nation would delay the results beyond the point
in time when they could be useful to U.S. EPA for rulemaking. In addition,
if airborne asbestos concentrations were found to be elevated in one school
system where asbestos-containing materials were present, that evidence would
be sufficient to alert other school systems of the potential problem.
One large urban school district had begun its own program to assess as-
bestos exposure. It had identified many schools that had asbestos present by
collecting and analyzing bulk samples. Also, it had scored areas within these
schools on algorithm factors. These data served as a base for constructing
stratification variables. A probability sample design was developed for this
school district.
B. Objective Two
The second objective of the study was to determine the best combination
of factors for predicting airborne asbestos levels in schools. The exposure
assessment tools to be considered in this study were based on the factors of
the algorithm (Table A-l) and a measurement of the percentage of asbestos
present in the site material by analyzing bulk samples taken from the site.
The primary goal in assessing exposure to asbestos in schools is to be able
to initiate corrective action in those situations that present an unreasonable
risk to the health of persons who use school buildings.
Of concern is the validity of exposure assessment (decision) rules re-
lated to taking a corrective action or deferring action to a later time. Two
general types of assessment rules are considered. The first is characterized
by an approximately continuous measurement scale that allows for differentia-
tion among air levels at all points on the scale. The second is characterized
by a discrete measurement scale that differentiates only between high and low
levels of airborne asbestos concentrations. To highlight the design require-
ments for the second objective, the framework underlying each type of rule is
described by specific examples of the two types of rules.
For the first case, the algorithm is used. A score is produced by sum-
ming factors 1 through 6 and multiplying the sum by factors 7 and 8. (Refer
to Table A-l for a description of the factors.) When considering this algo-
rithm, the decision rule is derived from the relationship between air levels
123
-------
Table A-l. Algorithm Factors and Their Weighted Scores
Factor
Weighted
score
1. Condition
No damage 0
Moderate damage 2
Severe damage 5
2. Accessibility
Not accessible 0
Rarely accessible 1
Accessible 3
3. Part of air moving system
No 0
Yes 1
4. Exposure
Material not exposed 0
10% or less of the material exposed 1
Greater than 10% of the material exposed 4
5. Water damage
No water damage 0
Minor water damage 1
Moderate or major water damage 2
6. Activity or movement
None or low activity level 0
Moderate activity level 1
High activity level 2
7. Friability
Not friable 0
Low friability 1
Moderate friability 2
High friability 3
8. Percentage asbestos
Less than or equal to 1% 0
Greater than 1% and less than or equal to 50% 2
Greater than 50% 3
124
-------
of asbestos and algorithm scores. Conceptually, the relationship takes the
form shown in Figure A-l. As the score increases, airborne asbestos concen-
tration increases. In the figure, A represents the airborne asbestos level
that defines the cutoff point between safe and unsafe levels. A in turn de-
termines S on the algorithm scale that represents the cutoff point defining
corrective action versus the deferral of corrective action. It is desirable
to be able to specify a value of A as a cutoff point. Therefore, validation
of the algorithm as an exposure assessment tool is based on demonstrating that
there is a high level of precision associated with the establishment of the
curve in Figure A-l.
For the second case, the decision tree algorithm (DTA) is considered as
an example. Each factor outcome is reduced to a dichotomy. The decision rule
for action or deferred action arises from the classification of sites into
one of two groups. For convenience, the groups are named Action (Group A)
and Deferred Action (Group DA). The site is classified into A or DA according
to the tree shown in Figure A-2. Using four factors to define the decision
tree, there are 16 types of sites shown in Table A-2. Five sites belong to
Group A; 11 sites belong to Group DA. From a conceptual perspective, there
is a distribution of airborne asbestos concentrations corresponding to each
group (see Figure A-3). Validation of DTA involves the comparison of the
characteristics of these distributions.
When confronted with both a validation and a development objective, as
is the case here, it can be difficult to specify a sample design that is si-
multaneously satisfactory for each. Since the assessment tools to be evalu-
ated will all be derived from the eight factors in Table A-l, it is intui-
tively appealing to suggest that air sampling data should be collected under
all combinations of the factors. Then the relationship between airborne as-
bestos concentrations and factors can be thoroughly investigated in search of
a "best" assessment tool. Unfortunately, this approach is unrealistic because
(1) the total number of combinations determined by these factors is 944,784,
which is unmanageable, and (2) it is not possible to know which combinations
actually exist in any specified study area until the sites are visited and
scored. Additionally, the distribution of the sample sites among the existing
factor combinations that is optimal (in terms of estimation precision) for
this second objective is not the same as the distribution that is optimal for
the first objective. This is because the algorithm factor combinations are
not present in equal proportions in the study population, and oversampling
certain subgroups (here, factor level combinations) to facilitate subgroup
estimation tends to reduce the overall estimation precision. Thus, the two
objectives have mutually conficting implications for the sample design. The
sample design employed for this survey involved a "compromise approach," im-
posing constraints towards satisfying both objectives. Of course, the estima-
tion precision associated with each objective is somewhat less than it would
be under a design developed to satisfy just the one objective.
125
-------
0)
I
**•
0)
_Q
O
_Q
c
O
O
C
ro
Ao, airborne asbestos level that defines cutoff
point between safe and unsafe level; So> airborne
asbestos level that defines cutoff point between
corrective action and deferral of corrective action
Deferred
action
region
Action region
Algorithm score
Figure A-l. Hypothesized relationship between air level and
algorithm score.
126
-------
Is the Material
Normally Exposed?
No (0.1)
Yes (4)
Yes (5)
Is the Material
Badly Damaged?
No (0,2)
Is the Material
Accessible?
No (0,1)
Yes (3)
Deferred Action
•>• Deferred Action
Yes (2, 3)
Is the K
Highly 1
Material
-riable?
No (0,1)
*> Deferred Action
Figure A-2. Original asbestos exposure assessment decision tree.
Applied to sites with > 1% asbestos in bulk sample. Numbers in
parentheses are weighted algorithm scores. (Source: decision tree
formulated in 1980 by Battelle researchers and U.S. EPA task leader.)
127
-------
a
Table A-2. Site Description for Decision Tree Algorithm
Site
no.
2
3
5
8
9
12
14
15
17
20
22
23
26
27
29
32
Factor
Friability
Low
Low
Low
Low
High
High
High
High
Low
Low
Low
Low
High
High
High
High
Condition
Good
Good
Bad
Bad
Good
Good
Bad
Bad
Good
Good
Bad
Bad
Good
Good
Bad
Bad
Exposure
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Accessibility
High
Low
Low
High
Low
High
High
Low
Low
High
High
Low
Low
Low
Low
High
Group
DA
DA
DA
A
DA
A
DA
A
DA
DA
DA
A
DA
DA
DA
A
a Applied to sites with > 1% asbestos in bulk sample.
b DA - deferred action; A - action.
128
-------
v
Group DA
\
\
Figure A-3. Hypothetical distribution of airborne asbestos concentration levels.
Group A, action; Group DA, deferred action.
-------
II. SAMPLE DESIGN
A. First-Stage Sample
1. Construction and Stratification of the First-Stage Frame
The first-stage frame consisted of all schools having student areas in
which material suspected of containing asbestos was present. This list of
schools was compiled on April 28, 1981, from the school district's asbestos
program information.
The first-stage frame was stratified into nine classes formed by consid-
ering three categories of asbestos content—low, high, and unknown—with three
categories of friability--low, moderate, and high. These nine classes, or
strata, are listed in Table A-3. Schools were classified according to the
asbestos content and friability of the sampling area. If a school contained
more than one sampling area and if those sampling areas differed as to asbes-
tos content and friability, the school was placed into the stratum (among
those strata corresponding to sampling areas in the school) that had substan-
tially fewer schools. If no stratum corresponding to a sampling area in the
school contained substantially fewer schools than the others, then the school
was stratified according to the sampling area of largest area.
Table A-3. First-Stage Strata, Constructed from Asbestos Content
and Friability Categories
Stratum
1
2
3
4
5
6
7
8
9
Asbestos
content
Low
Low
Low
High
High
High
Unknown
Unknown
Unknown
1
Friability
Low
Moderate
High
Low
Moderate
High
Low
Moderate
High
Asbestos content taken from school district's asbestos program laboratory
results known as of April 28, 1981.
