-------
TABLE OF CONTENTS
(Continued)
Page
Table 6-4. Predicted Effect of a 50% Decrease in Environmental-Lead Loadings
(//g/ft2) and Concentrations (/jg/g) Based on the Environmental-Lead
Pathways SEM for the CAP Study Data 53
Table 6-5. Predicted Effect of a 50% Decrease in EnvironmentaJ-Lead Loadings
(//g/ft2) and Concentrations (/jg/g) Based on the Environmental-Lead
Pathways SEM for the R&M Data 54
Table 6-6. Predicted Effect of a 50% Decrease in Environmental-Lead Loadings
(//g/ft2) and Concentrations (fjg/g) Based on the Environmental-Lead
Pathways SEM for the Rochester Data 55
Table 6-7. Structural Equation Modeling Results for the R&M Blood-Lead Pathways
Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (//g/g) 56
Table 6-8. Structural Equation Modeling Results for the Rochester Blood-Lead
Pathways Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations
(//g/g) 57
Table 6-9. Predicted Effect of 50% Decrease in Environmental-Lead Loadings
(//g/ft2) and Concentrations
-------
TABLE OF CONTENTS
(Continued)
Table B-4. Structural Equation Modeling Results for the R&M Environmental
Pathways, Including the R&R Exposure Pathways Dust-Lead
Loadings (//g/ft2) and Concentrations (//g/g) B-5
Table B-5. Structural Equation Modeling Results for the R&M Blood-Lead
Pathways, Including a Proportion of Carpeting and Renovation and
Remodeling Pathway Dust-Lead Loadings (//g/ft2) and Dust-Lead
Concentrations (//g/g) B-6
Table B-6. Predicted Effect of 50% Decrease in Environmental-Lead Loadings
(//g/ft2) and Concentrations (//g/g) and Blood-Lead Concentrations
(//g/dL) Based on the Blood-Lead Pathways SEM for the R&M Data,
Including Proportion of Carpet B-7
Table B-7a. R&M Data Correlation of Log-Transformed Blood-Lead Concentrations
(//g/dL), Dust-Lead Loadings (//g/ft2), and Water-Lead Concentrations
(//g/L) B-8
Table B-7b. Correlation of Log-Transformed Blood-Lead (//g/dL), Dust-Lead (//g/g),
and Water-Lead (//g/L) Concentrations B-9
Table C-1. Structural Equation Modeling Results for the Rochester Environmental-
Lead Pathways, Including the Window Paint and Door Paint Hazard
Score Pathways Dust-Lead Loadings (//g/ft2) and Dust-Lead
Concentrations (//g/g) C-2
Table C-2. Structural Equation Modeling Results for the Rochester Environmental
Pathways Model, Including Window and Door Paint XRF Measurement
Pathways Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations
(//g/g) C-3
Table C-3. Structural Equation Modeling Results for the Rochester Blood-Lead
Pathways, Including the Window Paint and Door Paint Hazard Score
Pathways Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations
(//g/g) C-4
Table C-4. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways
Model, Including Window and Door Paint XRF Measurement Pathways Dust-
Lead Loadings (//g/ft2) and Dust-Lead Concentrations (//g/g) C-5
Table C-5. Structural Equation Modeling Results for the Rochester Blood-Lead
Pathways, Including Hand Lead and Window and Door Paint Pathways
Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (//g/g) C-6
Table C-6. Structural Equation Modeling Results for the Rochester Blood-Lead
Pathways Model, Including Soil Coverage and Window and Door Paint
Pathways Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations
U/g/g) C-7
Table C-7. Structural Equation Modeling Results for the Rochester Blood-Lead
Pathways, Including Proportion of Carpeting and Indicator of Interior
Entryway Carpet Pathways Dust-Lead Loadings (//g/ft2) and Dust-
Lead Concentrations (//g/g) C-8
Table C-8. Structural Equation Modeling Results for the Rochester Blood-Lead
Pathways for Homes with Carpeted Bedrooms and Play Areas Dust-
Lead Loadings (//g/ft2) and Dust-Lead Concentrations (//g/g) C-9
VIII
-------
TABLE OF CONTENTS
(Continued)
Page
Table C-9. Predicted Effect of 50% Decrease in Environmental-Lead Loadings
U/g/ft2) and Concentrations (/ug/g) and Blood-Lead Concentrations
(A/g/dL) Based on the Blood-Lead Pathways SEM for the Rochester
Data, Including Carpet and Paint Pathways C-10
Table C-10. Structural Equation Modeling Results for the Rochester Blood-Lead
Pathways Model For African American Children Dust-Lead Loadings
(//g/ft2) and Dust-Lead Concentrations (//g/g) C-11
Table C-11. Structural Equation Modeling Results for the Rochester Blood-Lead
Pathways Model For Children of All Other Races Dust-Lead Loadings
U/g/ft2) and Dust-Lead Concentrations (ug/g) C-12
Table C-12a. Rochester Data Correlation of Log-Transformed Blood-Lead (ug/dL), Dust-Lead
Loadings U/g/ft2), Soil-Lead (ug/g) Concentrations, and Child and Housing
Characteristic Variables C-13
Table C-12b. Rochester Data Correlation of Log-Transformed Blood-Lead (//g/dL),
Dust-Lead (ug/g), and Soil-Lead (//g/g) Concentrations, and Child and
Housing Characteristic Variables. C-14
Table D-1. Description of Variables for Pathway Model from Exterior Surface Dust
Lead, Interior House Dust Lead and Childhood Lead Exposure in an
Urban Environment, by Bornschein, et al [2] D-2
Table D-2. Description of Variables for Pathway Model from Soil-Lead - Blood
Lead Relationship in a Former Lead Mining Town, by Bornschein, et al
[17] D-2
Table D-3. Description of Variables for Pathway Model from The Influences of
Social and Environmental Factors on Dust Lead, Hand Lead, and Blood
Lead Levels in Young Children, by Bornschein, et al 11] D-3
Table D-4. Description of Variables for Pathway Model from Pathways of Lead
Contamination for the Brigham and Women's Hospital Longitudinal
Study, by Menton, et al [7] D-3
Table D-5. Description of Variables for Pathway Model from Dust Lead
Contribution to Lead in Children, by Sayre[18] D-4
Table D-6. Description of Variables for Pathway Model from Pathways of Lead
Exposure in Urban Children, by Lanphear [19] D-5
LIST OF FIGURES
Figure 1-1. Pathways Diagram Assessing the Role of Floor Dust Lead as a Route
for Lead Dust from the Window Sill to the Child's Blood 2
Figure 5-1. General Pathway Diagram Based on Literature Review 34
Figure 6-1. Relationship of Blood-Lead Concentration (//g/dL) and Average Floor
Dust-Lead Loading (//g/ft2) for the R&M Data 45
Figure 6-2. Relationship of Blood-Lead Concentration (//g/dL) and Average Floor
Dust-Lead Concentration (ug/g) for the R&M Data 46
Figure 6-3. Significant Pathways for the CAP, R&M, and Rochester Environmental-Lead
Pathway Models Dust-Lead Loadings
-------
TABLE OF CONTENTS
(Continued)
Page
Figure 6-5. Significant Pathways for the R&M and Rochester Blood-Lead Pathway
Models Dust-Lead Loadings U/g/ft2) 58
Figure 6-6. Significant Pathways for the R&M and Rochester Blood-Lead Pathway
Models Dust-Lead Concentrations Ot/g/g) ^
Figure 6-7. Significant Pathways for the R&M and Rochester Environmental-Lead
Pathway Models Including Window and Door Paint Hazard Score
Pathways Dust-Lead Loadings (//g/ft2) 64
Figure 6-8. Significant Pathways for the R&M and Rochester Environmental-Lead
Pathway Models Including Window and Door Paint Hazard Score
Pathways Dust-Lead Concentrations (//g/g) 65
Figure 6-9. Significant Pathways for the Rochester Blood-Lead Pathway Model
Including Window and Door Paint Hazard Score Pathways Dust-Lead
Loadings (//g/ft2) and Concentrations U/g/g) 66
Figure 6-10. Significant Pathways for the R&M Blood-Lead Pathway Model Including
a Hand Lead and Window and Door Paint Pathways Dust-Lead
Loadings (//g/ft2) and Concentrations U/g/g) 67
Figure 6-11. Significant Pathways for the R&M Blood-Lead Pathway Model Including
a Proportion of Carpeting Pathway Dust-Lead Loadings (//g/ft2) and
Concentrations (//g/g) 69
Figure 6-12. Significant Pathways for the Rochester Blood-Lead Pathway Models
Including Proportion of Carpeting and Indicator of Interior Entryway
Carpet Pathways Dust-Lead Loadings (//g/ft2) and Concentrations
(//g/g) 71
Figure 6-13. Significant Pathways for the Rochester Blood-Lead Pathway Model For
Homes with Carpeted Bedrooms and Play Areas Dust-Lead Loadings
(//g/ft2) and Concentrations (//g/g) 72
Figure 6-14. Significant Pathways for the Rochester Blood-Lead Pathway Models For
African American and Other Race Children - Dust-Lead Loadings
(//g/ft2) 75
Figure 6-15. Significant Pathways for the Rochester Blood-Lead Pathway Models For
African American and Other Race Children Dust-Lead Concentrations
(//g/g) 76
Figure D-1. Statistically Significant Pathways from Exterior Surface Dust Lead,
Interior House Dust Lead and Childhood Lead Exposure in an Urban
Environment, by Bornschein, et al. [2] D-6
Figure D-2. Statistically Significant Pathways from Soil-Lead - Blood Lead
Relationship in a Former Lead Mining Town, by Bornschein, et al. [17] D-6
Figure D-3. Statistically Significant Pathways from The Influences of Social and
Environmental Factors on Dust Lead, Hand Lead, and Blood Lead Levels
in Young Children, by Bornschein, et al. [1] D-7
Figure D-4. Statistically Significant Pathways from Pathways of Lead
Contamination for the Brigham and Women's Hospital Longitudinal
Study, by Menton, et al. [7] D-7
Figure D-5. Statistically Significant Pathways from Dust Lead Contribution to Lead
in Children, by Sayre [18] D-8
Figure D-6. Statistically Significant Pathways from "Pathways of Lead Exposure in
Urban Children," by Lanphear et al. [19] D-8
-------
TABLE OF CONTENTS
(Continued)
LIST OF EQUATIONS
Equation Set 5-1. CAP Environmental-Lead Pathways Model 36
Equation Set 5-2. R&M Environmental-Lead Pathways Model 36
Equation Set 5-3. Rochester Environmental-Lead Pathways Model 36
Equation Set 5-4. R&M Blood-Lead Pathways Model 37
Equation Set 5-5. Rochester Blood-Lead Pathways Model 37
Equation Set 5-6. R&M Environmental-Lead Pathways Model - Assessing the
Impact of Paint-Lead Pathways 38
Equation Set 5-7. Rochester Environmental-Lead Pathways Model -
Assessing the Impact of Paint-Lead Pathways 38
Equation Set 5-8. Rochester Blood-Lead Pathways Model - Assessing the
Impact of Paint-Lead Pathways 38
Equation Set 5-9. Rochester Blood-Lead Pathways Model - Assessing the
Impact of Hand Dust-Lead 39
Equation Set 5-10. R&M Blood-Lead Pathways Model - Assessing the Impact
of a Carpeted Floors Pathway 40
Equation Set 5-11. Rochester Blood-Lead Pathways Model - Assessing the
Impact of a Carpeted Floors Pathway 41
Equation Set 5-12. Rochester Blood-Lead Pathways Model - Assessing the
Impact of Carpeted Bedroom and Play Area Floors 41
Equation Set 5-13. CAP Environmental-Lead Pathways Model - Assessing the Impact
of Recent Renovation and Remodeling Activities 42
Equation Set 5-14. R&M Environmental-Lead Pathways Model - Assessing the Impact
of Recent Renovation and Remodeling Activities 42
Equation Set 5-15. R&M Blood-Lead Pathways Model - 42
Equation Set 5-16. Rochester Blood-Lead Pathways Model - Assessing
Pathways for Different Races 43
Equation Set 5-17. CAP Study Environmental-Lead Pathways Model-Assessing
the Effect of an Air Duct Dust-Lead Pathway 43
XI
-------
This page intentionally left blank
XII
-------
EXECUTIVE SUMMARY
Despite reductions of the amount of lead in various sources over the past few decades,
elevated blood-lead concentrations in children continue to be a public health concern. Using data
from three environmental lead studies (the Comprehensive Abatement Performance (CAP) Study
[9,10,11], the Baltimore Repair and Maintenance (R&M) study [12], and the Rochester Lead-
in-Dust study [13]), the relationships among environmental lead levels and children's blood lead
concentrations were examined using structural equation models (SEM). SEM was used for this
analysis because the method accounts for the covariance among variables that allow direct and,
hi particular, indirect effects of various sources of lead to be assessed. Traditional multiple
regression only assesses the direct effects.
The primary analysis of this report focused on two types of structural equation models:
environmental-lead pathway models and blood-lead pathway models. The environmental-lead
pathway models were structured to assess the direct and indirect impact of several environmental
variables such as soil lead, window sill dust-lead, and window well dust-lead on floor dust-lead
so that comparisons could be made across the three studies (CAP, R&M, and Rochester). The
blood-lead pathway models were structured to focus on the direct and indirect impact of soil
lead, paint-lead, window sill dust-lead, window well dust-lead, and floor dust-lead on childhood
blood-lead concentration and to allow comparisons between R&M and Rochester. (The CAP
Study did not collect blood-lead data and therefore could not be included in the blood-lead
pathway model analysis). Because the three studies were not designed or conducted in the exact
same manner, there was some information collected in one study that was not collected in the
others. To take advantage of having additional information in the studies, several secondary sub-
analyses were performed. The sub-analyses focused on specific pathways of lead exposure
within either the environmental-lead pathway models or the blood-lead pathway models. Note
that for all analyses separate models were fit using dust-lead loading vacuum samples and dust-
lead concentration vacuum samples.
In the environmental-lead pathways analysis, three statistically significant direct
pathways of lead contamination were found for all three studies: 1) window well dust-lead
loading to window sill dust-lead loading, 2) window well dust-lead concentration to window sill
dust-lead concentration, and 3) exterior entryway dust-lead concentration to interior entryway
XIII
-------
dust-lead concentration. There were several indirect pathways which were the same for two out
of the three studies: 1) soil-lead concentration to window well dust-lead loading to window sill
dust-lead loading, 2) soil-lead concentration to window well dust-lead concentration to window
sill dust-lead concentration, 3) window well dust-lead loading to. window sill dust-lead loading to
the interior entryway dust-lead loading or the floor dust-lead loading, and 4) window well dust-
lead concentration to window sill dust-lead concentration to floor dust-lead concentration.
Despite the different study designs and dust vacuum collection methods, the results are quite
similar across the three studies.
The blood-lead pathways models fitted to the two data sets with blood lead measurements
(R&M and Rochester) had fewer consistent, statistically significant pathways across the studies.
In the lead loading models consistent pathways included direct pathways of lead exposure from
1) floor dust-lead and 2) children's mouthing habits to the blood, and indirect pathways from
1) interior entryway dust-lead, 2) dust on window wells and 3) dust on window sills to blood.
However, no consistent significant pathways to blood were found in the lead concentration
models. In the R&M concentration model, significant pathways included a direct pathway from
hiterior entryway dust-lead concentration to blood, and indirect pathways of exposure to the
blood from exterior entryway dust, window well dust, and the window sill dust-lead
concentration. In the Rochester concentration model, the only statistically significant pathway to
blood was a direct pathway from children's mouthing behavior to blood. There were no
statistically significant indirect pathways to blood in the Rochester concentration model.
An analysis of the Rochester data included a statistically significant hand dust-lead
pathway hi the blood-lead pathways model. Hand dust-lead was a statistically significant
pathway to blood-lead hi other studies analyzed by structural equations modeling. An analysis of
the Rochester data indicated that hand dust-lead was one of the direct pathways of lead exposure
to a child's blood, with environmental pathways directly and indirectly to the child's hand dust-
lead.
XIV
-------
1.0 INTRODUCTION
Although Federal and state regulatory agencies have succeeded in reducing lead in paint,
food, and ambient air, approximately one million children in the United States, ages one to five
years, have a blood lead concentration equal to or above the CDC threshold of 10 fig/dL [22].
The purpose of the work presented in this document is to understand the sources and
routes by which children are exposed to lead in their residences so that the optimal prevention
and remediation action can be carried out. A pathways analysis using structural equation models
(SEM) was performed to assess the different pathways by which a child may be exposed to lead.
Data from three studies, the Comprehensive Abatement Performance Study (CAP Study) [9,10,
11], the Baltimore Lead-Based Paint Abatement and Repair and Maintenance Study (R&M) [12],
and the Rochester Lead-in-Dust Study (Rochester) [13] were used to assess the pathway models.
Structural equation modeling was used for this analysis because the method takes into account
the covariance among the environmental and blood-lead variables. Taking into account the
covariance among the variables allows for the assessment of the direct effect or indirect effect of
one lead source on another lead source. Multiple regression, the traditional analysis method,
does not take into account the covariance structure among the variables. (An explanation of
direct and indirect effects is provided below).
1.1 DEFINITIONS
Throughout the document there are several terms which will be used. For clarity, the
definition of each term is given below.
Window Sill
"The portion of the horizontal window ledge that protrudes into the interior of the
room, adjacent to the window sash when closed [16]." Also referred to as
"window stool."
Window Well
"The portion of the horizontal window sill that receives both the upper and lower window
sashes when they are lowered, often located between the storm window and the interior
window sash [16]." Also referred to as "window channel" or "window trough."
-------
Structural Equation Models
Consider the pathway diagramed in Figure 1-1. This diagram shows directional pathways
of lead exposure as follows:
Window Sill Dust ^- Floor Dust > Blood
t
Figure 1-1. Pathways Diagram Assessing the Role of
Floor Dust Lead as a Route for Lead Dust
from the Window Sill to the Child's Blood.
Window sill dust is being assessed as to whether it directly impacts floor dust
and/or blood-lead (i.e., the arrows which point to blood and floor dust from
window sill dust),
Floor dust is being evaluated as to whether it is directly impacting blood-lead, and
Finally, window sill dust is being assessed as to whether it indirectly impacts
blood-lead via floor dust (i.e., the arrow which goes from window sill dust to floor
dust and then the arrow which goes to blood from floor dust).
These directional relationships illustrated in the diagram are represented by the following
two equations:
(1) Blood Lead = Floor Dust Lead + Window Sill Dust Lead
(2) Floor Dust Lead = Window Sill Dust Lead.
and a covariance matrix of all the variables in the diagram.
The directional nature of the diagram is illustrated through the equations by starting with
the highest numbered equation and working upwards, i.e., (2) and then (1). Equation (2)
represents the arrow from window sill dust to floor dust. Equation (1) represents the
arrow from floor dust to blood and the arrow from window sill dust to blood. Finally, the
indirect relationship of window sill dust to blood via floor dust is represented by both
equations (1) and (2) and the covariance matrix.
Evaluating equations (1) and (2) separately is similar to multiple regression or ANOVA,
i.e., both these methods can assess the direct effect of floor dust on blood in equation (1)
and window sill dust on floor dust in equation (2). What these methods cannot assess is
the indirect effect of window sill dust on blood via the floor dust. By evaluating both
equations (1) and (2) simultaneously and accounting for the covariance between these
variables, SEM allows for an assessment of the indirect effect.
-------
Environmental-Lead Pathways Model
The environmental-lead pathways model is a set of structural equation models that assess
the direct and indirect effects of lead hi environmental media, such as soil, window sill
dust, window well dust, floor dust, and paint. This model was developed for the R&M,
Rochester, and CAP Study data.
Blood-Lead Pathways Model
The blood-lead pathways model is a set of structural equation models that assess the
direct and indirect effects of lead in environmental media, such as soil, window well dust,
window sill dust, floor dust, water, and paint; child modifier variables, such as a child's
mouthing behavior; and indicator variables, such as the presence or absence of carpeting
in the entryway, on the blood-lead concentration of the child in the home. This model
was developed for the R&M and Rochester data.
Statistical Significance of Variables in the Structural Equation Models
A t-test was used to determine if the variable included in the model was statistically
significant. If the absolute value of the t-value was greater than or equal to 1.96 then the
parameter was considered to be statistically significant since the probability of observing
this extreme value (-1.96 or +1.96) for the t-test statistic is 0.05.
1.2 PEER REVIEW
Prior to publication, this report was reviewed by three individuals with knowledge and
expertise hi the subject matter of the report. These reviewers were not involved in the
development of the report, and conducted their reviews independently of each other. The
following is a summary of comments and responses to those comments which had an important
impact on the report or which are important for understanding and interpreting the report.
One reviewer strongly recommended that prediction limits be developed for the estimates
of percent change hi lead levels that resulted when the geometric mean of model input variables
was decreased. An appropriate methodology was developed and implemented, and prediction
limits were added to all tables which included a percent change. In response to another review
comment, an appendix was added to the report which describes the types of calculations made in
the pathways analysis and how the associated standard errors were estimated. This same
appendix include a discussion of goodness-of-fit statistics for the pathways models, hi response
to comments by the reviewers.
-------
A reviewer pointed out that hand wipe data were available for children in the Rochester
Study. A pathways analysis that included these data was added to the report. Another reviewer
asked whether bare soil or other "modifiers" should be included in the models. The Rochester
Study included a variable on vegetative soil cover, and an additional analysis was added which
incorporated this variable as a modifier to the soil lead level, in a manner similar to what was
done previously in the report for paint lead levels and paint condition.
Two reviewers commented on statements in the report about renovation and remodeling
in the Rochester Study. In fact, none of the houses in the Rochester Study had any renovation
and remodeling work done on them in the twelve months immediately prior to initiation of that
study. The language in the report was clarified on this point. However, information related to
renovation and remodeling was found for one of the other studies in the report. This information
was used to develop pathways analyses germane to the impact of renovation and remodeling for
a second study, using the same approach presented in the report at the time of peer review.
These additional pathways analyses were included in the final report.
Other review comments mentioned references for other lead pathway analyses that were
not included in the report. These other references were added to the discussion section of the
report, with appropriate text. Moreover, a reviewer pointed out that a published journal article
contained results which conflicted with the results of one of the pathway analyses in this report,
and requested double checking of the results in the report. An investigation showed that
differences in methodology were the reason for the differences between the two analyses. The
discussion section was revised to list the methodological differences.
One reviewer commented on the large number of pathways analyses, and stated that there
should be one environmental pathway analysis and one blood pathway analysis per study. The
same reviewer commented on the omission of lead from paint in a lead pathways analysis,
indicating that it would be far preferable to include lead from paint. The reviewer also stated that
there appeared to be a bias toward lead from soil. However, the report does contain analyses for
the Rochester study and for the Repair and Maintenance study which do include a measure of
lead hi paint for windows and doors. The text in the report was changed to clarify the goals of
the report. One of the primary objectives of the report was to examine lead pathways for three
different studies, and compare the results across the different studies. A secondary objective was
to examine other variables of interest that were not necessarily common to all three studies and
-------
to conduct pathways analyses with these additional variables. Paint and soil data were both used
in the report.
EPA has established a public record for the peer review under Administrative
Record 207. The administrative record is available in the TSCA Nonconfidential Information
Center, which is open from noon to 4 PM Monday through Friday, except legal holidays. The
TSCA Nonconfidential Information Center is located in Room NE-B607, Northeast Mall, 401 M
Street SW, Washington, B.C.
1.3 STRUCTURE OF THE REPORT
Section 2 discusses the conclusions drawn from the structural equation modeling.
Section 3 lists the quality assurance information for the data included in the analysis.
Background information specific to each study is provided in Section 4. Statistical methodology
used to assess the pathways models is discussed in Section 5. Results from the pathways
analysis are discussed in Section 6, and a discussion of the conclusions and results from this
report in comparison to published results is provided in Section 7. References are provided in
Section 8. Appendices A, B, and C present selected results from the CAP Study, R&M, and
Rochester data, respectively. Appendix D illustrates several pathway models which have been
analyzed and published in the literature. Finally Appendix E presents a discussion of model
specification, estimation criteria and goodness-of-fit tests associated with structural equation
modeling.
-------
2.0 FINDINGS AND CONCLUSIONS
This section presents conclusions drawn from the pathways models described in
Section 5 and the analysis results presented in Section 6. Though much of the same type of
information was collected in each of the three studies included in the analysis, there were a
number of variables not collected in some of the studies. For example, no blood-lead or water-
lead measurements were taken in the CAP Study. As a result, pathways including blood lead and
water lead could not be examined using the CAP Study data. When applicable, such exceptions
are noted. Because of these data limitations, there may be pathways other than those included in
the models which are significant contributors to either dust-lead levels or blood-lead
concentrations but were not investigated. Note that all the dust-lead samples included in the
analysis were all collected via vacuum.
2.1 CORRELATION ANALYSIS
Note: All correlations -were calculated within a study. The findings within a study are
compared across studies.
Correlation coefficients between window well and window sill dust-lead loadings
were larger than 0.5 for all three studies. Similarly, correlation coefficients for
window well and window sill dust-lead concentrations were larger than 0.5 for all
three studies. For one study, these correlation coefficients were larger than 0.8.
Water-lead samples in the R&M study were collected after a 2-hour fixed time
stagnation while the Rochester study water samples were collected after an 8-hour
stagnation period. The water-lead concentrations within the R&M study and within
the Rochester study were not highly correlated with any respective environmental
measure, or blood-lead concentration. [Note: Water and blood samples were not
collected in the CAP Study.]
For the Rochester study, blood-lead concentrations typically had a higher correlation
coefficient with dust-lead loadings than with dust-lead concentrations. For the R&M
study, blood-lead concentrations typically had a higher correlation coefficient with
dust-lead concentration than with dust-lead loading.
For the Rochester study, the environmental variables that had the highest correlation
coefficients with blood-lead were hand dust-lead (0.43), window well dust-lead
loading (0.37), soil-lead concentration (0.37), and door paint hazard score (0.36). For
the R&M study, the environmental variables mat had the highest correlation
coefficients with blood-lead were interior entryway dust concentration (0.56), floor
-------
dust concentration (0.53), and floor dust-lead loading (0.50). A number of other
environmental variables in the R&M study had correlation coefficients with blood-
lead that were larger than 0.40. (Hand dust-lead and soil-lead concentration were not
available for the R&M study.)
For the CAP study, floor dust-lead loadings were most highly correlated with interior
entryway dust-lead loadings (0.36). For the R&M study, floor dust-lead loadings
were most highly correlated with window sill dust-lead loadings (0.63). (The
correlation between floor dust-lead loading and blood-lead concentration was 0.50 for
the R&M study.) For the Rochester study, floor dust-lead loadings were most highly
correlated with hand dust-lead (0.36) and window sill dust-lead loading (0.35). (The
correlation between floor dust-lead loading and blood-lead concentration was 0.32 for
the Rochester study.)
2.2 ENVIRONMENTAL-LEAD PATHWAYS ANALYSIS
Note: The primary environmental-lead SEM was structured for comparison across the
three studies (CAP, R&M, and Rochester). The pathways to floor dust that were assessed
were interior entryway dust, exterior entryway dust, window sill dust, window well dust,
and soil. Because the CAP Study did not have paint data collected and more than half of
the R&M study homes did not have paint data, paint was not included as a pathway in the
primary analysis.
In all three studies, the significant direct pathways of lead exposure were: window
well dust lead to window sill dust lead for both loadings and concentrations, and
exterior entryway dust-lead concentration to interior entryway dust-lead
concentration.
In the R&M study, window well dust lead was an indirect pathway to interior
entryway dust lead via the window sill for both loadings and concentrations. In
addition, for the R&M study, window well dust-lead was an indirect pathway to floor
dust for both loadings and concentrations. In the Rochester study, window well dust
lead was an indirect pathway of lead exposure to floor dust lead via the window sill
for both loadings and concentrations. In the CAP study, window well dust-lead
concentration directly contributed to floor dust-lead concentration, while neither
window sill nor window well dust-lead loading were direct or indirect pathways of
lead exposure to floor dust-lead loading or interior entryway dust-lead loadings.
In the R&M study, exterior entryway dust lead loading was an indirect pathway to
floor dust lead via the interior entryway loadings. In the Rochester study, exterior
entryway dust-lead concentration was an indirect pathway to floor dust via the interior
entryway concentration. In the CAP study the exterior entryway dust-lead loadings
and concentrations were neither direct nor indirect pathways to the floor dust-lead
loadings or concentrations.
-------
In the CAP and Rochester studies, where soil-lead concentrations could be assessed,
soil-lead concentration was found to be an indirect pathway of lead to window sill
dust lead through the window well for both loadings and concentrations. (As noted
above, very few soil samples were collected in the R&M study and could not be
included as a pathway in the analyses.)
The following conclusions are drawn from the estimated environmental-lead pathways
models. These pathways models contained all variables, not just those that were significant. For
each study and model, one variable was assumed to have a 50% reduction in its geometric mean
lead level while all the other variables were held constant at their geometric mean lead levels.
In the CAP study, the largest reduction in floor dust-lead loadings or concentrations
occurred when the geometric mean soil-lead concentration was reduced 50%. In the
R&M study, the greatest decrease in the floor dust lead occurred when window well
dust lead was reduced by 50% for both loadings and concentrations. In the Rochester
study, a 50% reduction in the geometric mean soil-lead concentration had the greatest
effect on floor dust-lead loadings while a reduction hi interior entryway dust-lead
concentration had the largest effect on reducing the floor dust-lead concentrations.
In the CAP and Rochester studies, generally, the 50% reduction in the geometric
mean soil-lead concentration produced the largest reductions hi the other
environmental-lead levels (floor, interior entryway, window sill, and window well
dust-lead levels). In the R&M study, the largest reductions were achieved in window
sills when window well levels were reduced.
2.3 BLOOD-LEAD PATHWAYS ANALYSIS
Note: The primary blood-lead SEM was structured for comparison across the R&M and
Rochester studies. The pathways to blood that were assessed were floor dust, interior
entryway dust, exterior entryway dust, window sill dust, window well dust, water,
mouthing behavior, and soil. The CAP study data were not included because no blood
data were collected in the study. Paint pathways were not included in the primary
analysis because more than half the homes in the R&M study did not have paint data
collected.
In the R&M study, child's mouthing behavior, floor dust-lead loading, and interior
entryway dust-lead concentration were direct pathways of lead exposure to the child's
blood-lead concentration. In the Rochester study, the floor and window well dust-
lead loadings and child's mouthing behavior were direct pathways of lead to the
child's blood-lead concentration.
-------
For the R&M study, there were four indirect pathways of lead exposure to the blood
1) window well dust-lead loading to window sill dust-lead loading to interior
entryway dust-lead loading to floor dust-lead loading, 2) window well dust-lead
concentration to window sill dust-lead concentration to interior entryway dust-lead
concentration, 3) exterior entryway dust-lead loading to interior entryway dust-lead
loading to floor loading, and 4) exterior entryway concentration to interior entryway
concentration. For the Rochester study, two indirect pathways of lead exposure to
blood were 1) soil-lead concentration to window well dust-lead loading to window
sill dust-lead loading to floor dust-lead loading and 2) soil-lead concentration to
interior entryway dust-lead loading to floor dust-lead loading.
The following conclusions are drawn from the estimated blood-lead pathways models.
For each study and model, one variable was assumed to have a 50% reduction in its geometric
mean while all the other variables were held constant at their geometric means. For the
mouthing variable, a 50% reduction in mouthing activity was assessed.
For both the R&M and Rochester studies, the most significant reductions in blood-
lead concentration occurred when a child's mouthing habits were changed from
frequent mouthing behavior to infrequent mouthing behavior. For the Rochester
study only, the next largest reduction in blood-lead concentrations occurred when the
geometric mean window sill concentrations were reduced by 50%.
2.4 ASSESSMENT OF PAINT-LEAD INDICATORS
Paint-lead pathways were added to the environmental-lead pathways models discussed
above for both the R&M and Rochester studies and to the blood-lead pathways models for the
Rochester study. The paint-lead pathways were not added to the blood-lead pathways models for
the R&M data due to insufficient data. No paint information was collected in the CAP study.
The paint-lead pathways were represented by two types of paint-lead indicators:
paint-lead hazard score and average XRF measurement. For the Rochester study,
generally, the same statistically significant pathways were observed when either the
paint-lead hazard scores or the average XRF measurements were used as the paint-
lead indicator. For the R&M study, the number of statistically significant pathways
was the same when either average XRF measurements or paint-lead hazard scores
were used hi the concentration model. For the loading model, there were no
significant pathways with average XRF measurements in the model, whereas there
were two significant pathways when paint-lead hazard scores were used.
