Analysis of Lead Dust Clearance Testing Final Report


v EPA
  United States
  Environmental Protection
  Agency
Analysis of Lead Dust
Clearance Testing

Beam
Bedroom 1
BadroomZ
Bedrooms
Kitchen
Hallway
Dining Room
Living Room
Family Room
Housing
Elfiffii
25
SO
175
35
175
75
150
350
UnitA-1
Window Sill
50
200
700
400

200
300
400

Window Well
350
600
600
1000

550
400
600

-------

-------
                                                                EPA 747-R-01-005
                                                                  December 2001
V)
^
V
Analysis of Lead Dust
  Clearance Testing

                                     FINAL REPORT
                         Program Assessment and Outreach Branch
                           National Program Chemicals Division
                          Office of Pollution Prevention and Toxics
                           U.S. Environmental Protection Agency
                                Washington, D.C. 20460
                                               US EPA Headquarters Library
                                                     Mail code 3201
                                               1200 Pennsylvania Avenue NW
                                                  Washington DC 20460

-------
                                   DISCLAIMER
                                                                               /
      The material in this document has been subject to Agency technical and policy review.
Mention of trade names, products, or services does not convey, and should not be interpreted as
conveying, official EPA approval, endorsement, or recommendation.           r    '

                             FURTHER INFORMATION

      Information about other technical reports on lead can be found through the internet at the
address: http://www.epa.gov/lead.                                          '   .

      This report is copied on recycled paper.                       .. u

-------
                         CONTRIBUTING ORGANIZATIONS

      The methodology described in this report is part of a task funded by the U.S.
Environmental Protection Agency. The task was managed by the U.S. Environmental Protection
Agency. The task was conducted by Battelle Memorial Institute under contract with the U.S.
Environmental Protection Agency.
                             Battelle Memorial Institute

      Battelle Memorial Institute (Battelle) was responsible for acquisition of the data, data
management, development of statistical methods, and the writing of this report. The key Battelle
staff were Dr. Bradley Skarpness, Mr. Ying-Liang Chou, and Mr. Warren Strauss. Contributing
staff were Ms. Jennifer Holdcraft, Ms. Pamela Hartford and Mr. Matt Palmgren.
                       U.S. Environmental Protection Agency

      The U.S. Environmental Protection Agency (EPA) funded the task, managed the task,
reviewed task documents, and managed the peer review of this report. The EPA Work
Assignment Manager was John Schwemberger. The Project Officer was Sineta Wooten.
                                        tit

-------
                       ACKNOWLEDGMENTS

The following organizations contributed the data and information used in this report:

                    Atlantic City Housing Authority
                Cleveland Lead Hazard Abatement Center
                          Dewberry & Davis
                            Speedwell, Inc.
                       Dover Housing Authority
                Maryland Department of the Environment
            U.S. Department of Housing and Urban Development
        University of Cincinnati Department of Environmental Health
                                  IV

-------
                              TABLE OF CONTENTS

                                                                        Page

EXECUTIVE SUMMARY	 x

1.0    INTRODUCTION  	 1
      1.1   BACKGROUND	 1
      1.2   OBJECTIVES  	 2
      1.3   DEFINITIONS  	 4
      1.4   STRUCTURE OF THE REPORT 	 6
      1.5   PEER REVIEW	 7

2.0    FINDINGS AND CONCLUSIONS	 10
      2.1   OBJECTIVE 1:  CHARACTERIZATION OF THE NUMBER OF INDIVIDUAL
           SAMPLES, WORK AREAS, AND HOUSING UNITS THAT PASS OR FAIL
           CLEARANCE TESTING STANDARDS	 10
      2.2   OBJECTIVE 2:  CHARACTERIZATION OF THE DISTRIBUTION OF THE
           DUST-LEAD LOADINGS, GEOMETRIC MEAN DUST-LEAD LOADINGS,
           VARIABILITY WITHIN A HOUSING UNIT, AND VARIABILITY BETWEEN
           HOUSING  UNITS	 11
      2.3   OBJECTIVE 3:  CHARACTERIZATION OF THE CORRELATION BETWEEN
           COMPONENTS SAMPLED IN THE SAME WORK AREA	 12
      2.4   OBJECTIVE 4:  DEMONSTRATION OF THE IMPACT OF COMPOSITE
           SAMPLING ON PASS/FAIL RATES OF HOUSES	 12

3.0    OVERALL QUALITY ASSURANCE  	 14

4.0    DESIGN, DATA COLLECTION, AND CHEMICAL ANALYSIS	 16
      4.1   FHA SINGLE-FAMILY HOUSING PHASE OF THE HUD ABATEMENT
           DEMONSTRATION PROJECT	 16
      4.2   PUBLIC HOUSING ADMINISTRATION (PHA) PHASE OF THE HUD ABATEMENT
           DEMONSTRATION PROJECT  	 17
           4.2.1  The Albany Demonstration 	 18
           4.2.2  The Omaha Demonstration 	 19
           4.2.3  The Cambridge  Housing Authority Demonstration	 21
      4.3   MARYLAND DEPARTMENT OF THE ENVIRONMENT 	 22
      4.4   EVALUATION OF THE HUD LEAD-BASED PAINT HAZARD CONTROL GRANT
           PROGRAM	 23
      4.5   ATLANTIC CITY HOUSING AUTHORITY	 27
      4.6   CLEVELAND LEAD HAZARD ABATEMENT CENTER	 28
      4.7   DOVER HOUSING AUTHORITY, NEW HAMPSHIRE	 28

5.0    DATA METHODOLOGY AND ANALYSIS	 29
      5.1   OBJECTIVE 1:  CHARACTERIZATION OF THE NUMBER OF INDIVIDUAL
           SAMPLES, WORK AREAS, AND HOUSING UNITS THAT PASS OR FAIL
           CLEARANCE TESTING STANDARDS	 29
      5.2   OBJECTIVE 2:  CHARACTERIZATION OF THE DISTRIBUTION OF THE
           DUST-LEAD LOADINGS, GEOMETRIC MEAN DUST-LEAD LOADINGS,
           VARIABILITY WITHIN A HOUSING UNIT, AND VARIABILITY BETWEEN
           HOUSING  UNITS	 33
                                                    U S EPA Headquarters Library
                                                          Mail code 3201
                                                    1200 Pennsylvania Avenue NW
                                                       Washington DC 20460

-------
                               TABLE OF CONTENTS
                                    (Continued)
                                                                            Page
       5.3    OBJECTIVE 3: CHARACTERIZATION OF THE CORRELATION BETWEEN
             COMPONENTS SAMPLED IN THE SAME WORK AREA	  37
       5.4    OBJECTIVE 4: DEMONSTRATION OF THE IMPACT OF COMPOSITE
             SAMPLING ON PASS/FAIL RATES OF HOUSES	  39
             5.4.1  Construction of the Simulated Composite Samples	  40
             5.4.2  Composite Sample Clearance Criteria 	  42
             5.4.3  Methods for Summarizing the Simulated Composite Sample Results  ...  43

6.0    FINDINGS AND RESULTS	  48
       6.1    OBJECTIVE 1: CHARACTERIZATION OF THE NUMBER OF INDIVIDUAL
             SAMPLES, WORK AREAS, AND HOUSING UNITS THAT PASS OR FAIL
             CLEARANCE TESTING STANDARDS	  48
       6.2    OBJECTIVE 2: CHARACTERIZATION OF THE DISTRIBUTION OF THE
             DUST-LEAD LOADINGS, GEOMETRIC MEAN DUST-LEAD LOADINGS,
             VARIABILITY BETWEEN SAMPLES COLLECTED WITHIN HOUSING UNITS,
             AND VARIABILITY ACROSS HOUSING UNITS	  74
       6.3    OBJECTIVE 3: CHARACTERIZATION OF THE CORRELATION BETWEEN
             COMPONENTS SAMPLED FROM WITHIN THE SAME WORK AREA	  100
       6.4    OBJECTIVE 4: DEMONSTRATION OF THE IMPACT OF COMPOSITE
             SAMPLING ON PASS/FAIL RATES OF HOMES  	  108

7.0    REFERENCES  	  119

                                LIST OF APPENDICES

APPENDIX A
       Maryland Department of the Environment	A-1
APPENDIX B
       FHA Single-Family Housing Phase of the HUD Abatement Demonstration Project	B-1
APPENDIX C
       Public Housing Administration (PHA) Phase Of the HUD Abatement
       Demonstration Project 	C-1
APPENDIX D
       Evaluation of the HUD Lead-Based Paint Hazard Control Grant Program   	D-1
APPENDIX D1
       Grantees with Higher Floor Dust-Lead Clearance Standard in the HUD Grantee
       Program	D1-1
APPENDIX D2
       Grantees with Lower Floor  Dust-lead Clearance Standard in the HUD Grantee
       Program	D2-1
APPENDIX E
       Atlantic City Housing Authority	E-1
APPENDIX F
       Cleveland Lead Hazard Abatement Center	F-1
APPENDIX G
       Dover Housing Authority, New Hampshire	G-1
                                        VI

-------
TABLE OF CONTENTS
(Continued)
APPENDIX H
Additional Analysis Results on the Percentage of Housing Units That Passed
Clearance on the First Site Visit, by the Number of Individual Samples
Collected Within a Housing Unit
APPENDIX I
Additional Analysis Results
APPENDIX J
Comparison of Clearance Guidance
Page
H-1

1-1

J-1
LIST OF TABLES
Table ES-1 Geometric Mean, and 50th and 90th Percentiles for the First Site Visit, and
The 50th and 90th Percentiles for the Passed Clearance Visit, by Data Source
and Component Type xvi
Table 1-1. Summary of Lead Clearance Data Sources 3
Table 4-1. Interior Intervention Strategy Code Definitions 26
Table 5-1. Location of Answers to Questions on Pass/Fail Rates 31
Table 5-2. The Probability of Passing Each of the Three Composite Sample Clearance
Criteria Based on Multiple Simulated Composite Samples Constructed from
Hypothetical Sample Data 44
Table 5-3. Table Summarizing Pass/Fail Conclusions Made from Individual Sample Clearance
Results Compared to the Simulated Composite Sample Results 45
Table 6-1 a. Clearance Testing Results by Individual Sample, Room, and Housing Unit
for the First Site Visit 52
Table 6-1 b. Clearance Testing Results by Individual Sample, Room, and Housing Unit
for the Second Site Visit 54
Table 6-1 c. Clearance Testing Results by Individual Sample, Room, and Housing Unit
for the Third Site Visit 56
Table 6-1 d. Clearance Testing Results fay Individual Sample, Room, and Housing Unit
for the Fourth Site Visit 58
Table 6-1e. Clearance Testing Results by Individual Sample, Room, and Housing Unit
for the Fifth Site Visit 60
Table 6-1f. Clearance Testing Results by Individual Sample, Room, and Housing Unit
for All Site Visits 61
Table 6-1 g. Clearance Testing Results for the Floor Samples at Various Clearance
Standards for the First Site Visit 63
Table 6-1 h. Clearance Testing Results for the Window Sill Samples at Various Clearance
Standards for the First Site Visit 64
Table 6-1 i. Clearance Testing Results for the Window Trough Samples at Various
Clearance Standards for the First Site Visit 65
Table 6-2. Parameter Estimates and Associated Standard Errors from a Logistic
Regression Model of the Probability of a Residential Unit Failing Clearance
Testing Based on the Number of Samples Collected 66

-------
TABLE OF CONTENTS
(Continued)
Page
Table 6-3. Categorization of the Number of Parameter Estimates and Associated Standard
Errors from a Logistic Regression Model of the Probability of a Residential Unit
Failing Clearance Testing Based on the Number of Samples Collected 68
Table 6-4. Parameter Estimates and Associated Standard Errors from a Logistic
Regression Model of the Proportion of Samples within a Residential Unit
Expected to Fail Clearance Testing Based on the Number of Samples Collected. 70
Table 6-5. ' Categorization of the Number of Parameter Estimates and Associated Standard
Errors from a Logistic Regression Model of the Proportion of Samples within a
Residential Unit Expected to Fail Clearance Testing Based on the Number of
Samples Collected 72
Table 6-6a. Geometric Mean and Variance Component Estimates by Component Type
and Data Source for the First Site Visit 79
Table 6-6b. Geometric Mean and Variance Component Estimates by Component Type
and Data Source for the Second Site Visit 80
Table 6-6c. Geometric Mean and Variance Component Estimates by Component Type
and Data Source for the Third Site Visit 81
Table 6-6d. Geometric Mean and Variance Component Estimates by Component Type
and Data Source for the Fourth Site Visit 82
Table 6-7a. Percentage of Clearance Samples Below 50 //g/ft2, 75 jug/ft2, 100 //g/ft2,
and HUD Interim Guidance Clearance Standards by Data Source and Substrate
for the First Site Visit 83
Table 6-7b. Percentage of Clearance Samples Below 50//g/ft2, 75//g/ft2, 100//g/ft2,
and HUD Interim Guidance Clearance Standards by Data Source and Substrate
for the Second Site Visit 86
Table 6-7c. Percentage of Clearance Samples Below 50 //g/ft2, 75 //g/ft2, 100 //g/ft2,
and HUD Interim Guidance Clearance Standards by Data Source and Substrate
for the Third Site Visit 88
Table 6-7d. Percentage of Clearance Samples Below 50 //g/ft2, 75 //g/ft2, 100//g/ft2,
and HUD Interim Guidance Clearance Standards by Data Source and Substrate
for the Fourth Site Visit 90
Table 6-8. Percentage of Clearance Samples Below 50 //g/ft2, 75 //g/ft2, 100 //g/ft2,
and HUD Interim Guidance Clearance Standards by Data Source and Substrate
for the Passed Clearance Visits 91
Table 6-9a. Percentiles {//g/ft2) for Floor Dust-Loading by Data Source and Substrate for
the First Site Visit 94
Table 6-9b. Percentiles (//g/ft2} for Window Sill Dust-Loading by Data Source and Substrate
for the First Site Visit 95
Table 6-9c. Percentiles (//g/ft2) for Window Trough Dust-Loading by Data Source and
Substrate for the First Site Visit 96
Table 6-1 Oa. Percentiles (//g/ft2) for Floor Dust-Loading by Data Source and Substrate
for the Passed Clearance Site Visit 97
Table 6-1 Ob. Percentiles (//g/ft2) for Window Sill Dust-Loading by Data Source and
Substrate for the Passed Clearance Site Visit 98
Table 6-1 Oc. Percentiles (//g/ft2) for Window Trough Dust-Loading by Data Source and
Substrate for the Passed Clearance Site Visit 99
Table 6-11. Observed Within-Room Correlation Coefficients Between Dust-Lead Loading
Measurements Collected From Floors, Window Sills, and Window Troughs. . . 103
VIII

-------
TABLE OF CONTENTS
(Continued)
Page
Table 6-12a. Conditional Probabilities for Floors and Window Sills Clearance Testing
Estimated Using 2x2 Contingency Tables and Normal Probability Theory
Based on the First Site Visit Data 104
Table 6-12b. Conditional Probabilities for Floors and Window Troughs Clearance Testing
Estimated Using 2x2 Contingency Tables and Normal Probability Theory
Based on the First Site Visit Data 105
Table 6-12c. Conditional Probabilities for Window Sills and Window Troughs Clearance
Testing Estimated Using 2x2 Contingency Tables and Normal Probability
Theory Based on the First Site Visit Data 106
Table 6-12d. Conditional Probabilities for Floors, Window Sills, and Window Troughs
Clearance Testing Estimated Using 2x2 Contingency Tables and Normal
Probability Theory Based on the First Site Visit Data 107
Table 6-13. Number of Housing Units for Each Data Source that Contained (N) Individual
Clearance Samples of Each Component Type Based on the First Site Visit. ... 111
Table 6-14. Number of Housing Units that Passed or Failed Clearance on the First Site
Visit, Based on Individual Sample Clearance Results versus Simulated
Composite Clearance Results 112
Table 6-15. Performance Characteristics of Composite Clearance Criteria for Each Data
Source 114
Table 6-16a. Results from Fitting a Logistic Regression Model Which Investigates the
Relationship Between the Probability of Clearance and the Maximum
Individual Floor Sample Lead Loading, Under Three Different Composite
Sampling Decision Rules 116
Table 6-16b. Results from Fitting a Logistic Regression Model Which Investigates the
Relationship Between the Probability of Clearance and the Maximum Individual
Window Sill Sample Lead Loading, Under Three Different Composite Sampling
Decision Rules 117
Table 6-16c. Results from Fitting a Logistic Regression Model Which Investigates the
Relationship Between the Probability of Clearance and the Maximum Individual
Window Trough Sample Lead Loading, Under Three Different Composite
Sampling Decision Rules 118

LIST OF FIGURES

Figure ES-1. Geometric Mean, and the 50th and 901h Percentiles for the First Site Visit
by Data Source and Component Type xvii
Figure ES-2. The 50th and 90th Percentiles for the Passed Clearance Visit, by Data
Source and Component Type xviii
Figure 5-1. Histogram of Dust Lead Loadings from Floor Samples 34
Figure 5-2. Histogram of the Natural Logarithm of Dust Lead Loadings from Floor
Samples 34
Figure 5-3. Histogram of the Natural Logarithm of Dust Lead Loadings from
Window Sill Samples 35
Figure 5-4. Histogram of the Natural Logarithm of Dust Lead Loadings from
Window Trough Samples 35
IX

-------
                                EXECUTIVE SUMMARY
       The U.S. Environmental Protection Agency (EPA) gathered data about local clearance
dust results to characterize and assess the likelihood of achieving the HUD Interim Guidelines
Clearance Standards for a wide variety of housing units, components, and surface types. The
EPA obtained dust lead loading results from lead clearance dust wipe samples from three
federally funded projects: the Federal Housing Authority (FHA) and Public Housing Authority
(PHA) Lead-Based Paint Abatement Demonstration projects sponsored by the Department of
Housing and Urban Development (HUD) and HUD's Lead-Based Paint Hazard Control Grant
Program (HUD Grantee program), and from four state or local government authorities: the
Maryland Department of the Environment (MDE); Atlantic City Housing Authority, New Jersey;
Dover Housing Authority, New Hampshire; and the Cleveland Lead Hazard Abatement Center,
Ohio.
       The results presented in this report represent lead clearance testing activities that occurred
between 1989 and 1999 under the 1990 HUD Interim Guidelines [2], which set lead clearance
dust wipe standards at 200 ng/ft2 for bare floors, 500 ug/ft2 for window sills, and 800 ug/ft2 for
window troughs.  (These standards contrast with the 2001 EPA final dust-lead clearance
standards of 40, 250, and 400 ug/ft2, respectively [1].) The collected data include 39,301 lead
clearance dust wipe sample measurements taken on floors, window sills, and window troughs
from 4,518 single or multi-unit family housing units that had received some type of lead-based
paint abatement or other intervention. The samples were taken over a number of site visits since
a housing unit would be retested if at least one individual dust wipe sample had a lead-loading
result greater than the HUD Interim Guidelines Clearance Standards. Although the percentage of
individual dust wipe clearance sample results that passed clearance was above 90% for all three
surface areas, on average only 67% of the housing units passed clearance on the first site visit.
This indicates that the individual samples that fail clearance are not necessarily concentrated in
just a few homes. Eventually 4,095 housing units out of 4,518 (87%) were known to have passed
clearance.  Information on most of the remaining 13% of the housing units was lost during
follow-up ("lost to follow-up"), although some units might have simply failed clearance after
multiple attempts. It is likely that many of the units lost to follow-up eventually cleared,
although this assumption was not included in the data analysis.  The losses to follow-up may be

-------
attributed to families moving, litigation, or poor record keeping, or the fact that housing units
were still being retested at the time of data collection.
These data represent extensive dust-lead clearance testing, with results collected from
across the United States. Similar trends were observed in individual data sources. For example,
the distribution of clearance lead-loading results within each source of data was highly skewed to
high values, so a (natural) log transformation was applied to these results, which resulted in a log
normal distribution and was used in the analyses in this report, hi general, floor dust wipe
samples had a smaller geometric mean lead loading than window sills, which in turn had a
smaller geometric mean than window troughs. The geometric mean lead loading of the samples
collected within each component type generally increased from the first site visit to the third site
visit, since generally only homes that failed clearance testing were resampled in subsequent site
visits. For surfaces that failed clearance, it is presumed that they were re-cleaned prior to re-
testing in subsequent site visits. The variability of lead-loading results within a house was
slightly smaller than the variability of lead-loading results between houses for most data sources.
The pairwise correlations between floor, window sill, and window trough lead-loading results of
samples collected from within the same room during the first site visit were positive and
statistically significant. The correlation between window trough and window sill samples from
within the same room was highest among the three component types. Floor and window sill dust
wipe results had a slightly stronger correlation than floors and window troughs.
In general, accounting for longitudinal trends in a dataset is a difficult task even when the
data are collected under controlled conditions. The precise circumstances under which most of
these observational data were collected within each data source were very diverse and in many
cases did not span many years. Subsequently, only the Maryland Department of the
Environment data source was analyzed for longitudinal trends (Appendix A). These data
spanned five years, 1991-1995, and contained the largest number of housing units, 706, which
could be analyzed for trends over time. This analysis indicated no significant differences
between the geometric means of components across the years using 95% confidence intervals.
The HUD Guidelines [3] and 40 CFR Part 745 [1,4] indicate that composite samples of
up to four individual dust wipes collected from different locations of the same component type
within a housing unit can be analyzed together to yield a single dust lead-loading result for
clearance testing. To evaluate how composite sampling would affect clearance, composite
XI

-------
 sampling was simulated by grouping individual clearance dust wipe results.  Pass/fail rates of
 housing units associated with composite samples are different from those associated with
 individual samples.  Simulations indicated that composite sampling decreases the probability of a
 dwelling failing, if at least one of the individual sample results was above the HUD Interim
 Guidelines Clearance Standard.  The simulations were based on simulated composite samples
 constructed from measured individual lead-loading clearance testing results. Since lowering the
 HUD Interim Guidelines Clearance Standard associated with composite dust-wipe samples for
 each component type increases the sensitivity (the probability of a dwelling unit failing clearance
 based on simulated composite samples given that the unit would also have failed clearance based
 on the individual samples), two other clearance criteria with lower levels (Standard/n Rule and 2
 x Standard/n Rule)1 were defined and compared. The Standard/n criterion resulted in fewer false
 clearance passes than the HUD Interim Guidelines Clearance Standard in the simulation study, at
 the expense of increasing false clearance failures. The 2x Standard/n criterion struck a balance
 between the other two criteria.
       In addition to the above results and conclusions, the following are some of the results and
 conclusions drawn from the data in response to specific questions. Table ES-1 is the tabular
 display and Figures ES-1 and ES-2 are the graphical illustrations for the first three questions.

What Lead Levels For Floors. Window Sills,  and Window Troughs Are Typical in Dust
 Clearance Testing?
       As seen in Table ES-1, the range of geometric means across the eight data sources2 used
in this report were 8.8 to 57.6 ug/ft2 for bare floors, 11.3 to 461.6 ug/ft2 for sills, and 15.8 to
393.3 for troughs during the first clearance site visit.  The high window sill dust-lead loading of
       'A Standard/n Rule allows a composite sample to pass clearance only if the lead loading result is less than
the corresponding standard for individual samples divided by the number of subsamples in the composite sample. A
2>
-------
461.6 ug/ft2 may be due to the interior renovation strategy used in Dover, New Hampshire
(excluding Dover data, the range of window sill geometric mean dust-lead loading was 11.3 to
60.3 ug/ft2). The 90th percentiles for dust-lead loading clearance sample results from the first
site visit were between 68 and 418 ug/ft2,171 and 1624 ug/ft2, and 137 and 5810 ug/ft2, for bare
floors, window sills, and window troughs, respectively (excluding Dover data, the range for the
90th percentile window sill dust-lead loading was 171 to 714 ug/ft2). The 90th percentiles for
dust wipe clearance sample results from houses that passed clearance for standards of 200 ug/ft2,
500 ug/ft2, and 800 ug/ft2 were between 57 and 141 ug/ft2,141 and 443  ug/ft2, and 111 and 598
ug/ft2, for bare floors, window sills, and window troughs, respectively.

Are Window Sills Being Cleaned to the Same Level as Bare Floors?
       Table ES-1 shows that the geometric mean for the initial clearance test for floors was
between 8.8 and 57.6 ug/ft2; for window sills it was between 11.3 and 461.6 ug/ft2 (or 11.3  and
60.3 ug/ft2 excluding Dover's window sill data).  The 50th percentile for the first clearance  test
was between 5 and 48 ug/ft2 for floors, and between 17 and 52 ug/ft2 for window sills (excluding
Dover window sill data). The 90th percentile was between 68 and 418 ug/ft2 for floors and
between 171 and 714 ug/ft2 for sills (excluding Dover window sill data). For dust wipe
clearance sample results from houses that passed clearance for the standards of 200 ug/ft2 and
500 ug/ft2 for floors and window sills, respectively: the 50th  percentile was between 5 and 43
ug/ft2 for floors, and between 17 and 47 ug/ft2 for window sills (excluding Dover window sill
data); the 90th percentile was between 57 and 141 ug/ft2 for floors and between 141 and 263
ug/ft2 for sills (excluding Dover window sill data).  These results indicate that window sills
tended to have somewhat higher lead levels than floors after intervention and cleaning.
Are Window Troughs Being Cleaned to the Same Level as Window Sills?
       Table ES-1 shows that the geometric mean for troughs on the initial clearance test was
between 15.8 and 393.3 ug/ft2; for sills the corresponding figures were between 11.3 and 60.3
^ig/ft2 (excluding Dover's window sill data). The 50th percentile for troughs at the initial
clearance test was between 7 and 375 ug/ft2; for sills it was between 17 and 52 ug/ft2; the 90th
percentiles were between 137 and 5810 ug/ft2 for troughs and between 171 and 714 ug/ft2 for
sills (all excluding Dover window sill data).  For dust wipe clearance sample results from houses
                                           xiii
-------
that passed clearance for the standards of 500 ng/ft2 and 800 ug/ft2 for window sills and window
troughs, respectively: the 50th percentile for troughs was between 7 and 169 ^.g/ft2, for sills it
was between 17 and 47 ug/ft2 (excluding Dover window sill data); the 90th percentiles were
between 111 and 598 ng/ft2 for troughs and between 141 and 263 ug/ft2 for sills (all excluding
Dover window sill data). These results indicate that window troughs tended to have higher lead
levels than window sills after intervention and cleaning.

Do Window Sills  and Window Troughs. Bare Floors and Window Troughs, or Bare Floors
and Window Sills Pass Clearance Together or Fail Clearance Together?
       If a dust wipe lead loading collected from a component (bare floors,  window sills, or
window troughs) passed its clearance standard, there was a good chance the other components in
the same area passed. These probabilities were generally greater than 90%.  For example, the
probability that a window sill sample passed clearance given the window trough sample passed
clearance was greater than 92% from almost all data sources presented in this report.
       If one component failed its standard, it was not necessarily the case that another
component in the same  area (i.e., room equivalent) failed its standard. The probability that one
component failed given another had failed in the same area ranged from 0% to 79%. Generally,
these probabilities were under 30%. The exception occurred for the probability of a window sill
failing clearance given the window trough failed clearance. This estimated probability was
greater than 72% for some data sources.

Does Increasing the Number of Samples Taken Per Unit Increase the Probability of Failure
for the Whole Unit?
       Based only on the first site visit, as the number of samples taken within a given housing
unit increased, the chance of this given housing unit failing clearance increased. This could be
due to more samples being taken in the housing units that needed the most work or due to the use
of maximum of dust results as the criteria for passing or failing clearance.  Based on the logistic
regression model results, the increase was statistically significant. The estimated average
probability of clearance failure (across eight data sources) was 20% for two samples collected
from floors.  It increased to 22% when four samples were collected from floors and to 24% when
six samples were collected from floors. For samples collected from window sills, the estimated
                                          XIV
-------
average probability of failing clearance was 23% when two sill samples were collected; it
increased to 29% when four sill samples were collected and to 38% when six sill samples were
collected.  For samples collected from window troughs, the estimated average probability of
failing clearance was 14%, 19%, and 29% when two, four, and six window trough samples were
collected, respectively.
                                         XV
-------
 CO
JJ

'•M

 0

 S
 0)
Q.


b
0)

TJ

 ID


b
in
 a

 ~  0)

'
E E
il
o> a>
« u
•^ 3
C O
   s
.«

o >
   8
   C
   10
   -g
§ S
0) 
fl
§ «

73
0.
O O
11
SO
£
in
rst Site. •
E-
? -•

II
S
V


1
°I
11:
JBS-

) First Site.
'£•—
jf
S." • :
Bometric
O
' If


o _
? 55

' I
.||
-•**".
'15 '
i*


§
c
• if


$
o _
. .-'
i
c.
, 5 £'•
-11
3 *~ '

'.il
5 • .
:' I:
u.
Window
Trough •
'I
c 55
, i
• 1
u.


S §
Sl
1
§


^
S

.
o
u>


CM


(O
CM
^
CO



If)

09

CM
0>


00

a>
CM

5

^


1
5
4
03
O)
If)


n
(D

5
CB
ID


*


§
O
oo
it)



2

00
5

in
to


CM
If)

a>
CO
ri
8

o




10
CO
to


CO


in
CN
e>
09



h»

to

CO


00
CO

to
in
09
ID
en
CM

CO
CM


S*
**l
1(3 ^
S


O
00

§
o


«


o
§



§
CO

8

o


CO

0
en

co

in
00


.0
1 ^
5 o
1


o


CO
CO
CO


CO


o
CM
2



§

i

CO
CO


CO
CO

r-
CM
?
CD
r*.
CO

CO
CM

15
s
§
s
u
£


CO
?

s
r>


ID
CO


10
m



CM
to

oo

CO
in
o
§

in
to



1
O
   o
   o
   O
   H—


   1

O ~C
                                                               u
                                                                CO U
c *-
2 ro
CO ^.
® o
o 5
M i-
CO O
                                                               o o
•o -o
(D O)
« CO


*-• 
-------


I
1


oOO-
*OO-
300-
HOO-
LOO •
O-
Geometric Mean for the First Site

T

T

S » T
8 ?
Visit
S




1 P
Maryland HUT* HXJD Cranta* Granta* Atlantic Cl«v«land Doirar
rMA PHA HlflU low City


j Component Tyv* F — Floor S — Window SU1
T — Window Trougn |


S3-
i
S
1


OOO-
•too-
300-
SDO -
1OO-
D -

5Oth Percentile for the First Site

T

T T
f * $ F
rylaxxd HUD KXJD Crant*v Gz-ont** t
FKA PHA Htsti LAW
Data Sourev
\ CORiponont Tyyo T — Floor S — Window 9U1
Visit
S


i *
UCUyl<= Cl«v*l»n«l Dov«r

T -» WLn,
-------
    400
    300
    zoo-
    100-
      0-
      Maryland
                 50th Percentile for the Passed Clearance Visit
HUD
riu
                             i
                                S
                                F
                                S
                                ff
HTO
FHA
Grantee
 High
Grantee
  Law
Atlantic
  Citj
                                        Data Source
           Component Type  F — Floor
                                          S -
                                                   Sill
                                         T - Window Trough
                                        Cleveland    Dover
                90tb Percentile for the Passed Clearance Visit
    600
    SCO
    400
    300
 S  ZOO
    100
T


T

S

S
F «
F


T
S

T

T
? s
F s s
T
F F F F F
      Maryland
HUD
FHA
KUI>
PHA
Grantee
   gh

    Data Source
          Grantee
            Low
          AUaoUc
            Cfty
          Cleveland     Dover
           Component Tyj>=  F - Floor
                        S - Window SfU
                               T - Ttndow Trough
Figure ES-2.  The 50th and 90"1 Percent! les for the Passed Clearance Visit, by Data
              Source and Component Type
                                         XVIII
-------
1.0    INTRODUCTION

1.1    BACKGROUND
       One method of lowering the risk of childhood lead poisoning is to perform environmental
interventions which reduce or eliminate lead hazards (contaminated paint, dust, or soil) from the
environment of a child's primary residence.  Abatement and interim controls are interventions
that are used to control lead-based paint hazards in a child's environment. Some of these
approaches may potentially create a lead hazard in one or more environmental media while
reducing or eliminating the targeted hazard.  For example, the removal of deteriorated lead-based
paint will often generate lead contaminated dust.  Therefore, it is important that a thorough
survey of possible residual lead contamination in environmental media surrounding the work
area be conducted following any intervention.
       Clearance testing is a procedure that  is used to ensure that residents and unprotected
workers are not exposed to lead contaminated media following an environmental intervention.
Clearance testing is generally completed in two phases: a visual inspection, followed by a
detailed sampling and analysis of the dust lead levels surrounding each work area.  The visual
inspection is used to verify that the contracted work was completed as specified, and that there is
no visible settled dust or debris present in the work area(s). Once a given work area passes the
visual inspection, then samples of dust surrounding that work area are collected and analyzed for
lead content. The decision to allow either reoccupancy or further work by unprotected
contractors is determined by comparing the results of the sample analyses to specific clearance
criteria.
       A lead contractor's work is not complete until each work area tested by the clearance
inspector complies with the appropriate clearance standards. A particular work area may fail to
meet standards in either phase of the clearance testing process. If a work area fails visual
inspection, then the lead contractor must finish the work as specified, which usually includes a
thorough cleaning process.  If a work area passes visual inspection, but fails to meet the
appropriate clearance criteria by virtue of environmental sampling and analysis, then the
contractor must repeat the cleaning process on all areas represented by the failed samples until
the work area is clean enough to achieve the criteria.
-------
       Although most leaded dust and debris generated by intervention appear in the foim of
surface deposition in surrounding environmental media, it is not possible to determine the extent
to which lead contamination in media near the work area is directly attributable to the
intervention. For example, dust surrounding an interior paint abatement work area may have
contained high levels of lead prior to intervention. Nevertheless, the lead contractor is still
responsible for cleaning the entire work area until dust-lead levels meet appropriate clearance
criteria.
       The dust wipe clearance standards under which much of the work in this report was
performed were 200 ug/ft2 for bare floors, 500 ug/ft2 for window sills, and 800 ug/ft2 for window
troughs.  These standards originated in the State of Maryland and were adopted by HUD in 1990
for HUD's "Interim Guidelines for Hazard Identification and Abatement in Public and Indian
Housing" [2].  HUD [3] and EPA [5] later established 1994/1995 interim lead clearance
standards of 100 ug/ft2 for bare floors; 500 ug/ft2 for window sills; and 800 ug/ft2 for window
troughs.  Among 14 grantees participating in the HUD Grantee Program, 9 grantees used the
clearance standard for floors under the HUD Interim Guidelines (200 ug/ftz), 4 grantees used the
HUD and EPA 1995 floor clearance standard at 100 ug/ft2, and 1 grantee used 80 ug/ft2 as the
clearance standard for floors. In January 2001, under TSCA Title IV Section 403, the EPA
issued the final dust-lead clearance standards of 40,250 and 400 ug/ft2 for floors, interior
window sills, and window troughs, respectively [1],  The results presented in this report,
however, were not based on these new final clearance standards.
1.2    OBJECTIVES
       A thorough search for data sets that had field lead clearance dust wipe test results
revealed the following potential data sources:  the federally-funded FHA Lead-Based Paint
Abatement Demonstration, the PHA Lead-Based Paint Abatement Demonstration, and the HUD
Lead-Based Paint Hazard Control Grant Program; and data from four state or city programs:
Maryland Department of Environment; Atlantic City, New Jersey; Cleveland, Ohio; and Dover,
New Hampshire. Data were collected from all of these data sources (Table 1-1). The purpose of
this document is to investigate the clearance data from these sources.
       The clearance testing results correspond to measures of interior residential dust-lead
loading from bare floors, window sills, and window troughs, and are reported in units of
-------
micrograms of lead per square foot sampled (ug/ftz). Four objectives were pursued in the
statistical analysis of the clearance sampling dust-lead loading results. They are as follows:

                   Table 1-1.  Summary of Lead Clearance Data Sources.
Data Source
Maryland Department
of the Environment
HUD FHA
Demonstration Study
HUD Demo
Public Housing
Authorities
HUD Lead Hazard
Control Grant Program
High Group*
HUD Lead Hazard
Control Grant Program
Low Group**
Atlantic City Housing
Authority
Cleveland Lead
Hazard Abatement
Center
Dover Housing
Authority
Locations
Baltimore City and
Surrounding MD
Counties
Baltimore/DC
Birmingham, AL
Denver, CO
Indianapolis, IN
Seattle/Tacoma, WA
Albany, NY
Cambridge, MA
Omaha, NE
9 Grantees across
United States
5 Grantees across
United States
Atlantic City, NJ
Cleveland, OH
Dover, NH
Period of
Sample
Collection
1991-95
1989-90
1991-93
1994-99
1994-99
1994-95
1993-95
1992
Type of
Housing
Single Family
Mostly
Private
Single Family
Private
Multi-unit
Public
Single and
Multi-unit
Family
Single and
Multi-unit
Family
Multi-unit
Public
Single Family
Private
Multi-unit
Public
Number of
Dwelling
Units
706
149
119
2150
1038
160
38
158
Average
Number of
Samples Per
Unit
11
19
12
8
7
6
8
6
*  Grantees that used 200 u.g/ft2 as the clearance standard for floors.
** Grantees that used 100 jig/ft2 or 80 ng/ft2 as the clearance standard for floors.
Objective 1:   Within each component type (and across component types), characterize the
              number of individual samples, work areas, and housing units that pass or fail
              clearance testing standards at various stages (i.e., site visits) of the clearance
              process.
Objective 2:   Within each component type, characterize the distribution of dust-lead loadings,
              the geometric mean dust-lead loading, the variability between samples collected
-------
              from within the same housing unit, and the variability between samples collected
              from different housing units.

Objective 3:   Characterize the correlation in dust-lead loadings between samples collected from
              different component types within the same work area.

Objective 4:   Using lead-loading results from individual samples, demonstrate the impact that
              composite sampling would have on the pass/fail rate of houses by evaluating
              different types of clearance criteria placed on composite sample results.
       To assess these objectives, analyses were performed for each data source individually.

The analysis results from each of the data sources are summarized and discussed in the body of

the report. Results for individual data sources are presented in the appendices.


1.3   DEFINITIONS

       Throughout this document there are several special terms which will be used. For clarity,

the definition of each term is given below.
HUD Interim Guidelines Clearance Standards

       The clearance samples used in this report were, with one exception, collected in
accordance with the 1990 HUD Interim Guidelines Clearance Standards [2]. The following is a
brief description of what is included in the HUD Interim Guidelines Clearance Standards.

       The following steps were to be performed sequentially:  a "preliminary" final clean-up
       effort, a preliminary final visual inspection to insure that all surfaces requiring abatement
       had been addressed and cleaned of all visible dust and debris, painting and sealing of
       abated surfaces and of floors, a final clean-up, and post-abatement visual inspection.
       Following the post-abatement visual inspection, dust wipe samples were to be collected
       and compared to clearance standards of 200 ug/ft2 for floors, 500 ng/ft2 for window sills,
       and 800 ^g/ft2 for window wells (now called window troughs). Results above the
       clearance standard required recleaning and retesting. The numerical values of 200,500,
       and 800 jig/ft3 are known as the HUD Interim Guidelines Clearance Standards.

       These clearance standards, originating in the State of Maryland, were set at levels
considered to be feasible and with the belief that these levels would reduce the number of
children exposed after returning to a unit that had been abated.

       The HUD Interim Guidelines Clearance Standards required a careful reading as to what
action to take if a clearance sample was equal to the associated clearance standard. Since
subsequent EPA guidance and regulation stated that equaling a clearance standard meant failing
-------
 clearance3, this report interpreted "greater than or equal to" a clearance standard to mean that a
 sample failed clearance.

        One exception to the procedures in the HUD Interim Guidelines Clearance Standards was
 the set of clearance samples collected in the HUD FHA Demonstration Project. This project
 used the National Institute of Building Sciences guidelines [6] that existed at the time of the
 study. In the FHA Demonstration, clearance testing was conducted after "final cleaning," but
 before the application of primers, recoatings, or sealants. The numerical clearance standards,
 however, were the same as in the HUD Interim Guidelines Clearance Standards.

       Unless otherwise noted, the clearance standards used for the analyses in this report were
 the numerical standards in the HUD Interim Guidelines Clearance Standards.

 Site Visit

       After a housing unit had been abated and had passed a visual inspection the Ith site visit
 refers to the set of clearance dust wipe samples collected during that visit.  For example, the first
 site visit would refer to the set of clearance samples collected right after the housing unit was
 abated, while the second site visit would typically include only those samples from components
 that failed clearance on the first site visit and required a second site visit to have samples
 collected.

 Passed Clearance Data Set

       This term refers to the dust lead clearance sample results collected within a housing unit
 that led to the housing unit passing clearance standards.

       For example, consider a home where, during the first site visit, all floor and window sill
 dust wipe lead loading samples passed clearance, but two of the six window trough samples
 failed the clearance standard. As a result, the home failed clearance. After re-cleaning, during
 the second site visit two window trough samples were collected and they passed the clearance
 standard, thus passing the home. For this  home, the Passed Clearance data set would then
 include all floor and window sill samples  from the first site visit and the four window trough
 samples that passed clearance in the first site visit along with the two window trough samples
 from the second visit.

       This data set was developed to assess component-specific distributions of lead in dust
 wipe samples that passed clearance testing from housing units that passed clearance. These
 distributions were used to address an important question: Given that a specific component type
 (floor, window sill or trough) passed the HUD Interim Guidelines Clearance Standards, what
 percentage of components would have also passed clearance if the standard would have been
       3EPA's rule on the clearance standards uses "equals or exceeds" as the failure criterion at
40 CFR 745.227(e)(8)(vii) [1] and "above or equal" as the failure criterion in the notice
published at 60 FR 47248 [5].
-------
lower, and the same level of stringency of cleaning and visual inspection were used? For
example, what percentage of floor samples that had passed clearance testing at the 200 jig/ft2
standard would have also passed clearance if the standard had been 100 ug/ft2?

Component Type

       Component type(s) in this document include: Interior Bare Floor, Window Sill, or
Window Trough. Although there are other types of floor samples (e.g. carpeted floor samples),
only bare floor samples were included in the analyses in this report.

Statistical Significance

       Throughout this report geometric means are compared to determine if they are similar or
statistically different. For two independent samples,  the difference between the two geometric
means is statistically significantly different if neither of the 95% confidence intervals contain the
other (the level of significance is actually less than 5% for the comparison). All other statistical
tests have a 5% level of significance.

HUD Grantee Data

       This term refers to the Evaluation of the HUD Lead Hazard Control Grant Program data
which were collected through January 1999 and were managed by the University of Cincinnati
(UC) Department of Environmental Health. Fourteen grantees participated in the evaluation.

      HUD Grantee (High) represents data from a  group of nine grantees (Alameda County,
Baltimore, Boston, California, Massachusetts, Milwaukee, Rhode Island, Vermont, and
Wisconsin) that used the original HUD Interim Guidelines clearance standards, i.e., 200,500,
and 800 ug/ft2 for floors, window sills, and window troughs, respectively.  HUD Grantee (Low)
represents data from the other group of five grantees  (Cleveland, Chicago, New Jersey, New
York City, and Minnesota) that used a lower floor dust-lead clearance standard (i.e. 100 ug/ft2 or
80 ug/ft2).

      Dust lead samples from HUD Grantee homes that received a no-action interior
intervention strategy were not analyzed for the analysis in this report.

Work Area
      The term "work area" refers to the room or room-equivalent (e.g., hallway, stairway) in
which lead-hazard reduction activities had occurred within a residential unit.
1.4    STRUCTURE OF THE REPORT

       For each objective listed above, Section 2 discusses the conclusions drawn from the

statistical analyses. Section 3 lists the quality assurance information available from the data
-------
sources. Background information specific to each study is presented in Section 4. Similar to
Section 2, the statistical analyses used to assess the objectives are listed in Section 5 by
objective. Section 6 presents the summary results of the analyses across all the data sources by
objective. References are provided in Section 7. Appendices A through G present the results of
the statistical analyses for each data source separately. Appendix H presents additional analysis
results on the percentage of housing units that passed clearance on the first site visit, by the
number of individual samples collected within a housing unit, Appendix I presents additional
analysis results on the distributions of the clearance data, and Appendix J lists the key features of
EPA and HUD clearance guidance and regulations that have been published from 1990 to the
present.
1.5  PEER REVIEW
       The report was peer reviewed on two separate occasions by individuals with expertise in
the field. For the first peer review, five individuals were sent copies of the report and asked to
provide comments.  The following summary lists the major comments from the reviewers and
responses to those comments. One of the reviewers recommended that clearance data from the
HUD Grantee evaluation of interventions should be obtained, analyzed, and the results included
in the report. This data was obtained, and the results of the analysis of this data were included in
the revised report. Another reviewer commented that the types of housing covered in the report
were largely from the Northeast region of the country and largely representative of government
funded or subsidized housing. In particular, this reviewer felt government funded or subsidized
housing would contain more homogeneous surfaces than would likely be encountered in most
privately owned housing. The inclusion of the data from the HUD Grantee evaluation is
responsive to both these concerns. Another reviewer commented on the difficulty of seeing
similarities between individual source data and a combined set of all data from all sources. In
response, the combined set of all data was not used in the analysis for the revised report.
       Another reviewer commented on quality assurance aspects of data.  Quality assurance
information was reviewed.  For one data source, clarifying information regarding the data could
not be obtained, and all the data from that source was deleted in the revised report.   For another
data source, necessary clarifying information was identified and obtained. The report was
updated and revised where necessary. One reviewer noted that there were errors in certain
-------
formulas.  Appropriate corrections were made to the report. A reviewer commented on the
chemical analysis of composite samples. The composite samples in this report were created
mathematically from single samples for statistical analysis. Text was revised to make it clear
that the composite samples discussed in the report were constructed by mathematical
combinations.  A comment was received stating that there should be a literature search for the
report. However, the report already contained in an appendix a summary of pertinent guidance
and regulations regarding key features of clearance testing, and this summary of guidance and
regulations was regarded as appropriate for the scope of the report.
       In addition, there were numerous comments related to conclusions, presentation of data,
and clarity of exposition. Graphical presentations of data were added, conclusions restated, and
text revised in response to these comments.
       Because the inclusion of the HUD Grantee data and the deletion of the combined data set
across all sources were major changes to the report, a second peer review of the revised report
was conducted. Four experts in the field, one of whom was in the original group of reviewers,
were asked to peer review the report.
       There were far fewer comments in this peer review compared to the initial peer review.
The following summary is illustrative of the comments received in this peer review.  One
reviewer commented that the report should note the changes in clearance standards that would
result from EPA's publication of a new rule on dangerous levels of lead.  The report was updated
to make the reader aware of these changes.  (The data in the report itself were from clearance
evaluations conducted before the publication of the new EPA rule.)  A reviewer commented that
statistical significance of results was not addressed in the discussion of the relationship between
sample size and the probability of failing clearance at a house. Tables and text were revised to
make the cases of statistically significant results clear to the reader. A comment was received
regarding the use of the kappa statistic to adjust for expected chance concordance.  Upon review,
a determination was made that the analysis of conditional probabilities among the types of
components typically tested addressed the information need expressed by the reviewer, and
accordingly the kappa statistic was not used.
       EPA has established a public record for the peer review of this report under
Administrative Record 235. 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.
-------
2.0   FINDINGS AND CONCLUSIONS
       The data from eight data sources contain an extensive amount of dust-lead loading
clearance testing results for single and multi-family dwelling units. The data represent the
amount of lead left behind following lead-based paint abatement or other interventions once
attempts have been made by lead  contractors to meet the HUD Interim Guidelines Clearance
Standards. Four objectives were addressed in the statistical analysis of the clearance sampling
dust-lead loading results and were described in Section 1. In this section, an assessment of each
objective is presented. Later, in Section 5, details of the statistical analysis procedures used to
assess each of the objectives are presented.
       For the HUD Grantee data, two analyses were performed: one for a group of nine
grantees that used the HUD Interim Guidelines clearance standards (i.e., 200,500, and 800 fig/ft2
for floors, window sills, and window troughs, respectively); the other for a group of five grantees
that used a lower floor dust-lead clearance standard. The first group includes Alameda County,
Baltimore, Boston, California, Massachusetts, Milwaukee, Rhode Island, Vermont, and
Wisconsin. The second group includes Cleveland, Chicago, New Jersey, and New York City,
which use 100 ug/ft2 as the floor dust-lead clearance standard, and Minnesota, which uses 80
Hg/ft2.  The first group is labeled as "HUD Grantee (High)" and the second group is labeled as
"HUD Grantee (Low)" in this report.
       Each individual data source was analyzed separately.  The results of these independent
analyses of each data source are located in Appendices A through G.  The analysis results from
each of the data sources are summarized and discussed in the main body of this report.
Altogether there are eight data groups presented in this report, i.e., data from Maryland, HUD
FHA, HUD PHA, HUD Grantee (High), HUD Grantee (Low), Atlantic City, Cleveland, and
Dover.
2.1   OBJECTIVE 1:  CHARACTERIZATION OF THE NUMBER OF INDIVIDUAL SAMPLES.
      WORK AREAS, AND HOUSING UNITS THAT PASS OR FAIL CLEARANCE TESTING
      STANDARDS
      •   On average there were 8.0 dust wipe samples taken per housing unit on the first site
          visit.  Failure rates of individual samples, averaged across eight data sources, were
          lowest for floor samples at 8.7% (the range, over the eight data sources, is from 2.1%
          to 18.8%) followed by samples from window troughs at 10.4% (ranges from 1.9% to
                                          10
-------
           36.4% over eight data sources) and samples from window sills at 10.5% (ranges
           from 1.1% to 38.9% over eight data sources) during the first site visit. The
           proportion of housing units that failed clearance testing during the first site visit was
           higher, with an average of 16.8% of units failing based on high window trough dust
           lead levels (ranges from 3.4% to 56.5% over eight data sources), 23.0% failing
           clearance based on high floor dust lead levels (ranges from 7.3% to 46.2% over eight
           data sources), and 18.5% failing clearance based on high window sill dust lead levels
           (ranges from 3.0% to 41.7% over eight data sources). The higher failure percentages
           associated with housing units is due to that fact that a housing unit fails clearance if
           any individual dust wipe sample from a given component fails clearance, and in fact
           many of the units that failed clearance testing had high lead levels on more than one
           component type.