Low: > 1% and < 20% asbestos
High: ^ 20% asbestos
Unknown: Laboratory results not known as of April 28, 1981.
Friability ratings taken from worksheets prepared by school district
personnel during inspection of schools for its asbestos program.
Low: Rated 1
Moderate: Rated 2
High: Rated 3
130
-------
For stratification of the first-stage frame, asbestos content was taken
from the district's asbestos program laboratory results that were known as of
April 28, 1981. Low asbestos content was defined as asbestos present at a
concentration less than 20%. (Asbestos was said to be present if the average
asbestos concentration for the sampling area exceeded 1%.) High asbestos con-
tent was defined as asbestos concentration equal to or greater than 20%.
Friability ratings used in stratification were taken from worksheets prepared
by district personnel during their inspection of schools for their asbestos
program.
2. Allocation and Selection of the First-Stage Sample
The allocation of the 25 sample schools among the nine first-stage strata
is shown in Table A-4. Of the 25 sample schools, 14 were selected from strata
1 through 6 (asbestos content known), and 11 were selected from strata 7
through 9 (asbestos content unknown). This allocation between the two groups
is not in proportion to size or enrollment; an allocation proportional to en-
rollment would result in selecting 8 schools from strata 1 through 6 and 17
schools from strata 7 through 9. More schools were selected from strata 1
through 6, asbestos content known, to ensure the desired distribution of sam-
ple schools with respect to level of asbestos content in combination with
level of friability. Among strata 7 through 9, the 11 sample schools were
allocated in proportion to enrollment, subject to the restriction that at
least one school be selected from each first-stage stratum. Among strata 1
through 6, the 14 sample schools were allocated approximately in proportion
to enrollment but with slightly greater emphasis on high asbestos content as
opposed to low asbestos content.
Table A-4. Allocation of Sample Schools Among First-Stage Strata
Asbestos Number of
Stratum content Friability sample schools
1 Low Low 1
2 Low Moderate 3
3 Low High 3
4 High Low 1
5 High Moderate 4
6 High High _2
Total for strata 1 to 6 IA
7 Unknown Low 1
8 Unknown Moderate 9
9 Unknown High _1
Total for strata 7 to 9 11
TOTAL 25
131
-------
The required number of schools (Table A-4) was selected from each stratum
with probability proportional to size and without replacement. School enroll-
ment was used as a size measure. The preferred size measure would have been
the number of eligible sites suspected of containing asbestos per school.
However, this information was not readily available, and it was thought that
school enrollment would have a fairly strong positive relationship with this
number. A reason for selection with probability proportional to size is to
have selection probabilities at the first stage such that a self-weighting
(to the extent possible given the design constraints) sample can be obtained
at the final stage. For a discussion of selection with probability propor-
tional to size, see Kish (1965). Note that the design constraints cause this
sample not to be self-weighting. Hence, to obtain parameter estimates for
the school district, weighted statistical analysis is required. To obtain
valid estimates in a superpopulation setting requires that cell means be based
on weighted estimates.
3. First-Stage Selection Probabilities
The selection probability for a school is equal to the ratio of the
school's enrollment to the total enrollment for its stratum multiplied by the
number of schools selected from that stratum. Table A-5 gives selection prob-
abilities for the 25 sample schools. These selection probabilities were used
in calculating sampling weights for statistical analysis of the data, as de-
scribed in Section 7 of this report.
B. Second-Stage Sample
1. Construction and Stratification of the Second-Stage Frame
The second-stage frame consisted of all eligible sites suspected of hav-
ing asbestos-containing material in the first-stage sample of 25 schools.
These sites were counted and listed using the marked floor plans that were
prepared by school district personnel for the school district's asbestos pro-
gram. (Because of time constraints, no school visits for the purpose of
counting and listing eligible sites were possible.) The second-stage frame
was stratified according to the presence or absence of asbestos-containing
material.
Table A-6 gives the number of asbestos-containing material sites on the
second-stage frame by school and friability/condition/exposure/accessibility
category. These categories are described in Table A-7. A site was placed in
a category according to the rating its entire sampling area received from
school district personnel. Ratings for individual sites were not readily
available. With the exception of three sample schools, all eligible sites
from the same sample school fell into the same friability/condition/exposure/
accessibility category. This was, in part, because schools often had only
one or two eligible sampling areas, and because sites were classified by sam-
pling area ratings.
132
-------
Table A-5. Probabilities of Selection for Sample Schools
School
code
22
20
13
21
25
6
23
24
9
3
8
16
4
15
11
14
5
17
18
2
7
1
12
19
10
First-stage
stratum
1
2
2
2
3
3
3
4
5
5
5
5
6
6
7
8
8
8
8
8
8
8
8
8
9
Probability of
selection
.3202
.2147
.5228
.2147
.2953
.9130
.1830
1.0000
.7364
1.0000
.5186
.1815
.5726
.1554
.2464
.1982
.2028
.0785
.2660
.2265
.3005
.3915
.3968
.0536
.2855
133
-------
Table A-6. Number of Eligible Asbestos-Containing Material Sites
on Second-Stage Frame
School
code
22
20
13
21
25
6
23
24
9
3
8
16
4
15
11
14
5
17
18
2
7
1
12
19
10
Total
number
of sites
Friability/ condition/exposure/accessibility/ category
3
39
0
0
0
0
0
0
38
0
0
0
0
0
0
17
0
0
0
0
0
0
0
0
0
0
94
11
0
32
0
30
0
0
0
0
0
0
2
16
0
0
0
40
7
11
1
0
0
0
40
9
0
188
12
0
0
56
0
0
0
0
0
51
111
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
218
15
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
83
0
0
0
83
16
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
96
0
0
0
0
96
19
0
0
0
0
11
60
0
0
0
0
0
0
1
1
0
0
0
0
0
76
0
0
0
0
0
149
20
0
0
0
0
13
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
13
23
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
4
5
24
0
0
0
0
0
0
37
0
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
40
Total
number
of sites
39
32
56
30
24
60
37
38
54
111
2
16
2
1
17
40
7
11
1
76
96
83
40
9
4
886
Sites placed in categories according to sampling area ratings provided by
school district. Categories not included on this table contained no sites
from 25 sample schools, according to sampling area ratings. A description
of the categories is provided in Table A-7.
134
-------
Table A-7. Description of Friability/Condition/Exposure/Accessibility
Categories
Category
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Friability
Low
Low
Low
Low
Low
Low
Low
Low
Moderate
Moderate
Moderate
Moderate
Moderate
Moderate
Moderate
Moderate
High
High
High
High
High
High
High
High
Condition
Good
Good
Good
Good
Bad
Bad
Bad
Bad
Good
Good
Good
Good
Bad
Bad
Bad
Bad
Good
Good
Good
Good
Bad
Bad
Bad
Bad
Exposure
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Accessibility
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
a Friability: Low = 1, Moderate = 2, High
b Condition: Good =0,2, Bad = 5
c Exposure: Low = 1, High = 4
d Accessibility: Low = 0,1, High = 4
= 3
135
-------
2. Allocation and Selection of the Second-Stage Sample of
Asbestos-Containing Material Sites
The second-stage sample of 48 asbestos-containing material sites was al-
located among the 25 sample schools subject to the following restrictions:
(1) at least one site must be selected from each of the 25 schools in the
first-stage sample, and (2) at least one site must be selected from each non-
empty friability/condition/exposure/accessibility category. To the extent
possible, the second-stage sample was allocated among schools and friability/
condition/exposure/accessibility categories proportional to the number of
sites. Table A-8 displays this allocation.
As mentioned, eligible sites were listed using floor plans provided by
the school district. For each school (or school and each category given in
Table A-8), a starting point was selected on the school floor plan. From this
starting point (a site), eligible asbestos-containing material sites were
listed in order according to location. A random systematic sample of the re-
quired number of sites was then selected from this listing. This method was
used because of the suspected tendency of nearby sites to resemble each other
with respect to airborne asbestos levels. Systematic sampling assured a cer-
tain degree of spread in location among the selected sites within a school;
i.e., it provided the benefits of a location stratification. (As can be noted
from Table A-8, in 13 of the 25 sample schools, more than one site was
selected.)