-------
In the Rochester study, the door paint hazard score was a direct pathway of lead to
interior entryway dust-lead loading and concentration and the window paint hazard
score was a direct pathway to floor dust-lead concentration and both window well and
window sill dust-lead loadings and concentrations. In the R&M study, door paint and
window paint hazard scores were direct pathways of lead to floor dust-lead loading,
and window paint hazard score was a direct pathway to ulterior entryway dust-lead
concentration.
In the Rochester study, door paint score was an indirect pathway of lead to floor dust-
lead loading and concentration via interior entryway dust and window paint score was
an indirect pathway of lead to floor dust lead via window sill dust. No significant
indirect effects of paint scores to floor dust-lead were found in the R&M study.
In the Rochester study, both door paint score and window paint score were direct
pathways to blood-lead. Door paint score was also an indirect pathway to blood-lead
concentration through interior entryway dust-lead loading and floor dust-lead loading.
Window paint score was an indirect pathway to blood-lead concentration through
window well dust-lead loading, window sill dust-lead loading, and floor dust-lead
loading and through window well dust-lead concentration.
2.5 ASSESSMENT OF HAND DUST-LEAD
Hand dust-lead was collected only in the Rochester study. This pathway was added to the
Rochester study blood-lead pathways model which included paint hazard scores.
The hand dust-lead pathway explained additional variation in the model.
Hand dust-lead was found to be a direct pathway of lead exposure to blood.
Floor dust-lead loadings and window well dust-lead concentrations were direct
pathways of lead to the hands.
2.6 ASSESSMENT OF SOIL-LEAD COVERAGE
In the Rochester study only, a variable was available that described the soil coverage (i.e.,
grassy, bare soil, etc.) at the location where each soil sample was taken. Combining the soil
coverage and the soil-lead concentration variables, a soil-lead coverage variable was created.
This variable replaced the soil-lead concentration variable in the Rochester blood-lead models
that included the paint hazard score.
10
-------
Replacing the soil-lead concentration variable with the soil-lead coverage variable did
not change most of the statistically significant pathways.
In the model with soil-lead coverage replacing the soil-lead concentration variable,
the pathway from soil to interior entryway dust-lead loading is no longer significant.
In the model with soil-lead coverage, the pathways from soil to window well dust-
lead loading and soil to window well dust-lead concentration have much lower
parameter estimates than is the case in the model with the soil-lead concentration
variable.
These changes are interesting, but should be viewed with caution. Changing the input
variables to the model is not as convincing as, for example, estimating of the effects of bare
versus grass covered soil through a controlled study.
2.7 ASSESSMENT OF CARPETED FLOORS
The results of this section must be viewed with extreme caution since potential
confounding effects such as age of the home, type of home, or other socioeconomic variables
were not taken into account in the analysis.
Using the blood-lead pathway models discussed above, the effect of carpeting on floors in
the homes and blood-lead concentration was assessed for both the R&M and Rochester studies.
For the R&M study, an indicator variable of the proportion of rooms sampled in the home with
carpeted floors was added as a pathway in the models. The analysis using the Rochester study
data included a pathway that accounted for the proportion of carpeted floors sampled in the home
and also an indicator of whether the interior entryway was carpeted. There was no information
available on the absence or presence of carpeting at the interior entryway for the R&M study.
Also, a separate analysis of the Rochester study data was conducted that included only homes
which had carpeted bedrooms and play areas.
For the R&M study, the proportion of carpeted floors was a direct pathway to blood
lead for both the loading and the concentration models. For the Rochester study, the
proportion of carpeted floors was a direct pathway to blood for the loading model.
Also for the Rochester study, the indicator of whether the interior entryway was
carpeted was a direct pathway to blood-lead in both the loading and concentration
models.
In the Rochester study, the presence or absence of interior entryway carpeting was a
direct pathway to interior entryway dust-lead loading, and the proportion of carpeted
floors was a direct pathway of lead to the floor dust-lead loading.
11
-------
In the Rochester study, when only homes with carpeted bedrooms and play areas were
included, the significant pathways of lead exposure generally remained the same. The
notable absent pathway was the floor dust-lead loading pathway. When only homes
with carpeted floors were included in the analysis, floor dust-lead loading was no
longer a statistically significant pathway of lead exposure to blood.
The following conclusions were drawn from the estimated pathways models that included
the carpeted floors indicator variables. For each study and model, each environmental variable
was assumed to have a 50% reduction in its geometric mean while all other environmental
variables were held constant at their geometric means. Also, a 50% reduction in average
proportion of carpeted floors and the presence or absence of interior entryway carpeting was
assessed when each of the other environmental variables was held constant.
In the R&M study, decreasing the proportion of carpeted floors in the home by 50%
produced a large predicted increase in the blood-lead concentration. For the
Rochester study, decreasing the proportion of carpeted floors increased the predicted
blood-lead concentration. Moreover, for the Rochester study, going from a carpeted
interior entryway to an uncarpeted entryway also increased the predicted blood-lead
concentration.
In the R&M study, reducing the proportion of carpeted floors increased both the floor
dust-lead loadings and concentrations. In the Rochester study, reducing the
proportion of carpeted floors produced decreases in floor dust-lead loading and
increases in floor dust-lead concentration. Moreover, going from a carpeted to
uncarpeted interior entryway also produced decreases in floor dust-lead loading and
increases hi floor dust-lead concentration.
As indicated above, these results for carpets should be viewed with extreme caution.
Furthermore, adding or changing input variables to the model is not as convincing as estimating
the effects of carpets through a controlled study.
2.8 ASSESSMENT OF RECENT RENOVATION AND REMODELING
Both the R&M and CAP studies collected information that indicated whether renovation
or remodeling had taken place hi the home six months prior to sampling at the home. An
indicator variable of whether renovation or remodeling had taken place in the home was included
in the environmental-lead pathways model for both the R&M and CAP studies and in the blood-
lead pathways models for the R&M study. Since the occurrence of major renovation and
12
-------
remodeling in the home within twelve months of attempted recruitment into the study was an
exclusion criterion for the Rochester study, a renovation and remodeling pathway was not
assessed for the Rochester study.
Renovation and remodeling in the six months prior to environmental sampling was a
direct pathway of lead exposure to floor dust loadings and concentrations in the CAP
study. For the R&M study renovation and remodeling in the six months prior to
sampling was neither a direct nor indirect pathway of lead exposure to environmental-
lead loadings or concentrations.
In the R&M blood-lead pathway model, renovation and remodeling in the six months
prior to environmental and blood sampling was neither a direct nor indirect pathway
of lead exposure to environmental-lead loadings or concentrations or to blood-lead
concentration.
2.9 ASSESSMENT OF RACE
In the Rochester study, the race of the children was available for analysis. A pathways
model was fitted to two subsets of the Rochester study data, African-American children and all
other children in the study (e.g., Caucasian, Hispanic, and Puerto Rican children).
/
For African-American children, significant pathways of direct lead exposure to blood
were mouthing behavior in the concentration model and window well dust-lead
loading in the loading model. For all other race groups, window well dust-lead
loading and concentration, interior entryway dust-lead loading, floor dust-lead
loading, window paint hazard score and door paint hazard score were direct pathways
of lead exposure to blood.
Window paint hazard score and soil concentration were indirect pathways to blood-
lead concentration via both window well dust-lead loading and concentration for the
other race group. In addition, exterior entryway dust-lead loading and door paint
hazard score were indirect pathways to blood-lead concentration via the interior
entryway dust-lead loading for the other race group. There were no indirect pathways
of lead to blood-lead concentration for the African-American children.
Note that the differences observed in this analysis could be due solely to the limited
sample sizes that resulted when the data was subsetted by race group.
13
-------
2.10 ASSESSMENT OF AIR DUCTS
Air duct dust-lead loadings and concentrations were collected in the CAP study. An air
duct dust-lead pathway was added to the CAP environmental pathways models. In neither the
Rochester nor the R&M studies was air duct dust lead available for inclusion as a pathway.
The air duct dust-lead pathway was neither a direct nor indirect pathway of lead to
dust-lead measurements in the CAP study.
14
-------
3.0 OVERALL QUALITY ASSURANCE
For all aspects of data management and statistical analysis, the SASฎ software package
was used. All procedures and routines in this software have passed rigorous quality control
testing.
As discussed earlier, the data from three studies, CAP, R&M, and Rochester were
included in this report. These data sources are considered secondary data sources since the data
included in the analysis have already been subjected to quality assurance checks by the
respective study coordinators. For each study a report has been published discussing the
management of the data during the study and the results of the statistical analysis of the data
[9,10,11,12,13].
The data were available as SAS data sets. To prepare the data for the analysis conducted
in this report, the data were compared to the published results. For the CAP study, summary
statistics generated using the in-house data were compared to Table 1-7 in [10], for R&M, the
data were compared to tables in [12], and for Rochester, results were compared to tables
presented in [13]. Any differences were noted and resolved.
Next, several variables in each study were combined to allow similar comparisons across
the three studies. For the CAP study, a principal components analysis indicated that the mass-
weighted average of the exterior entryway, foundation, and boundary of the property soil samples
was a reasonable representation of the soil lead at each home in the study.
For the R&M and Rochester data, five similar variables were created to aid in the
comparison of the analysis results. For each study, a variable indicating children's mouthing
behavior was calculated from the available mouthing information. A categorical water-lead
variable was created for each study. The proportion of carpeted floors in each home was
calculated. Finally, two indicators of lead in the window paint and door paint were calculated for
each study. The first indicator was a hazard score which took into account the condition of the
paint as well as the XRF reading on the sampled surface. The second indicator was just the
average XRF reading from the windows in the home and the average XRF reading from the
doors in the home. Further descriptions of each of these variables are provided in Sections 4
and 5.
For each of the created variables, a hand check of the accuracy of the calculations was
conducted, and frequency, summary statistic, and graphical validations were performed.
15
-------
As the statistical analysis was performed, validations of the data included in the analysis
were periodically performed using frequency counts and summary statistics. PROC CALIS of
the SASฎ software system was used as the primary analysis procedure.
All tables and figures in the report were validated through visual inspection. When
possible, direct processing from SASฎ output to WordPerfect tables or figures was employed to
reduce any chances of error.
16
-------
4.0 DESCRIPTION OF ENVIRONMENTAL FIELD STUDIES
Data from the Baltimore Lead-Based Paint Abatement and Repair and Maintenance
Study (R&M), the Rochester Lead-in-Dust Study, and the Comprehensive Abatement
Performance Study (CAP) were used in the analysis. Brief descriptions of each data source are
provided below. Included in the descriptions are the purpose of the study, the data collected in
the study, the data used in the pathways analysis, and summary statistics for the data included in
the pathways analysis.
4.1 BALTIMORE LEAD-BASED PAINT ABATEMENT AND REPAIR AND MAINTENANCE
STUDY (R&M)
The purpose of the Lead-Based Paint Abatement and Repair and Maintenance (R&M)
study was to compare short-term (2 to 6 months) and long-term (12 to 24 months) efficacy of
comprehensive lead-paint abatement and repair and maintenance interventions for reducing lead
in settled house dust and children's blood. Five categories of vacant and occupied homes were
recruited into the study. The first three categories were Repair and Maintenance (R&M) I,
R&M n, and R&M m homes signifying the level of repair and maintenance efforts that would
be applied to the home. The repair and maintenance intervention homes were older, low-income
rental properties in Baltimore City. By study design, the R&M intervention homes were required
to be structurally sound, not excessively furnished, 800 to 1200 square feet in size, and
containing either lead-based paint (^ 0.7 mg Pb/cm2 or * 0.5 percent Pb by weight) on at least
one surface in a minimum of two rooms or be built prior to 1941. The other two categories
included "control" homes: modern urban and previously abated. The modem urban homes were
built after 1979 and were identified by house-to-house visits in areas where these newer homes
were clustered. The previously abated homes were chosen from homes that were abated in past
years as part of either the City of Baltimore or Kennedy Krieger Research Institute lead-based
paint abatement demonstration projects. The majority of the homes hi all five groups were
rowhouses.
For all five groups of homes, questionnaire data, blood-lead samples, and environmental-
lead samples were collected between January 1993 and November 1994. The data analyzed in
this report are "pre-intervention" data, collected prior to the implementation of the repair and
maintenance interventions in the study. Venous blood-lead samples were collected at Kennedy
17
-------
Krieger Research Institute Lead Clinic by a pediatric phlebotomist using 3 mL vacutainers. In
many homes, multiple children had blood samples taken. For the pathways analysis, only the
blood-lead sample of the youngest child in each home was considered. The average age of these
children was 2.2 years ranging from 6 months to 4.8 years of age. The settled house dust
samples were collected using the Baltimore Repair and Maintenance (BRM) vacuum. Three
i
composite floor dust samples - one across rooms with windows on the first story, one across
rooms with windows on the second story, and one from first and second story rooms without
windows - were collected in each home. Composite window sill and window well samples were
collected separately from all first and second story windows. Individual samples of settled dust
were collected from the interior entryway and exterior entryway. Two-hour fixed-time
stagnation drinking water samples were collected from the kitchen faucet the cold water was
run for at least two minutes to flush the pipes, then a sample from the first flush after a stagnation
period of two hours was collected. XRF measurements were taken on surfaces such as windows,
doors, walls and ceilings. Generally, in each home an XRF measurement from at least one
window and one door surface was taken. The other components were not consistently sampled
in each home. Therefore, for the pathways analysis only XRF samples from the windows and
doors were included in the analysis. Note that paint samples for this study were collected only
for the R&M I, H, and m homes. Of the 75 R&M I, H, and HI homes, 72 homes had paint
samples collected. Thirty-six of the 72 homes were not occupied, therefore only the 36 occupied
homes were included in the paint analyses. Additional details regarding the sampling protocol are
available in [12]. Table 4-1 provides descriptions of the variables used in the R&M pathways
analysis. Tables 4-2 and 4-3 present summary statistics for the variables included in the
pathways analysis.
The R&M homes, by definition, had lead-based paint on at least one surface in a
minimum of two rooms or were constructed prior to 1941. The control homes on the other hand
were previously abated or were built after 1979 [12]. Table 4-4 provides summary statistics for
the R&M and control homes separately. The R&M homes have higher concentrations and
loadings than the control homes for all media. The blood lead concentrations for the children in
R&M homes are higher, on average, than those for the children in the control homes. The
differences seen in the average levels were considered and assessed prior to the pathways
analysis. The statistical analysis results are presented in Section 6.
18
-------
Table 4-1. Description of Variables Used in the Baltimore Lead-Based Paint Abatement and
Repair and Maintenance (R&M) Pathways Analysis.
Analysis Variable ^i'fe
Blood
Water
Mouthing
Renovation &
Remodeling
Exposure
Window Paint
Hazard Score
Door Paint Hazard
Score
Proportion of Floors
that are Carpeted
Description , - Jt ^-yyyy '"y
Venous blood sample collected from children between the ages of 6 months and
4.8 years.
Individual sample collected from the kitchen faucet as a 2-hour fixed-time stagnation
sample.
The sample value was coded as 0 if the sample was less than or equal to the limit of
detect/on (LOD - 0.6 ug/L), 1 if the sample was greater than the LOD and less than
or equal to 2.6 ug/L, and 2 if the sample was greater than 2.6 fjg/L.
Indicator of how often a child puts their fingers, dirt, or paint chips into their mouth
or puts their mouth on the window sill.
The variable was coded as Off the child infrequently puts fingers, dirt, or paint into
mouth or mouth on the window sill U 1 day/week) or 1 if the child frequently puts
fingers, dirt, or paint into mouth or mouth on window sill (> 1 day/week).
Indicator of the renovation & remodeling activity in the home as indicated during
initial interview.
The variable was coded as 0 if no renovation or remodeling occurred in the 6
months previous to the interview and 1 if renovation or remodeling occurred within
the 6 months previous to the interview.
Arithmetic average of the product of the paint condition score and the XRF readings
taken from window wells, sashes, and sills in rooms throughout the house.
The paint condition score was 0 for intact paint and 1 for non-intact paint.
Arithmetic average of the product of the paint condition score and the XRF reading
taken from doors and door jambs throughout rooms in the home.
The paint condition score was 0 for intact paint and 1 for non-intact paint.
For all interior floors from which floor dust samples were collected, excluding the
interior entryway, this is the proportion of those floors which were carpeted.
Dust"1
Floor
Interior Entryway
Exterior Entryway
Window Sill
Window Well
Arithmetic average of composite dust samples collected: 1 in rooms with windows
on the first floor, 1 from rooms with windows on the second floor, and 1 from first
and second floor rooms without windows.
Arithmetic average of individual samples collected from the interior entryway of the
home.
Arithmetic average of individual samples collected from the exterior entryway of the
home.
Arithmetic average of composite dust samples collected from the window sills in
rooms on the first and second floors.
Arithmetic average of composite dust samples collected from the window wells in
rooms on the first and second floors.
Note: All dust lead loadings are area-weighted averages in jt/g/ft2 and dust lead concentrations are mass-
weighted averages in//g/g.
la) Samples collected using the BRM vacuum.
19
-------
Table 4-2. Summary Statistics for the Baltimore Lead-Based Paint Abatement and Repair
and Maintenance (R&M) Variables Included in the Pathways Analysis.
Location of
, Sample
i
N
Loadings (//g/ft2) * '
Geometric Mean
(GSD)*
Dust Levels
*
Floor
Interior Entryway
Exterior Entryway
Window Sill
Window Well
87
87
85
87
86
210 (5.63)
329 (8.28)
405 (9.44)
1,229(17.6)
37,035(21.0)
Minimum -Maximum
2.09 - 24,726
3.47-26,417
7.25- 196,752
2.10- 122,368
36.2 - 2,496,630
', Concentrations (f/glg)** , '
Geometric Mean
>? (GSD)*
'<'' * i
'- % ^ .< \
Minimum - Maximum
1,118(4.48)
1,459(4.50)
1,570(5.20)
5,411 (8.65)
8,452 (7.32)
40.4-64,109
39.8 - 42,625
17.6-89,505
7.25- 141,057
108- 191,480
Other Levels
Blood
Water
87
87
8.35 (2.08)
2.25 (3.98)
0.90-41.9
0.15-29.7
* GSD = Geometric standard deviation
** Blood-lead concentration is in ug/dL and the water-lead concentration is in pg/L.
Table 4-3. Summary of the Other Variables Included in the Baltimore Lead-Based Paint
Abatement and Repair and Maintenance (R&M) Pathways Analysis.
, Indicator
Variable , -
Water-Lead Level
Mouthing Behavior
R&R Exposure Indicator
' levels
0
1
2
0
1
0
1
' ' '5* ' , i V,s
1 \ ' '~ '' '* , >,
, * ' / , , , Description < <
* LOD (0.6 fjg/L)
> LOD and <; 2.6 uglL
> 2.6/jgJL
Infrequently puts fingers, dirt, or paint chips
in mouth or mouth on the window sill
Frequently (>, 1 day/week) puts fingers,
dirt, or paint chips in mouth or mouth on
the window sill
No renovation or remodeling occurred in 6
months previous to interview
Remodeling or renovation occurred 6
months previous to interview
%0l
Population
21
35
44
64
36
91
9
20
-------
Table 4-3. Summary of the Other Variables Included in the Baltimore Lead-Based Paint
Abatement and Repair and Maintenance (R&M) Pathways Analysis (Continued).
1 Indicator , .-
Variable
Variables
Door
Paint
Window
Paint
Hazard Score
XRF
Measurement
Paint Condition
Hazard Score
XRF
Measurement
Paint Condition
Proportion of Floors that are
Carpeted'01
>.ซ*. '.' \
t'-1 . Summary Statistics
Mean
1.29
6.70
0.18
5.19
7.16
0.71
0.29
Min
0.0
0.50
0.00
0.00
0.90
0.00
0.0
Max
15.0
23.8
1.00
21.8
38.7
1.00
0.88
Std
2.97
5.45
0.28
5.05
6.71
0.31
0.29
N
36ia'
36(a)
36(a)
36(a)
36(a)
36(a)
87
72 of the 75 R&M homes had XRF measurements collected. Only 36 of the homes were occupied resulting
in only 36 R&M homes being included in paint summaries.
(W The XRF measurements were not substrate corrected.
(c) On average 73% of the floors in a modern urban home were carpeted, while 20% of floors in a previously
abated home and 19% of floors in an R&M home were carpeted.
Table 4-4. Geometric Mean Lead Loadings (//g/ft2) and Concentrations (//g/g) for the R&M
and Control Homes Included in the R&M Pathways Analyses.
Location of
Sample
Loading
R&M Homes
(N = 56)
Dust - Geometric Mean (GSD)*
Floor
Interior Entryway
Exterior Entryway
Window Sill
Window Well
398 (3.66)
520 (6.44)
668 (8.95)
7,144 (3.35)
285,722(2.71)
s fc/g/ft2)
Control Homes
(N = 31 )
66.4 (6.24)
144(10.1)
155(7.82)
51.1 (8.66)
987 (6.10)
Concentrations U/g/g)
. R&M Homes
. (N = 56)
Control Homes
(N = 31 )
2,219(2.69)
2,575 (3.52)
2,810(3.59)
19,663(2.42)
29,428(2.16)
324 (4.39)
522 (3.96)
510(5.58)
526 (5.97)
924 (4.46)
Other Levels
Blood (ualdl)
9.98(1.87)
5.97 (2.27)
* GSD = Geometric standard deviation.
21
-------
4.2 ROCHESTER LEAD-IN-DUST STUDY
The Rochester Lead-in-Dust Study was designed to address several objectives: "to
determine whether dust-lead loading (ug/ft2) or dust-lead concentration (jig/g) is a better
predictor of children's blood lead levels; to investigate whether dust sampling using vacuum
methods or a wipe method is more predictive of children's blood lead levels; to identify which
interior household surface(s) should routinely be sampled for dust lead measurements; and to
estimate the probability of a child having an elevated blood lead level on the basis of a known
level of lead in house dust, controlling for other potential exposures. [13]"
Children 12 to 30 months of age who lived in the city of Rochester and had no known
history of elevated blood-lead concentrations were eligible for the study. Children were
excluded from the study if they had taken a prescribed iron supplement 2 months prior to
recruitment or if there had been major renovation hi their residence 12 months prior to
recruitment. The location of a child's residence, the blood-lead history of the child, and other
eligibility criteria were applied to control for the possibility of non-residential, non-typical
sources of lead affecting blood-lead concentrations [13]. Two hundred and five children were
enrolled into the study. To summarize the children and homes in the study, Table 4-5 shows the
distribution of the year in which a home was built, the average percentage of carpeted floors, and
the age of the child at blood collection. This table shows that 84% of the homes in the study
were built prior to 1940 and that 44% of the children in the study were between 12 months and
18 months of age.
Collection of questionnaire data, blood-lead samples, and environmental-lead samples
was performed between August and November, 1993. Venous blood.samples were obtained by a
certified pediatric phlebotomist during a home visit using lead-free containers provided by the
New York State Department of Health Clinical Laboratory Evaluation Program. Three dust
collection methods (one wipe method and two vacuum methods) were used to sample settled
house dust: wipe sampling, Dust Vacuum Method (DVM) sampling, and Baltimore Repair and
Maintenance (BRM) vacuum sampling [13]. To aid in comparisons with the CAP and R&M
results, only samples collected using the BRM vacuum sampling method were used in the
pathways analyses.
22
-------
Table 4-5. Distribution of the Year Homes were Built, Average Percentage of Carpeted
Floors, and Age of the Children for the Rochester Lead-in-Dust Study Data.
>H *>-v - - '^ '
X *
^ : , '
' N ->' '
ii<. >,
Percent of Population
i 1
Average, Percentage
of Carpeted Floors
Year in which home was built
Pre- 1900
1900-1909
1910- 1919
1920- 1929
1930- 1939
1940- 1959
1960- 1979
Post- 1979
19
32
40
61
20
19
4
10
9%
16%
20%
30%
10%
10%
2%
5%
36%
34%
38% '
38%
48%
32%
48%
36%
Age of children at time of blood collection
12-18 months
1 8 - 24 months
24 - 30 months
90
57
58
44%
28%
28%
Dust samples were collected from the window well, interior window sill, and floor in the
child's bedroom; the window well and floor in the kitchen; the window well, interior window sill,
and floor in the child's principal play area; the interior window sill and floor in the living room;
the interior entryway floor; and the exterior entryway floor. Floor samples were collected over a
1-ft2 area. Window well and interior window sill samples were collected over one-third of the
available surface area. Soil core samples, taken at a depth of Vz inch, were collected in two
distinct areas: the perimeter of the foundation and the child's principal outside play area.
Because a significant number of homes did not have play area soil samples taken, only the
foundation soil samples were used in the pathways analyses. Three core soil samples were taken
on each side of the house around the perimeter of the foundation and combined for a composite
foundation sample. The composite samples were sieved into fine and coarse samples, hi
addition, the amount of bare soil in the immediate area of the soil sample was characterized as
l=no bare soil, 2=small amount bare, 3=half bare, 4=mostly bare, and' 5=all bare. A pair of
composite hand dust-lead samples were taken from each child in the study. One was taken at the
beginning of the home visit, before any hand washing was done and the other was taken at the
23
-------
end of the home visit or two hours after the child's hands were washed, whichever came first.
Each composite hand dust-lead sample consisted of two wipe samples, one from each hand. Two
water samples were taken at each home. One sample was a first draw after a minimum 8-hour
stagnation period. The other was collected after a one minute flush. XRF paint measurements
were taken from components in a number of areas, such as the child's bedroom, the child's
principal play area, the kitchen, and the living room. Three XRF measurements were taken on
each surface and were averaged together. XRF measurements were not substrate corrected for
the pathways analysis. A visual inspection of each surface was also performed, and the paint
condition rated as poor, average, or good.
Floor dust-lead samples were collected from both carpeted and uncarpeted floors in some
homes. In order to account for potential differences in dust-lead exposure from carpeted floors
and uncarpeted floors, two average floor dust-lead values were calculated - one for carpeted
floors and one for uncarpeted floors. These two average values were then combined, weighting
the average floor dust-lead on carpeted floors by the proportion of rooms that were carpeted and
weighting the average floor dust-lead on uncarpeted floors by the proportion of rooms that were
uncarpeted. In addition, two variables related to the presence of carpeting in the home were
calculated and included in the pathway models: 1) the proportion of rooms mat were carpeted
and 2) an indicator of the presence of carpeting at the interior entryway. Table 4-6 provides a
description of the variables that were selected to be included in the pathways analysis.
Tables 4-7 and 4-8 provide summary statistics for the variables included in the pathways
analysis.
Note that some homes did not have all samples collected, resulting in less than 205
samples for many of the environmental variables. In the pathways analysis, if the value for a
variable in the model was missing for a home, the home was not included in the pathways
analysis.
24
-------
Table 4-6. Description of Variables Used in the Rochester Pathways Analysis.
Analysis Variable
Blood
Water
Mouthing
Indicator of whether
the Interior Entryway
is Carpeted
Proportion of Floors
that are Carpeted
Window Paint Hazard
Score
Door Paint Hazard
Score
Soil""
Soil-Lead Coverage
Description ~ ". '^-^^ ':'- =
Venous blood sample collected from children between the ages of 1 2 and 30 months.
Two individual water samples were collected from the kitchen faucet. One sample was a first
draw collected after an 8-hour stagnation period. The other sample was collected after a one
minute flush.
The sample value was coded as 0 if the sample was less than 0.5 ug/L and 1 if the sample was
greater than or equal to 0.5 ug/LM.
Indicator of how often a child puts their thumb, paint chips, or dirt into their mouth or puts their
mouth on the window sill.
The variable was coded as 0 if the child never, rarely, or sometimes puts their thumb, paint
chips, or din into their mouth or puts their mouth on the window sill and 1 if the child often or
always puts their thumb, paint chips, or dirt into their mouth or puts their mouth on the window
sill.
Indicates whether the interior entry way was carpeted.
The variable was coded as 0 if the interior entryway was not carpeted and 1 if the interior
entryway was carpeted.
For all interior floors from which floor dust samples were collected, excluding the interior
entryway, this is the proportion of those floors which were carpeted.
Arithmetic average of the product of XRF paint measurements and paint condition score from
window sills, wells, and sashes.
The paint condition score was coded as 0 if 0 % to less than 5% of paint on surface
characterized as deteriorated, 1 if greater than 5% of paint on the surface is characterized as
deteriorated.
Arithmetic average of the product of XRF paint measurements and paint condition score from
interior doors and jambs.
The paint condition score was coded as 0 if 0 % to less than 5% of paint on surface
characterized as deteriorated, 1 if greater than 5% of paint on the surface is characterized as
deteriorated.
Composite samples collected from the foundation of the home.
Arithmetic average of the product of soil lead measurements and soil cover score.
The soil cover score was 1 = no bare soil, 2= small amount bare, 3= half bare, 4= mostly bare,
and 5= all bare.
Dust*
Floor
Interior Entryway
Exterior Entryway
Hand"*
Window Sill
Window Well
Individual samples collected from carpeted and uncarpeted floors in the bedroom, play area,
kitchen, or living room.
Individual samples collected from the interior entryway of the home.
Individual samples collected from the driveway and porch.
Composite samples collected from the children's hands.
Individual samples collected from the interior window sills in the bedroom, play area, or living
room.
Individual samples collected from the interior window wells in the bedroom, play area, or
kitchen.
Note:
(a)
(b)
(c)
Id)
All dust-lead loadings are area-weighted averages in j/g/ft* and dust-lead concentrations are mass-weighted averages
in fjg/g.
72% of the one minute flush water lead samples were * 0.5 //g/L. The other 28% of the samples ranged from
1 i/g/l to 24 ugl\. with the exception of one sample at 157 g/L. Note that the definition of the presence of lead in
water for the Rochester data differs from that for the R&M data.
Composite of up to 12 samples taken at a depth of 0.5 inch sieved into coarse and fine fractions for analysis. An
average of the fine and coarse samples was used in the analysis.
Samples were collected using the BRM vacuum method. Floor samples were collected from 1-ft2 sample areas.
Window well and window sill samples were collected from % of the available surface area.
Pre and post interview samples were each a composite of two samples, one from each hand. An average of the pre
and post interview samples was used in the analysis.
25
-------
Table 4-7. Summary Statistics for the Rochester Variables Included in the Pathways
Analyses.
! '
Location of Sample
N*ป
Loadings (//g/fts)
Geometric
Mean (GSD)*
Minimum - Maximum
Dust
Exterior Dust
Floor Dust
Interior Entryway
Dust
Hand Dust (fjg}
Window Sill Dust
Window Well Dust
ii^^^m^^^*imittiHiiiHii^^m^^^iimiiiiiiiiiiiiiiimiiiiiiimiiiiiim
, "_ ^
%i ป
Other
Blood (jug/dL)
Soil (/jglg)
164
166
177
197
197
189
f^f^^Htuummm
515 (7.34)
100(4.34)
88.6(13.5)
2.26(2.12)
345 (10.5)
22,584(21.6)
^^^^^^^^^^^^^^^^H
0.08-51,012
3.49 - 37,093
0.30 - 32,040
0.38 - 25.85
0.68- 117,821
6.86-3,030,214
^^^^^^^v
-
f
N** :
. . Concentrations b/g/g)
Geometric Mean
(GSD)*
172
162
174
197
199
188
^^^i
N**
205
187
656 (5.35)
563 (4.05)
468 (4.90)
2,787 (8.44)
8,676 (10.7)
^^^^^^^^^^^^^^^1^^^
Concer
Geometric Mean
,'v:-.[GSD)* ,,
Minimum -
, i Maximum
0.16-44,854
21 .9 - 57,346
1.62-20,785
3.15-368,111
5.15-207,181
^^^H
itrations
Minimum - ; '
Maximum . :
6.38(1.85)
852 (3.83)
1.4-31.7
19.8-27,143
GSD = Geometric standard deviation.
* N = number of homes at which the sample was collected.
Table 4-8. Summary of Other Variables Included in the Rochester Pathways Analysis.
Indicator Variable
Water-Lead Level
Mouthing Behavior
Indicator of Whether the Interior
Entryway is Carpeted
Soil Coverage
Variable '::^-v:\v.^1
Door Paint Hazard Score
Door Paint XRF Measurements
Door Paint Condition Score
Window Paint Hazard Score
mHH^^HB^MWV^BVH^^Bl^H^^H^^^^^^BVB^^HM^B
Window Paint XRF Measurements
Window Paint Condition Score
Proportion of Sampled Floors that are
Carpeted'"
Levels
0
1
0
1
0
1
1
2
3
4
5
*;rN** '
196
196
196
199
199
199
205
.'>--. - - Description? ;ฐ;ซ:;:.v':";::,:,;:':--> "->;
> 0.5 A/g/L
s. 0.5 //g/L
Infrequently puts thumb, dirt, or paint chips in
mouth or mouth on the window sill
Frequently puts thumb, dirt, or paint chips in
mouth or mouth on the window sill
Uncarpeted interior entryway
Carpeted interior entryway
No bare
Small amount bare
Half bare
Mostly bare
All bare
^f Mean
0.97
3.47
0.16
3.48
5.81
0.44
0.38
Miri
0.00
0.50
0.00
0.00
HHM^BHMM^_^HB^^_H^^
0.50
0.00
0
Max '
24.8
48.4
1.00
33.0
33.0
1.00
1
%rof Population
27
73
81
19
64
36
1
86
13
0
0
StdPev
3.61
6.54
0.31
4.83
5.64
0.39
0.21
(a)
N = number of homes at which the sample was collected.