       •   Based on a logistic regression model of first site visit data, the estimated probability
           of failing clearance testing increases with the number of dust samples collected. For
           example,  among 4,487 residential units, it was estimated that approximately 25.4%
           failed clearance when 2 samples were collected, 27.3% failed when 4 samples were
           collected, and 28.9% failed when 6 samples were collected.  This trend may be
           attributable to more intensive sampling being conducted in residential units that
           either required more lead hazard control work, or that were in poor condition.

       •   On the first site visit, an average of 91.8% of the individual dust wipe sample results
           and 88.1% of the rooms tested passed the HUD Interim Guidelines Clearance
           Standards, while only 66.9% of the housing units passed clearance. Eventually, 87%
           of the housing units were known to have passed.  The remaining 13% of the units
           that did not pass clearance testing mostly were considered losses to follow-up,
           attributed to families moving, litigation, poor record keeping, housing units that were
           still in the process of meeting clearance, or housing units that never passed clearance
           testing; however, some of these probably eventually passed clearance. Since 87% of
           the housing units passed clearance and only 13% of the housing units were lost to
           follow-up or some might simply have failed clearance after multiple attempts, a
           conclusion was drawn that the clear majority of residential housing units could pass
           clearance if properly abated and allowances are made for multiple clearance testing
           site visits.
2.2   OBJECTIVE 2; CHARACTERIZATION OF THE DISTRIBUTION OF THE DUST-LEAD
      LOADINGS. GEOMETRIC MEAN DUST-LEAD LOADINGS. VARIABILITY WITHIN A
      HOUSING UNIT. AND VARIABILITY BETWEEN HOUSING UNITS

      •   For each of the first two site visits, the geometric mean dust wipe lead loading from
           floors was significantly lower than the geometric mean for window sills. In turn, the
           geometric mean of window sill samples was significantly lower than the geometric
           mean of the window trough samples in the first two site visits. This trend of
           geometric mean floor dust-lead loadings being lower than window sill dust lead
           loadings, and window sills being lower than window troughs continued in the third
                                          11
-------
           and fourth site visits; however the differences in these later site visits were not
           generally statistically significant due to smaller sample sizes.  These results suggest
           that window sills generally could not be cleaned to the same levels as floors, and
           window troughs could not be cleaned to the same levels as window sills.  However,
           these results do indicate that current lead clearance standards can be reasonably
           attained by lead abatement contractors.

       •   The geometric mean lead loading of the samples collected for each component type
           generally increased from the first site visit to the third site visit.  For example, the
           geometric mean floor dust lead loading from samples collected during the first site
           visit ranged from 8.8 to 57.6 ug/ft2 with an average of 27.0 ug/ft2, from 13.8 to 99.9
           ug/ft2 with an average of 40.2 tig/ft2 during the second site visit, and from 19.7 to
           213.4 ug/ft2 with an average of 77.0 ug/ft2 during the third site visit. This trend is
           expected since generally only homes that failed clearance testing were resampled in
           the subsequent site visits. Samples that were taken in the later site visits would be
           expected to have higher lead loadings because only the "dirtiest" homes would be
           included.

2.3    OBJECTIVE 3: CHARACTERIZATION OF THE CORRELATION BETWEEN
       COMPONENTS SAMPLED IN THE SAME WORK AREA

       •   In each of the first three site visits, within room correlation coefficients were
           estimated among dust-lead loadings from floors, window sills, and window troughs.
           In each of these site visits, the highest correlation was observed between window
           trough and window sill samples, followed by the correlation between floors and
           window sills, with the  lowest correlation being observed between floors and window
           troughs.

       •   If a dust-lead loading of one component type is below the HUD Interim Guidelines
           Clearance Standard, it  is very likely that the dust-lead loading for another component
           type within the same room will be below its respective clearance standard. For
           example, as seen in Table 6-12a, given that the window sill dust-lead loading is less
           than 500  ug/ft2, there is a high probability (> 84%) that a floor dust-lead loading
           sample collected from  within the same room will be less than 200 ug/ft2.

       •   The probability that a component will fail its HUD Interim Guidelines Clearance
           Standard given that a different component from the same room failed its respective
           clearance standard is generally below 50% (Tables 6-12a to 6-12c).  A notable
           exception is that when a window sill sample fails clearance, the probability that the
           window trough sample will also fail is greater than 72% for some data sources and
           still less than 30% for some other data sources (Table 6-12c).

2.4    OBJECTIVE 4:  DEMONSTRATION OF THE IMPACT OF COMPOSITE SAMPLING ON
       PASS/FAIL RATES OF HOUSES

       •   When comparing lead  loading results to the HUD Interim Guidelines Clearance
           Standards, it is very plausible that different clearance decisions (Pass/Fail) will be
                                          12
-------
           made based on the use of either individual samples or composite samples. For
           example, of the 2,092 homes in the HUD Grantee High data group, 1,872 homes
           would have passed clearance for floor dust based on simulated composite samples.
           Of the 1,872 homes, only 1,775 homes would have passed floor dust clearance if
           individual samples were considered (97 homes would have failed) (Table 6-14).

       •   The results of the analysis of simulated composite samples indicated that composite
           sampling is associated with an decrease of sensitivity when using the HUD Interim
           Guidelines Clearance Standard. (Sensitivity is the probability of a composite sample
           failing clearance when at least one of the subsamples within that composite had a
           lead-loading result above or equal to the HUD Interim Guidelines Clearance
           Standard for individual samples.  Higher sensitivity indicates a better possibility of
           making correct failing clearance decisions based on the composite samples.) For
           example, the probability ranges from 18.2% to 62.7% that a dwelling will fail
           clearance based on simulated floor composites where at least one of the individual
           samples in the composite is above or equal to  the floor clearance standard. The
           probability ranges from 25% to 100% that this will happen for window sills and
           ranges from 20% to 100% for window troughs. Most of the sensitivities are at or
           below 80% for three component types and across eight data sources.

       •   Comparison of composite sample lead-loading results to standards which are lower
           than the HUD Interim Guidelines Clearance Standard for each component type
           resulted in an increase in sensitivity. Two such composite clearance criteria were
           defined and compared.  The 2xStandard/n criterion resulted in fewer false clearance
           passes than the HUD Interim Guidelines Clearance Standard, and fewer false
           clearance failures than the Standard/n criterion.4 (The definitions of the criteria are
           provided in  Section 5.)

       •   The clearance testing results can be summarized with performance characteristics for
           each combination of component type and composite clearance criteria which are
           defined in terms of sensitivity, specificity, positive predictive value, and negative
           predictive value. (The definitions of performance characteristics are provided in
           Section 5.) Reducing the standard by using the 2xStandard/n criterion provided the
           best performance across all of the performance characteristics.
       4Refer to 40 CFR Part 745.227(e)(8)(vii) for the EPA regulation published in January
2001 regarding the interpretation of composite sample results for clearance testing and to the
Federal Register, Volume 66, No. 4, January 5,2001, page 1223 for the basis for the regulation.
                                          13
-------
3.0   OVERALL QUALITY ASSURANCE
       Lead dust clearance data were obtained from three federally funded projects: the HUD
FHA and PHA Lead-Based Paint Abatement Demonstration Projects and HUD's Lead-Based
Paint Hazard Control Grant Program (HUD Grantee Program) and from four state or local
programs: the Maryland Department of the Environment; the Atlantic City (New Jersey) Housing
Authority; the Dover (New Hampshire) Housing Authority; and the Cleveland (Ohio) Lead-
Hazard Abatement Center.
       The data represented clearance results corresponding to specific abatement or lead hazard
reduction interventions that took place between 1989 and 1999. Clearance activities in the HUD
FHA study followed the procedures in the NIBS Guidelines [6].  Clearance activities in the other
studies or projects generally followed the procedures in the HUD Interim Guidelines [2].  The
data collected included 39,301 lead-loading clearance dust wipe sample measurements on floors,
window sills, and window troughs from 4,518 dwelling units.
       The public housing authorities, city agency, and state  agency all indicated that they
followed the recommendations in the HUD Interim Guidelines [2] to have clearance samples
analyzed by laboratories that had appropriate credentials. The recommendations in the HUD
Interim Guidelines and associated list of laboratories for lead analysis were developed by EPA
for HUD. The recommended  laboratories in the Interim Guidelines were accredited by the
American Industrial Hygiene Association, or the American Association for Laboratory
Accreditation, or the EPA Contract Laboratory Program. The HUD FHA and PHA
Demonstrations were conducted with oversight by HUD. The grantees in the HUD Grantee
program were required to submit samples to laboratories that had passed  the proficiency testing
of the Environmental Lead Proficiency Analytical Testing (ELPAT) program [7j. (This program
later became part of the EPA National Lead Laboratory Accreditation Program (NLLAP) [8],
which was the successor to the list of laboratories recommended in the HUD Interim Guidelines.)
Hence all studies and programs which provided data for this analysis are regarded as having
followed adequate procedures and laboratory analysis for the purposes of dust clearance testing.
       Some of the data submitted were associated with a published study or had undergone
extensive quality assurance checks. Other data were submitted on hard copy forms and had to be
key entered. A quality assurance/quality control program was developed for this study to handle
                                          14
-------
each set of data in a way commensurate with its status and to establish consistency across sets of

data.  The following steps were taken as part of the quality assurance/quality control program:


       •   Data Entry Using Paradox Interface - Data received in hard-copy format were
           entered into a database via a Paradox interface. Important alphanumeric
           characterization variables were checked against the look-up tables to ensure accuracy
           of entry.

       •   Data-Checking Programs - Programs were written to screen the data for obvious
           errors and systematic coding mistakes. These programs included a series of range
           checks, frequency counts, and scatterplots that were reviewed by the Quality
           Assurance Team.

       •   Outlier Analysis - An outlier analysis was performed to determine if any of the lead
           loading measurements fell significantly outside the expected range of the clearance
           results.

       •   Data Audit - At least 5-10% of the data were subjected to a data audit. Data values
           taken from the study database were hand-checked against the original hard-copy or
           electronic data files. Of the data that were checked, less than 1% was found to be
           problematic. The records in question were identified in the database, located in hard
           copy reports, and corrected or verified.


       After resolution of all issues and records in question identified in the steps above, the
statistical analysis proceeded. Results of statistical analyses were double checked and revised as

necessary. Software was written to reduce transcription errors from the results of analysis

programs to the tables in the report.  The report was reviewed extensively for accuracy and

consistency. Finally, two peer reviews were conducted to  ensure that conclusions and
methodology were sound.
                                           15
-------
4.0   DESIGN, DATA COLLECTION, AND CHEMICAL ANALYSIS
       Below is a summary of available background information for each data source. When
possible, information is included in the descriptions pertaining to incidence and levels of lead-
based paint in the houses, age of units, the abatements or interventions that were performed,
circumstances of cleaning, use of sealants, and clearance testing specifics. The FHA and PHA
federally funded housing projects have extensive details that are part of published reports [9-12].
The summary of available background information for the HUD Grantee data is based on the
Fifth Interim Report of the evaluation of the HUD Grantee Program [13]. Information on the
housing units collected from the state and local housing authorities is not as detailed. Because
much of the data collection was performed several years ago and/or the original data collectors
could not be contacted, some background information was not available.
4.1   FHA SINGLE-FAMILY HOUSING PHASE OF THE HUD ABATEMENT
      DEMONSTRATION PROJECT
      The HUD Abatement Demonstration Project was conducted from 1989 - 1990 to assess
the costs and benefits associated with several lead-based paint abatement methods [14].  In the
FHA single-family housing phase, the demonstration was conducted in HUD-owned vacant
single-family dwelling units [9]. The FHA held title to the houses. A total of 304 units from
several different U.S. cities were screened for leaded paint, and 172 units were found with
sufficient amounts of lead-based paint to warrant abatement. Paint abatement methods
performed during the FHA phase included enclosure, encapsulation, chemical removal, removal
with heat gun, removal and replacement, and abrasive removal.
      Following abatement, clearance testing was conducted using individual wipe samples in
149 units. Note that three of the 172 units were used as pilot units and 20 units had only exterior
abatement. Wipe dust-lead loading results from all clearance samples were available in the HUD
Demonstration Lead-Based Paint (HUDLBP) Database [15]. Clearance dust-wipe samples were
collected  after the unit had been "final cleaned," but prior to recoating or priming of any surfaces.
Housing units passed clearance testing if all surfaces sampled resulted in dust-lead loadings
below the HUD Interim Guidelines Clearance Standards of 200 ug/ft2 for floors, 500 ug/ft2 for
window sills, and 800 ug/ft2 for window troughs.  If these dust lead standards were not met, the
study protocol required the housing unit to undergo more extensive cleaning and repeated
                                          16
-------
clearance wipe sampling until the unit met the standards, for up to three iterations of cleaning
and retesting. However, data from the HUDLBP Database indicated that several units underwent
more than three clearance testing iterations.  Additional information on the HUD Demonstration
Lead-Based Paint Database can be  found in [15].

4.2    PUBLIC HOUSING ADMINISTRATION (PHA) PHASE OF THE HUD ABATEMENT
       DEMONSTRATION PROJECT
       (ALBANY, OMAHA, CAMBRIDGE)
       The three PHAs selected to participate in the HUD Demonstration were Albany, New
York; Omaha, Nebraska; and Cambridge, Massachusetts [14].  A total of 109 units were included
at these three demonstration sites. The following material is comprised of excerpts from
published reports [10-12].
       The PHA Phase of the HUD Abatement Demonstration Project was performed to assess
the costs and benefits associated with performing lead-based paint abatement in multifamily
housing. The project in Cambridge involved two garden apartment buildings, each with 24
housing units.  Paint abatement was conducted using chemical methods for the housing units in
one building, and abrasive methods were used for the housing units in the other building. In
Albany, there were also two apartment buildings, each with 18 housing units. Paint abatement
was performed in the first building  using encapsulation and enclosure systems, and chemical
stripping was used for housing units in the second apartment building. The apartment complex
in Omaha consisted of brick faced townhouses, which were abated using component removal and
replacement.
       Following abatement, clearance testing using individual wipe samples was conducted in
the housing units of each building.  Wipe dust-lead loading results from all lead clearance
samples were available from records collected during the study. Housing units passed clearance
testing if all surfaces sampled resulted in dust-lead loadings below the HUD Interim Guidelines
Clearance Standards of 200 ng/ft2 for floors, 500 ug/ft2 for window  sills, and 800 fig/ft2 for
window troughs. If these dust lead standards were not met, the study protocol required the
housing unit to undergo more extensive cleaning and repeated clearance wipe sampling until the
unit met the standards. All units abated passed clearance.
                                          17
-------
4.2.1  The Albany Demonstration
       The project in Albany involved a major redesign of two 18-unit, three-story buildings
which are part of a 6-building, 108-unit project. Because the Albany Housing Authority (AHA)
chose to carry out almost complete redesign of the buildings, many building components were
removed and replaced.

Background
       The Robert Whalen Homes Development, Albany's first public housing project, was
constructed in 1950, and consisted of six similar three-story buildings, each with six units per
floor, or eighteen units per building. Four of the buildings (Buildings A-D) are in a cluster. The
buildings are brick with concrete slab floors, steel frame windows, and wood interior doors,
shelves and trim. There was a single entry and central stairway in each building. Interior
hallways provided access to the individual units. The demonstration was carried out in only two
buildings.
       All six of the Robert Whalen Homes buildings were deleaded and modernized.  The
rehabilitation of the demonstration buildings called for the central boilers to be replaced with
modular boilers in each building.  The electrical service system was to be replaced. Vinyl tile
flooring was to be installed over existing vinyl asbestos tiles and asbestos-containing pipe
insulation was removed throughout the building. The fire alarm system was replaced, a new
sprinkler system was installed, and all doors were replaced. Kitchen cabinets and sinks, and
bathroom fixtures and toilet accessories that were installed during the previous modernization
were saved and re-used.
       The buildings were also redesigned so that ground floor units would have their own
outside entry. The second and third floors were used to create two-story units.
The Extent of Lead-Based Paint in the Demonstration Buildings
       The demonstration buildings were tested for lead-based paint over the period March -
April 1990. The XRF testing was performed according to the HUD Interim Guidelines using
Warrington Microlead I Revision 4 portable XRF analyzers. Paint samples were obtained from
all substrates for atomic absorption spectroscopy (AAS) testing, whether or not the XRF results
indicated a need for confirmatory AAS testing. Because the buildings were occupied during the

                                          18
-------
testing, the area where paint samples were obtained was immediately resealed. The XRF and
AAS test results were similar in both buildings.
       Lead-based paint, as measured by XRF analyzers, was mostly limited to closet shelves
and shelf supports, bathroom doors, limited areas of walls and ceilings, window sills, and stair
treads and risers on the interior steel common staircases.  The XRF results indicated a high
percentage of window sills (76%) and risers (64%) contained lead-based paint, but this finding
was in conflict with laboratory AAS test results (2% and 0%, respectively), which are considered
to be more precise and accurate.  "This difference probably reflects the measurement error
associated with XRF  testing, particularly on metal surfaces.  In all cases in which the XRF and
AAS findings differed substantially, the XRF results indicated a higher incidence of lead-based
paint hazards" [11].
       A total of 3,034 XRF readings were obtained and 12% were above or equal to the
threshold of 1.0 mg/cm2. In addition, 1,226 AAS samples were obtained and 4% were above or
equal to the threshold of 1.0 mg/cm2.

4.2.2  The Omaha Demonstration
       The Omaha Demonstration buildings are two-story town house structures constructed in
1938. The six buildings contain a total of 49 units ranging in size from 1 to 5 bedrooms.
Construction is concrete frame floors with brick on the exterior walls, and partition walls are
plaster over clay tile.  Interior stairs are metal with concrete treads.  During the course of
previous modernization work,  all windows and window frames had been replaced, and new
soffits, fascias, and porches had also been installed. Doors, door frames, window sills, shelves,
shelf-cleats, and baseboards were painted wood.
       The Stage IV modernization plans for the Southside Terrace buildings consisted  of seven
types of improvement: (1) enclosing or removing asbestos tile floors and removing asbestos in
crawl spaces and basements, (2) insulating and furring interior and exterior walls and covering
interior walls with new sheetrock, (3) reconfiguring the upstairs of most units to change the
number of bedrooms, moving partitions in the washer/dryer, kitchen, and pantry areas, and
eliminating exterior storage sheds, (4) replacing plumbing and plumbing fixtures in bathrooms,
(5) installing new ceilings and new overhead lighting, (6) replacing existing cabinets,  installing
                                          19
-------
new closets, and improving access to under-stair storage area, and (7) replacing existing hot
water heating system with electric radiant heat.
       The painted surfaces in these units consisted of baseboards, window sills, interior doors
and door frames, exterior doors and door frames, shelves, shelf-cleats/chair rails, interior walls
and ceilings. Planned modernization of these units would have effectively abated most of the
lead-based paint even without the lead abatement objective.
The Extent of Lead-Based Paint in Demonstration Buildings
       Systematic testing of the six Demonstration buildings for the presence of lead-based paint
was performed using protocols described in the Research Design. The first four buildings (the
Group 1 buildings) were tested using Wanington Microlead I Revision 4 portable XRF
analyzers. Paint samples were taken from substrates which could not be tested with the XRF
machines and subjected to laboratory analysis using atomic absorption spectroscopy (AAS).
AAS tests were also performed, when XRF readings were in the range where confirmatory AAS
testing was required by the HUD Interim Guidelines, and when confirmatory testing was
required by the Research Design for the Demonstration. The two buildings added later (the
Group 2 buildings) were tested for lead-based paint in the same way, except that the portable
XRF analyzer used was a Scitec instrument.  Testing in all buildings was carried out after the
units had been vacated. The Douglas County standard of 0.6 mg/cm2 and the HUD Interim
Guidelines standard of 1.0 mg/cm2 of lead were used as criteria for classifying a reading as
positive. There were 2,190 XRF readings from Building Group 1 and 1,522 from Building
Group 2.
       The use of the more stringent 0.6 mg/cm2 standard (compared to the 1.0 mg/cm2) for
abatement increased the percentage of all building components testing positive on the XRF
reading from 8% to 17% in Building Group 1 and from 34% to 43% in Building Group 2.  The
percentage of all building components testing positive on the AAS samples increased from 4% to
6% in Building Group 1  and from 21% to 27% in Building Group 2. For some components, the
percentage testing positive was almost the same under both standards (i.e., baseboards, shelf-
cleats, doors, and door frames on the XRF readings and door frames and window sills on the
AAS test), while for others, the difference in the percentage testing positive was quite striking
(i.e., ceilings and walls on the XRF readings and ceilings on the AAS test).
                                          20
-------
4.2.3  The Cambridge Housing Authority Demonstration
       The CHA's Newtowne Court project contains 282 units, configured into a number of
three story walk-up buildings. The project was constructed in 1938. The building construction
consists of brick exteriors, load bearing 4" block interior walls covered with plaster, and concrete
floor slabs. Each building also has a full basement, with some being used for storage and
laundry facilities.
       The condition and layout of the units in Newtowne Court was consistent with public
housing units built in the late 1930s - deteriorated interior surfaces, small and inefficient
kitchens, outdated baths, and numerous coats of paint, applied both by the CHA and tenants.
CHA's initial modernization plans included the construction of a bay window, or extension, in
the kitchen to provide needed additional space. CHA also planned on modernizing the
bathrooms and restoring the remainder of the units to present-day standards. Extensive site
improvements were also planned.  A new heating system had been installed in 1985; however the
buildings' plumbing and electrical systems required a full upgrade. All asbestos-containing pipe
insulation was to be removed during the modernization.
       CHA's strategy for modernization involved vacating two 12-unit buildings, completing
modernization in these buildings, and then using the completed buildings as a temporary "hotel"
for the residents, until their buildings were modernized. These two initial buildings were .
selected for the demonstration.  Each building has two central stairways,  each having access to
two units on each of the three floors.

The Extent of Lead-Based Paint in the  Demonstration Buildings
       The two demonstration buildings were tested for the presence of lead-based paint during
the spring of 1990. The XRF testing protocols were the same as those in the HUD Interim
Guidelines except that five repeated measurements per test point were taken using a Warrington
Microlead IML-1, Revision 4 direct reading XRF machine, versus three measurements specified
in the HUD Interim Guidelines5. Comprehensive AAS sampling was also performed whether or
not the XRF results indicated a need for  confirmatory AAS testing. Confirmatory testing
       5 Massachusetts regulations required that the average of 5 repeated measurements taken at the same point
be used to determine the need for abatement.
                                           21
-------
involves the extraction of paint chip samples on substrates where the XRF readings were within a
defined range around the abatement threshold of 1.0 mg/cm2. This process is detailed in the
report [10], and was done to permit comparisons of XRF and AAS measurements on a wide
range of lead paint concentration levels. The only other variance from the HUD Interim
Guideline protocols was CHA's decision to test both sides of each door frame by AAS because
different sides of the door frames may have had different paint histories.
       A total of 2,194 XRF readings were obtained; 16% indicated lead levels above or equal to
the 1.0 mg/cm2 threshold. In addition, 896 AAS samples were obtained and 11% were above or
equal to the 1.0 mg/cm2 threshold.
4.3    MARYLAND DEPARTMENT OF THE ENVIRONMENT
       The Lead Enforcement Group within the Maryland Department of the Environment
(MDE) has been actively conducting post abatement clearance testing since 1988 [16].  When
they first started their clearance testing program, there were no protocols for sampling or
standards developed for the purpose of clearance testing. Based on scientific evidence, this
program developed protocols for the collection of dust wipe samples from floors, window sills,
and window troughs in rooms that had received abatement. The clearance standards of 200
|ig/ft2 for floors, 500 ug/ft2 for window sills and 800 ug/ft2 for window troughs also originated
from within the MDE Lead Programs. These standards were based on pilot data, and were
designed so that a work area would pass clearance through reasonable cleaning efforts on the part
of the lead contractor.  These clearance standards also represented levels of post-abatement lead
on floors, sills and troughs that were far below the levels of lead that would typically be observed
on these components prior to abatement.
       The lead enforcement group has archived data on dust-lead loading results since they
started conducting clearance sampling in 1988. These data existed only hi hard copy form, and
were made available to EPA for the purposes of this project. All clearance testing sample results
collected between January 1,1991 and January 30,1995 were entered into an electronic database
for the purposes of this investigation. The statistical analysis of data from MDE represents these
four years of clearance testing results throughout the state of Maryland.
       If the results from  a dust-lead loading sample exceeded the clearance standards, the area
had to be recleaned and retested until acceptable results were obtained for each unit. However,
                                           22
-------
there are no records of additional testing for approximately 23% of the units which failed
clearance. These losses to follow-up may be attributed to families moving, litigation, or poor
record keeping. Some of these may have been houses that had ongoing recleaning and retesting
after January 1995. Since the Lead Enforcement Group within MDE is a state regulatory
program, and not a research program, this loss to follow-up within the data could not be avoided.

4.4    EVALUATION OF THE HUD LEAD-BASED PAINT HAZARD CONTROL GRANT
       PROGRAM

       The following summary is based on the "Evaluation of the HUD Lead-Based Paint
Hazard Control Grant Program: Fifth Interim Report, Progress as of September 1,1997," which
was published in March 1998 [13].
       The overall purpose of the evaluation of the HUD Lead-Based Paint Hazard Control
Grant Program ("HUD Grantee Program") is to measure the relative cost and effectiveness of the
various methods used by State and local government grantees to reduce lead-based paint hazards
in housing [17]. Approximately 2,900 dwelling units were followed for 12 months and
approximately 750 units were followed for 36 months [13]. Although national in scope, this
program is locally driven and implemented. Grantees design their own programs, including the
methods of recruitment and the treatments that they carry out.
       There are 14 grantees in the HUD Grantee Program, with 11 starting in FY 1992 (first
round, required to participate) and 3 starting in FY 1993  (second round, voluntary participation).
The administration, sampling targets, and abatement methods that were used by each grantee are
summarized in Table D-l. Program standard forms and procedures were developed by the
University of Cincinnati (UC) Department of Environmental Health and the  National Center for
Lead-Safe Housing (NCLSH). UC and NCLSH are also responsible for data analysis and
reporting, training, and support of grantee data collection and recording. Data management
quality control was done by UC.
      Data being collected in the HUD Grantee Program are environmental, biological,
demographic, housing, cost, and hazard-control information. Measurements of lead in dust,
paint,  soil, and blood are collected, though not all grantees collect all these measurements.  Pre-
intervention, immediate  post-intervention (clearance results), and 6- and 12-month
                                          23
-------
post-intervention data are collected. However, only selected homes (estimated 800 homes) are

included in the data collection for the 24- and 36-month post-interventions.

       This report analyzes the latest set of HUD Grantee data, which was collected through

January 1999 and was released by the University of Cincinnati hi June 1999.

       Based on the HUD Grantee Program's Fifth Interim Report [13] (data collected through

September 1,1997), the following provides basic information on building characteristics,

occupancy status, environmental sampling, and clearance of the dwelling units in the program.
Building Characteristics

       (1) Types
          •  Single family detached (32%)
          •  Single family attached (25%), rowhouses - mostly hi Baltimore
          •  2-4 Unit Multi-family (37%)
          •  > 4 Unit Multi-family (6%), 85% of NY enrollments were of this type, averaging
             14 units per building

       (2) Age of Housing
             Less than 1% of the enrolled buildings built after 1959
             90% pre-1940
             Median for Cleveland, Massachusetts, Milwaukee, Minnesota and Vermont is pre-
             1910
             Median for Baltimore, Chicago and Rhode Island was in 1920's
             Median for California was in 1930's
Occupancy Status
             20% vacant prior to intervention
             Baltimore had 60% pre-intervention vacancy rate
             Vermont, NY, NJ had from 24% to 34% pre-intervention vacancy rates
             All others had below 14% pre-intervention vacancy rate
Environmental Sampling
             Dust is collected from 7-9 locations during each phase of the evaluation.
             Single-surface dust wipe samples are collected from:
             •  Floor (bare or carpeted):  interior entry, kitchen, child's play room (or living
                room), youngest child's bedroom (or smallest room), next youngest child's
                bedroom (if present).  Note that only bare floor dust samples were included in
                the analyses in this report.
             •  Interior window sill: kitchen, youngest child's bedroom (or smallest room)
                                          24
-------
             •  Window trough:  child's play room (or living room), next youngest child's
                 bedroom (if present)

Clearance of the Dwelling Units
       •  Program requires clearance after intervention
       •  Clearance standards follow 1990 HUD Interim Guidelines
          Floors:        200 ug/ft2
          Sills:         500 ug/ft2
          Troughs:      800 ug/ft2
       •  Since 1994-1995, grantees have been allowed to use 200 ug/ft2, 100 ug/ft2 or use a
          locally established level as the clearance level for floors. Five grantees chose to use
          lower clearance levels for floors.
          •  Cleveland, Chicago, New Jersey, New York City used 100 ug/ft2
          •  Minnesota used 80 ug/ft2
       •  28% of the dwelling units failed the initial clearance dust lead test. There is a wide
          variation in the clearance rates for the grantees, with rates of initial failure ranging
          from 8 to 50% [13].

       All dwellings in the HUD Grantee program are required to "meet clearance" after the
intervention is complete. In other words, dust wipe tests must demonstrate that the amount of
leaded dust on components in all treated rooms does not exceed levels designated by HUD. For
the first and second rounds of the HUD Grantee Program, clearance levels were set at 200, 500,
and 800 ug/ft2 for floors, window sills, and window troughs, respectively. In 1994-1995, HUD
and EPA released new guidance that lowered the clearance level on floors, from 200 fig/ft2 to
100 fig/ft2.  Since then, HUD has allowed grantees to use either 200 ug/ft2,100 ug/ft2, or a
locally established level if less than 200 ug/ft2. Five grantees used clearance levels less than 200
ug/ft2 for floors:  Cleveland, Chicago, New Jersey, and New York City used 100 ug/ft2, while
Minnesota used 80 ug/ft2.
       A main attribute of the HUD Grantee program is the flexibility that grantees have to
select the lead treatments for any particular unit. The grantees have the freedom to treat all areas
of the property or treat only some locations (interior, exterior, and/or soil). The grantees also
decide on the  intensity of the treatment.  Possible treatment intensities range from cleaning with
some spot painting to partial abatement to full abatement.  The grantees have been encouraged to
experiment with different levels of lead hazard control activities. Grantees are allowed to
experiment because there is no consensus on a single state-of-the-art intervention to control lead-
based paint hazards. For example, some grantee programs require that windows containing
                                           25
-------
lead-based paint be replaced to protect the health of residents.  Others contend that by using
lower level/lower cost treatments, more residents can be served, while still protecting then-
health.  Grantees have even decided to leave some lead-based paint hazards untreated, under the
assumption that these hazards have no or limited immediate impact on the residents' health.
       Grantees report the intensity of the interior intervention as strategy code (level 01 to 07).
Higher strategy levels reflect more intensive treatment. Each dwelling unit was assigned only
one interior intervention strategy. The interior intervention strategies are summarized in Table 4-
1. Grantees, as a whole, have selected a diversity of strategies. On an individual basis, however,
most grantees have tended to select one or two dominant strategies [13].

                 Table 4-1.  Interior Intervention Strategy Code Definitions
               6
                     No Action
                     Cleaning, Spot Paint Stabilization Only
                     Level 02 plus
                     Complete Paint Stabilization, Floor Treatments
                     Level 03 plus
                     Window Treatments
                     Level 04 plus
                     Window Replacement, Wall Enclosure/Encapsulation
All Lead-Based Paint Enclosed, Encapsulated, or Removed
(Meets Public Housing Abatement Standards)
                     All Lead-Based Paint Removed
          Note:   Dust lead samples from grantee houses that received a no action interior
                  intervention strategy were not analyzed for the analysis in this report.
                                             26
-------
Glossary of HUD Grantees Interior Intervention Strategies:
       Encapsulation - The application of a covering or coating that acts as a barrier between
          lead-based paint and the environment, the durability of which relies on adhesion and
          which has an expected life of at least 20 years.
       Enclosure - The application of rigid, durable construction materials that are mechanically
          fastened to the substrate to act as a barrier between lead-based paint and the
          environment.
       Paint Stabilization - The process of repainting surfaces coated with lead-based paint,
          that includes the proper removal of deteriorated paint and priming.
       Paint Removal - The complete removal of lead-based paint by wet scraping, chemical
          stripping, or contained abrasives.
       Removal/Replacement - The removal/replacement of a building component that was
          coated with lead-based paint.
       Window Treatments - The process of eliminating lead-containing surfaces on windows
          that are subject to friction or impact through the removal of paint or enclosure or
          certain window components.
4.5    ATLANTIC CITY HOUSING AUTHORITY

       The Atlantic City Housing Authority provided data from a comprehensive rehabilitation

project performed on public housing buildings containing 10 to 23 multi-family housing units

[18]. These two- to three-story brick buildings had lead-based paint on the doors, windows,

radiators, trim, and stairwells. The Atlantic City Housing Authority removed the doors and

windows and replaced them along with most of the trim during abatement. Walls were enclosed

by drywall, then painted, and steel panels were placed on the walls in the stairwells.  There were

two phases to the lead clearance process: worker entry clearance and re-occupancy clearance.

During worker entry clearance, the first phase, clearance samples were collected after protected

workers had finished the abatement job. The clearance samples collected during the second

phase, the re-occupancy cleanup, were taken after the renovation was completed but before the

unit was reoccupied.  Only samples collected during the second phase, or re-occupancy cleanup,

were available and used in the analysis for this report. Within each completed unit, one dust

sample was collected from each room or area where abatement occurred. Sample locations were

randomly distributed between floors and window troughs. Since the abatement process included

removal of many windows, few window sill and window trough samples were tested. The HUD

Interim Guidelines Clearance Standards thresholds of 200 ug/ft2 for floors, 500 ug/ft2 for

window sills, and 800 jig/ft2 for window troughs were utilized in determining whether dust wipe
                                          27
-------
 samples passed or failed. In all cases where results exceeded the clearance thresholds, the areas
 were re-cleaned and re-tested until acceptable results were obtained.

 4.6    CLEVELAND LEAD HAZARD ABATEMENT CENTER
       The lead abatement program managed by the Cleveland Lead Hazard Abatement Center
jointly with the Cleveland Department of Public Health Environmental Health Watch recruited
 houses where children with elevated blood lead levels lived [19].  Most of the units were large
 single- family houses built before 1950. The level of intervention depended on the age and
 number of children in the house. Many of the homes were completely abated.  Chipped paint
 was wet sanded and repainted with a primer and a coat of paint. Carpets were removed and the
 floors refinished. Windows were replaced in some units. Vinyl siding was applied to homes with
 lead-based painted wood siding. Porches were repaired, deteriorated sections replaced, and other
 surfaces scraped and repainted. If lead  levels in the soil were high, the soil was removed and
 replaced.  Sod, woodchips, or some other form of landscape coverage was placed over the
 exposed soil. This project was continuing at the time of the data analysis for this report.

 4.7    DOVER HOUSING AUTHORITY. NEW HAMPSHIRE
       One-hundred and eighty-four units in 49 multi-family buildings were abated and
renovated [20]. Forty-three of the buildings had four units per building while the other six
 buildings had two units per building. The exterior siding and window sashes of the buildings
 contained lead-based paint, so siding and windows were removed and replaced. Interior lead-
 based paint was found only on water radiators.  Thirty-one of the units had cast iron radiators.
 The radiators were removed from the apartments, sand blasted, off-site repainted, and reinstalled.
 Even though there was very limited abatement activity inside the units, dust wipe samples from
the window trough and sill along with floor samples were collected for lead clearance following
 cleaning of the work area.  The HUD Interim Guidelines Clearance Standards of 200 ug/ft2 for
 floors, 500 ug/ft2 for window sills, and  800 ug/ft2 for window troughs were utilized in
 determining whether dust wipe samples passed or failed.
                                           28
-------
5.0   DATA METHODOLOGY AND ANALYSIS
       The four objectives listed in Section 1 were addressed in the statistical analysis of the
clearance sampling dust-lead loading results for each source of data separately.  The statistical
methods corresponding to each of these objectives are discussed below.

5.1    OBJECTIVE 1: CHARACTERIZATION OF THE NUMBER OF INDIVIDUAL SAMPLES.
       WORK AREAS. AND HOUSING UNITS THAT PASS  OR FAIL CLEARANCE TESTING
       STANDARDS
       The clearance testing data investigated in this analysis represent post cleanup dust-lead
loading results of wipe samples collected from floors, window sills, and window troughs in and
around abatement work areas.  The samples were collected and chemically analyzed as
individual samples. For the purposes of this analysis, the dust lead loading results from all data
sources were assumed to represent a clearance testing program which compared lead loading
results to the HUD Interim Guidelines Clearance Standards. The dust lead loading results were
compared to clearance standards of 200 jig/ft2 for bare floors, 500 ng/ft2 for window sills, and
800 ug/ft2 for window troughs. When a unit failed clearance, additional cleanup was performed.
For those units that failed, a second site visit followed the first site visit, with cleanup and
sampling being performed generally only in the areas which failed clearance standards.  Results
of these sample analyses determined whether the unit passed or failed clearance at this stage.
Data were available for up to five visits for some residences. However, there were very few
dust-lead loading results after the third site visit, so statistical results for the fourth and fifth site
visits are not considered reliable in most cases.
       Within each clearance testing program investigated, an analysis was conducted to
summarize the results of clearance testing with regard to the HUD Interim Guidelines Clearance
Standards. This analysis was designed to provide answers to the following questions:

       1.  During each site visit, how many individual samples passed/failed clearance?
              la. During each site visit, how many floor samples passed/failed clearance?
              Ib. During each site visit, how many window sill samples passed/failed
                clearance?
                                           29
-------
       Ic. During each site visit, how many window trough samples passed/failed
          clearance?

2. During each site visit, how many work areas (rooms) passed/failed clearance testing?

       2a. During each site visit, how many work areas would have passed/failed
          clearance based only on floor sample results (ignoring results from window
          sills and window troughs)?

       2b. During each site visit, how many work areas would have passed/failed
          clearance based only on window sill sample results (ignoring results from
          floors and window troughs)?

       2c. During each site visit, how many work areas would have passed/failed
          clearance based only on window trough sample results (ignoring results from
          floors and window sills)?

3. During each site visit, how many housing units passed/failed clearance testing?

       3a. During each site visit, how many housing units would have passed/failed
          clearance based only on floor sample results (ignoring results from window
          sills and window troughs)?

       3b. During each site visit, how many housing units would have passed/failed
          clearance based only on window sill sample results (ignoring results from
          floors and window troughs)?

       3c. During each site visit, how many housing units would have passed/failed
          clearance based only on window trough sample results (ignoring results from
          floors and window sills)?

4. After all sampling was completed, how many individual samples, work areas, and
   housing units passed/failed clearance?

       4a. After all sampling was completed, how many samples, work areas, and
          housing units passed/failed clearance based only on floor sample results
          (ignoring results from window sills and window troughs)?

       4b. After all sampling was completed, how many samples, work areas, and
          housing units passed/failed clearance based only on window sill sample
          results (ignoring results from floors and window troughs)?

       4c. After all sampling was completed, how many samples, work areas, and
          housing units passed/failed clearance based only on window trough sample
          results (ignoring results from floors and window sills)?
                                    30
-------
       For each source of data, specific subsets of data were used to summarize the answers to
these questions. These subsets are represented in Table 5-1.  Each cell in Table 5-1 corresponds
to a specific subset of data; the question which the data in a given cell addresses is indicated
within the cell.
       The preliminary analyses also suggested that for the majority of the data, only one sample
of each component type was sampled following cleanup in most of the abated work areas. Thus,
with regard to the responses to the above questions, Questions la, Ib, and Ic are essentially
equivalent to questions 2a, 2b, and 2c within each site visit for each source of data.
             Table 5-1.  Location of Answers to Questions on Pass/Fail Rates.
                        (i.',,;—' a.'(»cw n!iiisJ.-.*i'41'-.i'.r
 A Specific
  Site Visit
              Floors
               Sills
             Troughs
               AH
1a
1a
1a
2a
2a
2a
Ib
Ib
Ib
2b
2b
2b
1c
1c
Ic
2c
2c
2c
3a
3b
3c
3a
3b
3c
3a
3b
3c
All Site
Visits
Combined
Floors
Sills
Troughs
All
4a
4b
4c
4
4a
4b
4c
4
4a
4b
4c
4
4a
4b
4c
4
4a
4b
4c
4
4a
4b
4c
4
4a
4b
4c
4
4a
4b
4c
4
4a
4b
4c
4
       There are many factors that can influence the number of clearance samples that are
collected from within a residential unit (size of house, extent of lead hazard, extent of abatement
and sampling protocol). However, information on these factors was generally not available for
this statistical analysis. A similar, related question of interest is: Does the probability that a
housing unit will fail clearance testing increase as the number of clearance dust-wipe samples
collected increases? Four logistic regression models were used to answer this question. The first
logistic regression model was developed to estimate the probability of a housing unit failing
clearance as a function of the number of samples collected.  The model appears as:
                                            31
-------
where
       logit ( ) is the logit function, logit (P) = log (P/(1-P))
       Ttj  is the expected probability of clearance failure for the z* housing unit,
       HJ  is the number of clearance samples collected in the i* housing unit,
       P0  is the intercept parameter, and
       P,  is the slope parameter.

       For the second logistic regression model the numbers of samples collected within a
housing unit were collapsed into five successive categories, as follows:
where
           is the categorized number of samples collected in the i* housing unit, where c{ ranges
           in values from 1 to 5 as ^ takes on values of (1-4 samples, 5-8 samples, 9-12 samples,
           13-16 samples, and greater than 16 samples) respectively, and
          p0 and P, are defined the same as in the previous logistic regression model.
       Both models were applied to clearance testing data collected during the first site visit, and
were fitted both over all component types and separately for each component type (floors, sills,
and troughs).
       The third and fourth logistic regression models are similar to those described above but
were developed using the proportion of samples that failed within a housing unit as the response
variable. n{ was defined as the proportion of dust wipe samples expected to fail clearance testing
in the Ith housing unit.
       Tables were generated that provide estimates of P0 and p,, and the associated standard
errors as well as the estimated probability of failing clearance (or estimated proportion of
samples expected to fail) associated with a housing unit which collected 2,4, and 6 wipe samples
of the associated component type.
                                            32
-------
5.2    OBJECTIVE 2:  CHARACTERIZATION OF THE DISTRIBUTION OF THE DUST-LEAD
       LOADINGS. GEOMETRIC MEAN DUST-LEAD LOADINGS. VARIABILITY WITHIN A
       HOUSING UNIT, AND VARIABILITY BETWEEN HOUSING UNITS
       As shown in Figure 5-1 below, the distribution of the clearance sample results tends to
have a small number of extremely large values. These large dust loading sample observations
are thought to be the result of many minor influences which combine in a multiplicative way.
The multiplicative influences have a greater effect on large dust-lead loadings than on smaller
values. In situations where the error has a multiplicative effect on the true lead level, a natural
logarithm transformation (In-transformation) is recommended prior to performing parametric
statistical procedures based on normally distributed data. If the distribution of the dust-lead
loading clearance sample results is log-normal, then the actual In-transformed results will be
normally distributed. Histograms of the In-transformed floor, window sill, and window through
dust-lead loading data are given in Figures 5-2,5-3, and 5-4, respectively.  The normal
distribution has been overlaid on  the empirical distribution.  The spike in the plot in Figure 5-2
between 1 and 2 on the x-axis is likely due to a large number of results reported at the detection
limit. Review of the data for floor samples suggested that the detection limit for these samples
varied from 5 to 25 micrograms per square foot (ug/ft2).  Except for the spike that appears in
Figure 5-2, the log-normality assumption appears to be reasonable for these data. None of the
inferential statistical protocols in this report are overly sensitive to the assumption of log-
normality. Therefore, the statistical models used in this report are based on the natural logarithm
transformed dust-lead loadings.
       Because the dust-lead clearance sample results were assumed to follow the log-normal
distribution, the geometric mean was used as the measure of central tendency. (The geometric
mean is the inverse-logarithm transformation of the arithmetic mean of the In-transformed dust-
lead clearance samples.)
       For each data source separately, the following statistical model was applied to the dust-
lead loading results to characterize within-unit (owithjn) and between-unit (aBetweeJ variability. It
was fitted separately for each combination of site visit (j =1,2, or 3) and component type (/ =
floors, window sills, or window troughs) (i.e. levels ofy and / were held fixed):
                       \n(PbDiji(1J =
where
                                          33
-------
        30
        20
        10
             0   100200300400500600700800900  1000
                                  Pb loading, //g/ft2
  Figure 5-1.  Histogram of Dust Lead Loadings from Floor Samples.
       20
       15
       10
          0123
4567
 In (Pb loading) fig/ft2
8     9    10
Figure 5-2.  Histogram of the Natural Logarithm of Dust Lead
            Loadings from Floor Samples.
                                    34
-------
12-
                        In  fPto Loading) uo/ft-"*2
                                                         no    11    12
 Figure 5-3.   Histogram of the Natural Logarithm of Dust Lead Loadings
              from Window Sill Samples
                                                       1O   11   12
                        In  (Rt> Loading) uo/tt~*2
 Figure 5-4.    Histogram of the Natural Logarithm of Dust Lead Loadings
              from Window Trough Samples

                                   35
-------
           House,,
             =   geometric mean dust-lead loading across housing units for site visit
                 j and component /,

             =   random effect for the average log(dust-lead loading) within
                  housing unit i for site visit y and component /,

             =   random (within house) error term associated with \og,(PbDijklm),

i=l....N     =   N = the number of housing units sampled,

k=l,..,,Ki    =   £,=the total number of work areas sampled in unit i,

m=l,...,MiJkl  —   Mija = the number of replicate samples of component / collected in
                  work area k of housing unit i during they1*  site visit.
House,-^ is assumed to follow a normal distribution with mean zero and standard deviation

^Between- Error(VbM0;; is assumed to follow a normal distribution with mean zero and standard

deviation o*wilhin. (Note: Because very few results were recorded for the fifth site visits, the

analysis was performed using only the information provided for the first, second, third, and

fourth site visits.)
       The analysis is based on three methods of assessing the distribution of the individual dust

lead loading samples:
       1.   Consider the 50th, 80th, 90th, 95th, and 99th percentiles of the dust wipe samples for
           the first or initial site visit and the final site visit,

       2.   Consider the percentage of individual dust wipe sample results that fell below 50, 75,
           and 100 jig/ft2, and the interim standard for each surface, and

       3.   Consider box and whisker plots for each surface type and for the dust wipe samples
           collected during the first site visit and the dust wipe samples collected during the site
           visit that passed clearance.