For each of the selected sites, alternate sites were also selected.
Field personnel were to substitute an alternate site for a selected site when-
ever a selected site was found to be not eligible or whenever there was non-
response (failure to obtain the required observation) at a selected site. A
selected asbestos-containing material site was to be classified as not eligi-
ble if, contrary to floor plan information, the site was not being used as a
student activity area or the site did not actually contain asbestos material.
Nonresponse could occur, for example, if a teacher or school official would
not permit air sampling at the selected site or if air sampling was not possi-
ble at the site because of lack of an electrical outlet, repeated vandalism,
etc. An alternate site could not be substituted for a selected site simply
because it might be more convenient for air sampling. These procedures for
substitution were carefully followed by field personnel.
A summary of site selections, including alternate sites, is given in
Table A-9. Of the 48 selected sites, nonresponse occurred at 6 sites, a non-
response rate of 12%.
136
-------
Table A-8. Allocation of Second-Stage Sample of Asbestos-Containing Material
Sites Among Schools and Friability/Condition/Exposure/Accessibility
Categories
School
code
22
20
13
21
25
6
23
24
9
3
8
16
4
15
11
14
5
17
18
2
7
1
12
9
10
Number of
sample
sites
Q
Friability/ condition/ exposure/accessibility/ category
3
2
0
0
0
0
0
0
2
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
5
11
0
2
0
2
0
0
0
0
0
0
1
1
0
0
0
3
1
1
1
0
0
0
2
1
0
15
12
0
0
3
0
0
0
0
0
0
4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7
15
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3
0
0
0
3
16
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3
0
0
0
0
3
19
0
0
0
0
2
3
0
0
0
0
0
0
1
1
0
0
0
0
0
4
0
0
0
0
0
11
20
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
23
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
24
0
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
Number of
sample
sites
2
2
3
2
3
3
1
2
1
4
1
1
1
1
1
3
1
1
1
4
3
3
2
1
1
48
Categories not included on this table contained no sites from 25 sample
schools, according to sampling area ratings. A description of the
categories is provided in Table A-7.
137
-------
Table A-9. Asbestos-Containing Material Sites and Selection Probabilities
in Second Stage of Selection
School Number of eligible sites with Site
code asbestos-containing material code
22
20
13
21
25
6
23
24
9
3
8
16
4
15
11
14
5
17
39
32
56
30
11
13
60
37
38
3
111
2
16
1
1
17
40
7
11
1
2C
1
2
1
2
3d
1
2
1
2
3
1C
2
3°
1
1
2e
1
I8
2
3
4
1
1
1
1
1
1
2
3
1
1
Probability,
of selection
.0513
.0513
.0625
.0625
.0536
.0536
.0536
.0667
.0667
.1818
.1818
.0769
.0500
.0500
.0500
.0270
.0526
.15796
.0185f
.07208
.0360
.0360
.0360
.5000
.0625
.5000f
1.0000
.0588
.0750
.0750
.0750
.1428
.0909
(continued)
138
-------
Table A-9 (continued)
School
code
Number of eligible sites with
asbestos-containing material
Site
code
Probability,
of selection
18
2
12
19
10
1
76
96
83
40
9
4
1
2
3
1
2
3
21
1
1.0000
.0526,
.1053h
.0526
.0526
.0312
.0312
.0312
.0361
.0361
.0361
.0500.
.20001
.1111
.5000h
a According to school district floor plan information.
b Probability is conditional on first-stage sample of schools.
c Second alternate site was used. Nonresponse occurred at selected and
first alternate sites.
d First alternate site was used. Nonresponse occurred at the selected site.
e Selected and first alternate sites were not eligible. Second alternate
site was used, and selection probability was adjusted accordingly.
f Selection probability was adjusted to account for the nonrepresented
category (see Tables A-6 and A-8) at this school.
g Nonresponse occurred at selected site. First and second alternates were
not eligible. A third alternate site was used, and selection probability
was adjusted accordingly.
h Selected site was not eligible. First alternate site was used, and
selection probability was adjusted accordingly.
i Selected, first alternate, and second alternate sites were not eligible.
A third alternate was used, and selection probability was adjusted
accordingly.
139
-------
3. Second-Stage Selection Probabilities for the Asbestos-
Containing Material Sites
Conditional on the first-stage sample of schools, the selection probabil-
ity of an asbestos-containing material site is calculated as the number of
asbestos-containing material sites selected from the school (or school and
each category given in Table A-8) divided by the number of eligible asbestos-
containing material sites in the school (or school and each category given in
Table A-8). The number of eligible asbestos-containing material sites counted
using the marked school floor plan might not be the actual number of such
sites. Some selected sites were in fact found to be not eligible, and alter-
nate sites were substituted (Table A-9). In such a case, the number of sites
selected divided by the number of sites listed is multiplied by the number of
sites (selected and alternate sites) inspected to locate an eligible site sus-
pected of containing asbestos. This value estimates the probability of site
selection, conditional on the first-stage sample of schools. For example,
consider school 10. Four sites were listed, and a site was randomly selected.
Upon inspection, it was found that the selected site did not have asbestos-
containing material. The first alternate was used for data collection. The
site selection probability was then estimated using the following equation:
(1 site selected/4 sites listed) x 2 sites inspected = (1/4) x 2 = .5000.
This method arose from estimating the number of sites on the list that were
eligible and suspected of containing asbestos by dividing the number of sites
on the list by the number of chosen sites inspected to locate an eligible site
suspected of containing asbestos.
The selection probabilities calculated here are conditional on the first-
stage sample of schools. Thus a site's selection probability (conditional on
the first-stage sample of schools) cannot be multiplied by the selection prob-
ability of its school to yield the overall selection probability of the site.
4. Selection of the Second-Stage Sample of Control Sites
At each of the 25 schools in the first-stage sample, a control site was
selected. For a sample school, all student activity areas without asbestos-
containing material were listed using the marked floor plan provided by the
school district. A control site was randomly selected from this list. All
sites on the list had equal probabilities of being selected. Alternate con-
trol sites were also selected. An alternate was used to replace a selected
control site if, upon inspection by field personnel, it was found that the
selected site was actually not eligible or if there was nonresponse. A se-
lected site was ineligible if, contrary to floor plan information, the site
was no longer being used as a student activity area or if the site did have
material suggested of containing asbestos. (This process facilitates the es-
timation of list incompleteness for eligible sites suspected of containing
asbestos.) Nonresponse refers to situations such as teacher or administrator
refusal to allow air sampling at the selected site, lack of an electrical out-
let needed to conduct air sampling, etc.
In one sample school, the marked floor plans indicated that there were
no student activity areas without asbestos-containing material. In this case,
all areas without asbestos-containing material were listed, though these areas
140
-------
were not student activity areas. A control site was randomly selected from
this list. In a few other sample schools, the marked floor plans indicated
that the schools contained no areas, student activity or not, without asbestos-
containing material. In this situation, field personnel inspected the school
to find an area without asbestos-containing material for conducting air sam-
pling.
Table A-10 is a summary of control site selection. The probabilities of
selection were calculated in the same way that the second-stage selection
probabilities (see Table A-9) were calculated. There was no requirement that
control sites be a specified distance from material suspected of containing
asbestos. In other words, a control site could be immediately next door to a
room with asbestos-containing material, or it could be on a floor entirely
free of asbestos-containing material.
5. Selection of Outdoor Ambient Sites
At each school an outdoor site was selected to sample ambient air. This
site was on the roof of the school building at a location remote from any
school airflow ventilation exhaust point. No randomization process was in-
volved in selecting a site's location outside the school building.
6. Selection of Bulk Sampling Locations
Three bulk samples of asbestos-containing material were collected at each
of the 48 study sites. Bulk sampling locations were selected from each site
according to the USEPA method (1980a, 1980b). This method involves use of a
random number pair procedure to select a simple random sample of bulk sampling
locations; all locations in the area of interest have equal probabilities of
selection. Additionally, one of the three bulk sampling locations at each
site was randomly selected for collection of duplicate (side-by-side) bulk
samples.