On average, 39% of sampled floors in a home built before 1940, 32% of the sampled floors in a home built
between 1940 and 1959, 48% of the sampled floors in a home built between 1960 and 1979, and 36% of the
sampled floors in a home built after 1979 were carpeted (Table 4-5).
26
-------
4.3 COMPREHENSIVE ABATEMENT PERFORMANCE (CAP) STUDY
There were four objectives for the Comprehensive Abatement Performance (CAP) study.
They were:
(1) Assess the long-term efficacy of two primary abatement methods;
(2) Characterize lead levels in household dust and exterior soil in unabated homes and
homes abated by different abatement methods;
(3) Investigate the relationship between lead hi household dust and lead from other
sources, in particular, exterior soil and air ducts, and
(4) Compare dust lead loading results from cyclone vacuum sampling and wipe
sampling protocols [9,10,11].
This study was a follow-up to the HUD Lead-Based Paint Abatement Demonstration [5]
project which assessed the costs and short-term efficacy of alternative methods of lead-based
paint abatement in 169 homes in five urban areas: Washington/Baltimore, Birmingham, Denver,
Indianapolis, and Seattle/Tacoma. The CAP study, conducted two years after the completion of
the HUD Abatement Demonstration, evaluated the longer-term performance of the abatement
strategies employed in the Demonstration project.
Thirty-five homes which were abated in the HUD Abatement Demonstration project and
were located in Denver were included hi the CAP study. To assess the performance of the
abatement methods, pre- and post-abatement soil and dust-lead levels were needed. Only
foundation soil samples and a limited number of dust samples were collected prior to abatement
in these 35 homes. Therefore, to provide a comparison of the abatement performance lead levels
with other environmental levels, 17 unabated homes hi Denver that were previously tested by
XRF hi the HUD Demonstration and were found to be relatively free of lead-based paint were
included in the CAP study. For these 52 homes, 35 abated homes plus 17 unabated homes,
environmental samples were collected during March and April of 1992.
In general, two to three rooms in each home were selected for testing. Dust samples were
collected on the perimeter of the floors, from the window sills, the window troughs, and the air
ducts in each of these selected rooms. Additionally, dust samples were collected from inside and
outside the entryways. Most dust samples were collected by a vacuum method, and only vacuum
27
-------
dust samples were included in the analysis for this report. Soil core samples were collected from
the entryways adjacent to the home, the foundation of the home, and the boundary of the
property. The lead in the dust samples were reported both as loadings (ng/ft2) and concentrations
(jig/g) while the lead from the soil samples were reported as concentrations (ng/g). In addition,
the information on whether renovation and remodeling had taken place in the home six months
prior to the environmental sampling was collected. Additional sampling protocol details are
available in [9, 10,11].
The samples included in the pathways analysis are described in Table 4-9. Since the CAP
study houses were vacant prior to the performance of the abatements, no blood-lead
measurements were collected.
Table 4-9. Description of Variables Used in the Comprehensive Abatement Performance
(CAP) Study Pathways Analysis.
! Analysis Variable
Dust"1
Floor
Interior Entryway
Air Duct
Window Sill
Window Well
Exterior Entryway
"" , ' * ' "T
Description , ' ' - - " - !
Arithmetic average of dust samples collected from the perimeter of carpeted or
uncarpeted floors in the kitchen, bedroom, living room, etc.
Note: Two to three rooms, in general, were chosen for each home.
Arithmetic average of dust samples collected immediately inside the front and rear
entryways of the home.
Arithmetic average of dust samples collected from the air ducts in the rooms
selected for sampling.
Arithmetic average of dust samples collected from the window sills in the rooms
selected for sampling.
Arithmetic average of dust samples collected from the window wells in the rooms
selected for sampling.
Arithmetic average of dust samples collected immediately outside the front and
rear entryways of the home.
Soil""
Exterior Entryway
:oundation
Boundary
Arithmetic average of soil samples collected immediately outside of the front and
rear entryways.
Arithmetic average of soil samples collected at the foundation of the home.
Arithmetic average of soil samples collected at the property boundary
Note: Dust-lead loadings are area-weighted averages in //g/ft2 and dust-lead concentrations are mass-weighted
averages in //g/g.
Samples were collected using a cyclone style vacuum. Floor and entrance samples were collected from
1 - ft2 sample areas. Other samples were collected from the entire accessible surface.
At each location, a composite of three core samples was collected. The three composite soil samples were
averaged together for this analysis using a mass-weighted average as indicated by a principal components
analysis.
la)
ft)
28
-------
For each analysis variable and each home in the study, an average lead level in the home
was calculated. This was done to provide comparable information from each home. Summary
statistics over all 52 homes are provided in Table 4-10, while Table 4-11 presents summary
statistics for the unabated and abated homes separately. The geometric mean dust-lead loadings
and concentrations are lower in the unabated homes than in the abated homes. The same holds
true for the soil-lead concentrations. Note that the abated homes generally had paint lead levels
greater than or equal to 1 mg/cm2 prior to being abated while the unabated homes were found to
be relatively free of lead-based paint. In addition, the abated homes on average were built in
1926 while the average year in which the unabated homes were built was 1943 [10]. The
differences seen in the average levels were considered prior to the pathways analysis. This is
covered in Section 6.1.
Table 4-10. Summary Statistics for the Lead Loadings (//g/ft2) and Concentrations (//g/g)
for All Homes Included in the CAP Study Pathways Analysis.
- 3 s *1 '
^s^i
lr
Location of Sample
r.
N*ป
Dust
Floor
Interior Entryway
Air Duct
Window Sill
Window Well
Exterior Entryway
52
52
52
52
49
51
. Loadings U/g/ftz)
Geometric Mean
(GSD)ป
48.8(5.01)
342 (4.57)
156(8.16)
107(6.81)
3,230 (6.32)
574 (4.00)
Minimum -
Maximum
Concentrations (fjg/g)
, Geometric Mean
(GSD)*
Minimum -
' Maximum
2.02 - 2,640
1 .66 - 4,870
4.13-25,400
2.76-16,700
45.0 - 93,500
17.2-10,700
174(2.91)
201 (2.44)
485 (2.23)
778 (4.08)
1,577(4.65)
261 (2.68)
SoU
Exterior Entryway
Foundation
Boundary
Indicator
Variable
R&R Exposure
Indicator
52
52
52
Levels
0
1
157 (2.24)
196(2.52)
132(1.94)
i " , * "" "
Description
No renovation or remodeling occurred in 6 months prior to
interview
Remodeling or renovation occurred 6 months prior to
interview
37.9 - 5420
9.65 - 4,940
1 22 - 2,920
45.8-17,000
133-22,900
20.7 - 2,820
19.7-644
11.0-1810
24.1 - 606
" ' %of i
, Population
75
25
*GSD = Geometric standard deviation.
** N = Number of homes at which the sample was collected.
Note: Average levels were calculated for each home separately. The summary statistics were then
calculated using the average levels. This differs from the results presented in Table 1-7 in the CAP
study report [10] where the summary statistics were calculated using all samples individually.
29
-------
Table 4-11. Geometric Mean Lead Loadings (/^g/ft2) and Concentrations (//g/g) for Abated
and Unabated Homes Included in the CAP Study Pathways Analyses.
w
Location of
Sample
Dust
Floor
Interior Entryway
Air Duct
Window Sill
Window Well
Exterior Entryway
Loadings (fig/ft2)
Unabated Homes*
(N = 17)
Abated Homes* ~<
(N=35)
42.3 (6.24)
200 (7.40)
58.3 (7.24)
52.7 (5.14)
2,600 (6.46)*
351 (4.43)*
52.4 (4.55)
444 (3.22)
252 (7.48)
152(7.17)
3,588 (6.38)*
718(3.67)
Concentrations (j/g/g)
Unabated Homes*
(N = 17)
Abated Homes* -;
(N=35) ',
-
143(2.66)
164(2.25)
449 (2.38)
487 (3.95)
1,050(4.15)*
204 (2.67)*
192(3.03)
222 (2.52)
503(2.18)
977 (3.99)
1,923(4.82)'
292 (2.66)
Soil
Exterior Entryway
Foundation
Boundary
122(2.47)
107 (2.43)
99.1 (2.06)
177(2.09)
263 (2.22)
151 (1.82)
* Geometric standard deviation is given in parentheses.
t Only 16 samples were included in the calculations.
t Only 33 samples were included in the calculations.
30
-------
5.0 STATISTICAL METHODOLOGY
Structural equation modeling (SEM) is similar to multiple regression, factor analysis, and
analysis of variance (ANOVA) approaches in that these approaches are based on linear statistical
>
models. A difference between SEM and these other approaches is that SEM requires the formal
specification of a model that includes both direct and indirect effects. The other approaches
require specification of only direct effects.
A direct effect is a directional relation between two variables, typically the type of
relationship evaluated using ANOVA or multiple regression. Within a model, the direct effect
characterizes the relationship between an independent variable and the dependent variable in the
equation. The indirect effect, which can be specified in a SEM, characterizes the effect of an
independent variable on a dependent variable through intervening or mediating variables [3]. For
example, consider the following pathway diagram:
t
This diagram shows directional pathways of lead exposure as follows:
Window sill dust is being assessed as to whether it directly impacts floor dust
and/or blood-lead (i.e. the arrows which point to blood and floor dust from
window sill dust.),
Floor dust is being evaluated as to whether it is directly impacting blood-
lead, and
Finally, window sill dust is being assessed as to whether it indirectly impacts
blood-lead via floor dust (i.e., the arrow which goes from window sill dust to
floor dust and then the arrow which goes to blood from floor dust).
The directional relationships illustrated in the diagram are represented by the following
two equations:
(1) Blood Lead = Floor Dust Lead + Window Sill Dust Lead
(2) Floor Dust Lead = Window Sill Dust Lead
and a covariance matrix of all the variables in the diagram.
31
-------
The directional nature of the diagram is illustrated through the equations by starting with
the highest numbered equation and working upwards, i.e., (2) and then (1). Equation (2)
represents the arrow from window sill dust to floor dust. Equation (1) represents the arrow from
floor dust to blood and the arrow from window sill dust to blood. Finally, the indirect
relationship of window sill dust to blood via floor dust is represented by both equations (1) and
(2) and the covariance matrix.
Evaluating equations (1) and (2) separately is similar to multiple regression or ANOVA,
i.e., both these methods can assess the direct effect of floor dust on blood in equation (1) and
window sill dust on floor dust in equation (2). What these methods cannot assess is the indirect
effect of window sill dust on blood via the floor dust. By evaluating both equations (1) and (2)
simultaneously and accounting for the covariance between these variables, SEM allows for an
assessment of the indirect effect.
SEM has been used to examine pathways of lead in a number of studies. Section 5.1
discusses several examples published in the literature that illustrate pathway models analyzed
using SEM. Section 5.2 presents the various environmental-lead and blood-lead pathway models
assessed in this report using SEM.
Note that the mechanism by which lead is transported from one location to another is
beyond the scope of this report. The analyses are designed only to evaluate the association
between lead at one location with lead at another location, not how the lead is transported via the
statistically significant pathway.
5.1 OVERVIEW OF THE PATHWAYS THAT HAVE BEEN INVESTIGATED IN THE
LITERATURE
Several structural equation models describing the pathways by which environmental lead
exposure occurs during childhood have been published in the literature. Table 5-1 below lists
several papers that were identified as having assessed pathway models associated with an
environmental-lead study. The first column in the table lists the title of the paper and the author,
the second column lists the study from which the analyzed data were obtained, and the third
column identifies the pathway diagram in Appendix D. The diagrams in Appendix D illustrate
all pathways considered in each analysis. The significant pathways are indicated by solid lines
32
-------
Table 5-1. Identified Papers Which Assessed Pathways.
" \^ -. ' \ ' ป ' ->.
" '' : ' N,^ ... ปN ป ' V*"*'- ' * ^ J
Title of Paper /Author(s)
Exterior Surface Dust Lead, Interior House Dust
Lead and Childhood Lead Exposure in an Urban
Environment [2]
Bornschein RL, Succop PA, Kraft KM, Clark CS,
Peace B, and Hammond PB
Soil Lead - Blood Lead Relationship in a Former
Lead Mining Town [17]
Bornschein RL, Clark CS, Grote J, Peace B, Roda S,
and Succop P.
The Influence of Social and Environmental Factors
on Dust Lead, and Blood Lead Levels in Young
Children [1]
Bornschein RL, Succop PA, Dietrich KN, Clark CS,
Que Hee S, and Hammond PB.
Pathways of Lead Contamination for the Brigham
and Womens Hospital Longitudinal Lead Study [7]
Menton R, Burgoon DA, and Marcus AH.
Dust Lead Contribution to Lead in Children [18]
Sayre J
Racial Differences in Urban Children's
Environmental Exposures to Lead [6] Lanphear, BP,
Weitzman, M, and Eberly, S
Pathways of Lead Exposure in Urban Children [19]
Lanphear, BP, and Roghmann, KJ
Evaluation of the HUD Lead-Based Paint Hazard
Control Grant Program [20]
U.S. Department of Housing and Urban
Development, Fifth Interim Report
Reid Study(s) Used to Assess the
Hypothesized Pathways
Cincinnati Lead Study
Telluride, Colorado
Cincinnati Lead Study
Boston Hospital for Women Lead
Study
Rochester Lead Study (1973)
Rochester Lead Study (1993)
Rochester Lead Study (1991-92)
Field studies conducted in
Alameda County, CA, Baltimore,
MD, Boston, MA, California,
Chicago, IL, Cleveland, OH,
Massachusetts, Minnesota, New
Jersey, New York City, NY, Rhode
Island, Vermont, Wisconsin, and
Milwaukee, Wl, during
1996-1997.
Pathway
Diagram
Rgure Number
D-1
D-2
D-3
D-4
D-5
NA
D-6
NA
Note: The Sayre paper and the Lanphear paper on racial differences utilized multiple regression and not
structural equation modeling to analyze the data.
while the pathways not found to be significant in the analysis are indicated by a dotted line.
Tables D-1 through D-6 in Appendix D explain each of the variables used in the pathways
analysis. Because the report Evaluation of the HUD Lead-Based Paint Hazard Control Grant
Program [20] is an interim report, its analyses will not be covered in this report.
Among these studies, pathways of exposure from lead in paint, dust, and soil were often
found to be statistically significant. Moreover, lead dust on children's hands was a statistically
33
-------
significant direct pathway to blood lead in each study where hand lead data were collected. A
direct pathway of lead exposure from renovation and remodeling activities (also called
"Refinishing") to a child's blood-lead concentration was found to be significant in one of the
papers (Figure D-4). Similarly, several child modifiers such as age of child (Figure D-2), pica
habits (Figure D-5 and D-6), and socioeconomic status (Figure D-3) were found to be direct
pathways of lead exposure to blood. A pathway of lead exposure from air was addressed in two
papers (Figure D-4 and Figure D-5). Because each study was designed differently, the pathways
investigated varied somewhat from study to study.
Based on the various pathway models described in the literature and common
perceptions, a general pathway diagram was designed. This diagram is presented in Figure 5-1.
Water
Figure 5-1. General Pathway Diagram Based on Literature Review.
The diagram in Figure 5.1 forms the basis for the pathway models analyzed in this report.
Many of the illustrated pathways could not be included in the analyses because data necessary for
those pathways were not available.
34
-------
5.2 PATHWAYS TO BE INVESTIGATED
As mentioned earlier, the data used for the pathway analysis in this report were collected
in three separate studies with three different objectives. Pieces of information collected in one
study were not necessarily collected in another study. For instance, no blood-lead or water-lead
concentrations were collected in the CAP study; very few soil samples and air duct dust-lead
samples were collected in the R&M study; and no air duct dust-lead samples were collected in
the Rochester study.
Because of the differences in the data collected among the studies, several pathways that
can be tested using data from one study cannot be tested using data from another study. Below
are descriptions of the pathways that were tested. The primary analysis centered around two sets
of pathways models: environmental-lead and blood-lead pathways models. The environmental-
lead pathways were designed to be similar across studies so that comparisons could be made
among the CAP, R&M, and Rochester data. Similarly, the blood-lead pathways for the R&M
and Rochester data were similar to facilitate comparisons.
In addition to the environmental and blood-lead pathway models, several sub-analyses
were conducted. For the R&M data and the Rochester data, sub-analyses were conducted to
compare the effect of using a paint hazard score versus an average XRF measurement and to
assess whether an indicator of carpeting hi the home is informative. The significance of an air
duct dust-lead pathway was addressed using the CAP data and the significance of recent
renovation and remodeling in the home was analyzed using the CAP and R&M study data.
Finally, differences in exposure pathways for African American children and children from all
other races, the effect of hand dust-lead, and the significance of soil vegetative cover were
examined using the Rochester data.
All of the pathway models described below were analyzed using SEM, and the results of
the analyses are presented in Section 6.
5.2.1 Environmental-Lead Pathways Equations
As discussed previously, because of the differences hi data collection, the pathways that
could be statistically evaluated were restricted. Equation Sets 5-1, 5-2, and 5-3 present the
environmental-lead pathway models evaluated using the CAP, R&M, and Rochester data,
35
-------
respectively. To allow comparisons between the studies, the pathways for each study were the
same with one major exception. Because very few samples were collected, the soil pathway was
not included in the R&M model.
Equation Set 5-1. CAP Environmental-Lead Pathways Model
Floor Dust
Interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Interior Entryway Dust + Exterior Entryway
Dust + Window Well Dust + Soil
= Exterior Entryway Dust + Window Sill Dust
+ Soil
= Window Well Dust -I- Soil
= Soil
Dust + Window Sill
+ Window Well Dust
Equation Set 5-2. R&M Environmental-Lead Pathways Model
Floor Dust
= Interior Entryway Dust + Exterior Entryway Dust + Window
Sill Dust -I- Window Well Dust
Interior Entryway Dust = Exterior Entryway Dust + Window Sill + Window Well Dust
Window Sill Dust = Window Well Dust
Equation Set 5-3. Rochester Environmental-Lead Pathways Model
Floor Dust
Interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Interior Entryway Dust -I- Exterior Entryway Dust + Window
Dust + Window Well Dust + Soil
= Exterior Entryway Dust -t- Window Sill Dust + Window Well
Dust + Soil
= Window Well Dust + Soil
= Soil
Sill
5.2.2 Blood-Lead Pathways Equations
Child blood-lead concentrations were not collected in CAP study. Therefore, pathway
models to assess the pathway of environmental lead sources to a child's blood were developed
only for the R&M and Rochester studies and are presented hi Equation Sets 5-4 and 5-5,
respectively.
36
-------
Equation Set 5-4. R&M Blood-Lead Pathways Model
Blood
Floor Dust
Interior Entryway
Dust
Window Sill Dust
= Floor Dust+ Interior Entryway Dust + Exterior Entryway
Dust + Window Sill Dust + Window Well Dust + Water +
Mouthing Behavior
= Interior Entryway Dust + Exterior Entryway Dust + Window
Sill Dust + Window Well Dust
= Exterior Entryway Dust + Window Sill Dust + Window Well
Dust
= Window Well Dust
Equation Set 5-5. Rochester Blood-Lead Pathways Model
Blood
Floor Dust
Interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Floor Dust + Interior Entryway Dust + Exterior Entryway
Dust + Window Sill Dust + Window Well Dust + Water +
Mouthing Behavior
= Interior Entryway Dust + Exterior Entryway Dust + Window
Sill Dust + Window Well Dust + Soil
= Exterior Entryway Dust + Window Sill Dust + Window Well
Dust +Soil
= Window Well Dust + Soil
Soil
As discussed above, no soil pathway is included in the R&M model because of limited
soil data. The main difference between the blood-lead pathways and the environmental-lead
pathways is the inclusion of an equation in the blood pathways model which assesses the direct
effect of the environmental dust-lead levels on the childhood blood-lead concentration. Included
in this equation are the mouthing behavior of the child and water-lead concentration.
5.2.3 Paint-Lead Indicators
Two types of paint indicators were considered for this analysis. The first included paint
hazard scores for interior windows and doors, while the second was the average XRF
measurement for interior windows and doors. This analysis was carried out for the
environmental-lead models for both the R&M data and the Rochester data and for the blood-lead
pathway model for the Rochester data. There were not enough occupied homes with XRF
measurements to be able to carry out the blood-lead pathway model for the R&M data.
37
-------
The CAP data were not included in this paint analysis. In the CAP study, the 17 unabated
homes had relatively few cases of lead-based paint and the 35 abated homes had the lead-based
paint abated two years prior to the environmental sampling.' Equation Sets 5-6 and 5-7 show the
environmental-lead pathway models that included window and door paint for the R&M and
Rochester data, respectively. Equation Set 5-8 shows the Rochester blood-lead pathway model
that included window and door paint.
Equation Set 5-6. R&M Environmental-Lead Pathways Model - Assessing the Impact
of Paint-Lead Pathways.
Floor Dust
^^^^^^^
Interior Entryway
Dust
Interior Entryway Dust + Exterior Entryway Dust + Window Sill
Dust + Window Well Dust + Window Paint + Door Paint
Exterior Entryway Dust + Window Sill Dust + Window Well Dust
+ Window Paint + Door Paint
Window Sill Dust
Window Well Dust + Window Paint
Window Well Dust
Window Paint
Equation Set 5-7. Rochester Environmental-Lead Pathways Model - Assessing the
Impact of Paint-Lead Pathways.
Floor Dust
Interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Interior Entryway Dust + Exterior Entryway Dust + Window Sill
Dust + Window Well Dust + Soil + Window Paint + Door Paint
= Exterior Entryway Dust + Window Sill Dust + Window Well Dust
+ Soil + Window Paint + Door Paint
= Window Well Dust + Soil + Window Paint
= Soil + Window Paint
Equation Set 5-8.
Rochester Blood-Lead Pathways Model - Assessing the Impact of
Paint-Lead Pathways.
Blood
Floor Dust
Interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Floor Dust + Interior Entryway Dust + Exterior Entryway Dust +
Window Sill Dust + Window Well Dust + Water + Mouthing
Behavior + Window Paint + Door Paint
= Interior Entryway Dust + Exterior Entryway Dust + Window Sill
Dust + Window Well Dust + Soil + Window Paint + Door Paint
= Exterior Entryway Dust + Window Sill Dust + Window Well Dust
+ Soil + Window Paint + Door Paint
= Window Well Dust + Soil + Window Paint
= Soil + Window Paint
38
-------
The results from the SEM analysis of Equation Sets 5-6 and 5-7 were compared to
Equation Sets 5-2 and 5-3 to assess the effect the paint pathways have on the environmental-lead
pathway models. Similarly, results for Equation Set 5-8 were compared to results for Equation
Set 5-5. In addition, results for models with paint hazard scores were compared to results for
models with average XRF measurements.
5.2.4 Hand Dust-Lead
The average of pre- and post-interview hand dust-lead levels was included as a pathway
in the Rochester study blood-lead pathway models that included the paint-lead pathways. This
pathways model is similar to Equation Set 5-8, except a separate equation for the average hand
dust-lead level and a pathway from hand dust-lead to blood-lead are included. Equation Set 5-9
illustrates the hand dust-lead model evaluated for the Rochester study data. No hand dust-lead
was collected in the R&M and CAP studies.
Equation Set 5-9. Rochester Blood-Lead Pathways Model - Assessing the Impact of Hand
Dust-Lead.
Blood
Hand Lead
Floor Dust
Interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Hand Lead + Floor Dust + Interior Entryway Dust + Exterior
Entryway Dust + Window Sill Dust + Window Well Dust + Water
+ Mouthing Behavior + Window Paint + Door Paint
= Floor Dust + Interior Entryway Dust + Exterior Entryway Dust +
Window Sill Dust + Window Well Dust + Soil + Mouthing
Behavior + Window Paint + Door Paint
= Interior Entryway Dust + Exterior Entryway Dust + Window Sill
Dust + Window Well Dust + Soil + Window Paint + Door Paint
= Exterior Entryway Dust + Window Sill Dust + Window Well Dust
+ Soil + Window Paint + Door Paint
= Window Well Dust + Soil + Window Paint
= Soil + Window Paint
The results from the SEM analysis of Equation Set 5-9 were compared to the results for
Equation Set 5-8.
5.2.5 Soil-Lead Coverage
A variable indicating the extent of grass covering at the site of each soil sample was
available in the Rochester study data. To assess how the grass covering may affect the pathways
of soil-lead exposure, a variable that was a combination of the soil coverage variable and the
associated soil-lead concentration replaced the soil-lead variable hi the blood-lead pathways
39
-------
model of Equation Set 5-8. A comparison of the parameter estimates of the two sets of models
was made.
5.2.6 Carpeted Floors
Equation sets for an assessment of the effect of carpeting on floor dust-lead loading and
concentration and blood-lead concentration are presented in this section. Equation Sets 5-10 and
5-11 illustrate me models evaluated for the R&M and Rochester data, respectively. These
models are very similar to the blood-lead pathways described hi Equation Sets 5-4 and 5-5, but
they include an indicator of the proportion of floors from which samples were taken that were
carpeted. For the Rochester data, the window and door paint hazard scores and a variable
indicating whether the interior entryway was carpeted were included as pathways. Information
on the presence of carpeting in the interior entryway was not available for the R&M data, and
only a subset of R&M homes were both occupied and had paint information available. Hence
because of these data limitations, the indicator of whether the interior entryway was carpeted and
the door and window paint pathways were not included in the R&M model.
An additional analysis using only homes in the Rochester study in which the floors in the
bedroom and play area were carpeted was conducted. The equations for this pathway analysis
are illustrated in Equation Set 5-12. The results from mis analysis were compared to the results
for Equation Set 5-11 to assess the effect of sampling only from carpeted surfaces.
Equation Set 5-10. R&M Blood-Lead Pathways Model - Assessing the Impact of a Carpeted
Floors Pathway.
Blood
Floor Dust
Interior Entryway
Dust
Window Sill Dust
ป Floor Dust + Interior Entryway Dust + Exterior Entryway Dust
+ Window Sill Dust + Window Well Dust + Water + Mouthing
Behavior + Proportion Carpeted
= Interior Entryway Dust + Exterior Entryway Dust + Window
Sill Dust + Window Well Dust + Proportion Carpeted
= Exterior Entryway Dust + Window Sill Dust + Window Well
Dust
= Window Well Dust
40
-------
Equation Set 5-11.
Rochester Blood-Lead Pathways Model - Assessing the Impact
of a Carpeted Floors Pathway.
Blood
Floor Dust
Interior Entryway
Dust
^^^v^ _^M^BMK^BซBBMMIIBซBlli^HI
Window Sill Dust
Window Well Dust
= Floor Dust + Interior Entryway Dust + Exterior Entryway Dust
+ Window Sill Dust + Window Well Dust + Water + Mouthing
Behavior + Window Paint Hazard Score + Door Paint Hazard
Score + Proportion Carpeted + Indicator Interior Entrance
Carpeted
= Interior Entryway Dust + Exterior Entryway Dust + Window Sill
Dust + Window Well Dust + Soil + Window Paint Hazard
Score + Door Paint Hazard Score + Proportion Carpeted +
Indicator Interior Entrance Carpeted
= Exterior Entryway Dust + Window Sill Dust + Window Well
Dust + Soil + Window Paint Hazard Score + Door Paint Hazard
Score + Indicator Interior Entrance Carpeted
= Window Well Dust + Soil + Window Paint Hazard Score
= Soil
Equation Set 5-12. Rochester Blood-Lead Pathways Model - Assessing the Impact of
Carpeted Bedroom and Play Area Floors.
Blood
Floor Dust (Bed/Play
Carpeted)
Interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Floor Dust(Bed/Play Carpeted) + Interior Entryway Dust +
Exterior Entryway Dust + Window Sill Dust + Window Well
Dust + Water + Mouthing Behavior + Window Paint Hazard
Score + Door Paint Hazard Score + Indicator Interior
Entrance Carpeted
= Interior Entryway Dust + Exterior Entryway Dust + Window
Sill Dust + Window Well Dust + Soil + Window Paint
Hazard Score + Door Paint Hazard Score + Indicator
Interior Entrance Carpeted
= Exterior Entryway Dust + Window Sill Dust + Window Well
Dust + Soil + Window Paint Hazard Score + Door Paint
Hazard Score + Indicator Interior Entrance Carpeted
= Window Well Dust + Soil + Window Paint Hazard Score
Soil
5.2.7 Renovation and Remodeling Activities
A variable indicating whether renovation and remodeling had been performed in the
home six months prior to environmental sampling was collected in the R&M and CAP studies.
In the Rochester study, the occurrence of renovation and remodeling any time twelve months
prior to recruitment into the study was an exclusion criterion.
To compare the effect of recent renovation and remodeling in the R&M and CAP studies,
the renovation and remodeling variable was included in the environmental-lead pathways model.
In addition, to assess the effect of recent renovation and remodeling on a child's blood-lead
41
-------
concentration, the renovation and remodeling variable was included hi the R&M blood-lead
pathways model that included proportion of carpeting.
Equation Set 5-13 and 5-14 show the environmental pathways models; these models are
similar to those presented in Equation Sets 5-1 and 5-2, respectively. Equation Set 5-15 presents
the blood-lead pathway models evaluated for the R&M data. The results from this model were
compared to those from Equation Set 5-10.
Equation Set 5-13. CAP Environmental-Lead Pathways Model - Assessing the Impact of
Recent Renovation and Remodeling Activities.
Floor Dust
Interior Entryway
Dust
Window Sill Dust
Window Well
Dust
= Interior Entryway Dust + Exterior Entryway Dust + Window
Dust + Window Well Dust + Soil + Recent R&R
= Exterior Entryway Dust + Window Sill Dust + Window Well
+ Soil + Recent R&R
= Window Well Dust + Soil
= Soil
Sill
Dust
Equation Set 5-14. * R&M Environmental-Lead Pathways Model - Assessing the Impact of
Recent Renovation and Remodeling Activities.
Floor Dust
= Interior Entryway Dust + Exterior Entryway Dust- + Window Sill
Dust + Window Well Dust + Recent R&R
Interior Entryway
Dust
Exterior Entryway Dust + Window Sill + Window Well Dust
+ Recent R&R
Window Sill Dust = Window Well Dust
Equation Set 5-15. R&M Blood-Lead Pathways Model - Assessing the Impact of Recent
Renovation and Remodeling Activities.
Blood
Floor Dust
Interior Entryway
Dust
Window Sill Dust
= Floor Dust + Interior Entryway Dust + Exterior Entryway Dust
+ Window Sill Dust + Window Well Dust + Water + Mouthing
Behavior + Proportion Carpeted + Recent R&R
= Interior Entryway Dust + Exterior Entryway Dust + Window Sill
Dust + Window Well Dust + Proportion Carpeted + Recent
R&R
= Exterior Entryway Dust + Window Sill Dust + Window Well
Dust + Recent R&R
= Window Well Dust
42
-------
5.2.6 Race
Using the Rochester data and multivariate regression, Lanphear et al. [6] reported a racial
difference in the sources of environmental-lead exposures in urban children. A pathways
analysis, subsetting the Rochester data into two racial groups, African-American children and
children of all other races, was performed to estimate whether there is a difference in the
pathways of lead exposure between different race groups. The model used in this analysis is
described in Equation Set 5-16.
Equation Set 5-16. Rochester Blood-Lead Pathways Model - Assessing Pathways for
Different Races.
Blood
Floor Dust
interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Floor Dust + Interior Entryway Dust + Exterior Entryway Dust
+ Window Sill Dust + Window Well Dust + Water + Mouthing
Behavior + Window Paint Hazard Score + Door Paint Hazard
Score
= Interior Entryway Dust -f- Exterior Entryway Dust + Window Sill
Dust + Window Well Dust + Soil + Window Paint Hazard
Score + Door Paint Hazard Score
= Exterior Entryway Dust + Window Sill Dust + Window Well
Dust + Soil + Window Paint Hazard Score + Door Paint Hazard
Score
= Window Well Dust + Soil + Window Paint Hazard Score
= Soil + Window Paint Hazard Score
5.2.7 Air Ducts
No air duct dust samples were collected in the Rochester study and very few samples
were collected in the R&M study. Only the CAP study had enough air duct dust samples to
perform a pathways analysis. Equation Set 5-17 presents the pathways to evaluate the effect of
air duct dust-lead. The pathways are very similar to the pathways presented in Equation Set 5-1.