           Box and whisker plots describe the distribution of the data in the following manner.
                                            36
-------
                IX
               i:
                 Extreme Values {> Upper Adjacent Value)
                • Upper Adjacent Value (Upper Quartile +1.5 interquartile range*)
                • Upper Quartile (75th percentile)
                       ' Sample Median (50th percentile)
               i
                • Lower Quartile (25th percentile)

                ' Lower Adjacent Value (Lower Quartile -1.5 interquartile range*)
                • Extreme Values (
-------
 (or fail) given that another component in the same room either passed or failed. This type of
 information could be used to determine whether it is necessary to sample from all three
 component types within a single work area or room. These conditional probabilities were
 calculated in two ways using first-site-visit data only, first by empirical evaluation of 2X2 tables,
 and second by using normal probability theory.
       To illustrate the use of these two approaches, an example is provided for calculating the
 conditional probability that a floor would pass clearance given that a window sill in the same
 room passed clearance. Both methods require that the first site visit data be organized to include
 paired observations of floors and sills observed in the same room:

       Floor,-,- represents the geometric mean floor lead loading in they* room of the i* housing
       unit, and
       Sill,., represents the geometric mean window sill lead loading in they* room of the z*
       housing unit.

 The empirical evaluation of 2><2 tables would then organize this data in the following manner:
                                   Floor^ < 200 ug/ft2   Floor;,. ;> 200 ug/ft2
                 Sill,.,-< 500 ng/ft2
                 Silly ;> 500 jig/ft2
a
c
b 1
d 1
where 'a' represents the number of rooms in which both the floor and window sill passed
clearance, 'b' represents the number of rooms in which the sill passed and the floor failed, 'c'
represents the number of rooms in which the floor passed and the sill failed, and 'd1 represents the
number of rooms in which both the floor and the sill failed clearance.  Thus, the conditional
probability that a floor would pass clearance given that a window sill in the same room passed
clearance would be estimated as a/(a+b).
       For the second approach, using the multivariate normal theory approach [21], the data
would be organized as follows:
                                           38
-------
                                             "
                                                  r 2
Modeling the multivariate distribution between floors, window sills, and window troughs using
the multivariate normal distribution provides interested parties the means of estimating these
conditional probabilities without having to directly access the data.
       The parameters of the above multivariate normal distribution are calculated using the
observed (Floor^, Sill,j) pairs. The conditional distribution of floor dust-lead loading given sill
dust-lead loading is
                                            tts
            fr) is defined as the density function of the above normal distribution, then using
standard normal theory the conditional probability of a floor passing clearance given that a
window sill in the same room passed clearance would be estimated as
                                                    ln(200)  ln(500)
P(Floor <
                                                     J     J fF,s(f^dsdf
                                 Sill <
                                                     <»   In(SOO)
                                                     J    J  fF>s(f,s)dsdf
                                                     -08    -00
We also applied similar methods for calculating more complicated conditional probabilities, such
as the conditional probability of a floor passing clearance given that a window sill and trough in
the same room passed clearance.

5.4    OBJECTIVE 4: DEMONSTRATION OF THE IMPACT OF COMPOSITE SAMPLING ON
       PASS/FAIL RATES OF HOUSES
       Under EPA's Section 402 Rule (40 CFR Part 745) [4], composite sampling of dust is
permitted for clearance following an abatement. Additionally, the HUD Guidelines indicate that
                                          39
-------
composited dust samples may be used for conducting clearance testing.  A composite wipe
sample is a group of individual wipe samples collected from multiple locations of the same
component type, which are combined into a single analytical sample. Thus, a composite sample
result could be interpreted as the (area weighted) average dust-lead loading across all locations
included in the sample.
       Analytical results from composite samples have different statistical properties from those
for individual samples. For example, error associated with lead  loading results from composited
wipe samples differs from error associated with lead loadings found in individual wipe samples.
The error attributed to spatial variability in sampling is reduced in composite samples, due to the
averaging of multiple sampling locations. Differences between individual and composite
samples with regard to variability caused by laboratory measurement error has not yet been
established. However, due to the reported difficulties in digesting large numbers of dust wipes
for chemical analysis, it is recommended [3] that composite wipe samples contain at most four
wipes, thereby limiting the number of sampling locations to at most four.
       As an example of the difference in statistical interpretation between a composite sample
and individual samples, consider four individual floor wipe samples with lead loadings of 150,
175,100, and 275 jig/ft2, each collected from one square foot. If the composite sample is the
area weighted average of these four loadings, then the composite lead loading result would be
175 jig/ft2. If the clearance testing is based solely on the composite sample, the home would pass
at the clearance level of 200 ng/ft2.  Yet, if the individual samples were used for clearance
testing, the home would fail based on the sample at 275 ng/ft2.

5.4.1  Construction of the Simulated Composite Samples
       Because most of the samples were collected during the first site visit, only the first site
visit individual samples were considered for the composite sample analysis. During the first site
visit, often more than four dust wipe samples were collected from a single component type
within a housing unit.  For example, in clearance testing from the Maryland Department of the
Environment, the number of individual floor samples collected from each housing unit during the
first site visit ranged from one to fifteen. When there were four  or fewer individual samples from
a component type within a housing unit, the simulated composite sample included all samples.
Therefore, in the cases of 1 to 4 individual samples collected within a housing unit for a
                                           40
-------
component type, these 1,2,3, or 4 individual samples were considered as one single simulated
composite sample. Even when there was only one individual sample collected within a housing
unit for a component type, this individual sample was still considered as one simulated
composite sample in the analysis. When five or more individual samples from a certain
component type were collected within a housing unit, multiple simulated composite samples
were created. For example, consider five floor samples, 100,150,175,150, and 250 ug/ft2, that
were collected in a home. Two composite samples could be created, one with four samples and
one with one sample. For the five samples, five different combinations could be used to create
the two composite samples:
EXAMPLE 1:
       Combination
          1
          2
          3
          4
          5
Composite Sample 1
(Individual
Sample Loadings)
100, 150, 175, 150
150, 175, 150, 250
175, 150, 250, 100
150, 250, 100, 150
250, 100, 150, 175
Composite Sample 1
(Average Loading
fqr^Composite)
    143.75
    181.25
    168.75
    162.50
    168.75
Composite Sample 2
(Loading for
Individual Sample)
      250
      100
      150
      175
      150
Therefore, 2 simulated composite samples could be constructed in the cases of 5,6,7, or 8
individual samples which were collected within a housing unit for a component type: one with
four individual samples and the other one with 1,2, 3, or 4 individual samples, respectively. To
construct 3 simulated composite samples, 9 to 12 individual samples were needed. In the cases
of 9 individual samples collected within a housing unit for a component type, 3 simulated
composite samples would be constructed: two simulated composite samples each would have 4
individual samples and the third simulated composite sample would have only one individual
sample. As many as 630 different combinations could be constructed from the 3 simulated
composite samples for the cases of 9 individual samples collected within a housing unit for a
component type.
       The different combinations or multiple composite samples that are possible add another
level of complexity to the analysis of composite samples which must be addressed. Each
combination can result in a different outcome (pass/fail).
                                          41
-------
 5.4.2  Composite Sample Clearance Criteria
       Since composite samples can pass the HUD Interim Guidelines Clearance Standards
 when one or more individual samples within the composite fail the standards, it is important to
 compare and assess the failure rate of composite samples under different clearance criteria.
 Three criteria were evaluated in this report and are as follows:

       Criterion 1:  Standard Rule
       The first criterion states that a composite sample will pass clearance only if its lead
 loading is less than the corresponding clearance standard for individual samples. This criterion,
 referred to as the "Standard Rule," implies that the average lead-loading for all subsamples
 included in the composite sample must be lower than the individual sample clearance standard
 for the given component in order to pass clearance. For instance, in Example 1 above, only
 Composite Sample 2 in Combination  1 will fail the home.  All other composite samples would
be below 200 ug/ft2, passing the clearance standard under this criterion. Because this criterion
only considers the overall result of the composite sample, it is liberal in that it will often allow
 individual locations to pass clearance  testing, even though certain subsamples may have loadings
 that exceed the standards.
       Criterion 2:  Standard/n Rule
       The second criterion allows a composite sample to pass clearance only if the lead loading
result is less than the corresponding standard for individual samples divided by the number of
subsamples in the composite sample.  Thus, assuming all subsamples represent equal surface
areas, the total amount of lead in all subsamples must be lower than the individual sample
clearance standard. For example, 200 ug/ft2^ = 50 fig/ft2.  For instance, using Example 1, the
home would fail floor clearance testing because of both Composite Samples in Combination 1,
and because of Composite Sample 1 for all the rest of the combinations. This criterion, referred
to as the "(Standard/n) Rule", is conservative in that a composite sample result can fail the
criterion even though all subsamples may have had lead loadings that were below the standards
for individual samples.  The "(Standard/n) Rule" will never pass a composite sample result
whenever at least one single subsample has a lead loading which exceeds the individual sample
standard.
                                          42
-------
       Criterion 3: 2xStandard/n Rule
       The third criterion is a compromise between the liberal "Standard Rule" and the
conservative "(Standard/n) Rule."  This criterion states that if a composite sample consists of n
subsamples (where 2^n^4), its lead loading must be less than the corresponding individual
sample clearance criterion multiplied by 2/n. For example, 2 x 200 ug/rW4 = 100 ug/ft2. In
Example 1, the home fails clearance testing in all cases. This criterion, referred to as the
"(2 x Standard/n) Rule", will result hi fewer false clearance passes than the "Standard Rule," and
fewer false clearance failures than the "(Standard/n) Rule."
       To assess the effect of the different composite combinations on the pass/fail outcome for
a home, the probability of a home passing clearance was calculated under the different criteria,
considering all the possible combinations of composite samples. Note that the compositing was
done separately for each component type since actual composite samples would not contain
subsamples from different component types. Also, for calculating the simulated composite
samples an assumption was made that all dust samples from a given component type within a
home were collected from equal sized surface areas.
       For various examples of hypothetical data for individual samples, Table 5-2 presents the
pass probability associated with each of the three composite sample clearance criteria when
considering all possible ways to construct multiple simulated composite samples.

5.4.3  Methods for Summarizing the Simulated Composite Sample Results
       The statistical methodologies discussed in this subsection were used to summarize the
simulated composite sample results.  These methods, frequency table performance characteristics
and logistic regression, were applied separately to each combination of component type and
composite clearance criteria.
       For each combination of component type and clearance criterion, the simulated composite
sample results within each housing unit were categorized into three groups: those that passed
clearance (Pr(Pass)=l), those that failed clearance (Pr(Pass)=0), and those that were inconclusive
(0
-------
 «
 "a.
 5

 i

 o

 v
 v>
 a
 00


 I
 •c

 «2
 O 10

 i°
 V CO

1 «
 E .o
 to "^
(A Q>

5 I
 w a.
 o >
o E
o £
9 *•
9 "O
55
« ^
ui "a
o> E
5  °»
To (0
o

£  o
so


1'
o  *•
t.  a
Q-  3
t- (0
in

9
:. i *p --»
'."•.;2-;.(ft -*S
wo S
•*T" f
we*
•'..'I-3 |
' (0 i1^ tfl
•5 3 "»
'••.8.8-8
Q.- o c
-•^s = <5
"'!•£:'* •"§
"Ill
ct ,g 35
"
1.
a .
•9 9
. g|
W
X
N
|
J|
••!•*
.*
••g
ig _g
5>
J'jj
»J jl
3'^^W
Z :£r£ :.
*^ , :;'S H- J
*•.._ >: *•>. -'I
„ - I- "•>-;. •''•'".•




•§ 's V - :.
II : -.^
-»% ^'5 L"g
£ g :^
" J !o '•'•'
— *- •*•$. Q ^ —
• ' " ' *
(


• . . .

••• if li
i S W m W
c ^ **""*" Si
— • v . .
. "li
I*.


*~




CM
O

_


in










j5
*~
o"

o
in
o
in
10


^.
d




o


,-


in








o
CM
d
in
d
0
^
in

0
tn
o"
m



O




O


d


in








o
o
CO
o
in
CM
in
£
d
in


8

o"
co
M
*•*
i


o




o


CO
(D
d


in
co







o
O
CO
d
in
CM
d
O
CM
d
in

d
o

in
PX
in
r-


0




o


CO
d


o



o
o
co
%
0
in
CM
8
CM
d
in
*~
8

o"
o
T™
in
fs.
in
^

O
O
CM
en
o
O
u.

CO
d




o


00
d


in









o
o
(D
O*
o
CO
o"
o
CM

d
o

d
o



O




o


_


tn









O
in
O
in
^
O*
^

o
o

o
o
CO


o




o


CO
d


in








8
s
d
in
CO
d
in
CO

0
0
CO

d
in
CM
N
*-<
S—
s


O




O


o


in








o
o
00
d
o
(D
g
in



0
in

8



o




o


d


m
CO





o
o
o>
d
o
00
d"
0
CM
o"
in

d
o


o
o

o
o


O
O
in
o »
TJ =
.£
-------
composite sample = 250 ug/ft2), the other 4 combinations of simulated composite samples passed
clearance. This resulted in inconclusive determination of failing or passing clearance for this
house, based on the construction of multiple simulated composite samples. Inconclusive results
were only possible for those housing units which contained more than four individual samples;
the many possible ways of combining five or more individual sample lead-loadings into multiple
simulated composite samples had the potential of resulting in a variety of different outcomes
under each of the three composite clearance criteria, and therefore, "inconclusive" results.

Performance Characteristics
       Frequency tables were generated to compare the clearance testing outcome within a
residential unit based on simulated composite clearance results versus individual sample
clearance results. An example of such a table is found in Table 5-3. Consider the cell labeled
'b'. The value of *b' indicates the number of units that would pass clearance based on analysis of
simulated composite samples, but would fail clearance had the results for individual subsamples
been considered.
Table 5-3.   Table Summarizing Pass/Fail Conclusions Made from Individual Sample
             Clearance Results Compared to the Simulated Composite Sample Results.
« \ "C />h **„ '
"^ ^£ * (
«T - t- ~' v. _ „
vsa -- „ " * * „,
•f.
Simulated
Composite
Samples
Pass
Inconclusive
Fail
Individual Samples" >
_"*•*• .. *
" . Pass '
a
c
e
-: -: Fail" "
b
d
f
       A quantifiable method for evaluating the comparison of the individual sample results to
the simulated composite results is to assess the performance characteristics of sensitivity,
specificity, positive predictive value (PPV), and negative predictive value (NPV). Using the
structure of Table 5-3, the performance characteristics are calculated as follows:
                                           45
-------
       Sensitivity   =
       Specificity
       PPV
       NPV
Probability of a dwelling unit failing clearance testing based on
simulated composite samples given that the unit would have failed
clearance testing based on the results of individual samples.
Sensitivity is estimated by f/(b+d+f).

Probability of a dwelling unit passing clearance testing based on
simulated composite samples given that the unit would have passed
clearance testing based on the results of individual samples.
Specificity is estimated by a/(a+c+e).

Probability of a dwelling unit failing clearance testing based on
individual samples given that the unit would have failed clearance
testing based on the results of composite samples. PPV is estimated
by f/(e+f).

Probability of a dwelling unit passing clearance testing based on
individual samples given that the unit would have passed clearance
testing based on the results of composite samples. NPV is estimated
by a/(a+b).
Contrary to the classical epidemiological use of these measures, the performance characteristics

do not include any measures of actual lead exposure. Rather they compare the performance of

each composite sample criterion to the clearance decision based on individual sample clearance

testing results.

       The performance of the simulated composite samples was made using the

epidemiological terms of sensitivity, specificity, NPV and PPV because an appropriate

composite sample clearance criterion could then be selected based on maximizing the

epidemiological performance characteristics. Such a composite criterion would result in less

costly clearance testing and clearance decisions, while making decisions similar to those based

on individual samples.  On the other hand, certain performance statistics may be of higher

importance than others. For example, to protect children from potential lead exposure, one may

decide to maximize the sensitivity of the composite clearance criteria while sacrificing higher

specificity.


Logistic Regression

       All three composite clearance criteria have different specificity and sensitivity error rates.

These rates correspond to the consistency between clearance decisions and the true lead hazards
                                           46
-------
present in the various locations included as part of the composite sampling scheme. Since
clearance testing is a procedure that is used to ensure that unprotected workers and residents are
not exposed to lead contaminated media following an environmental intervention, the
performance of each composite clearance criterion was characterized based on the maximum lead
hazard that is likely to be left behind. This was done by estimating pass probabilities as a
function of the maximum lead loading result.
       Logistic regression models were developed that estimated the probability of passing
clearance within a housing unit (as estimated by the simulated composite sample results) based
on the maximum individual sample lead-loading result. This model appears as follows:
where
           logit (  ) is the logit function, logit (P) = log (P/( 1 -P))
           TCiJk     is the estimated probability of clearance for component(/) in house(z) under
                  composite criterion (A),
           Max,-,   is the maximum individual sample lead loading result in house(i) for
                  component(/),
           Po      is the intercept parameter, and
           Pi      is the slope parameter.

       Tables were generated that give the estimates of P0 and P,, and associated standard errors
from the logistic regression models, as well as estimates of the probability of passing clearance
(using composite samples) when the maximum lead loading among all locations included in the
composite sampling scheme is equal to !/2,1,2, and 4 times the associated HUD Interim
Guidelines Clearance  Standard for individual samples. Figures were also produced that illustrate
the relationship between the probability of passing clearance (using composite samples) and the
maximum lead loading among all locations included in the simulated composite samples.
                                           47
-------
 6.0   FINDINGS AND RESULTS
       Based on the analysis methods outlined in Section 5.0, data from seven data sources were
 analyzed and discussed in this chapter. For the HUD Grantee data, two analyses were
 performed: one for a group of combined grantees that used the HUD Interim Guidelines
 clearance standards (i.e., 200,500, and 800 \ig/fi2 for floors, window sills, and window troughs,
 respectively); the other for a group of combined grantees that used a lower floor dust-lead
 clearance standard. The first group includes Alameda County, Baltimore, Boston, California,
 Massachusetts, Milwaukee, Rhode Island, Vermont, and Wisconsin. The second group includes
 Cleveland, Chicago, New Jersey, and New York City, which use 100 p-g/ft2 as floor dust-lead
 clearance standard, and Minnesota which uses 80 ^g/ft2.  The first group is labeled as "HUD
 Grantee (High)"; the second group is labeled as "HUD Grantee (Low)" in the analysis summary
 tables in this chapter.  Altogether, analysis results from eight data groups are discussed: data
 from Maryland, HUD FHA, HUD PHA, HUD Grantee (High), HUD Grantee (Low), Atlantic
 City, Cleveland, and Dover.
       The range of analysis results across data sources are summarized and discussed in the
 following four subsections. The analysis results on the individual data sources are provided in
Appendices A through G.
       Unless otherwise noted, the clearance standards used for the analyses in this report are the
numerical standards in the HUD Interim Guidelines clearance standards. That is, clearance
standards are 200, 500, and 800 ug/ft2 for floors, window sills, and window troughs, respectively.

 6.1    OBJECTIVE 1: CHARACTERIZATION OF THE NUMBER OF  INDIVIDUAL SAMPLES.
       WORK AREAS. AND HOUSING UNITS THAT PASS OR FAIL CLEARANCE TESTING
       STANDARDS
       Tables 6-la to 6-le present the number of samples, rooms, and housing units in which the
clearance criteria for individual samples were met ("Pass") or were not met ("Fail") for a given
site visit and component type. Table 6-lf presents similar results for all site visits combined.  In
addition, the number of samples, rooms, and housing units sampled are provided in the "Total"
column. Tables 6-lg to 6-li provide similar results as Table 6-la for the first site visit data only,
except that while in Table 6-la, pass/fail was based on  HUD Interim Guidelines clearance
standards, in Tables 6-lg, 6-lh, and 6-li, pass/fail was  based on various floor, window sill, and
window trough clearance standards.

                                         48
-------
       As seen in Table 6-la, for the first site visit and all components tested, across eight data
sources, 2.7% to 20.8% of the individual dust samples, 3.6% to 31.7% of the rooms, and 13.3%
to 65.1% of the housing units failed the clearance standards. Note that if an individual sample
did not meet the clearance standard, then the room and housing unit did not meet the standard.
Yet, even if the individual sample passed clearance, the room or housing unit associated with the
individual sample did not necessarily pass clearance. There were 35,800 dust wipe clearance
samples collected in 4,487 housing units across eight data sources for the first site visit. On
average, there were about 7.97 dust wipe clearance samples from each housing unit collected
immediately after intervention.
       The failure rate for individual samples during the first site visit for floor samples ranges
from 2.1% to 18.8%, from 1.1% to 38.9% for window sill samples, and from 1.9% to 36.4% for '
window trough samples. For the first site visit, the failure rates for housing units based on the
results solely from individual components were as follows: 3.4% to 56.5% of the housing units
would have failed based on the results of window trough samples, 3.0% to 41.7% of the housing
units would have failed based on the results of window sill samples, and 7.3% to 46.2% of the
housing units would have failed based on the results of floor samples.
       For houses that did not pass the clearance standard the first time, failure rates were
higher for the subsequent site visits in some cases. This can be seen in Tables 6-lb to 6-le,
which present clearance testing results for the second, third, fourth, and fifth site visits,
respectively.  Table 6-1 f presents clearance testing results for all site visits data combined. It
shows that, for all components, across the eight available data sources, 2.7% to 26.1% of wipe
dust-lead loading clearance samples failed the clearance standards, which resulted in 3.4% to
25.5% of the houses failing clearance.
       As expected, Tables 6-lg to 6-li show that the failure rate increased as the clearance
standard was set lower. The failure rates for housing units were between 17.5% and 72.4%,
29.2% and 81.5%, and 36.2% and 87.4%, based on the floor standards at 100 ug/ft2, 50 ug/ft2,
and 40 ug/ft2, respectively. Based on window sill standards at 500 ug/ft2, 250 jig/ft2, and 125
jig/ft2, the failure rates for housing units were between 3% and 41.1%, 12.4% and 83.3%, and
24.9% and 91.7%, respectively. Based on window trough standards at 800 fig/ft2, the failure
rates for housing unit were between 3.4% and 56.57%; at 400 ug/ft2, they were between 5.5%
and 69.4%.
                                          49
-------
       The results of the logistic regression analyses for determining the probability of a housing
unit failing clearance testing based on the number of samples collected for the housing unit are
presented in Tables 6-2 through 6-5. In Table 6-2, many slope estimates for individual
components and for all components together are positive and statistically significant. Several
slope estimates are negative; however, they are not statistically significant.  The positive slopes
indicate an increase in the probability of the home failing clearance as the number of samples
increases. The second half (right-hand side) of Table 6-2 shows that when two samples were
collected, the estimated probability of failing clearance ranges from 0.04 to 0.49 for samples
collected from floors, from 0.03 to 1.00 for window  sill samples, from 0.03 to 0.41 for window
trough samples, and from 0.16 to 0.50 for all three components combined. When 6 samples were
collected, the estimated probability of a home failing clearance ranges from 0.11 to 0.44, from
0.04 to 1.00, from <0.01 to 0.71, and from 0.13 to 0.46 for samples collected from floors,
window sills, window troughs, and all components, respectively.
      Table 6-3 provides  similar  logistic regression results when considering the number of
samples in categories of 1 to 4 samples and 5 to 8 samples collected. Table 6-3 also shows that
many of the slope estimates are positive and significant. For three very negative slope estimates
that were significant (-24.044 and -23.998 for window sills in Atlantic City and Cleveland data,
and -21.247 for window troughs in Atlantic City data), almost all housing units in these cases
passed clearance, regardless of the number of clearance samples collected. All other negative
slope estimates are not significant. Within a housing unit when 1 to 4 samples are collected,
Table 6-3 shows that the estimated probability of failing clearance ranges from 0.05 to 0.42 for
samples collected from floors, from 0.03 to 0.42 for  window sill samples, from 0.03 to 0.52 for
window trough samples, and from 0.17 to 0.45 for all three components  combined.
        Tables 6-4 and 6-5 present the results of the logistic regressions run to predict the
proportion of samples within a housing unit expected to fail clearance testing based on the
number of samples collected. Table 6-4 shows that only the HUD Grantee Low data group
presents any case of a positive and statistically significant slope estimate. The proportion of
samples expected to fail clearance  increases as the number of samples collected increases.  All
other significant slope  estimates are negative, indicating in these cases that a greater proportion
of samples pass clearance when more clearance samples are collected. Similar results are seen in
Table 6-5: the HUD Grantee Low data group has the only case of a statistically significant
                                           50
-------
positive slope estimate when the number of samples is treated as a categorical variable. Note
that many slopes are negative in Table 6-5; the estimated proportions of samples failing
clearance testing tend to decrease as the number of samples increases in these cases. Overall, the
estimated proportion of samples failing clearance ranged from 0.01 to 0.19 for floors, from 0.01
to 0.39 for window sills, from 0.01 to 0.44 for window troughs, and from 0.05 to 0.24 for all
components when 1-4 samples were collected.  When 5-8 samples were collected, the estimated
proportions of samples failing clearance testing were ranging from 0.04 to 0.19 for floors, from
O.01 to 0.39 for window sills, from <0.01 to 0.32 for window troughs, and from 0.03 to 0.23 for
all components.
                                           51
-------
 (0
>
 o
 +*

 o
D
 O)

'to
 3
 O


•o

 a

 E
 o
 o
e
 a

 a
CO

~a

IB


I


 X

 CO
 10
 CO
 CO
 u
 (0

 i
 CO

£
 to

!'
f
8
£
•• :, ;


CO
0..
w
a
I;
••7*..;,

''.--
H-1'
ce
"5.
'CD
W
'
e •
2'.
S,
u

i
•1:
1

u.
1
75
?
S
1:
u. •
"^
^ '":
'•"a-.
• U.
?*•'<'
'•'.;-•:
w
f
•'?'
S
iB.
•s.
.#,

,. „
2

i
a.
: ''.S1". '
' 3.^
.CO -.
1':
.D











^
g
LL














CO
co
CM
CM

o
CM
in
CO
in
CO
CM

eri


CO
CM


CO
5
CM
O
in
p~
CM

,_
en



O)

o
in
tv


Maryland
S
CM
CO

CD
CO
PS
CM
CM
cn

O)
CO


pt


00
co
cn

00
00



CN
00

in
co
p-


u.
0
I
cn

to
CM

CO
CO
CO
O
cn

co


S


CO
00
in
in

CM
O



tn

O
in


Q.
O
X
CN
cn
o
CM
CN
in

P»
co
in
CM
in

CO


O)
in


CO
CO
O
PS.
o>
CO

r«
in



CO
CO

E

2
i
6 •-.
§ °>
—) ^
CM
O
q
-

CM
cn
O
Oi
CO
in
co

CO


in


o
CO
CM
CD
CO

CO
<*



CO
in

CO
CO
CO

4)
C

CD



00
CO

CO
in

£*
Atlantic Ci
CO
CO
co
CD
CO

2
CM
in

p>
CM


O
C\


CO
CO
m

PS
CM



o
CM

00
CO


1
ID
0>
O
LO
CO


^
o
CO
CM
CD

CM


CO


O
CO
CO
CO
CD

,_
CN



CO

0
CM
CO


CO
o
Q











55
g
•o
i













CM
CD
CN
CO
CN

CM
0
CO
CM

in
CD


O
in
CM


CM
CM
in
CD
CM

CO
o



in
CM

r"
CM
CM


1 Maryland
-------









•D
(It

C
O
u

a

3
u.
#


1
u.

8
05
Q.




§'

•i.
<0
2
a













f
i

1




















CO
CO
in
^
CO


cn
cn
rx
CO
CO
tn
co
in
CM


cn

in
in


in
00
cn


CO
o
CM



co
o

rx
m






•o
c
CO
(3

00
O
in
CD
in


CO
IX
0
t
^
CD
CO


O
CO

O
CO
CM


CM


*
CO
CO



CO

CO
CM






^
u.
Q
15
X

O
in



in
in
o
CO
in
CO
O
CM


rx

CO
CO


CO


cn
—



-

co
CO






I
CL.
O
I

0)
CO
CO
CO


in
CO
t
in
CO
in
00
CM

CO


CO
rx

CO
00
CO
CM


O
O
CO


CO
in



in

CM
00
CM



,


**
C
to
t5 •_
if

CO
CO
CO
^


0
CM
cn
co
co
CO
CO


00

CM
CM


CO
CO


CO
CO



oo

00
CM
CM





CD
B
2 .
CD •
Q 5

CO
r-
CM



tn
s
in
CM

CM


to

CO
CM


CO
in
CM


CO
CM



CO

CM
in
CM






*-<
CJ
o
c.

CM
CO
00


CM
CM
CM
CQ
co

tn


CM

rx
CO


CO
CO


^
in



CM

CO






•o
c
cu
0

CO

CO


tn
5
co
CM
0)
•-


in

CO
in
CM


in
CO
CM


cn
-



in

O
CO
CM







o
0










t


co
**
C
e

U




















CO
o
IX
r
CO


o
CO
CM
o
o
in
co
CO
cn


CO
00
co

CM
co
rx
CM


o
o
CM


rx
CM



cn

0)
co
CM
CO






c
ra
I

cn
t
^
in
CO


IX
CT>
CM
in
CM
CM
o
^
CO


a
CO

CO
cn


IX
IX
O
CM


00
o
CM



co
CO

I






5
O
I.

cn
cn
CO


CO
CO
CO
CD
CO

^


CO
in

*«•
CO


o
CO


^
in



CO

CO
CM






o.
Q
X

CM
CO
CM
CO
CM


CM
in
CO
o
CD
1

^


a>

§
in
cn


cn
cn
in


CO
in



O
CO

cn
CO
in





0)
0)
+•*
c
CD
CD •_
It

tn
CM
0
co
CM
CM


cn
CM
CM
CO
O>
fx
CM
CM
in
CO



CO
00

CQ


in
CO


cn
^j-



00
in
00

CN
CO
r.
CO





0)
4-»
CD
CD :
o 5
i 3

8
^.
CM


cn
CO
T—
CM
CM
CO
|x
00
co


?

s


ro
CO
oo


^
in



CO

oo






.*;
O
o
S

1
00
CO
in
cn
co


in
CO
CM
CD
CO
^
-


CO
CM

CO


0
cn
CM


cn
IX



oo
CM

rx
co
CM






•o
ID
CD
O

00
in
*••
CO
co


CM
rx
CO
o
CD
CO


in
CM

CD
IX
CO


CO
CO


^
CM



in
CM

o
co







o
O















o
•a
10

O (0
•o £
>- (0
CD CD
co tn

g *
I s
o O
•5 eo
ru ~~
\ "
0 0
o o
CN *-
•a -o
SCO
CO
3 3
CO 09
£ .C
in w
CD CD
* *
53
-------

 to
CO

•o

 o
 u
 o
w

 o
 c
 o


•o
 c
 ID




 i
DC

 D
"5.


 a
(0

 o

TJ
A

 W
*J


 (0
 a>

CC

 O)


"•P
 (O
 a>
 u
U
to

JO

a
 CO

^
0>
1
e
U)
0

•-••-.'
18
•a
s
u»
10
R
o
£
>



8
f
a
CO
e
0
I
o
...
1
£
£
»ft

"5
UL
1
|
1
i2
0s


•'•*
u.

-• ;
S.
a
$
£
=1,
u.
*


u.

s
£
1


o
CO
B
a















o
Z
















CO
in
in


in
CM
00
8
CM
0
^




CM


CO
IX
IX
O
CM

CO
O



CM
CM

in
co




1
a
5
O
rx
^
in


CD
CO
•*
CO
CD
CM
CO
in
CO




CO


5
CO
CM

q
in



CM
oo

CM
in




I
u.
O
D
I
fx
CM
^
<~


co
CM
CO
CO
O
CM




CD


CO
CO

in
CM



2

o




I
a
o
i
CO
CO
CO
CM
CO
**•


-
en
CO
CM
o
CO
CO
CD




in


§
in
00
o
CO

in
en



co
in

0
in
in

O1


c
§ j
2
O)
0


CD
CO
in
CO
CM

CO




en


co
CM
in
CM

CO



CO

CO
CM
CM

to


S.
Q ?
CD
o
in
CM


'
CN
co
^
CM
CO




CM


in
CM
CM

q
CO



CO

CO
CN



.**
u
o
c
J2
CO
^
CO
CM


CO
o
co
^
CO




CO


in
CO

rx
CO



CO

in




•o
c.
ID
0)
I
O














to
$
•§
5















CO
rx
^
J


-
rx
CO
|x
CO
CO
CM




CO
CM


?
CO

CO
CM



in
CM

CM
CM




•o
c
(D
to
in
CM
5


CM
o
CO
CO
o
CD
in
CM




CO
CM


O
CO
-

CM
in
CM



co
CM

CO
00




u.
O
I
0)
o
0


o
en
o
O
O




O


o
O

o
O



o

o




I
fl
Q
I
CO
0
CM

CO


CO
0
03
in
IX
CM
CO
IX




O
CM


in
in
CM
00
CM

IX



o
CM

CD
CM

Q)


c
to
11
rx
00
o


CO
co
CO
c?
o
CO




00


CM
CO
CO

CO
CO



CO

co
co

0


(C
a:
0 §
iS
o
o
o


o
o
o
o
0




o


o
o

o
o



0

o




o
o
c
15
-
o
o


o
-
-
o
o




o


<-
1"-

0
o



o

-




T3
(D
ffi
I
o













.c
U)
i
H
o
I















in
en
o
CM


CM
CO
o
CO
CM
^
00




IX


CO
CM
to
rx
CN

in
rx



00

fx
CM
CM




•o
CO
I
CO
in
CM
in


CM
CO
CO
CM
CO
CO
CD
£




O
00


00
CO
00
CD

CO
rx



o
CO

CO
00




u.
Q
I
in
0
0
't


CM
CO
IX
CD
CO
CM




CM


in
IX

co
CO
CM



CM

in




0.
a
i
in
in
rx



CM
co
1
O
CO




CM
V*


CO
CO
CM
O
CN

CT>
in



CM

!

©


c
(Q
li
                                                             54
-------











T3
jjj
C

0
o
.a

CD
.O
'














»
a
CB
CO
1
'o
-.'



"'
a.
i
CO
w
e
§
cc



...
T;
••8
•a

at
S
c
i
1
o

I
•f'
£
?fi

1
'X : •
.0)
"5
2
£
.3
1

- '*•

1,


1

1
1
£
«R



1
• .
to
8


o
CO
o
CM
ro
in
00


-

in
r-.
00
O

i-.
CO



oo


o
en

to
0
r-.
CO




00

O
CO

ID
0)

to
o "5
15
-
CO
CM


CO

00
£

CO
£



CO


2

r*.
co
£




CO

=


^*

O
.y
c
CM
O
O


O

CM
CM

O
O



O


CN

CM
O
O




0

CM




[[Cleveland










i '•



CO
CD
o
b

o















•
*•
CM


CO
CO

CM
CO
CM
CM

00
£



in
fx


CO

CO
co
^
J




in

00
in




I Maryland
in
0)

CM
CO


in

CO
CO
CO
CO

••-
5



o
CD


CO
O
CM

CO
in
o
r-
co




o
O)

CO
CM
CO




U-
0
D
in
CO
CO
*


in

o
CO
CO
in

co
en



T—


in

o
^
CM




in

in
in




1
0
I
00
CO
CO


00
CO

o
CM
o
00
00

r»
O)



in
00


m
0)

o>
O
CM
00




g

O
o



**
c
CD
li
00
CM
CM

r-


05
CO

o>
CO
en
CO

CM
-



't


o
in
co

3
_
o




in

O)
o>
CO

CO
CO

c
CO
Q "g
O
CM
O
in
CO


r-

CO
to

CO
c\
CO



m


co

O)
in
^^

 tQ    CO
 §
(5    (D
to
cp
-------
 CA

 O

•o
 (0
  %


 i
GC
 a

 CO
r°
I
 in
+*

 3
 co
 CD
cc

 O)

V
 CA
 w


 0>
 U

 CO

 CO
 0)
(O

 01

TO-
"S.
ta
o

V
•'(D
a
a
U)
J-
-I""
.
,.<
, r

a.
a
(O
CO

g
1
U
•-.-
,""*
:?|-
u.
,#
I
L- "^ -
•M'
t
gK
•*.
f
* ,

«B
Q.
1
S
3
£
eP


,v. '.
2

l;
"I'.'
§
a
.: ^ -0 ••-• '










o
EL
















o
o
d

o
o
co
CM
o
d


0


CO
CM
CO
01

o
o




0

CO
CM


•D
C
<0
I
in
CO
in

CO

oi
OI

-
CO
o>
CO
T""


-


CD
co
in
oo

en
01




^

S

(D
£
11

2.
0 5
O
O
d

o
o
0
o
d


o


o
o

o
o




o

o

2"
o
o
n
§
-
O
d

o
-
r-
o
d


o


*~
-

o
o




o

-


•o
ID
a>
J
O









•
(6
CM
oi
in

-
O
CO
CO
co
in
in


CN


IX
CO

CO
m
in




CN

fx


u.
Q
D
I
O
0
d

o
o
0
o
d


o


o
o

o
o




0

0


I
a.
o
I
in
CO
in

CN
CO
co
^
in


CM


01
3

in
-t




01

0)

c
if
in
O
d

O
in
CO
o
d


O


CO
co

0
0




o

00
a>
a
c
a 5
S3
o
o
d

o
o
o
o
d


o


o
o

o
o




o

o

~
o
o
1
5
O
o
d

o
o
o
o
d


o


o
o

o
o




o

o


?
n

CO
0|


|X


0)
CM

CD
a>
CM




00

o>


•o
(Q
£•
Si
o
CO
o
d
CD

00
CN
o
co
co
CD
in


in


in
CO
CM
00

CO
in




5

ID
CO


U.
Q
I
CM
O
d

o
CM
01
O
d


0


CM
OI

O
O




0

OJ


a.
a
i
co
CO
-------








•o
a
c
1
0
o
u

to
J>
(0













••.$
2
a
o
§
o
.4


•d
o
"5.
i
to
»'
§
cc



-S4-
.^5
09
"5.
£
C».
earanc
O

.-~.
52"
'£:'
I
*
'."5"
u.
Hi-.
1
0
H. :
IL.i

U.


s
"«
a.
.i-«
1
of
3 -
iS
^<
. :••
£j
:•,'
s.
§..

' 3'. .
• '


o>
CO

o
00

in
CM

0

0



Maryland

in
in
00
(O
CO
5
00
in

in
in

00


r-

o
o
CM

O
in

CM
0

00
O)



X
U-
Q
I

in
o
o
to
CO
CM
,_

in

.


in

in

co
in

co

rv



I
CL
O
I

(O
CO
o
CM
CO
in
co

0)

2


o

S

in
CD

*

CO
CO
T-*


ID
c

O
1
+*

-
o
o
o
-


o
d

o


-

»-

o
o

o

-



1 Cleveland
                      o

                     .2


                      o
                     *^
                I   I
                      

                 £    =1

                 S   o
                "o   oo


                 %    o
o   o
o   o
CM   i-


 V
 in
                      09
                 C

                 (D
      c
      (D

     a
57
-------

 9
I

 O
 O)

'5

 o


•o

 CO
  ki


 O
 o
EC

 to

 a.


 a
CO

"5
 CO
 o
OC

 U>

*&
 m
 a>




 1

 I
 (0
 to
•B
t—
 i
(O
.a
 ra
-
",- . *'
w
•?"••>
'it
•a.
CO
3
O
.-
'*

'*
•b
a.
to
(O
E-
o

DC
V
|
v;
--
to
a

CO
•8
-•1
IB
xO,
U
v-
;>'
"5
0
. H
;i"3
: U.
''*
._
«•
"'8
• B.
1
J-,
'a
•S

* w^
t.
,
M*
CO
1^

.
s
O-
.'3
£
. J3
i£
^


: •- ••
: U.
'
J
0.
' :' ^ ^

'!:
Q




















§
UL


















T-

O
o


o




o
0




o



ro

CO

o
o




o

CO



•o
c
n)
ra

CD

in
CO


CO
in

in

r*.
CO
CM




^



,-

co

O
in
CM




"t

CM



U.
Q
D
I

CO

0
o


o


to

o
o




o



co

oo

0
o




o

00



Q.
o
I

en

0
0


o
0)

.

o
o




o



CD

o

0
0




o

o


0>
c
09
It

^t

o
o


0


r-

o
o




o



^

-

o
o




0



0>
CD
ra
Q §
S3


















M
CO
c


















»—

o
o


0


CO

o
o




0



CO

CO

o
o




o

co



•D
CD
CO
5

CO

^
CO
CD


CM


CM

CO
co
in




r»



in

CM

CO
CO
in




i-

in



LL
Q
I

0

O
O


o
o

o

o
o




o



o

0

o
o




o

o



Q.
a
D

CM

O
O


O
CM

(N

O
O




O



CM

«

O
O




o

CM


-------










•o
9
"3

O

T3

(9
£1
H










*
•

-
'"1
w
i
o
•
-
:^ -:.•;
1
.£*".
-iVy ,•
^i
"o
§*;-
--*•*
.r'-^-.r-
;i|Eji

:":•':"
•i
'-'-•«D
"to
8
1


3
*H
* r
'S
;?
u.
.£
'JU.
"(A
W
IB
'0.
;|
i
Ku.
. fi !-.- •
'.~-"o::
;i
-•-•<--•
-::*?
":-'*>•
'•^2'
••-.-..- :
1
•ii-
u.
rjp"
• - =,•

-1:
•;s
ft.
a -'
S -.
• '§•'•

a>
c
(D
If
CO



O
d

o
CO

-
0
d


o



-

-
o
o



0
^_


(D
c
a
ol>
O S
          o
          o
          (0
          •a


     8   I


    1
    ;   i
     i-   <0
     
     IB    in
     ^^    (D
     w   „

     S    5

          s
 a
 o
•j;


 S
          O
          oo
    1   s
    O    o
    O    O
    eg    ,-

    "S    ?
     0)    W
     (D    (0


     «"*    ^^

     10    CD


     
     *^    -^

     C    C

     (D    10
59
-------
 0)
tMMW


 O
*<
O)
u.
 o
 O)
 c
'«
 o
X
•o
 a

 o
 o
 a

 (0
w
To
S3
I
 O)
 CO
 o
 u
 n
 i5
 o>
 e
 •
(O
 tt
Z
 a
• •' t
?£
:i
'S.
w
a>
to
*
'.'••
:r
'A'
'ix
JO.
• «.
• o

3*
f-^S'*
- • .
;ri.
V-
•>:
"8
a
'I
S
2
ffl
O
o
•--;.
• "5 .
•1
*i
"'' "5
u.
;JA

•' 35
u.
; O.
'^
0
/ii
";£
"*.
.-
'••- '.-.














CM
O
O
in


*~
-


m
*


*~




to

t*-
CO
2



»—


(O

1
0
13
I









..


JI
§
J-
5
" O
TJ

















o
0


o
-


o
o


o




*

'
o
0



o


•>

I
U-
Q
D
I








-'



OJ
c
§
O
O
3















CM
O
C)
O


CM
O
O

O
O
CO


CO




f-

in
o
O
CM



co


CM

£
o

I
                                                                              10
                                                                             •a
                                                                         o    
-------
 V)
|S

 o


5
 o
D
 (0

 o
•o
 a

 o
"S.
 E
 a
CO
 (0
*J
 v>
I

s
•+•
 v>
 CO
 CO
3
u
(O
.o
 a
''••"'
•g
Q.

' '-:;•.:.',
•-.«


(O
f
Q
CO
iearanc
u.

t
•!
SB

5K-
'! r!;';
I
• to
f
"3-
1
f
^s


'•'•:''
u.