C. Long-Term Sampling
At three sites randomly selected from the 48 sample sites, air sampling
was conducted for three consecutive periods of 5 school days. The purpose of
this sampling was to investigate the variability of airborne asbestos levels
over time.
To select the three sites, the list of 48 sample sites was partitioned
into three strata. Sites were first classified as to type of room. Auditor-
iums, libraries, cafeterias, and gymnasiums formed the first stratum. These
types of rooms were selected from the standpoints of size and activity level.
The remaining sites were classified according to asbestos content and friabil-
ity, based on the school district's information at the sampling area level.
The second stratum contained sites with high friability or a combination of
high asbestos content and moderate friability. The third stratum contained
the remaining sites. Tables A-11 through A-13 display the three strata.
One site was selected from each of the three strata, and within a stratum
all sites had an equal probability of selection. The sites selected for long-
term sampling are indicated in Tables A-ll through A-13.
141
-------
Table A-10. Summary of Control Site Selection
School
code
22
20
13
21
25
6
23
24
9
3
8
16
4
15
11
14
5
17
18
2
7
1
12
19
10
Number of eligible sites without
asbestos-containing material
a
22
32
2
10
69
2
1
5
a
d
5
80
20
73
a
e
f
18
22
14
51
16
9
82
Probability of selection
from its school
a
.0454
.0312
.5000
.1000
.0290C
.5000
1.0000
.2000
a
d
.2000
.0125
.0500
.0274°
a
e
f
.0555
.0909°
.0714
.0196
.0625
.1111
.0122
According to school district's floor plan information, school contained no
areas without asbestos-containing material. Field personnel inspected
school and found a non-student area without asbestos-containing material
for control air sampling.
First alternate was used as substitute for selected control site, due to a
nonresponse situation.
First alternate was used as substitute for selected control site because
selected site was no longer being used as student area. Selection
probability was adjusted accordingly.
No school floor plan was available. Field personnel inspected school and
located control site in a student area.
No school floor plan was available. Field personnel inspected school and
located control site in a non-student area.
According to school district's floor plan information, school contained no
student areas without asbestos-containing material. A non-student area
was randomly selected to be used as a control site.
142
-------
Table A-ll. Selection of Long-Term Sampling Sites: Stratum 1
School
code
25
9
3
8
4
15
5
18
10
a Large student activity areas (auditoriums,
cafeterias, and gymnasiums).
b Site selected for long-term sampling.
Site
code
1
1
2
1
1
1
1
1
1
libraries ,
Table A-12. Selection of Long-Term Sampling Sites: Stratum 2a
School Site
code code
25 2
3
6 1
2
3
23 1
3 3l
4
16 1
a Classrooms and corridors with high friability or a combina-
tion of high asbestos content and moderate friability.
b Site was selected for long-term sampling.
143
-------
Table A-13. Selection of Long-Term Sampling Sites: Stratum 3
School
code
22
20
13
21
24
11
14
17
2
Site
code
1
2
1
2
1
2
3
1
2
1
2
1
1
2
3
1
1
l<
7 1
2
3
1 1
2
3
12 1
2
19 1
Classrooms and corridors with low friability, low asbestos
content and moderate friability, or unknown asbestos content
and moderate friability.
Alternate, large student activity area used to replace this
site.
Site was selected for long-term sampling.
144
-------
D. Sampling Period
Air sampling for this study was conducted during a period of approxi-
mately 4 weeks. Except at the long-term sites, air sampling was conducted
for 5 consecutive school days at each site when possible. A randomization
procedure was used to determine the order for monitoring the 22 schools that
did not contain long-term sites. For operational efficiency, the schools were
first paired according to location. The pairs were then randomly ordered, and
this determined the order of field visits to the schools for air sampling.
REFERENCES
Nicholson WJ. 1978. Mount Sinai Sch. of Medicine, City of New York. Control
of sprayed asbestos surfaces in school building. Washington, DC: Natl. Inst.
of Environmental Health Sciences. Contract N01-ES-7-2113.
USEPA. 1980a. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Asbestos-containing materials in school buildings: guidance for
asbestos analytical programs. Washington, DC: USEPA. EPA 560/13-80-017A.
PB81 24358 6.
USEPA. 1980b. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Asbestos-containing materials in school buildings: guidance for
asbestos analytical programs. Washington, DC: USEPA. EPA 560/5-80-17B.
USEPA. 1980c. U.S. Environmental Protection Agency. Office of Pesticides
and Toxic Substances. Measurement of asbestos air pollution inside buildings
sprayed with asbestos. Washington, DC: USEPA. EPA 560/13-80-026. PB81
14700 1.
USEPA. 1981. U.S. Environmental Protection Agency. Office of Toxic Sub-
stances. Airborne asbestos levels in schools: a design study. Washington,
DC: USEPA. EPA 560/5-81-006.
145
-------
APPENDIX B
PROTOCOL FOR AIR SAMPLING
147
-------
PROTOCOL FOR AIR SAMPLING
Airborne asbestos sampling will be conducted according to the general
procedure outlined by Price et al. (1980). This will involve samples taken
at each site (room or hallway) as designated by the site selection team. In
addition, one outside ambient sample and one inside control sample (at a site
chosen under the guidance of the selection team) will be taken at each group
of sites (building). Field and laboratory blanks will be taken and analyzed
for quality control.
Each sample system will be equipped with automatic timers and set to op-
erate during hours of normal school activity (0730 to 1530) over a period of
five consecutive school days, and the sample rate will be 5 L/min. The col-
lection substrate will be 0.45-(Jm, 47-mm Millipore HA cellulose acetate mem-
brane filters.
I. SELECTION OF SAMPLE LOCATION
A. Inside Samplers
Once a site has been identified, the indoor sample system is placed,
within practical constraints, so as to collect a representative sample of the
entire site. The filter is to be placed at a height of 1.5 m and in the least
conspicuous location possible in order to minimize disruption of normal activ-
ity. The sample should not be located in a high activity area (doorway),
within 2 ft of a wall or within 5 ft of a window. (Typically, a sample is
placed on a table in the back of a classroom or on a table against the wall
at the midpoint of a hallway.) Attention should also be given so as to ensure
that the sampler in operation does not create an unsafe situation (e.g., ex-
tension cord across a doorway).
B. Outside Ambient Sampler
The location of the outside ambient sampler is important in obtaining a
representative background measure. The sampler, thus, should be placed upwind
of the building it is to represent such that no bias is created by identifi-
able local sources (parking lots, highways, building exhaust). With regard
to the above considerations, as well as power requirements and anticipated
accessibility to vandals, the upwind side of a building roof may be the most
desirable location.
II. SAMPLE SETUP
The sampling system consists of:
1. A Gelman magnetic-type open-face filter.
2. A critical flow orifice.
3. A pump with muffler.
4. Associated plumbing and stand.
5. A 7-day timer.
148
-------
The sampler setup is schematically represented as follows.
1
1
Filter
Holder
/\
V
Criticial
Flow
Orifice
rv~*
\ )
^~s
Pump
With
Muffler
1
1 1
Timer
Electrical Power
Source
III. PROTOCOL
1. Clean and dry filter holder.
2. Place filter in holder, assuring proper position (see filter han-
dling section).
3. Mount filter holder such that filter is in a vertical position
(perpendicular to ground).
4. Check plumbing for any leaks.
5. Check flow with flowmeter using manual control of pump.
6. Set automatic timer to desired on-off time settings.
7. Make appropriate logbook entries.
8. Conduct sampling.
9. After sampling period, check flow.
10. Remove filter and place in Millipore petri dish for proper handling
and transport.
IV. FILTER HANDLING
During loading and unloading of the filter holder, the filters are han-
dled by forceps (not with fingers). When a filter is removed after exposure,
it is placed in the petri holder exposed side up and maintained in that posi-
tion during the handling and transport of samples back to the laboratory.
The samples will be hand-carried to Battelle Columbus Laboratories by Battelle
149
-------
field personnel in a container that will keep the petri dish in a horizontal
(flat) position at all times (handling, transport, and storage).