The results from mis analysis were compared to the results of the Equation Set 5-1 analysis.
Equation Set 5-17. CAP Study Environmental-Lead Pathways Model-Assessing the Effect
of an Air Duct Dust-Lead Pathway.
Floor Dust
Interior Entryway
Dust
Window Sill Dust
Window Well Dust
= Interior Entryway Dust + Exterior Entryway Dust + Window
Sill Dust + Window Well Dust + Soil + Air Duct Dust
= Exterior Entryway Dust + Window Sill Dust + Window Well
Dust + Soil + Air Duct Dust
= Window Well Dust + Soil
= Soil
43
-------
6.0 FINDINGS AND RESULTS
Presented in this section are the results and findings from the analysis of the pathway
models illustrated in Section 5. Detailed tables of the results for the CAP, R&M, and Rochester
data are presented in Appendices A, B, and C, respectively. Note that log-transformed dust-lead
loadings and concentrations, soil-lead concentrations, and blood-lead concentrations are used in
the analyses.
6-1 CORRELATION ANALYSIS RESULTS
Prior to performing the structural equation modeling, an assessment of the correlation
structure of the data included in the analysis was conducted for each study individually. The
Pearson correlation coefficients for the CAP data are presented in Tables A-4a and A-4b. Tables
B-7a and B-7b list the correlation results for the R&M data and Tables C-l la and C-l Ib present
correlation results for the Rochester data. Note that the first table for each study contains
correlations for the dust-lead loadings while the second table includes correlations for the dust-
lead concentrations. All other variables are the same.
After a visual assessment of the CAP study data, it was decided that data from both
abated and unabated homes in the CAP study would be included in the pathway analysis. The
visual assessment indicated similar relationships among the media for both types of homes.
The largest correlations for the CAP data occurred among the three soil-lead
concentrations: exterior entryway and foundation soil had a Pearson correlation coefficient (r) of
0.53; exterior entryway and boundary soil had a coefficient of 0.64; and foundation and boundary
soil had a correlation of 0.56. Other correlated variables included the window well and window
sill dust-lead loadings (r = 0.55) and concentrations (r = 0.59) and the interior and exterior
entryway dust-lead concentrations (r = 0.56). Because of the high correlation seen among the
soil samples, a principal components analysis of the three log-transformed soil-lead
concentrations was performed to represent the information from the soil samples with a single
linear combination. Over seventy percent of the variability in the soil samples was explained by
the first principal component which weighted each soil sample nearly equally. Therefore a mass-
weighted average soil-lead concentration of exterior entryway, foundation, and boundary
soil-lead concentration was used in the analysis. Table A-5 in Appendix A lists the results of the
principal components analysis.
44
-------
Similarly, a decision on whether to include the control homes in the analysis with the
R&M homes was made. Figures 6-1 and 6-2 illustrate the relationship between blood-lead
concentration and floor dust-lead loading and concentration, respectively. Though Table 4-4
shows that the control homes have consistently lower floor dust-lead levels and blood-lead
concentrations than the R&M homes, the figures illustrate that the relationships between the
floor dust-lead levels and the blood-lead concentrations are similar for both types of homes.
Hence both sets of homes were included in the analysis.
1000.0 i
100.0-
10.0:
1.0:
0.1-I
o o
o
ฐ0
0*0
10
100 1000 10000 100000
Floor Dust-Lead LoadingCug/fC)
Control Homes
o o o RM Homes
1000000 10000000
Figure 6-1. Relationship of Blood-Lead Concentration U/g/dL) and Average Floor Dust-
Lead Loading (//g/ft2) for the R&M Data.
45
-------
1000.0 i
100.0:
10.0^
O.H
TTT
10
100
1000
10000
100000
1000000 10000000
Floor Dust-Lead Concentration(ag/gm)
Control Homes
0 ฐ ฐ RM Homes
Figure 6-2. Relationship of Blood-Lead Concentration (//g/dL) and Average Floor Dust-
Lead Concentration (//g/g) for the R&M Data.
The R&M analysis included the dust-lead concentrations and loadings and water-lead
concentration. The largest correlations occurred between window well and window sill dust
loadings (r = 0.84) and concentrations (r = 0.83). Blood-lead was moderately correlated wife the
dust measures (r was between 0.34 and 0.50 for loadings and between 0.40 and 0.56 for
concentrations).
The highest correlations in the Rochester data occurred between window well and
window sill dust loadings (r = 0.56) and window well and window sill dust concentrations
(r=0.55). Blood-lead was somewhat correlated with sills and wells (r between 0.34 and 0.37 for
loadings and between 0.21 and 0.24 for concentrations), soil (r = 0.37), and door paint hazard
score (r =0.36).
46
-------
6.2 ENVIRONMENTAL-LEAD PATHWAYS RESULTS
The environmental-lead pathway models, presented in Equation Sets 5-1,5-2, and 5-3 of
Section 5 were analyzed for the CAP Study, R&M, and Rochester data, respectively. Parameter
estimates for each of the data sets are listed in Tables 6-1,6-2, and 6-3 for the CAP study, R&M,
and Rochester data, respectively. Note that two sets of models were run for each study. One set
of models utilized dust-lead loadings and the other employed dust-lead concentrations. All other
variables remained the same in the two analyses. For comparison purposes, the significant
pathways in the dust loading and the dust concentration models are illustrated below in Figures
6-3 and 6-4.
Across all three studies, the dust loading models indicated one consistent, statistically
significant direct pathway of lead contamination: window well dust directly impacted the
window sill dust. An additional statistically significant direct pathway from the interior
entryway dust to the floor dust was observed in the R&M analysis and the Rochester analysis.
hi the R&M Study, a statistically significant indirect pathway of lead contamination from
exterior entryway dust to floor dust through ulterior entryway dust was observed. In both the
CAP and Rochester data, the indirect pathway of lead contamination from soil to window well
dust to window sill dust was found to be statistically significant. A statistically significant
indirect pathway from window well dust to window sill dust to ulterior entryway dust to floor
dust was observed in R&M, while window well dust indirectly impacted floor dust through
window sill dust and soil indirectly impacted floor dust via interior entryway dust, as well as
through window wells and sills, in Rochester.
Similar statistically significant pathways were observed when the dust-lead
concentrations were included in the analysis. For all three data sets, two significant direct
pathways, 1) window well dust to window sill dust and 2) exterior entryway dust to interior
entryway dust, were observed. For the CAP and Rochester studies, a statistically significant
indirect pathway of lead contamination from soil to window well dust to window sill dust was
observed. Note that when the dust-lead concentrations were included hi the analysis the direct
pathway from interior entryway dust to floor dust was no longer statistically significant for the
R&M data.
47
-------
Using the parameter estimates presented in Tables 6-1,6-2, and 6-3, the estimated
decreases in environmental lead-loadings and concentrations when the geometric mean of each
environmental variable is decreased by 50% were calculated and provided in Tables 6-4, 6-5, and
6-6 below.
Table 6-1. Structural Equation Modeling Results for the CAP Study Environmental
Pathways Dust-Lead Loadings if/gift2) and Dust-Lead Concentrations U/g/g).
Variables
^NW Independent
Dependent ^-s,
Direct Effect Parameter Estimates (t-value)
Interior
Entryway
Dust
* i
Window
Sill Dust
Window
Well Dust
Exterior
Entryway
Dust
1
'- Soil
-' R* r ''
Dust-Lead Loadings toig/ft2) - ' ' * .-
Floor
Interior Entryway Dust
Window Sill
Window Well
f *
Floor
Interior Entryway Dust
Window Sill
Window Well
0.2942
(1.91)
0.2275
(1.15)
0.1504
(1.05)
0.2330
(1.78)
-0.0610
(-0.42)
-0.0109
(-0.08)
0.5291*
(4.06)
0.1224
(0.75)
0.2145
(1.43)
Dust-Lead Concentrations U/g/g)
-0.1807
(-1.45)
0.0007
(0.01)
0.2647*
(2.18)
0.0207
(0.23)
0.5389*
(4.56)
-0.0838
(-0.48)
0.4580*
(4.09)
0.0979
(0.28)
0.3512
(1.08)
0.4513
(1.29)
0.8059*
(2.15)
" ,''
0.1694
10.71)
0.1960
(1.13)
0.0810
(0.31)
0.8479*
(2.81)
0.17
0.18
0.33
0.09
-.'>
0.16
0.34
0.35
0.14
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-values * 1.96 and
<; -1.96 are significant at the 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/
concentrations and soil-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9996, and 0.9934 for the
dust-lead concentration model.
48
-------
Table 6-2. Structural Equation Modeling Results for the R&M Environmental Pathways
Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (j/g/g).
Variables
^Slndependent
Dependent >^
5 , Direct Effect Parameter Estimates (t-value)
i~
* * **
Interior
Entryway,
Dust
'V->
h
^<1
'^
Window
Sill Dust
Window
Well Dust
Exterior
Entryway
Dust
Dust-Lead Loadings (pg/ft2)
Floor
Interior
Entryway
Window Sill
0.2114*
(2.77)
0.1577
(1.71)
0.3943*
(3.13)
0.1507
(1.75)
-0.1432
(-1.17)
0.7958*
(13.86)
0.0325
(0.47)
0.2210*
(2.30)
"^ *,
Rz
0.46
0.25
0.70
:-?y- -,..,,:- . .:...,.:';' \^;:V:;^v. '' Dust-Lead Concentrations (fjgfg) /,", ,, " > ,
Floor
Interior
Entryway
Window Sill
0.2035
(1.88)
0.3082*
(3.51)
0.1870*
(2.16)
0.1737
(1.71)
-0.0415
(-0.40)
0.9059*
(13.38)
-0.0281
(-0.29)
0.5486*
(6.95)
0.59
0.57
0.68
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value * 1.96 or
ฃ -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/
concentrations and soil-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9569, and 0.9001 for the dust-
lead concentration model.
49
-------
Table 6-3. Structural Equation Modeling Results for the Rochester Environmental
Pathways Dust-Lead Loadings (//g/ftz) and Dust-Lead Concentrations (//g/g).
Variables
?\lndependent
Dependent ^v
Direct Effect Parameter Estimates (t-value)
Interior
Entryway
Dust
! f
i
Window
Sill Dust
i ',
Window
Well Dust
i 1
Exterior
Entryway
Dust
Soil
,' * -,ป >si
i t
i
- R2
- Dust-Lead Loadings (j/g/ft2)
Floor
Interior
Entryway
Window Sill
Window Well
0.1743*
(3.29)
0.1795*
(2.73)
0.0077
(0.07)
-0.0118
(-0.22)
0.0595
(0.66)
0.4121*
(6.56)
0.0383
(0.62)
0.1950
(1.95)
0.1424
(1.24)
0.4767*
(2.58)
0.2057
(1.4067)
1.1057*
(6.20)
0.21
0.12
0.34
0.23
Dust-Lead Concentrations U/g/g) ',', . ' - - '
Floor
Interior
Entryway
Window Sill
Window Well
0.3795*
(5.17) '
0.2002*
(3.14)
-0.0284
(-0.37)
-0.0338
(-0.54)
0.1862*
(2.50)
0.4695*
(6.26)
-0.0828
(-1 .24)
0.2006*
(2.54)
-0.0938
(-0.92)
0.0942
(0.76)
0.1591
(1.15)
0.9787*
(7.05)
0.25
0.16
0.35
0.28
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value *1.96 or
^-1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/
concentrations and soil-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9878, and 0.9869 for the
dust-lead concentration model.
50
-------
CAPS
(a) Soil
fi&M
Window Well
Dust
Exterior
Entryway Dust
Rochester
Soil
Window Well
Dust
Window Sill
Dust
Interior
Entryway Dust
.^.Window.
Well
Window
Sill
Interior
Entryway Dust
Window Sill
Dust
Floor Dust
* Floor Dust
Figure 6-3. Significant Pathways for the CAP, R&M, and Rochester Environmental-Lead
Pathway Models Dust-Lead Loadings (//g/ft2).
51
-------
CAPS
la) "nil * Window Well ^ pioor Du3t
Dust
i
Window Sill
Dust
(b) Exterior Entryway Interior Entryway
Uust " '" * Dust
R&M
Exterior Entryway
Dust
Window Well ^ Window Sill Interior Entryway
Dust * Dust * Dust
Rochester
Exterior
Entryway Dust
4
Qrtil - fc \A/inrfn\A/ \A/pll h lrป1-r>rTr*r
1 Entryway Dust
Window Sill > Floor Dust
Figure 6-4. Significant Pathways for the CAP, R&M, and Rochester Environmental-Lead
Pathway Models Dust-Lead Concentrations O^g/g).
52
-------
Table 6-4. Predicted Effect of a 50% Decrease in Environmental-Lead Loadings (j/g/ft2) and
Concentrations (//g/g) Based on the Environmental-Lead Pathways SEM for the
CAP Study Data.
*-l * *
5 "I" l
J
Sample Location
^
Interior Entryway
Window Sill
Window Well
Exterior Entryway
Soil1"
01 -;v ;
i
Geometric
Mean
k
342
107
3,230
574
157
i
50%
Decrease in
GM
Predicted Percent Change (%)
(95% Lower Prediction Bound,
95% Upper Prediction Bound) tw
Floor
Dust-Lead Loading (/u
171
54
1,615
287
79
-18 (-40, 13)
-14 (-36, 16)
-4 (-29, 30)
-12 (-37, 24)
-21 (-62, 63)
Interior
Entryway
Window
Sill
Window
Well
3/ft*)
NA
-15 (-34, 10)
-7 (-29, 22)
-14(-36, 16)
-32 (-64, 30)
NA
NA
-31 (-48, -8)
NA
-46I-75, 17)
> - > n. >
-. _.j---- -ฃซ ,- . ! , " Dust-Lead Concentration (//g/g) ,
Interior Entryway
Window Sill
Window Well
Exterior Entryway
Soil1"
201
778
1,577
261
157
101
389
789
131
79
-15 (-35, 12)
-13 (-27, 3)
-1 1 (-25, 5)
-7 (-27, 19)
-21 (-43, 10)
NA
0{-9, 10)
-1 (-10, 9)
-27 (-35, 18)
-14 (-28, 3)
NA
NA
-31 (-43, 16)
NA
-31 (-55, 6)
NA
NA
NA
NA
-43 (-77, 44)
"ซ*
NA
NA
NA
NA
-44 (-69, 2)
(al Average soil-lead concentration in //g/g.
"* The prediction intervals or forecasting intervals presented are confidence intervals for the actual or future
value of a response. Note that the upper and lower 95% prediction bounds are based on the direct effects
only.
NA Indicates that the fitted pathway model did not include this variable.
As seen in Table 6-4, for the CAP data, the largest reduction in floor dust loading or
concentration, a 21% reduction, occurred when the geometric mean of the soil was reduced by
50%. Similar results were seen for floor dust loading in the Rochester model, as shown in
Table 6-6. For the Rochester floor dust concentration, only a 6% reduction in concentration was
seen when the soil was reduced by 50%, but a 23% reduction was seen when the interior
entryway dust concentration was lowered. Generally, for both the CAP and Rochester data, a
50% reduction in the geometric mean soil-lead concentration produced the largest reductions in
lead levels for floors, interior entryways, window sills, and window wells, ranging from a 6%
decrease in floor dust concentration to a 54% decrease in window well dust loading for the
Rochester data, and a 14% decrease in interior entryway dust concentration to a 46% decrease in
window sill dust loading for the CAP data.
53
-------
Table 6-5. Predicted Effect of a 50% Decrease in Environmental-Lead Loadings (i/g/ft2) and
Concentrations (j/g/g) Based on the Environmental-Lead Pathways SEM for the
R&M Data.
Sample Location .
i
Geometric
Mean
50% Decrease, in
, GM
1 : Predicted Percent Change (%) ,
' (95% Lower Prediction Bound, , .
i . ' - / 95% Upper Prediction Bound) ""
Floor
Dust-Lead Loading (//g/ft2)
Interior Entryway
Window Sill
Window Well
Exterior Entryway
329
1,229
37,035
405
165
615
18,518
203
-14 (-25, -2}
-15 (-28, 0)
-19 (-30, -6)
-5 (-16, 7)
Interior , "*
Entryway _,
Window
Sill , :\
' i
i
NA
-24 (-44, 4)
-11 (-35,21)
-14 (-32, 9)
NA
NA
-42 (-49, -34)
NA
,' " ' Dust-Lead Concentration (j/g/g)
Interior Entryway
Window Sill
Window Well
Exterior Entryway
1,459
5,411
8,452
1,570
730
2,706
4,226
785
-13 (-24, 0)
-21 (-30, -11)
-28 (-37, -18)
-6 (-17, 7)
NA
-12 (-22, -1)
-8 (-20, 5)
-32 (-39, -25)
NA
NA
-47 (-53, -41)
NA
(al The prediction intervals or forecasting intervals presented are confidence intervals for the actual or future
value of a response. Note that the upper and lower 95% prediction bounds are based on the direct effects
only.
NA Indicates that the fitted pathway model did not include this variable.
Because very few soil samples were available for analysis from the R&M homes, a
comparison to the CAP and Rochester reductions was difficult. In R&M study, the exterior
entryway dust, considered to be a surrogate for soil, does show decreases in other environmental
variables when its geometric mean was reduced, although not to the same degree as seen for soil
in the CAP and Rochester studies. The largest decreases in floor dust-lead levels, 28% and 21%,
occur when window well dust-lead concentration and window sill dust-lead concentration were
decreased.
54
-------
Table 6-6. Predicted Effect of a 50% Decrease in Environmental-Lead Loadings (//g/ft2) and
Concentrations (//g/g) Based on the Environmental-Lead Pathways SEM for the
Rochester Data.
r > '** *
, VJ,-*f. _ , ,
.-vo^'*"* ""x '
^ ,-. ' * ซซ - , '
Viป A_t- ,, m,v,i
'*'-"' A *" " ^
*X* x" v**^u^^'
.-^ ^
Sample Location
-
Interior Entryway
Window Sill
Window Well
Exterior Entryway
Soil"1
assaaaa
w
Geometric
Mean
=====
50%
Decrease in
GM
Predicted Percent Change (%)
(95% Lower Prediction Bound,
95% Upper Prediction Bound) IW
Floor ,
Dust-Lead Loading (u
88.6
345
22,584
521
852
44.3
173
11,292
258
426
-11 (-20, -1)
-12(-23, 0)
-5 (-15, 6)
-5(-16,7)
-21 (-37, -1)
Interior '
Entryway
Window
Sill
3/ft2)
NA
-1 (-30, 40)
-4 (-28, 28)
-13 (-37, 20)
-32 (-62, 23)
NA
NA
-25 (-36, -12)
NA
-37 (-57, -8)
Window
Well
NA
NA
NA
NA
-54 (-76 -13)
. Dust-Lead Concentration (j/g/g) ,
Interior Entryway
Window Sill
Window Well
Exterior Entryway
Soil"1
468
2,787
8,676
656
852
234
1,394
4,338
328
426
-23 (-32, -13)
-12 (-21, -2)
-8(-17, 2)
0(-10, 12)
-6(-20, 11)
NA
2 (-13, 19)
-1 1 (-23, 3)
-13 (-26, 2)
-161-34,7)
NA
NA
-28(-39, -14)
NA
-35 (-53, -11)
NA
NA
NA
NA
-49 (-65, -25)
M Average soil-lead concentration in //g/g.
"" The prediction intervals or forecasting intervals presented are confidence intervals for the actual or future
value of a response. Note that the upper and lower 95% prediction bounds are based on the direct effects
only.
NA Indicates that the fitted pathway model did not include this variable.
6.3 BLOOD-LEAD PATHWAYS RESULTS
The results of the SEM for the blood-lead pathway models illustrated in Equation Sets
5-4 and 5-5 are presented in Tables 6-7 and 6-8 using the R&M and Rochester data, respectively.
The significant pathways observed in Tables 6-7 and 6-8 are diagramed in Figures 6-5 and 6-6
for easy comparison.
Four similar direct pathways of lead exposure/contamination were observed in Figure 6-5
(i.e., dust-lead loading model): 1) window well dust to window sill dust, 2) interior entryway
dust to floor dust, 3) floor dust to blood, and 4) mouthing habits of the child to the blood. For the
Rochester data, there was another significant direct pathway from window well dust to blood.
No additional significant direct pathways to blood were observed in the R&M data.
55
-------
Table 6-7. Structural Equation Modeling Results for the R&M Blood-Lead Pathways
Dust-Lead Loadings (/ig/ft2) and Dust-Lead Concentrations U/g/g).
Variables > -
N^ Independent
Dependent >v
Direct Effect Parameter. Estimates (t-value) ' .
Interior
Entryway
Dust
Window
Sill Dust
- ..
Blood
Floor
Interior
Entryway
Window Sill
0.0591
(1.57)
0.2114*
(2.77)
-0.0089
(-0.20)
0.1577
(1.71)
0.3943*
(3.13)
Window
Well Dust
Exterior
Entryway
Dust
l
Floor
Dust
Water
?
Mouthing
Dust-Lead Loadings
-------
Table 6-8. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways -
Dust-Lead Loadings (j/g/ft2) and Dust-Lead Concentrations fo/g/g).
^^=^^=s=
Variables
\lndependent
Dependent N^
i ~^^^^^^^^^^^^^""11'^^^^^^^^^^^^^B^^HiH3S5S55S^5HB55SS^SSS^^5SS3i!
Direct Effect Parameter Estimates (t-value) -,
<
Interior
Entryway
Dust
^
v ' !
'Window
Sill Dust
.'*-'' *
Blood
Floor
Interior
Entryway
Window Sill
Window Well
Blood
Floor
Interior
Entryway
Window Sill
Window Well
-0.0275
(-1.26)
0.1813ป
(3.39)
0.0233
(0.87)
0.1726*
(2.61)
0.0146
(0.13)
Window
Well Dust
Exterior'
Entryway
Dust
Dust-Lead Loadinc
0.0620*
(2.97)
-0.0182
(-0.32)
0.0748
(0.81)
0.4186*
(6.44)
-0.0101
(-0.41)
0.0406
(0.66)
0.1909
(1.90)
[
Soil
Floor
Dust
Water
js (/ug/ft2)
0.1671
(1.41)
0.4273*
(2.25)
0.2067
(1.37)
1.1669*
(6.64)
>- , Dust-Lead Concentrations (jjgl
-0.0283
(-0.67)
0.3854*
(5.20)
0.0273
(0.80)
0.1917*
(3.00)
-0.0196
(-0.25)
0.0386
(1.27)
-0.0437
(-0.69)
0.1794*
(2.39)
0.4595*
(6.00)
0.0204
(0.59)
-0.0862
(-1.29)
0.1948*
(2.46)
-0.0596
(-0.57)
0.0868
(0.69)
0.1884
(1.33)
1.0160*
(7.30)
0.1096*
(3.19)
-0.0691
(-0.64)
, i $
Mouthing
s^s^s^
f
Rz
'"-.."; :&.
0.3933*
(3.11)
0.27
0.22
0.12
0.35
0.25
a) . ,
0.0654
(1.40)
0.0083
(0.07)
0.4788*
(3.37)
0.15
0.25
0.16
0.35
0.30
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value 2 1.96 or
ฃ -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/
concentrations, soil-lead concentrations, and blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The gqodness-of-fit index (GFI) for the dust-lead loading model is 0.9751, and 0.9661 for the
dust-lead concentration model.
57
-------
R&M
Exterior
Entryway
Dust Mouthing
1 1
Window Well Window Sill Interior cioor
Dust * Dust > Entryway * Dust * B'ฐฐd
Dust
Rochester
Interior
__- -*" Entryway Dust
Soil > Window Well Window Sill Roor
Dust * Dust * Dust * | r
Mouthing
-
Figure 6-5. Significant Pathways for the R&M and Rochester Blood-Lead Pathway
Models Dust-Lead Loadings {//g/ft2}.
A comparison of Figure 6-6 (i.e., dust-lead concentration model) to Figure 6-5 (i.e., dust-
lead loading model) shows that there were some common statistically significant pathways in the
concentrations and loadings models. For instance, the R&M data had three direct pathways that
were the same for dust loadings and concentrations: 1) exterior entryway dust to interior
entryway dust, 2) window well dust to window sill dust, and 3) window sill dust to interior
entryway dust. The Rochester data shows three pathways which were the same for the dust
loadings and concentrations: 1) children's mouthing to blood, 2) interior entryway dust to floor
dust, and 3) soil to window well dust to window sill dust to floor dust.
58
-------
R&M_
Exterior Entryway
Interior Entryway
Dust
Blood
Window Well
Dust
Window Sill
Dust
Floor Dust
Rochester
(1) Mouthing
(2) Soil
Exterior Entryway
Dust
Blood
Window Well
Dust
Window Sill
Dust
I
I
Interior Entryway
Dust
Floor Dust
Figure 6-6. Significant Pathways for the R&M and Rochester Blood-Lead Pathway
Models Dust-Lead Concentrations (ug/g).
One notable difference between the loading and concentration models occurred in the
Rochester data. In the loading model the floor dust, window well dust, and child mouthing
directly impacted the blood. In the concentration model only the child mouthing directly
impacted the blood.
Tables 6-9 and 6-10 list the predicted effect of a 50% decrease in the geometric means of
the environmental variables on the blood-lead concentrations and the environmental-lead
loadings and concentrations. The parameter estimates from Tables 6-7 and 6-8 were used to
calculate the predictions. Also listed in the table are the 95% prediction intervals about the
predicted effect. The prediction intervals are based only on the direct effects.
59
-------
Table 6-9. Predicted Effect of 50% Decrease in Environmental-Lead Loadings (pg/ft2) and
Concentrations (//g/g) and Blood-Lead Concentrations d/g/dL) Based on the
Blood-Lead Pathways SEM for the R&M Data.
Sample Location
Geometric
Mean
50% Decrease
, in GM
Predicted Percent Change (%}
(95% Lower Prediction Bound,
95% Upper Prediction Bound) lb>
Blood
; Floor
Interior
Entryway
Window
Sill
Dust-Lead Loadings (jug/ft2)
Floor
Interior Entryway
Window Sill
Window Well
Exterior Entry
/
Mouthing""
Water""
t >? -
Floor
Interior Entryway
Window Sill
Window Well
Exterior Entry
'-,$"":'" r vi:-."i-' - ' -: "' .:':. 'i.
Mouthing1"'
Water"1
210
329
1,229
37,035
405
Exposure
1
1
105
165
615
18,518
203
No Exposure
0
0
-8(-12, -4)
-6 (-9, -3)
-3 (-6, 1)
-6 (-9, -3)
-3 (-6, 0)
NA
-14(-25,2)
-18 (-30, 4)
-14 (-26, 0)
-51-16,7)
NA
NA
-231-44,5)
-19 (-40, 10)
-14 (-32, 9)
NA
NA
NA
-43 (-50, -35)
NA
^
-30 (-40, -19)
4 (-6, 15)
NA
NA
NA
NA
NA
NA
Dust-Lead Concentrations (/yg/g) ' " &
1,118
1,459
5,411
8,452
1,570
'' Exposure
1
1
559
730
2,706
4,226
785
No Exposure
0
0
-8 (-13, -3)
-12(-16, -7)
-8(-12, -4)
-6(-11,-1)
-7 (-11, -3)
NA
-131-24,0)
-21 {-30, -11)
-201-30, -9)
-61-17,8)
NA
NA
-12 (-22, 1)
-1 1 (-22, 2)
-32 (-39, -25)
NA
NA
NA
-47 (-53, -41)
NA
- >,'< ! ,, . K * 1 f
<* -*% , ^-- r +. ^
-26 (-35, -16)
9(0, 19)
NA
NA
NA
NA
NA
NA
'*' Variables were treated as categorical.
*' The prediction intervals or forecasting intervals presented are confidence intervals for the actual or future
value of a response. Note that the upper and lower 95% prediction bounds are based on the direct effects
only.
NA Indicates that the fitted pathway model did not include this variable.
60
-------
Table 6-10. Predicted Effect of 50% Decrease in Environmental-Lead Loadings (j/g/ft2) and
Concentrations fo/g/g) and Blood-Lead Concentrations (//g/dL) Based on the
Blood-Lead Pathways SEM for the Rochester Data.
^SB=^==
Sample
Location
BSB^S
Geometric
Mean
50%
Decrease
in GM
===!=:===^^===s=!=^===^=i^^=^=^^=^^=^=:^=^^=^^=i
Predicted Percent Change {%)
(95% Lower Prediction Bound, 95% Upper Prediction Bound)""
Blood
Floor
_
Window
Sill
Window
Well
>j AKl Dust-Lead Loading U/g/ft2)
Floor Dust
Interior
Entryway
Dust
Window Sill
Window Well
Soil"1
Exterior Dust
S,- .' "* A
Mouthing (W
Water""
fn "ซ* "jig.
Floor Dust
Interior
Entryway
Dust
Window Sill
Window Well
Soil
Exterior Dust
- ., *
Mouthing (b)
Water
100
89
345
22,584
852
515
Exposure
1
1
50
45
173
11,292
426
258
No
; Exposure
0
0
-7 (-9, -5)
-1 (-3, 1)
-3 (-5, -1)
-4 (-6, -2)
NA
0 (-2, 2)
-<-:
-33 (-39, -26)
7 (-2, 16)
NA
-12 (-21, -2)
-11 (-22,1)
0(-11, 11)
-13 (-31, 10)
-5 (-16, 7)
-
NA
NA
NA
NA
-1 (-30, 40)
-5 (-29, 28)
-7 (-50, 71)
-12 (-36, 21)
NA
NA
NA
NA
NA
-25 (-37, -11)
-38 (-58, -8)
NA
NA
NA
NA
NA
NA
NA
-55 (-76, r17)
NA
-
NA
NA
Dust-Lead Concentrations (i/g/g)
563
468
2,787
8,676
852
656
1
1
282
234
1,394
4,338
426
328
No;
0
o
-4 (-8, 0)
0 (-4, 4)
-31 (-33, -29)
-21 (-23, 19)
NA
-1 (-4, 2)
NA
-23 (-32, -13)
-12 (-21, -2)
-11 (-20, -1)
-13 (-27, 3)
1 (-9, 13)
"
-38 (-45, -30)
-1 (-11 10)
NA
NA
NA
NA
1
-121-24,2)
-11 (-31, 14)"
-13 (-26, 2)
" '""-
NA
NA
NA
NA
NA
0(-16, 19)
-51 (-65, -32)
NA
<
NA
NA
NA
NA
NA
NA
-51 (-66, -29)
NA
NA
NA
(al Soil lead is measured as a concentration.
Variables were treated as categorical.
(ci The prediction intervals or forecasting intervals presented are confidence intervals for the actual or future
value of a response. Note that the upper and lower 95% prediction bounds are based on the direct effects
only.
NA Indicates that the fitted pathway model did not include this variable.
61
-------
Tables 6-9 and 6-10 show that for both data sets, the largest decreases in blood-lead
occurred when the mouthing variable was changed from exposure to no exposure, ranging from
an estimated 26% decrease to a 38% decrease. Reduction of all other variables, in general,
produced lower decreases in blood-lead levels, less than 12% and typically hi the single digits.
In a few cases there were no decreases or small positive increases in the blood-lead
concentration. Two exceptions to the general rule of small predicted percent change occurred in
the concentration model for the Rochester data. Reducing the geometric mean window sill dust-
lead concentration by 50% resulted in a 31% reduction in blood-lead concentration. Reducing
the geometric mean window well dust-lead concentration by 50% resulted in a 21% reduction in
blood-lead concentration.
6.4 RESULTS OF ASSESSING THE PAINT-LEAD INDICATORS
The results of two analyses that assessed two types of paint-lead indicators are presented
below. The first set of results assessed the impact of paint in the R&M and Rochester data
environmental pathway models described in Equation Sets 5-2 and 5-3. The second set of
analysis results assessed the impact of paint in me Rochester data blood-lead pathway model
illustrated in Equation Set 5-8.
The parameter estimates for the environmental pathway models for R&M and Rochester
are given in Tables B-l and B-2 of Appendix B and Tables C-l, C-2, C-3, and C-4 of
Appendix C for the R&M and Rochester data, respectively.
Hazard Score versus XRF Measurement
The tables listed above show several differences in the statistically significant pathways
when the paint pathway in the model was the paint hazard score versus the average XRF
measurement. As shown in Table B-l, when the hazard scores were used in the R&M loading
model, both the window paint and the door paint were statistically significant pathways of lead to
floor dust. However, when the average XRF measurement was used in the R&M loading model,
there were no statistically significant pathways to floor dust, as shown in Table B-2. In the R&M
concentration model, the statistically significant pathways were generally the same regardless of
the type of paint pathway.