• '-*.
;•(;«:'
O.
t
1
i2<
*•
1

o>
I
'-•.;..•• T"
o
c/a
S:
Q















0













CM
CO
CD
O


O
CM
O
in
Cn
CO
co
CM
o>




CO



CD
in
in
CM
CO
CO
cn
CM

cn

CM

CM
CM


T3
C
(D
<6
in
q
^


CO
CM
r»
cn
CO
CD




CD



o
cn
o
CO
„.
CO
CM

O
ro

cn
cn


u_
0
I
o>
't
co


*
in
CM
cn
CM




CO



CO
00
CM
CO
CD
CO
CM

co
r-

in
in


n
0
D
I
CM
O
CM
CO
CM


CO
t
't
in
0
CM
CM
CM
CO
CO
in




o



CM
m
CO
r-.
CM
CO
CO
CO
o
CD

r-.
CO
in

in
CO
CO


c
£
§|
o
co
O

CM


in
CM
in
8
co
o
09
CO
5




CO
in



o
in
CO
CO
CM
cn
CO
in
••t

CO

cn
CO


c
(0
o 3
0)
in
in
2


CO
CM
CO
CO
cn
CO
m
CO




CO



CM
o
in
(O
at
in
co
CO

m

in
m

51
o
*
+•*
CO
CO
on
r-


CO
m
co
in
in
CM




«



CO
in
£
o
CO

CO
CM

in


•o
10
1
O
in
CO
r*


-
o
CO
CM
co
CM




CO



O
CO
CO
CO
CD

CM

CO

O
CM
CO








^
CO
CQ
CO
CO

^

-



Q)
O
O
                                                           61
-------










•o
§

*




O
CO




a>
Q)
c
CD
a |
CM
CM

CO
CO
-
00
CO
CO
CM
CO
CM




CO


O
CO
CM

CD
CM
CO
CO




cn
CD
CM




>
O
o
c
_co
*•*
CM

o
o
o
CM
cn
CO
O
O




o


cn
CO

'
cn
•*




CM
cn
CO





•o
i
CO
O
co

^j.
CO
in
#•»
CD
CM
cn
^l




.m


co
tn
CM

in
CD
CM
cn





in
o
co
CM






CO
o
Q












CO
s
o
a
•E
o
U
















1

«-
CO
CM
CO
CD
CO
in
CM
O
in
CO
i
CM
*~



tn
CM


r-
r-
o
CO

CO
cn
00
CM




T"*
O
o
in
cn
CD





•o
c
CO
co
5
cn

tn
in
CM
CO
CO
jr
CD
8
in
*—
^



in
CM


CO
cn

in
00
CM
^
CD
CM




in
cn
0
CM





u.
o
I
cn

^.
co
-
to
CM
00
cn
O




*


00

CO
CO

co




CD
cn
r-
co
CO





0.
Q
D
O
in
CM

^
*
O
O
o
m
o
CM
CM
0
CM
CM
P^




§
CO


CO
CO
CO
o

CO
CM
in
in




cn
Cs.
CO
CM
CD




CO
CO
c
CD
I!
00
CO
o

in
co
r-
co
cn
CD
co
cn
cn
CO




CO


m
cn
tn
tt

in
CO
CO
co
tn




1
o
CD
CM
r-




*
c

0





in
CD
CO
tn
CD




>-
O
c
JS
CO
CO

cn

CO
in
CO
CD
cn
0
CM




*


CM
cn

CM
CO
CO
CO




CD
CM
CO
co
CM





ID
0)
o
CO
in

CO
CO
CM
CO
s
CO
CO




tn
CM


cn
CO

CO
cn

CM




in
CM
cn
00






CO
o
a














I*
M
o
*~
CD
,_: c
O CD
T «
CD 

0 4-
o S;
C D>
CO 5.
CO
CD O
"o oo
8 °
8 8
CM «-
co e
at ca
co n
•£ •£
O CO
+* 4-r
C C
S 2
O O
*
* *
62
-------
 o
 **
 W
 «<
 «A
 il
 W
•o
 CD
•o
 a
**
(0
 a
 o
._
 a
  .
 
O

 O)
^
CO
':'i
:,

o

;.••'

"S
D.
W
F
0
S.

' ,

- '
.\.
"'.'-
0>


c
'Clearan


c





1
5
1
;1
8
£
I
?
. a
1


i


..
1
;•.'•'
"1
:£
.3
,£
-
1
8
£



i













«
S
o
<«•


tandard

c
E













i
cn
CO
tn
•
CO
in
CM
CD
00
in
CM
O
CM
CM
CM
CO

.-
in


00
o
in



5

rx


^
in
in

CD
CO
in
CO




I
U_
Q
I
CO
„
rx
00
0
in
0
CO

CO
rx
in


r—
00
CN



§
CM

CO
in
in

CO
tn

O
CM
CO
00
co
CM




Q.
o
I
CM
0)
O
CM
CO
5
o
CO
CM
in
CM
in
IX,

CM
CM


IX,
CM
00



in
in

CO
fx
T—
CO

^
CO
CM

CO
O)
co
CD
CM
(D

'±
a>
0)
c
<0
5
o
T.
CM
O
CM
CD
CO
O
IX
CO
in
CD
CD
tn
CO

ro
CO


cn
00
CO



in
00
CM

CM

^
o»

CO
o
CO
CO
CM
.
J
O
o
2
O
O
I
CD
to
CM
CM
IO
co
00
CD
CO
in

^
o
CO


fx
in
*—



CO
to
CO

10
to

CO
CO
CM

CD
r-
g
co




U
o

O
in
CO
CM
in
O)
CM
rx
CO
CM
CO

CO
CO
CM


00



fx

CO
CO
CO

IX
CO
CM

O
in
CO
00





o
o
o










*
S
o
if.

a
tandard
W
o
S
u.













CO
CO
0
CM
in
CO
rx
CO
CO
in
CO
CN

i-
co


in
CM
00



CO
00

g

CD
O
CO

CO
CO
0
CO




to
>
<0
in
fx
§
IX
00
CM
CM
CM
O

0
cn

r*
tn
03
rx
CO
CM
CM
O
CO

00
00


CD
CO
CM



in
CM

00
ID
in

CM
00

O)
CD
CN
CD
00
CM




0.
O
i
CM
CO
8

O)
CO
IX
CO
to
IX,
CN
CM
in

«-
O
CM


in



in
CM
o
CD

O
OO

^
CD

CO
00
in
(D
CD
10
CD
2
if


1
o
1.
CN
O
CM
O>
CM
CO
CO
CM
CO
CM
CD
in
CO

0
in


CO
tn



O
O
CO

CM
i

*-
in

3
in
CO
CD
O
CO
?
o
Q)

c
<0
CT
o
3
CO
in
CO
in
CM
rx
CO
CO
in

00
CM


00
CM



OO
oo
CO

in
in

10
CO
CM

O
CO
S
"4-




CJ
.9
C

00
CO
„
g
CO
CM
CM
rx
ID

<•
CO


in



CO
o

00
ID

CM
CO

S
s




10
CO
ID
0
ID

*
CM
CO
00
CO
CM
CO

rx
CD


o
V"



CO
r"
tn

CO
CO
CO

CO
CO

in
0
00
CM
in





0>
s










s.
o



•g
a
•ts
V)
o
1













5

CD
in

CM
rx
tn
0
§
CM
CM
cn

CO
CO
CO


fx,



to
CO

IX,
CO
CO

ID
CO
CO

CM
CO
CO
CO




u_
Q
D
O)
00
CO
in
S
in
in
o
CO

CO
CO
CM


^



CO
fx
CO

00
in
in

00
CO
CM

co
co
in
CM




Q_
O
D
CM
CO
S
fx
in
CM
rx
CO
in
in
in
in
CM
in
rx

in
^


cn
CO
00



CO
fx
CO
CO

CO
rx
oo

o
^

rx
CO
00
CM
00
CM
2
'£
09

1
(T
O
D
CM
O
in
IX.
CO
rx
CM
00
CD
IO
CO

o
CO


tn
03
CM



cn
IX,
CM
w

CM
CD
CO

O
00

co
CM
in
CO
CO
¥
o
-J
CO

c
CO
o
D
cn
in
fx
IX
CM
*
in
CO
in

*
CO


CO
-------

 o
 *J
 55

 2
 \L
 o
 eo
•o

 (0
•o

 a
4-1
V)

 cs
 U

 CO

 <0
 (0

 o
•E
 a
 en
 a
"5.
 E
 (0
 o
"O
 £
 0)
*tf


 0)
 a>
CC
 co

•2

 CD
 U

 CO

 CO
 OJ
(0

13
£
E
CO
o

1
9
1












M
£


in
cs
_
o
•D
(0
£5
I
I
















01
CO

•*
CD


00
Ol
CO
01
CO
V—
IX
CO
01
in
o!



o
00
in


03

in
CD
Ol

S


03
in

00







Maryland

CO
0|

«-
rx
in


01
IX
in

CO
CO
00
CO



CM


CO
in

03
CO
CO
O
Ol
CO


ft
CN

1







£
Q
y:

E

CD
03
CO


5
IX
CO

03
CO
,_
*



in


CD
CO
CO

5
CD
*


O
CO

in
CO







CL
Q
I

1

03
CN


CO
O
in
in
ot
in
O
03
in
01
in



CO
en
CO


03
00
CO

00
0>
IX
t
IX
T"


CO
o
IX

Ot
0>
S




O>
T
ID
C
10
Q
D
I

00
CO
03

CO
in
CM


in
CM
CO
CN
IX
CO
01
CM
rx
in



CO
CO


03
00
rx

IX
CN
rx
in


00
CO
CO

03
O
00






CO
-
Atlantic Ci

CO
CO

•tf
CO
co


CM
CN

CO
0)
„
O
01



03


rx

CO
03
^
0
CM


CO

S







Cleveland

Ol

rx
03


T*
-

00
f-
^
0>



T™"


-

00
^.
O3


|x

-







0
1
















8
CN
^J
a
CO
to
to
1
















CN
CO

00
CO
CO


§
in

IX
CO
CM
rx
co



IX
03
CO


|x
03
r~

IA
CD
CM
^
CO


0

i







Maryland

5

o
9


oo
in
00
CD

CO
CO
0
01



CD
tf
f»


00
in

00
CD
CD
Ol
01
01


00

o
01
in







u.
O
T.

Z

oo
03


01
01
03
00

03
CO

CO



S


IX
CO
CO

5
co
co


rx
CM

CO







a

CO
o
Ol

rx
in


CO
CO
Ol
rx
O
03
in
in
CO



03
00


I

00
0>
Ol
03


in
03
CO

CO
o




CD
I
CO
a>
c
ID
5
a
I

oo
CO
03

^f
OJ


§
CO
CO
CO
01
CN
co
CO



o


CO
oo
03

rx
CN
CD
CO


5

CO
8
CM




I

05
03
C
-
Atlantic Ci

CO
CO

CM
in


in
00
CM

CO
03
^
in



in


00
00

CO
03

in


in

00
00







Cleveland

CM

CO
CO
00


0
01

03
CO
CO
00



in


C*3

00
CO
CO
00


in

CO







g
Q












N
£
b
4
S
^
aj
T>
S
to
55
1
















01
CO

Ol
CO
01


CM
O
IX

IX
CO
01
in
o



o
in
01


CN
Ol

in
CO
ct
CO
o


in
CN

CM
Ol







Maryland

co
CM

in
CO
CO


*
O
00

CO
CO
co
CO



CO
00


CO
rx
in

00
CO
CO
in
CO


0
03

00
rx
in







u.
0
I

£

01
rx


00
CO
o

O3
CO
o
CM



00


CO
S

5
CN
CM


03

CM
§







D.
0

CO
O
CM

rx
00


IX
T—
in
oo
O
CO
in
CO
*



CD
CO


o>
CO

00
o>
IX

•*


rx
o>

*•*
O
CO
Tf




CD
^
CO
c
ID
c5
Q
X

00
CD
03

CO
rx


rx
rx
03
00
CO
01
IX
CO



CO
|X


•*
O
CM

IX
CN
rx
CO


03
rx

00
CD
O
01




1

CO
c
ID
Q
I

O
CN

0
CD
Ol


*
CO

in
00
03



in


co

in
00
03


in

CO






>-
Atlantic Ci

CO
CO

o
CO


>-
Ol
CO

CO
03

*-



-


CM
03

CO
03

»-


-

CN
CO







Cleveland

CM

(X
5


in
rx

00
03
CO
CO



rx


-

03
03
00
CO


^

'-







I
Q



















IM.
•S
re
3. w
n U
*^ ^
•B TO
10 0)
£ "

03 *;
O 5;
2 §
 -------
 w

 >

 V
 +*

 W
 **
 M

 il

 o>
 CO
•E
 (0
•o


w

 I

 I
 CO
 o
 CO



 a


%
 (0

"a

 CO
CO

£
 a
 5
 o
•c
 o
«*-
 n


 »
 ro
 a>


 a>
 u


 I
 CO
_»

U
(O
_
.Q
 a
-
T
Q.
09
.8
«9
0
" 'k .
Jv;
••<'-;
:-«.
Q.
08
w
. C
o
o
cc








E0
•i
I
to
•8
I
• a
.5?
O
>•••:

S
1
-*

I
8
£
1
"' TTt
s
a

CM


in
CD
in
CM


CM
8
CO



en
en
00
en
CM


0
|X
CM
i
CO
CO
4-1
c
a
Q
I
CO
00
IX
en


CM
O
en
CO
CO

CO
in



in
O
CM


CO


CD



CN
in
m
o
CM


5
?
_i
CO


c
to
5
a
i
en
en
O


CO
co
O
in
CM

en
in



in
*""


en
CO
CM


CO
in
CM



00
in
in
*~


co




O
c
to
<
CM
CD


*
O
CM
en
CO

CO
CM



in



CO


en
CO



00
CM
in



CO




leveland
o
CD
in
in


00
CO
CO
CO
CM

»_
CO



CO



CO
m
CM


in
CO
CM



O
CO
00



|X
in
CM




CO
§
O








M
1
O
o
on
o

k»
CO
1
CO
a

2
O

c
5












co
CO
in
CO


en
en
|X
co
co
•<*
in
00

in
CM



en
en
CO


m
m


in
00
en



CO
o
CM
CO
^


in




TO
c
JO
^
00
o
in
CO
10


CO
5
O

^
co
CO



o
CO
fH.


o
00
CN


CN



Tfr
CO
CO
co
^


co
CM




LL
a
i
o
in


in
in
O
CD
in
CO

O
CM



,,.



en
CO


CO



en





S
CO




I
a.
a
i
en
CO
en
00


in
in
en
in
CO
CM

,_
CD



CO
T""


CD
CO
CO
CM


CM
O
O
CO



00
in
in
^


CM
00
fM
D)
i
ID


§
O
I
CO
00
co


o
CM
en
co
co

00
00



CO
ir--


CM
CM


CD
CO



00
CO
oo
^


oo
CM
CM
1
ID
a>
4J
i
5
o
I
en
CM


in
s
CM

,4.
CM



CO



CD
CN


00
in
CM



CO
CM
CO



CM
in
CM




O
o
'*•!
C

CM
CO
CO


CN
CM
CM
en
00

,_
in



CM



IX
CO


en
CO




in
CM



IX
CO




eveland
O
CO
co


in
5
CO
CM

0)




in



CO
in
CM


in
CO
CM



en
*~
in



O
CO
CM




I
o
o
                                                                                                                       o
                                                                                                                       o
                                                                                                                      •g

                                                                                                                       CO

                                                                                                                  ^    c
                                                                                                                  o    **•
                                                                                                                  o    t!

                                                                                                                 "T    co
                                                                                                                  o    £
                                                                                                                 *•    £
                                                                                                                 -o    2
                                                                                                                 I    g


                                                                                                                  (0    in
                                                                                                                      o>
                                                                                                                  in    v>
                                                                                                                  3    3
                                                                                                                  *^    4^
                                                                                                                  

                                                                                                                  
-------
Table 6-2.    Parameter Estimates and Associated Standard Errors from a Logistic
             Regression Model of the Probability of a Residential Unit Failing Clearance
             Testing Based on the Number of Samples Collected.
*. ^ 1
•v J
• Data Source
*"4 "
**. ' "
J
f
Number of -
Residential
Units
>
| Parameter Estimates;
'TN IT ^ «
I . m*
Po
s.e. (Po)
/P.. :
s.e. (Pi)
Estimated Probability of Failing Clearance
Testing When the Number3of Samples
Collected Within a Residential Unit-is:
-•- t
m=2
*^
, T , \ t\
B.m=4 ' 1 ,
^"'if-fVir^
^*ni=6,=f
Floor
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee (Low) "•
Atlantic City
Cleveland
Dover
661
145
119
2092
1021
159
38
151
-2.0752
(0.2128)
-0.8300
(0.4811)
-1.8649
(0.5205)
-2.2242
(0.1331)
-2.7460
(0.2728)
-1 .6335
(0.4505)
0.4396
(1.2155)
-3.7027
(1.3086)
0.1725t
(0.0408)
0.1014
(0.0673)
0.1680
(0.0935)
0.1223t
(0.0276)
0.1753T
(0.0655)
0.0719
(0.1122)
-0.2387
(0.2891)
0.2692
(0.2872)
0.15
0.35
0.18
0.12
0.08
0.18
0.49
0.04
0.20
0.40
0.23
0.15
0.11
0.21
0.37
0.07
0.26
0.44
0.30
0.18
0.16
0.23
0.27
0.11
Window Sill
Maryland
HUD FHA
HUD PHA
HUD Grantee {High}*
HUD Grantee (Low)*»
Atlantic City
Cleveland
Dover
612
126
111
2031
968
20
33
12
-1.1974
(0.2145)
-1.5216
(0.4949)
-2.2334
(0.9141)
-2.9473
(0.2002)
-4.0901
(0.3921)
-3.3278
(1.6858)
-3.8361
(2.7169)
-28.508
(0.8018)
0.0001
(0.0477)
0.1787t
(0.0829)
-0.0892
(0.2381)
0.2431 1
(0.0718)
0.6520T
(0.1471)
0.6941
(0.5227)
0.1274
(0.8415)
27.2553T
(0.0000)
0.23
0.24
0.08
0.08
0.06
0.13
0.03
1.00
0.23
0.31
0.07
0.12
0.19
0.37
0.03
1.00
0.23
0.39
0.06
0.18
0.46
0.70
0.04
1.00
                                         66
-------
Table 6-2. (Continued)
~ Data Source
Number of
.Residential
Units
Parameter Estimates

PoV
s.e. {Pol
', - 	
-: -••&. ;-::•::
s.e;:(3i)":
Estimated Probability of Failing Clearance
Testing When the Number of Samples
'Collected Within "a Residential Unft:is
£•>/ n&2±
-------
Table 6-3.    Categorization of the Number of Parameter Estimates and Associated
             Standard Errors from a Logistic Regression Model of the Probability of a
             Residential Unit Failing Clearance Testing Based on the Number of Samples
             Collected.
<•-
Data Source
•Number of
Residential
< Units
Parameter Estimates ,
-; • -- " , v*
- <• ••) > •!,,*,'
Po'
s.e. (Po> '
3,
r S:e.(Pt|
' Estimated Probability,of , Failing „ '
Clearance Testing Wheri the Number
' of Samples Collected Within aj|
-- • Residential Unites Between P-?
- *• » f
,,;, 'iSmS4
5*mS8 /,;
Floor
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee {Low)**
Atlantic City
Cleveland
Dover
661
145
119
2092
1021
159
38
151
-2.2958
(0.2642)
-1.1854
<0.5953)
-1.6011
(0.5853)
-2.4331
(0.1690)
-2.6618
(0.2904)
-1.6540
(0.5459)
0.3245
(0.9610)
-4.3550
(0.9710)
0.6575T
(0.1587)
0.4900
(0.2702)
0.3679
(0.3539)
0.4935t
10.1057)
0.43771
(0.2042)
0.2144
(0.3911)
-0.6315
(0.6719)
1.3932T
(0.6457)
0.16
0.33
0.23
0.13
0.1
0.19
0.42
0.05
0.27
0.45
0.30
0.19
0.14
0.23
0.28
0.17
Window Sill
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee (Low!**
Atlantic City
Cleveland
Dover
612
126
111
2031
968
20
33
12
-0.8946
(0.2680)
-1.7193
(0.6237)
-2.2893
(1.1202)
-3.3281
(0.3425)
-4.4770
(0.8672)
22.7218
(0.5627)
20.6307
(1.0171)
-0.3365
(0.5855)
-0.2257
(0.1893)
0.66341
(0.3337)
-0.2094
(0.8448)
0.9344t
(0.3130)
1.9016T
(0.8404)
-24.044 1
(0.0000)
-23.998T
(0.0000)
0.0000
(0.0000)
0.25
0.26
0.08
0.08
0.07
0.21
0.03
0.42
0.21
0.40
0.06
0.19
0.34
<0.01
<0.01
0.42
                                        68
-------
Table 6-3.  (Continued)
Data Source "V
r
* * -
'Number, of
Residential
- Units
-I »- ?-3st '*' '
- , "• Parameter Estimates^*;/-''
. i/t • ' ,y$J3fcr~ ;,'•••
IV '.
s.e. (Po)
Pi'v'X-
s.e. (Pi) I"'_i :
Estimated Probability of Failing
Clearance Testing -When the" Number
of Sample? Collected Withirra
. Residential Unit is Between ;
" 	 	 ,\ yH -:§i-
1ini*4^''.'^ii:
'•.'{l;^5friii8 ' ' •
Winrinw Trntinh
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover
536
108
110
1697
873
119
24
146
-1.0503
(0.2259)
-0.3354
(0.5799)
-4.8865
(1.5555)
-2.8219
(0.6447)
-3.3893
(1.2431)
18.1294
(0.4570)
-2.3979
(0.7385)
-3.3393
(0.4551)
- All Con
Maryland
HUD FHA
HUD PHA
HUD Grantee {High}*
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover
706
149
119
2132
1025
160
38
158
-1.1864
(0.2021)
-0.8280
(0.4665)
-1.5737
(0.5485)
-1.4722
(0.1527)
-1.5823
(0.2778)
-2.0008
(0.5285)
-0.0544
(1.1759)
-1.2594
(1.3119)
0.3923t
(0.1546)
0.4150
(0.3825)
1.2755
(0.9376)
0.4880
(0.6272)
1.3481
(1.2293)
-21.247
(0.0000)
0.0000
(0.0000)
0.0000
(0.0000)
0.34
0.52
0.03
0.09
0.11
0.04
0.08
0.03
0.43
0.62
0.09
0.14
0.33
<0.01
0.08
0.03
innnente
0.3145t
(0.0641)
0.4263T
(0.1304)
0.2486
(0.1521)
0.1521T
(0.0618)
0.1561
(0.1232)
0.5086
(0.2818)
-0.1529
(0.4638)
-0.3203
(0.6757)
0.29
0.40
0.21
0.21
0.19
0.18
0.45
0.17
0.36
0.51
0.25
0.24
0.22
0.27
0.41
0.13
*  Grantees that used 200 jug/ft2 as clearance standard for floor.
** Grantees that used 100  //g/ft2 or 80 ;/g/ft2 as clearance standard for floor.
t  Slope estimate is statistically significantly different from zero at the 0.05 level.
                                              69
-------
Table 6-4.    Parameter Estimates and Associated Standard Errors from a Logistic
             Regression Model of the Proportion of Samples within a Residential Unit
             Expected to Fail Clearance Testing Based on the Number of Samples
             Collected.
" -* " -
»V J.
Data Source
rf >. *
Number of
Residential
Units
* «, »
^ ^ * ,
-Parameter Estimates ~-
-3 ^ " —
Po' -
s.e. (Po)
- Pt "
s.e. (P,"| *
' Flnnr
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee
(Low)'*
Atlantic City
Cleveland
Dover
661
145
119
2092
1021
159
38
151
-2.2355
(0.1647)
-1.1178
(0.2601)
-2.4463
(0.4158)
-2.5207
(0.1168)
-2.9566
(0.2529)
-1.8520
(0.4094)
-0.2857
(0.9032)
-4.0533
(1.2004)
Win
Maryland
HUD FHA
HUD PHA
HUD Grantee (High) *
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover
612
126
111
2031
968
20
33
12
-1.2893
(0.17001
-2.2012
(0.3326)
-2.3940
(0.8824)
-2.9782
(0.1933)
-3.5886
(0.1592)
-2.1174
(1.3324)
-3.6164
(2.8720)
-1.5343
(1.03791
-0.0133
(0.0282)
-0.0456
(0.0331)
0.0466
(0.0659)
-0.0572T
(0.0222)
-0.0360
(0.0579)
-0.1886
(0.0996)
-0.3840
(0.2144)
0.0425
(0.2617)
Estimated Proportion of Samples Failing .
Clearance -Testing When the«Number of
Samples Collected "Within"^ Residential c
' Unitis'" "^.-" v
<i=4;/;
•Jtf-t-K, "
,-*?& .

0.09
0.23
0.09
0.07
0.05
0.10
0.26
0.02
0.09
0.21
0.09
0.06
0.04
0.07
0.14
0.02
0.09
0.20
0.10
0.05
0.04
0.05
0.07
0.02
rlnw Sill ' - . ' - '
-0.1854t
(0.0353)
0.0529
(0.0477)
-0.3540
(0.2280)
-0.0635
(0.0668)
0.1 141 1
(0.0348)
-0.0337
10.4149)
-0.2851
(0.8954)
0.4964
(0.4056)
0.16
0.11
0.04
0.04
0.03
0.10
0.01
0.37
0.12
0.12
0.02
0.04
0.04
0.10
0.01
0.61
0.08
0.13
0.01
0.03
0.05
0.09
< 0.01
0.81
                                         70
-------
Table 6-4.  Continued
* > W . ™
•Data Source

Maryland
HUD FHA
HUD PHA
HUD Grantee (High)»
HUD Grantee
{Low)*»
Atlantic City
Cleveland
Dover
Number of '
Residential _
- Units ,
«. * * ,- ', ~* *
V S»
Parameter Estimates •
. H*"- ,. > - , > »< »
* ' R ** "
.Po -
s.e. (Pol
..-, 3, : •
s.e. fa
.Estimated Proportion of Samples Failing .
Clearance Testing When the Number of
•:. . . . .i.f.r.,Ld.,:-:.v. »-4'.n.-.,-', -..-MI',.-:..W'-.'''-' ••
Samples Collected Within a. Residential •.
1 ; y;: ;:V^iyntill^i?:s^.-pi5'
hi =2 •'.•-.'•
?::;;f;:^?lS:f?rf
X.,.;~™Sj*J:C:?;
Winrinw Trrmnh
536
108
110
1697
873
119
24
146
-1.1680
10.1059)
0.0062
(0.3009)
-4.1917
(1.2484)
-2.2143
(0.1862)
-2.4714
(0.2308)
-4.6205
{1.2611)
-1.5352
(1.6569)
-1.0855
{1.5360)
All C.n
Maryland
HUD FHA
HUD PHA
HUD Grantee {High}*
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover
706
149
119
2132
1025
160
38
158
-1.6733
(0.0846)
-1.4924
(0.1596)
-2.2296
(0.3438)
-2.4963
(0.1002)
-3.1217
(0.1531)
-2.5903
(0.4176)
-1.1633
(0.7233)
-1.1184
(0.8542)
-0.0360 t
{0.01811
-0.1154T
(0.0588)
0.0588
{0.2870)
-0.2745 1
(0.0865)
0.0679
(0.1093)
0.3490
(0.4482)
-0.7794
{0.97691
-1.6073
(0.9097)
0.22
0.44
0.02
0.06
0.09
0.02
0.04
0.01
0.21
0.39
0.02
0.04
0.10
0.04
0.01
< 0.01
Xi!.6';.;;.i

0.20
0.33
0.02
0.02
0.11
0.07
< 0.01
< 0.01
mpnnpnts
-0.01 97t
(0.0060)
0.0091
(0.0086)
-0.0469
(0.0251)
-0.0457 1
(0.0111)
0.0214
(0.0182)
-0.0356
(0.0641)
-0.1595
(0.0896)
-0.4140t
(0.1467)
0.15
0.19
0.09
0.07
0.04
0.07
0.19
0.12
0.15
0.19
0.08
0.06
0.05
0.06
0.14
0.06
0.14
0.19
0.08
0.06
0.05
0.06
0.11
0.03
*  Grantees that used 200 /yg/ft2 as clearance standard for floor.
** Grantees that used 100 jt/g/ft2 or 80 {jglft2 as clearance standard for fioor.
t  Slope estimate is statistically significantly different from zero at the 0.05 level.
                                             71
U.S. EPA Headquarters Library
       Mail code 3201
1200 Pennsylvania Avenue NW
   Washington DC 20460
-------
Table 6-5.    Categorization of the Number of Parameter Estimates and Associated
             Standard Errors from a Logistic Regression Model of the Proportion of
             Samples within a Residential Unit Expected to Fail Clearance Testing Based
             on the Number of Samples Collected.
-- - ^
«..». "
— Data Source
V
j. »
Number of ;'
.'Residential
Units ,
• r - '".- ; Pafamef e/ Estimates jfv-
• -' ' '•:•%; ii '• *•'' •;*";$•«. :;/ -.-"'t. "

.-. ' , P«:V^
s.e.'lPo)

•T..T-. .PV"-... -.-'
•-•.^(W:-'
v, ;. Estimated Proportion of Samples .,
::' Failing: Clearance Testing Whenithe
• -i - •; ^n£ '^•-*'*-- •*'...<~*''*'-iis- •• •'..•.•' -ti- i
Number -of Samples Collected ^Within .
V i a Residentiaf Unit is .Betwreen* -:: ":

^.•.!^4^pf*
-' •">•" .- ., ,-1,. .-- •
^j-^BSiMSS *y ' =
Floor
Maryland
HUD FHA
HUD PHA
HUD Grantee (High}»
HUD Grantee
|Low)*»
Atlantic City
Cleveland
Dover
661
145
119
2092
1021
159
38
151
-2.5154
(0.1925)
-1.4789
{0.3043}
-2.1458
(0.4304)
-2.5980
(0.1356)
-3.0986
(0.2332)
-1.9776
(0.4594)
-0.9233
(0.6693)
-5.1085
(0.8678)
Wind
Maryland
HUD FHA
HUD PHA
HUD Grantee (High}»
HUD Grantee
(Low>»*
Atlantic City
Cleveland
Dover
612
126
111
2031
968
20
33
12
-1.1391
(0.1870)
-2.0246
(0.3982)
-2.3408
(1.0466)
-3.3402
(0.2779)
-3.7198
(0.2141)
21.7156
(0.4737)
19.2362
(1.0065)
-0.4520
(0.4835)
0.1205
(0.1032)
0.0074
(0.1261)
-0.01 59
(0.2343)
-0.1256
(0.0792)
-0.0058
(0.1563)
-0.4473
(0.3166)
-0.6927
(0.4549)
0.9188
(0.5639)
0.08
0.19
0.10
0.06
0.04
0.08
0.17
0.01
0.09
0.19
0.10
0.05
0.04
0.05
0.09
0.04
ow Sill ' •
-0.6903 1
(0.1251)
0.0835
(0.1923)
-1.0932
(0.8080)
0.1747
(0.2459)
0.4153t
(0.1562)
-23.820t
(0.0000)
-23.580t
(0.0000)
0.0000
(0.0000)
0.14
0.13
0.03
0.04
0.04
0.11
0.01
0.39
0.07
0.13
0.01
0.05
0.05
<0.01
<0.01
0.39
                                        72
-------
Table 6-5.  Continued
" tnl'llv;'" L^'l'"' fY'*'*, .*"" 1
^;^|jrv ,,-... ..
.;-i*-X"». :!•
• *5i -^ •' ~ " • t " * •*
A*!. ., /K' „>!.•_,- •'. '.I* .
>• ir-ri"- ':?/^; -f
iNiimbero^
.••Residential!:
."".•"-Units-''."'-
. ,* ' "''- '** "
U ., •./* .;!Ej^^=±^;jf'ta,^-,'V ''^Mjf'^ ).; il'JT
..- ' v *„./• ""t ^isliT ^L^r^1'^' '• * i.LWf*;lii'v^ 'tir&r'i.
/••'f^Par'am^ter^Estimafes-i^-v
«:•' ' j?V'f i';:>v,i!tftj;;-,,'f-it' '':"," -••;•.'"'"' ' '?' '*'•
:'• $*$§&$&£ l^*:-1^^ 3 >
,.-V'-"p6V:^a
;... -: ." ';•},.. -.j.
s.e.(Po» .'
T"$kfi&$&
.v Estimated Proportion of Samples"
*'% "i r ! ' " ; > . V* "-jF ; "•;•'* '" " ' *r**s* .*• V:~ -"'• •
"\ Failing Clearance Testing Whehithe •.
• -: • -N»-NW . '.-. - -M?,, • **?;• * ~ *>** ;fc^r,- jf^i^^i. i^.t
Number of Samples Collected Within
.: ^_ a Residential UniriSiBetween'pife
:;.^?.ii^-S^
3;.- v ,=-r-;,v-v--S»»5"» -• , «•".
• Window Troiinh ' - • :. .-•,,,
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover
536
108
110
1697
873
119
24
146
-1.2361
(0.1244)
0.2344
(0.3207)
-5.6392
(1.4921)
-2.7639
(0.4205)
-2.0399
(1.0530)
18.4886
(0.4132)
-2.9178
(0.7260)
-3.9512
(0.4515)
-0.0708
(0.0686)
-0.4885 1
(0.1902)
1 .0594
(0.8432)
-0.0175
(0.3973)
-0.2996
(1.0396)
-22.2061-
(0,0000)
0.0000
(0.0000)
0.0000
(0,0000)
0.21
0.44
0.01
0.06
0.09
0.02
0.05
0.02
0.2
0.32
0.03
0.06
0.07
<0.01
0.05
0.02
' . - ; • All Conrmonents • .. . '
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee
(Low)"
Atlantic City
Cleveland
Dover
706
149
119
2132
1025
160
38
158
-1.8066
(0.1128)
-1.1257
(0.2019)
-2.0704
(0.3687)
-2.3494
(0.1173)
-2.7902
(0.2156)
-2.7647
(0.4212)
-1.2386
(0.8123)
-0.3746
(1.1666)
-0,0360
(0.0309)
-0.0505
(0.0474)
-0.21 30t
(0.0985)
-0.2129T
(0.0450)
-0.0719
(0.0912)
-0.0241
(0,2101)
-0.4689
(0.3130)
-1.6367t
(0.6075)
0.14
0.24
0.09
0.07
0.05
0.06
0.15
0.12
0.13
0.23
0.08
0.06
0.05
0.06
0.10
0.03
*  Grantees that used 200 //g/ft2 as clearance standard for floor.
** Grantees that used 100  fjg/ft2 or 80 jug/ft2 as clearance standard for floor.
t  Slope estimate is statistically significantly different from zero at the 0.05 level.
                                              73
-------

 6.2   OBJECTIVE 2:  CHARACTERIZATION OF THE DISTRIBUTION OF THE DUST-LEAD
       LOADINGS, GEOMETRIC MEAN DUST-LEAD LOADINGS. VARIABILITY BETWEEN
       SAMPLES COLLECTED WITHIN HOUSING UNITS. AND VARIABILITY ACROSS
       HOUSING UNITS
       The geometric mean, the house-to-house variability (aBetween), and the room-to-room
variability within a home (owjthin) by site visit and component tested are listed in Tables 6-6a to 6-
6d, for the first site visit to fourth site visit, respectively. In addition, these tables also present the
lower and upper 95% confidence bounds for the geometric mean and the 95% prediction bounds
which were calculated using the two sources of variability, o^^^ and awithin.  Section 5.2
discusses the equations used to estimate ^Between and owithin from the In-transformed data. Note
that the length of the confidence intervals for any given component type increases successively
from the first to the fourth site visit. This increase in length is due primarily to the decrease by
orders of magnitude in the sample size used to estimate the variance components.
       Generally, for a given component type, geometric means of the samples collected
increase from the first to the third site visit, then decrease from the  third to the fourth site visit.
In the first site visit, Table 6-6a shows that geometric mean floor, window sill, and window
trough dust-lead loadings range from 8.8 to 57.6 ug/ft2,11.3 to 461.6 ug/ft2, and  15.8 to 393.3
ug/ft2, respectively. While the HUD Grantee High data group presents the lowest geometric
mean floor and window sill dust-lead loadings at 8.8 and 11.3  ug/ft2 across eight data sources,
the HUD FHA data group presents the highest geometric mean floor and window trough dust-
lead loadings at 57.6 and 393.3 ug/ft2.
       The Dover data  group presents the highest geometric mean window sill dust-lead
loadings and the lowest geometric mean window trough dust-lead loadings.  This may be
attributed to the interior renovation strategy used by Dover Housing Authority, which consisted
of removing lead painted radiators generally located under a window and window replacement.
Window replacement may not necessarily include replacing window sills.
       The percentages of dust-lead clearance samples below 50,75,100 ug/ft2 and the
applicable interim clearance standard (200 fig/ft2 for floor, 500 ug/ft2 for window sill, and 800
ug/ft2 for window trough) by component type and substrate are presented in Tables 6-7a (first
site visit) to 6-7d (fourth site visit)  and 6-8 (passed  clearance visits). The 50th, 80th, 90th, 95th,
and 99th percentiles for the dust-lead clearance samples for the first site visit and the passed
clearance visits and combinations of data source and substrate are presented in Tables 6-9a to 6-
9c and 6-10a to 6-10c.  Additional tables listing the percentiles by data source are provided in
                                           74
-------
Appendix I. Box and whisker plots that present the distribution of dust-lead loadings from the
first and passed clearance visits by component type and substrate are displayed in the appendices
for each individual data source.
       Using the results in Tables 6-6 through 6-10, Tables 1-1 through 1-4 and Figures 1-1 to I-
18 in Appendix I, and the box and whisker plots displayed in Appendices A to G, the following
comparisons were made:
What lead levels for floors, window sills, or window troughs are typical in dust clearance
testing?
       Table 6-6a shows that the geometric mean dust-lead loading from floors ranged from 8.8
to 57.6 M-g/ft2 with an upper 95% prediction bound ranging from 68 to 509 ng/ft2 for the first site
visit.  Although the geometric mean lead loading for floors was significantly less than the 200
Hg/ft2 clearance standard, some of the 95% upper prediction bounds were above the 200 ug/ft2
level. Similar results were obtained for window sills and troughs. The geometric mean lead
loading for window sills ranged from 11.3 to 461.6 ug/ft2 with a 95% upper prediction bound
ranging from 89 to 7559 ug/ft2 for the first site visit compared to the 500 ug/ft2 clearance
standard.  The geometric mean lead loading for window troughs was between 15.8 and  393.3
ug/ft2, and the upper 95% prediction bound was between 137 and 4119 ug/ft2 for the first site
visit compared to the 800 ug/ft2  clearance standard.
       Another way to answer the question is to consider the percentiles of the dust wipe sample
results presented in Tables 6-9a  to 6-9c and 6-10a to 6-10c, and Tables 1-1 through 1-4 in
Appendix I.  These results indicate that for the first site visit between 90 and 95% of the floor
dust wipe sample results and 80 to 90% of the window sill and window trough clearance samples
were below the Interim EPA/HUD clearance standards.  However, as seen in Tables 6-9a to 6-9c,
there are some extremely high values associated with the 95th and 99th percentiles. Most of
these extreme values are from the Maryland and the FHA HUD Demonstration data. Both sets
of results were obtained after the housing units had passed a visual inspection. The FHA dust
wipe samples were obtained before  a sealant was applied to the sampled surface, while  the
Maryland data were obtained by following the HUD Interim Guidelines Clearance Standards [2]:
applying a sealant to the designated surfaces and then performing clearance  sampling.
       A data set was constructed which included lead loading results only from housing units
that passed lead clearance (passed clearance data set). An analysis of the distribution of dust
                                          75
-------
wipe clearance sample results for this data set would indicate the levels attained for all the
housing units that passed the clearance standards. As shown in Tables 6-lOa, when considering
the category of all substrates, the 80th, 90th, and 95th percentile lead levels for bare floor dust
wipe samples ranged from 34 to 102 ug/ft2, from 57 to 141 ug/ft2, and from 82 to 165 ug/ft2,
respectively.  For all substrates in window sills (Table 6-1 Ob), the 80th, 90th, and 95th
percentiles ranged from 72 to 443 ug/ft2, from 141 to 443 ug/ft2, and from 215 to 467 ug/ft2,
respectively, and from 59 to 431 ug/ft2, from 111 to 598  ug/ft2, and from 157 to 678 ug/ft2 for all
substrates in window troughs (Table 6-10c), respectively.

For each site visit,  how do the dust-lead levels compare across component type?
       Tables 6-6a to 6-6d show that for all site visits, generally for most of the data sources, the
geometric mean dust-lead loading for floors was less than the mean for window sills. In turn, the
geometric mean dust-lead loading for window sills was less than the mean for window troughs.
One obvious exception is Dover's window sill and trough data.

For each component, how do the dust-lead levels compare across site visits?
       As shown in Tables 6-6a to 6-6c, for each component type, the geometric mean lead-
loading increased from the first to the second site visit, and then again from the second to the
third site visit. This trend may be due to the fact that typically only those homes which did not
pass clearance standards on the previous site visit were included in the subsequent site visit (i.e.,
only the dirtiest homes were included).
What are the problem areas in terms of cleaning? Is one area more difficult to clean than
another?
       As shown in Table 6-6a, the geometric mean dust-lead loading ranged 8.8 to 57.6 .ug/ftz
for floors, 11.3 to 461.6 ug/ft2 for window sills, and 15.8 to 393.3 ug/ft2 for window troughs for
the first site visit. The geometric mean lead loading for floors, window sills, and window
troughs increased from the first site visit to the second site visit, and then again from the second
site visit to the third site visit.  The geometric means decreased from the third site visit to the
fourth site visit, but there are very few samples to estimate the geometric mean during the fourth
site visit. The window trough results were higher than the window sill or floor results, except for
during the fourth site visit. This was determined using a comparison of the 95% confidence
                                           76
-------
intervals constructed from the variance component models (the level of significance is actually
less than 5% for the comparison).
       The higher dust wipe sample results for window troughs were discussed with experts in
the field. They collectively indicated that window troughs are the most difficult of the three
surfaces to pass clearance. Many of them indicated that contractors will first, if possible, replace
old windows with either vinyl or aluminum windows, and second, put a laminated aluminum coil
or insert over an enclosure. Their third option is to refinish the window trough using more
involved on-site paint removal techniques. However, this third option is less desirable because it
creates leaded debris, and there is an experience factor required to do the job adequately.

Can window sills be cleaned to the same levels as floors?
       Note that the available data were not designed to answer this question directly. The data
can provide insights however.
       Table 6-6a shows that for some data sources (e.g., HUD FHA, HUD Grantee High, and
Cleveland), the 95% confidence intervals for geometric mean dust-lead loading from floors and
from window sills overlap. The 95% confidence interval for geometric mean dust-lead loading
from window sills is even lower than that from floors for the HUD PHA data. (Note that the
level of significance is actually less than 5% for the comparison.)
       Table 6-7a shows that for the first site visit at the 50, 75, and 100 ug/ft2 standard levels,
over all substrates, the percentage of floor-lead loading results below the standard exceeds the
percentage of window sill results below the standard by approximately 10%.
       As seen in Tables 6-9a, 6-9b, 6-1 Oa, and 6-1 Ob, and  Tables 1-1 through 1-4 in Appendix I,
at each percentile (50th, 80th, 90th, 95th, and 99th) the floor dust-lead loadings are generally
lower than the corresponding window sill dust-lead loadings.  This indicates that floors were
generally cleaned to a lower level (i.e., cleaner, or less lead  dust remaining) than window sills.

Can window troughs be cleaned to  the same levels as window sills?
       Again, as stated above,  while the available data in this report were not designed to answer
this question directly, the data can provide insights.
       Table 6-6a shows that the geometric mean dust-lead loading from window sills ranged
from 11.3 to 461.6 ug/ft2 for the first site visit.  The geometric mean lead loading for window
troughs ranged 15.8 to 393.3 ug/ft2 for the first site visit.  Their lower and upper 95% confidence
                                           77
-------
bounds do overlap for some data sources. (Note that the level of significance is actually less than
5% for the comparison.)
       Tables 6-7a to 6-7d and 6-8 show that a greater proportion of window sill dust-lead
loadings fell below the standards of 50,75, and 100 ug/ft2 than the window trough dust-lead
loadings for most of the larger data sources (specifically, Maryland, HUD FHA, HUD Grantee
High and Low). Assessing the first site visit at the interim clearance standards of 50,75, and 100
jig/ft2, approximately 10% more of the window sill dust-lead loadings were below each standard
than the associated window trough dust-lead loadings for those larger data sources. As seen hi
Tables 6-9b, 6-9c, 6-10b, and 6-10c, and Tables 1-1 through 1-4 in Appendix I, at each percentile
(50th, 80th, 90th, 95th, and 99th) the window sill dust-lead loadings are generally lower than the
corresponding window trough dust-lead loadings. This indicates that window sills were
generally cleaned to a lower level (i.e., cleaner, or less lead dust remaining) than window
troughs.

How do the distributions of floor, window sill, and window trough lead  loadings compare
for first site and passed clearance site visits?
       The first set of box and whisker plots that are presented in the appendices for each
individual data source display the distributions of dust lead loadings from the first and passed
clearance site visits for floors, window sills, and window troughs on a log scale. These results
clearly indicate that the distribution of dust-lead loadings for each component type during the
first site visit and passed clearance site visits are significantly skewed and variable. The
distributions of dust-lead loading results from the "passed clearance" data for floors, window
sills, and window troughs are displayed on an untransformed scale in the second set of box and
whisker plots in the appendices for each individual data source. The skewness of the
distributions of dust-lead loading results for window  sills and window troughs for the passed
clearance site visit data are better displayed on the original lead loading, ug/ft2, scale. In both
figures, the floor dust wipe sample results are generally lower than sample results for the window
sills, which are lower than window trough sample results.
                                            78
-------
*»
"55
>
o
w
CO
II
a>
£
^
o
0)
g
o
w
H
(0
O
(0
o
a.
+*
c
o
s
E
u
O
A
0
«B
1
U
c
3J
c
o
a
i
o
U
a
u
m
°-

85?
T* 5 £
5=5|
J
SI,
ill
S § u
so
£8
tower 9
Confide
Bouni

«
i "
t



ft
° e
„
• e
|i
O



»
"Jl
4k


e -o
If
° CO

15 s
,*. £ E
I (0




i
1
I






















j
u.



























to
PH



CO


IO
CM



O
CM


IO
0>
O



CO


„
CM
CM





to
to




IO
to
to
CM



g
CM






1
£
ID

cn
o
10



**


£



5


IO
<=




in


to
IO





to




CM
CM
o>



to
cn






<
u.
*

CO



(£>


5



(O
Tf


g
O



10
10


cn
CM
IO





;




o
cn



CO
in






<
X
a.
n
3
X

CO
IO



"~


o



CO


CO
*""



CO



S)





CM
CM




CM
in



01
CO




CD
X
3 Grantee
3
X

to
cn



CM


2



-


"
«—



5


CO
CM





CM
0





Q S
i£

U)



CM


CM
CM




f£
U
.a
e
•^
<

CO
CD
CM



CO


CO



CM
CM


IO
O
*"~



r*.
CO


,>.
cn
CM





CO
CO




in



CO






•o
c
a
1
w
O

CM



CM


to



CM


(N
O
^



5


10
CO





u>




CO
CM
to



CO
CO







*




















ss
CO
1



























CO
CO
CO



IO


5



CO


*
*~



o
in


r^
;





CM




CO
CM



IO
to
CM






tJ
1
i

IO



to


o
CO



IO


CM
CO
^



to


CO
6
CO





ID
CM




I



CO
CO
to






X
u.
n
s
X

CO
CN



*


cn
CO



CO
CM


ID
cn
o



CO


,>,
CM
CO





£




a>
CO



;






I
n
3

o>
CO



*~


CO



o


CO
**



n
q


CO
£





CO
o
CM




O
cn
in



CO
O!



Jf
cs
X
D Grantee
3

0>
cn



CO


CO
CM



CM
CM


S
**



cn
*


cn
CM





cn




CO
CM
CM



CM






a
1
O-_
a i


cn
-------
 •o


 §
 9
 a>
 e

 o
(0
 a
O

•a

 CD

 03
 a
 9
 Q.


 O
 w
 a>
•»*
 co


*!
 OT
UJ




I
 O
 a.


 o
O
.S

 ro


•o

 (0
 a
 a>
 u
•c
    .52
    >
 *•
# e-
U) 0--D

i S °
Upper 96%
Confidence
Bound
*l
if1
j|.
i\.
«? — •
s

u
Geometri
Mean
z E
If
a
"5 •>
-ii
5
I
O











o
LL











CO
CM


CO

m

cv
CN
q
8.




r%
aj
CM
o>
O
o
CM


fx
o
CM

|| Maryland

IO
CM
o


o

o
•9

r*
cn
O
CM
CM




e>
cn
o
r*.
CO
CM


CO
CM

I
U.
o
I

CO
CN
IO


«

o
CO

CO
to
h.
6
§




o
IO
CM
CO


CO
to

|| HUD PHA

IO


CM

CM
CM

"
CO
IO
o




CD
CO
CO
CO
o
to
10


00
o
to
X
|| HUD Grantee

CO


—

en

o
IO
CM
IO




CO
CO
s
to
CO
CM


IO
CM

UHUD Grantee
(Lowt

CO
IO


CM

0
CO

8
ID
cn
6
£




o
CM
•«•
to
CO


CM

|| Atlantic City

CO
IO


—

CD
CD

CO
oo
CO
6
01
CM




CO
CM
CO
CO


CO

|| Cleveland










S

1
"*










o
CO


IO

CO
co

CO
CM
CO
d
to
to




10
10
IO
05
CO


01

| Maryland

!