The MRI field test leader will give the filters that are collected by
MRI field personnel to the Battelle field crew chief. The Battelle crew chief
will place the samples in a shipping container and transport them to the
Battelle Columbus Laboratories.
The chain-of-custody system will be followed at all times. A chain-of-
custody record, therefore, will be kept on each filter (or sample).
One field blank will be randomly selected at each location. Any dropping
or mishandling of a filter after collection must be recorded. Each filter
holder is labeled according to a coding system.
V. LOGBOOK ENTRIES
An important part of any successful field program is the accurate obser-
vations and records of the field team. At a minimum, logbook entries will
include:
1. Name of field operator.
2. Date of record.
3. Number and location of site.
4. Position of sampler within site.
5. Brief description of site.
6. Corresponding filter number.
7. Sample flow rate at start of sampling period.
8. Start time.
9. Stop time.
10. Sample flow rate at end of sampling period.
11. Comments.
REFERENCES
Price B, Melton C, Schmidt E, Townley C. 1980. Battelle Columbus Labs. Air-
borne asbestos levels in schools: a design study. Report. Washington, DC:
Office of Pesticides and Toxic Substances, U.S. Environmental Protection
Agency. Contract 68-01-3858.
150
-------
APPENDIX C
PROTOCOL FOR CREATING AND MAINTAINING CHAIN OF CUSTODY
151
-------
PROTOCOL FOR CREATING AND MAINTAINING CHAIN OF CUSTODY
The objective of this protocol is to establish a chain of custody for
the handling of samples, from which the results may be introduced as evidence
in legal proceedings. The chain-of-custody protocol described is based on
MRI's standard laboratory practice for the handling of field samples. The
manner in which the chain of custody is maintained may be altered by the proj-
ect monitor.
Whenever physical evidence is introduced in a trial, it must be shown
conclusively that the evidence introduced is in fact the evidence taken at
the scene (in the field). The results of chemical analysis of a sample may
be considered physical evidence and must be accounted for, back to the taking
of the original sample. If necessary everyone who was in possession of the
sample may be required to testify.
The chain of custody, then, consists of everyone who has possession of a
sample. The chain of custody begins with the person who collects the sample.
The person doing the sampling must actually collect, or witness the collection
of, the sample. The sample is then labeled with the following information:
a unique identifying code, the date of collection, the exact time of collec-
tion, and the sampler's signature. If a field logbook is being used, it
should be bound. The identifying number, location, date, time, and descrip-
tion of the sample are entered in ink, as is the signature or initials of the
person entering the foregoing data for each sample. A chain-of-custody form
is also filled out in ink showing the sampler as having received the sample.
From this point the sampler must be the only person with access to the sample
until it is released to the next link in the chain of custody. A good system
is to place the sample in a locked container or the trunk or interior of a
vehicle, which must remain locked when the sampler is not present. The person
having custody of the sample should retain possession of the keys. If the
person having custody of the sample relinquishes custody, it must be noted on
the chain-of-custody form, and the person receiving custody must note receipt
on the chain-of-custody form.
The Battelle crew chief will be responsible for all air samples taken by
the Battelle field crew. Samples taken by the MRI field crew will be given
to the Battelle crew chief as soon as possible after the samples have been
taken. The Battelle crew chief is totally responsible for sample custody upon
transfer of the MRI collected samples.
I. SAMPLE TRANSPORT
When samples are transported from the field to the laboratory, the ship-
ping container shall be securely fastened. Before closing the container, the
original copy of the chain-of-custody form is signed and enclosed. If pos-
sible, the container should be transported by a method in which everyone re-
ceiving custody of the container must sign for it.
The air samples (filters) that are collected will be hand-carried by
Battelle field crew personnel to Battelle Columbus Laboratories where they
152
-------
will be analyzed. These samples will be kept in a horizontal position at all
times.
When the container arrives at the laboratory, a quality control represen-
tative, not the person who will be preparing the samples, will take custody
of it. The quality control representative first examines the shipping con-
tainer for any evidence of damage or tampering. After a check of the outside
of the container, the examiner opens the container and checks each sample.
The quality control representative shall note any damage or indication of tam-
pering on the enclosed chain-of-custody forms and shall sign them. The ship-
ping invoice showing everyone who received the container will be attached to
one of the chain-of-custody forms.
The quality control representative has sole custody of the sample until
it is relinquished, though someone else may examine or prepare the sample for
analysis in full view of the quality control representative. A requirement
here is a locked storage area to which only the person having custody has ac-
cess. Notes concerning changes, manipulations, and storage conditions should
be entered in ink in the logbook, initialed, and dated.
Whenever custody of the sample is transferred from one person to another,
the person relinquishing custody of the sample must sign the chain-of-custody
form and note the time and date of the transfer. The person receiving the
sample must do the same. The person having custody of the sample has sole
control of access to the sample.
153
-------
APPENDIX D
PROTOCOL FOR THE SAMPLING AND ANALYSIS OF INSULATION
MATERIAL SUSPECTED OF CONTAINING ASBESTOS
155
-------
PROTOCOL FOR THE SAMPLING AND ANALYSIS OF INSULATION MATERIAL
SUSPECTED OF CONTAINING ASBESTOS
Bulk samples pf asbestos-containing material will be taken after air sam-
pling is completed at a site. The specific points where these samples will
be taken will be designated by the statistician at Research Triangle Institute
who was involved in the air sampling survey design.
I. SAMPLING
The bulk sampling procedure will be based on that presented in EPA docu-
ment entitled "Asbestos-Containing Materials in School Buildings—Guidance
for Asbestos Analytical Programs" (USEPA 1980). Three sampling points will be
selected from each air sampling site; two samples taken side by side at one
of the three points provide a duplicate for quality assurance. This procedure
eliminates the necessity of splitting samples at a later time.
An identification number will be assigned to each sample. This number
will also appear on the sampler container, in the field logbook along with
descriptive information, and on the chain-of-custody records. These numbers
will be sent to the field on preprinted replicate gum labels that have other
pertinent information on them.
II. SAMPLE HANDLING
The samples will be shipped by the field crew to the attention of the
quality control representative at MRI, who will log them in and assign them
permanent numbers on a random basis. The quality control representative will
then identify and remove the duplicates and, from this set of duplicates,
choose, on a random basis, a number of them for analysis by an external lab-
oratory. The remaining duplicates will be put back with the remaining sam-
ples, and all of these will be given to the MRI analyst for analysis. The
basis for the number of duplicate samples to be sent to the external labora-
tory is given later in this protocol under Quality Assurance.
III. ANALYSIS
The samples will be analyzed by polarized light microsocpy including dis-
persion staining. The procedure will closely follow that given in The Asbestos
Particle Atlas (McCrone 1980). The procedure is summarized in Figure D-l.
Quantitation of the asbestos content of the samples also will follow the pro-
cedure given in The Asbestos Particle Atlas.
IV. QUALITY ASSURANCE
Of the three bulk samples taken from each sampling site, one of the sam-
ples will be taken in duplicate for quality assurance.
Of the duplicate quality assurance samples, a number equal to 20% of the
first 100 samples (not including the duplicates) and 10% of the remainder will
be selected at random for analysis at a laboratory other than MRI. The remain-
ing duplicate samples will be analyzed at MRI. All samples for analysis will
have no identification other than the random identification number (OTS-XXXX).
156
-------
MOUNT A REPRESENTATIVE SAMPLE IN CARCILLE HIGH DISPERSION LIQUID nf) •= 1.550
ISOTROPIC
GLASS WOOL (106)a
Straight uniform
diameter cylinders.
XQ » 700 nm
MINERAL WOOL (111)
"Exotic" shapes, fibers
variable n (1.50-1.70)
PUMICE (226)
Fire-polished flakes
with vesicles, XQ » 700 nra
PERLITE (529)
Thin glass films,
foamed glass bubbles,
X0 J> '00 n»
DIATOMS (5)
Organized, pitted, flat.
sometimes elongated.