62
-------
For the Rochester data, comparisons of Tables C-l to C-2 and C-3 to C-4 show that
significant pathways were generally the same when either paint hazard score or average XRF
measurement was used in the model.
The small sample size available for the R&M data may explain the differences observed
in the R&M results. Since the hazard score is thought to be more representative of the paint
hazard present at the time of blood sampling and since the significant pathways in the Rochester
analysis were generally the same whether hazard score or average XRF measurement was
included, the paint hazard scores were included in all subsequent models that had a paint
pathway.
Note that in some cases estimated coefficients for the XRF variables or the paint hazard
scores were statistically significant and negative. The negative parameter estimates may seem
non-intuitive since paint-lead has been shown to be a leading cause for increases in blood-lead
concentration. Perhaps these negative parameter estimates are due to not having substrate
corrected XRF measurements, not sampling all painted surfaces in the residence, or having
subjectivity in the evaluation of paint condition.
Environmental-Lead Pathways
Figures 6-7 and 6-8 present the significant pathways for the environmental pathway
model paint assessment using results for the models which included paint hazard score as
pathways. The figures show that there were far more significant pathways for the Rochester data
than for the R&M data. This may be due to the fact that there were only 36 homes available for
paint analysis in the R&M data and there were up to 205 homes available for analysis in the
Rochester data.
For the Rochester data, window paint directly impacted window sill and window well
dust loading and concentration and floor dust concentration. Door paint directly impacted
interior entryway dust loading and concentration. All other significant pathways remain
unchanged to those illustrated in Figures 6-3 and 6-4.
63
-------
For the R&M data, window paint directly impacted the floor dust loading and interior
entryway dust concentration while the door paint directly impacted floor dust loading. Note that
inclusion of the paint hazard score in the R&M environmental models and the decrease in sample
size resulted in major changes in the statistically significant pathways as presented in Figures 6-3
and 6-4.
R&M
Window Paint Hazard Score ^^^
Floor Dust
Door Paint Hazard Score
Rochester
_ . Interior Entryway
Door Paint ^ . p|oor QU5t
Hazard Score ^^ \
Soil k. Window Well k. Winrlnw ^ill
Window Paint ^
Hazard Score
Figure 6-7. Significant Pathways for the R&M and Rochester Environmental-Lead Pathway
Models Including Window and Door Paint Hazard Score Pathways Dust-Lead
Loadings (pg/ft2).
64
-------
,1
R&M
(a) Window Sill
Window Paint Hazard Score "**
,.
^*
_
Rochester
Door Paint Hazard Score
Exterior Interior
Entryway *" Entry
Dust Du
i
VVdy F i luui
\ / \
Window Paint
Hazard Score
/ \
Dust
L
low Sill
^^^^^^mmm
Figure 6-8. Significant Pathways for the R&M and Rochester Environmental-Lead Pathway
Models Including Window and Door Paint Hazard Score Pathways Dust-Lead
Concentrations fc/g/g).
Blood-Lead Pathways
Figure 6-9 presents the significant pathways for the blood-lead pathways assessed for the
Rochester data and Table C-3 in Appendix C presents the model parameter estimates.
Figure 6-9 illustrates that the window and door paint directly impacted the blood in the
loading model while only the door paint contributed directly to blood in the concentration model.
In addition, for both the loading and concentration models the window paint directly impacted
the window sills and window wells dust while the door paint directly impacted the interior
entryway dust.
65
-------
Lead Loadings Ijjg/ft2)
Door Paint
Hazard Scor
^---
Soil
Exterior Entryway Dust
1
,, Interior
e _^~ Entryway -^^_^
J-^-" Dust " ' -^~^_
Well \ / Sill Dust /**
Hood
Concentrations (fjg/g)
Exterior
Dust
^nll
Door Paint Hazard
Interior
k* r +
Dust
i i
*
ocore v
\ Mouthing
j
Window Paint
Hazard Score
/ \
Dust \ 1
' \ T
Blood
t L
.hi Winrinuu \A/nll . ... fc \A/inrir>iA/ lill
1
Rgure 6-9. Significant Pathways for the Rochester Blood-Lead Pathway Model Including
Window and Door Paint Hazard Score Pathways Dust-Lead Loadings (//g/ft2)
and Concentrations d/g/g).
6.5 RESULTS OF ASSESSING THE HAND DUST-LEAD
Since only the Rochester study had hand dust-lead information, this analysis assesses the
impact of including hand dust-lead as a pathway in the Rochester blood-lead pathways model
that includes paint. The pathways model assessed was described in Equation Set 5-9, the
parameters are presented in Table C-5 in Appendix C, and Figure 6-10 presents the statistically
significant pathways. Comparing Figure 6-10 to Figure 6-9 shows that when hand dust-lead is
included in the analysis, the pathways in Figure 6-10 are almost the same as those in Figure 6-9,
but hand-lead is now a direct pathway to blood-lead in both the loading and concentration
models. In addition, in the loading model, floor dust-lead is a direct pathway to hand dust-lead.
In the concentration model, window well dust-lead concentration is a direct pathway to hand
dust-lead.
66
-------
Door Paint
Hazard Score
Interior
Entryway
Dust
ป Window -
Sill
Window Paint Hazard Score -
Mouthing
Floor Dust
Hand
II
Blood
Concentrations (jyg/g)
Exterior
Entryway
Dust
Door Paint Hazard Score
I
Interior
* Entryway
Dust
Floor Dust
Window Paint
Hazard Score
\
Mouthing
I
Blood
Hand Dust
Soil
Window Well
Window Sill
Figure 6-10. Significant Pathways for the Rochester Blood-Lead Pathway Model Including
a Hand Lead and Window and Door Paint Pathways Dust-Lead Loadings
U/g/ft2) and Concentrations (//g/g).
Finally, comparing the blood-lead equation in Table C-3 to Table C-5 shows that the R2
for the blood-lead equation increases when hand dust-lead is added as a pathway, indicating that
additional variability is explained when the hand dust-lead variable is included in the model.
6.6 RESULTS OF ASSESSING SOIL-LEAD COVERAGE
A variable was collected in the Rochester study which described the soil coverage at the
site of the soil sample, i.e., whether the area was grass covered, bare, etc. This variable was
combined with the soil-lead concentration by multiplying the coverage indicator times the soil-
67
-------
lead concentration. This new variable was then substituted for the soil-lead concentration in the
Rochester blood-lead pathways model illustrated in Equation Set 5-8. Table C-6 in Appendix C
lists the parameter estimates. The significant pathways were similar to those shown in
Figure 6-9, the original model that included soil-lead concentration. However, in the model with
the soil-lead coverage variable replacing the soil-lead concentration variable, the pathway from
soil to interior entryway dust-lead loading is no longer significant. Moreover, in the model with
soil-lead coverage, the pathways from soil to window well dust-lead loading and soil to window
well dust-lead concentration have much lower parameter estimates than is the case in the model
with soil-lead concentration.
6.7 RESULTS OF ASSESSING CARPETED FLOORS
This section discusses the effect of carpeting. Note that the results in this section must be
viewed with extreme caution since potential confounding effects such as age of the home, type of
home, or other socioeconomic variables were not taken into account in the analysis.
For the R&M model, a pathway for the proportion of floor samples which were carpeted
was included in the blood-lead pathway model. Equation Set 5-10 presents the exact model
assessed. Table B-3 in Appendix B lists the parameter estimates and Figure 6-11 illustrates the
statistically significant pathways for the loading and concentration models.
A comparison of Figure 6-11 to Figures 6-5 and 6-6 shows that the statistically
significant pathways generally remained the same. The carpet indicator was an added pathway
of lead exposure to blood in both the loading and concentration models. All other pathways in
the loading model remained the same. In the concentration model, the window sill dust to floor
dust pathway was replaced by the window well to floor dust pathway while all other pathways
remained the same. Most of the parameter estimates in Table 6-7 remained about the same when
the carpet indicator was included in the model, as can be seen by comparing Table 6-7 to
Table B-3.
68
-------
Loadings (fjg/ft2)
^ ^
Window Well Window Sill
Dust *" Dust
^~^~ ^
Exterior
Entryway
Dust
1
I
Interior
Entryway
Dust
^ __ __^_^
Mouthing
1
T
-+ "oor ^ Blood
Dust
Proportion
Carpet
Concentrations (fjg/g)
Exterior
Entryway Dust
\
Window Well
Dust
Window Sill
Dust
Interior
Entryway Dust
Blood
I
Proportion
Carpet
Figure 6-11. Significant Pathways for the R&M Blood-Lead Pathway Model Including a
Proportion of Carpeting Pathway Dust-Lead Loadings (j/g/ft2) and
Concentrations (f/g/g).
The models were used to assess the effect of a 50% decrease in geometric mean
environmental-lead levels on blood-lead concentrations and other environmental levels.
Table B-6 in Appendix B presents the results for the R&M study data and shows the largest
decreases in blood-lead concentrations, a 27% decrease for the loading model and a 25%
decrease for the concentration model, occurred when the mouthing habits of the child were
reduced. There was an estimated 104% increase, for the loading model, and an 81% increase, for
the concentration model, in the blood-lead concentration when the proportion of carpeted floors
in the home was reduced. Again, extreme caution should be exercised in drawing any
conclusions from mis result. The results may be highly confounded. The R&M homes had
higher lead levels and on average 20% of the sampled floors were carpeted. The control homes
69
-------
had lower lead levels and about 47% of the sampled floors in a control home were carpeted.
Moreover, 73% of the sampled floors in the modem urban control homes were carpeted
Two additional analyses were run for the Rochester data. The first analysis (see Equation
Set 5-11) added a variable for the proportion of floors carpeted and an indicator variable of
whether the interior entryway was carpeted to the blood-lead pathway model illustrated in
Equation Set 5-8. The second analysis (see Equation Set 5-12) essentially applied the blood-lead
pathway model, which included the door and window paint hazard score variables, to a subset of
homes that had both a carpeted bedroom and a carpeted play area. For the subset of homes, the
average floor dust sample for each home was the average of the bedroom and the play area floor
dust samples only. Presence of carpeting in the interior entryway was accounted for by an
indicator variable. The purpose of the analysis of the subset data was to determine if carpeting in
the rooms where a child spends a lot of time reduces the child's blood-lead concentration.
Figures 6-12 and 6-13 diagram the statistically significant pathways for the two analyses. Tables
C-7 and C-8 in Appendix C present the parameter estimates for the models.
Comparing Figure 6-12 to Figure 6-9 shows the statistically significant pathways were
nearly the same. However, in Figure 6-12, the indicator of interior entryway carpet and the
proportion carpet variable are now direct pathways to blood-lead in the loading model. The
indicator of entryway carpet is a direct pathway to blood-lead in the concentration model. In the
loading model, the indicator of entryway carpet is a direct pathway to entryway dust-lead, and
the proportion carpet variable is a direct pathway to floor dust-lead.
The results of the second analysis on the subset of homes are presented in Figure 6-13
and Table C-8. Note that the proportion carpet indicator was omitted from the second analysis
since all floors other than entryways were carpeted in the analysis. Comparison of Figure 6-13 to
Figure 6-12 shows that floor dust-lead loading is no longer a direct pathway to blood-lead when
only carpeted (non-entry) floors are included in the model. Additionally, ulterior entryway and
window sill dust concentrations are no longer direct pathways to floor dust concentration.
70
-------
Loadings (jug/ft2)
Exterior
Entryway
Dust
Indicator Interior-
Entryway Carpet
Interior
Entryway
Dust
t
Door Paint
Soil.
Window
Sill
Window Paint
Hazard Score
Blood
Proportion
Carpet
\
Mouthing
Concentrations (fjg/g)
Door Paint
Hazard Score
Indicator Interior
Entryway Carpet
Exterior
Entryway
Dust
Interior
Entryway
Dust
Floor Dust
I
Blood
Window Paint
Hazard Score
\
\
Mouthing
Soil
Window
Well
-> Window Sill
Figure 6-12. Significant Pathways for the Rochester Blood-Lead Pathway Models Including
Proportion of Carpeting and Indicator of Interior Entryway Carpet Pathways -
Dust-Lead Loadings (//g/ft2) and Concentrations U/g/g).
71
-------
Loadings (jug/ft2)
Indicator Interior
Entryway Carpet
I
Exterior
Entryway
Dust
Interior
Entryway
Dust
Soil
Window Paint
Hazard Score
Window
Well
Blood
Concentrations (yg/g)
Exterior
Entryway
Dust
Soil
Interior
Entryway
Dust
Window
Well
Door Paint
Indicator Interior Hazard Score
Entryway Carpet
Blood
Window Paint
Hazard Score
Mouthing
Figure 6-13. Significant Pathways for the Rochester Blood-Lead Pathway Model For
Homes with Carpeted Bedrooms and Play Areas Dust-Lead Loadings
(//g/ft2) and Concentrations (//g/g).
72
-------
To assess the effect of the carpeting and paint pathways, an analysis of the effect of 50%
decreases in the reported geometric means of environmental variables was performed. Table C-9
in Appendix C shows that for both the loading and the concentration models, when the
proportion of carpeting is reduced (i.e., more uncarpeted floors) the blood-lead concentration
increased: a 27% increase for the loading model and a 40% increase for the concentration model.
The similarities between the models of the effect of decreasing carpeting in the home did not
hold for floor dust. Decreasing the proportion of carpeting decreased the floor dust-lead loading
by 81% but increased the floor dust-lead concentration by 19%.
These results must be interpreted cautiously. They do not necessarily mean that
children's blood-lead concentrations are higher in homes with uncarpeted floors. There may be
socioeconomic variables which have not been included in the analysis that could be confounding
the analysis results. In particular, for the R&M homes, the modern urban control homes are
likely to be heavily influencing the outcome. Also, the BRM vacuum may collect dust deep in
carpet piling that is not normally accessible to a child. A controlled study would likely be
necessary to determine the effect of carpeting.
6.8 RESULTS OF ASSESSING RENOVATION AND REMODELING
The R&M and CAP studies collected information on the conduct of renovation or
remodeling (R&R) in the home, where the R&R activities were performed six months prior to
the environmental sampling. A pathway for the R&R activities was included in the
environmental pathway models for CAP and R&M studies. In addition, a pathway for R&R
activities was included in that R&M blood-lead pathway model which had the proportion of
carpeting as a pathway. Equation set 5-13,5-14, and 5-15 present the exact models assessed.
Tables A-l in Appendix A, and Tables B-4 and B-5 in Appendix B list the parameter estimates.
For the CAP study, renovation and remodeling was a significant pathway to both floor
dust loading and concentration, whereas for R&M environmental model, R&R was not a
significant pathway to floor dust. Recent R&R recent activities occurred among 25% of the
homes for the CAP study, but only 10% of R&M homes had recent renovation and remodeling
activities performed. This may explain the difference in results between the two studies.
73
-------
The parameter estimates for the blood-lead model with R&R activity for R&M, shown in
Table B-5 of Appendix B, indicate that the R&R pathway was not a significant pathway of lead
exposure. Not surprisingly, the significant pathways highlighted in Table B-5 are nearly the
same as those highlighted in Table B-3. There is a change in the loading model and a change in
the concentration model. In the loading model, window sill dust to floor dust becomes a
significant pathway when R&R activity is included in the model. In the concentration model, a
significant pathway of window sill dust-lead concentration to floor dust-lead concentration
replaces a pathway from window well dust-lead concentration to floor dust-lead concentration
when R&R activity is included hi the model.
Note that the nonsignificant R&R pathway in the R&M blood-lead pathway model is
different from published results for the Brigham and Women's Hospital Longitudinal Lead Study
[7] which indicated a statistically significant R&R (or "Refinishing") pathway to blood-lead
concentration. Differences in the results could be due to study differences, such as the
percentage of houses with recent renovation and remodeling activities (10% for the R&M data
and 25% for the Brigham and Women's data) or the number of observations used hi the analysis
(84 for R&M and about 180 for Brigham and Women's).
6.9 RESULTS OF ASSESSING RACE
This section discusses the significant direct and indirect pathways of lead exposure in the
Rochester study for two different race groups: African-American children and children from all
other race groups, including Caucasian, Hispanic and Puerto Rican. The average blood-lead
concentrations were 9.2 ug/dL for African-American children and 4.9 ng/dL for children of all
other races, so it was conjectured there may be different pathways of lead exposure.
Equation Set 5-16 presents the pathways model analyzed for the two race groups.
Figures 6-14 and 6-15 present the statistically significant pathways for the lead loading and lead
concentration models, respectively. The parameter estimates for the models are provided in
Tables C-10 and C-ll of Appendix C.
74
-------
African-American
Window
Well Dust
-> Blood
Window Paint
Hazard Score
Soil
Window
Sill Dust
Interior
Entryway
Dust
Floor Dust
Other Races
Window
Sill Dust
Floor Dust
t
I
Soil
Window
Well Dust
-> Blood
t
Window Paint
Hazard Score
Door Paint
Hazard Score
Exterior
Entryway
Dust
Interior
-> Entryway
Dust
Figure 6-14. Significant Pathways for the Rochester Blood-Lead Pathway Models For
African-American and Other Race Children - Dust-Lead Loadings (//g/ft2).
75
-------
African-American
Mouthing
(1)
(2)
I
Blood
Interior
Entryway Dust
(3) Soil
Floor Dust
Window
Well Dust
Window Paint
Hazard Score
Window
Sill Dust
Other Races
Soil
Window Paint
Hazard Score
Window
Well Dust
Window
Sill Dust-
Interior
Entryway Dust
t
Exterior
Entryway Dust
Blood
Door Paint
Hazard Score
Floor Dust
Figure 6-15. Significant Pathways for the Rochester Blood-Lead Pathway Models For
African-American and Other Race Children Dust-Lead Concentrations
76
-------
For the loading models, the statistically significant direct pathways of lead exposure to
the blood are different for the two race groups. For the African-American children, window well
dust-lead loading was the only direct pathway of lead to blood, while for the other race children,
the statistically significant direct pathways to blood included interior entryway dust, window
well dust, floor dust, window paint hazard score, and door paint hazard score. Blood-lead was
indirectly impacted by exterior entryway dust and soil for other race children. There were no
statistically significant indirect pathways of lead to blood for African-American children.
In the concentration model, the mouthing behavior was the only direct pathway of lead to
blood, for the African-American children and there were no statistically significant indirect
pathways. For the other race children, window well dust and door paint hazard score were
statistically significant direct pathways to blood. Soil and window paint hazard score indirectly
impacted blood via window well dust for these children.
Note that there were 86 African-American children and 119 children in the other race
group. The differences observed in this analysis could be due solely to the limited sample sizes
that resulted when the data was subsetted by race group.
6.10 RESULTS OF ASSESSING AIR DUCTS
The results of an analysis using air duct data are presented in Table A-2 of Appendix A
and show that the air duct dust-lead pathway is not significant. A comparison of the parameter
estimates presented hi Tables 6-1 and A-2 show that there were minimal changes when the air
duct dust-lead pathway was added to the model.
The estimated change in environmental lead levels are presented in Table A-3 of
Appendix A. This table shows that the reduction of the soil-lead concentration geometric mean
by 50% is estimated to produce the greatest reductions in floor dust-lead - a 21% reduction for
both the loading and concentration models.
77
-------
7.0 DISCUSSION
In the environmental-lead pathways analysis, three statistically significant direct
pathways of lead contamination were found for all three studies: 1) window well dust-lead
loading to window sill dust-lead loading, 2) window well dust-lead concentration to window sill
dust-lead concentration, and 3) exterior entryway dust-lead concentration to interior entryway
dust-lead concentration. There were several indirect pathways which were the same for two out
of the three studies: 1) soil-lead concentration to window well dust-lead loading to window sill
dust-lead loading, 2) soil-lead concentration to window well dust-lead concentration to window
sill dust-lead concentration, 3) window well dust-lead loading to window sill dust-lead loading to
the interior entryway dust-lead loading or the floor dust-lead loading, and 4) window well dust-
lead concentration to window sill dust-lead concentration to floor dust-lead concentration.
Despite the different study designs and dust vacuum collection methods, the results are quite
similar across the three studies.
The blood-lead pathways models fitted to the two data sets with blood lead measurements
(R&M and Rochester) had fewer consistent, statistically significant pathways across the studies.
In the lead loading models consistent pathways included direct pathways of lead exposure from
1) floor dust-lead and 2) children's mouthing habits to the blood, and indirect pathways from
1) interior entryway dust-lead, 2) dust on window wells and 3) dust on window sills to blood.
However, no consistent significant pathways to blood were found in the lead concentration
models. In the R&M concentration model, significant pathways included a direct pathway from
interior entryway dust-lead concentration to blood, and indirect pathways of exposure to the
blood from exterior entryway dust, window well dust, and the window sill dust-lead
concentration. In the Rochester concentration model, the only statistically significant direct
pathway to blood was a pathway from children's mouthing behavior to blood. There were no
statistically significant indirect pathways to blood in the Rochester concentration model.
In one of the sub-analyses, two paint-lead metrics were separately included in the models
as ulterior door and window paint-lead pathways: 1) a hazard score defined as the product of
paint condition and XRF measurement, and 2) an average XRF measurement. In the
concentration models for the R&M and Rochester data, the hazard score and the average XRF
measurement generally yielded the same statistically significant pathways, indicating little
difference in the explanatory power of either metric. In the loading model for Rochester, again,
78
-------
similar statistically significant pathways were observed when either metric was included in the
model. In contrast, in the loading model for the R&M data, no statistically significant pathways
were observed when the average XRF measurement was used, but both the door and window
paint were statistically significant pathways of lead exposure to blood when the hazard scores
were used.
A sub-analysis of the effect of interior entryway carpeted floors in blood-lead
concentrations was performed using the Rochester study data. An indicator variable representing
whether the interior entryway was carpeted or not was included in the analysis. The analysis
showed that the presence of interior entryway carpeting in a home was associated with lower
blood-lead concentrations. Although the result should be interpreted with caution, this result
may indicate the usefulness of interior entryway mats in the home for reducing the soil and dust
tracked into the home. A study by Roberts et al., discussed in an EPA literature review which
identified 59 articles on dust and lead exposures associated with residential carpet, indicated that
a walk-off mat present at the entry resulted in a 6.4-fold reduction in dust-lead loadings in
carpets [21].
In the Rochester data, floor dust-lead loadings were on average higher on carpeted
surfaces than on uncarpeted surfaces and the floor dust-lead concentrations were lower on the
carpeted surfaces than on the uncarpeted surfaces. In the EPA literature review report, it was
found that carpets can have high dust-lead loadings relative to other surfaces but only moderate
dust-lead concentrations. One study reported geometric mean dust-lead loadings from carpet to
be approximately 18 times higher than loadings from uncarpeted floors and reported lower dust-
lead concentrations on carpets relative to other surfaces [21].
For both the R&M and Rochester studies, a variable representing the proportion of
sampled floors that were carpeted was a significant pathway to blood; as the proportion of
carpeted floors increased the blood-lead concentration decreased. In the Rochester analysis
discussed above, an indicator of interior entryway carpet was found to be a significant direct
pathway to blood. In an effort to further address the carpeted versus uncarpeted floors issue, an
analysis was performed on Rochester homes where both the bedroom and play area floors were
carpeted. The analysis showed that the floor dust-lead loading was no longer a significant
pathway of lead exposure to blood. These results should be interpreted cautiously. There may
be socioeconomic variables which have not been included in the analysis that could be
79
-------
confounding the analysis results. In particular, for the R&M houses it seems likely that the
modem urban control homes are strongly influencing the outcome. Also, the BRM vacuum may
be collecting dust from carpets that is not normally accessible to a child. It is likely a controlled
study would be necessary to estimate the effect of carpets.
As discussed earlier, there have been several studies in which pathways of lead exposure
to blood have been examined. Six sets of pathways, as reported in the literature, are diagramed
in Appendix D and the variables used in the analysis are described in the tables in Appendix D.
There were three sets of analyses performed by Bomschein, et al., using environmental
lead, demographic, and blood lead information for 18-month old children in the Cincinnati Lead
Study (2 sets of analyses) and 6 to 72-month old children in the Telluride, Colorado smelter
study. Menton, et al., assessed pathway models for 24-month old children in Boston. Using
stepwise multiple regression techniques, Sayre assessed the blood lead, environmental lead, and
demographic information for 18 to 71-month old children living in central Rochester, New York.
An SEM analysis was performed by Lanphear et al. using data from the 1991-1992 Rochester
study.
In all five of the analyses which included hand dust-lead hi the analysis, hand dust-lead
was found to be a statistically significant direct pathway of lead exposure to the blood, and
interior dust was an indirect pathway to blood via hand dust. Similarly, the sub-analysis of the
Rochester hand dust-lead performed for this report indicated significant pathways of hand lead to
the blood-lead concentration and interior dust to hand lead. No other sub-analyses could be run
for hand dust-lead since hand dust-lead was not available for assessment in the R&M study.
Both the Sayre and Lanphear analyses found pica habits to be statistically significant
direct pathways of lead exposure to blood. This is similar to results seen for mouthing behavior
hi the blood-lead pathway models analysis for the R&M and Rochester study data. These
similarities are observed even though the definition of pica was slightly different across the
analyses. Lanphear defined his pica variable as putting soil or dirt in the mourn, and Sayre used
finger sucking, mouthing of toys, coins, pencils, or articles of clothing as his pica variable. Pica
was defined in R&M and Rochester for this report by an indicator of how often a child puts a
thumb, paint chips, or dirt into his/her mouth or mouths a window sill.
In Bornschein (see Figure D-l), the paint hazard score was a direct pathway to floor dust
and an indirect pathway to blood-lead through floor dust and hand-lead. However, when
80
-------
maximum XRF measurements were used in another Bomschein pathways model (see
Figure D-2), an indirect pathway from exterior XRF to floor dust or blood-lead through soil was
observed. In the Lanphear pathways analysis, paint lead, the average of XRF measurements,
contributed to blood-lead levels indirectly through dust-lead and hand-lead. Significant
pathways in the Rochester analyses in this report were generally the same whether hazard score
or average XRF was included. But in the R&M analyses, both the window and door paint hazard
scores were statistically significant pathways of lead to floor dust loading. However, when the
average XRF measurements was used in the R&M loading model, there were no statistically
significant pathways. The small sample size available for the R&M data may explain the
differences observed in the R&M results.
In the Rochester data, the average blood-lead concentrations for African-American
children were 9.2 ng/dL and 4.9 ug/dL for children of all other races (Caucasian, Hispanic, and
Puerto Rican). Because of these differences in blood-lead concentrations, it was thought that
there may be different pathways of lead exposure for each race group. One pathway model was
assessed separately for the African-American children and for the children of all other races.
This assessment showed one common direct pathway of lead exposure to blood-lead
concentration and several different statistically significant pathways for each race. For African-
American children, the loading model had just one direct pathway to blood, window well dust
loading. The concentration model had only mouthing as a direct pathway to blood. There were
no indirect pathways to blood for the African-American children. For the analysis of all other
race groups, the loading model included the following direct pathways to blood: floor dust
loading, window well dust loading, window paint hazard score, door paint hazard score, and
interior entryway dust loading. In the concentration model, window well concentration and door
paint hazard score were the only direct pathways to blood. There were a number of indirect
pathways to blood in both the loading and concentration models for the other race group.
Lanphear, et al. [6,19] also assessed the relationship of race group to blood-lead
concentrations. Two types of analysis were performed to address this issue: SEM [19] and
multivariate regression [6]. Lanphear's SEM model did not split the data into two distinct data
sets as was done for the Rochester analysis performed here, but assessed the data for African-
American children and Caucasian children together by including an indicator variable for race in
the model. In the multivariate regression analysis Lanphear split the data into separate data sets:
81
-------
African-American children and Caucasian children. In the multivariate regression analysis,
Lanphear found that for African-American children, lead interior to the homesdust-lead
loadings, condition and lead content of the painted surfaces, and water lead concentrationswere
the most significant predictors of blood-lead concentration. For Caucasian children, lead exterior
to the home-soil lead concentration, mouthing of dirt or soil, and the amount of time spent
outdoors- were significant predictors of blood-lead concentration. In the SEM analysis,
Lanphear showed that African-American race was a significant direct pathway to blood-lead
concentration.
A comparison between Lanphear's multivariate results and SEM analysis performed here
indicates that the exposure from lead interior to the home for African-American children and
exposure from lead exterior to the home for Caucasian children found in Lanphear's multivariate
regression analysis [6] work did not hold in the SEM analysis performed in this report.
Methodological differences between Lanphear's multivariate regression analysis and the SEM
analysis in this report are the likely explanation for the differences in results. Besides the
difference between multivariate regression analysis and structural equations modeling, there are
the following differences: 1) Lanphear's analysis relies on wipe samples while the analysis in
this report used vacuum samples, and 2) the sample size in Lanphear's analysis was 86 African-
American children and 86 White children whereas the sample size for the analysis hi this report
was 86 African-American children and 119 children from all other race groups. Finally, the
differences in pathways between the African-American children and the children of all other race
groups that were observed in the SEM analysis in this report could be due to the smaller sample
sizes that resulted from subsetting by race group.
82
-------
8.0 REFERENCES
[1] Bomschein, R.L., Succop, P., Dietrich, K.N., Clark, C.S., Que Hee, S., and Hammond,
P.B. (1985) "The Influence of Social and Environmental Factors in Dust Lead, Hand
Lead, and Blood Lead Levels in Young Children." Environmental Research 38-108-
118.
[2] Bomschein, R.L., Succop, P.A., Kraffi, K.M., Clark, C.S., Peace, B., and Hammond, P.B.
(1986) "Exterior Surface Dust Lead, Interior House Dust Lead and Childhood Lead
Exposure in an Urban Environment." in Trace Substances in environmental Health, II,
1986. A symposium edited by D.D. Hemphill (University of Missouri, Columbia).'
[3] Hoyle, R.H., ed. (1995) Structural Equation Modeling: Concepts. Issues, and
Applications. Sage Publications, Inc., 1995.
[4] HUD (1990) "Comprehensive and Workable Plan for the Abatement of Lead-Based Paint
in Privately Owned Housing: Report to Congress." U.S. Department of Housing and
Urban Development, pp. 2-18, December 7,1990.
[5] HUD (1991) "The HUD Lead-Based Paint Abatement Demonstration (FHA)." Office of
Policy Development and Research, U.S. Department of Housing and Urban
Development, HUD-1316-PDR, August 1991.
[6] Lanphear, B.P., Weitzman, M., and Eberly, S. (1996) "Racial Differences in Urban
Children's Environmental Exposures to Lead." American Journal of Public Health 86
(10): 1460-1463.
[7] Menton, Ronald G., David A. Burgoon, and Allan H. Marcus, (1993) "Pathways of Lead
Contamination for the Brigham and Women's Hospital Longitudinal Lead Study" in
Beard, Michael E. and S.D. Allen Iske, ed., Lead in Paint. Soil, and Dust: Health Risks.
Exposure Studies. Control Measures. Measurement Methods, and Quality Assurance, and
STP; 1226,1993 ASTM Boulder Conference on Lead in Paint, Soil, and Dust.
[8] SAS/STAT User's Guide, Version 6, Fourth Edition, Volume 1. "Proc Calis". pp. 292-
365.
[9] USEPA (1995) "Comprehensive Abatement Performance Pilot Study." Office of
Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency,
EPA 747-R-93-007, February 1995.
[10] USEPA (I996a) "Final Report for the Comprehensive Abatement Performance Study
Volume I: Summary Report", Office of Prevention, Pesticides and Toxic Substances,
U.S. Environmental Protection Agency, EPA 230-R-94-013a, April 1996.
83
-------
[11] USEPA (1996b) "Final Report for the Comprehensive Abatement Performance Study
Volume E: Detailed Statistical Results", Office of Prevention, Pesticides, and Toxic
Substances, U.S. Environmental Protection Agency, EPA 230-R-94-013b, April 1996.
[12] USEPA (1996c) "Lead-Based Paint Abatement and Repair and Maintenance Study in
Baltimore: Pre-Intervention Findings." Office of Prevention, Pesticides and Toxic
Substances, U.S. Environmental Protection Agency, EPA 747-R-95-012, August 1996.
[13] The University of Rochester School of Medicine and The National Center for Lead-Safe
Housing (1995) "The Relation of Lead-Contaminated House Dust and Blood Lead
Levels Among Urban Children: Volumes I and n." Departments of Pediatrics,
Biostatistics, and Environmental Medicine, The Rochester School of Medicine,
Rochester, New York, and the National Center for Lead-Safe Housing, Columbia,
Maryland, June, 1995.
[14] Lanphear, B.P., Weitzman, M., Winter, N.L., Eberly, S., Yakir, B., Tanner, M.,
Emond, M., and Matte, T. (1996) "Lead-Contaminated House Dust and Urban
Children's Blood-Lead Levels." American Journal of Public Health. 86(10): 1416-1421.