*

GO
CM

ID
O
CO
in
°




01
to
CO
0


^

|| HUD FHA

CO
cn



41
2
C
d>
0 |









S
o
§
^



o
-
_


*~

|| Cleveland









"
|

1
i










CM
CO


-

CD
IO

CO
CO
to
,




q
IO
o
to
CM


to
CM

|| Maryland

o
CD
CO
to


g

CO

CO
to
CO
CO
01
CO




CO
CM
IO
IO
IO
CO
CD


00
co

I
u.
o
I

10925


,-

o
CO
IO

CM
CM
CO
o
CO




co
CM
CM
to
r*-


^

X
CL
I

09

§


CO
o

o
4>
C
-------
 0
 *->

 w

 •a
 I
 O
 a>
 *•
 a
 a

 •o

 a

 e
 Q.
 0)
 C
 o
 a


 o
O
S
a



a>
UJ
o
o
O

o
o
a

•o
§


i

^o
1-1
o

o
9
 -O
Ml.
fil m'
10 3 -a-
|If
111"
to c -o
» jTc
fci ^* 3
'*•-.-• ;r_.
"A
•>]t
;M|
1
' . !
of Houses., i
Sampled- <
• * '.'•<:
E TJ •
*. '
fof
	 Individual
Samplos ,
§
o ••
(O





,J!

*
i.

;••'.
1
S


'







8
CD
CM
CM
CD

10
CO
CO
O
CO
CD
O
CO
d
O

C3

CM


a
1

o
o>
10
en
S

-
o

10
CD
O
i
ID

10

0
00


1
Q
I

CS
en
o
to
10

O
CO
10
o
CO
o
n
CO

en

2


x
CL
a
3

CS
CM
CO

CO
CO
CO

en
CD
o
o
oi
CM
CO

en

10
CD
2
o
I
o
a
i
<•>
Q
3

I
-
?

en
en
CO

w

cn
to
OJ

S

s

V
s
o f






o
o
o
o
o
o
o
-

^

-


1
u

•?

.?


: t.

i
1

..",



-,!





CO
CO
00
o
CO
CO

CO
10
CO
o
to

CN
10
10

CM
CN

O
n


1
I

10242
CO
m
CM

CM
en
o

CO

CO
§
10
CM

CO
CO

CO
n


u.


ID
'
10
00

o
n
CN
CO

t
e>
o
CM
d
ID
10
CO

?

5
f
0
a
<5


CO
10
-
10
m

—
u>
o
„
CM
CD
CO
CO
10

00

00

8
c
(D
o 5

i






i
t*.
s
','•!.










s
o
in
-
10

rx
CO
CO
CS

CN
CO

10
0>
CM

CO
CM

CM


tj
0
£
i

10
IS
CO
00
o
CM

O
IX
n
00
o

CO
CM

CO
s

o
CO

CM
CO


I
U.
o
I
o
1.0993SE
o
3333S

o
en
CO
o
o
o
o
o>
co
CO
CN

CM

CM


1
a
X

CO
CO
(S
10

o
10
v

o
o
o
CO
CO
CO

CO

CD
s
1
c
a
n
o
X

ID
CN
*
en
n

00
CN
O
en

CO
o
CM
en
CM
CM

CO
CN

n
CM

I
s.
o s






o
c
c
o
c
o
c
-

J_

-

>
c
c
1
                                                                  o
                                                                  o
                                                                  CO
11
s °
"-«
8§
                                                                3 3
                                                                II
                                                                s s
                                                                o o
                                           81
-------
 LL



 £
 O
 U
 CO
O

•o

 o

 o
 a



**

 0

 o
 o
u
 CO
 o
 *-•
 a
 
•o
(O

(O

 0)
Upper 96% I
Prediction I
Bound |
I!1
Hi
30
* 8'
|||
'is 3
••'*%-'

	 "*••'
:4;
• " 4
•ta -r
'• 5 C:' '
:- •:'! 1"' !
• . 0. . • •
•ll-
-'s S •
~-> o £
i
s
o
(O
/'•I
..'

•-







,


















t

^

o
o
6


o>




co
p>


Maryland
CM
CO
CM
O


CO

CM
CM

CO
cn
o


^
ID
10

CO

10
r-


cn
o
f
a
X
s
HUD Gran
10
8
o
10
CO


-

CO
to
o
(O


CO
T

,

-
-

*
V
2.
&
§1


••







S
•i


















10
CM

O


co
<:



CO
CO


|| Maryland
251066
-
CO
cn
CO
CO


IO
1C

in

o
CO
6


CO
CO

CO

CM
CM


U.
O
X
232810507 |
o
CO
s


cn

o
CM
O
o
o
o


CO
10

CM

CM
CM
I
B
C
s
a
s
X
»'





,.•



• £
;l
'.•0
;•§,
..•.,-.:


1


;



CM
cn
a>
cn
-
CO
n


CO
O

CM
10
O
cn
CM
6


10
<0
CO

CO

10
in


|| Maryland
in
00
-
CO
CM


t

co
o>
O
CM
CM


*
5

..

CM
tN


u.
Q
X
3346S62914
o
CO
o


o

o
10
o
cn
CM
o


«
IO

CM

CM
CM
•±
S
c
o
X
r-
co
10
10
10


§

o
o
o
6


—
o
CO

,

-
-

s
c
Q 1
X =
   •o
                                                                                                  •g-
 2 o
•5 «
•o -a

in ai
 c c
 2 2
a o
                                                                 82
-------
Table 6-7a. Percentage of Clearance Samples Below 50 A/g/ft2, 75 A/g/ft2, 100//g/ft2, and
           HUD Interim Guidance Clearance Standards by Data Source and Substrate for
           the First Site Visit.
Data Source
*
Substrate '
Sample Size

Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)'
HUD Grantee
(Low)"
Atlantic City
Cleveland
Dover
All
Wood
Vinyl
Other
Unknown
All
Unknown
Al)
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
All
Unknown
All
Unknown
All
Wood
Vinyl
Unknown
2750
1082
660
19
989
967
967
558
558
8179
2027
4067
2076
9
3642
1774
1510
358
554
554
158
158
633
3
625
5
'/ -Percentage of Samples Below
50 //g/ft2
Floor
69
71
68
37
70
51
51
52
52
81
75
83
81
89
85
86
87
70
77
77
66
66
83
67
84
80
75 f/g/ft2
100 A/g/ft2
Interim
Clearance
Standard

77
79
75
42
78
61
61
66
66
86
82
89
86
89
90
92
91
78
84
84
77
77
91
100
91
80
82
84
80
68
82
66
66
76
76
89
85
91
89
100
92
93
93
81
87
87
82
82
95
100
95
100
91
91
90
84
91
81
81
90
90
94
92
95
95
100
96
97
96
91
93
93
87
87
98
100
98
100
                                         83
-------
Table 6-7a.  Continued
Data Source
Substrate
-; .";
Sample Size
•- •' •• "Percentage of Samples Below
-t-wwm*
'TS/JB/ft2
Window SiB
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)'
HUD Grantee
(Low)"
Atlantic City
Cleveland


Dover
All
Wood
Vinyl
Aluminum
Other
Unknown
AH
Unknown
AH
Unknown
AH
Wood
Vinyl
Other
Unknown
Alt
Wood
Vinyl
Other
All
Unknown
All
Unknown
All
Wood
Vinyl
Aluminum
Unknown
2465
1687
97
40
18
623
668
668
411
411
4798
636
37
4120
5
2147
507
35
1605
51
51
93
93
18
9
1
4
4
60
62
81
73
78
51
48
48
66
66
71
56
78
73
60
57
49
71
59
57
57
61
61
6
0
0
25
0
68
70
88
75
83
58
59
59
79
79
78
65
81
80
60
68
62
74
70
65
65
69
69
6
0
0
25
0
100 //a/tt2
Interim
Clearance
Standard

72
74
90
83
83
63
64
64
83
83
82
72
84
84
60
78
71
77
80
75
75
78
78
6
0
0
25
0
90
92
97
98
100
82
87
87
98
98
96
92
92
97
80
96
96
94
97
90
90
99
99
61
44
100
100
50
                                         84
-------
Table 6-7a.  Continued
; Data Source
«. . . .' •'-••••.•-
Substrate
Sample' Size
Percentage of Samples* Below . .. .-. ...
...• SOifg/ft*," :
, 75>/g/ft2 .
. .100 «B/ft? .
<• , Interim ,
• , Clearance . ".
: Standard '
'"-•••-'•-•/.".' : 'Window Trough .'•• . '. • . -• ^ :
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High}'
HUD Grantee
it i * *
(Low)
Atlantic City
Cleveland
Dover
All
Wood
Vinyl
Aluminum
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
All
Unknown
All
Unknown
All
Vinyl
Aluminum
Other
Unknown
1985
458
551
286
690
442
442
371
371
3002
34
1626
1341
1
1346
65
238
1043
258
258
39
39
265
185
11
2
67
41
27
54
50
36
17
17
58
58
61
47
69
51
0
36
32
69
29
75
75
64
64
77
74
91
50
85
47
32
60
57
42
22
22
75
75
68
59
76
59
0
48
37
75
42
81
81
67
67
83
79
91
50
94
51
36
65
62
46
25
25
82
82
73
65
81
65
0
56
40
79
52
84
84
72
72
88
85
91
50
96
79
67
92
88
74
64
64
98
98
94
94
97
91
0
91
78
94
91
98
98
95
95
98
97
100
100
100
   Grantees that used
   Grantees that used
200 //g/ft2 as clearance standard for floor.
100  j/g/ft2 or 80 fjglft1 as clearance standard for floor.
                                                     85
-------
Table 6-7b. Percentage of Clearance Samples Below 50//g/ft2, 75 j/g/ft2, 100//g/ft2, and
           HUD interim Guidance Clearance Standards by Data Source and Substrate for
           the Second Site Visit.
• . ' **'•
Data Source- <•

Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)"
HUD Grantee
(Low)"
Atlantic City
Cleveland
J
•5 ' '
..V-' - -:
Substrate
Sample Size. •
'«'."-.•' " • Parentage of Samples Below \: '.-'..,
jsomift?
: • . . .Floor
All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Unknown
All
Unknown
All
Unknown
All
Unknown

Maryland
HUD FHA
HUD PHA
All
Wood
Vinyl
Aluminum
Unknown
All
Unknown
All
Unknown
207
71
52
3
81
234
234
53
53
608
608
245
245
42
42
18
18

194
102
12
3
77
111
111
10
10
59
56
50
33
69
31
31
36
36
69
69
76
76
33
33
61
61
75 jig/ft2 .
vipbj#m*-;'
' n- :
Interim
Clearance
Standard

69
65
62
33
79
38
38
45
45
77
77
82
82
40
40
72
72
Window Sill
52
46
92
67
53
26
26
80
80
59
52
92
100
62
33
33
80
80
76
75
67
67
84
46
46
57
57
82
82
86
86
40
40
72
72
89
85
88
100
94
65
65
75
75
90
90
92
92
69
69
83
83
-
65
58
92
100
69
40
40
90
90
87
80
100
100
94
75
75
100
100
                                        86
-------
Table 6-7b.  Continued
Data Source
Substrate
Sample Size
: Percentage of Samples Below ' . ";
50^9/ft2
75^9/ft2
lOO/fo/ft2
. ...*..; ' '
Interim" '. . ..
Clearance
Standard '
Window Sill •:-.!• • • r ; -,
HUD Grantee
(High)'
HUD Grantee
(Low)"
Cleveland
All
Unknown
All
Unknown
All
Unknown
281
281
91
91
1
1
54
54
57
57
100
100
63
63
64
64
100
100
68
68
66
66
100
100
93
93
91
91
100
100
, • Window Trough ' .'.'Z' ' ' ~'~; " • '• - - , ;?.
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)'
HUD Grantee
(Low)"
Atlantic City
All
Wood
Vinyl
Aluminum
Unknown
AD
Unknown
All
Unknown
All
Unknown
All
Unknown
All
Unknown
275
83
72
30
90
168
168
7
7
202
202
108
108
17
17
35
20
47
47
34
10
10
29
29
48
48
35
35
47
47
44
34
56
53
40
14
14
57
57
58
58
44
44
59
59
49
36
63
67
43
15
15
71
71
63
63
50
50
59
59
83
73
93
97
78
52
52
71
71
94
94
83
83
82
82
*  Grantees that used 200 //g/ft2 as clearance standard for floor.
** Grantees that used 100 pg/ft2 or 80 //g/ft2 as clearance standard for floor.
                                                     87
-------
Table 6-7c. Percentage of Clearance Samples Below 50 j/g/ft2, 75 jig/ft2, 100//g/ft2, and
           HUD Interim Guidance Clearance Standards by Data Source and Substrate for
           the Third Site Visit.
Data Source
Substrate
F
Sample. Size
•/'''." ' - ".-.•• ..- •'•.''.
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)*
HUD Grantee
(Low)"
Cleveland

Maryland
HUD FHA
HUD Grantee
(High)'
HUD Grantee
11 »• •
{Low)
All
Wood
Vinyl
Other
Unknown
Atl
Unknown
All
Unknown'
All
Unknown
All
Unknown
All
Unknown
23
2
11
1
9
80
80
13
13
85
85
33
33
1
1
• .-. Percentage of Samples- Below ' '
SOj/9/ft2
: Floor
74
100
55
100
89
6
6
8
8
54
54
64
64
100
100
'• > . • .. Window Sill .
All
Wood
Aluminum
Unknown
All
Unknown
All
Unknown
All
30
18
3
9
38
38
44
44
18
73
78
33
78
8
8
36
36
44
75pB/ft*
iOOjig/ft*
Interim
Clearance
.Standard
*'.'-'' • • ? '". 4. -';•:-: : \n'. . • : "./•-•/•'•' >
83
100
73
100
89
21
21
8
8
71
71
73
73
100
100
87
100
82
100
89
34
34
8
8
73
73
82
82
100
100
100
100
100
100
100
58
58
38
38
87
87
91
91
100
100
- • : •-.."• ,• ,
77
78
33
89
13
13
45
45
50
77
78
33
89
18
18
57
57
56
93
94
67
100
45
45
95
95
100
                                          88
-------
Table 6-7c.  Continued
" " 1 ''"• '•
•'; **• :' '
• Data Source ;
Substrate -,.
' Sample Size
. ^ • Percentage of Samples Below .. :
sdjjgfft*
Window Trough,
Maryland
HUD FHA
HUOPHA
HUD Grantee
(High)'
HUD Grantee
i* «*•
(Low)
Atlantic City
All
Wood
Vinyl
Aluminum
Unknown
All
Unknown
All
Unknown
All
Unknown
All
Unknown
All
Unknown
27
8
3
3
13
82
82
2
2
18
18
23
23
1
1
30
0
33
0
54
10
10
50
50
33
33
35
35
100
100
75/Jflrtt* -
100/*g/ft2
Interim ..'.-
Clearance •
Standard ;'

33
0
67
0
54
13
13
SO
50
44
44
39
39
100
100
37
0
TOO
0
54
17
17
50
50
56
56
39
39
100
100
70
50
100
33
85
43
43
100
100
94
94
83
83
100
100
   Grantees that used 200 //g/ft2 as clearance standard for floor.
   Grantees that used 100 f/g/ft2 or 80 pg/ft2 as clearance standard for floor.
                                                     89
-------
Table 6-7d.  Percentage of Clearance Samples Below 50//g/ft2, 75/ig/ft2,  100//g/ft2, and
             HUD Interim Guidance Clearance Standards by Data Source and Substrate for
             the Fourth Site Visit.
" •'", ".
Data Source
* Substrate'
Sample Size
..... .'.'-'.
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High!'
HUD Grantee
(Low>"
All
Unknown
All
Unknown
All
Unknown
All
Unknown
All
Unknown
3
3
16
16
8
8
10
10
7
7
• * • .' .
Maryland
HUD FHA
HUD Grantee
(High)'
All
Unknown
All
Unknown
Ail
Unknown
3
3
12
.12
2
2
."'..'•-,_ . Percentage of Samples Below .•;.•/.... * ?>£'£*'.£'"'
50Vg/ft*
75//8/ft2
.Floor .*
33
33
44
44
25
25
90'
90
100
100
33
33
63
63
63
63
100
100
100
100
Window Sill
33
33
0
0
0
0
33
33
0
0
0
0
-.'"'. : , : Window Trough "
Maryland
HUD FHA
HUD Grantee
(High)'
HUD Grantee
(Low)"
All
Wood
Unknown
All
Unknown
All
Unknown
All
Unknown
5
1
4
21
21
2
2
4
4
0
0
0
10
10
50
50
0
0
0
0
0
24
24
50
50
0
0
IOOj/9/ft2
- interim'; •>'
:• Clearance
~ -Standard ni
•'. 	 	 ;..-.. * .•'••-•'••"' ' ' '••".:il'^-,':-'iJ'™.!«i,
100
100
63
63
75
75
100
100
100
100
100
100
75
75
100
100
100
100
100
100
•'••'. '-•.•• '• • • :_:• '..\'.;..'..":!'.l:' •^•f'i'
67
67
0
0
0
0
100
100
42
42
100
100
..•'... -"> •' :• •••••r-"' ~.:v
0
0
0
29
29
100
100
0
0
80
100
75
71
71
100
100
100
100
  Grantees that used
  Grantees that used
200 ;/g/ft: as clearance standard for floor.
100 //g/ft2 or 80 pg/tt2 as clearance standard for
floor.
                                             90
-------
Table 6-8.  Percentage of Clearance Samples Below 50 /ig/ft2, 75 /ig/ft2, 100 /ig/ft2, and
           HUD Interim Guidance Clearance Standards by Data Source and Substrate for
           the Passed Clearance Visits.
Data Source
Substrate
Sample Size
Percentage of Samples Below -•*':.-•
SOpg/ft1
75/fg/ft2 .
100 jig/ft2
. ' Interim
Clearance
- : Standard. •
. ' " ; ' - '•'-•,' 'Floor . • .•'•.''•' ~ '•• ' ;.•': .- • . '•
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High!*
HUD Grantee
(Low)"
Atlantic City
Cleveland


Dover

All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Unknown
2712
1052
651
20
989
997
997
554
554
8345
1859
3875
1968
643
3749
1716
1446
324
263
545
545
154
154
620
3
612
5
76
77
74
45
76
58
58
56
56
85
82
87
85
75
88
89
91
77
82
80
80
75
75
85
67
85
80
85
86
82
50
85
71
71
72
72
91
90
93
90
85
93
95
95
86
88
89
89
88
88
93
100
93
80
90
92
88
80
90
79
79
83
83
94
93
96
94
90
96
97
97
90
93
92
92
94
94
97
100
97
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
                                         91
-------
Table 6-8.  Continued
Data Source
Substrate ;
'• '. > i
!' >
Sample' Size ;
?
...': .. Percentage. of Samples Below :V/'-:.
'SOpg/ft2 ,
75j/g/ft*
; . .,. '»
100^/g/ft2
. -• Interim -.
Clearance
. 'Standard
Window Sill / . . : : i
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)'
HUD Grantee
(Low)"
Atlantic City
Cleveland

Dover
All
Wood
Vinyl
Aluminum
Other
Unknown
All
Unknown
AH
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Aluminum
Unknown
2411
1651
106
44
18
592
689
689
412
412
4906
587
34
3976
309
2170
485
33
1550
102
46
46
93
93
11
4
1
4
2
67
67
85
73
78
61
52
52
68
68
73
61
85
76
56
59
52
76
61
59
63
63
62
62
9
0
0
25
0
75
75
91
77
83
71
63
63
81
81
81
70
88
83
64
71
64
79
73
66
72
72
70
70
9
0
0
25
0
80
80
92
84
83
77
69
69
85
85
85
78
91
87
71
80
74
82
83
70
83
83
80
80
9
0
0
25
0
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
                                         92
-------
Table 6-8.  Continued
Data Source ' ;
• ' • , '• ' • • *;•- .' • ''
• -Substrate
Sample Size:
;';..\:--.. ',K '•.'.}•[ " .>'•• -; Percentage of Samples Below ••..!_> v..-~, :_. . •' ,-.
. ' *• .( .''•... *„; :
..,.••:; '•' .; * • .
;;;:-*:Sp:f(B/ft* '•'••"
' 75\j/g/ft* _'", ;
•:'. ; •-„*,.-*
•>'.•'-.' .-"''•••'' •'- i ''..,',•" '•••'"-'"' '' Window Trough '_ '..'.: /-' '-^ -.
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)*
HUD Grantee
{Low)"
Atlantic City
Cleveland
Dover
All
Wood
Vinyl
Aluminum
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Vinyl
Aluminum
Other
Unknown
1827
373
575
282
597
423
423
371
371
3036
32
1577
1218
209
1341
51
224
953
113
267
267
39
39
260
180
11
2
67
50
38
58
55
48
24
24
58
58
64
50
71
56
49
40
41
74
32
41
76
76
64
64
79
76
91
50
85
58
47
65
63
56
33
33
77
77
72
63
78
65
61
52
47
79
46
50
83
83
67
67
85
82
91
50
94
.. lOOJig/ft*.
'. . ; . ':~i ' - .'.
> • A . : _ .

64
52
71
70
61
37
37
84
84
77
69
83
72
67
61
51
83
56
56
85
85
72
72
89
87
91
50
96
Interim;, ~.
• Clearance <
-Standard

. -.fZ', . . . , * **
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
   Grantees that used
   Grantees that used
200 fjg/tl2 as clearance standard for floor.
100 ,ug/ft2 or 80 jug/ft2 as clearance standard for floor.
                                                      93
-------
Table 6-9a. Percentiles (fig/ft2) for Floor Dust-Loading by Data Source and Substrate for
              the First Site Visit.
''A'..' >T"'v ;:''. r. ;'
1 • .-Data Source ,
*•< -• :? . - -
Maryland
HUD FHA
HUD PHA
HUD Grantee

Percentile-
181
159
200
725
182
418
418
200
200
110
163
87
110
88
76
62
68
189
130
130
190
190
68
61
68
96
!£$*£*£
iPercentite'r
330
297
360
1110
347
871
871
312
312
227
293
189
206
88
168
134
164
359
230
230
320
320
107
61
107
96
. ,-sJsT -£«f i:"»'" *"•"•
••SSjpSSth;::. ;
...a**"***^ v..
"•Percentile
1128
892
1399
1110
2580
3119
3119
600
600
1030
1430
917
754
88
664
505
569
1162
700
700
440
440
317
61
317
96
*  Grantees that used 200 jig/ft2 as clearance standard for floor.
** Grantees that used 100 //a/ft2 or 80 yg/ft2 as clearance standard for floor.
                                                  94
-------
Table 6-9b. Percentiles (pg/ft2) for Window Sill Dust-Loading by Data Source and Substrate
             for the First Site Visit.
Data Source
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)'
HUD Grantee
(Low)"
Atlantic City
Cleveland
Dover
Substrate
All
Wood
Vinyl
Aluminum
Other
• Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
All
Unknown
All
Unknown
Al)
Wood
Vinyl
Aluminum
Unknown
Sample
Size
2465
1687
97
40
18
623
668
668
411
411
4798
636
37
4120
5
2147
507
35
1605
51
51
93
93
18
9
1
4
4
50th
Percentile
28
24
12
10
12
48
52
52
28
28
17
41
20
. 16
29
38
51
16
35
31
31
33
33
443
718
212
443
993
80th
Percentile
181
144
47
84
55
439
271
271
80
80
84
186
58
72
2066
106
129
110
100
150
150
125
125
1263
1263
212
467
2276
90th
Percentile
511
376
116
186
327
1351
714
714
175
175
202
390
477
169
3901
171
215
277
158
380
380
190
190
1624
1624
212
467
2276
95th v
Percentile
1373
864
420
280
406
2785
1678
1678
281
281
418
787
2580
359
3901
356
411
592
312
660
660
250
250
2276
1624
212
467
2276
. 9§th
Percentile'
5646
3593
6619
4078
406
9500
9054
9054
1175
1175
1870
3650
23400
1509
3901
1568
1179
5337
1506
790
790
1700
1700
2276
1624
212
467
2276
  Grantees that used
  Grantees that used
200 //g/ft2 as clearance standard for floor.
100 i/g/ft2 or 80 /jg/ft2 as clearance standard for floor.
                                                 95
-------
Table 6-9c. Percentiles (j/g/ft2) for Window Trough Dust-Loading by Data Source and
             Substrate for the First Site Visit.
'Data Source
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)'
HUD Grantee
it » •*
(Low)

Atlantic City
Cleveland


Dover


• Substrate '
All
Wood
Vinyl
Aluminum
Unknown
All
Unknown
Ail
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
All
Unknown
All
Unknown
All
Vinyl
Aluminum
Other
Unknown
Sample
Size
1985
458
551
286
690
442
442
371
371
3002
34
1626
1341
1
1346
65
238
1043
258
258
39
39
265
185
11
2
67
50th
Porceittilo
92
283
43
51
133
375
375
42
42
30
53
21
46
1727
81
172
17
97
10
10
33
33
7
7
10
136
7
80th
Percentile
844
2160
239
337
1372
2569
2569
85
85
150
189
95
255
1727
285
919
123
312
70
70
170
170
65
77
16
239
12
' 90th ' s
: Percentile
3124
7299
684
983
5695
5810
5810
167
167
396
440
227
722
1727
692
1570
307
680
190
190
530
530
137
152
16
239
68
.• .95th:;-v:
•'Percentile- <
9482
16642
1530
2581
15066
9150
9150
268
268
937
1719
450
1833
1727
1527
1868
1200
1505
450
450
850
850
202
222
106
239
88
;:••: 'i®*'.'--;'
•> Percentile •
49803
59606
10187
38896
63784
50534
50534
1089
1089
7600
7600
1875
12452
1727
7088
17588
5708
9019
2130
2130
2900
2900
1829
2040
106
239
643
  Grantees that used
  Grantees that used
200 /ig/ft2 as
100 //g/ft2or
clearance standard for floor.
80 pg/ft2 as clearance standard for floor.
                                                96
-------
Table 6-1 Oa.  Percentiles dig/ft2} for Floor Dust-Loading by Data Source and Substrate for
               the Passed Clearance Site Visit.
Data Source
Maryland
HUD FHA
HUD PHA
HUD Grantee
{High)*
HUD Grantee
(Low)"
Atlantic City
Cleveland
Dover
Substrate
All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Unknown
Sample
She
2712
1052
651
20
989
997
997
554
554
8345
1859
3875
1968
643
3749
1716
1446
324
263
545
545
154
154
620
3
612
5
50th
Percentile
16
15
18
67
15
40
40
43
43
12
13
10
11
17
15
16
14
16
16
10
10
20
20
5
31
5
9
80th
Percentile
60
58
69
105
56
102
102
93
93
39
44
34
35
60
34
33
31
56
46
46
46
54
54
43
61
43
72
90th
Percentile
98
88
106
116
98
141
141
118
118
70
76
57
72
99
57
52
48
104
79
80
80
78
78
61
61
60
96
95th
Percentile
128
121
135
118
134
165
165
148
148
108
120
92
110
135
90
79
77
140
113
120
120
120
120
82
61
81
96
•:;99th
Percentife
181
180
181
120
184
191
191
185
185
173
180
169
170
179
163
161
151
171
163
170
170
177
177
145
61
145
96
  Grantees that used
  Grantees that used
200 0g/ft2 as
100 pg/ft2 or
clearance standard for floor.
80 fjg/tt* as clearance standard for floor.
                                                97
-------
Table 6-1 Ob.  Percentiles (//g/ft2) for Window Sill Dust-Loading by Data Source and
               Substrate for the Passed Clearance Site Visit.
Data-Source
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)'
HUD Grantee
(Low)"
Atlantic City
Cleveland
Dover
Substrate
All
Wood
Vinyl
Aluminum
Other
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Wood
Vinyl
Aluminum
Unknown
Sample
Size
2411
1651
106
44
18
592
689
689
412
412
4906
587
34
3976
309
2170
485
33
1550
102
46
46
93
93
11
4
1
4
2
•i' 50th
Percentile
21
20
12
11
12
32
47
47
27
27
17
33
19
. 14
40
36
47
14
32
37
30
30
33
33
361
262
212
443
398
80th
Percentile
101
99
37
83
55
119
164
164
68
68
72
116
37
60
128
99
114
77
92
190
90
90
120
120
443
376
212
467
434
90th
Percentile
201
193
66
149
327
238
263
263
150
150
141
224
77
124
213
145
161
127
130
309
180
180
170
170
443
376
212
467
434
95th .r
Percentile.
295
295
167
234
406
315
358
358
222
222
236
309
133
212
296
215
222
277
182
360
270
270
230
230
467
376
212
467
434
99th
• Percentile
457
454
420
325
406
475
481
481
378
378
410
472
477
406
381
372
393
366
368
426
380
380
388
388
467
376
212
467
434
  Grantees that used 200 //g/ft2 as clearance standard for floor.
  Grantees that used 100 pg/ftj or 80  pg/ftz as clearance standard for floor.
                                                98
-------
Table 6-1 Oc.  Percentiles (/tig/ft2) for Window Trough Dust-Loading by Data Source and
               Substrate for the Passed Clearance Site Visit.
Data .Source
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)*
HUD Grantee
(Low)"
Atlantic City
Cleveland
Dover
Substrate
All
Wood
Vinyl
Aluminum
Unknown
AH
Unknown
All
Unknown
Alt
Wood
Vinyl .
Other
Unknown
All
Wood
Vinyl
Other
Unknown
All
Unknown
All
Unknown
All
Vinyl
Aluminum
Other
Unknown
Sample
• V'Size- . ''•
1827
373
575
282
597
423
423
371
371
3036
32
1577
1218
209
1341
51
224
953
113
267
267
39
39
260
180
11
2
67
50th
Percentile
50
88
38
38
53
169
169
41
41
27
49
20
37
50
68
77
16
84 •
69
10
10
33
33
7
7
10
136
7
80th
Percentile
231
338
163
168
257
431
431
81
81
116
155
82
150
207
198
267
78
214
319
66
66
170
170
59
67
16
239
12
90th
Percentile
400
530
308
322
412
598
598
138
138
237
210
164
304
391
346
490
158
352
516
170
170
490
490
111
140
16
239
68
95th
Percentile
571
655
515
456
604
678
678
213
213
389
440
298
464
560
512
646
284
505
581
320
320
540
540
157
169
106
239
88
99th
. 'Perceiitile
753
747
753
778
754
760
760
399
399
691
571
585
722
694
725
769
619
725
735
650
650
648
648
404
• 404
106
239
643
  Grantees that used
  Grantees that used
200 //g/ftz as clearance standard for floor.
100 #g/ft2 or 80 fjg/il2 as clearance standard for floor.
                                                99
-------
 6-3   OBJECTIVE 3: CHARACTERIZATION OF THE CORRELATION BETWEEN
       COMPONENTS SAMPLED FROM WITHIN THE SAME WORK AREA
       The relationships among floor, window sill, and window trough dust-lead wipe samples
 collected from within the same room are an important aspect of examining a clearance testing
 program. One method of assessing the relationships is to estimate the linear correlation
 coefficients. Table 6-11 below displays the Pearson product-moment correlation coefficients and
 the associated sample sizes for the In-transformed loading measurements of individual floor,
 window sill, and window trough samples taken within the same room during a given site visit by
 data source.
       For the first site visit, floor sample results were positively correlated with both window
 sill and window trough sample results. The correlation between window sills and window
 troughs was also positive and was generally the strongest of the three correlations from the first
 site visit.  Second site visit correlations follow the same general pattern as those observed during
 the first site visit, but the linear relationships are not as strong.  The correlations from the third
 site visits are difficult to interpret because of the small number of samples from which they were
 estimated. No correlations are displayed for the fourth and fifth site visits because very few
 samples were available.
       The conditional probabilities of a sample passing or failing a HUD Interim Guideline
 clearance standard are given in Tables 6-12a to 6-12d.  These analyses were conducted on two
 different  sets of data, the first set using all possible paired observations from within the same
 room.  The second set of data restricted the analyses to rooms in which floor, window sill, and
 window trough lead loadings were simultaneously observed.
       As shown in Tables 6-12a to 6-12c, results from contingency table estimates and normal
 theory estimates are consistent. For both types of estimates and both subsets of data included in
 the analysis, the probability that one component passes clearance given that another component
passes clearance ranged from 60% to 100% between floors and window sills, from 66% to 100%
 between floors and window troughs, and from 34% to 100% between window sills and troughs.
       For both types of estimates and both sets of data included in the analysis, the probability
 that one component fails clearance testing given that another component in the same room fails
 clearance testing ranged from 0% to 54% between floors and window sills, from 0% to 59%
 between floors and window troughs, and from 0% to 79% between window sills and troughs.
 The probability that  a sample collected from a window trough is greater than 800 fig/ft2,  given
                                          100
-------
that a sample collected from the window sill is greater than 500 ug/ft2, is fairly high for some
data sources, ranging from 72% to 79% for Maryland and HUD FHA data.
       Another use of the conditional probabilities is to estimate the probability that samples
from a given component type will pass clearance, given that samples collected from both of the
other two component types in the same room had already passed clearance.  For example,

               Pr (Troughs < 800 ug/ft21 Floors < 200 ug/ft2, Sills < 500 ug/ft2)

is such a conditional probability.
       Table 6-12d shows these types of conditional probabilities. To estimate these conditional
probabilities, another set of clearance data was developed, consisting of clearance dust-lead
loadings from rooms in which samples from all three component types (floors, sills, and troughs)
were collected simultaneously during the first site visit.  The conditional probabilities were
estimated using two different statistical methods.  The first method used the clearance data to
derive an empirical estimate of the conditional probabilities (i.e., count the number of rooms in
which all three component types passed clearance, and divide by the number of rooms in which
floors and sills passed clearance). The second method was to assume that the natural log
transformed (In-transformed) lead loadings from floors, sills, and troughs jointly follow a
multivariate normal distribution, and calculate the conditional probabilities.  Due to
complications associated with computing the tail area probabilities for multivariate normal -
distributions using triple integrals, the second method conditional  probability estimates were
calculated via simulation.
       When estimating the conditional probabilities using the second method, tests were
conducted to determine whether the observed dust-lead loading data from floors, sills, and
troughs met the assumption of jointly following a multivariate normal distribution.  Due to the
fact that there were many observations that were below the detection limit in these data, there
were departures from normality in each of the marginal distributions. Thus, the joint data from
all three component types fail the assumption of following a multivariate normal distribution.
The large proportion of non-detects is likely to bias the second method conditional probability
estimates. Therefore, the first method empirical procedures should provide more accurate
estimates of these conditional probabilities.
                                          101
-------
       For those clearance dust-lead loadings from rooms in which samples from all three
component types were collected simultaneously during the first site visit, when none were greater
than or equal to HUD Interim Guideline clearance standards, zero probabilities were shown in
the contingency table estimates column in Table 6-12d.  However, non-zero probabilities for the
normal theory estimates were obtained since they were distributional estimates and were
calculated by simulation.
                                          102
-------
Table 6-11.   Observed Within-Room Correlation Coefficients Between Dust-Lead Loading
               Measurements Collected From Floors, Window Sills, and Window Troughs.
• . ,. ;. ' A"
Data Source

Maryland
HUD FHA
HUD PHA
HUD Grantee (High)'
HUD Grantee (Low)"
Atlantic City
Cleveland
Dover
."; Observed Within Room Correlation Coefficients Between • " ' :, " !
Floors and Sills
P noon ON
n
Floors and Troughs
P flooffTnnglii
n
-, Sills and Troughs
P Skin*.
n -•" •
; First Site Visit
0.534'
0.388 '
0.417*
0.464 '
0.477 '
0.323
0.142
0.155
1739
581
213
3021
1687
26
64
13
0.41 5 '
0.360 '
0.352 '
0.395 '
0.402 '
0.646 '
0.297
0.117
1340
415
192
1558
809
33
24
184
'."•"- r .'.- : • :. Second Site. visit .
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee (Low)"
Atlantic City
Cleveland
0.313'
0.356 "

0.297 '
0.202


86
45
1
101
21
0
0
0.277 •
0.119
-0.793
-0.048
0.526 *
0.099
.
84
61
3
51
16
8
0
0.694 *
0.651 '
0.580*
0.394 *
0.593 "

0.467
1.000
1489
375
201
324
20
0
13
2
•.' . ' . - *
\ - • ." r . . . ,*. . .. "
0.554*
0.200
.
0.696 *
-1 .000


123
55
1
15
2
0
0
: - : : ' Third Site Visit
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee (Low)"
Atlantic City
Cleveland
0.322
-0.042

0.311
1.000

.
7
12
0
18
2
0
0
-0.343
-0.308
.
-0.087
.
.
.
7
14
0
5
1
0
0
0.412
0.586 "
.



.
16
17
0
0
0
0
0
• Statistically significantly different from zero at the 0.01 level.
" Statistically significantly different from zero at the 0.05 level.

*  Grantees that used 200 //g/ft2 as clearance standard for floor.
** Grantees that used 100 pg/ft2 or 80 pg/ft2 as clearance standard for floor.
                                                103
-------
Table 6-12a. Conditional Probabilities for Floors and Window Sills Clearance Testing
              Estimated Using 2x2 Contingency Tables and Normal Probability Theory
              Based on the First Site Visit Data.
Data 'Source

Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)*
HUD Grantee
(Low) * *
Atlantic City
Cleveland
Dover
•""
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High) «
HUD Grantee
(Low) * *
Atlantic City
Cleveland
Dover
Contingency Table*
• Estimates "
All
Possible
Pairs
F, S,andT
within the
same room
. P{F<200
0.927
0.855
0.933
0.961
0.974
0.956
0.921
1.000
0.936
0.843
0.935
0.955
0.900

0.875
1.000
NormalTheory :
Estimates '
All
Possible
Pairs
F, S; and T
within the
same 'room.
S<500) - '
0.939
0.850
0.964
0.937
0.961
0.998
0.951
0.955
0.942
0.854
0.966
0.935
0.919
.
0.930
•
--P(S<500tF<200) .. .: „.;.;••'
0.927
0.882
0.989
0.970
0.971
0.880
1.000
0.692
0.930
0.860
0.994
0.965
0.900
.
1.000
1.000
0.943
0.889
0.998
0.938
0.935
0.957
0.994
0.599
0.947
0.875
0.999
0.953
0.885

0.983

Contingency Table
Estimates
All
Possible
Pairs
F, S.andT
within the
same room'
P(F 2200
0.358
0.313
0.000
0.211
0.132
0.000

0.000
0.379
0.300
0
0.125
0.500


•
• Normal Theory ;."; ..j
'•'• .Estimates :" >';;i
All
Possible
Pairs
F, S, ahdTf
within the L
same room
Sisoo) • ••>;
0.323
0.374
0.277
0.275
0.214
0.009
0.095
0.075
0.3194
0.3873
0.4015
0.3282
0.2882

0.2342

p(s*5oo|Fi20o» :" ;ffi
0.358
0.265
0
0.169
0.142
0.000
0.000
.
0.403
0.272
0.000
0.100
0.5

0.000
.
0.306
0.297
0.020
0.273
0.320
0.232
0.012
0.535
0.3017
0.3453
0.0123
0.2580
0.3746

0.0649
.
*  Grantees
** Grantees
that used 200//g/ft2 as clearance standard for floor.
that used 100 pg/ft2 or 80 //g/ft2 as clearance standard for floor.
                                           104
-------
Table 6-12b.  Conditional Probabilities for Floors and Window Troughs Clearance Testing
              Estimated Using 2x2 Contingency Tables and Normal Probability Theory
              Based on the First Site Visit Data.
i-
Data Source
." * *
• -
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)*
HUD Grantee
{Low}**
Atlantic City
Cleveland
Dover
K- "
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)*
HUD Grantee
(Low) * *
Atlantic City
Cleveland
Dover
Contingency Table
Estimates ;
All
Possible
.Pairs
-
0.941
0.858
0.941
0.964
0.960
0.935
0.956
0.983

0.829
0.668
0.977
0.951
0.918
0.935
0.956
0.977
F, S, and T
within the
same room .
Normal Theory
Estimates ;
All
Possible
Pairs
F,:S; and T
within the
same room
P(F<200|T<800)
0.942
0.873
0.939
0.951
0.900

0.875
1.000
0.951
0.869
0.968
0.936
0.953
0.960
0.964
0.988
0.953
0.878
0.967
0.929
0.939

0.936
.
P{t<800|F<200) "•'.>••
0.823
0.664
0.977
0.965
0.900
.
1.000
1.000
0.819
0.681
0.992
0.930
0.891
0.995
0.967
0.997
0.806
0.676
0.991
0.951
0.799

0.962

Contingency Table •. •
Estimates
All" ^
Possible
Pairs' '•„
F, S, and T
/within the '• "
same room
: Normal Theory /
• . ; Estimates ,v-; •—'
•'•'.>" AiT' V
Possible
Pairs
•rF,:S.jiiid!iT™"
within the
: same; room
P(Fi2bOLTi800)
0.224
0.279
0.200
0.109
0.137
0.000
0.000
0.000
•,•:••',
0.487
0.537
0.083
0.145
0.256
0.000
0.000
0.000
0.238
0.268
0.200
0.000
0.500


.
0.174
0.299
0.175
0.223
0.173
0.506
0.121
0.028
0.1826
0.2935
0.1903
0.2524
0.2657

0.2531

'."' ' P
-------
Table 6-12c.  Conditional Probabilities for Window Sills and Window Troughs Clearance
              Testing Estimated Using 2x2 Contingency Tables and Normal Probability
              Theory Based on the First Site Visit Data.
. Data Source

Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)»
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover
•^T1 • • . -
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High) *
HUD Grantee
(Low) * *
Atlantic City
Cleveland
Dover
. Contingency Table -
Estimates •'''••
•All
Possible
Pairs
F. S. and T
within the:
s&rno room
Normal Theory
Estimates
VVA..-;':'
-Possible
Pairs
rVti.andT
within the
same room
P(S<500|T<800)
0.972
0.928
0.994
0.971
0.941
.
1.000
0.500
0.969
0.930
0.994
0.965
0.900

1.000
1.000
0.978
0.945
0.999
0.949
0.926

0.986
0.339
0.978
0.943
0.999
0.945
0.922

0.990

PtT<80d[S<500)
0.843
0.700
0.975
0.971
0.941
.
1.000
1.000
0.853
0.693
0.973
0.970
0.900
.
1.000
1.000
0.820
0.712
0.991
0.949
0.852

0.963
1.000
0.823
0.709
0.990
0.949
0.815
.
0.963

. Contingency Table
Estimates
..VA»>£'
Possible
Pairs J
< F. S; and T .:
within the •
same room
Normal Theory ' •
Estimates ./
( -All
"Possible
Pairs
F,S;iandT
whhin the
same- room •
P(S iSOOIT 2800)
0.353
0.318
0.000
0.100
0.666
.
.

0.372
0.318
0.000
0.142
0.500
.
.
.
0.251
0.349
0.043
0.202
0.366

0.108
1.000
0.2518
0.3501
0.0350
0.2072
0.3496
.
0.2292

•:•'.--•''- "'••:.'•'"' • P(T s806|S:a500| ^"' .''!••- ^'^-
0.782
0.721
0.000
0.100
0.666
.