*0 » 700 nm
ANISOTROPIC
FIBROUS
CHRYSOTILE (122)
XQ - 600-700 nm (blue
i length; 500-600 (n)
WOOD FIBERS (70-73)
Blue (1 length).
yellow ( n length) , pitted
POLYESTER (100)
Cylindrical,
high birefringence
n,, •= 1.71, nA - 1.54
n"s > 1.55 (pale yellow colors)
Mount in 1.605 HD liquid
> 1 XQ < 700 nm
TREHOLITE (205)
Oblique extinction
view (15-20°) usually
shows yellow (n) and
blue (1); H extcn. :
yellow (n), magenta (1)
ANTHOPHYLLITE (121)
All views n extcn. ,
usually pale yellow (n);
golden-yellow to blue-
magenta (1)
ACTINOLITE (671)
Like tremolite, but all
AQ'S < 450 nm
WOLLASTONITE (735)
Not so fibrillar,
XQ'S (480-530 nm),
(+) and (-) elongation
All X0's < 4<>0 (yellows);
mount in 1.68
AMOSITE (120)
Yellow (II length)
magentas and blues
(1 length), (+)
elongation
CROCIDOLITE (123)
Yellow ( II length) ,
golden yellow (1 length),
(-) elongation; pleochroic:
gray-blue (1) and blue (ll)
with one polar and no stops
NON-FIBROUS
Xg 700 nm)
(pale blues)
GYPSUM (151)
Low birefringence,
often tabular with
oblique extinction
XQ Colors
in visible
QUARTZ (183)
Glassy flakes,
in (blue) , c
(blue-magenta)
LIZARDITE (710)
Lamellar aggre-
gates, undulose
extinction, blues
and magentas
ANTIGORITE (117)
Yellow (H) to
golden magenta
(1) rods
VERMICULITE (207)
Very thin sheets.
nearly isotropic,
XQ'S in yellow.
turned up edges
usually give blue
crosswise, yellow
lengthwise but n's
vary
XQ'S < 400 (•,))
(pale yellows, white)
CALC1TE (133)
Very high bire-
fringence
DOLOMITE (140)
Like calcite,
u> = 1.679
MAGNESITE (164)
Like calcite,
a. - 1.694
TALC
Lamellar aggre-
gates, pale
yellows, plate
view; blue (1
plate)
a Particle Atlas Numbers, Vols II and III.
ALL IIISri'.RSION COLORS HIVEN ARF. FOR TIIF, CENTRAL STOP
Figure D-l. Procedure for analysis of asbestos materials.
-------
REFERENCES
McCrone WC. 1980. The asbestos particle atlas. Ann Arbor, MI: Ann Arbor
Science, 122 pp.
USEPA. 1980. U.S. Environmental Protection Agency. Office of Toxic
Substances. Asbestos-containing materials in school buildings: guids
asbestos analytical programs. Washington, DC: USEPA. EPA 560/13-80-
PB81-24358 6.
158
-------
APPENDIX E
ANALYTICAL PROTOCOL FOR AIR SAMPLES
159
-------
ANALYTICAL PROTOCOL FOR AIR SAMPLES
Air samples will be analyzed by transmission electron microscopy (USEPA
1980). Select one filter from each box of 24 0.45-pm, 47-mm Millipore HA mem-
brane filters to serve as a laboratory blank. Use all filters from the same
production lot number, if possible. Prior to field sampling, determine if
the laboratory blank filters are asbestos free by ashing followed by trans-
mission electron microscope examination. Record filter box and lot number.
Upon receipt of filters from the sampling team, record in a laboratory
logbook the sample numbers, date they were received, and any macroscopic iden-
tifying characteristics of particular filter samples. This includes damaged
or smudged areas on the filter surface, lack of uniform sample deposition,
unattached particulate or debris, unusually heavy-appearing deposit concentra-
tion, etc.
Measure precisely the diameter of the effective filter area. Mount on
glass slides with double-sided adhesive any damaged areas removed prior to
sample preparation and carefully measure. The total effective filter area
and damaged areas of sample removed should be accurately recorded for purposes
of calculation procedures.
In the original sample dish cut a 90° radial section of the original
47-mm filter sample with a clean, single-edged razor blade. Transfer the
quarter section with stainless steel forceps to a clean, 1x3 in. glass slide,
and cut again into smaller wedges to fit into the glass ashing tube (approxi-
mately 15 mm diameter x 150 mm long). Transfer the wedges by forceps to clean,
numbered ashing tubes. Place the tubes in an LFE 504 low temperature plasma
oven, one sample tube and one laboratory control tube per ashing chamber.
The laboratory control tube may either contain a blank Millipore filter or be
run as an empty tube. Maintain the ashing process at 450 watts for 2 hr.
Upon removal from the oven, treat the ashing tubes as follows. Place
the tube in an ultrasonification bath. Pour 1 to 2 ml of 0.22-|jm filtered
Millipore-Q water into the tube from a clean 100-ml graduated cylinder. Soni-
cate (at 40 milliamperes) the sample vigorously for approximately 5 min and
transfer it to a clean 150-ml glass beaker. Rinse the tube by additional
ultrasonification two or three times more using a few milliters of filtered
water each time, and transfer the contents to a 150-ml sample beaker. Add
the remaining volume (up to 100 ml) of filtered water and sonicate again the
entire suspended sample or blank, so that the total time of dispersion in the
sonicator takes at least 20 min. Use a clean glass rod to stir the suspended
sample while it is being sonicated.
Divide the 100-ml fraction into three aliquots: 10, 20, and 70 ml, pre-
pared in that order. Using a 25-mm Millipore filter apparatus, place a 0.2-(Jm
Nuclepore polycarbonate filter on top of an 8.0-[Jm mixed cellulose ester Milli-
pore backup filter. Wet the filters by aspirating approximately 10 ml of fil-
tered deionized water. Stop aspiration, pour in the first sample aliquot or
portion thereof, and begin the aspiration procedure again. Carefully add the
remaining sample volume without disturbing the flow across the Nuclepore fil-
ter surface. The suspended sample may be resonicated or stirred between fil-
tration of the aliquots.
160
-------
When the sample is deposited, carefully transfer the Nuclepore filter to
a clean, labeled (sample number, date, and aliquot size) 1x3 in. glass slide
Discard the Millipore backup filter.
When dry, attach the 0.2-|Jm Nuclepore filter tautly to the slide with
transparent tape. Coat the filter with an approximately 40-nm-thick carbon
film (National Spectroscopic Laboratories carbon rods) by vacuum evaporation.
The film thickness need be sufficient only to provide support for the deposit
sample.
Transfer the polycarbonate filter deposit to a 200-mesh electron micro-
scope copper grid (E. F. Fullam) by first cutting a 3-ml-square portion from
the filter using a clean, single-edged razor blade. Place this deposit side
down on the electron microscope (EM) grid which, in turn, has been set upon a
small, correspondingly labeled portion of lens tissue paper. Place the film,
grid, and lens paper on a Jaffe dish consisting of a copper screen supported
on a bent glass rod in a covered 90-mm glass petri dish. Pour reagent grade
chloroform (J. T. Baker Company) into the dish to saturate the lens paper
without submersing the grid and sample. Keep the dish covered at room tem-
perature for 2 hr. Shift the prepared sample to a clean petri dish with fresh
chloroform. Heat to 40°C for 10 min to provide a washing procedure.
While it is still wet, place the sample grid in a small gelatin capsule.
Tape the capsule to the slide that has the remaining coated polycarbonate fil-
ter, and store until analysis.
Starting with the 70-ml fraction filter, examine the EM grid under low
magnification in the transmission electron microscope to determine its suit-
ability for high magnification examination. Ascertain that the loading is
suitable and is uniform, that a high number of grid openings have their carbon
film intact, and that the sample is not contaminated excessively with extrane-
ous debris or bacteria.
Scan the EM grid at a screen magnification of 20,OOOX. Record the length
and breadth of all fibers that have an aspect ratio of greater than 3:1 and
have substantially parallel sides. Observe the morphology of each fiber
through the 10X binoculars and note whether a tubular structure characteristic
of chrysotile asbestos is present. Switch into selective area electron dif-
fraction (SAED) mode and observe the diffraction pattern. Note whether the
pattern is typical of chrysotile or amphibole, ambiguous, or neither chryso-
tile nor amphibole. Use energy dispersive X-ray analysis where necessary to
further characterize the fiber. Take pictures as desired representing the
sample type, fiber/particulate distribution, or characteristic SAED patterns
of chrysotile and specific amphibole types.