[15] Emond, M.J., Lanphear, B.P., Watts, A., Eberly, S. (1996) "Measurement Error and Its
Impact on the Estimated Relationship Between Dust Lead and Children's Blood Lead."
Environmental Research. 72: 82-92.
[16] HUD (1995) "Guidance for the Evaluation and Control of Lead-Based Paint Hazards in
Housing." U.S. Department of Housing and Urban Development. June, 1995.
[17] Bomschein, R.L., Clark, G.S., Grote, J., Peace, B., Roda, S., and Succop, P. (1988)
"Soil Lead - Blood Lead Relationship in a Former Lead Mining Town." Lead in Soil.
149-160.
[18] Sayre, J. (1981) "Dust Lead Contribution to Lead in Children." Environmental Lead.
23-35.
[19] Lanphear, Bruce P. and Klaus J. Roghmann, (1997) 'Tathways of Lead Exposure in
Urban Children.: Environmental Research 74, 67-73.
[20] HUD (1998) "Evaluation of the HUD Lead-Based Paint Hazard Control Grant Program."
U.S. Department of Housing and Urban Development, Fifth Interim Report, pp. 80-85,
March 1998.
[21] USEPA (1997) "Summary and Assessment of Published Information on Determining
Lead Exposures and Mitigating Lead Hazards Associated with Dust and Soil in
residential Carpets, Furniture, and Forced Air Ducts." Office of Pollution Prevention and
Toxics, U.S. Environmental Protection Agency, EPA 747-S-97-001, December 1997.
[22] CDC (1997) Centers for Disease Control, MMWR 46:141-146.
84
-------
APPENDIX A
Results from the Pathways
Analyses of the CAP Study Data
A-1
-------
Table A-1. Structural Equation Modeling Results for the CAP Study Environmental
Pathways, Including an R&R Exposure Pathway Dust-Lead Loadings
(pg/ft2) and Dust-Lead Concentrations (//g/g).
Variables
\. -Independent
Dependent >s,
Direct Effect Parameter Estimates (t-value) '
Interior
Entryway
Dust
j
Window
Sill Dust
i <
Window
Well Dust
i
Exterior
Entryway
Dust
V !
Soil
Dust-Lead Loadings (jug/ft2) , ,
Floor
Interior Entryway
Dust
Window Sill
Window Well
0.2911*
(1.98)
0.1296
(0.95)
0.2323
(1.77)
0.0099
(0.07)
-0.0086
(-0.06)
0.5450*
(4.08)
0.1062
(0.68)
0.2140
(1.42)
-0.0194
(-0.06)
0.3475
(1.05)
0.4154
(1.17)
0.8831*
(2.41)
**
fi&R
Exposure
1
1.9042*
(2.13)
0.0344
(0.07)
0.2918
(0.52)
-0.9733
(-1.63)
,' . ' " '
fs '
J ^
^ ! 1
R*l"'<*
4
0.24
0.18
0.14
0.34
<'.. "- , .,- - Dust-Lead'Concentrmti6ns;j(^g/g).v^:::i'?;;*:Xi.:,;i'/ "--': "''~;Ky~':i <^fll/:tM
Floor
Interior Entryway
Dust
Window Sill
Window Well
0.0923
(0.47)
-0.2099
(-1.78)
-0.0174
(-0.20)
0.3395*
(2.88)
0.0653
(0.74)
0.5498*
(4.55)
-0.1164
(-0.70)
0.3994*
(3.62)
0.1224
(0.54)
0.1504
(0.89)
0.0584
(0.22)
0.9074*
(3.05)
0.8651*
(2.42)
0.1688
(0.41)
-0.7523
(-1.55)
0.26
0.39
0.19
0.36
Notes: I.Bolded and a * indicate parameter estimates are significant at the 0.05 level. T- values * 1.96 and
ฃ -1.96 are significant at the 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/
concentrations and soil-lead concentrations.
3.First number is estimated parameter; second number is corresponding t-value.
4.The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9869, and 0.9772 for the dust-
lead concentration model.
A-2
-------
Table A-2. Structural Equation Modeling Results for the CAP Study Environmental
Pathways, Including an Air Duct Pathway - Dust-Lead Loadings tog/ft2) and
Dust-Lead Concentrations (fjglg).
==I^=^==3=
Variables
\^ Independent
^''C^xf*-
Dependent >. ^^
%
Floor
Interior Entryway
Dust
Window Sill
Window Well
===^
! v
t * ^
^; *
Interior
Entryway
Dust
0.2968
(1.93)
Direct Effect Parameter Estimates (t-value)
Window
Sill Dust
Window
Well Dust
Dust-Lead Loadin
1.428
(1.0)
0.2424
(1 .86)
-0.0617
(-0.43)
-0.0097
(-0.07)
0.529 1ป
(4.06)
! i.
Exterior
Entryway
'Dust
s U/g/ft2)
0.1109
(0.67)
0.2294
(1.49)
Air Duct
Soil
^^^==
R2
0.0330
(0.29)
-0.0449
(-0.42)
0.0985
0.28
0.3491
(1.07)
0.4513
(1.29)
0.8059*
(2.15)
0.17
0.19
0.33
0,09
- - "- ; Dust-Lead Concentrations U/g/g) , ?' ;
Floor
Interior Entryway
Dust
Window Sill
Window Well
0.2128
(1.070)
-0.1934
(-1.56)
-0.0167
(-0.18)
0.2627*
(2.17)
0.0176
(0.20)
0.5389*
(4.56)
-0.0806
(-0.46)
0.4531*
(4.09)
0.0980
(0.54)
0.1352
(1.02)
0.1726
(0.72)
0.1964
(1.14)
0.0810
(0.31)
0.8479*
(2.81)
0.17
0.35
0.35
0.14
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-values * 1.96 and
ฃ -1.96 are significant at the 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/
concentrations and soil-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9538, and 0.9918 for the dust-
lead concentration model.
A-3
-------
Table A-3. Predicted Effect of a 50% Decrease in Environmental-Lead Levels Based on the
CAP Study Environmental-Lead Structural Equation Model, Including the Air
Duct Pathway - Dust-Lead Loading Qvg/ft2) and Concentrations (i/g/g).
Sample Location
Geometric
Mean
' y
50%
Decrease in
GM
r Predicted Percent Change {%) , , -
(Lower Prediction Interval, Upper Prediction Interval) (bl '
Floor
Interior
- Entryway , ,
Window
Sill
Window
Well
Dust-Lead Loading U/g/ftz)
Interior Entryway
Window Sill
Window Well
Air Duct
Soil'31
Exterior
Entryway
342
108
3,229
156
157
572
171
54 -
1,615
78
79
287
-19<-41 , 12)
=14 (-36 , 16)
-3 (-28,31)
-1 (-22 , 26)
-21 (-62 , 64)
-12 (-38, 25)
NA
-15 (-34, 10)
-8 (-21 , 20)
-3 (-22 , 20)
-32 (-64 , 30)
-15 (-37, 15)
NA
NA
-31 (-48 , -8)
NA
-46 (-75, 17)
NA
NA
NA
NA
NA
-43 (-77, 44)
NA
a,, . " j Dust-Lead Concentration (//g/g) ,', , Jl . *'* "'
Interior Entryway
Window Sill
Window Well
Air Duct
Soil'31
Exterior
Entryway
201
778
1,577
485
57
261
101
389
789
243
79
130
-14 (-34, 13)
1 5 (-3 , 36)
-1 1 (-24 , 5)
-8 (-28, 18)
-21 (-43 , 9)
-1 (-22 , 26)
NA
1 (-8, 11)
-17 (-24 , -9)
-9 (-20 , 4)
-13 (-27 , 3)
-27 (-35, -18)
NA
NA
-3 (-20, 17)
NA
-31 (-55 , 5)
NA
NA
NA
NA
NA
-44 (-69, 1)
AM
(al Average soil-lead concentration in
-------
Table A-4a. Correlations for Log-Transformed Dust-Lead Loadings (/ig/ft2) and Soil-Lead Concentrations (pg/g) Used in the CAP Study
Pathways Analysis.
Floor Dust
Air Duct Dust
Interior Entryway Dust
Window Well Dust
Window 51 11 Dust
Exterior Entryway Dust
Exterior Entryway Soil
Foundation Soil
Boundary Soil
Mass Weighted Average Soil
Floor
Dust
1.00000
0.0
52
0.15743
0.2650
52
0.35648
0.0095
52
0.10123
0.4889
49
0.21898
0.1188
52
0.18105
0.2036
51
0.20107
0.1529
52
0.06759
0.6340
52
0.23057
0.1001
52
0.18763
0.1829
52
A1r
Duct
.Dust
0 .15743
0 .2650
52
1.00000
0.0
52
0.05862
0.6798
52
0.13728
0.3469
49
0.11246
0.4273
52
0.24169
0.0875
51
0.14830
0.2941
52
0.05701
0.6881
52
0.20016
0.1548
52
0.17672
0.2101
52
Interior
Entryway
Dust
0 .35648
0 .0095
52
0.05862
0.6798
52
1.00000
0.0
52
0.19758
0.1736
49
0.32992
0.0169
52
0.22004
0.1208
51
0.28452
0.0409
52
0.13223
0.3501
52
0.18554
0.1879
52
0.25688
0.0660
52
Window
Well
Dust
0.10123
0.4889
49
0.13728
0.3469
49
0.19758
ฃ.1736
49
1.00000
0.0
49
0.55256
0.0001
49
0.02957
0.8419
48
0.28873
0.0442
49
0.15568
0.2854
49
0.25439
0.0777
49
0.29788
0.0376
49
Window
S111
Dust
0.21898
0.1188
52
0.11246
0.4273
52
0.32992
0.0169
52
0.55256
0.0001
49
1.00000
0.0
52
0.00309
0.9829
51
0.22846
0.1033
52
0.12126
0.3918
52
0.29352
0.0347
52
0.24304
0.0825
52
Exterior
Entryway
Dust
0.18105
0.2036
51
0.24169
0.0875
51
0.22004
0.1208
51
0.02957
0.8419
48
0.00309
0.9829
51
1.00000
0.0
51
0.19018
0.1813
51
0.18760
0.1874
51
-0.02470
0.8634
51
0.20299
0.1531
51
Exterior
Entryway
Soil
0 .20107
0 .1529
52
0.14830
0.2941
52
0.28452
0.0409
52
0.28873
0.0442
49
0.22846
0.1033
52
0.19018
0.1813
51
1.00000
0.0
52
0.52970
0.0001
52
0.64037
0.0001
52
0.85763
0.0001
52
Foundation
Soil
0.06759
0.6340
52
0.05701
0.6881
52
0.13223
0.3501
52
0.15568
0.2854
49
0.12126
0.3918
52
0.18760
0.1874
51
0.52970
0.0001
52
1.00000
0.0
52
0.55853
0.0001
52
0.80579
0.0001
52
Mass Weighted
Boundary Average
Soil Soil
0.23057
0.1001
52
0.20016
0.1548
52
0.18554
0.1879
52
0.25439
0.0777
49
0.29352
0.0347
52
-0.02470
0.8634
51
0.64037
0.0001
52
0.55853
0.0001
52
1.00000
0.0
52
0.79196
0.0001
52
0.18763
0.1829
52
0.17672
0.2101
52
0.25688
0.0660
52
0.29788
0.0376
49
0.24304
0.0825
52
0.20299
0.1531
51
0.85763
0.0001
52
0.80579
0.0001
52
0.79196
0.0001
52
1.00000
0.0
52
Notes: 1. Correlations were calculated for the natural logarithm transformed dust-lead loadings and soil-
2. First number is the Pearson correlation coefficient, the second number is the p-value, and the
lead concentrations.
third number is the number of observations.
-------
o>
Table A-4b. Correlations for Log-Transformed Dust-Lead Concentrations (fjglg) and Soil-Lead Concentrations (pglg) Used in the CAP
Study Pathways Analysis
Floor Dust
Air Duct Dust
Interior Entryway Dust
Window Well Dust
Window Sill Dust
Exterior Entryway Dust
Exterior Entryway Soil
"
Foundation Soil
Boundary Soil
Mass Weighted Average Soil
Notes: 1 . Correlations
2. First number
Floor
i.oBBBS
0.0 0
52
0.14612
0.3013
52
0.22249
0.1129
52
0.28349
0.0484
49
0.01190.
0.9333
52
0.09858
0.4913
51
0.19160
0.1736
52
0.13421
0.3428
52
0.33594
0.0149
52
0.21932
0.1182
52
were calculated
Air
Duct
o.iS?!S
.3013
52
1.00000
0.0
52
0.19240
0.1718
52
0.17034
0.2419
49
0.20521
0.1445
52
0.05998
0.6759
51
0.18726
0.1837
52
-0.01820
0.8981
52
0.18599
0.1868
52
0.12774
0.3668
52
Interior
Entryway
0.2S5IS
0.1129
52
0.19240
0.1718
52
1.00000
0.0
52
0.14587
0.3173
49
0.00561
0.9685
52
0.55763
0.0001
51
0.37130
0.0067
52
0.17489
0.2150
52
0.23164
0.0985
52
0.33668
0.0147
52
Window
Well
0.2?SIS
0.0484
49
0.17034
0.2419
49
0.14587
0.3173
49
1.00000
0.0
49
0.59493
0.0001
49
0.11442
0.4387
48
0.39132
0.0054
49
0.14883
0.3074
49
0.30910
0.0307
49
0.36555
0.0098
49
for the natural logarithm transformed
is the Pearson correlation
coefficient, the
Window
Sill
o.oSSlS
0.9333
52
0.20521
0.1445
52
0.00561
0.9685
52
0.59493
0.0001
49
1.00000
0.0
52
-0.04710
0.7428
51
0.24320
0.0823
52
0.08412
0.5532
52
0.23619
0.0918
52
0.21682
0.1226
52
Exterior
Entryway
O.flS^iS
0.4913
51
0.05998
0.6759
51
0.55763
0.0001
51
0.11442
0.4387
48
-0.04710
0.7428
51
1.00000
0.0
51
0.31996
0.0221
51
0.18571
0.1920
51
0.03995
0.7807
51
0.31660
0.0236
51
dust-lead concentrations and
Exterior
Bntryway
o.JSfti
0.1736
52
0.18726
0.1837
52
0.37130
0.0067
52
0.39132
0.0054
49
0.24320
0.0823
52
0.31996
0.0221
51
1.00000
0.0
52
0.52970
0.0001
52
0.64037
0.0001
52
0.85763
0.0001
52
Mass Weighted
Foundation
o.iflil
0.3428
52
-0.01820
0.8981
52
0.17489
0.2150
52
0.14883
0.3074
49
0.08412
0.5532
52
0.18571
0.1920
51
0.52970
0.0001
52
1.00000
0.0
52
0.55853
0.0001
52
0.80579
0.0001
52
Boundary
0.3งงli
0.0149
52
0.18599
0.1868
52
0.23164
0.0985
52
0.30910
0.0307
49
0.23619
0.0918
52
0.03995
0.7807
51
0.64037
0.0001
52
0.55853
0.0001
52
1.00000
0.0
52
0.79196
0.0001
52
Average
0.2fSงJ
0.1182
52
0.12774
0.3668
52
0.33668
0.0147
52
0.36555
0.0098
49
0.21682
0.1226
52
0.31660
0.0236
51
0.85763
0.0001
52
0.80579
0.0001
52
0.79196
0.0001
52
1.00000
0.0
52
soil-lead concentrations.
second number is the p-value, and the third number is
the number of
observations.
-------
Table A-5. Principal Components Analysis of the Entryway, Foundation, and Boundary
Soil-Lead Concentrations (j/g/g) from the CAP Study.
Mean
StD
Simple Statistics
UCBDY UCFDN
4.881270657
0.664755800
5.278522566
0.923506587
WCEWY
5.053991088
0.806544484
UCBDY
WCFDN
WCEWY
Correlation Matrix
UCBDY UCFDN
1.0000
0.5585
0.6404
0.5585
1.0000
0.5297
WCEWY
0.6404
0.5297
1.0000
PRIN1
PRIN2
PRIN3
Eigenvalues of the Correlation Matrix
Eigenvalue Difference Proportion
2.15371
0.48896
0.35733
1.66475
0.13162
0.717903
0.162986
0.119111
Cumulative
0.71790
0.88089
1.00000
UCBDY
UCFDN
WCEWY
Eigenvectors
PRIN1 PRIN2
0.592676
0.554944
0.583757
-.297676
0.824371
-.481457
PRIN3
-.748415
0.111578
0.653778
UCBDY = Mass-weighted boundary soil concentration
UCFDN = Mass-weighted foundation soil concentration
UCEUY = Mass-weighted exterior entryway soil concentration
A-7
-------
This page intentionally left blank.
-------
APPENDIX B
Results from the Pathways
Analyses of the R&M Data
B-1
-------
fable B-1. Structural Equation Modeling Results for the R&M Environmental Pathways, Including the Window Paint and
Door Paint Hazard Score Pathways - Dust-Lead Loadings (jig/ft2) and Concentrations (//g/g).
Variables
vs^ Independent
Dependent ^\^
" Direct Effects Parameter Estimates (t-value)
Interior '
Entryway
Dust
Window
Sill Dust
* ?
Window
Well Dust
Exterior
Entryway
Dust
>
Window Paint
Hazard Score
Door Paint
Hazard Score
, Rz
Dust-Lead Loadings (//g/ft2)
Floor
Interior Entryway
Window Sill
Window Well
0.0121
(0.11)
0.0458
(0.33)
0.1796
(0.83)
0.2983
(1.73)
0.0940
(0.35)
-0.0216
(-0.10)
-0.0497
(-0.58)
0.0582
(0.44)
0.0960*
(2.59)
0.0065
(0.11)
0.0144
(0.32)
0.0013
(0.04)
-0.1652 *
(-2.67)
-0.0364
(-0.38)
0.31
0.03
0.003
0.00004
: ~- ' - ' > Dust-Lead Concentrations (//g/g)3 - .
Floor
Interior Entryway
Window Sill
Window Well
0.0200
(0.15)
0.3463*
(2.25)
-0.2012
(-1.03)
0.2774
(1.54)
-0.1843
(-0.80)
-0.1940
(-0.97)
-0.1292
(-0.90)
0.6414*
(4.33)
0.0525
(1.67)
-0.0793*
(-2.08)
0.0142
(0.45)
0.0199
(0.73)
-0.0894
(-1.87)
-0.0439
(-0.72)
0.28
0.40
0.03
0.02
ro
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value z 1.96 or z -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9768, and 0.9053 for the dust-lead concentration model.
-------
Table B-2. Structural Equation Modeling Results for the R&M Environmental Pathways, Including the Average Window Paint
and Door Paint XRF Measurement Pathways Dust-Lead Loadings (//g/ft2) and Concentrations (//g/g).
I) Variables
Ipx. Independent
II Dependent ^\
Direct Effects Parameter Estimates (t-value) , / ,, .
Interior
Entryway , -
Dust
K
' ' Window
Sill Dust
1
Floor
II Interior Entryway
Window Sill
1 Window Well
0.0453
(0.37)
-0.0045
(-0.03)
0.1589
(0.75)
. '-.r-
-*
Window
Well Dust ,
Just-Lead Loading
0.2657
(1.41)
0.1280
(0.48)
-0.0203
(-0.09)
,ป flr '
Exterior
Entryway"
Dust
i ' * _
Window Paint
, XRF
Measurement
^
Door Paint
, XRF
Measurement
s fc/g/ft2) .-"
-0.0395
(-0.43)
0.0603
(0.46)
0.0631
(1.70)
0.0286
(0.55)
0.0034
(0.10)
-0.0066
(-0.24)
0.0019
(0.04)
-0.0815
(-1.29)
-
_< ,_
f *
R2
t.
0.16
0.08
0.001
0.002
||v-- ' vi i .. , Dust-Lead,dQlricentrationi!;U'g/g) , , T
|| Floor
|| Interior Entryway
Window Sill
1 Window Well
0.0350
(0.23)
0.4161*
(2.49)
-0.2907
(-1.63)
0.2627
(1.37)
-0.2363
(-1.14)
-0.1830
(-0.93)
-0.1401
(-0.89)
0.6516*
(4.87)
0.0259
(0.91)
-0.0233
(-0.75)
0.0112
(0.47)
0.0002
(0.01)
0.0001
(0.004)
-0.0927*
(-2.50)
0.22
0.53
0.03
0.000002
ซp
CO
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value a 1.96 or s -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9810, and 0.9097 for the dust-lead concentration model.
-------
Table B-3. Structural Equation Modeling Results for the R&M Blood-Lead Pathways, Including a Proportion of Carpeting
Pathway Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations fcig/g).
Variables
N\^ Independent
Dependent ^x
Direct Effects Parameter Estimates (t value) ,
Interior
Entryway
Dust
Window
Sill Dust ,
Window
.Well Dust;
^
Exterior
Entryway
, ,Dust
Proportion
Carpet
i_
Floor
Dust
Water
.
Mouthing
R*
Dust-Lead Loadings fag/ft2)4 ~ ' - *
Blood
Floor
Interior Entryway
Window Sill
0.0472
(1.36)
0.2075ซ
(2.73)
0.0333
(-0.81)
0.1400
(1.51)
0.3943*
(3.13)
0.0212
(0.52)
0.1504
(1.63)
-0.1432
(-1.17)
0.7958*
(13.86)
0.0439
(1.43)
0.0367
(0.53)
0.2210*
(2.30)
-1.0059*
(-3.84)
-0.2813
(-0.47)
0.1048*
(2.18)
-0.0489
(-0.45)
0.3152*
(1.97)
0.44
0.46
0.25
0.70
Dust-Lead Concentrations (//g/g)4 . , ,- .
Blood
Floor
Interior Entryway
Window Sill
0.1292*
(2.04)
-0.5533
(-1.24)
0.0300
(0.56)
0.1793
(1.67)
0.1870*
(2.16)
-0.0591
(-0.94)
0.2890*
(3.32)
-0.0415
(-0.40)
0.9059*
(13.37)
0.0183
(0.32)
0.1455
(1.36)
0.5486*
(6.95)
-0.8432*
(-3.21)
-0.0154
(-0.16)
0.1005
(1.58)
-0.1053
(-0.97)
0.2930
(1.89)
0.43
0.59
0.57
0.68
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value * 1.96 or ฃ -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations and blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9524, and 0.9106 for the dust-lead concentration model.
-------
Table B-4. Structural Equation Modeling Results for the R&M Environmental Pathways, Including the R&R Exposure
Pathways Dust-Lead Loadings (/sg/ft2) and Concentrations (//g/g).
Variables
^^v^^ Independent 5
Dependent *\.
/ ,. Direct Effects Parameter Estimates (t -value) .. . , '
S-
N
* J
Interior
Entryway
Dust
> .' "
- i
< ' ^
.Window
Sill Dust
(. ,
R&R Exposure
ป*'-
-ฃ X
t
* \
R2 f ,
Dust-Lead Loadings (//g/ft2)3
Floor
Interior Entryway
Window Sill
' r; , ., ..:,. .' .' :: '":.', '.'*'<
Floor
Interior Entryway
Window Sill
0.2222*
(2.96)
-
0.1853*
(2.05)
0.3753*
(2.99)
0.1063
(1.25)
-0.1158
(-0.93)
0.7958*
(13.86)
0.0282
(0.42)
0.2222*
(2.32)
0.8665
(1.82)
-0.5179
(-0.75)
0.48
0.25
0.70
Dust-Lead Concentrations (//g/g)3 m
0.2186*
(2.0288)
0.3221*
(3.71)
0.1739*
(2.02)
0.1258
(1.20)
-0.0035
(-0.03)
0.9059*
(13.39)
-0.0128
(-0.13)
0.5302*
(6.45)
0.4953
(1.30)
-0.3877
(-1 .00)
0.60
0.58
0.68
DO
01
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value 2 1.96 or <. -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. Goodness-of-fit index (GFI) for dust-lead loadings is 0.9538 and GFI for dust-lead concentration is 0.9120
-------
Table B-5. Structural Equation Modeling Results for the R&M Blood-Lead Pathways, Including a Proportion of Carpeting
and Renovation and Remodeling Pathway - Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (//g/g).
Variables
x^ Independent
Dependent >s,
Direct Effects Parameter Estimates (t-value)
* i f ' ^3k.& * %
Interior
Entryway
'* Dust
i * I*-
Window
Sill Dust
Window
Well Dust
Exterior
Entryway
. Dust
Proportion
Carpet
f.
>l ',
* 3
Floor
Dust ,
?'
^ ^ * ^
f -f
Water
Mouthing
R&R
Exposure
** .*
}
,Ra
' i \ -Dust-Lead Loadings (//g/ft2)
Blood
Floor
Interior Entryway
Window Sill
0.0467
(1.33)
0.2194ป
(2.93)
-0.0343
(-0.84)
0.1802*
(2.00)
0.3753*
(2.99)
0.0223
(0.54)
0.1001
(1.09)
-0.1158
(-0.93)
0.7958*
(13.85)
0.0439
(1.43)
0.0310
(0.46)
0.2222*
(2.32)
-1 .0084*
(-3.85)
-0.1828
(-0.31)
0.1058*
(2.17)
-0.0507
(-0.46)
0.3162*
(1.97)
-0.0250
(-0.11)
0.8529
(1.80)
-0.5179
(-0.75)
0.44
0.48
0.25
0.70
,, Dust-Lead Concentrations (jug/g) *
Blood
Floor
Interior Entryway
Window Sill
0.1417*
(2.24)
0.1955
(1.83)
0.0409
(0.76)
0.3037*
(3.53)
0.1739*
(2.02)
-0.0810
(-1.28)
0.1046
(0.96)
-0.0035
(-0.03)
0.9059*
(13.40)
0.0250
(0.44)
-0.0027
(-0.03)
0.5302*
(6.50)
-0.8123*
(-3.12)
-0.4973
(-1.12)
0.0901
(1.42)
-0.0879 '
(-0.82)
0.2847
(1.86)
0.2744
(1.23)
0.4529
(1.21)
-0.3877
(-1.02)
0.44
0.60
0.58
0.68
Notes: 1. Bolded and a * Indicate parameter estimates are significant at the 0.05 level. T-value > 1.96 or <, -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations and blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9512, and 0.9155 for the dust-lead concentration model.
-------
Table B-6.
Predicted Effect of 50% Decrease in Environmental-Lead Loadings fo/g/ft2) and
Concentrations U/g/g) and Blood-Lead Concentrations fc/g/dL) Based on the
Blood-Lead Pathways SEM for the R&M Data, Including Proportion of Carpet.
^
, ! ,V " -" ^ ,
>>
V fe
" -. s,
* - ' , n
Sample Location
Geometric
, Mean
=====
50%
Decrease in
GM
=======5=== =S=====^ =^==5= = |
Predicted Percent Change (%)
(Lower Prediction Interval. Upper Prediction) ซ">
Blood
* 'Floor
Interior
En t rvwd v
Window
Sill
' /; \ Dust-Lead Loadings U/g/ft2)
Floor
Interior Entryway
Window Sill
Window Well
Exterior Entry
Proportion Carpet
-,
Water"1
Mouthing'"
210
319
1,229
37,035
405
0.29
Exposure
1
1
105
165
615
18,518
203
0.15
No Exposure
0
0
-7 (-10, -4)
-5 (-8, -2)
1 (-2, 4)
-1 (-4, 2)
-4 (-6, -2)
104(67, 148)
NA
-14 (-25, -2)
-12 (-25, 3)
-14 (-27, 1)
-5 (-16, 7)
149(59,219)
NA
NA
0 (-27, 37)
8 (-20, 47)
.-14 (-32, 9)
NA
NA
NA
NA
-42 (-49, -34)
NA
NA
"-.. ..:;:;%>: A*^.. -VI -,
5 (-3, 14)
-27 (-35, -18)
NA
NA
NA
NA
NA
NA
-;:.[" -i " , Dust-Lead Concentrations U/g/g) >
Floor
Interior Entryway
Window Sill
Window Well
Exterior Entry
Proportion Carpet
,, ซ- "~*'- -
Water'81
Mouthing'31
1,118
1,459
5,411
8,452
1,570
0.29
Exposure
1
1
559
730
2,706
4,226
785
0.15
No Exposure,
0
0
-6 (-10, -1)
-10 (-14, -6)
-5 (-9, -1)
1 (-4, 6)
-7 (-11, -3)
81 (49, 120)
NA
-13 (-24, 0)
-19 (-28, -9)
-23 (-33, -12)
-7 (-18, 5)
41 (-21, 150)
NA
NA
0(-11, 12)
0(-13, 15)
-31 (-38, -23)
NA
NA
NA
NA
-47 (-53, -41)
NA
NA
g f H *
9(1, 18)
-25 (-33, 16)
NA
NA
NA
NA
NA
NA
-------
Table B-7a. R&M Data Correlation of Log-Transformed Blood-Lead Concentrations (//g/dL), Dust-Lead Loadings fag/ft2), and
Water-Lead Concentrations (pg/L).
Blood
Blood 1.00000
0.0
87
Floor Oust 0.49977
0.0001
87
Interior Entryway Oust 0.40889
0.0001
87
Window Well Dust 0.41655
0.0001
86
Window S111 Dust 0.43834
0.0001
ce,
CO
Exterior Entryway Dust 0.33690
0.0016
85
Water -0.05243
0.6296
87
Door Paint Hazard Score -0.17675
0.3024
36
Door Paint XRF -0.14461
0.4001
36
Window Paint Hazard Score -0.11413
0.5075
36
Window Paint XRF -0.30488
0.0706
36
Floor
Dust
0.49977
0.0001
87
1.00000
0.0
87
0.48834
0.0001
87
0.57258
0.0001
86
0.62981
0.0001
87
0.36254
0.0007
85
-0.06280
0.5633
87
-0.30151
0.0739
36
0.20076
0.2404
36
0.28776
0.0888
36
0.29448
0.0813
36
Interior
Entryway
Dust
0.40889
0.0001
87
0.48834
0.0001
87
1.00000
0.0
87
0.31023
0.0036
86
0.46245
0.0001
87
0.39209
0.0002
85
0.04191
0.6999
87
-0.04979
0.7731
36
-0.20211
0.2372
36
-0.01012
0.9533
36
-0.06624
0.7011
36
Window
Well
Oust
0.41655
0.0001
86
0.57258
0.0001
86
0.31023
0.0036
' 86
1.00000
0.0
86
0.83763
0.0001
86
0.37982
0.0004
84
0. 07790
0.4759
86
0.10948
0.5313
35
0.08804
0.6150
35
0.00600
0.9727
35
-0.04169
0.8120
35
Window
S111
Dust
0.43834
0.0001
87
0.62981
0.0001
87
0.46245
0.0001
87
0.83763
0.0001
86
1.00000
0.0
87
0.47593
0.0001
85
0.09091
0.4024
87
0.14931
0.3848
36
-0.01875
0.9136
36
0.05881
0.7333
36
0.02234
0.8971
36
Exterior
Entryway
Dust
0:33690
0.0016
85
0.36254
0.0007
85
0.39209
0.0002
85
0.37982
0.0004
84
0.47593
0.0001
85
1.00000
0.0
85
-0.08369
0.4464
85
-0.00524
0.9758
36
0.03416
0.8432
36
0.14555
0.3970
36
0.13428
0.4349
36
Water
-0.05243
0.6296
87
-0.06280
0.5633
87
0.04191
0.6999
87
0.07790
0.4759
86
0.09091
0.4024
87
-0.08369
0.4464
85
1.00000
0.0
87
0,28190
0.0958
36
0.02113
0.9026
36
-0.20182
0.2379
36
-0.20550
0.2292
36
Door Paint
Hazard
Score
-0.17675
0.3024
36
-0.30151
0.0739
36
-0.04979
0.7731
36
0.10948
0.5313
35
0.14931
0.3848
36
-0.00524
0.9758
36
0.28190
0.0958
36
1.00000
0.0
36
0.51022
0.0015
36
0.16400
0.3392
36
0.02130
0.9019'
36
Door
Paint
XRF
-0.14461
0.4001
36
0.20076
0.2404
36
-0.20211
0.2372
36
0.08804
0.6150
35
-0.01875
0.9136
36
0.03416
0.8432
36
0.02113
0.9026
36
0.51022
0.0015
36
1.00000
0.0
36
0.63836
0.0001
36
0.59231
0.0001
36
Window
Paint
Hazard
-0.11413
0.5075
36
0.28776
0.0888
36
-0.01012
0.9533
36
0.00600
0.9727
35
0.05881
0.7333
36
0.14555
0.3970
36
-0.20182
0.2379
36
0.16400
0.3392
36
0.63836
0.0001
36
1.00000
0.0
36
0.86920
0.0001
36
Window
Paint
XRF
-0.30488
0.0706
36
0.29448
0.0813
36
-0.06624
0.7011
36
-0.04169
0.8120
35
0.02234
0.8971
36
0.13428
0.4349
36
-0.20550
0.2292
36
0.02130
0.9019
36
0.59231
0.0001
36
0.86920
0.0001
36
1.00000
0.0
36
Notes: 1. Correlations were conducted on the natural logarithm transformed dust-lead loadings and blood-lead and water-lead concentrations.