0.000
0.767
0.733
0.000
0.125
0.500

.
.
0.765
0.789
0.283
0.201
0.556

0.253
0.297
0.7651
0.7869
0.3359
0.1916
0.5923
.
0.5316

*  Grantees
** Grantees
that used 200 //g/ft2 as clearance standard for floor,
that used 100 ^g/ft2 or 80  //g/ft2 as clearance standard for floor.
                                          106
-------
Table 6-12d. Conditional Probabilities for Floors, Window Sills, and Window Troughs
              Clearance Testing Estimated Using 2x2 Contingency Tables and Normal
              Probability Theory Based on the First Site Visit Data.
f, •' . '
-Data Source

Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)»
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover
Maryland
HUD FHA
HUD PHA
HUD Grantee
(High)*
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover

Maryland
HUD FHA
HUD PHA
HUD Grantee
(High) *
HUD Grantee
(Low)**
Atlantic City
Cleveland
Dover
. Contingency Table
Estimates
;-••• An •
Possible
Pairs
F, S, and T
within the
same room
Normal Theory
Estimates
All
Possible
Pairs
F, S> and T
within the
same room
P
-------
6.4    OBJECTIVE 4: DEMONSTRATION OF THE IMPACT OF COMPOSITE SAMPLING ON
       PASS/FAIL RATES OF HOMES

       Table 6-13 lists the number of housing units investigated by component type and the
number of individual samples collected within a housing unit.  For example, Table 6-13 shows,
for the Maryland data, that there were five floor dust-wipe, window sill dust-wipe, and window
trough dust-wipe samples taken during the first site visit for 88, 77, and 67 housing units,
respectively. As explained in Section 5.4, within a component type, multiple simulated
composite samples were created for housing units containing more than four samples. When
there were four or fewer individual samples from a component type within a housing unit, the
simulated composite sample included all samples. As can be calculated in Table 6-13, for the
HUD Grantee High group  data, five or more floor dust samples were collected from 766 of 2,092
homes (or 37% of the homes). Similarly, five or more samples were collected from window sills
and window troughs, of 3% (66/2,031) and 1% (22/1,697) of the homes, respectively.
       For each combination of component type and composite clearance criterion (Standard
Rule, Standard/n Rule, and 2xStandard/n Rule), each housing unit either passed clearance, failed
clearance, or yielded inconclusive results. The numbers of housing units in these categories are
presented in Table 6-14. Inconclusive results were only possible for those housing units which
contained more than four individual samples; the many possible ways of combining five or more
individual sample lead-loadings into multiple simulated composite samples had the potential of
resulting  in a variety of different outcomes under each of the three composite clearance criteria.
However, Table 6-14 shows that the percentage of housing units with greater than four
individual clearance samples of a given component type that resulted in inconclusive results was
low, ranging from 1.5% (4/263, HUD Grantee Low data, 2xStandard/n Rule) to 19.2% (24/125,
HUD FHA data, Standard/n Rule) for the evaluation of floor samples. (The HUD Grantee Low
data group had 263 housing units and the HUD FHA data had 125 housing units which contained
more than four individual floor samples;  these were shown in Tables D2-5 and B-5,
respectively.) A more detailed description of the methods used to categorize a simulated
composite result as pass, fail, or inconclusive is provided in Section 5.4.
       Table 6-15 presents the performance characteristics used to evaluate the composite
clearance criteria for each component type in terms of sensitivity, specificity, positive predictive
value (PPV), and negative predictive  value (NPV). The calculation of the performance statistics
is discussed in Section 5.4. These performance characteristics are sensitivity, specificity,
                                         108
-------
positive predictive value (PPV), and negative predictive value (NPV), which are defined as
follows:

       Sensitivity = Pr(Fail Composite Sample Clearance | Fail Individual Sample Clearance)

       Specificity = Pr(Pass Composite Sample Clearance | Pass Individual Sample Clearance)

       PPV = Pr(Fail Individual Sample Clearance | Fail Composite Sample Clearance)

       NPV = Pr(Pass Individual Sample Clearance | Pass Composite Sample Clearance)

       If the primary purpose of conducting clearance testing is to prevent potential lead
exposure for residents, then one might choose a composite sample clearance criterion with high
sensitivity at the expense of high specificity. By design, the Standard/n Rule always sacrifices
specificity for sensitivity; the sensitivity for the Standard/n Rule is always 1.000 while the
specificity for this rule is estimated at 0.487 to 0.944 for floors, 0.875 to 1.000 for window sills,
and 0.723 to 1.000 for window troughs.
       On the other hand, the Standard Rule always sacrifices sensitivity for specificity. The
specificity for the Standard Rule is always 1.000 while the sensitivity for this rule is estimated at
0.182 to 0.627 for floors, 0.250 to 1.000 for window sills, and 0.200 to 1.000 for window
troughs.  The 2*Standard/n Rule attempts to maximize both sensitivity and specificity
simultaneously. For the 2>-------
in Section 5.4 was fitted for each combination of component type and composite clearance
criteria to describe the relationship between the probability of passing clearance and the
maximum lead-loading present in all of the sampling locations tested in a housing unit.
       Tables 6-16a to 6-16c list, for floor, window sill, and window trough dust-lead loadings,
and for different composite clearance criteria, the parameter estimates and associated standard
errors for the logistic regression model. In addition, the estimated probability of clearance for a
component when the maximum individual sample lead loading is greater than or equal to 1/2,1,
2, and 4 times the associated HUD Interim Guidelines Clearance Standard [2] is presented.
Figures which illustrate the predicted relationships for each component and each composite
clearance criterion are presented in appendices for each individual data source. Tables 6-16a to
6-16b show that, for most of the data sources, under the Standard Rule for composite sampling,
there is at least a 50% chance that a housing unit will pass clearance testing for floor and window
sill samples even when a lead loading from an individual sample location is twice the HUD
Interim Guidelines Clearance Standard. Only two data sources, Maryland and HUD PHA, show
this phenomenon for the window trough samples.  For the conservative Standard/n Rule the
probability of passing clearance is low when the maximum individual sample lead loading is
greater than or equal to the HUD Interim Guidelines Clearance Standard, but the probability of
passing clearance when the maximum individual sample lead loading is equal to *A the HUD
Interim Guidelines Clearance Standard is between 0.56 and 0.89 for the floor samples, for
example. The 2xStandard/n Rule is shown to be a compromise between the Standard and
Standard/n Rules.  At 1A the HUD Interim Guidelines Clearance Standard, the estimated
probability of passing clearance testing under the 2 x Standard/n Rule is nearly one (1.00). At
twice the HUD Interim Guidelines Clearance Standard, the estimated probability of passing
clearance testing under the 2>
-------
Table 6-13.   Number of Housing Units for Each Data Source that Contained (N) Individual
               Clearance Samples of Each Component Type Based on the First Site Visit.
. Data Source
Number of Individual Samples -> • ' •
. 1
V.
3
4
5
6
7-
8
.,g+;:.
Total;
-..."'. ' • ..• '•• ' • Floor •. . . • ;... . ' .•-•'. : %f '•. ,: -. : , ''.-.:
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)*
HUD Grantee
.. , •*
(Low)
Atlantic City
Cleveland
Dover
65
2
6
236
129
3
0
3
127
7
12
413
124
57
1
2
101
3
32
334
159
33
13
6
108
8
16
343
346
26
10
111
88
36
7
258
201
23
10
17
54
17
8
288
50
10
3
8
54
17
25
142
7
2
0
1
39
23
8
36
2
2
0
3
25
32
5
42
3
3
1
0
661
145
119
2092
1021
159
38
151
.,;..": „•/ , •:.:.'..• ~. . window SUE . _. - .'',:•;. -: -.-.;
Maryland
HUD FHA
HUD PHA
HUD Grantee (High}*
HUD Grantee
* i i **
(Low)
Atlantic City
Cleveland
Dover
61
8
3
263
120
3
7
9
75
5
30
1068
587
9
4
1
121
21
23
501
214
3
13
1
151
8
23
133
40
4
6
1
77
23
7
41
4
1
3
0
46
23
24
20
2
0
0
0
45
22
0
2
0
0
0
0
23
9
1
2
0
0
0
0
13
7
0
1
1
0
0
0
612
126
111
2031
968
20
33
12
;.".'•" . , , • ' Window Trough - " . • . \; .'.'."' •
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)'
HUD Grantee
(Low)"
Atlantic City
Cleveland
Dover
79
13
17
732
495
14
14
32
99
8
15
727
302
79
6
110
106
19
26
170
60
19
3
3
101
21
18
46
13
6
1
1
67
24
32
13
3
1
0
0
28
14
1
7
0
0
0
0
28
6
0
1
0
0
0
0
13
2
1
1
0
0
0
0
15
1
0
0
0
0
0
0
536
108
110
1697
873
119
24
146
* Grantees that used 200 A/g/ft2 as clearance standard for floor.
* * Grantees that used 100 A/g/ft2 or 80 fig/ff as clearance standard for floor.
                                             111
-------
r
            Table 6-14.  Number of Housing Units that Passed or Failed Clearance on the First Site
                         Visit, Based on Individual Sample Clearance Results versus Simulated
                         Composite Clearance Results.
'Data
.. Source
Individual
Sample
Clearance
Results
-: -/ ... "
Maryland
HUD FHA
HUD PHA
HUD
Grantee
(High)'
HUD
Grantee
. »•
(Low)
Atlantic
City
Cleveland
Dover
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
. : •; Composite Sample Clearance Results 	 	 .';_,.
' . . Standard - • :. •
Pass | Inconclusive
Fail

521
42
78
7
88
20
1775
97
909
39
127
16
27
7
140
5
0
36
0
18
0
5
0
63
0
18
0
6
0
2
0
3
0
62
0
42
0
6
0
157
0
55
0
10
0
2
0
3
'.- Standard/n
Pass
Floor .
434
0
38
0
52
0
1622
0
858
0
119
0
20
0
124
0
Inconclusive
.. : i 	 . '
32
0
24
0
9
0
78
0
14
0
3
0
2
0
2
0
Fail
',
55
140
16
67
27
31
75
317
37
112
5
32
5
11
14
11
2xStandard/n . . . '/.
. Pa**.
•,;
494
5
68
0
75
2
1730
8
896
7
126
4
26
3
138
1
inconclusive
.Faih
^-' . • : • . ' _..{ -,; V-; — -
14
28
9
10
8
7
28
62
4
13
1
6
1
0
0
1
13
107
1
57
5
22
17
247
9
92
0
22
0
8
2
9
.;'••• ' , ' '••'••' '. ...::; ' '• Window Sill •••;''.•• ' . :. ••-.' .'':. '• •'--U^--
Maryland

HUD FHA
HUD PHA
HUD
Grantee
(High)'
HUD
Grantee
{Low)"
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
470
35
80
6
103
2
1854
72
897
0
14
0
13
0
1
0
7
0
0
93
0
27
0
5
0
98
0
430
0
70
0
97
0
1797
0
883
18
0
5
0
2
0
2
0
0
22
142
5
46
4
8
55
177
14
467
8
78
0
100
0
1829
8
892
2
9
2
7
2
1
0
2
0
1
125
0
39
1
7
25
167
5
                                                    112
-------
Table 6-14.  Continued
Data
Source
Atlantic
City
Cleveland
Dover
Individual
Sample
Clearance •
Results
Pass
Fail
Pass
Fail
Pass
Fail

Maryland
HUD FHA
HUD PHA
HUD
Grantee
(High)'
HUD
Grantee
(Low)"
Atlantic
City
Cleveland
Dover
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Pass
Fail
Composite Sample Clearance Results . .
Standard
Pass
16
3
32
0
7
0
Inconclusive
0
0
0
0
0
0
Fail
0
1
0
1
0
5
Standard/n
Pass-"'
15
0
30
0
7
0
.. * Window Trouj
337
39
47
6
105
2
1546
33
772
20
114
1
22
0
141
2
0
16
0
7
0
2
0
1
0
0
0
0
0
0
0
0
0
144
0
48
0
1
0
117
0
81
0
4
0
2
0
3
300
0
34
0
104
0
1527
0
757
0
109
0
22
0
141
0
Inconclusive '
0
0
1
0
0
0
Fail
1
4
1
1
0
5
2xStandard/n <
Pass
15
2
32
0
7
0
Inconclusive,.
0
0
0
0
0
0
•'Fan
1
2
0
1
0
5
h ' •. . . .•,.;.: ;.
10
0
4
0
1
0
1
0
1
0
0
0
0
0
0
0
27
199
9
61
0
5
18
151
14
101
5
5
0
2
0
5
324
6
45
2
105
0
1533
2
762
0
111
0
22
0
141
0
5
15
1
5
0
1
0
1
0
0
0
0
0
0
0
0
8
178
1
54
0
4
13
148
10
101
3
5
0
2
0
5
   Grantees that used 200 /sg/ft2 as clearance standard for floor.
   Grantees that used 100 j/g/ft2 or 80 ;/g/ft2 as clearance standard for floor.
                                                     113
-------
 o
 s

 o
(0
 CD

Q
 o

 (D

UJ
JO


 <0
*^


O
 to
 a
 «
 o
 a.


 o
 o

 o
•c
 o
ts
O




1
(Q
(O

 0

2

.2




i£
• '•«
:e
w
w







e'
"g
1
w







;• )•
1


"I
e
a
5




i.



n.

^
&

:i

e
- «o
1
"5
C
(0
; ;-z
"I

^
3
t"

CO

• -fr
-'•'C'
•9
W
" >
o.

• •: v
•O.

•»
"Z5
M

S
C
* o
1
o .
to •
« *
S
-.












•V"
•• •' i




g
11






'• ..':


•••
.


^




o
en
on
O
CM
en
CO
O
00
S
o

to

0

o
8
00
r-
o

CO
CO
CO
o



8
q

in
CM
CD
o

o
O
q

O
8

CO
5
o

•a
c
CO
2-
(D
2
8
o
*""
CO
CD
O
CM
I*.
CO
O

in
CO
0

o
8
o
CO
o

Is*
CO
d



o
8

CO
en
o

o
o
q

O
o
q


CM
CO
o

I
UL
Q
I
h-
co
0
in
00
o
CM
in
00
O

0
r-.
O

Q
8
CO
tn
0

CO
in
0



O
o
q

in
CO
0

8
q

8
q


en
o

i
a.
a
i
in
CD
CO
0
to
CO
en
0
in
Is.
a>
o

CD
ps.
O

0
o
q
g
CO
0

CO
0



o
0
q

CO
o>
0

o
o
q

o
o
q

in
?
O

"£
10

O
ej

JO
«"•
rs.
CO
00
O
8
o
*°
CO
to
CD
O

CM

0

O
8
00
CO
to
o

*
d



8
q

en

O

O
o
q

O
8

CM
CO
O

1
-2
I

o

O
o
q

o
8

in
m
CO
0

•D
c
CO
CO
o
8
*~
0
O
o
*~
in
CD
o

CD
CO
0

!
CM
O
O)
o

in
PS
CO
d



o
8

o
CO
en
0

8
q

8
q

r-
00
in
o

I
LL.
a
o
8
*"
in
CO
O

en
O

in
fs.
CO
O

8
q
i
0

CM
d



o
8

CO
en
0

1

|

in
CM
to
O

<
GL
0
CO
en
en
O
O
CO
O
rs.
CO
en
o

«
CD
0

O
8
CO
CD
fs.
o

en
to
en
o



0
8

CO
CO
en
d

O
8

8
q


in
in
O
0)
a>
c
-------






































•a
0
3
1
O
u
in
(D
o
S
eo



-'


._c
•P
•«
•o
a
CO
X
. CM





;



•B
'.•8
c

.1
S
CO
•

'' f
r
,


£
• "£'
} 'Z,

^^
(X
' *%•


- ^




" *?'
:•>•
i
- c
«
1
»!»'*
-• Gk
-**

*"';* •"-<
• -S-
T»-
a1
>•
•S.
'.O'
•m.
*• ;
£

"'I;
••:
"3
'.*
"^
1
I
.'.-;;
2 "
o
u
S
9" -

•..





,.<

-


,

L



°;
-^





' .JB
£

"'1
J

".-:
. •



*">

; i
^ •




CM
00
O
fx
to
CO
o



,_
to
o>
0


00
0

o
o
q
-
00
o


0
00
o

o
8
^™
(O
o>
00
0
o
o
q
1


«M
Px
0


[Maryland
Px
10
CO
o
CM
OO
CO
o



(V.
10
o>
o


00
00
0

o
o
q
Px
00
o


n
IN
fx
o

o
o
o
^~
^
09
CO
0
0
-
1


00
Px
0


u.
o
13
I
0
8

0
o
o




o
0
o


o
§
0

o
o
q
o
O
o
^


0
CD
O

o
o
o
*~

00
CD
0
0
o
q
1


o
S
0


0.
O
I
CD
O)
CD
O
CD
f_-
CD
O



CM
CD
CD
0

O
00
CO
0

o
o
q
CO
O)
00
O


00
00
CO
o

o
8
*~
00
Px
00
0
8
q
o
o
q


10
Px
0

0)
09
S
II
0
o
o

0

CD
o



rx
00
CO
0

o
0
q
^*

8
q
00
Px
oo
O


eo
CO
o

o
8
*~
in
Px
o>
0
8
q
o
8


i
0

0)
S
Q ?
§3
o
o
o

in
CM
to
0



^
Px
CO
o

o
o
0
q—

!
8
10
o


(0
to
CO
o

8
o
^

o>
en
o
8
q
0
o
q


8
o

>
O
£
C:
JO
O
O
o

o
o
o




o
o
0


o
o
0
f—

8
q
g
o
*-


o
o
o


8
o
^
o
8
"~
8
q
0
o
q


8
0



Cleveland
o
o
O

o
a
o




o
o
0


o
o
0
i—

O
8
8
o
»-


o
o
o
T"

0
o
o
*"
(O
8
o
8
q
0
o
q


0
o
(0
o


o
O
115
-------
 .
O
cn
cn
o
i
O
to
o
9

00
IN
O
en
CM
fx
ID
CM
CO
TO
co
in
Standard

o
o
O
o
0
co
CD
O
CD
cn
O
8
o
o
CD
S
o

rx
in
O
S
r»
\t)
§
en
3
2xStandard/n
|| Maryland
0
o
c
0
CN
o
o
in
rx
O
g
o
o
CO
g
0

CD
g
o
cs
01
CO
CO
in
cn
CO
Standard/n

O
O
,_
o
rx
cn
o
8
0
8
o
o
eo
O
9

o
rx
^
cs
00
in
es
in
CO
•o

8
o
o
o
CD
CO
O
%
0
in
O
o
O
in
cs
O
O

CD
5
^
o
o
CD
00
\o
00
co
2xStandard/n
u.
O
3
8
o
o
o
o
0
0
CD
O
cs
O
O
CD
in
O
o

in
CO
^
in
ct

0
8

8
o
CD
0
0

s
^
CD
**"
CD
1
Standard

8
0
S
o
co
o
co
cn
0
I
O
o
es
O
O

cs
rx
O
en
en
*
CM
"
2xStandard/n
1
Q
I
8
o
8
o
8
o
CD
in
0
co
O
O
CD
O
O

rx
^
00
CO
CO
to
GO
in
cs
in
Standard/n

o
o
o
in
O
CD
cn
O
cn
en
0
8
o
in
o
O

00
CO
o
o
en
cs
to
10
cs
CO
Standard

o
o
o
CN
o
in
o
O>
o
i
o
CO
o

cn
CO
o
in
CN
CO
CO
CN
00
CO
2xStandard/n
1 HUD Grantee (High)'
c
o
o
o
o
s
o
en
eo
O
8
o
00
9

I
o
§
CD
CS
cn
CO
cs
CS
CD
Standard/n

8
o
CD
in
0
cn
o
cn
CD
O
8
o
to
o
o

00
00
to
o
CO
cn
CO
in
in
00
en
Standard

O
o
o
8
0
oo
CO
o
CD
O
CO
O
O
o
8
o

cn
CO
CD
o
00
8
CD
3
1
.C
•o
CO
•a
x
rs
1 HUD Grantee (Low)"
§
o
o
o
o
CO
0
0
s
o
in
o
o
o
CN
in
o
o

8
CO
o
CN
O
CD
cn
00
in
00
Standard/n

o
o
CO
o
oo
cn
0
»
o
|
o
0
0

CS
in
^
in
CD
CD
O
CO
•g
OT

8
o
o
o
00
0
8
o
CD
8
o
CM
0
0

§
*™
O>
CO
cs
CM
o
CO
I
•o
I
X
CM
I Atlantic City
8
o
8
o
S
0
cn
CO
O
CN
o
o
S
o
o

IX
rx
^
00
0
PX
CO
cn
Standard/n

8
o
CO
cs
0
8
"
8

CN
8
o
5
o
o

CN
IX
O
-
fx
CO
CO
r-
es
co
Standard

8
o
8
o
8
"
o
o

01
in
cs
O
CO
00
cs
o

§
o
00
CO
o
cn
00
00

a>

c
                                                                  116
-------
to
(0
"•I
I5"
i 1 i
11
S i „
11
Si
5 .3- *
•O ^ "
I
" 1
> 5


(Mi 1
' 1
' 1
£ *
ijjk
>. Nu(nb«r'6
*, .jwlth'Siri
'Composite;
l*V$®
£
i J.BJ
B
i 'Data Source : :-
11
«!
11

§1
z!
o •=
*«
*s
-*
83
**>
i

c£
*
£•
:!.
4.
-
O
O
o
in
O
CO
cn
o
cn
CO
o
§
o
CO
o
o
9
§
o
o
rx
CM
to
CO
CD
in
S
Standard

8
o
o
o
i
o
CD
CD
O
CN
8
o
CM
O
9
CO
o
§
rx
CD
CM
in
rx
2xStandard/n
Maryland
8
o
c
o
§
o
§
o
i
o
CM
O
9
<*
o
8
in
rx
CD
0
CO
Standard/n

m
o
o
o
IX
O
to
o
00
o
i
o
8
9
IX
s

•^
CM
in
in
e»
CO
co
2xStandard/n
i.
u.
O
X
c
o
o
o
o
o
o
g
o
1
o
CO
CM
O
9
IX
CM
in
CO
in
O
rx
Standard/n

8
o
to
co
0
co
cn
o
8

IN
O
O
O
O
O
9
CO
CM
CO
in
**
in
in
o
Standard

8
o
0
0
in
O
09
CO
O
I
O
5
9
g
CO
"
IX
rx
CO
00
o
o
c
1
1
a.
o
I
8
o
8
0
00
o
o
cn
0
to
8
o
CD
O
9
00
*"
1
"""
CM
IX
CO
Standard/n

8
o
o
in
0
ix
CD
o
CD
co
0
8
o
IX
o
o
9
s
<*
o
CO
cn
to
CO
CO
CO
CD
CM
0)
Standard

8
o
8
0
5
o
8
o
CM
8
o
rx
o
9
CD
CD
CO
O
«s
CO
s
rx
n
CO
2xStandard/n
HUD Grantee (High)'
8
o
8
0
IX
O
O
cn
0
CM
8
o
«
o
9
IX
CO
CO
o
to
ra
CM
fx
CD
rx
Standard/n

8
o
in
in
0
CO
cn
o
8

8
o
0
9
rx
CO
o
in
rx
CO
m
en
Standard

8
o
8
0
in
O
0>
cn
0
i
o
CO
5
9
CO
CO
*"*
-
cn
o
CO
cn
00
c
•g
IB
-a
S
Si
X
CM
HUD Grantee (Low)"
8
o
8
0
cn
o
to
co
0
CM
8
o
CO
0
9
00
o
-t
in
IX
IX
CO
CO
CO
co
Standard/n

8
o
CO
0
in
en
o
OJ
cn
0
cn
8
o
§
o
9
CN
in
o
CO
CD
IX
-
9)
Standard

8
o
M
0
0
o
o
co
0
m
8
o
S
o
9
Tf
in
"
CO
to
CO
«
-
2xStandard/n
Atlantic City
8
o
8
0
in
o
CO
CO
0
IX
8
o
0
9
CD
CD
in
CM
CM
CO
CO
in
in
in
Standard/n

8
o
in
CO
0
8
"
o
o

co
00
in
rx
in
o
9
CO
UJ
»
CM

rx
s
in
,-
CM
Standard

8
o
in
CO
0
8
"
o
0

CO
CO
CD
00
in
IX
in
O
9
CO
UJ
CM

rx
§
in
-
CM
CO
2xStandarrJ/n
1 Cleveland
8
o
8
0
00
o
CD
CO
0
cn
8
o
in
0
9
in
CM
in
CM
CN
IX
rx
in
CM
8
Standard/n

8
o
8
0
8
"
o
0

-------
-;
J|;
3*X\*
:|1 --
:||-
11 -
~^~g w -
i
-J& -
"£~
: | "
t •*
:'T%
•*!• ;
'vlf-r
?%.
g g"
1 1 I**'
5 £ S to
»«
- - ,
- ,-i
V!
, Q
*5* *
. I
o|
* £
" s
II
?•!
gl
;.*
II
^*
"2
T


,g:
M"
*' *
«£
"2
*_%g
.»_
•a."
— r"
8
o
CM
in
O
CO
0
en
en
0
o
O
o
s
o
9
§
O

CD
t
CD
CO
Standard

8
o
o
o
0
o
en
O
§
O
CD
8
9
in
to
O
IV
CO
in
CD
GO
CO
2xStandard/n
l| Maryland
8
o
o
o
o
g
o
00
oo
o
8
o
o
o
co
to
o
CM
CO
ID
CD
CM
CM
O
CO

o
o
o
to
o
to
en
0
*
o
i
o
§
o
o
00
00
«•*
«
CO
03
CO
Standard

8
o
o
o
co
o
co
en
O
g
O
o
to
o
o
9
CO
CO
CD
^
rv
CM
CO
sn
in
rv
c
S
in
X
CM
1
o
X
o
o
o
o
o
o
o
o
co
0
8
o
o
9
CO
IV
CO
^
co
CM
in
o
co
Standard/n

CM
O
O
CO
co
o
en
en
O
o
0

CM
8
o
o
o
o
CO
in
C"J
CM
_
,-
0
•o

in
•g
to

o
o
o
8
o
«o
o
8

8
o
o
CO
0
9
rv
CO
00
*™
CM
CD
CM
to
in
CO
in
2xStandard/n
| HUD Grantee (High}'
8
O
o
o
o
rv
o
8

8
o
o
en
o
o
o
09
CM
CO
CO
en
rv
CM
in
Standard/n

o
o
o
CM
O
O
CM
cn
O
o
0

8
o
§
o
9
rv
CO
*""
rv
in
00
CO
CM
en
IV
Standard

o
o
o
8
o
o
CM
O
8

n
o
o
o
rv
0
9
CM
CM
0>
in
CM
£
CM
co
rv
2xStandard/n
o
«
^
ID
5
a
X
8
0
8
o
S
o
en
en
O
i
O
rv
o
o
00
cn
'"f™"
8
-
in
rv
in
rv
CO

8
o
8
o
0
o
~
o
o

00
o
in
in
CM
CO
00
O
o
CO
UJ
en
in
CO
cn
CO
CM
'
in
T~
Standard

o
o
o
8
o
,v
0
in
O
§
O
CM
O
9
o
CO
CM
CO
CM
CD
fv
CO
-
C
1
CO
X
CM
| Atlantic City
o
o
o
8
o
o
o
a>
00
o
(V
O
0
o
rv
o
o
o
CM
in
CO
CM
CO
0
en
O
Standard/n

8
o
8
0
o
o
o
o
0

to
UJ
CM

in
rv
-------
7.0    REFERENCES
[1]    U.S. Environmental Protection Agency, "Lead; Identification of Dangerous Levels of
      Lead; Final Rule," Federal Register, 66 FR 1206-40,40 CFR 745 subparts D, L, and Q;
      January 5,2001.

[2]    U.S. Department of Housing and Urban Development, "Lead-Based Paint: Interim
      Guidelines for Hazard Identification and Abatement in Public and Indian Housing,"
      Office of Public and Indian Housing, 1990.

[3]    U.S. Department of Housing and Urban Development, "Guidelines for the Evaluation and
      Control of Lead-Based Paint Hazards in Housing," Office of Lead-Based Paint
      Abatement and Poisoning Prevention, HUD-1539-LBP, July 1995.

[4]    U.S. Environmental Protection Agency, "Lead; Requirements for Lead-Based Paint
      Activities in Target Housing and Child-Occupied Facilities; Final Rule," Federal
      Register, 61  FR 45777-4583 O.August 29,1996.

[5]    U.S. Environmental Protection Agency, "Guidance on Identification of Lead-Based Paint
      Hazards," Federal Register, 60 FR 47248, FRL-4969-6, September 11,1995.

[6]    National Institute of Building Sciences. Guidelines for Testing, Abatement, Clean Up,
      and Disposal of Lead-Based Paint in Housing, Washington, D.C.  1989.

[7]    NIOSH (National Institute for Occupational Safety and Health), "Laboratory Evaluations
      and Performance Reports for the Proficiency Analytical Testing (PAT) and
      Environmental Lead Proficiency Analytical Testing (ELPAT) Programs," U.S.
      Department of Health and Human Services, Centers for Disease Control and Prevention,
      DHHS (NIOSH) Number 95-104, November 1994.

[8]    U.S. Environmental Protection Agency, "National Lead Laboratory Accreditation
      Program; Notice of Availability of Requirements," Federal Register 58 FR 38656, July
      19,1993.

[9]    U.S. Department of Housing and Urban Development, "The HUD Lead-Based Paint
      Abatement Demonstration (FHA)," Office of Policy Development and Research, HUD-
      1316-PDR, August 1991.

[10]   U.S. Department of Housing and Urban Development, "The HUD Public Housing Lead-
      Based Paint Demonstration in Omaha," letter and report to Albert Gore, Jr., President of
      the Senate from Henry G.  Cisneros, Secretary of U.S. Department of Housing and Urban
      Development, April 30,1993.
                                         119
-------
[11]  U.S. Department of Housing and Urban Development, "The HUD Public Housing Lead-
      Based Paint Demonstration in Albany," letter and report to Albert Gore, Jr., President of
      the Senate from Henry G. Cisneros, Secretary of U.S. Department of Housing and Urban
      Development, November 23,1994.

[12]  U.S. Department of Housing and Urban Development, "The HUD Lead-Based Paint
      Abatement Demonstration, Cambridge Housing Authority Case Study," February 1995.

[13]  The National Center for Lead-Safe Housing and The University of Cincinnati Department
      of Environmental Health, "Evaluation of the HUD Lead-Based Paint Hazard Control
      Grant Program," Fifth Interim Report, Progress as of September 1,1997, March 1998.

[14]  Dewberry & Davis and Speedwell, Inc., 1991 (HUD FHA and PHA Data Source).

[15]  Breen, Joseph, J. and Cindy R. Stroup, ed. "Information Collected in the HUD Abatement
      Demonstration Program and Its Application in Planning a Follow-on Study." Lead
      Poisoning:  Exposure. Abatement. Regulation, pp. 37-48, CRC Press, Inc. 1995.

[16]  Maryland Department of the Environment, Lead Enforcement Group, Susan Guyaux,
      Baltimore, MD, 1995 (Data Source).

[17]  University of Cincinnati Department of Environmental Health, 1999 (HUD Grantee Data
      Source).

[18]  Atlantic City Housing Authority, James P. Walsh, Atlantic City, NJ, 1995 (Data Source).

[ 19]  Cleveland Lead Hazard Abatement Center, Stuart Greenberg, Cleveland, OH, 1995 (Data
      Source).

[20]  Dover Housing Authority,  Jack Buckley, Dover, NH, 1995 (Data Source).

[21]  Anderson, T.W., An Introduction to Multivariate Statistical Analysis. John Wiley &
      Sons, Inc. N.Y., 1958.

[22]  U.S. Department of Housing and Urban Development, "Requirements for Notification,
      Evaluation and Reduction of Lead-Based Paint Hazards in Federally Owned Residential
      Property and Housing Receiving Federal Assistance; Final Rule," Federal Register, 64
      FR 50139-50231, September 15,1999.

[23]  U.S. Department of Housing and Urban Development, "Requirements Notification,
      Evaluation, and Reduction of Lead-Based Paint Hazards in Federally Owned Residential
      Property and Housing Receiving Federal Assistance; Proposed Rule," Federal Register,
      61 FR 29169-29232, June 7,1996.

[24]  U.S. Environmental Protection Agency, "Lead; Identification of Dangerous Levels of
      Lead; Proposed Rule," Federal Register, 63 FR 30301-30355, June 3, 1998.

                                        120
-------
            APPENDIX A
Maryland Department of the Environment
                A-1
-------
                                     APPENDIX A
                 MARYLAND DEPARTMENT OF THE ENVIRONMENT

       The Lead Enforcement Group within the Maryland Department of the Environment
(MDE) has been actively conducting post abatement clearance testing since 1988. When the
Lead Enforcement Group first started its clearance testing program, there were no protocols for
sampling or standards developed for the purpose of clearance testing. Based on scientific
evidence, this program developed protocols for the collection of dust wipe samples from floors,
window sills, and window troughs in rooms that had received abatement. The clearance
standards of 200 ng/ft2 for floors, 500 ng/ft2 for window sills, and 800 ng/ft2 for window troughs
also originated from within the MDE Lead Programs [16]. These standards were based on pilot
data, and were designed so that a work area would pass clearance through reasonable cleaning
efforts on the part of the lead contractor. These clearance standards also represented levels of
post abatement lead on floors, sills and troughs that were far below the levels of lead that would
typically be observed on these components prior to abatement.  As a result of adopting these
standards, Maryland had fewer cases of re-poisoning following abatement.
       The Lead Enforcement Group has archived data on dust-lead loading results since they
started conducting clearance sampling in 1988. These data existed only in hard  copy form, and
were made available to EPA for the purposes of this project. All clearance testing sample results
collected between January 1,1991 and January 30,1995 were entered into an electronic database
for the purposes of this investigation. Thus, the statistical analysis of data from  MDE represents
these four years of clearance testing results throughout the state of Maryland.
       If the results from a dust-lead loading sample exceeded the clearance standards, the area
had to be recleaned and retested until acceptable results were obtained for each unit.  There are
no records of additional testing for approximately 23% of the units which failed clearance.
These losses to follow-up may be attributed to families moving, litigation, poor  record keeping,
or on-going clearance testing. Since the Lead Enforcement Group within MDE  is a state
regulatory program, and not a research program, this loss to follow up within the dataset could
not be avoided.
                                          A-2
-------
       Statistical results presented in the main body of this report summarize the four years of
clearance testing as a single unit. Descriptive modelling of each calendar year separately is
presented in Section A-2.

A-1.   Objective 1: Characterization of the Number of Individual Samples, Work Areas, and
       Housing Units That Pass or Fail Clearance Testing Standards
       As part of the State of Maryland clearance testing program, 7967 individual dust wipe
samples were collected from floors, window sills and window troughs within 3502 rooms in 706
residential units during the four years from January 1991 through January 1995. Table A-1
presents the number of individual samples, work areas and residential units that passed or failed
clearance testing within each combination of component type and site visit. Approximately 90%
(7200/7967) of the dust samples were collected during the first site visit to a residential unit.
Although 87% (6289/7200) of the dust samples fell below the clearance standards of 200 ug/ft2
for floors, 500 ug/ft2 for window sills and 800 ug/ft2 for window troughs, only 80% (2762/3450)
of the rooms and 57% (402/706) of the residential units passed clearance on the first site visit.
This increase in the failure rate from individual samples to rooms and residential units is
attributable to  the fact that if any individual sample exceeds the standard and fails clearance, then
both the room  and residential unit also fail clearance.  Of the 304 residential units that failed
clearance on the first site visit, 168 (55%) of these residential units were revisited for a second
clearance testing.  Eventually, 543 of the 706 residential units (77%) are known to have passed
clearance testing.
       The failure rate  for individual samples during the first site visit was highest for window
trough samples at 21%  (408/1,985), followed by window sills at 10% (254/2,465) and floors at
9% (249/2,750). This pattern is reflected in the failure rates for residential units based on results
from individual components: 37% (199/536) of the residential units would have failed based on
the results of window trough samples, 23% (142/612) of the residential units would have failed
based on the results of window sill samples, and 21% (140/661) of the residential units would
have failed based on the results of floor samples.
                                          A-3
-------
Table A-1. Clearance Testing Results by Individual Sample, Room, and Dwelling Unit for
          the State of Maryland Regulatory Program (January 1991 - January 1995}

  First
             Floor
             Sill
            Trough
             All
2501
249
2750
2413
243
2656
521
2211
254
2465
2121
250
2371
470
1577
408
1985
1455
399
1854
337
6289
911
7200
2762
688
3450
402
140
142
199
304
661
612
536
706
 Second
            Floor
             Sill
           Trough
             All
 185
 22
 207
 178
 22
 200
 82
 169
 25
 194
 164
 23
 187
 67
227
 48
 275
 213
 47
 260
 91
581
 95
 676
 347
 75
 422
132
 15
 11
 24
 36
 97
 78
115
168
  Third
            Floor
             Sill
           Trough
             All
 23
        23
         23
                23
               10
 28
        30
         20
                22
  19
  8
  27
  19
        26
 70
 10
  80
  39
        48
         14
                        10
                        12
                19
  Fourth
            Floor
             Sill
           Trough
             All
 10
         11
Total
Floor
Sill
Trough
All
2712
2411
1827
6950
271
281
465
1017
2983
2692
2292
7967
2558
2243
1649
3077
131
166
256
425
2689
2409
1905
3502
592
525
434
543
70
89
107
163
662
614
541
706
                                         A-4
-------
A-2.   Objective 2: Characterization of the Distribution of the Dust-lead Loadings.
       Geometric Mean Dust-lead Loadings. Variability Within a Housing Unit, and
       Variability Between Housing Units
       Preliminary assessment of the data indicated that the distribution of dust-lead loading
clearance sample results was highly skewed.  A natural logarithm transformation was applied to
the data.
        Table A-2 gives the geometric means for each component type by site visit for data
collected in Maryland between January 1991 and January .1995. A 95% confidence interval for
the geometric mean dust-lead loading for each combination of site visit  and component type is
also provided.  In general, the within-house (or room-to-room) variability is smaller than the
between-house variability; however, all of the variance components are  within a single order of
magnitude. The length of the confidence intervals for any given component type increases
successively from the first site visit to the third.  This increase in length is due primarily to the
decrease by orders of magnitude in the sample size used to estimate the  variance components
between site visits.
       The geometric means and their 95% confidence intervals are used to compare trends
between site visits for a given component tested, and to compare average results between
component types within a site visit. If the 95% confidence intervals on two geometric means do
not overlap then these two geometric means are statistically significantly different at a level less
than 0.05. For the Maryland data, the floor dust-lead loading clearance geometric means are less
than the window sills, which are in turn less than the window troughs within a site visit.
However, the geometric means for a given component are not significantly different across the
three site visits based on comparing the 95% confidence intervals.  These results also generally
hold across individual years, as seen in Tables A-2.1 through A-2.5.
       Figures A-l and A-2 contain box and whisker plots that present the distribution of dust-
lead loadings from the first and passed clearance visits by component type. Figures A-3 to A-5
contain box and whisker plots that present the distribution of dust-lead loadings from the first
visit for floor, window sill, and window trough, respectively; Figures A-6 to A-8 contain box and
whisker plots that present the distribution of dust-lead loadings from the passed clearance visits.
                                          A-5
-------
 o>
 Q.
 2
 o
 Q.
•o

 CO
•M
"w
w

m
 10
 o
*<
 CO
 E
'•5
LU
<->
 e

 i
o
 S  «
 C  Q-
•o  g
 §1
 o  >
o  2
 row
 o  £
 O  i.
                                   CO
                00
                CO
           n
           CO
CO
in
«N
co
                                   r>
                                        CM
                                        CM
                CM
                CM
                                                   co
                co
                CO
                                in
                                        r>
                                        CO
                     CM
                     01
                     (O
                     
                                                        D
                                                        O
                                O
                                O
                                                   05
                                                   to
                                                                  
-------
 CD


 o
 Q.


 O
U

•o

 (8

4j

*»

>

 0>
+••

35


A

 en
 0
 v>
Ul
i->

 V


 i


L

 »  £
 u  0)


1  =
 o  «
    >T3
    ®


1-
«<

 §2
S  il
 a  «
•c  .2
*••  f
 a>  t

I  &
2|
O  O-

g>?
•I  «

S  £"
H  5
 «  *-

II
o  h-


CN
a
                               co
                                    03
                                    in
                                         CM
                                         o
                                         co
                                              CO
                    CO
                    en
                    CM
                                                        r-
                                                        r«.
                                                        co
                                                             co
                                                             co
                                                                  ;

                              o
                                                                  
                              _o
                              u.
                                                             g
                                                  t3


                                                   O
                                                   U
                                                   m
                                                  CO
                                                A-7
-------
 0>

 o
 a

 o
u
•o

 a
w

a
 m
LLJ
                                GO
                                CO
                                    00
                                                 to
                                                          a>
                                                          LLJ
                                                          GO
                                                          a>
                           CO
                                n
                                01
                                             CQ
                                                 (O
   80)
   0)

1  c
 (o ZI
> ^
•o
                                CO
                                O
                                        CO

                                        d
                                                      o
                                                      o
                                                          IT)
                                                          (O
o
o
   2
-2 S
** 35

                                                 CO
                                                 o
                                                          es
                                oo
                                tn
                                    

                                                 o
                                                      o
                                                      o
                                                          a>

                                                          o
                                             -o

                                             o
                                           A-8
-------
 03

 O
 a

 o
O
•o

 a
 f
*«
 5
m
 n
 a
 »*
 
O I-
M
                                 CM
                          
                               co
                                    co
     o

     CM
                                              in
                                                        co
                                           o
                                           in
               O
               CM
                                                          CO
                                                                     CO
                                                                          in
00
in
     (O
     in
          O
          tn
                     CA
                                                          
                                                          (O
                              CO
                                              CO
                                                   co
o
o
          CO


          I
                o
               _o
               u.
                                                           2
                               o
                               _o
                               u.
     O5


     O
                                                         3
                                                         O
                                                      8
                                                      ID
                                                     CO
                                    •g

                                    IE
                                                    o
                                                   A-9
-------
 CD

 O
 O
 O
 •o

 5
 s
 w

 J3
 W
UJ

 O  0)
 u  «
'=  .2

 •  I

 I  &
 S  s
•   «
    -
 O  g

 03  *-


II
o  £
 a
                               oo
S
                                   CO
                                   00
                                   00
                                       CO
                                       CD
                                   0)

                                   il
                                           in
                                               in
                                               oo
                                                   in
                                               •o
                                               o

                                               I
                                                       8
                                                       CM
                                                      CM
                                                      CO
                                                           a>
                                                          a>
                                                          Oi
                                       O
                                       a.

                                       o
                                       O
                                       •o

                                       §
                                       w


                                       J3

                                       to
                                       o
                                                                      «o
                                                                     Ul
                                                                      e

                                                                      o
                                                                      O id
                                                                      O 0>
                                                                      o o>
                                                                      5 8
                                       u  «
                                      •= .£
                                      ** 4^
                                       o  fc
                                           Ea>
                                       ^  o.
                                       o  o
                                                                      «
                                                                      o  i-

                                      to
                                      tN

                                      <

                                       «
                                      £
                                       a
                                                                                                           CD
                                                                                                                in
                                                                                                           CD
                                                                                                                CM
                                                    A-10
-------
             100.0
              10.0
               1.0
               0.1
                    Fkns      WMmrSfc    Wndw Troughs     Ftas       WMwSfc    WMw loughs

                              First Visit                         Passed Clearance
                                              Surface
Figure A-1.    Comparison of In-Transformed Dust-Lead Loadings from the First Site Visit
              vs. Passed Clearance Results on an Expanded Scale.
                                          A-11
-------
             800


             700
          eg
          1  500
          £
300


200


100


  01
                  Floors
                               Window Sills
                                 Surface
Window Troughs
Figure A-2.   Box and Whisker Plots of the Distribution of Dust-Lead Loadings from the
              Passed Clearance Data.
                                           A-12
-------
             10000
              1000
          CM
          <
               100
          D
          Q.
                10
                      AU
Wood
Vty
Aluminum
Other
Unknown
Figure A-3.   Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the First Site Visit.
                                             A-13
-------
              1000000
               100000
                10000
                1000
                 100
                  10
                                  Wood        Vnyl       Aluminum
                                                 Substrate
Other
Unknown
Figure A-4.    Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
               by Substrate for the First Site Visit.

                                             A-14
-------
             100000
              10000
          CM
          <
          0)

          1
          3

          £
               1000
100
                 10
                                 Wood        Vhyl      Alunmum       Oiher       Unknown


                                               Substrate
Figure A-5.   Box and Whisker Plots of the Distribution of Window Trough Dust-Lead

              Loadings by Substrate for the First Site Visit.
                                            A-15
-------
             1000
          CM   100
          CD
          o.   10
                    Al
Wood
Vnyl
Alutrtntm
Other
Unknown
Figure A-6.   Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the Passed Clearance Visits.
                                           A-16
-------
             1000
          CM   100
          01


          I

          3

          n
          Q.
10
                               Wood        Vnyl       Aluminum       Other


                                              Substrate
Figure A-7.   Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings

              by Substrate for the Passed Clearance Visits.
                                            A-17
-------
             1000
          w   100
               10
                    AO
Wood
Vnyl
AMnun
Other
Unknown
Figure A-8.   Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the Passed Clearance Visits.
                                           A-18
-------
A-3.   Objective 3: Characterization of the Correlation Between Components Sampled in
       the Same Work Area
       The relationships among floor, window sill, and window trough wipe samples are another
important aspect of examining a clearance testing program.  One method of assessing the
relationship between individual floor, window sill, and window trough wipe samples collected
from the same room is to estimate linear correlation coefficients.  Table A-3 displays the Pearson
product-moment correlation coefficients and the associated sample size for the log lead loading
measurements of individual floor, window sill, and window trough samples taken within the
same room. Results from the first site visit show that floor samples are positively correlated with
both window sill and window trough samples. The correlation between window sills and window
troughs is also positive and is the highest of any observed correlation from the first site visit. All
observed correlations from the first site visit are significant at the 0.01 level.  The correlations
from the second site visit follow the same general pattern as those observed during the first site
visit, but are not as strong. These correlations observed at the second site visit are all  significant
at either the 0.05 or 0.01 level. The correlations from the third site visit are difficult to interpret
because of the small number of samples from which they were estimated.
       The conditional probabilities of a sample passing or failing a standard are given in
Table A-4. These analyses were conducted on two different sets of data, the first set using all
possible paired observations  from within the same room. The second subset of data restricted the
analyses to rooms in which floor, window sill, and window trough lead loadings were
simultaneously observed.
       As shown in Table A-4, results from contingency table estimates and normal theory
estimates are consistent. For both types of estimates and both data sets included in the analysis,
the probability that one component passes clearance given that another component passes
clearance ranges from 81% to 98%. The probability that a window sill sample is less than
500 ug/ft2, given that the samples from the window troughs are less than 800 ug/ft2, is among the
highest values at 97 to 98%.
       For both types of estimates and both data sets included in the analysis, the probability that
one component fails clearance testing given that another component in the same room fails
clearance testing ranged from 17% to  78%. The probability that a sample collected from a
                                         A-19
-------
window trough is greater than or equal to 800 ug/ft2, given that a sample collected from the
window sill is greater than or equal to 500 ug/ft2, is fairly high (greater than 76%).
Table A-3.  Observed Within-Room Correlation Coefficients Between Dust-Lead Loading
            Measurements Collected From Floors, Window Sills and Window Troughs For
            the Maryland Data.
                        llliiObsere^
       First
0.53*
1739
0.42*
1340
0.69s
1489
     Second
0.31'
 86
0.28"
 84
0.55'
 123
      Third
0.32
           -0.34
                       0.41
                         16
       '  Statistically significantly different from zero at the 0.01 level.
       b  Statistically significantly different from zero at the 0.05 level.
A-4.   Objective 4: Demonstration of the Impact of Composite Sampling on Pass/Fail Rates
       of Houses
       First site visit data from the Maryland Department of the Environment included 7,200
individual sample lead loading clearance sample results collected from floors, window sills, and
troughs in 3,450 rooms within 706 dwelling units. Individual sample lead loading results from
each component type within a residential unit were combined to construct simulated composite
sample results.
       Table A-5 provides, for each component type, the number of residential units that were
investigated by the number of individual samples that were collected. For example, there were
88 residential units which included five floor dust-wipe samples on the first site visit.
Residential units containing more than four samples from a component type resulted in the
estimation of multiple simulated composite sample results in this analysis.  Therefore, summing
all the units that had five or more individual samples, approximately 40% (260/661), 33%
(204/612), and 28% (151/536) of the residential units resulted in the estimation of multiple
simulated composite samples  from floors, window sills, and window troughs, respectively.
       When there were four  or fewer individual samples from a component type within a
housing unit, the simulated composite sample included all samples.
                                          A-20
-------
s
 CO
SI

 8
0.
?
 (0

 (O
 OT


^



I

CM

 X

CM
•o
 V
+*
 (0


"•p
 co
Ul

 O)

v=
 w
 e
 S
 e  si
 tO  (A
Bjg
    .
.2  i^
£  eo

I-

    1
"•s  >>
O
(0
           •5ZJ
     M
     6
          s
g
6
          CM
          6
                              to
                              IN
                              00
                                8
                                m
                                6
               6
                          ra
                          IX
                          1
                          o
     n

     §
     o
     N
     IO

     O
                    IO
                    IO
                    10

                    6
     w
     o
CM
CM

o
     n
     a>
     n
     o
     S
      Al
     10
     EL
          o
          s
«

o
               S
               cq
               o
          S
               GO
               6
                    IB
                    CO
                        m
                        CO
                                    IB
                                    cn
                                    6
               to
               e
               CO
               6

     O
                                         O)
                                         CM
                                              cn
                                              6
                          v
                          CO
     in
     V
     CO
     s:
V
t-

e


v
IL
sr
                                         CO
                                         v
                                         i-
          U)

          V
          U)
          V
          CO

          g
          CO
          V
                              V

                              —
                              o
                   V
                   co
                   E
                                        §
                                        IO
                                        V
                                        co
                                  g
                                  CO
                                  V
                                                            A-21
-------
       For each component type within a residential unit, the set of individual clearance sample
lead loading results was used to construct simulated composite samples for the purpose of
evaluating the three composite sample clearance criteria (Standard, Standard/n, and
2xStandard/n).  Table A-6 shows, for each combination of component type and composite
clearance criterion, the number of residential units that passed clearance, failed clearance, or
yielded inconclusive results.  Inconclusive results were only possible for those residential units
which contained more than four individual samples; the many possible ways of combining five
or more individual sample lead-loadings into multiple simulated composite samples had the
potential of resulting hi a variety of different outcomes under each of the three composite
clearance criteria.  However, the percentage of residential units with greater than four individual
clearance samples of a given component type that resulted in inconclusive results was low,
ranging from 5.4% (11/204) for the evaluation of window sill samples under the 2>
-------
Table A-6.   Individual Sample Clearance Results versus Simulated Composite Clearance
             Data from the Maryland Data.