Count the fibers in the grid openings until at least 100 fibers, or the
fibers in a maximum of 10 grid openings, have been counted. Once counting'of
fibers in a grid opening has started, the count shall be continued though the
total count of fibers may be greater than 100.
To ensure uniformity of grid opening dimensions, examine several 200-mesh
grids by optical microscopy and measure roughly 10 openings per grid. Average
these dimensions to provide a standard grid opening area.
161
-------
Calculate from the following equation fiber number concentration ex-
pressed as the total number of fibers/original filter.
Fibers/m3 = number of fibers counted [area factor*][Nuclepore dilution
factor x Millipore section dilution factor] 1
volume, m3
Calculate fiber mass for each type of asbestos in the sample by assuming
that the breadth measurement is a diameter; thus, the mass can be calculated
from
Mass (|Jg) = - . (length, |Jm) . (diameter, |Jm)2 . (density, g/cm3) . 10
4
-6
The density of chrysotile is assumed to be 2.6 g/cm3, and of amphibole,
3.0 g/cm3. The mass concentration for each type of asbestos is then calcu-
lated from
Mass Concentration Total Mass of All
((jg/m3) of a = Fibers of that Type (|Jg)
Particular Type Volume of Air Sampled (m3)
Record the fiber bundles and clusters as such, but do not include them
in the mass calculation or the fiber count. Three reasons for not counting
the fiber clusters and fiber bundles in the mass calculation are as follows:
(1) it is difficult to assign the third dimension to the two-dimensional ob-
servation of the aggregates; (2) it is difficult to determine void space
within bundles and clusters; and (3) since the bundles and clusters make up
about 2% of the item count, one cannot be certain of the even distribution
throughout the filter. The calculation assumes even distribution because to
go from the counted area to filter area is a factor of about 106 depending on
how many grid openings are actually counted.
REFERENCE
USEPA. 1978. U.S. Environmental Protection Agency. Office of Research and
Development. Electron microscope measurement of airborne asbestos concentra-
tions: a provisional methodology manual. Research Triangle Park, NC: USEPA
EPA 600/2-77-178.
Total Effective Filter Area, cm2
Average Area of an EM Grid Opening, crrr'
162
-------
APPENDIX F
FRIABLE MATERIAL SITE-SPECIFIC RATING FORMS
163
-------
RATER'S NAME: DATE!
Specific Site Cover Sheet
Full Name of School:
Location of Site:
Circle one: a Site b Indoor Control
1. Room Dimensions:
a. length: feet
b. width: feet
c. height: feet
CIRCLE ALL THAT APPLY:
2. Where is the suspicious material located?
a on an exposed ceiling
b above a suspended ceiling
c on the walls
CIRCLE ONE ANSWER FOR EACH OF THE FOLLOWING QUESTIONS:
YES NO 3. Is the room carpeted?
YES NO 4. Is the floor tile or linoleum?
YES NO 5. Is the floor wood?
YES NO 6. Is a suspended ceiling present?
7. Circle all types of cleaning practices that apply to the area
of the site and the frequency for each.
a Sweeping 1 never 2 monthly 3 weekly 4 daily
b Wet Mopping 1 never 2 monthly 3 weekly 4 daily
c Dry Mopping 1 never 2 monthly 3 weekly 4 daily
d Vacuuming 1 never 2 monthly 3 weekly 4 daily
164
-------
RATER'S I1AME: DATE:
Circle: Site or Control
ALGORITHM SCORING SHEET
SCHOOL NAME:
Room (number):
1. Condition (CLEARLY CIRCLE ONE SCORE FOR EACH FACTOR)
0 no damage
2 moderate damage
5 severe damage
2. Accessibility
0 not accessible
1 rarely accessible
3 accessible
3. Part of Air Moving System (Plenum)
0 no
1 yes
4. Exposure
0 material is not exposed
1 10% or less of the material is exposed
4 greater than 10% of the material is exposed
5. Water Damage
0 no water damage
1 minor water damage
2 moderate or major water damage
6. Activity or Movement
0 none or low activity level
1 moderate activity level
2 high activity level
7. Friability
0 not friable
1 low friability
2 moderate friability
3 high friability
8. Percentage Asbestos
0 1% or less
2 greater than 1% and less than or equal to 50%
3 greater than 50%
165
-------
RATER'S NAME:
SCHOOL NAME: _
Room (number):
DATE:
QUESTIONNAIRE; Circle the appropriate response to each question.
YES NO 1. Is there any evidence of "fallout" by
accumulation of debris on any horizontal
surfaces?
YES NO 2. Is the material covered with paint or other type
of protective or decorative coating?
YES NO 3. Is the material subject to vibration due to
machinery, traffic, airplanes, etc?
Is this material damaged or deteriorating?
Is there evidence that water has degraded the
integrity of the material?
Is the coating material exposed?
Is this material accessible?
Is there normally activity in (or around) the
inspection area that could cause deterioration
of the coating material?
YES NO 9. Is the material located in an air plenum or air
stream?
YES NO 10. Is the coating material friable?
YES
YES
YES
YES
YES
NO
NO
NO
NO
NO
4.
5.
6.
7.
8.
For the next section, use a scale of A, B, C, D, to rate the
degree to which each factor applies. A is used to indicate the
lowest rating for any factor while D is used to indicate the
highest rating possible. Circle the letter which is applicable
for each.
A B C D 11. Material condition (this factor includes
deterioration due to water damage and other
factors such as vandalism which have
influenced the integrity of the material).
A=excellent condition
A B C D 12. Exposed surface area. A=not exposed.
A B C D 13. How accessible is the material? A=not
accessible.
B C
A B
14. What is the normal level of activity in the
inspection area? A=no activity.
15. How friable is the material? A=not friable.
166
-------
APPENDIX G
CLASSIFICATION ANALYSES BASED UPON THE 50TH PERCENTILE
(84 ng/m3) OF THE AIRBORNE CHRYSOTILE DISTRIBUTION
167
-------
Table G-l. Predicting Low/High Airborne Chrysotile Concentration
with Three Regression-based Dichotomies
Distribution of the 48 Asbestos-containing Friable Material Sites:
Observed Airborne Chrysotile Concentration Versus Predicted
Observed airborne
chrysotile
concentration (ng/m3)
Low (g 68. 6b)
High (> 68.6)
Total
rt
Predicted by seven algorithm factors
(Regression model
Low
21
18
39
I, Table 31)
High
3
6
9
Total
24
24
48
% Correct = 27/48 = 56.2%
Sensitivity = 6/24 = 25.0%
Observed airborne
chrysotile
concentration (ng/m3)
Predicted by six algorithm .factors
and releasability
(Regression model II, Table 31)
Low
High
% Correct = 30/48 = 62.5%
Sensitivity = 11/24 = 45.8°/|
Total
Low (g 68.6)
High (> 68.6)
Total
19
13
32
5
11
16
24
24
48
Observed airborne
chrysotile
concentration (ng/m3)
Predicted by releasability, room
volume, and water damage
(Unweighted regression model, Table 34)
Low High
Total
Low (^ 68.6)
High (> 68.6)
Total
18
U
29
19
24
24
48
% Correct = 31/48 = 64.6%
Sensitivity = 13/24 = 54.2%
Of the eight algorithm factors, exposure was not included because of zero
variability. Bulk sample chrysotile content was included as the actual
percentage instead of the original algorithm categorization.
68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
Of the eight algorithm factors, exposure was not included because of zero
variability, and friability was replaced by releasability category. Bulk
sample chrysotile content was included as the actual percentage instead of
the original algorithm categorization.
Releasability category.