2. First number is the Pearson correlation coefficient, the second number is the p-value, and the third number is the number of observations
-------
Table B-7b. Correlaton of Lg-Transormed Blood-Lead (//g/dU, Dust-Lead (f/g/g), and Water-Lead l/yg/L) Concentrations.
ซP
CO
Blood
Floor Dust
Interior Entryway Oust'
Window Nell Dust
Window S111 Dust
Exterior Entryway Dust
Water
Door Paint Hazard Score
Door Paint XRF
Window Paint Hazard Score
Window Paint XRF
Bl ood
1.00000
0.0
87
0.53302
0.0001
87
0.56015
0.0001
87
0.40277
0.0001
86
0.48946
0.0001
87
0.4674S
0.0001
85
-0.05243
0.6296
87
-0.17675
0.3024
36
-0.14461
0.4001
36
-0.11413
0.5075
36
-0.30488
0.0706
36
Floor
Dust
0.53302
0.0001
87
1.00000
0.0
87
0.60667
0.0001
87
0.68940
0.0001
86
0.74478
0.0001
87
0.58587
0.0001
85
0.01357
0.9007
87
-0.27486
0.1047
36
0.01775
0.9181
36
0.22005
0.1972
36
0.15373
0.3707
36
Interior
Entrway
Dust
0.56015
0.0001
87
0.60667
0.0001
87
1.00000
0.0
87
0.52726
0.0001
86
0.65675
0.0001
87
0.75587
0.0001
85
0.01781
0.8700
87
-0.13834
0.4210
36
-0.38511
0.0204
36
-0.18070
0.2916
36
0.22767
0.1817
36
Window
Well
Dust
0.40277
0.0001
86
0.68940
0.0001
86
0.52726
0.0001
86
1.00000
0.0
86
0.83469
0.0001
86
0.59699
0.0001
84
0.00512
0.9627
86
0.21543
0.2139
35
0.09517
0.5865
35
0.12491
0.4746
35
0.00126
0.9943
35
Window
S111
Dust
0.48946
0.0001
87
0.74478
0.0001
87
0.65675
0.0001
87
0.83469
0.0001
86
1.00000
0.0
87
0.73108
0.0001
85
0.05961
0.5834
87
-0.21133
0.2160
36
-0.15752
0.3589
36
0.09436
0.5841
36
0.11403
0.5079
36
Exterior
Entryway
Dust
0.46745
0.0001
85
0.58587
0.0001
85
0.75587
0.0001
85
0.59699
0.0001
84
0.73108
0.0001
85
1.00000
0.0
85
0.07175
0.5140
85
0.01723
0.9206
36
0.09880
0.5665
36
0.29418
0.0816
36
0.25049
0.1406
36
Water
-0.05243
0.6296
87
0.01357
0.9007
87
0.01781
0.8700
87
0.00512
0.9627
86
0.05961
0.5834
87
0.07175
0.5140
85
1.00000
0.0
87
0.28190
0.0958
36
0.02113
0.9026
36
-0.20182
0.2379
36
-0.20550
0.2292
36
Door
Paint
Hazard
-0.17675
0.3024
36
-0.27486
0.1047
36
-0.13834
0.4210
36
0.21543
0.2139
35
-0.21133
0.2160
36
0.01723
0.9206
36
0.28190
0.0958
36
1.00000
0.0
36
0.51022
0.0015
36
0.16400
0.3392
36
0.02130
0.9019
36
Door
Paint
XRF
-0.14461
0.4001
36
0.01775
0.9181
36
-0.38511
0.0204
36
0.09517
0.5865
35
-0.15752
0.3589
36
0.09880
0.5665
36
0.02113
0.9026
36
0.51022
0.0015
36
1.00000
0.0
36
0.63836
0.0001
36
0.59231
0.0001
36
Window
Paint
Hazard
-0.11413
0.5075
36
0.22005
0.1972
36
-0.18070
0.2916
36
0.12491
0.4746
35
0.09436
0.5841
36
0.29418
0.0816
36
-0.20182
0.2379
36
0.16400
0.3392
36
0.63836
0.0001
36
1.00000
0.0
36
0.86920
0.0001
36
Window
Paint
XRF
-0.30488
0.0706
36
0.15373
0.3707
36
-0.22767
0.1817
36
0.00126
0.9943
35
0.11403
0.5079
36
0.25049
0.1406
36
-0.20550
0.2292
36
0.02130
0.9019
36
0.59231
0.0001
36
0.86920
0.0001
36
0.00000
0.0
36
Notes: 1 Correlations were conducted on the natural logarithm transformed dust-lead, blood-lead, and water-lead concentrations.
2. First number is the Pearson correlation coefficient, the second number is the p-value, and the third number is the number of observations
-------
This page intentionally left blank.
-------
APPENDIX C
Results from the Pathways
Analyses of the Rochester Data
C-1
-------
Table C-1. Structural Equation Modeling Results for the Rochester Environmental-Lead
Pathways, Including the Window Paint and Door Paint Hazard Score Pathways
Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (fjglg).
Variables
\lndependent
Dependent \
Direct Effect Parameter Estimates (t-value) :
Interior
Entryway
Dust
Window
Sill Dust
Window
Well Dust
- ' Dust-Lead
Floor
Interior
Entryway
Window Sill
Window Well
0.1909*
(3.48)
0.2014*
(2.94)
-0.0642
(0.57)
0.0033
(0.06)
0.0560
(0.60)
0.3450*
(5.09)
Exterior
Entryway
Dust
Soil
Window
Paint
Hazard
Score
Door
Paint
Hazard
Score
-, /
^ ' .
' I* * " *
t
Ra
Loadings (yq/ft2)
0.0399
(0.63)
0.1978
(1.95)
0.1511
(1.24)
0.5000*
(2.57)
0.1115
(0.72)
0.7866*
(4.09)
0.0102
(0.38)
0.0206
(0.47)
0.0907*
(2.69)
0.1591*
(3.77)
-0.0476
(-1.33)
0.1292*
(2.25)
0.25
0.16
0.33
0.25
# " tf Dust-Lead Concentrations (A/g/g) , , 'M
Floor
Interior
Entryway
Window Sill
Window Well
0.4103*
(5.29)
0.1582*
(2.27)
-0.0513
(-0.62)
-0.0486
(-0.73)
0.1915*
(2.48)
0.3740*
(4.75)
-0.0855
(-1.25)
0.1803*
(2.27)
-0.1314
(-1.23)
0.0940
(0.75)
0.0115
(0.08)
0.6619*
(4.43)
0.0456*
(1.97)
-0.0233
(-0.85)
0.0965*
(3.38)
0.0945*
(2.93)
-0.0280
(-0.90)
0.1084*
(3.05)
0.26
0.19
0.33
0.24
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value ^1.96 or
ฃ -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/
concentrations and soil-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9845, and 0.9897 for the dust-
lead concentration model.
C-2
-------
Table C-2. Structural Equation Modeling Results for the Rochester Environmental Pathways
Model, Including Window and Door Paint XRF Measurement Pathways - Dust-
Lead Loadings fc/g/ft2) and Dust-Lead Concentrations (/t/g/g).
Variables
N. Independent
Dependent \
^=^=^^^^=
4l
Interior
Entryway
Dust
====!S!S====:=S=a========B=^^
Direct Effect Parameter Estimates (t-value)
Window
Sill Dust
- . .
Floor
Interior
Entryway
Window Sill
Window Well
0.1786*
(3.22)
0.1943*
(2.86)
0.0218
(0.20)
Window
Well Dust
- S
Exterior
Entryway
Dust
Soil
Window
Paint
XRF
Door
Paint
Dust-Lead Loadings (//g/ft2)
0.0081
(0.14)
0.0604
(0.66)
0.3705*
(5.51)
0.0408
(0.64)
0.1700
(1.67)
0.1785
(1.40)
0.5836*
(2.92)
0.0778
(0.48)
0.7549*
(3.69)
-0.0072
(-0.26)
-0.0712
(-1.64)
0.0597*
(1.96)
0.1131*
(2.88)
-0.0043
(-0.16)
0.1055*
(2.49)
=====
0.23
0.16
0.32
0.22
,,--,' . Dust-Lead Concentrations U/g/g) , -
Floor
Interior
Entryway
Window Sill
Window Well
0.3923ซ
(5.05)
0.1627*
(2.38)
-0.0126
(-0.16)
-0.0494
(-0.74)
0.1647*
(2.12)
0.3936*
(4.89)
-0.0941
(-1.37)
0.1729*
(2.17)
-0.1481
(-1.35)
0.1402
(1.08)
-0.0080
(-0.06)
0.6169*
(3.98)
0.0399
(1.71)
-0.0437
(-1.60)
0.0645*
(2.42)
0.0847*
(2.87)
-0.0029
(-0.12)
0.0802*
(3.04)
0.26
0.18
0.30
0.24
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value *1.96 or
z -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/
concentrations and soil-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9909, and 0.9884 for the dust-
lead concentration model.
C-3
-------
Table C-3. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways, Including the Window Paint and
Door Paint Hazard Score Pathways - Dust-Lead Loadings (/ig/ft2) and Dust-Lead Concentrations U/g/g).
Variables
\lndependent
Denendent x
Interior
EntryWay
Dust
Window
Sill Dust
, ,t , > -
Window
Well Dust
^ J i '
" 'J
Exterior
Entryway
Dust
^Floor*
Dust
eter Estimate* (t-walua)
Soil
i !
window
paint ,
Hazard
Score
Door
Paint
Hazard
Score
Water
Mouthing
! f "
R*
; l Dust-Lead Loadinos (un/fta) "
Blood
Floor
Interior
Entrvwav
Window
Sill
Window
Well
-0.0405
(-1.90)
0.1941ซ
(3.49)
0.0022
(0 08)
0.1975*
(2.85)
-0.0843
(-0.75)
0.0877*
(4 2O)
-0.0058
(-0 10)
0.0841
(0.88)
0.3532*
(5.041
-0.0177
(-O 74)
0.0401
(0 63)
0.2016*
(1.99)
0.1125*
(3 38)
0.1763
(1.40)
0.4648*
12.32}
0.1121
(0.70)
0.8601*
(4.50)
-0.0236*
(-2.30)
0.0087
(0.32)
0.0188
(0.42)
0.0895*
(2.65)
0.1483*
(3.57)
0.0687*
(4.28)
-0.0405
(-0.96)
0.1543*
(2.30)
-0.0924
(-0.90)
0.2307
(1.83)
0.35
0.25
0.16
0.34
0.27
,''''!., " Dust-Lead Concentrations (i/ti/a) - - ~ -
Blood
Floor
Interior
Entrywav
Window Sill
Window Well
-0.0665
(-1.571
0.4094*
(5.28)
0.0045
(0.13)
0.1515*
(2.16)
-0.0514
(-0.61)
0.0621*
(1.96)
-0.0586
(-0.87)
0.1905*
(2.42)
0.3676*
(4.581
0.0111
(0.33)
-0.0898
(-1.31)
0.1799*
(2.24)
0.0667
(1.46)
-0.0974
(-0.89)
0.0963
(0.73)
0.0360
(0.25)
0.7035
(4.65)
-0.0138
(-1.17)
0.0436
(1.86)
-0.0234
(-0.83)
0.0949*
(3.30)
0.0886*
(2.75)
0.0704*
(3.83)
-0.0233
(-0.64)
0.1079*
(2.56)
-0.0093
(-0.08)
0.3236*
(2.24)
0.21
0.26
0.17
0.32
0.25
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value ;> 1.96 or <, -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations, soil-lead concentrations, and
blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9649, and 0.9635 for the dust-lead concentration model.
-------
Table C-4. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways Model, Including Window and
Door Paint XRF Measurement Pathways Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (//g/g).
II Variables
| \lndependent
|| , \. ,
|l Dependent \
Direct E
.
<*
Interior
Entryway
Dust
^ s
Window
SHI Dust
E A
Window
Well Dust
!
Exterior
Entryway
Dust
fects Paramf
Floor
Dust
tar Estimates
Soil
A
(t-value) '
: " *
,
Window
Paint
XRF
'
Door
Paint
XRF *
-'
Water
* *
Mouthing
r *
'-.
J
,-
Rซ -1
Dust-Lead Loadings (i/o/ft*l
Blood
B
1 Floor
| Interior
II Entrvwav
Window
Sill
I Window
I) Well
-0.0314
(-1.411
0.1813ป
J3.251
0.0331
(1.211
0.1877*
12.741
0.0190
(0.171
0.0803*
(3.621
0.0008
(0.01)
0.0847
10.88)
0.3755*
(5.37)
-0.0164
(-0.65)
0.041 1
(0.641
0.0713
(1.67)
0.0993*
(2.85)
0.2111
(1.61)
0.5451*
(2.64)
0.0872
(0.531
0.8233*
(4.10)
-0.0233*
(-2.22)
-0.0113
(-0.411
-0.0750
(-1.681
0.0581
(1.90)
0.1142*
(2.99)
0.0159
(1.45)
0.0056
(0.19)
0.1075*
(2.371
^0.0776
(-0.7?!
0.3291*
(2.53)
0.29 I
0.24
0.15 ||
0.32 I
0.25
o
61
Dust-Lead Concentrations (i/o/ol
[ Blood
1
I Floor
II
11 Interior
| Window Sill
Window Well
-0.0436
(-0.99)
0.3935*
(S.08)
0.0310
(0.86)
0.1532*
(2.241
-0.0117
(-0.14)
0.0570
(1.70)
-0.0641
(-0.95)
0.1628*
(2.06)
0.3807*
(4.63)
0.0156
(0.441
-0.0997
(-1.451
0.1725*
(2.14)
0.0669
(1.39)
-0.1053
(-0.941
0.1337
(1.001
0.0245
(0.171
0.6565*
(4.231
-0.0169
M.40I
0.0387
(1.63)
-0.0402
(-1.431
0.0652*
(2.431
0.0852*
(2.921
0.0142
(0.13)
0.0034
(0.141
0.0742*
(2.61)
-0.0066
1-noKi
0.4362*
(2.941
=^
0.14 |
0.26 If
0.17
0.30
0.26 |
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value a 1.96 or s-1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations, soil-lead concentrations, and
blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9759, and 0.9673 for the dust-lead concentration model.
-------
Table C-5. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways, Including Hand Lead and Window
and Door Paint Pathways Dust-Lead Loadings (j/g/ft2) and Dust-Lead Concentrations (fjglg).
Variables
Nsjndependent
Deoendent x
Interior
Entryway
Dust
Window
Sill Dust
- b
a
( 5
Window
Well Dust
/ |
Exterior
Entryway,
Dust
rect Effects
T.
3
t, ^
f J
f
V Floor/
Dust
^
i
J
Hand'
" Lead
Soil
Window
Paint ''j-
Hazard
Score
i
Door
Paint'
'" Hazard
Score
Water
'!
Mouthlnq
'C
R"
. Dust-Lead Loadfnas (//a /ft2!
Blood
Hand Lead
Floor
Interior
Entrywav
Window Sill
Window Well
-0.0264
(-1.271
-0.0321
(-1.10)
0.2029*
(3.561
0.0162
(0 63)
-0.0311
(-0 87)
0.2037*
(2.87)
-0.0918
(-0.80)
0.0805*
(3.81)
0.0510
(1 69)
0.0122
(0.20)
0.0598
(0.59)
0.3362*
(4.49)
-0.0272
(-1 16)
0.0527
(1 63)
0.0459
(0.70)
0.1940
(1.85)
0.0825*
(2.51)
0.1049*
(2 32)
0.2163*
(3 27)
0.0518
(0 81)
0.1714
(1.33)
0.4653*
(2.26)
0.1097
(0.67)
0.7485*
(3.93)
-0.0278*
(-2.78)
0.0105
(0.75)
0.0030
(0.10)
0.0239
(0.52)
0.0927*
(2.66)
0.1529*
(3.74)
0.0578*
(3.57)
0.0371
(1.64)
-0.0473
(-1 .04)
0.1694*
(2.35)
-0.0591
(-0.59)
0.2563*
(2.06)
-0.0566
(-0.33)
0.40
0.17
0.25
0.15
0.31
0.26
' ' ' ': * > ' , - Dust-Lead Concentrations (oci/ft2) ** --*'- ,. :
Blood
Hand Lead
Floor
Interior
Entrvwav
Window Sill
Window Well
-0.0523
(-1.26)
-0.0308
(-0.56)
0.4228*
(5.07)
0.0195
(0.59)
-0.0390
(-0.91)
0.1441*
(2.01)
-0.0226
(-0.28)
0.0316
(1.0O)
0.0843*
(2.0H
-0.0594
(-0.84)
0.1235
(1.55)
0.3804*
(4.45)
-0.0022
(-0.07)
0.0422
(1.00)
-0.0888
(-1.27)
0.1708*
(2.19)
0.0484
(1.131
0.0519
(0.93)
0.3135*
(4.33)
0.0701
(1.04)
-0.1084
(-0.97)
0.1255
(1.00)
0.0263
(0.18)
0.6311*
(4.23)
-0.0187
(-1.68)
0.0118
(0.80)
0.0473*
(1.96)
-0.0267
(-0.98)
0.0949*
(3.22)
0.0904*
(2.87)
0.0593*
(3.26)
0.0345
(1.45)
-0.0356
(-0.91)
0.1311*
(3.05)
0.0458
(0.41)
0.3457*
(2.48)
-0.0918
(-0.50)
*
0.31
0.12
0.25
0.16
0.32
0.24
o
en
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value ^1.96 or ^1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations, soil-lead concentrations, hand-lead
measurements, and blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9666, and 0.9660 for the dust-lead concentration model.
-------
Table C-6. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways Model, Including Soil Coverage
and Window and Door Paint Pathways Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (//g/g).
II Variables
1 Ntadependent
II Denend.e.nl \
*'"'*< Direct ฃ
*
Interior
Entryway
Dust
V
x ,
4,
Window
" Sill Dust
"" t
>k
Window
Well Dust
*\ ;*
, Exterior
Entryway
Dust
1 D
Blood
1 Floor
II Interior
II E^frvwav
1 Window
Sill
1 Window
| Well
1
|| Blood
| Floor
|| Interior
| En*rvwav
1 Window Sill
| Window Well
-0.0405
(-1.901
0.2036*
(3.68)
0.0022
(0.08)
0.2066*
(2.98)
-0.0617
(-0.54)
0.0877*
(4.20)
0.0099
(0.17)
0.1236
(1.32)
0.3715 *
(B.59)
-0.0177
(-0.75)
0.0449
(0.71)
0.2206*
(2.16)
Dus1
-0.0600
(-1.30)
0.4482ป
(533)
,
0.0055
(0,16)
0.1552*
(2,211
-0.0594
(-0.77)
0.0602
(1.88)
-0.0867
1-1,33)
0.2045*
(2.94)
0.3767*
(4.961
0.0089
(0.26)
-0.1122
(-1,64)
0.2077*
(2.831
fects Pararnt
iU
ซf . , -
i-
"- -, <
. V
Floor
-QllSi
tar Estimates
Soil
Coverage
It-value)
Window
Paint
Hazard
Score
<4
i
1 I 4f-
Door
Paint
Hazard
Score
1
Water
i"
- ;
Mouthlna
xi" . 'll
ฐ,ซH H
*-ฃ?> * I
"1
R* I
4 ' ~ II
jst-Lead Loadings (oa/fta) l|
0.1125*
(3.39)
-lead Conce
0.0647
(1.40)
0.0000
(0.81)
0.0001
(1.62)
0.0000
(0.02)
0.0002*
(2.33)
-0.0236*
(-2.31)
0.0115
(0.42)
0.0260
(0.58)
0.0922*
(2.74)
0.1808*
(4.21)
0.0687*
(4.29)
-0.0499
(-1.19)
0.1320
(1.94)
-0.0924
(-0.90)
0.2307
(1.83)
itratlons (ua/al w;. .-.. "i. .'. : -''.:, ,.,
0.0000
(-0.33)
0.0001
(1.37)
-0.0000
1-0.02)
0.0001*
(2.371
-0.0135
(-1.14)
0.0425
(1.831
-0.0256
(-0,93)
0.0958*
(3.341
0.1180*
(3.52)
0.0698*
(3.79)
-0.0227
(-OR3I
0.0993*
(2551
-0.0080
(-0.071
0.3256*
(2.2S)
1
0.36 II
0.24 j
0.13 I
0.33 I
0.18
II
0.21 I
0.26 |
0.21 Jj
0.32
0.16 ||
o
ij
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value si.96 or <. -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations, soil coverage values, and
blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9686, and 0.9705 for the dust-lead concentration model.
-------
Table C-7. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways, Including Proportion of Carpeting
and Indicator of Interior Entry way Carpet Pathways Dust-Lead Loadings U/g/ft2) and Dust-Lead
Concentrations (//g/g).
Variables
\lndependent
Dependent N,
-. Direct Effects Parameter Estimates (t-value)
Interior
Entryway
' Dust
Window
Sill Dust
' Window
Well Dust
Blood
Floor
Interior
Entryway
Window Sill
Window Well
Blood
Floor
Interior
Entryway
Window Sill
Window Well
0.0071
(0.28)
0.1862ป
(2.69)
0.0114
(0.47)
0.1624ป
(2.47)
-0.0801
(-0.93)
0.0789*
(4.08)
0.0117
(0.21)
0.0436
(0.60)
0.3532*
(5.04)
^ t, ,
J<*
Exterior
Entryway
Dust
f
V *'> .=!
"s
i
5 -Floor
..-, Dust
V
Soil
i
Window
, Paint
' Hazard
Score
Dust-Lead Loadings (//g/ft2)
-0.0395
(-1 .74)
0.0660
(1.07)
0.2350*
(3.03)
0.1428*
(4.41)
0.1933
(1.61)
0.3190*
(2.07)
0.1121
(0.70)
0.8601*
(4.50)
-0.0246*
(-2.58)
0.0084
(0.32)
0.0187
(0.55)
0.0895*
(2.65)
0.1483*
(3.57)
*-*
1 /
Y ^ ^
Door
Paint
Hazard ,
Score ,,'
, ' ;
11 i
Water
j.
X
Mouthing
T
Proportion
Carpet
Indicator
Interior
Entryway
Carpet
R*
* ฃ s i
0.0455*
(2.91)
-0.0033
(-0.08)
0.1943*
(3.78)
-0.0496
(-0.52)
0.2795*
(2.39)
-0.6866*
(-2.78)
2.3914*
(3.68)
x , ' Dust-Lead Loadings (//g/ft2) _ -
-0.0665
(-1.63)
0.4076*
(5.21)
0.0058
(0.17)
0.1514*
(2.17)
-0.0527
(-0.64)
0.0792*
(2.61)
-0.0507
(-0.75)
0.2036*
(2.62).
0.3676*
(4.58)
0.0006
(0.02)
-0.0953
(-1.37)
0.1781*
(2.24)
0.0554
(1.27)
-0.0964
(-0.87)
0.1084
(0.83)
0.0360
(0.25)
0.7035*
(4.65)
-0.0133
(-1.18)
0.0434
(1.85)
-0.0231
(-0.83)
0.0949*
(3.30)
0.0886*
(2.75)
0.0599*
(3.36)
-0.0279
(-0.76)
0.1026*
(2.45)
0.0289
(0.25)
0.3570*
(2.58)
-0.4672
(-1.69)
-0.2565
(-0.44)
-0.3368*
(-2.65)
-0.2026
(-0.58)
3.1860*
(9.40)
"._'_
-0.2732*
(-2.29)
-0.1047
(-0.41)
-0.4282
(-1.53)
0.45
0.32
0.51
0.34
0.27
0.30
0.27
0.19
0.32
0.25
9
00
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value ;>1.96 or s -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations, soil-lead concentrations, and
blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9665, and 0.9611 for the dust-lead concentration model.
-------
Table C-8. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways for Homes with Carpeted
Bedrooms and Play Areas Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (//g/g).
[Variables
1 \lndependent
llDeDendent > \
' '* DIM
Interior
Entryway
Dust
11 ^
Blood
II
Floor -
If
HCaroeted
|
|| Interior
II Entrvwav
| Window Sill
1 Window Well
0.0523
(1.82)
-0.0813
(-0.961
Window Sill
Dust '
IS *
Window
Well Dust
Exterior
Entryway
Dust
~" ^
tct Effegts Parameter Estimates (t-1
j
\
i ^
Carpeted
floor,
<* Diist
"" i
Soil
'4
t
Window
' Patnl :>'
Hazard ,
>'i * >
Score
* Dust-Lead Loadings (ua/ft1)
0.0406
(1.50)
0.0708
(0.89I
-0.1383
(-1.55)
0.0858*
(4.O21
-0.0257
(-0.391
0.0391
(0.531
0.3288*
(4.52)
-0.0343
(-1.39)
0.0807
11.14)
0.2178*
(2.82)
-0.0456
(-1.39)
-0.1007
(-0.69)
0.3086
(1.91)
0.1754
(1.03)
0.9342*
(4.54)
-0.0296*
(-2.83)
0.0037
(0.12)
-0.0277
(-O.82)
0.0933*
(2.73)
0.1375*
(3.20)
ktaluel " '~ ' '
(
Door
i Paint
Hazard
Score
1
Water
*
Mouthina
Indicator
Interior*
- Entryway
Garnet
0.0790*
(3.68)
-0.0243
(-0.40)
0.1683*
(2.49)
0.0780
(0.71)
0.2298
(1.80)
-0.5029*
(-3.49)
0.7413
(1.79)
3.3162*
(9 64)
" 1
I
0.41 I
II
0.05 I
0.54
0.35 I
0.28
o
cb
y Dust-Load Concentrations li/o/o)
r 1
| Blood
I Floor -
II Interior
llFntrvwnY
Window Sill
||W!ndow Well
-0.0481
(-1.261
0.0868
(1.03)
-0.0019
(-0.06)
0.0846
(1.12)
0.0024
(0.031
0.0721*
(2.31)
-0.0486
(-0,65)
0.2144*
(2.57)
0.3220*
I3.67)
-0.0017
f-0.051-
-0.0461
(-0.62)
0.1762*
(2.08)
0.0195
(0.44)
0.1304
(1.041
0.0092
(0.061
0.0654
(0.411
0.7592*
(4.70)
-0.0140
(-1.19)
0.0108
(0.42)
-0.0161
(-O.R4I
0.1019*
(3.33)
0.0813*
(2.441
0.1028*
(4.30)
-0.0195
(-0.371
0.0929
(1.55)
0.1543
(1.211
0.3316*
(2.271
-0.3938* '
(-3.25)
0.5005
(1.891
-0.2101
(-O.B9I
_-
0.29 |
0.08 |]
0.16 |
0.30
0.27 |
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value a 1.96 or <; -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations, soil-lead concentrations,
and blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9713, and 0.9809 for the dust-lead concentration model.
-------
Table C-9. Predicted Effect of 50% Decrease in Environmental-Lead Loadings U/g/ft2) and
Concentrations U/g/g) and Blood-Lead Concentrations (//g/dL) Based on the
Blood-Lead Pathways SEM for the Rochester Data, Including Carpet and Paint
Pathways.
Sample Location
Geometric
Mean
50%
Decrease in
GM
Predicted Percent Change (%) "
(Lower Prediction Interval, Upper Prediction Interval) lbl
Blood
Floor
Interior
Entryway
Window
Sill
Window
Well
Dust-Lead Loading (/t/g/ft2)
Floor Dust
Interior Entryway
Dust
Window Sill
Window Weil
Soil""
Exterior Dust
Window Paint
Door Paint
Proportion
Carpet
* *. *
Indicator Interior
Entryway Carpet
*
Mouthing
Water
-,-' " : 1 ; . ":. -ry."
Floor Dust
Interior Entryway
Dust
Window Sill
Window Well
Soil
Exterior Dust
Window Paint
Door Paint
Proportion
Carpet
Indicator Interior
Entryway Carpet
Mouthing
Water
100
89
345
22,584
852
515
4.9
1.5
0.38
Presence
1
Exposure
1
1
50
45
173
11,292
426
258
2.5
0.75
0.19
Absence
0
No Exposure
0
0
-9 (-11, 7)
-2 (-4, 0)
-2 (-4, 0)
-6 (-7, -5)
NA
2 (0, 4)
0(-1,1)
-4 (-5, -3)
27 (8, 50)
NA
-12 (-23, 0)
-10 (-20, 2)
-5 (-14, 5)
-21 (-37, -1)
-7(-17, 4)
-2 (-7, 3)
-2 (-9, 6)
-81 (-94, -36)
f r '
29 (19, 40)
-32 (-65, 30)
": ' * --
-24 (-30, -18)
5 (-2, 12)
NA
NA
Dust-Lead Concentrations U/g/g
563
468
2,787
8,676
852
656
4.9
1.5
0.38
Presence
1
Exposure
1
1
282
234
1,394
4,338
426
328
2.5
0.75
0.19
Absence
0
No Exposure
0
0
-4 (-7, -17)
3 (-4, 0)
-1 (-4, 2)
-5 (-7, -3)
NA
1 (-2, 4)
0(-1, 1)
-4 (-5, -3)
40 (13, 74)
NA
-25(-34,-15)
-91-19,2)
-5 (-15, 6)
-1 (-17, 19)
2 (-9, 14)
-4 (-8, 0)
-1 (-7, 5)
19 (-55, 213)
NA
NA
6 (-14, 30)
-1 (-17, 18)
-20 (-45, 1 6)
-1 5 (-30, 3)
-1 (-9, 8)
-13 (-23, -1)
NA
NA
NA
NA
-22 (-35, -7)
-25 (-50,
13)
NA
-9 (-17, -1)
NA
NA
NA
NA
NA
NA
-45 (-7 1,4)
NA
-10 (-27, 3)
NA
NA
" ' "'*ซ,
-96 (-98, -
91)
'
NA
NA
)
NA
NA
4 (-12, 22)
-12 (-24, 3)
-15 (-34, 10)
-12 (-25, 3)
1 (-4, 7)
-7 (-14,1)
NA
30(19,43)
32(-13, 100)
-30 (-37, -22)
-3 ( -1 1 , 6)
NA
NA
53 (-11,
164)
'-
NA
NA
NA
NA
ซJ->
NA
NA
NA
NA
'
NA
NA
NA
-22 (-35, -7)
-18(-40,12)
NA
-8 (-14, -2)
NA
NA
NA
NA
NA
NA
NA
NA
NA
-39 (-58, -11)
NA
-6 (-1 3, 2)
NA
NA
NA
NA
NA
" Soil lead is measured as a concentration.
(b)
A prediction interval or forecasting interval is a confidence interval for the actual or future value of a response, which is
the mean value plus error. Here the upper and lower 95% prediction intervals are based on the direct effects only.
NA Indicates that the fitted pathway model did not include a pathway from the sample location.
C-10
-------
Table C-10. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways Model For African-American
Children Dust-Lead Loadings (/vg/ft2) and Dust-Lead Concentrations (/ig/g).
[variables -
I Njndependent
II Deoendent \
^-~> v - Dfrfe*i=
i
Interior
Entry way
' Dust
Window
Sill Dust
>
Window
Well Dust
V ซ
. *
'Exterior
Entryway
Dust
facts Paramf
Floor
Dust
ter Estimates
}
' E
Soil
,* ฃ ' 'V^~- **, ' ''
(t-valua) ' F-t?-
'' >
Window
Paint
Hazard
Score
T /
Door
Paint
Hazard
Score ^_
:t
.. Water
Mo'uthlno
R* '- , l|
Dust-Lead Loadings (uo/ft'l
Blood
Floor
It Interior
I Entrvwav
Window
Sill
Window
(I Well
-0.0061
(-0.18)
0.3301*
(4 09)
-0.0015
(-0.04)
0.2725*
12.88)
-0.2919
(-1 .70)
0.0630*
(2.15)
-0.0525
(-0.68)
0.0678
(0.47)
0.2588*
(2.1SI
0.0017
(0.05)
-0.0399
(-0.51)
-0.0158
(-0.11)
0.0596
(1.O4)
0.0267
(0.14)
0.9957*
(2.96)
0.2138
(0.74)
0.5628
(1.60)
0.0019
(0.07)
0.1158
(1.82)
0.1315
(1.13)
0.2110*
(2.82)
0.1431
(1.57)
0.0439
(1.54)
-0.0736
(-1.08)
0.0156
(0.121
-0.2349
(-1.221
0.3693
(1 87)
H '-"--."' ''-'' -' :' '" i. ^ '..;:-:/- /I': ?; '- .-^-vU: .-.:'v:.\V': Dust-LsaH Concentrations luala) ' ' ^:"-'i':-"" '.:. :. v;_: -y - ",-'- ,:
I Blood
| Floor
|| Interior
||
|| Window Sill
| Window Well
-0.0164
(-0.31)
0.2906*
(2.32)
0.0210
(039)
0.0318
(053)
-0.1649
(-1 001
0.0232
(O.R71
0.0644
(0.62)
0.2081
(1.68)
0.2345*
(2.13)
0.0170
(0.40)
-0.0802
(-0.831
0.1035
(0.88)
0.0661
(1,06)
-0.1594
(-0.80)
0.2844
(1.181
0.2057
(0.93)
0.7935*
(2.80)
0.0161
IQ.5.9)
0.0599
(0.95)
-0.0517
<-0,fi7)
0.1287*
(2,37)
0.1407
(1.93)
0.0339
11.09)
0.0043
(O.Ofi)
0.1541
(1.86)
-0.3033
(-1.42)
0.4809*
(2.24)
0.38 |
0.50
0.24
0.32 |
0.13
0.35
0.21 I
0.24 1
0.34 H
0.26 I
o
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value 21.96 or s -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations, soil-lead concentrations, and
blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9508, and 0.9000 for the dust-lead concentration model.