      Standard
                   Pass
Inconclusive
                   Fail
                    521
        42
        470
35
         36
                 14
                             62
                          93
337
39
                 16
                                  144
     Standard/n
                   Pass
Inconclusive
                   Fail
                    434
                 430
32
         18
                    55
        140
         22
142
                          300
         10
27
199
    2 x Standard/n
                   Pass
Inconclusive
                   Fail
                    494
                 467
14
28
                    13
        107
                 125
                         324
                 15
                 178
       The performance characteristics of each combination of component type and composite
clearance criteria are presented in Table A-7 in terms of sensitivity, specificity, positive
predictive value (PPV), and negative predictive value (NPV).  These performance statistics are
based on empirical estimates of the following conditional probabilities:

Sensitivity    =  Pr(Fail Composite Sample Clearance J Fail Individual Sample Clearance)
Specificity    =  Pr(Pass Composite Sample Clearance | Pass Individual Sample Clearance)
PPV         =  Pr(Fail Individual Sample Clearance j Fail Composite Sample Clearance)
NPV         =  Pr(Pass Individual Sample Clearance | Pass Composite Sample Clearance)

If the primary purpose of conducting clearance testing is to prevent potential lead exposure for
residents, then one might choose a composite sample clearance criteria with high sensitivity at
the expense of high specificity.  By design, the Standard/n Rule always sacrifices specificity for
sensitivity; the sensitivity for the Standard/n Rule is always 1.00 while the specificity for this
rule is estimated at 0.83 for floors, 0.92 for window sills, and 0.89 for window troughs.  On the
other hand, the Standard Rule always sacrifices sensitivity for specificity.  The specificity for
the Standard Rule is always 1.00 while the sensitivity for this rule is estimated at
                                         A-23
-------
 Table A-7.   Performance Characteristics of Composite Clearance Criteria Based on the
              Maryland Data.


                    Sensitivity
0.44
1.00
0.76
       Floors
                    Specificity
1.00
0.83
0.95
                    PPV
1.00
0.72
0.89
                    NPV
0.93
1.00
0.99
                    Sensitivity
0.66
1.00
0.88
      Window
        Sills
                    Specificity
1.00
0.92
0.99
                    PPV
1.00
0.87
0.99
                    NPV
0.93
1.00
0.98
                    Sensitivity
0.72
1.00
0.89
Window
Troughs
                    Specificity
1.00
0.89
0.96
                    PPV
1.00
0.88
0.96
                    NPV
0.90
1.00
0.98
0.44 for floors, 0.66 for window sills, and 0.72 for window troughs. The 2xStandard/n Rule
attempts to maximize both sensitivity and specificity. For the 2>
-------
where n^ is the estimated probability of clearance for component(/) in house(i) under composite
criterion (£), and Max,-, is the maximum individual sample lead loading result in house(i) for
component(/).
       For each component type, the estimated relationship between the probability of passing
clearance and the maximum lead-loading is presented graphically in Figure A-9. In this figure,
the solid, long dashed, and finely dashed lines represent the estimated relationship for the
Standard,  Standard/n, and 2*Standard/n Rules respectively.
       Table A-8 provides parameter estimates and associated standard errors from the logistic
regression models, as well as estimates of the probability of passing clearance (using composite
samples) when the maximum lead loading among all locations included in the composite
sampling scheme is greater than or equal to '/z, 1,2 and 4 times the associated HUD standard
for individual samples.  Results from fitting these logistic regression models to clearance data
from the Maryland Department of the Environment show that under the Standard Rule for
composite sampling, there is a better than 50% chance that a residential unit will pass clearance
testing when there is an individual sample location which has a lead-loading level that is equal
to twice the HUD recommended clearance standard.  The estimated relationship for the
Standard/n Rule demonstrates this rule's high sensitivity (probability of passing is low when the
maximum individual sample lead loading is greater than or equal to the HUD Standard) along
with the loss in specificity for this rule (probability of passing is between 0.75 and 0.90 when
the maximum individual sample lead loading is equal to l/z HUD Standard). Once again, the
2 x Standard/n Rule is shown to be a compromise between the Standard and Standard/n Rules.
At 1A HUD Standard, the estimated probability of passing clearance testing under the
2 x Standard/n Rule is nearly one, and at 2>
-------
                         Probability of Passing Clearance Testing Using Composite Floor Samples
                                                                Standard Rub
                                                        	(2 x Standard/n)Rute
                                                                (Standard/n) Rule
                                    200     300    400    500     600
                                   Maximum Individual Sample Floor Pb Loading
                             700
                              800
                       Probability of Passing Clearance Testing Using Composite Window Sill Samples
                                                               Standard Run
                                                        	(2 x Standard/n) Rub
                                                        	(Standard/n)Ruto
                                   500
750
1000   1250
                                                                1500
1750    2000
                                 Maximum Individual Sample Window SID Pb Loading
                      Probability of Passing Clearance Testing Using Composite Window Well Samples
                                                               Stavntwd Ruto
                                                       	(2 x Standard In) Rub
                                                       	(Standard/n) Rule
                   0.0
                                          1200   1600   2000    2400    2600   3200

                                 Maximum Individual Sample Window Well Pb Loading
Figure A-9.   Estimated Relationship Between the Probability of a Residential Unit Passing
               Clearance Testing versus the Maximum Individual Lead-Loading Result by
               Component Type Based on Simulated Composite Samples from the
               Maryland Department of the Environment.
                                               A-26
-------
    «  3
 0}
K O
00
.Q
 a
                                 §
                                     o
                                     o
                                     CL
                                                       CO
•
.£ 55
                                                            
                                                                     (O
                                                                     r-
                                                                     (O
                                                                     CM
                                                                     CM
                          in
                          ro
                          o>



                          io
                          D

                          10
                                                                           o
                          (5


                          (Q

                                                                           Q.



                                                                            O
                                                                            o
                                                                            o
                                                                           cc
                                                                           LLJ
                                                                           O
                                                                            O
                                                                            ID
                                                                            •O
                                                                           I
                                                                           2
   .

1!
                                                  A-27
-------
-------
            APPENDIX B

FHA Single-Family Housing Phase of the
HUD Abatement Demonstration Project
                B-1
-------
                                   APPENDIX B
                  FHA SINGLE-FAMILY HOUSING PHASE OF THE
                  HUD ABATEMENT DEMONSTRATION PROJECT
       The HUD Abatement Demonstration Project was performed from 1989 -1990 to assess
the costs and benefits associated with several lead-based paint abatement methods [14]. In the
FHA single-family housing phase, the demonstration was conducted in HUD-owned vacant
single family dwelling units. The Federal Housing Administration (FHA) held title to the
houses. A total of 304 units from several different U.S. cities were tested for leaded paint, and
172 units were identified as having sufficient amounts of lead-based paint to warrant abatement.
Paint abatement methods performed during the FHA phase included enclosure, encapsulation,
chemical removal, removal with heat gun, removal and replacement, and abrasive removal.
       Following abatement, clearance testing using individual wipe samples was conducted in
149 units.  Three of the 172 units were used as pilot units and another 20 had only exterior
abatement.  Wipe dust-lead loading results from all clearance samples were available in the
HUD Demonstration Lead-Based Paint (HUDLBP) Database maintained by Battelle [15].
Original data were obtained from Speedwell Inc., located in Boston, Massachusetts. Following
the NIBS Guidelines [6], clearance dust-wipe samples were collected after the unit had been
"final cleaned," but prior to recoating or priming of any surfaces [2]. Residential units passed
clearance testing if all surfaces sampled resulted in dust-lead loadings below the HUD
recommended standards of 200 ug/ft2 for floors, 500 ng/ft2 for window sills, and 800 ug/ft2 for
window troughs.  If these dust lead standards were not met, the study protocol required the
residential unit to undergo more extensive cleaning and repeated clearance wipe sampling until
the unit met the standards, for up to three iterations of cleaning and retesting. However, data
from the HUDLBP Database indicated that several units underwent more than three clearance
testing iterations.  Note that although clearance was not attained in every unit, any failed surface
was eventually sealed.
                                        B-2
-------
B-1.   Objective 1:  Characterization of the Number of Individual Samples. Work Areas,
       and Housing Units That Pass or Fail Clearance Testing Standards
       Individual dust wipe samples were collected from floors, window sills, and window
troughs as part of the HUD Demonstration Project. A total of 2,854 samples were taken from
1,086 rooms in 149 residential units. Table B-1 presents the number of individual samples,
work areas, and residential units that passed or failed clearance testing within each combination
of component type and site visit.  Approximately 73% (2,077/2,854)  of the dust samples were
collected during the first site visit to a residential unit Of the 97 residential units that failed
clearance on the first site visit, 95 (98%) of these residential units were revisited for a second
clearance testing.  This high return rate to residential units which failed is again observed in the
next round as 93% (55/59) of the residential units which failed after a second site clearance
testing site visit were revisited a third time. The pattern does not continue, however, since only
32% (11/34) of the sites failing clearance on the third site visit were revisited for additional
clearance testing.  Eventually, 111 of the 149 residential units (74%)  passed clearance testing.
On the first site visit, 79% (1,644/2,077) of the dust samples fell below the clearance standards
of 200 pg/f? for floors, 500 ug/ft2 for window sills and 800 ng/ft2 for window troughs while
68% (698/1,022) of the rooms and 35% (52/149) of the residential units passed clearance.  This
sharp increase in the failure rate from the percentage of individual samples that fail to the
percentage of rooms and residential units that fail is attributable to the fact that if any individual
sample exceeds the standard and fails clearance, then both the room and residential unit also fail
clearance. This data set is particularly prone to exhibit this type of trend since, on average, there
were more individual samples taken per residential unit than in any of the other data sets
analyzed (13.9 per unit (2,077/149) during the first site visit).
       The failure rate for individual samples during the first site visit was higher for this data
set than for any other.  The failure rate was highest for window trough samples at 36%
(161/442), followed by floors at 19% (182/967) and window sills at 13% (90/668).  The failure
rate for individual samples showed an increase from the first site visit to the second site visit
and again from the second visit to the third site visit. This trend indicates that units identified as
a problem on the first site visit were very likely to continue to be a problem even after several
attempts at clearance.
                                           B-3
-------
Table B-1. Clearance Testing Results by Individual Sample, Room, and Residential

           for the HUD Demonstration Project.
                                                        Unit
                                                            JiX~«; in
                                                            retail
  First
             Floor
              Sill
            Trough
              All
785
182
 967
748
174
922
78
578
 90
 668
576
 88
664
80
281
161
442
280
160
440
47
1644
433
2077
698
324
1022
52
67
46
61
97
145
126
108
149
 Second
             Floor
              Sill
            Trough
              All
152
 82
234
141
78
219
34
 83
 28
 111
 80
28
 108
30
 88
 80
 168
 88
80
 168
29
323
190
513
203
160
363
36
36
21
32
59
70
51
61
95
  Third
             Floor
             Sill
            Trough
             All
 46
 34
 80
43
32
 75
19
 17
 21
 38
 17
21
 38
10
 35
47
 82
35
45
 80
12
 98
102
200
71
87
158
21
16
11
18
34
35
21
30
55
 Fourth
             Floor
             Sill
            Trough
             All
 12
        16
         11
               15
                12
                              12
 15
        21
         15
               21
 32
 17
 49
23
11
 34
             11
  Fifth
             Floor
             Sill
            Trough
             All
                                      4
 6
 12
        15
                       10
Total
Floor
Sill
Trough
All
997
689
423
2109
304
147
294
745
1301
836
717
2854
910
674
416
961
61
21
56
125
971
695
472
1086
129
119
94
111
16
14
17
38
145
133
111
149
                                         B-4
-------
B-2.   Objective 2; Characterization of the Distribution of the Dust-lead Loadings,
       Geometric Mean Dust-lead Loadings. Variability Within a Housing Unit, and
       Variability Between Housing Units
       Preliminary assessment of the data indicated that the distribution of dust-lead loading
clearance sample results were highly skewed.  A natural logarithm transformation was applied
to the data.
       For the clearance data collected in the HUD Demonstration Project, Table B-2 lists the
geometric mean lead loading and the 95% confidence interval about the mean for each site visit
and component type. In general, the within-home (room-to-room) variability is smaller than the
between homes variability; however, all of the variance components are within a single order of
magnitude of each other. Also notice that the lengths of the confidence intervals increase
successively from the first site visit to the third site visit for any given component. The increase
in the length of the confidence intervals is due primarily to the decrease in the number of
samples used to estimate the variance components.
       The geometric mean lead loadings and their 95% confidence intervals are used to
compare trends between site visits for a given component tested and to compare average results
between component types within a site visit.  If the 95% confidence intervals on two geometric
means do not overlap then these two geometric means are statistically significantly different at
a level less than 0.05. A comparison across site visits shows that the geometric means for floors
and window sills increase significantly from the first site visit to the second site visit, 58 ug/ft2
to 100 ug/ft2 and 60 ug/ft2 to 172 ug/ft2, respectively. The increase from the second site visit to
the third site visit is not statistically significant. The geometric means from the first site visit to
the third site visit for the window troughs, 393 ug/ft2,563 ug/ft2, 644 ug/ft2, are not statistically
different.  A comparison of the components tested within a site visit show that the floor dust-
lead loadings are on average less than the window sill dust-lead loadings, which are in turn less
than the window trough dust-lead loadings. Note that estimates from the fourth and fifth site
visits are associated with very small sample sizes and may not be very reliable for establishing
trends.
       Figures B-l and B-2 contain box and whisker plots that present the distribution of dust-
lead loadings from the first and passed clearance visits by component type. Figures B-3 to B-5
contain box and whisker plots that present the distribution of dust-lead loadings from the first
visit for floor, window sill, and window trough, respectively; Figures B-6 to B-8 contain box
                                           B-5
-------
 
2
 o

 e
O
a
D
X
 M
 0
 **
 a


I
u
 +*


 §
 o
 a.

 o
O

 8

.s
 <5


•o

 a

 a
 o>
_

1
 E

 2
O
GO

 O

3
                                          O
                                          00
            (O
            (O
            in
            0>
            CO
                                                           00
                                                           
                                                      in
                                                      
co
co
O)


CN
                       co
                       CD
           in
           |x
           oo
           CO
            o
            00
                                   CN
                                         
-------
              100.0
          Sf
          <   10.0
           o>
          I   1.0
               0.1
                     floors       WhkwSBs    Window Troths     Hoors      Window Sfc    Window txtft,

                                RistVisI                          Passed Ctearance

                                                Surface
Figure B-1.  Comparison of In-Transformed Dust-Lead Loadings from the First Site Visit vs.

             Passed Clearance Results on an Expanded Scale.
                                              B-7
-------
             700
          CM
          <  500
          JP4QO
          TJ
             300


             200


             100


              0
                  Floors
Window Sills
  Surface
Window Troughs
Figure B-2.   Box and Whisker Plots of the Distribution of Dust-Lead Loadings from the
             Passed Clearance Data.
                                           B-8
-------
              10000
               1000
           g    100
           I
           .0
           CL
                10
                      AO         Wood        Vnyt       Akunhm      Other      Urftnwn


                                                Substrate
Figure B-3.   Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by

              Substrate for the Rrst Site Visit.
                                              B-9
-------
              100000
              10000
               1000

                100
                 10
                       AI        Wbod        Vhyl      Aluminum       Other       Unknown
                                                Substrate
Figure B-4.   Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
              by Substrate for the First Site Visit.
                                             B-10
-------
              1000000
               100000
           CM    10000
                1000
                 100
                  10
                        Al
Wood
Vinyl
Aluminum
Other
Unknown
Figure B-5.   Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings fay Substrate for the First Site Visit.
                                             B-11
-------
             1000
          w  100
           0)

          1
          3

          £
10
                1
                               Wood
                           Vnyl
Aluminum
Other
Unknown
Figure B-6.  Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
             Substrate for the Passed Clearance Visits.
                                           B-12
-------
             1000
          CM  100
              10
Afl         Wood        Wr/1       Aluminum

                        Substrate
                                                               Other
Unknown
Figure B-7.   Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
             by Substrate for the Passed Clearance Visits.
                                          B-13
                                                                U.S. EPA Headquarters Library
                                                                       Mail code 3201
                                                                1200 Pennsylvania Avenue NW
                                                                   Washington  DC  20460
-------
             1000
          CM  100
              10
                    AJ
Wood       Vnyl       Aluminum       Otter       Unkmm

               Substrate
Figure B-8.   Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
             Loadings by Substrate for the Passed Clearance Visits.
                                           B-14
-------
and whisker plots that present the distribution of dust-lead loadings from the passed clearance
visits.

B-3.   Objective 3: Characterization of the Correlation Between Components Sampled in
       the Same Work Area
       The relationships among floor, window sill, and window trough wipe samples from
within the same room are another important aspect of examining a clearance testing program.  By
estimating linear correlation coefficients, the strength of these relationships can be assessed.
Table B-3 displays the Pearson product-moment correlation coefficients and the associated
sample size for the natural logarithm transformed dust-lead loading measurements of individual
floor, window sill, and window trough samples taken within the same room. The data from the
first site visit show that floor samples are positively correlated with both window sill and
window trough samples. The correlation between window sills and window troughs is also
positive and is the highest of any observed correlation from the  first site visit. All observed
correlations from the first site visit are significant at the 0.01 level. The correlations from the
second site visit are again all positive, but are not as strong as those observed during the first site
visit. Only the correlation coefficient between floors and window sills during the second site
visit is significant at the 0.05 level.  The correlations from subsequent site visits are difficult to
interpret because of the small number of samples from which they were estimated.
       The conditional probabilities of a sample passing or failing a standard are given in
Table B-4.  These analyses were conducted on two different sets of data, the first set using all
possible paired observations from within the same room. The second subset of data restricted the
analyses to rooms in which floor, window sill, and window trough lead loadings were
simultaneously observed.
       As shown in Table B-4, results from contingency table estimates and normal theory
estimates are consistent. For both types of estimates and both data sets included in the analysis,
the probability that one component passes clearance given that another component passes
clearance ranges from 66% to 95%. The probability that a window sill sample is less than
500 fig/ft2, given that the samples from the window troughs are  less than 800 ug/ft2, is among the
highest values at 93 to 95%.
       For both types of estimates and both data sets included in the analysis, the probability that
one component fails clearance testing given that another component in the same room fails
clearance testing ranged from 27% to 79%. The probability that a sample collected from a
                                          B-15
-------
window trough is greater than or equal to 800 ug/ft2, given that a sample collected from the
window sill is greater than or equal to 500 ug/ft2, is fairly high (greater than 72%).
Table B-3. Observed Within-Room Correlation Coefficients Between Dust-Lead Loading
           Measurements Collected From Floors, Window Sills and Window Troughs for
           the HUD Demo Data.
       First
       Second
       Third
       Fourth
       Fifth
                                   • ••* i_» " f*9J^~~ r*'y*i?™w *«»g"H",,T »ai^u.'f^'9'~'"i:-fr^L^ffti''r'-lm^
                                   Within Room CorrelatiomCoefficients

0.39*
0.37"
-0.04
0.98
581
45
 12
0.36"
0.12
-0.31
          0.75
415
 61
 14
                                                                ijrauglit?
0.65'
0.20
0.59"
                       0.73
                                           0.47
375
55
 17
       * Statistically significantly different from zero at the 0.01 level.
       " Statistically significantly different from zero at the 0.05 level.
B-4.   Objective 4:  Demonstration of the Impact of Composite Sampling on Pass/Fail
       Rates of Houses
       Table B-5 provides for each component type the number of residential units that were
investigated by the number of individual samples that were collected. Residential units
containing more than four samples from a component type were included in the estimation of
multiple simulated composite sample results. From Table B-5, the number of homes with five or
more samples was  125 for floors, 84 for window sills, and 47 for window troughs.  Therefore,
summing all the units that had five or more individual samples, approximately 86% (125/145),
67% (84/126), and 44% (47/108) of the residential units were included in the estimation of
multiple simulated composite samples from floors, window sills, and window troughs,
respectively. When there were four or fewer individual samples from a component type within a
housing unit, the simulated composite sample included all samples. The percentage of
residential units with more than four individual samples of a component type was much higher in
the HUD Demonstration Project than in the other sources of data considered in this analysis.
This is probably attributable to the fact that the HUD Demonstration Project was research
oriented.
                                          B-16
-------
•o

 «
 (A
.32
15
 CO


 fr

 4)
 D>


I
 O
O
CM

 X
CM


f
 «>
D
•o
 o
to
Ul

O)
 10
£
 9
 O  .
 c *;
 a m

1!
o «2
 ow
* I
 c il
I e

1"
Jo e
1 °
^^
1«
 §OQ
S ^
•o 5
 C Q>
 O £
O H
^1
m
j>
^
(0
                           IN

                           O
                                    o\
                                    oo

                                    o
                                          I
                       8
                                       8
                   8
                           /M



                           I
                                    I/I
                                    o
                                    o
                                    00
                                          g
                                          V}
                                          eu
                         a.
                         o
                         o
                         00
                         /M
                    00
                    o
                       oo
                       o
          8
          vq
          O
                                          I
o
S
o
                                    cs

                                    r^
oo
CS
             I
S
o
8
V
m
a
                          8

                          I
                          9
                                    8
                                             8
                                                B-17
-------
       For each component type within a residential unit, individual clearance sample lead
 loading results were used to construct simulated composite samples for the purpose of evaluating
 the three composite sample clearance criteria (Standard, Standard/n, and 2* Standard/n). For
 each combination of component type and composite clearance criterion, each residential unit
 either passed clearance, failed clearance, or yielded inconclusive results. Inconclusive results
 were only possible for those residential units which contained more than four individual samples.
 From Tables B-5 and B-6, the percentage of residential units with five or more samples collected
 for a component that resulted in inconclusive results under the Standard/n Rule ranged from 6%
 (5/84) for window sills to 19% (24/125) for floor samples.
Table B-5.  Number of Residential Units that Contained (N) individual Samples of Each
           Component Type Based on First Site Visit Clearance Testing Data from the HUD
           Demonstration Project.
                     3 +
                    Total
                Total with Nj>5
                                   36
                                   17
                                   17
                                   23
 32
145
125
                                                21
              23
              23
              22
126
84
                              19
                                                                21
                24
                14
108
47
       The performance characteristics of each combination of component type and composite
clearance criteria are presented in Table B-7 in terms of sensitivity, specificity, positive
predictive value (PPV), and negative predictive value (NPV). By design, the sensitivity for the
Standard/n Rule is always 1.00 while the specificity for this rule is estimated at 0.49 for floors,
0.88 for window sills, and 0.72 for window troughs. In contrast, the specificity of the Standard
Rule is always 1.00 while the sensitivity for this rule is estimated at 0.63 for floors, 0.59 for
window sills, and 0.79 for window troughs. The 2xStandard/n Rule attempts to maximize both
                                          B-18
-------
sensitivity and specificity.  For the 2xStandard/n Rule, the values of sensitivity are higher than
those calculated for the Standard Rule, while the values of specificity are higher
than those calculated for the Standard/n Rule. Estimates of sensitivity and specificity in these
examples are always conservative, because inconclusive composite test results factor into the
denominator for each estimate, but never factor into the numerator.
Table B-6. individual Sample Clearance Results Versus Simulated Composite Clearance
           Results Based on Data from the HUD Demonstration Project.
                                                      idu'ajySarnple (
      Standard
                    Pass
Inconclusive
                    Fail
                     78
                80
        18
13
                             42
                        27
        47
                  48
     Standard/n
                    Pass
Inconclusive
                    Fail
                     38
                70
24
                      16
       67
46
        34
61
   2 x Standard/n
                    Pass
Inconclusive
                    Fail
                     68
                78
        10
                             57
                        39
        45
                  54
       Table B-7.  Performance Characteristics of Composite Clearance Criteria Based on
                   Data from the HUD Demonstration Project.
       Floors
                   Sensitivity
                   Specificity
                   PPV
                   NPV
                        0.63
                         1.00
                         1.00
                        0.92
                     1.00
                     0.49
                     0.81
                     1.00
               0.85
               0.87
               0.98
               1.00
      Window
       Sills
                   Sensitivity
Specificity
                   PPV
                   NPV
                        0.59
                                            1.00
                         1.00
                        0.93
                     1.00
                     0.88
                     0.90
                     1.00
               0.85
               0.98
               1.00
               1.00
      Window
      Troughs
                   Sensitivity
Specificity
PPV
NPV
                        0.79
                                            1.00
                                            1.00
                                            0.89
                     1.00
                     0.72
                     0.87
                     1.00
               0.89
               0.96
               0.98
               0.96
                                            B-19
-------
       The above performance characteristics estimated from the HUD Demonstration Project
help illustrate that the three composite clearance testing criterion have different specificity and
sensitivity rates. These rates correspond to the consistency between clearance
decisions and the true lead hazards present in the various locations sampled (assuming the
individual sample lead-loading results are representative of these lead hazards).  To further
characterize the performance of each of the three composite clearance criteria, the following
logistic regression model was fitted to clearance data from the HUD Demonstration Project for
each combination of component type and composite clearance criteria:
where TC^ is the estimated probability of clearance for component(/) in house(i) under composite
criterion (K), and Max^, is the maximum individual sample lead loading result in house(i) for
component(/).
       For each component type, the estimated relationship between the probability of passing
clearance and the maximum lead-loading is presented graphically in Figure B-9.  In this figure,
the solid, long-dashed, and finely-dashed lines represent the estimated relationship for the
Standard,  Standard/n, and 2*Standard/n Rules, respectively.
       For each combination of component type and composite clearance criteria, Table B-8
provides parameter estimates and associated standard errors from fitting the logistic regression
models to data from the HUD Demonstration Project. Table B-8 also presents estimates of the
probability of passing clearance (using composite samples) when the maximum lead loading
among all locations included in the composite sampling scheme is greater than or equal to 1A, 1,
2, and 4 times the associated HUD interim standard for individual samples. Results from these
logistic regression models show that there is a better than 70% chance that a residential unit  will
pass clearance for floors and window sills under the Standard Rule for composite sampling when
there is an individual sample location which has a lead-loading level that is equal to twice the
HUD clearance standard. Estimates for the Standard/n Rule demonstrate this rule's high
sensitivity (probability of passing is close to zero when the maximum individual sample lead
loading is greater than or equal to the HUD Standard) along with a loss in specificity (probability
of passing is between 0.67 and 0.90 when the maximum individual sample lead loading is equal
to !/2 HUD Standard).  Once again, the 2*Standard/n Rule is shown to be a compromise between
the Standard and Standard/n Rules.  At Vz HUD Standard, the estimated probability of passing
clearance testing under the 2xStandard/n Rule is nearly one, and at 2xHUD Standard, the
estimated probability of passing clearance testing under the 2*Standard/n Rule is close to  zero.
                                          B-20
-------
                          Probability of Passing Clearance Testing Using Composite Floor Samples
                                                                Standard Ruto
                                                           	(2 x Standard/n) Rule
                                                           - — (Standard In} Ruto
                              100     200     300    400    500    600
                                    Maximum Individual Sample Floor Pb Loading
700
        800
                        Probability of Passing Clearance Testing Using Composite Window Sill Samples
                                                                Standard Ruto
                                                          	(2 x Standard/n)Rute
                                                          	{Standard/n) Ruto
                       0     250    500     750    1000   1250    1500    1750   2000

                                  Maximum Individual Sample Window SHI Pb Loading


                       Probability of Passing Clearance Testing Using Composite Window Wed Samples
                                                                  Approach

                                                        	 Standard Rub
                                                        	(2 x Standard/n) Ruto
                                                        	(Standard/n)Rute
                             400    600    1200   1600   2000    2400    2800

                                  Maximum Individual Sample Window Well Pb Loading
      3200
Figure B-9. Estimated Relationship Between the Probability of a Residential Unit Passing
             Clearance Testing versus the Maximum Individual Lead-Loading Result by
             Component Type Based on Simulated Composite Samples from the HUD
             Demonstration Project.
                                                B-21
-------
*;  0>
 O  C


II
£  (0

*»
 2.5
Q.  0)

Ja
-  E
 c  o
 o O
tS
c %
o 2
 c °
— B
 O CD

S "3.

 S E
_ a
 §
   £. O

sl 1
yi ^ q)

£ cQ
 O X •*-
2 ^ B)
.2 2 .£
 O    W

31 =
 (0 ^ »


 l«l
   »g

e o Q
CO

CO

o

Z
                                 CO
                                          o
                                          o
                                     o
                                     o
 o
 o
                                                   
0>
          CO

          0
         CO


         6
                                                       o

                                                       q

                                                       6
                                                             8
                I
                                                                     o
                                                       CD
                                                       
      i>.

      oi
                                                            CO
                                                            oo
                                                            oo
    oo
    CO
    
                                                                     CM
                                                                     in
                                                       in
                                                             CO
                                                             •*
                in
                in
                                                                 (O
                                                       O
                                                       r*
           co
           in
                                                       c

                                                       T!
                                                       m
                                                       TJ

                                                       m
                                                       *•
                                                       CO
           to
           •o
           to
           z
    .c

    •o

     CD
    T3


     g


    (/)

     X

    OJ
                                                                      re
                                                                     •o
i.
•o =
.£ c/5
                                                                5
                                                                o

                                                                ~
   ~° 3
   .E o
                                                    B-22
-------
               APPENDIX C

 Public Housing Administration (PHA) Phase
of the HUD Abatement Demonstration Project
          (Albany, Omaha, Cambridge)
                   C-1
-------
                                   APPENDIX C
                PUBLIC HOUSING ADMINISTRATION (PHA) PHASE
             OF THE HUD ABATEMENT DEMONSTRATION PROJECT
                           (ALBANY, OMAHA, CAMBRIDGE)
       The Public Housing Administration (PHA) Phase of the HUD Abatement Demonstration
Project was performed to assess the costs and benefits associated with performing lead-based
paint abatement in multifamily housing [14]. The demonstration was conducted in multi-unit
apartment complexes in Cambridge, Massachusetts; Albany, New York; and Omaha, Nebraska.
The project in Cambridge involved two garden apartment buildings, each with 24 residential
units. Paint abatement was conducted using chemical methods for the residential units in one
building, and abrasive methods were used for the residential units in the other building. In
Albany, there were also two apartment buildings, each with 18 residential units. Paint abatement
was performed in the first building using encapsulation and enclosure systems, and chemical
stripping was used for residential units in the second apartment building. The apartment
complex in Omaha consisted of brick faced townhouses, which were abated using component
removal and replacement.
      Following abatement, clearance testing using individual wipe samples was conducted in
the residential units of each building. Wipe dust-lead loading results from all clearance samples
were available in hard copy form from records collected during the study.  Residential units
passed clearance testing if all surfaces sampled resulted in dust-lead loadings below the HUD
recommended standards of 200 ug/ft2 for floors, 500 ug/ft2 for window sills, and 800 ug/ft2 for
window troughs. If these dust lead standards were not met, the study protocol required the
residential unit to undergo more extensive cleaning and repeated clearance wipe sampling until
the unit met the standards.
      The following excerpts [10-12] give detailed information on the abatement outcomes for
Albany, Omaha, and Cambridge.
Albany Abatement Outcomes: Dust Lead
      The interiors of all the demonstration units were wipe-tested for lead dust after the units
had been deleaded and cleaned, and met visual abatement clearance standards. More wipe tests
(296) were taken in Building E, where paint removal methods were widely employed, than in
Building A (124), where encapsulation of lead hazards was emphasized.  The Interim HUD
                                        C-2
-------
Guidelines required that three wipe samples be obtained in each area where on-site paint removal
is used, whereas only one wipe sample was required in each area where replacement and/or
encapsulation methods are employed.
       All but one of 113 window sills and all but two of 98 window troughs met the wipe test
clearance standard on the first attempt. The failure rate for floors on initial wipe tests was also
low at 3.3% (7/202).  The very low failure rate on floor wipe tests indicates the effectiveness of
the cleaning, considering the polyethylene floor covering was used only on selected areas during
abatement.
       There were 10 cases where initial wipe test clearance standards were not achieved. Nine
of these 10 cases, and five of the six units, were in Building E, where on-site paint removal
methods were emphasized. All six units met the clearance criteria on the second attempt.

Omaha Abatement Outcomes: Dust Lead
       The interiors of all the Demonstration dwelling units were wipe-tested for the presence of
dust lead after lead-contaminated building components had been removed and the units had been
cleaned and before the remainder of the modernization was started. The methods of cleaning
were those prescribed in the HUD Interim Guidelines, as were the dust lead clearance standards.
All 161 window troughs and 158 out of 165 window sills met the wipe test clearance standard on
the first attempt These very low failure rates may have reflected the fact that all the windows
were fairly new. The failure rate for floors on the initial wipes was significantly higher at 12.9%
(22/148). Sixteen of the 49 dwelling units failed to meet wipe test clearance standards at one or
more interior test locations. These units were then cleaned one more time, and all of them passed
when retested.
       The interiors of all the Demonstration dwelling units were again wipe tested when the
renovation of the buildings was complete and prior to reoccupancy.  A total of '277 wipe samples
were obtained and analyzed and all passed the clearance standards. On the post-abatement wipe
tests, 26% of the samples had less than 25 ng/ft2, compared to 86% on the post-renovation wipes.
The post-renovation samples were not included in this analysis.

Cambridge Abatement Outcomes:  Dust Lead
       A visual inspection was performed to ensure that all lead-based paint had been abated in
accordance with the contract specifications. This inspection was  followed by a cleaning process
outlined in the HUD Interim Guidelines which consisted of a thorough High Efficiency
                                         c-3
-------
 Particulate Accumulator (HEPA) vacuuming of all surfaces, a trisodiuni phosphate (TSP) wash
 down of all surfaces, and a final HEPA vacuuming of all surfaces. The interior of all the units
 and selected locations in the stair systems were wipe-tested for the presence of lead dust
 Generally this process called for the collection of surface dust from a horizontal surface of at
 least one window sill and window trough, and a one square foot area on the floor of each room or
 space after abatement was completed.  The collection of dust was accomplished by wiping the
 surface with a moistened cloth wipe (a baby wipe). The wipes were then analyzed by a
 laboratory to determine if there was any lead present. To be acceptable, results had to be below
 200 ug/ft2 for floors, 500 fig/ft2 for window sills, and 800 ug/ft2 for window troughs.
       Over 98 percent (142 of 144) of window sills and over 95 percent (114 of 119) of the
 window troughs met wipe test clearance standards on the first attempt. The success rate for
 floors on the initial wipe clearance test was somewhat lower at 87 percent (148 of 170)1.
 Building A had a higher failure rate on initial floor wipe test (19.6 percent) than did Building B
 (10.2 percent), but the difference was not statistically significant.  The success in achieving
 clearance indicated the effectiveness of the  specifications and the diligence of the contractors'
 personnel at final cleaning.
       There was a statistically significant difference between Building A and Building B when
 wipe test results are compared in terms of the number of attempts required to achieve wipe test
 clearance at each floor location.  It was harder to achieve wipe test clearances on floors in
Building A (where chemical stripping was used) than in Building B (where abrasive methods
were used). The differences between buildings in achieving wipe test  clearances on window sills
 and window troughs were not statistically different.  All units passed clearance wipe testing.

 C-1.   Objective 1: Characterization of the Number of Individual Samples. Work Areas, and
       Housing Units That Pass or Fail Clearance Testing Standards
       Individual dust wipe samples were collected from Albany, Cambridge, and Omaha as part
of the PHA Phase of the HUD Abatement Demonstration Project. From March 1991 through
April 1993,1433 individual dust wipe samples were collected from floors, window sills and
window troughs within 821 rooms in 119 residential units. Floors were the most frequently
        Some difficulty was experienced in achieving wipe test clearance on concrete landings, but the problem was
overcome by more frequent changes of water.
                                           C-4
-------
Table C-1. Clearance Testing Results by Individual Sample, Room, and Residential Unit for
           the PHA Data.
    First
              Floor
               Sill
             Trough
               All
             501
57
 558
443
47
490
88
            402
      411
        383
               391
                103
            364
      371
        349
               356
                105
            1267
73
1340
761
58
819
81
31
38
119
                      111
                      110
119
   Second
              Fioor
               Sill
             Trough
               All
             40
13
  53
 34
         43
         24
              10
       10
         10
                 10
              55
15
  70
 45
11
 56
30
              27
       35
   Third
              Floor
Trough
               All
                            13
                            15
                              11
   Fourth
              Floor
               All
Total
Floor
Sill
Trough
All
554
412
371
1337
78
9
9
96
632
421
380
1433
486
391
356
814
6
1
0
7
492
392
356
821
115
110
110
115
4
1
0
4
119
111
110
119
sampled component (632 samples) followed by window sills (421 samples) and window troughs
(380 samples). Table C-1 presents the number of individual samples, work areas, and residential
units that pass or fail clearance testing within each combination of component type and site visit.
Although 95% (1,2677,1340) of the dust samples collected on the first site visit fell below the
clearance standards of 200 u,g/ft2 for floors, 500 ng/ft2 for window sills and 800 jig/ft2 for
window troughs and 93% (761/819) of the rooms passed clearance on the first site visit, only
68% (81/119) of the residential units passed.  This increase in the failure rate from the percentage
of individual samples that fail to the percentage of rooms and residential units that fail is
attributable to the fact that, on average,  1.6 individual samples were taken per room (1,340/819)
                                          C-5
-------
 and 11.3 individual samples were taken per residential unit (1,340/119) during the first site visit.
 In general, the probability that a residential unit fails clearance increases as the number of
 samples collected increases.
       The failure rate for individual samples taken during the first site visit was much higher
 for floor samples, 10% (57/558), than it was for either window sills, 2.2% (9/411) or window
 troughs, 1.9% (7/371). In addition, more individual floor samples, 4.7 (558/119) were collected
 per residential unit than for window sills, 3.7 (411/111) or for window troughs, 3.4 (371/110).
 Considering each component individually, 28% (31/119) of the residential units would have
 failed based only on the results of floor sampling. If only window sill samples were considered
 then 7% (8/111) of the residential units would have failed and 5% (5/110) of the units would
 have failed based only on the window trough samples.

 C-2.   Objective 2: Characterization of the Distribution of the Dust-lead  Loadings.
       Geometric Mean Dust-lead  Loadings. Variability Within a Housing Unit,  and
       Variability Between Housing Units
       Preliminary assessment of the data indicated that the distribution of dust-lead loading
 clearance sample results was highly  skewed.  A natural logarithm transformation was applied to
 the data.
       For the clearance data collected in the PHA phase of the HUD Demonstration Project,
 Table C-2 lists the geometric mean lead loading and the 95% confidence interval about the mean
 for each site visit and component type. The within-home (room-to-room) variability is larger
than the between homes variability for the first site visit for all three components. The results are
mixed for the second and third site visits. Notice that the lengths of the confidence intervals
 increase successively from the first site visit to the third site visit for any given component. The
 increase hi the length of the confidence intervals is due primarily to the decrease in the number of
samples used to estimate the variance components.
       The geometric mean lead loadings and their 95% confidence intervals are used to
compare trends between site visits for a given component tested and to compare average results
between component types within a site visit.  If the 95% confidence intervals on two geometric
means do not overlap then these two geometric means are statistically significantly different at a
 level less than 0.05. A comparison across site visits shows that,  for a given component, the
geometric means are not significantly different from the first site visit to the third site visit. A
 comparison of the components tested within a site visit show that the floor dust-lead loadings
                                          C-6
-------
 o
 a


 o
U

•o

 10
**
"55
 9
 +•

35
 to
 O
 M

UJ

«-•


 0)
 a

 o
 o
 o
•o

 (0


 a
 C0
 o
•c
*"
 CD


 o
 a>

0   d


 "
 g  o
O  H
CSI

6
 a
                                     (O
                                           o>
                                           CO
             tn
                   o
                   oo
      en
      
                              CD
to
      oo
             oo
                   CM
                         O)
            CM
            CM
                        o
                        CO
      in
      CO

      6
(-.

6
in


O
      CO
      •*

      d
                        oo
                        in
            cq

            o
                  oo
                  CM
                  oo
                  in
                                     ID
                                     co

                                     ci
ro

d
            co
            CQ
8
            oo
            CM
O
CO

6
                        o
                        o
o
o
      ro
      ro
o
in
            in
            CM
CN
CN
                  C9
                  1^

                  CM
                                           CD
                                                       t-.
                                                       CM
                                     co
                                     O
                                     en
                                     Ti-
            CO
            in
            co
co
                  en
                              CO
00
in
in
                                                 r-
                                                 co
co
in
                                                                         CO
                                                                                     00
                                     o
                                     o
                                                 o
                   o
                   o
             a

             o
                         §
                        EE
                  g
                  E

                                                              o
                                                              o
                                                              CO
                                                             co
                                                                            •a
                                                                            .ir
                                                                     C-7
-------
are statistically higher than the window sill dust-lead loadings for the first site visit.  Significant
differences are not shown between floor and window trough dust-lead loadings and between
window sill and window trough dust-lead loadings for the first site visit.  Also, the differences
are not statistically significant for the second site visit among three components.  Note that
estimates from the third and fourth site visits are associated with very small sample sizes and
may not be very reliable for establishing trends.
       Figures C-l and C-2 contain box and whisker plots that present the distribution of dust-
lead loadings from the first and passed clearance visits by component type.  Figures C-3 to C-5
contain box and whisker plots that present the distribution of dust-lead loadings from the first  .
visit for floor, window sill, and window trough, respectively; Figures C-6 to C-8  contain box and
whisker plots that present the distribution of dust-lead loadings from the passed clearance visits.

C-3.   Objective 3: Characterization of the Correlation Between Components Sampled in
       the Same Work Area
       For the PHA Phase of the HUD Abatement Demonstration, Table C-3 displays the
Pearson product-moment correlation coefficients and the associated sample size for the log lead
loading measurements from the first site visit in individual floor, window sill, and window
trough samples taken within the same room. The data show that floor samples are positively
correlated with both window sill and window trough samples. The correlation between window
sills and window troughs is also positive and is the highest of any observed correlation from the
first site visit.  All correlations from the first site visit are significant at the 0.01 level.  The
correlations observed during the first site visit for this data set are similar in magnitude to those
observed for both the Maryland data set and the PHA Phase of the HUD Abatement
Demonstration Project data set.
       The conditional probabilities of a sample passing or failing a standard are given in
Table C-4. These analyses were conducted on two different sets of data, the first set using all
possible paired observations from within the same room. The second subset of data restricted the
analyses to rooms in which floor, window sill, and window trough lead loadings were
simultaneously observed.
       As shown in Table C-4, results from contingency table estimates and normal theory
estimates are generally consistent on the left side of the table, but differ in some cases on the
                                           C-8
-------
             100.0
          sr
          <   10.0
          0)
          .Q

          I   1.0
              0.1
                    tons      WMmrSfe    Widow Tratfis     Rons      WWowSfe   WntowTtogis


                              RstVisi

                                              Surface
Figure C-1.    Comparison of In-Transformed Dust-Lead Loadings from the First Site Visit

              vs. Passed Clearance Results on an Expanded Scale.
                                           C-9
-------









N
<
ra
3
d
i
5
£



800

700



600



500


400

300
200
100
0


















i




1*
!
•
1
i
i
I


1
1 i








1














                  Floors
Window Sills
  Surface
Window Troughs
Rgure C-2.    Box and Whisker Plots of the Distribution of Dust-Lead Loadings from the
              Passed Clearance Data.
                                           C-10
-------
             10000
              1000
          OJ
          <
               100
                10
                     Al
Wood        Vnyl       Akmfoum      Other

                Substrate
Unknown
Figure C-3.   Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the First Site Visit.
                                            C-11
-------
              100000

               10000
                1000
           0
           I
           £
                 100
                 10
                        All         Wood       Vnyl       Ahirinim      Other      Unknown
                                                 Substrate
Figure C-4.    Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
               by Substrate for the First Site Visit.
                                              C-12
-------
              1000000
               100000
           CM    10000
           I
           £1
           Q.
                1000
100
                  10
                        AD         Vlfood        Vmyl      AMum      Other       Unknown

                                                Substrate
Figure C-5.    Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
               Loadings by Substrate for the First Site Visit.
                                             C-13
-------
             1000
          CM   100
          o>
          $.    10
                    AH
Wood
Vinyl
Aluminum
Other
Unknown
Figure C-6.   Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the Passed Clearance Visits.
                                           C-14
-------
             1000
          CM  100
          0)


          TJ
.Q
Q.
               10
                     A3         Wood         Vir^       Aluminum      Other


                                              Substrate
                                                                  Unknown
Figure C-7.   Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings

              by Substrate for the Passed Clearance Visits.
                                            C-15
-------
             1000
          N  100
          0>
               10
                     All
Wood        Vnyl       Aluminum       Other
               Substrate
Untoxjwn
Figure C-8.   Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the Passed Clearance Visits.
                                            C-16
-------
right side. For both types of estimates and both data sets included in the analysis, the probability
that one component passes clearance given that another component passes clearance is very high
(greater than 93%), while the probability that one component fails clearance testing given that
another component in the same room fails clearance testing is fairly low, ranging from 0% to
40%.

Table C-3. Observed Within-Room Correlation Coefficients Between Pb Loading
           Measurements Collected from Floors, Window Sills and Troughs for the PHA
           Data.
         First

                ^Biiii
0.42'
213
0.35'
192
0.58'
201
       a Statistically significantly different from zero at the 0.01 level.