168
-------
Table G-2. Predicting Low/High Airborne Chrysotile Concentration with the
Original Decision Tree Based upon Proportion of Material Exposed,
Material Condition, Accessibility, and Friability
(0,0)a
0,lb
Exposure
i (24,24;
Condition
1,2 | (24,24;
Lo-« (21'23)
0,1
Accessibility
(3,1)
(0,0)
(0,0)
0,1
Friability
(3,1)
2,3
High
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
Chrysotile
concentration (ng/m3)
Low (£ 68. 6C)
High (> 68.6)
Total
Predicted by
Low
21
23
44
% Correct = 22/48 = 45.
Sensitivity = 1/24 = 4.
tree
High
3
1.
4
8%
2%
Total
24
24
48
a Number of sites with low airborne chrysotile concentration (^ 68.6 ng/m ),
number of sites with high airborne chrysotile concentration (> 68.6 ng/m )
b Algorithm factor codes.
c 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
169
-------
Table G-3. Predicting Low/High Airborne Chrysotile Concentration with
a Decision Tree Based upon Releasability and Water Damage
I r»i»r "* - -
(12, 7)°
,b
Releasability
(3,9) r
3
Low
(9,8)
High
« (6,4)
Oc
Water Damage
(3,4)
1,2
High
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
chrysotile
concentration (ng/m3)
Predicted by tree
Low
High
% Correct = 31/48 = 64.6%
Sensitivity = 13/24 = 54.2%
Total
Low (
High
Total
S 68. 6d)
(> 68.6)
18
11.
29
6
13
19
24
24
48
a Number of sites with low airborne chrysotile concentration (^ 68.6 ng/m3),
number of sites with high airborne chrysotile concentration (> 68.6 ng/m3)
b Releasability codes: 1 = ranks 1-4, 2 = ranks 5-6, 3 = ranks 7-9.
c Water damage codes.
d 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
170
-------
Table G-4. Predicting Low/High Airborne Chrysotile Concentration with
a Decision Tree Based upon Releasability and Air Stream Status
(12,7)°
lb
Releasability
(3,9)
3
I
(9,8)
(4,3)
Oc
Air Stream Status
(5,5)
1
hi Hlnl-i
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
chrysotile
concentration (ng/m3)
Predicted by tree
Low
High
Total
Low (^
High (>
Total
68. 6d)
68.6)
16
10
26
8
14
22
24
24
48
% Correct = 30/48 = 62.5%
Sensitivity = 14/24 = 58.3%
a Number of sites with low airborne chrysotile concentration (^ 68.6 ng/m3),
number of sites with high airborne chrysotile concentration (> 68.6 ng/m3)
b Releasability codes: 1 = ranks 1-4, 2 = ranks 5-6, 3 = ranks 7-9.
c Air stream status codes.
d 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
171
-------
Table G-5. Predicting Low/High Airborne Chrysotile Concentration with
a Decision Tree Based upon Releasability and Bulk Sample
Chrysotile Content
(12,7)°
lb
Releasability
(3,9)
3
High
(9,8)
(8,6)
<20%
Bulk Sample Chrysotile Content
(1.2)
>20%
High
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
chrysotile
concentration (ng/m3)
Predicted by tree
Low
High
Total
Low (g 68.6)
High (> 68.6)
Total
20
13
33
4
11.
15
24
24
48
% Correct = 31/48 = 64.6%
Sensitivity = 11/24 = 45.8°/!
a Number of sites with low airborne chrysotile concentration (^ 68.6 ng/m3),
number of sites with high airborne chrysotile concentration (> 68.6 ng/m3).
b Releasability codes: 1 = ranks 1-4, 2 = ranks 5-6, 3 = ranks 7-9.
c 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
172
-------
Table G-6. Predicting Low/High Airborne Chrysotile Concentration with
a Decision Tree Based upon Releasability, Water Damage (or
Material Condition), and Air Stream Status
(12,7)0
lb
(6,6)
WD=0C
(MC=0)
(2,1)
0
Releasability (3,9)
/ _ ^ (((gl,
2 | (9,8)
Water Damage
or
(Material
Condition )
WD = 1,2
(MC=2,5) (*>2>
^ >
Air Stream Stnhis * L fe 1 IT ,-.!-,
Distribution of the 48 Asbestos-containing Friable Material Sites:
Low/High Airborne Chrysotile Concentration - Observed Versus Predicted
by Decision Tree
Observed airborne
chrysotile
concentration (ng/m3)
Low (^ 68. 6d)
High (> 68.6)
Total
Predicted by
Low
20
14
34
% Correct = 30/48 =
Sensitivity = 10/24 =
tree
High
4
10
14
62.5%
41.7%
Total
24
24
48
a Number of sites with low airborne chrysotile concentration (£ 68.6 ng/m ),
number of sites with high airborne chrysotile concentration (> 68.6 ng/m3)
b Releasability codes: 1 = ranks 1-4, 2 = ranks 5-6, 3 = ranks 7-9.
c Water damage and (material condition) codes.
d 68.6 ng/m3 = 50th percentile for the 48 asbestos-containing friable
material sites.
173
-------
50272-101
REPORT DOCUMENTATION 11-_REPORT NO.
PAGE J EPA 560/5-83-003
4. Title and Subtitle
Airborne Asbestos Levels in Schools
7. Author(s) Paul C. Constant,3 Fred Bergman,3 Donna Rose,3 "Caylord
Atkinson,3 Donna Wattsxb__Everett__L^ggue_t_b Ty_ Hartwell,0 Bertram
9. Performing Organization Name and Address Price C
a. Midwest Research Institute, 425 Volker Blvd. Kansas City,
MO 64110
b. Research Triangle Institute, Research Triangle Park,
NC 27709
c. Battelle Columbus Laboratories, Columbus, OH 43201
12. Sponsoring Organization Name and Address
Environmental Protection Agency
Office of Toxic Substances
Exposure Evaluation Division
401 M Street. SW, Washington. D.C. 10460,
3. Recipient's Accession No.
5. Report Date
_June 1983
6.
8. Performing Organi7ation Rept. No.
4901-A(41)
10. Proiect/Task/Work Unit No.
11. Contract(C) or Grant(G) No.
(0 68-01-5915 and
(G) 68-01-5848
13. Type of Report & Period Covered
Task Final
April 81 - April 8_2...
14.
IS. Supplementary Notes
is. Abstract (Limit: 200 words) Air (116) and bulk (192) samples were collected from 58" units at 25
different schools of an urban independent school district. These were analyzed respec-
tively by transmission electron microscopy and polarized light microscopy techniques for
asbestos fiber concentrations. The new factor of releasability (of fibers) rating re-
sulted from the bulk fiber analysis. Each sampling site was rated by a special five-
person team for assessment tools: algorithm, condition, accessibility, part of air moving
system, material exposure, friability and water damage, and people's activity at the site.
The results were statistically analyzed to document potential exposure to airborne asbes-
tos resulting from the friable, asbestos-containing material in schools and to develop an
exposure assessment tool that would be based on the above stated factors. The principal
conclusions related to the first study objective are: (1) airborne asbestos levels inside
school buildings with asbestos are significantly higher than outdoor ambient levels due to
the release of asbestos fibers from asbestos-containing materials inside those buildings;
(2) within a school building, asbestos fibers are transported from rooms having asbestos-
containing materials to rooms without these materials. The principal conclusions related
to the second study objective are: (1) the existing algorithm is not a valid predictor of"
exposure to airborne asbestos levels; (2) the amount of asbestos in the bulk material is
not a valid predictor of exposure to airborne asbestos levels; (3) the releasability rat-
ing system developed in this study is related to levels of airborne asbestos. (Additional
studies are underway.) .. ..._
17. Document Analysis a. Descriptors
Asbestos, asbestos sampling, asbestos analysis, asbestos algorithm, asbestos releas-
ability, survey design, TEM, PLM
b. Identifiers/Open-Ended Terms
c. COSATI Field/Group
18. Availability Statement
19. Security Class (This Report)
Unclassified
20. Security Class (This Page)
11
21. No. of Pages
178
22. Price
(See ANSI-Z39.18)
See Instructions on Reverse
OPTIONAL FORM 272 (4-77)
(Formerly NTIS-35)
Department of Commerce
t U . S . GOVERNMENT PRINTING OFFICEi 1984-421-545/11833
------- |