-------
Table C-11. Structural Equation Modeling Results for the Rochester Blood-Lead Pathways Model For Children of All Other
Races Dust-Lead Loadings (//g/ft2) and Dust-Lead Concentrations (f/g/g).
^independent
Dependent ^^
^^^^^"M
Interior
Entry way,
Dust
IMHMBMMB^MIM
Window
Sill Dust
^BH^^_H^^_~^M^_
Window
Well Dust
r > Direct E
Exterior
Entfyway
* Dust
facts Parami
Floor
Dust
' Dust-Lead Loa
Blood
Floor
Interior
En^rvvtfaY
Window
Sill
Window
Well
-0.0684ป
(-2.82)
0.0958
11.261
0.0036
(0.12)
0.1387
(1.43)
0.1043
(0.72)
f ,-:.-... ,; ';';'.' ''I' -'". ;.*
Blood
Floor
Interior
Entrywav
Window SHI
Window Well
-0.0685
(-1.32)
0.5122ป
(4.90)
-0.0414
(-1.141
0.2076*
(2.58)
0,041 2
(0.46)
0.0828*
(3.4O1
0.0649
(0.79)
0.0210
(0.17)
0.4270*
(5.09)
-0.0335
(-1.O5)
0.1163.
(1.181
0.4522*
(3.281
0.1478*
(4.13)
tar Estimate!
Soil
Window
Paint Hazard
. Score ,
I
Door
Paint
Hazard
ป Score
M^^M^_ซซ^BPซM_>
Water
^MIIM
Mouthina
Hlnns d/n/ft'l -
0.1612
(1.011
0.1922
(0.81)
-0.0038
(-0.02)
0.9828*
(4.44)
' r Dust-Lead Concentrations (1/9
0.1409*
(3.89)
-0.1575
(-1.79)
0.0976
(1.00)
0.4901*
(4.31)
-0.0539
(-1.171
-0.1075
(-0.991
0.4095*
(3.651
0.0950
(1.881
-0.0491
(-0.36)
0.0005
(0.00)
-0.1025
(-0.55)
0.6872*
(3.97)
-0.0215*
(-2.23)
-0.021 1
(-O.69)
0.0320
(0.70)
0.0456
(1.30)
0.1500*
(3.38)
0.0640*
(3.41)
-0.0703
(-1.21)
0.1766*
(2.09)
-0.1719
(-1.621
0.1941
(1.44)
B1
0.43
0.20
0.22
0.40
0.36
a) -3". '''.. * :'' . ;- 'vi,;*K
-0.0198
(-1.85)
0.0388
(1.56)
-0.0109
(-0.39)
0.0848*
12.49)
0.0701*
(2.06)
0.0587*
(2.84)
-0.0526
(-1.121
0.0912
(1.77)
-0.1160
(-0.97)
0.1828
(1.22)
0.28
0.34
0.23
0.34
0.27
o
M
Notes: 1. Bolded and a * indicate parameter estimates are significant at the 0.05 level. T-value ;> 1.96 or s -1.96 significant at 0.05 level.
2. Pathways analyses were conducted on the natural logarithm transformed dust-lead loadings/concentrations, soil-lead concentrations, and
blood-lead concentrations.
3. First number is estimated parameter; second number is corresponding t-value.
4. The goodness-of-fit index (GFI) for the dust-lead loading model is 0.9388, and 0.9541 for the dust-lead concentration model.
-------
Table C-12a.
Rochester Data Correlation of Log-Transformed Blood-Lead (//g/dL), Dust-Lead Loadings (fig/ft2), Soil-Lead (/sg/g)
Concentrations, and Child and Housing Characteristic Variables.
o
CO
Blood
Igiaaaa
vvvW
0.0
205
0.43006
0.0001
197
0.32454
0.0001
203
0.15113
0.0446
177
0.34288
0.0001
197
0.37437
0.0001
189
0.11389
0.1357
173
-0.22194
0.0014
205
0.36924
0.0001
187
0.36258
0.0001
196
0.17198
0.0159
196
0.11088
0.1190
199
0.07151
0.3155
199
0.04832
0.4947
202
Hand Lead
0.43006
0.0001
197
1.00000
0.0
197
0.36410
0.0001
195
0.15941
0.0379
170
0.25869
0.0003
189
0.32547
0.0001
181
0.2S563
0.0009
166
0.03467
0.6286
197
0.33040
0.0001
179
0.23871
0.0010
188
0.22575
0.0018
188
0.27208
0.0001
191
0.24014
0.0008
191
0.04415
0.5410
194
Floor
Oust
0.32454
0.0001
203
0.36410
0.0001
195
1.00000
0.0
203
0.31822
0.0001
177
0.34589
0.0001
196
0.24292
0.0008
188
0.21412
0.0049
171
0.23506
0.0007
203
0.26879
0.0002
185
0.13810
0.0548
194
0.14096
0.0499
194
0.23847
0.0007
197
0.16850
0.0179
197
-0.06289
0.3763
200
Interior
Entry
Oust
0.15113
0.0446
177
0.15941
0.0379
170
0.31822
0.0001
177
1.00000
0.0
177
0.22144
0.0035
172
0.21090
0.0067
164
0.21183
0.0088
152
-0.00978
0.8972
177
0.33916
0.0001
161
0.22591
0.0031
169
0.25711
0.0007
169
0.18080
0.0180
171
0.14346
0.0612
171
0.00412
0.9569
175
Window
sin oust
0.34288
0.0001
197
0.25869
0.0003
189
0.34589
0.0001
196
0.22144
0.0035
172
1.00000
0.0
197
0.55848
0.0001
184
0.26942
0.0004
169
-0.08404
0.2403
197
0.37066
0.0001
181
0.25226
0.0005
188
0.22337
0.0021
188
0.38279
0.0001
193
0.41298
0.0001
193
0.05496
0.4454
195
Window
Well Dust
0.37437
0.0001
189
0.32547
0.0001
181
0.24292
0.0008
188
0.21090
0.0067
164
0.55848
0.0001
184
1.00000
0.0
189
0.23358
0.0030
160
-0.13024
0.0741
189
0.44630
0.0001
174
0.16490
0.0270
180
0.12814
0.0865
180
0.44204
0.0001
186
0.41139
0.0001
186
-0.06286
0.3940
186
Exterior
Entry
Dust
0.11389
0.1357
173
0.25563
0.0009
166
0.21412
0.0049
171
0.21183
V OBuu
152
0.26942
0.0004
169
0.23358
0.0030
160
1.00000
0.0
173
-0.12940
0.0897
173
0.20777
0.0082
161
0.12076
0.1223
165
0.19326
0.0129
165
0.11331
0.1436
168
0.17624
0.0223
168
0.01958
0.7999
170
Indicator of
Proportion
Carpeted
-0.22194
0.0014
205
0.03467
0.6286
197
0.23506
0.0007
203
-0.00978
0.8972
177
-0.08404
0.2403
197
-0.13024
0.0741
189
-0.12940
0.0897
173
1.00000
0.0
205
-0.09477
0.1970
187
-0.14334
0.0450
196
-0.15836
0.0266
196
-0.06647
0.3509
199
-0.10256
0.1494
199
0.04522
0.5228
202
Soil
0.36924
0.0001
187
0.33040
0.0001
179
0.26879
0.0002
185
0.33916
0.0001
161
0.37066
0.0001
181
0.44630
0.0001
174
0.20777
0.0082
161
-0.09477
0.1970
187
1.00000
0.0
187
0.10603
0.1578
179
0.17742
0.0175
179
0.24743
0.0008
182
0.35131
0.0001
182
0.07652
0.3006
185
Door
Paint
Hazard
Score
0.36258
0.0001
196
0.23871
0.0010
188
0.13810
0.0548
194
0.22591
0.0031
169
0.25226
0.0005
188
0.16490
0.0270
180
0.12076
0.1223
165
-0.14334
0.0450
196
0.10603
0.1578
179
1.00000
0.0
196
0.53604
0.0001
196
0.26511
0.0002
190
0.19748
0.0063
190
-0.06131
0.3970
193
Door
Paint
XRF
0.17198
0.0159
196
0.22575
0.0018
188
0.14096
0.0499
194
0.25711
0.0007
169
0.22337
0.0021
188
0.12814
0.0865
180
0.19326
0.0129
165
-0.15836
0.0266
196
0.17742
0.0175
179
0.53604
0.0001
196
1.00000
0.0
196
0.30336
0.0001
190
0.36710
0.0001
190
-0.05903
0.4148
193
Window
paint
Hazard
Score
0.11088
0.1190
199
0.27208
0.0001
191
0.23847
0.0007
197
0.18080
0.0180
171
0.38279
0.0001
193
0.44204
0.0001
186
0.11331
0.1436
168
-0.06647
0.3509
199
0.24743
0.0008
182
0.26511
0.0002
190
0.30336
0.0001
190
1.00000
0.0
199
0.77743
0.0001
199
-0.00707
0.9216
196
Window
Paint
XRF
0.07151
0.3155
199
0.24014
0.0008
191
0.16850
0.0179
197
0.14346
0.0612
171
0.41298
0.0001
193
0.41139
0.0001
186
0.17624
0.0223
168
-0.10256
0.1494
199
0.35131
0.0001
182
0.19748
0.0063
190
0.36710
0.0001
190
0.77743
0.0001
199
1.00000
0.0
199
0.01139
0.8742
196
Water
0.04832
0.4947
202
0.04415
0.5410
194
-0.06289
0.3763
200
0.00412
0.9569
175
0.05496
0.4454
195
-0.06286
0.3940
186
0.01958
0.7999
170
0.04522
0.5228
202
0.07652
0.3006
185
-0 06131
mf EFU &W 1
0.3970
193
-0 05901
V * V^J7KIซJ
04140
*r JL*f O
193
-0.00707
0921fi
^** 4 v
196
0
a B74ป
196
i aaaaa
* . VVVVV
0a
V
202
Blood
Hand Lead
Floor Dust
Interior Entryway Dust
Window Sill Dust
Window Well Dust
Exterior Entryway Oust
Indicator of
Proportion Carpeted
Soil
Door Paint Hazard Score
Door Paint XRF
Window Paint Hazard
Score
Window Paint XRF
Water
Notes: 1. Correlations were conducted on the natural logarithm transformed hand-lead measurements, dust-lead loadings, and blood-lead, soil-lead and
2. First number is the Pearson correlation coefficient, the second number is the p-value, and the third number is the number of observations
water-lead concentrations.
-------
Table C-12b.
Rochester Data Correlation of Log-Transformed Blood-Lead (/ig/dL), Dust-Lead if/gig), and Soil-Lead
and Child and Housing Characteristic Variables.
Concentrations,
Blood
Hand Lead
Floor Dust
Interior Entryway
Dust
Window S111 Oust
Window Well Dust
Exterior Entryway
Dust
Indicator of
Proportion Carpeted
Soil
Door Paint Hazard
Score
Door Paint XRF
Window Paint Hazard
Score
Window Paint XRF,
Water
Blood
1.00000
0.0
205
0.43006
0.0001
197
0.13216
0.0602
203
0.08323
0.2749
174
0.23639
0.0008
199
0.21112
0.0036
188
0.11436
0.1352
172
-0.22194
0.0014
205
0.36924
0.0001
187
0.36258
0.0001
196
0.17198
0.0159
196
0.11088
0.1190
199
0.07151
0.3155
199
0.04832
0.4947
202
Hand Lead
0.43006
0.0001
197
1.00000
0.0
197
0.12492
0.0819
195
0.13622
0.0783
168
0.19449
0.0070
191
0.30279
0.0001
180
0.15863
0.0419
165
0.03467
0.6286
197
0.33040
0.0001
179
0.23871
0.0010
188
0.22575
0.0018
18
0.27208
0.0001
191
0.24014
0.0008
191
0.04415
0.5410
194
Floor
Oust
0.13216
0.0602
203
0.12492
0.0819
195
1.00000
0.0
203
0.34997
0.0001
173
0.29123
0.0001
197
0.21547
0.0031
186
0.13863
0.0714
170
-0.07530
0.2857
203
0.09887
0.1806
185
0.07819
0.2785
194
0.11357
0.1149
194
0.21859
0.0020
197
0.23000
0.0011
197
0.06806
0.3383
200
Interior
Entry
Dust
0.08323
0.2749
174
0.13622
0.0783
168
0.34997
0.0001
173
1.00000
0.0
174
0.24679
0.0012
170
0.23814
0.0024
160
0.23838
0.0035
148
-0.08393
0.2709
174
0.24524
0.0018
159
0.23521
0.0023
166
0.31201
0.0001
166
0.08317
0.2838
168
0.09677
0.2121
168
0.11659
0.1277
172
Window
S111 Dust
0.23639
0.0008
199
0.19449
0.0070
191
0.29123
0.0001
197
0.24679
0.0012
170
Iftaaaa
vvvvv
0.0
199
0.55223
0.0001
183 1
0.27469
0.0003
168
-0.03245
0.6491
199
0.33537
0.0001
182
0.22162
0.0021
190
0.19938
0.0058
190
0.37154
0.0001
194
0.33780
0.0001
194
0.02750
0.7013
197
Window
Well Dust
0.21112
0.0036
188
0.30279
0.0001
180
0.21547
0.0031
186
0.23814
0.0024
160
0.55223
0.0001
183
1.00000
0.0
188
0.21672
0.0062
158
-0.00234
0.9745
188
0.49852
0.0001
173
0.10959
0.1442
179
0.12936
0.0844
179
0.36497
0.0001
185
0.35457
0.0001
185
0.03503
0.6359
185
Exterior
Entry
Dust
0.11436
0.1352
172
0.15863
0.0419
165
0.13863
0.0714
170
0.23838
0.0035
148
0.27469
0.0003
168
0.21672
0.0062
158
1.00000
0.0
172
-0.15739
0.0392
172
0.28968
0.0002
160
0.10459
0.1826
164
0.14948
0.0561
164
0.07633
0.3269
167
0.14292
0.0654
167
0.03524
0.6492
169
Indicator of
Proportion
Carpeted
-0.22194
0.0014
205
0.03467
0.6286
197
-0.07530
0.2857
203
-0.08393
0.2709
174
-0.03245
0.6491
199
-0.00234
0.9745
188
-0.15739
0.0392
172
1.00000
0.0
205
-0.09477
0.1970
187
-0.14334
0.0450
196
-0.15836
0.0266
196
-0.06647
0.3509
199
-0.10256
0.1494
199
0.04522
0.5228
202
Soil
0.36924
0.0001
187
0.33040
0.0001
179
0.09887
0.1806
185
0.24524
0.0018
159
0.33537
0.0001
182
0.49852
0.0001
173
0.28968
0.0002
160
-0.09477
0.1970
187
1.00000
0.0
187
0.10603
0.1578
179
0.17742
0.0175
179
0.24743
0.0008
182
0.35131
0.0001
182
0.07652
0.3006
185
Door
Paint
Hazard
Score
0.36258
0.0001
196
0.23871
0.0010
188
0.07819
0.2785
194
0.23521
0.0023
166
0.22162
0.0021
190
0.10959
0.1442
179
0.10459
0.1826
164
-0.14334
0.0450
196
0.10603
0.1578
179
1.00000
0.0
196
0.53604
0.0001
196
0.26511
0.0002
190
0.19748
0.0063
190
-0.06131
0.3970
193
Door
Paint
XRF
0.17198
0.0159
196
0.22575
0.0018
188
0.11357
0.1149
194
0.31201
0.0001
166
0.19938
0.0058
190
0.12936
0.0844
179
0.14948
0.0561
164
-0.15836
0.0266
196
0.17742
0.0175
179
0.53604
0.0001
196
1.00000
0.0
196
0.30336
0.0001
190
0.36710
0.0001
190
-0.05903
0.4148
193
Window
Paint
Hazard
Score
0.11088
0.1190
199
0.27208
0.0001
191
0.21859
0.0020
197
0.08317
0.2838
168
0.37154
0.0001
194
0.36497
0.0001
185
0.07633
0.3269
167
-0.06647
0.3509
199
0.24743
0.0008
182
0.26511
0.0002
190
0.30336
0.0001
190
1.00000
0.0
199
0.77743
0.0001
199
-0.00707
0.9216
196
Window
Paint
XRF
0.07151
0.3155
199
0.24014
0 . 0008
191
0.23000
0.0011
197
0.09677
0.2121
168
0.33780
0.0001
194
0.35457
0.0001
185
0.14292
0.0654
167
-0.10256
0.1494
199
0.35131
0.0001
182
0.19748
0.0063
190
0.36710
0.0001
190
0.7774
0.0001
199
1.00000
0.0
199
0.01139
0.8742
196
Water
0.04832
0.4947
202
0.04415
0.5410
194
0.06806
0.3383
200
0.11659
0.1277
172
0.02750
0.7013
197
0.03503
0.6359
185
0.03524
0.6492
169
0.04522
0.5228
202
0.07652
0.3006
185
-0.06131
0.3970
193
-0.05903
0.4148
193
-0.00707
0.9216
196
0.01139
0.8742
196
1.00000
0.0
202
Notes: 1. Analyses were conducted on natural logarithm transforrned hand-lead measurements, and dust-lead, soil-lead, and blood-lead concentrations.
2. First number is the Pearson correlation coefficient, and the second number is the number of observations.
-------
APPENDIX D
Pathway Diagrams
Identified in the Literature
D-1
-------
Table D-1. Description of Variables for Pathway Model from Exterior Surface Dust Lead,
Interior House Dust Lead and Childhood Lead Exposure in an Urban
Environment' bY Bornschein, et al [2].
Variable' ~
XRF Hazard
PbSS
PbD
PbH
PbB
... B m ' . . ,- . . , * * "*
Description ''
Paint hazard score derived from a linear combination of the product of a maximum
of 1 5 XRF measurements and the condition code (0 to 10) value for the painted
surface.
Soil surface scrapes were taken from 1 ) surfaces either paved with asphalt, 2)
concrete or brick, or 3) were composed of hard-packed soil devoid of vegetation.
Collected from areas where child played and/or immediately outside the dwelling
unit entry.
Floor dust lead in //g/g.
Hand lead samples taken from the surface of the child's hands by three repeated
wipings of each hand with a total of six wet wipes.
Blood lead concentrations collected by venipuncture.
Table D-2. Description of Variables for Pathway Model from Soil-Lead - Blood Lead
Relationship in a Former Lead Mining Town, bY Bornschein, et al [17].
Variable
XRF-EXT
XRF-INT
PbSS
PbSC
DIST. 2
PbD Floor
PbD Window
Age (yr)
PbH
Age x PbH
PbB
Description , * :*'.ซ
Maximum exterior XRF/house in mg/cm2
Maximum interior XRF/house in mg/cm2
Median soil surface scrapings from exposed soil in play areas, paths through yards
or playgrounds, and from paved areas immediately outside the house entry-
Median 1 inch soil core lead in //g/g.
Location of dwelling-proximity to the old railway right of way which originated at
the mill on the east side of town and ran parallel to the San Migual river on the
southern boundary of town.
Floor dust lead in //g/g.
Window sill dust lead in //g/g.
Age of child in years.
Hand lead in fjg for two hands from handwipes.
Interaction between hand lead and age of the child.
Blood lead concentrations collected by venipuncture.
D-2
-------
Table D-3. Description of Variables for Pathway Model from The lnfluences ofSocfa/and
Environmental Factors on Dust Lead, Hand Lead, and Blood Lead Levels
inYoung Children' by Bornschein, et al [1].
Housing quality including the age, type of home, and condition of the home
H.O.M.E.
Home Observation for Measurement of the Environment (HOME) used to
quantitate various aspects of the child's rearing environment.
SES
Socioeconomic status (SES) of families made through the use of the Hollingshead
Four-Factor Scale.
PbD
Interior surface dust collected by three sweeps of a defined area using a 2-liter/min
vacuum.
PbH
Hand lead recovered from surface of child's hands by repeated wipings of both
hands with a total of six wet wipes.
PbB
Blood lead concentration obtain via venipuncture.
Table D-4. Description of Variables for Pathway Model from Pathways of Lead
Contamination for the Brigham and Women's Hospital Longitudinal Study
Menton, et al [7].
Variable "
Refinishing
PbAir
Pb Dust Floor
Pb Dust Window Sill
Pb Soil
Pb Blood
, - ^^j
Description x1 <
An indication of the presence of any refinishing or renovation and remodeling
activity within the preceding 6 months.
Amount of paniculate lead U/g/m3) in air sampled at 24 months.
Amount of lead U/g) in floor dust wipe at 24 months.
Amount of lead U/g) in window sill dust wipe at 24 months.
Average concentration of lead (//g/g) in soil.
Concentration of lead in blood Qug/dLI at 24 months.
r^^^===^s^^=^^=^==^^^^
D-3
-------
Table D-5. Description of Variables for Pathway Model from Dust Lead Contribution to
Lead in Children- by Sayre[18].
Variable
Pb-Containing Paint
Airborne Pb
Pb-Contaminated Outside Dirt
Interior Dust
Hands
Play Objects
Pb B
* , n *
Description ' , ., '
Loose or peeling
paint from any
area where loose
paint could be
seen, regardless
of its accessibility
to the child.
Pica: Parent indicated the child ate paint chips.
Paint containing lead becomes dust which
contributes to the interior house dust.
Explanation not provided.
Soil taken from
area reported to
be used most
frequently by the
child.
Parent indicated child ate dirt.
Soil containing lead which contributes to the
interior house dust.
Towel wipe taken from window sill and floor in area child commonly
plays.
Towel wipe, rubbing both front and back surfaces of both hands of
child.
Explanation not provided.
Blood lead concentration collected at health center within one year prior
to environmental sampling.
D-4
-------
Table D-6. Description of Variables for Pathway Model from Pathways of Lead Exposure
in Urban Children' by Lanphear [19].
Variable -- f
Dust Lead
Hand Lead
Soil Lead
Paint Lead
Blood Lead
Black Race
Income Level
Playing Outside
Ingestion of Soil
Description I . ,
Dust lead level, as measured by wipe sampling, was the average of all
transformed measures across the four surfaces (carpeted floors,
uncarpeted floors, interior window sills, and window troughs).
Explanation not provided.
Composite sample of three core samples taken on each side of the
house around the perimeter of the foundation.
Average of all interior paint XRF values.
Venous blood sample collected from children between the ages 1 2 and
30 months.
An indicator variable coded 1 for African American and 0 for Caucasian.
Gross Income levels were categorized as income below $15,000, and
above $ 1 5,000.
The amount of time spent playing outdoors (# hours).
How often a child puts dirt or sand in his/her mouth.
D-5
-------
XRF Hazard
PbD
-ป. PbH
+~ PbB
PbSS
Note: Solid lines represent significant pathways and dotted lines represent non-significant pathways.
Figure D-1. Statistically Significant Pathways from Exterior Surface Dust Lead,
Interior House Dust Lead and Childhood Lead Exposure in an Urban
Environment, by Bornschein, et al. [2]
Age x PbH
Note: Solid lines represent significant pathways. Arrows represent presumed direction. No arrow means no
direction was presumed. No line means the pathway was not significant.
Figure D-2. Statistically Significant Pathways from Soil-Lead - Blood Lead Relationship in
a Former Lead Mining Town, by Bornschein, et al. [17].
D-6
-------
SES
PbD
PbH
H.O.M.E..
House
Note: Solid lines represent significant pathways and dotted lines represent non-significant pathways.
Figure D-3. Statistically Significant Pathways from The Influence of Social and
Environmental Factors on Dust Lead, Hand Lead, and Blood Lead Levels in
Young Children, by Bornschein, et al. [1].
Pb *-
,* Air X
Refinishing
Pb
Soil
Pb Dust
* Window
Sill
Pb Dust
Floor
A
Note Diagram was selected as most representative of the analyses done in this article. Solid lines
represent significant pathways and dotted lines represent non-significant pathways.
Figure D-4. Statistically Significant Pathways from Pathways of Lead Contamination
for the Brigham and Women's Hospital Longitudinal Study, by Menton,
et al. [7].
D-7
-------
Airborne Pb
Pb Contaminated
Outside Dirt+
To
To
Pb-Containing
Paint +
To
Interior
Dust +
To
Play Objects
Pica*
Hands *
Mouthing +
PbB
Ingestion *
Note: Diagram was selected as most representative of the analyses done in this article. Solid lines and plus
signs represent significant correlations and dotted lines represent hypothesized pathways.
Figure D-5. Statistically Significant Pathways from Dust Lead Contribution to Lead in
Children, by Sayre [18].
Paint Lead
Soit Lead
->Dust
Play Outside-
-> Eats Soil
Hand
Blood
Black Race
Income
Note: Solid lines represent the pathways represented and assumed to be statistically significant.
Figure D-6. Statistically Significant Pathways from Pathways of Lead Exposure in Urban
Children, by Lanphear et al. [19].
D-8
-------
APPENDIX E
Structural Equation Modeling
(The CALIS Procedure)
E-1
-------
This section briefly describes the SAS procedure, PROC CALIS (CALIS), that was used
in all the structural equation modeling (SEM) for this report.
Background
The SEM models divide the explanatory variables into exogenous and endogenous
variables. The endogenous variables in the analyses of this report were blood-lead, hand dust-
lead, floor dust-lead, interior entryway dust-lead, window sill dust-lead and window well dust-
lead, while the exogenous variables were exterior entryway dust-lead, soil-lead, mouthing habits
of children, an indicator of water-lead, an indicator of renovation and remodeling activities, air
duct dust-lead, paint-lead indicators, and the carpeting indicator variables.
The Model
In CALIS there are several options for calculating the parameter estimates in the model.
For this analysis, the LINEQS method was chosen using the Newton-Raphson optimization
method with maximum likelihood estimation. The structure of the SEM is
where P and y are coefficient matrices and TJ and ฃ are vectors of random variables. The
components of i] correspond to the endogenous variables expressed as a linear combination of the
remaining endogenous variables, of the exogenous variables of |, and of a residual component in
ฃ. The coefficient matrix P describes the relationships among the endogenous variables in i\.
The coefficient matrix y explains the relationships between the endogenous variables of TJ and the
exogenous and error components of 4-
The parameters are estimated using maximum likelihood estimation (MLE) criterion and
an iterative a non-linear optimization algorithm (Newton-Raphson) that optimizes a goodness-of-
fit criterion F. The fit criterion for the maximum likelihood estimation is
FML = Tr [SO"1] - n + In det [C] - hi det
E-2
-------
where n is the number of variables, S is the sample covariance matrix, and C denotes the
predicted moment matrix. This can also be expressed by the generally weighted least-squares
criteon:
OWLS =
where W is the weight matrix, S is the sample covariance matrix, and C is the predicted moment
matrix. For the normal theory maximum-likelihood, W is the iteratively updated predicted
moment matrix C. The values of the maximum-likelihood function F^ and the generally
weighted least-squares criterion FGWLS with W = C are asymptotically equivalent. Then the
approximate standard errors can be computed as the diagonal elements of the matrix
NM
where N is the sample size, NM = N -1 if the correlation or covariance matrix is analyzed, H is
the approximate Hessian matrix of F evaluated at the final estimates, and c = 2 for the maximum
likelihood method. If a given correlation or covariance matrix is singular, PROC CALIS
computes a generalized inverse of the information matrix either by the Moore-Penrose inverse or
a G2 inverse method, depending on the G4 specification. The Moore-Penrose inverse uses an
eigenvalue decomposition and the G2 inverse is produced by sweeping the linearly independent
rows and columns and zeroing out the dependent ones.
Goodness-of-Fit
To evaluate the models, a goodness-of-fit index (GFI) was calculated and assessed. The
GFI computed by CALIS for the maximum likelihood estimation method is given by
,l-Tr[W-l(S-C)2]
Tr[W-lS]2
If the GFI is between 0 and 1 then the fit is considered to be good. If the GFI is negative or
much larger than 1 then the data is considered to not fit the specified model.
E-3
-------
Prediction Interval
The prediction intervals presented in Tables 6-4, 6-5, 6-6, 6-7, etc. are confidence
intervals for the estimates of percent change in blood-lead levels that resulted when the
geometric mean of one of the model input variables was decreased by 50%. Due to the
complexity of the pathway models, the confidence bounds were based on the direct effects only.
A two-sided 100(l-a)% confidence interval for the estimated percent change in blood-lead level
was calculated as
Pi,
where
K$ = Estimated percent change in blood-lead level.
P = Estimated value of the regression coefficient for the model input variable.
S(P) = Standard error of P .
d = Residual Mean Square (MSB)
K = log(D) with D being the ratio of the model input variable value at a 50%
reduction of the geometric mean to the geometric mean.
The upper and lower bounds were transformed back to the original scale to facilitate
physical interpretation.
E-4
-------
REPORT DOCUMENTATION PAGE
Form Approved
i No 0704-0188
Public reporting burden for this collection of information is estimated to averge 1 hour per response, including the time for reviewino
'
Waston DC 20BQ3 222Q2^Q2' and to the Off'ce ฐf Management and Budget, Paperwork Reduction Project (0704-0188),
. AGENCY USE ONLY (Leave blank) | 2. REPORT DATE I 3. REPORT TYPE AND DATES COVERED
December 2000 Final Report
4. TITLE AND SUBTITLE 5. FUND.NG NUMBERS
Analysis of Pathways of Residential Lead Exposure in Children
6. AUTHOR(s)
P.A. Hartford and J. Nagaraja
C: 68-D5-0008
7. PERFORMING ORGANIZATION NAME(s) AND ADDRESS(ES)
Battelle Memorial Institute
505 King Avenue
Columbus, Ohio 43201
8. PERFORMING ORGANIZATION
REPORT NUMBER
Not Applicable
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES)
U.S. Environmental Protection Agency
Office of Pollution Prevention and Toxics
401 M Street SW (7401)
Washington, D.C. 20460
10. SPONSORING/MONITORING AGENCY
REPORT NUMBER
EPA 747-R-98-007
11. SUPPLEMENTARY NOTES
Other Battelle staff involved in the production of this report included Y.L. Chou, N. McMillan, and R. Menton.
12.a DISTRIBUTION/AVAILABILITY STATEMENT
12b. DISTRIBUTION CODE
13. ABSTRACT (Maximum 200 words)
This report presents the results of the pathways analysis of data from three major studies: the Rochester Study, the
Repair and Maintenance Study, and the Comprehensive Abatement Performance Study. Data from the studies were analyzed
via structural equations models to determine the significant pathways of lead in residential settings. Both environmental-lead
pathways and blood-lead pathways were analyzed. Results of the environmental-lead pathways modeling were quite similar
across all three studies. Blood-lead pathways results were less similar across studies. For the blood pathways models, results
for models with lead loadings were very different from results for models with lead concentration. Additional analyses were
conducted to examine pathways from paint on windows and doors, renovation and remodeling, carpeted floors, air ducts, and
other variables for which data was available from at least one but not all of the three studies.
14. SUBJECT TERMS
Structural Equations Modeling, Lead in Dust and Soil, Blood Lead Levels in Children,
Environmental-Lead Pathways Analysis, Blood-Lead Pathways Analysis, Rochester Study, Repair
and Maintenance Study, Comprehensive Abatement Performance Study
15. NUMBER OF PAGES
144
16. PRICE CODE
7. SECURITY CLASSIFICATION
OF REPORT
Unclassified
18. SECURITY CLASSIFICATION
OF THIS PAGE
Unclassified
19. SECURITY
CLASSIFICATION OF
ABSTRACT
Unclassified
20. LIMITATION OF
ABSTRACT
NSN 7540-01-280-5500
Standard Form 298 (Rev 2-89)
Prescribed by ANSI Std. Z39-18
298-102
-------