C-4.   Objective 4: Demonstration of the Impact of Composite Sampling on Pass/Fail Rates
       of Houses
       By combining individual sample lead loading results from each component type within a
residential unit, simulated composite sample results were constructed.
       Table C-5 shows the number of residential units that were investigated by the number of
individual samples collected for each component type. For example, there were 25 residential
units which included seven floor dust-wipe samples on the first site visit. Each residential unit
which contained more than four samples from a particular component type resulted in the
estimation of multiple simulated composite sample results in this analysis. For this data set,
summing all the units  that had five or more individual samples, approximately 45% (53/119),
29% (32/111), and 31% (34/110) of the residential units resulted in the estimation of multiple
simulated composite samples from floors, window sills, and window troughs, respectively.
When there were four or fewer individual samples from a component type within a housing unit,
the simulated composite sample included all samples.
       The simulated composite samples constructed by combining the individual samples for
each component type within a residential unit were used for the purpose of evaluating the three
composite sample clearance criteria (Standard, Standard/n, and 2>
-------
C-18
-------
Table C-5.  Number of Residential Units that Contained (N) Individual Samples of Each

           Component Type Based on First Site Visit Clearance Testing Data from the PHA

           Phase of the HUD Demonstration Study.
                      9 +
                      Total
                  Total with N 26
                                     \2
                                     16
                                    25
                   119
                   53
                                30
                                                 23
                                23
                                                 24
   111
    32
                                                             17
                 15
                                            26
                 18
                                                             32
        110
         34
Table C-6. Individual Sample Clearance Results Versus Simulated Composite Clearance

          Results Based on Data from PHA Phase of the HUD Demonstration Study.
                                                             -=
                                            Individual|Sample\Clearance
     Standard
    Standard/n
                 Pass
Inconclusive
                 Fail
                 Pass
Inconclusive
                 Fail
                    88
20
103
                    0
                   ——


                    52
        97
                    27
31
105
                104
                 Pass
   2 x Standard/n
Inconclusive


Fail
                    75
        100
                                             22
                105
                                         CM 9
-------
passed clearance, failed clearance, or yielded inconclusive results based on the simulated
composite samples.  Inconclusive results could only occur in those residential units for which
more than four individual samples were collected for a component type. The many possible
ways of combining five or more individual samples into multiple simulated composite samples
had the potential for creating uncertainty in the decision rule for that residential unit (Table C-6).
The highest inconclusive rates were observed for floors where the maximum was observed when
using the 2x Standard/n Rule, which resulted in 28.3% (15/53) of the houses with more than four
individual samples yielding inconclusive results.  The rates for the other components ranged
from 2.9% (1/34) for window troughs using the Standard/n Rule and the 2x Standard/n Rule to
9.4% (3/32) for window sills using the 2x Standard/n Rule.
       The performance characteristics of each combination of component type and composite
clearance criteria are presented in Table C-7 in terms of sensitivity, specificity, positive
predictive value (PPV), and negative predictive value (NPV). The sensitivity for the Standard/n
Rule is always 1.00 while the specificity for this rule is estimated at 0.59 for floors, 0.94 for
window sills, and 0.99 for window troughs. The specificity for the Standard Rule is always 1.00
while the sensitivity  for this rule is estimated at 0.19 for floors, 0.63 for window sills, and 0.20
for window troughs.  For the 2>
-------
where nijk is the estimated probability of clearance for component^") in house(z) under composite
criterion (fc), and Max,y is the maximum individual sample lead loading result in house(i) for
component(/).
Table C-7. Performance Characteristics of Composite Clearance Criteria Based on Data
           from the PHA Phase of the HUD Demonstration Study.
                                                  Composite Clearance Criteria
                   Sensitivity
                       0.19
                  1.00
                 0.71
       Floors
                   Specificity
                        1.00
                  0.59
                 0.85
                   PPV
                        1.00
                  0.53
                 0.82
                   NPV
                       0.82
                  1.00
                 0.97
                   Sensitivity
                       0.63
                  1.00
                 0.88
      Window
        Sills
Specificity
                                          1.00
                  0.94
                 0.97
                   PPV
                       1.00
                  0.67
                 0.88
                   NPV
                       0.98
                  1.00
                 1.00
                   Sensitivity
                       0.20
                  1.00
                 0.80
      Window
      Troughs
Specificity
                                          1.00
                  0.99
                 1.00
PPV
1.00
1.00
1.00
                   NPV
                       0.98
                  1.00
                 1.00
       For each component type, the estimated relationship between the probability of passing
clearance and the maximum lead loading observed for the individual samples is illustrated in
Figure C-9. The relationship using the Standard/n Rule for window troughs could not be
estimated. It is inestimable because of the special structure of the data. For this data, all of the
simulated composite samples passed when the maximum lead loading was at or below 401
ug/ft2, and all of the simulated composite samples failed when the maximum lead loading was at
or above 813 ug/ft2 with one inconclusive composite result at 524 ug/ft2.  The estimation
problem occurs because between 401 and  813 jig/ft2 there are an infinite number of potential
regression curves which can maximize the likelihood equations; hence the relationship is
inestimable.
       Table C-8 provides parameter estimates and associated standard errors from the logistic
regression models, as well as estimates of the probability of passing clearance (using composite
samples) when the maximum lead loading among all locations included in the composite
sampling scheme is greater than or equal to '/2,1, 2, and 4 times the associated HUD standard for
individual samples.  The Standard Rule's lack of sensitivity is apparent since the chance of a
                                          C-21
-------
residential unit passing clearance is estimated to be greater than 2/3 when there is an individual
floor or sill sample location which has a lead-loading level that is equal to twice the HUD
recommended clearance standard. Conversely, the estimated relationship for the Standard/n Rule
demonstrates this rule's high sensitivity since the probability of passing clearance given a
maximum lead loading of 2 times the HUD interim standard is always near zero. The low
specificity of this rule is also apparent with estimated probability of passing clearance for floor
composite samples as only 0.56 when the maximum individual sample lead loading is equal to
1A HUD Standard.  Of course, because the 2xStandard/n Rule is a compromise between the
Standard and Standard/n Rules, the estimated probability of passing clearance under this rule is
almost always between the estimated probabilities under the other two.
                                          C-22
-------
                          Probability of Passing Clearance Testing Using Composite Floor Samples
                                                          	(2 x Standard/n} Rule
                                                                  (Standard/n} Rute
                              100
 200     300     400     500    600
Maximum Individual Sample Floor Pb Loading
                         Probability of Passing Clearance Testing Using Composite Window Sill Samples
                                                                    Approach

                                                          	Standard Rute
                                                          	(2 x Standard/n) Rub
                                                          	(Standard /n)Ruto
                              250
 500
750    1000   1250    1500   1750   2000
                                   Maximum Individual Sample Window SID Pb Loading
                        Probability of Passing Clearance Testing Using Composite Window Well Samples
                                                                 Standard Rute
                                                         	(2 x Standard/n) Rute
                                                         r	(Standard/n) Rute
                              400    800    1200    1600    2000    2400   2800
                                   Maximum Individual Sample Window Well Pb Loading
                                           3200
Figure C-9.  Estimated Relationship Between the Probability of a Residential Unit Passing
              Clearance Testing versus the Maximum Individual Lead-Loading Result by
              Component Type Based on Simulated Composite Samples from the PHA Data.
                                                 C-23
-------
H_ O)
 o .S

ti
5 E
   l


fS
(0 g>
C 0
   g
CC -g
V> O)

« c
*< ._
   a
   _
   a

$ E
_ a
« o)
^ ^
o a
   P
ox**
•-P «  o>

.2 § .E
« -p S



I   !
M C O
±S « ^
•=
W
«

CC
o

a>
   5 O

   O O

                           8
                           to
                              in
                              CN
                              in
                              1
                                  8
                                  en
                                      8
8
                                      8
                                      • ••

                                      8
                                      in
                                          00
                                          co
                                          CO
                                         co
                                         co
                                             CN
                                             en
                                         O V)
                                         in
               8
                                                     o
                                                     o
                                                     in
                                                     
-------
                        APPENDIX D

Evaluation of the HUD Lead-Based Paint Hazard Control Grant Program
                    (HUD Grantee Program)
                             D-1
-------
                                     Appendix D
        Summary of Background Information for the HUD Grantee Data

       The overall purpose of the Evaluation of the HUD Lead-Based Paint Hazard Control
Grant Program ("HUD Grantee Program") is to measure the relative cost and effectiveness of the
various methods used by State and local government grantees to reduce lead-based paint hazards
in housing [17]. Although national in scope, this program is locally driven and implemented.
Grantees design their own programs, including the methods of recruitment and the treatments
that they carry out.
       There are 14 grantees in the HUD Grantee Program, with 11 starting in FY 1992 (first
round, required to participate in the Evaluation) and 3 starting in FY 1993 (second round,
voluntary participation). Table D-l summarizes the administration, sampling targets, and
abatement methods used for each grantee. Program standard forms and procedures were
developed by the University of Cincinnati (UC) Department of Environmental Health and
National Center for Lead-Safe Housing (NCLSH). UC and NCLSH are also responsible for data
analysis and reporting, training, and support of grantee data collection and recording.  Data
management quality control was done by UC.
       Data being collected in the HUD Grantee Program are environmental, biological,
demographic, housing, cost, and hazard-control information. Measurements of lead in dust,
paint, soil, and blood are collected, though not all  grantees collect all these measurements. Pre-
intervention, immediate post-intervention (clearance results), and 6- and 12-month post-
intervention data are collected. However, only selected homes (estimated 800 homes) are
included in the data collection for the 24- and 36-month post-interventions.
       The data available to this investigation were collected through January 1999 and were
released by the University of Cincinnati in February 1999. This appendix contains summaries
and analysis results for this set of HUD Grantee data.
       Based on the HUD Grantee Program's Fifth Interim Report (data collected through
September 1,1997), the following bulleted list provides building characteristics, occupancy
status, environmental sampling, and clearance of dwelling units in the program [13].
                                          D-2
-------
Table D-1.  Administration, Sampling Targets, and Abatement Methods by Grantee

Alameda County,
CA""
Baltimore, MD
Boston, MA
California
Chicago, IL
Cleveland, OH
Massachusetts
Minnesota
New Jersey
'•i'j- & ^**r ^^* sWfllSilfcafe'
||NumDer^^
ISubgranteesl
1
1
1
4»i
1fc)
1
4«»
3">
7'"

High risk cities, many homes with
lead poisoned child
Three target communities, two
mainly investor-owned, one mainly
owner-occupied. In all three,
many vacant and many
row/houses
Most have received an order to
abate based on presence of a lead
poisoned child
Older homes in low-income
neighborhoods
Selection based on report of lead
poisoned child and compliance
interview
Scattered-Site Program (SSP):
targets lead poisoned children
Intensive Neighborhood Program
(IMP); less intensive and less
costly remediation, contiguous
houses, some previously or
concurrently rehabilitated houses
managed by a large non-profit
agency, high rates of childhood
lead poisoning in these
neighborhoods
Many homes under orders to
abate based on presence of a lead
poisoned child.
Two subgrantees target only lead
poisoned children, one targets
homes with deteriorated
conditions
Concurrent renovation/
rehabilitation, many vacant homes

Abatement and interim control
work on interior, minimal
exterior work, extensive soil
treatment
Abatement and interim
controls, window replacement,
no soil treatment
State regulations dictate
significant abatement. No soil
treatment.
Interior and exterior abatement,
window replacement.
Infrequent soil treatment.
Abatement and interim control
work
SSP: abatement and interim
control work, including soil
INP: contiguous houses get
emphasis on education and
involvement, rehab houses get
additional lead remediation
State regulations dictate
significant abatement. No soil
treatment.
Interim control (paint
stabilization and friction
controls on all deteriorated
LBP). Some have exteriors
enclosed with vinyl siding and
other coverings
Most get full abatement, some
get complete removal of all
lead painted components
                                        D-3
-------
Table D-1. Continued
  New York City, NY
           Neighborhoods with highest
           percentage of lead poisoning,
           usually multifamily (6 to 20 units),
           one administrator targeted
           newborns with deteriorated
           housing, other administrator
           enrolled houses that were in  an
           ongoing rehabilitation program
                                                                            Interim controls of
                                                                            deteriorated, friction or impact
                                                                            surfaces.  No soil treatment.
  Rhode Island
           Must meet Section 8 Housing
           Quality Standards, owner must
           own no more than 12 units
                                    Abatement (window
                                    replacement and soil
                                    treatment), interim controls
                                    (correct friction surfaces and
                                    defective paint)
  Vermont
           Each source has own target:
           poisoned children, non-profit
           developers, unsolicited
           applications
                                                                    lead
                                    Developers get substantial
                                    rehabilitation, private units get
                                    less work, one neighborhood
                                    gets cleaning, rarely performs
                                    soil intervention (but does have
                                    pre-intervention  soil
                                    measurements)
                                        Each subgrantee has own criteria
 Wisconsin
                                               Depends on subgrantee.
                                               Pre-intervention soil
                                               measurements but no soil
                                               remediation
 Milwaukee
NA
Lowest income neighborhoods,
lead poisoned children, projects
receiving HUD rehabilitation funds,
direct outreach
Cleaning and education, interim
controls, abatement, housing
rehabilitation, no soil treatment
(a)  Four cities:  Alameda, Berkeley, Emeryville, and Oakland.
(b)  Four subgrantees: Los Angeles, San Francisco, San Diego, and Visalia (Central Valley).
(c)  Five different neighborhoods each do their own enrollment.
(d)  Four subgrantees: Brockton, Chelsea, Lawrence, and Worcester.
(e)  Three subgrantees: Duluth, Minneapolis, and Saint Paul.
(f)  Eleven original subgrantees, four of which withdrew from the program. Seven subgrantees are still in the program:
    Camden, Elizabeth, Englewood, Irvington, Jersey City, Newark, and Paterson.
(g)  Two different administrating organizations for three targeted neighborhoods: Brooklyn, the Bronx, and Manhattan.
(h)  21 local offices, one administrator.
(i)  Enrolled units from several sources.
(j)  Not including Milwaukee. One administrator, but many decisions made locally.  Twelve subgrantees:  Chippewa County,
    Eau Claire County, Madison, Manitowoc, Oshkosh, Richland County, Rock County, Sheboygan, Superior, Wausau, West
    Allis, and Wisconsin Rapids.
(k)  Descriptions of the lead hazard control options use the terms "abatement" and "interim control" as they are used in Title X
    of the 1992 Housing and Community Development Act.  Abatement denotes that class of treatment that permanently
    remove or cover lead-based paint hazards. HUD defines permanent treatment as treatments expected to last at least 20
    years. Building component removal, surface enclosure, and paint removal are common methods of abatement. Interim
    controls include treatments that eliminate immediate lead-based paint hazards, but do so in a manner that is not expected to
    last 20 years. Repainting, friction reduction on windows and doors, and cleaning are common interim control methods.
                                                    D-4
-------
Building Characteristics
       (1) Types
          •   Single family detached (32%)
          •   Single family attached (25%), rowhouses - mostly in Baltimore
          •   2-4 Unit Multi-family (37%)
          •   > 4 Unit Multi-family (6%), 85% of NY enrollments were of this type, averaging
              14 units per building
       (2) Age of Housing
              Less than 1% of the enrolled buildings built after 1959.
              90% pre-1940
              Median for Cleveland, Massachusetts, Milwaukee, Minnesota and Vermont is pre-
              1910
              Median for Baltimore, Chicago and Rhode Island was in 1920's
              Median for California was in 1930's

Occupancy Status
          •   20% vacant prior to intervention
          •   Baltimore had 60% pre-intervention vacancy rate
          •   Vermont, NY, NJ had from 24% to 34% pre-intervention vacancy rates
          •   All others had below 14% pre-intervention vacancy rate

Environmental Sampling
          •   Dust is collected from 7-9 locations during each phase of the Evaluation.
          •   Single-surface dust wipe samples are collected from:
              •  Floor (bare or carpeted): Interior Entry, kitchen, child's play room (or living
                 room), youngest child's bedroom (or smallest room), next youngest child's
                 bedroom (if present). Note that only bare floor dust samples were included in
                 the analyses in this report.
              •  Interior window sill: kitchen, youngest child's bedroom (or smallest room)
              •  Window trough: child'splay room (or living room), next youngest child's
                 bedroom (if present)

Clearance of the Dwelling Units
       •  Program requires clearance after intervention
       •  Grantees followed 1990 HUD Interim Guidelines
          Floors:        200 ug/ft2 (lowered to 100 ug/ft2 in 1994-1995)
          Sills:         500 ug/ft2
          Troughs:       800 ug/ft2
       •  Exception:
          Floors: Grantees used 200 ug/ft2, 100 ug/ft2 or a locally established level
          •   Cleveland, Chicago, New Jersey, New York City used 100 ug/ft2
          •   Minnesota used 80 ug/ft2
       •  28% of the dwelling units failed the initial clearance dust lead test There is a wide
          variation in the clearance rates for the grantees, with rates of initial failure for units
          ranging from 8 to 50 %.
                                         D-5
-------
       All dwellings in the HUD Grantee program are required to "meet clearance" after the
intervention is complete. In other words, dust wipe tests must demonstrate that the amount of
leaded dust on components in all treated rooms does not exceed levels designated by HUD.  For
the first and second rounds of the HUD Grantee Program, clearance levels were set at 200,500,
and 800 ug/ft2 for floors, window sills, and window troughs, respectively. In 1994-1995, HUD
and EPA released new guidance that lowered the clearance level on floors, from 200 ng/ft2 to
100 ug/ft2. Since then, HUD has allowed grantees to use either 200 ug/ft2,100 ug/ft2, or a
locally established level if less than 200  ug/ft2. Five grantees used clearance levels less than 200
ug/ft2 for floors:  Cleveland, Chicago, New Jersey, and New York City used 100 jig/ft2, while
Minnesota used 80 ug/ft2.
       Analyses of the HUD Grantee data were performed separately for two groups of grantees
and presented in the following sub-Appendices: Dl and D2.  Sub-Appendix Dl presents analysis
results from nine grantees (Alameda County, Baltimore, Boston, California, Massachusetts,
Milwaukee, Rhode Island, Vermont, and Wisconsin) that used the original HUD Interim
Guidelines clearance standards, i.e., 200, 500, and 800 ug/ft2 for floors, window sills, and
window troughs, respectively. Sub-Appendix D2 presents analysis results from the other five
grantees (Cleveland, Chicago, New Jersey, New York City, and Minnesota) that used a lower
floor dust-lead clearance standard (i.e., either 100 ug/ft2 or 80 ug/ft2).
       Immediately following lead hazard control intervention (within three working days), dust
wipe samples were taken and tested as part of the standard HUD clearance procedure. These
clearance testing results were recorded in the "Dust Sample Collection Form." If the amount of
lead in dust on any tested component in a dwelling unit exceeded the clearance level, re-cleaning
and re-testing of the failed surface would be performed. These re-testing clearance results were
recorded in the "Clearance Dust Re-testing after Failure Form." Each retest was numbered: the
first retest was recorded as Failure #1, the second as Failure #2, and so forth.
       To be consistent with other data sources presented in this report, the clearance testing
dust-lead data taken from the "Dust Sample Collection Form" (immediately following
intervention) were labeled as "First Site Visit" data. The clearance re-testing data taken from the
"Clearance Dust Re-testing after Failure Form" were labeled as "Second Site Visit" for the first
retested dust clearance data (Failure #1), as 'Third Site Visit" for the second retested dust
clearance data (Failure #2), and so on.
                                          D-6
-------
       Note that in the available HUD Grantee data of January, 1999, thirty-one (31) dwelling
units had dust clearance re-testing results recorded in the "Clearance Dust Re-testing after Failure
Form" but no clearance data shown in the "Dust Sample Collection Form." Therefore, while
these 31 units were not included in the first site visit analysis, they were included in the second
or third site visit analysis. Eighteen (18) of these 31 dwelling units belong to the first analysis
group (grantees that used 200 ng/ft2 as floor dust-lead clearance standard) and 13 units belong to
the second analysis group (grantees used 100 fig/ft2 or 80 Jig/ft2 as floor dust-lead clearance
standard). There is also a case where dwelling units had records of Failure #2 but no records for
Failure #1 in the "Clearance Dust Re-testing after Failure Form." Among the 113 dwelling units
which had this case, 63 belong to analysis group 1 and 50 belong to analysis group 2. These
dwelling units were not included in the second site visit analysis but they were included in the
third site visit analysis. However, they all had first site visit data and therefore were included in
the first site visit analysis.
                                            D-7
-------
-------
           APPENDIX D1

Grantees with Higher Floor Dust-Lead
      Clearance Standard in the
       HUD Grantee Program
               D1-1
-------
                                    APPENDIX D1
                   GRANTEES WITH HIGHER FLOOR DUST-LEAD
                           CLEARANCE STANDARD IN THE
                              HUD GRANTEE PROGRAM
       Appendix Dl presents analysis results on the clearance data collected from nine grantees
that used the HUD Interim Guidelines clearance standards, i.e., 200,500, and 800 fig/ft2 for
floors, window sills, and window troughs, respectively. This group includes grantees Alameda
County, Baltimore, Boston, California, Massachusetts, Milwaukee, Rhode Island, Vermont, and
Wisconsin
       Note that, as explained in Appendix D, 18 dwelling units did not have first site visit dust-
lead clearance data and were not included in the first site visit analysis. However, since these
dwelling units had other site visit data, they were included in those analyses. Sixty-three (63)
dwelling units did not have second site visit data, but did have other site visit data (including first
site visit data) and were included in those site analyses.

D1-1.  Objective 1: Characterization of the Number of Individual Samples. Work Areas.
       and Housing Units That Pass or Fail Clearance Testing Standards
       Individual dust wipe samples were collected from floors, window sills and window
troughs as part of the HUD Grantee Program. A total of 17,231 samples were taken from 11,202
rooms in 2,150 residential units from nine grantees using the higher floor dust-lead clearance
standard of 200 jig/ft2. Table Dl-1 presents the number of individual samples, work areas and
residential units that passed or failed clearance testing within each combination of component
type and site visit. Approximately 93% (15,979/17,231) of the dust samples were collected
during the first site visit to a residential unit.  On the first site visit, 95% (15,139/15,979) of the
dust samples fell below the clearance standards  of 200 ng/ft2 for floors, 500 jig/ft2 for window
sills and 800 ng/ft2 for window troughs while 92%  (9,508/10,299) of the rooms and 75%
(1,608/2,132) of the residential units passed clearance. The increase in the failure rate from the
percentage of individual samples that fail to the percentage of rooms and residential units that
fail is attributable to the fact that if any individual sample exceeded the standard and failed
clearance, then both the room and residential  unit also failed clearance.
                                         D1-2
-------
Table D1-1.  Clearance Testing Results by Individual Sample, Room, and Residential Unit
              from the Grantees with Higher Floor Dust-lead Clearance Standard in the
              HUD Grantee Program

First
Second
Third
Fourth
All Visits
£i,compone?
*nf~TesfedJ
I •*-•;:?•:" ::'«v*4
Floor
Sill
Trough
All
Floor
Sill
Trough
All
Floor
Sill
Trough
All
Floor
Sill
Trough
All
Floor
Sill
Trough
All


7711
4601
2827
15139
550
261
190
1001
74
42
17
133
10
2
2
14
8345
4906
3036
16287

468
197
175
840
58
20
12
90
11
2
1
14
0
0
0
0
537
219
188
944
Sofi
8179
4798
3002
15979
608
281
202
1091
85
44
18
147
10
2
2
14
8882
5125
3224
17231
&&>£&^*$*&^?&&^£.r'2.Jlf*:. •*'i:£W^
^€ RoomstSampIed &$&
-j5p^.i?Wvr%'»~"TJ!*> '- *«**•*& .iC"! .»Ls' "r ' ?*•• -*!V '
Ufesil^r
SsPass:&
7083
4394
2686
9508
506
255
188
795
68
42
17
104
9
2
2
13
7652
4683
2889
10396
Jfe»ilS£
459
196
173
791
54
20
12
85
11
2
1
14
0
0
0
0
470
199
174
806
•ySS&SSwtf
fTotal;
7542
4590
2859
10299
560
275
200
880
79
44
18
118
9
2
2
13
8122
4882
3063
11202
^i<^Hous^^iai9|>!e^^^
^PSsislt
1775
1854
1546
1608
289
190
143
420
37
33
15
51
9
2
2
12
2054
2004
1678
2050
pill
317
177
151
524
44
19
12
68
11
2
1
13
0
0
0
0
48
38
30
100

2092
2031
1697
2132
333
209
155
488
48
35
16
64
9
2
2
12
2102
2042
1708
2150
       The failure rate for individual samples during the first site visit was at 5.7% (468/8,179)
for floors, at 4.1% (197/4,798) for window sills, and at 5.8% (175/3,002) for window trough
samples. The failure rate for individual floor samples showed an increase from the first site visit
(5.7%, 468/8,179) to the second site visit (9.5%, 58/608) and again from the second visit to the
third site visit (12.9%, 11/85).  This trend is not as obvious in window sill and window trough
samples.
       Of the 2,132 dwelling units in Table Dl-1 for which initial post-intervention clearance
sampling (first site visit) data were available, 75.4% (1,608/2,132) passed on the first attempt. A
                                          D1-3
-------
total of 15.2% (317/2,092) units had at least one floor location with a dust-lead loading above the
clearance level. This failure rate was higher than that reported for window sills (8.7%,
177/2,031) or for window troughs (8.9%, 151/1,697).

D1-2.  Objective 2:  Characterization of the Distribution of the Dust-lead Loadings.
       Geometric Mean Dust-lead Loadings. Variability Within a Housing Unit, and
       Variability Between Housing Units
       Preliminary assessment of the data indicated that the distributions of dust-lead loading
clearance sample results were highly skewed. Therefore, a natural  logarithm transformation was
applied to the data.
       For the clearance data collected from nine grantees with the higher floor dust-lead
clearance standard, Table Dl-2 lists the geometric mean dust-lead loading, variance components
(within-home variability and between-home variability associated with these loadings), and a
95% confidence interval for the geometric mean, calculated for each site visit and component
type. Notice from this table that all of the variance components are within a single order of
magnitude of each other. Also, the  lengths of the confidence intervals increase from the first site
visit to the third site visit for any given component. The increase in the length of the confidence
intervals is due primarily to the decrease in the number of samples  used to estimate the variance
components.
       The geometric mean lead loadings and their 95% confidence intervals were used to
compare trends between site visits for a given component type and to compare average results
between component types within a site visit. If the 95% confidence intervals on two geometric
means do not overlap, then these two geometric means are statistically significantly different at a
level less than 0.05.  A comparison across site visits shows that the geometric means, from the
first site visit to the second site visit increase significantly: 8.8 ug/ft2,17.6 jig/ft2 for floors, 11.3
ug/ft2, 33.7 ug/ft2 for window sills, and 26.5 ug/ft2,45.6 ug/ft2 for window troughs. A
comparison of the components tested for the first site visit shows that the floor dust-lead  loadings
                                         01-4
-------
 0
 u
 c
 2
 (D
 9
O
•o
 a
 CD
 o
 o
 a>

.?
 CO
 o
 I
o

I
 CO
 o
 w
01
**

 *  c
 S  E
 i  a
    E2
    D)
 O  o

?*
 U  Q*
>
 g  J
 a>  «
 E  c

a  w

c\i

5
1=
                                   CO
                                   CO

                                   OJ
      00
      LO

      CN
      CO
      oJ
      CM
           to
           to
                                                    CN
           CO
           CO
           to
      LO
      o>
                                                               
O
O
CD
                                                                                          00
                                                                                          CO
                                                                                                CO
                                                                                                CN
                                                                                                O
CO
IX
      00
           
                00
           O)
           CO
           o
                                                                          O)
                                                                          6
                                                                               o
                                                                               o
                 o
                 o
                                                                                          o
                                                                                          o
                                                                                          o
           O)
           CM
                                  CO

                                  CO
                                  CM
                                  O)
                                  o
                                  CN
                                        CO
      CO
      O
      CM
           LO
           CD
           CM
      IX
      O
      CO
           CD

           IX
     CO
     CO
     CO
                 CO
                 CO
§
CM
                CD
                LO
                t
                                                              10
                                                              10
           cn
           CM
                                                                          CN
                                                                          o
                                                                          10
           00
           IX
           00
                            CO

                            CO
                                                                                          CO

                                                                                          LO
           CO
           *t
           10
                                  00
      10
      CO
                                             CO
                                                                                     CO
                                                                                          CN
                                                                                                CN
                                  LO
                                  |X
                                        o
                                        o>
                                        to
           O)
           to
           CO
           CN
           o
           CO
           LO
           to
           rx
           CN
      o
      o
      CM
                                                                    O)
                                                                    IX
                                                                               00
                      CM
                            CM
                                  CO
                                  IX

                                  CO
      CO
      o>
      |X
     CM
     §
     CO
     00
     o
     CO
CO
CM
                                                              CM
                                                              o
                                                              CM
LO
00
                      CO
                                 CN
                                       CM
                                   o
                                   o
                                        C/5
                 o
                 o
                            o
                            o
                                                                          CO
                                  cn
                                  D
                                  o
                            o
                            o
                                                                                          
                                                         CO
                                                        3
                                                        o
                                                            D1-5
-------
were significantly less than the window sill dust-lead loadings, which were in turn less than the
window trough dust-lead loadings. Note that estimates from the fourth site visit were associated
with very small sample sizes and may not be reliable for establishing trends.
       Figures Dl-1 and Dl-2 contain box and whisker plots that present the distribution of
dust-lead loadings from the first and passed clearance visits by component type. Figures D1-3 to
Dl-5 contain box and whisker plots that present the distribution of dust-lead loadings from the
first visit for floor, window sill, and window trough, respectively; Figures Dl-6 to Dl-8 contain
box and whisker plots that present the distribution of dust-lead loadings from the passed
clearance visits.
D1-3.  Objective 3:  Characterization of the Correlation Between Components Sampled in
       the Same Work Area
       The relationships in dust-lead loadings among floor, window sill, and window trough
wipe samples collected from within the same room are another important aspect of examining a
clearance testing program. By estimating linear correlation coefficients, the strength of these
relationships can be assessed. Table Dl-3 displays the Pearson product-moment correlation
coefficients and associated sample sizes for the log-transformed dust-lead loading measurements
of individual floor, window sill, and window trough samples taken within the same room. For the
first site visit, dust-lead loadings for floor samples were significantly positively correlated (at the
0.01 level) with loadings for both window sill and window trough samples, as was the correlation
between window sills and window troughs. The correlations from the second site visit were
again significantly positive between floor and sill dust-lead loadings and between  sill and trough
dust-lead loadings. The correlation between floor and trough samples from the second site visit,
however, was not significantly different from zero.  The correlations from the third site visit
were not statistically significant.
       The conditional probabilities of a sample passing or failing a standard are given in
Table Dl-4. These analyses were conducted on two different sets of data, the first set using all
possible paired observations from within the same room.  The second subset of data restricted the
analyses to rooms in which floor, window sill, and window trough lead loadings were
simultaneously observed.
       As shown in Table Dl-4, results from contingency table estimates and normal theory
estimates are consistent on the left side of Table Dl-4, but are not always consistent
                                          Dl-6
-------
             100.0
          <   mo
           01

          T)
          fe   1.0
               0.1
                     Floors      WhfawSK    Window Troughs      ROOTS       WndowSfc    Wndow Troupes
                               RrdVtsit                         Passed Clearance
                                               Surface
Figure D1-1.  Comparison of In-Transformed Dust-Lead Loadings from the First Site Visit
              vs. Passed Clearance Results on an Expanded Scale.
                                            D1-7
-------
             800




             700




             600



          CM

          I  500
          °>400
          X)
          n
300




200]




100




  0
                  Floas
                               Window Sills


                                 Surface
Window Troughs
Figure D1-2.  Box and Whisker Plots of the Distribution of Dust-Lead Loadings from the

              Passed Clearance Data.
                                           D1-8
-------
             10000
              1000
          CVi
          <
          3

          9   100
               10
                               Wood
Vinyl      Alurrmum      Other      Unknown
Figure D1-3.  Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the First Site Visit.
                                           D1-9
                     U.S. EPA Headquarters Library
                            Mail code 3201
                     1200 Pennsylvania Avenue NW
                        Washington DC 20460
-------
              100000
              10000
               1000
                100
                 10
                  1
                                 Wood       Vnyi      Akmhin
                                                Substrate
Other
Unknown
Figure D1-4. Box and Whisker Plots of the Distribution of Window Siti Dust-Lead Loadings
              by Substrate for the First Site Visit.
                                            D1-10
-------
             1000000
              100000
          M   10000
                1000
          S.     100
                  10
                   1
                        Al         Wtood        Vinyl      Aluminum
                                                Substrate
Other
Unknown
Figure D1-5.  Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the First Site Visit.
                                            D1-11
-------
             1000
          CM   100
          o>
          3
          £    10
                    Al
Wood        Vnyf       Alumirun       Other
                Substrate
Unknown
Figure D1-6.  Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the Passed Clearance Visits.
                                           DM2
-------
             1000
          CM   100
                    AO
Wood        Vinyl       Aluminum      Other
                Substrate
Unknown
Figure D1-7.  Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
              by Substrate for the Passed Clearance Visits.
                                           D1-13
-------
             1000
          CM  100
               10
                     AB
Wood
Vnyl
Aluminum
Other
Unknown
Figure D1-8.  Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the Passed Clearance Visits.
                                           D1-14
-------
Table D1-3.   Observed Within-Room Correlation Coefficients Between Dust-Lead Loading
              Measurements Collected From Floors, Window Sills and Window Troughs
              from the Grantees with  Higher Floor Dust-lead Clearance Standard in the
              HUD Grantee Program.

                                        C'-'iii-.ij:* *--•• --••
                                        ;$'FIoors;and
                                        ---~-   - '-
                                                      „	^i1-*1"? -3i:
     First
0.46'
3021
0.39a
1558
0.39"
324
   Second
0.30a
 101
-0.05
 51
0.70"
 15
    Third
0.31
 18
-0.09
       * Statistically significantly different from zero at the 0.01 level.
       * Statistically significantly different from zero at the 0.05 level.
on the right side of the table. For both types of estimates and both data sets included in the
analysis, the probability that one component passes clearance given that another component
passes clearance is very high (greater than 92%), while the probability that one component fails
clearance testing given that another component in the same room fails clearance testing is fairly
low, ranges from 0% to 33%.

D1-4. Objective 4: Demonstration of the impact of Composite Sampling on Pass/Fail
       Rates of Houses
       Table Dl-5 provides, for each component type, the number of residential units having a
given number of samples collected from a given component type.  Across component types, most
of the residential units had fewer than four clearance samples collected. Data for residential units
having more than four samples collected from a given component type were used in constructing
multiple simulated composite samples.  When there were four or fewer individual samples from
a component type within a housing unit, the simulated composite sample included all samples.
From Table Dl-5, the number of homes with five or more samples was 766 for floors, 66 for
window sills, and 22 for window troughs. Therefore, data for approximately 37% (766/2,092),
3% (66/2,031), and 1% (22/1,697)  of the residential units were used when constructing multiple
simulated composite samples from floors, window sills, and window troughs, respectively.
                                         DM5
-------
i
O>
o
o

CM

X

CM

O)


'I
z>

•o
0>
w
Ul
D>

^P
«)

£
O
o
c **
(0 <0

«>
•2 »
O 2

0W

is
•* _
09 E

i o

I5
•S §
2-S
°-^

g g
gffl

I -
•g o
C O
O £
o h-
t
»-
O

_o
A
 CO
         I
            VjjSrtv;-..

             '!'"
            11:

                          rs
                          IT)
                          cs
                                CM
                                o
                                CM

                         8
                             §
81
mil

                                 _
                             in
                             o>
                                      s
                                      in
                                      A[
                                      52.

                                      §
                                      CM
                                         §
                                         05

                                         cl
                                         CO
                                         in
                                         o>

                                            o
                                            §11
                       col
                       m\\
                       roll

                       oil
                               0)
                               If
                               o>

                               o
                      ir>
                      ID I
                      o>
                                u>
                                o
                                en

                                6
               ^t
               in
               OJ

               O
                      g
                             
-------
Table D1-5.   Number of Residential Units that Contained (N) Individual Samples of Each
              Component Type Based on First Site Visit Clearance Testing Data from the
              Grantees with Higher Floor Dust-lead Clearance Standard in the HUD Grantee
              Program.
                                                  inap^siiiifei
                              236
                      263
                      732
                              413
                      1068
                      727
                              334
                      501
                      170
                              343
                      133
                       46
                              258
                       41
                       13
                              288
                       20
                               142
          8
 36
         9 +
 42
         Total
2092
2031
1697
    Total with N 2:5
 766
 66
 22
       For each component type within a residential unit, the set of individual clearance sample
lead loading results for the first site visit were used to construct simulated composite samples for
the purpose of evaluating the three composite sample clearance criteria introduced hi Section
5.4.2 (Standard, Standard/n, and 2>
-------
 Table D1-6.  Numbers of Residential Units that Pass or Fail Clearance, Based on Individual
              Sample Clearance Results vs. Simulated Composite Clearance Results, Using
              Data from Grantees with Higher Floor Dust-lead Clearance Standard.
                                            IndiviauaUSamplelClearanca
                Pass
    Standard
Inconclusive
                Fail
                   1775
         97
         1854
 72
         63
                             157
                            98
1546
33
                                     117
                Pass
   Standard/n
Inconclusive
                Fail
                   1622
                  1797
78
                    75
        317
          55
177
                           1527
 18
                                      0
151
                Pass
  2xStandard/n
Inconclusive
                Fail
                   1730
          8
         1829
 8
28
62
                    17
        247
          25
167
1533
 13
148
       Values of the four performance characteristics defined in Section 5,4.3 (sensitivity,
specificity, positive predictive value, negative predictive value) are presented in Table Dl-7 for
each combination of component type and composite clearance criterion. By design, the
sensitivity for the Standard/n Rule is always 1.00 (as all sets of simulated composite samples
would fail if at least one individual sample result failed) while the specificity for this rule is
estimated at 0.91 for floors, 0.97 for window sills, and 0.99 for window troughs. In contrast, the
specificity of the Standard Rule is always 1.00 (as all sets of simulated composite samples would
pass if all individual samples passed) while the sensitivity for this rule is estimated at 0.50 for
floors, 0.55 for window sills, and 0.77 for window troughs. The 2>
-------
Table D1-7.   Performance Characteristics of Composite Clearance Criteria Based on Data
              from Grantees with Higher Floor Dust-lead Clearance Standard.
      Floors
                   Sensitivity
                   Specificity
                   PPV
                   NPV
0.50
1.00
1.00
0.95
1.00
0.91
0.81
1.00
0.78
0.97
0.94
1.00
   Window Silts
                   Sensitivity
                   Specificity
                   PPV
                   NPV
0.55
1.00
1.00
0.96
1.00
0.97
0.76
1.00
0.94
0.99
0.87
1.00
                   Sensitivity
 Window Troughs
                   Specificity
                   PPV
                   NPV
0.77
1.00
1.00
0.98
1.00
0.99
0.89
1.00
0.98
0.99
0.92
1.00
       The above estimates of the performance characteristics illustrate that the three composite
clearance testing criteria have different specificity and sensitivity rates.  These rates correspond
to the consistency between clearance decisions and the true lead hazards present in the various
locations sampled (assuming the individual sample lead-loading results  are representative of
these lead hazards). To further characterize the performance of each of the three composite
clearance criteria, the following logistic regression model was fitted to clearance data for each
combination of component type and composite clearance criterion:
where nijk is the estimated probability of clearance for component(/) in house(i) under composite
criterion (£), and Max,-, is the maximum individual sample lead loading result in house(z) for
component(/).
                                         D1-19
-------
       For each component type, the estimated relationship between the probability of passing
clearance and the maximum lead-loading is presented graphically in Figure Dl-9.  In this figure,
the solid, long-dashed, and finely-dashed lines represent the estimated relationship for the
Standard,  Standard/n, and 2>
-------
                          ProtebBy of Pasak^j Ct
                       1.0
 S UshQ Composite Floor Sempte*
                              100
                                     200    300    400    coo    000
                                     Maximum kxMduaJ Sampto ROOT Pb Loading
                                                                    TOO
                                                                          BOO
                        Pretoobtitty of Porafcig
U*ig Comports Window SB Samptaa
                              2SO
                                    500     78D   1000   12GO    1600
                                  MEDdmum HMdUal Sarr^n. Wnttow SB F* baadUg
                                                                   17GO
                                                                         sooo
                        ProbabBy of PBSS*« CteennM lattng Udr« CompoBte WMow WM SenfUae
                                                         	(2 x SuintKd / n) HUa
                                                                   n) FU»
                                                      2000
                                                            2400
                                                                   2800
                                                                         3200
                                  Maximum MMduaJ San** WMnv Wai Pb Laadng
Figure D1-9.  Estimated Relationship Between the Probability of a Residential Unit Passing
               Clearance Testing versus the Maximum Individual Lead-Loading Results by
               Component Type Based on Simulated Composite Samples from the Grantees
               with Higher Floor Dust-lead Clearance Standard in the HUD Grantee Program.
                                             D1-21
-------

Dl-22
-------
          APPENDIX D2

Grantees with Lower Floor Dust-Lead
     Clearance Standard in the
       HUD Grantee Program
               D2-1
-------
                                    APPENDIX D2
                    GRANTEES WITH LOWER FLOOR DUST-LEAD
                           CLEARANCE STANDARD IN THE
                              HUD GRANTEE PROGRAM
       Appendix D2 presents analysis results on the clearance data collected from five grantees
that used a lower floor dust-lead clearance standard (i.e. 100 ug/ft2 or 80 ug/ft2).  The clearance
standards still remained at 500 ug/ft2 and 800 ug/ft2 for window sills and window troughs,
respectively.  The five grantees are Cleveland, Chicago, New Jersey, and New York City which
used 100 ug/ft2 as a floor dust-lead clearance standard, and Minnesota which used 80 ug/ft2.
       Note that, as explained in Appendix D, 13 dwelling units did not have first site visit dust-
lead clearance data and were therefore not included in the first site visit analysis. However, since
these dwelling units had data for other site visits, they were included in those analyses. Fifty
(50) dwelling units did not have second site visit data but did have data for other site visits
(including first site visit data) and were included in those analyses.
       To be consistent with other data sources presented in this report, a floor dust-lead
clearance sample in the analyses presented below was labeled as a "pass" if its loading was
below 200 ug/ft2 and  as a "fail" if its lead loading was greater than or equal to 200 ug/ft2, despite
the lower clearance standard used by the grantee.

D2-1.  Objective 1:  Characterization of the Number  of Individual Samples. Work Areas.
       and Housing Units That Pass or Fail Clearance Testing Standards

       Individual dust wipe samples were collected from floors, window sills and window
troughs as part of the  HUD Grantee Program. A total of 7,664 samples were taken from 4,936
rooms in 1,038 residential units from five grantees using the lower floor dust-lead clearance
standard (either 100 or 80 ug/ft2). Table D2-1 presents the number of individual samples, work
areas and residential units that passed or failed clearance testing within each combination of
component type and site visit. Approximately 93% (7,135/7,664) of the dust samples were
collected during the first site visit to a residential unit. On the first site visit, 95% (6,782/7,135)
of the dust samples fell below the clearance standards of 200  ug/ft2 for floors, 500 ug/ft2 for
window sills and 800 ug/ft2  for window troughs while 93% (4,191/4,522) of the rooms and 78%
                                         D2-2
-------
(796/1,025) of the residential units passed clearance.  The increase in the failure rate from the

percentage of individual samples that fail to the percentage of rooms and residential units that

fail is attributable to the fact that if any individual sample exceeded the standard and failed

clearance, then both the room and residential unit also failed clearance.

 Table D2-1. Clearance Testing Results by Individual Sample, Room, and Residential Unit
              from the Grantees with Lower  Floor Dust-lead Clearance Standard in the HUD
              Grantee Program
W&r^
ilvisfe^
First
Second
Third
Fourth
Total
iXw^o^sssst
|Componenti
M^fce'stealfl
Floor
Sill
Trough
All
Floor
Sill
Trough
All
Floor
Sill
Trough
All
Floor
Sill
Trough
All
Floor
Sill
Trough
All
lISXvu*"**i'''l^i?"**ii*il*s"'!B!*
^fe?OT« SamRlesJf
Ipall
3486
2068
1228
6782
226
83
90
399
30
18
19
67
7
0
4
11
3749
2169
1341
7259
3sf&*"*l?ij!|l
156
79
118
353
19
8
18
45
3
0
4
7
0
0
0
0
178
87
140
405
SSSSfflSs*:
STotal?
3642
2147
1346
7135
245
91
108
444
33
18
23
74
7
0
4
11
3927
2256
1481
7664
's&BSf™™''''**? !'": •« • j'-'~ :.-'
-MJRoomslSampled.f^
••%t&l^*&-'-^'-&$#y&^r-^''*^'Z'''!:' •sftjitvs-

3410
2044
1221
4191
216
82
90
350
30
18
19
64
7
0
4
11
3650
2138
1333
4595
^Fail-T?
154
79
118
331
19
8
18
44
3
0
4
7
0
0
0
0
156
81
124
341
KS!
3564
2123
1339
4522
235
90
108
394
33
18
23
71
7
0
4
11
3806
2219
1457
4936
' K^Wi* * " Ta* Jgg*BliWj«!S3«»
•^Houses iSampJedjp

909
897
772
796
131
66
75
189
23
15
18
45
4
0
4
8
1005
951
849
971
2WMBI
112
71
101
229
16
8
17
39
3
0
4
7
0
0
0
0
25
22
32
67
^Totals
1021
968
873
1025
147
74
92
228
26
15
22
52
4
0
4
8
1030
973
881
1038
                                          D2-3
-------
       The failure rate for individual samples during the first site visit was at 4.3% (156/3,642)
for floors, at 3.7% (79/2,147) for window sills, and at 8.8% (118/1,346) for window trough
samples. The failure rate for individual samples showed an increase from the first site visit to the
second site visit and again from the second visit to the third site visit.
       Of the 1,025 dwelling units in Table D2-1  for which initial post-intervention clearance
sampling (first site visit) data were available, 77.7% (796/1,025) passed on the first attempt.  A
total of 11.0% (112/1,021) had at least one floor location with a dust-lead loading above the
clearance level. This failure rate was higher than that reported for window sills (7.3%, 71/968)
and lower than that reported for window troughs (11.6%, 101/873).

D2-2.  Objective 2:  Characterization of the Distribution of the Dust-lead Loadings.
       Geometric Mean Dust-lead Loadings. Variability Within a Housing Unit, and
       Variability Between Housing Units
       Preliminary assessment of the data indicated that the distributions of dust-lead loading
clearance sample results were highly skewed. Therefore, a natural logarithm transformation was
applied to the data.
       For the clearance data collected from five grantees with the lower floor dust-lead
clearance standard, Table D2-2 lists the geometric mean dust-lead loading, variance components
(within-home variability and between-home variability) associated with these loadings, and a
95% confidence interval for the geometric mean, calculated for each site visit and component
type. No sill dust sample results were available for the fourth site visit. Notice from this table
that all of the variance components are within a single order of magnitude of each other. Also,
the lengths of the confidence intervals increase from the  first site visit to the third site visit for
any given component. The increase in the length of the confidence intervals is due primarily to
the decrease in the number of samples used to estimate the variance components.
       The geometric mean lead loadings and their 95% confidence intervals were used to
compare trends between site visits for a given component type and to compare average results
between component types within a site visit. If the 95% confidence intervals on two geometric
means do not overlap, then these two geometric means are statistically significantly different at a
level less than 0.05. A comparison across site visits shows that the geometric means, from the
first site visit to the third site visit, increase gradually but were not statistically significant:
                                         D2-4
-------
 o
 o

 I
 CB
 
O  o
"I
if
II
«  o
o —
«  n
E  c
sl
O (A
CN

CN
O

                            in
                            to
     o>
     o>
                                 
          O)
                                                                   CO
                                       CO
                                       CO
          o
          CM
          CO
          CM
               CM
               CM

               IO
               CO
          O>
                                                                             in
                                                 o>
                                                 o
CO
     10
     10
     O)
     CO
     CO
     CM
LO
O
LO
CO
0)
CO
                                  IO
                                  6
     o
     O)
                                                                        CO
                                                                        IO
§
                                 O)
                                      O)
                                      10
                                           LO
                   CO
                   CO

                   CN
                   00
                   CM
                    CM
                    10
               CM
               CM

               CN
               CO
               q
               CM
                                                                        CO
                            CO

                            CM
     O>

     it
     CM
     IO

     CO
     (O
     CO

     CO
IO
o
CO
q
o>
O)
PX

O>
to
CO
CO
                                                                        CO
                                                                   o
                                                                             o
                                                                             CO
CM
o
CO
CO
0)
CO
r-
CO
     CM
     O)
     CO
     CM
                                  LO
          CM
          CM
                            
-------
 12.3 fig/ft2,13.8 ng/ft2,19.7 ug/ft2 for floors and 24.9 ug/ft2, 30.5 ug/ft2, 38.6 ng/ft2 for window
 sills.  The geometric means for window troughs increase from 66.5 ug/ft2 to 99.0 ug/ft2 from the
 first site visit to the second site visit but decrease to 97.1 ng/ft2 on the third site visit. A
 comparison of the components tested for the first site visit shows that the floor dust-lead loadings
 were significantly less than the window sill dust-lead loadings, which were in turn significantly
 less than the window trough dust-lead loadings. Note that estimates from the fourth site visit
 were associated with very small sample sizes and may not be reliable for establishing trends.
       Figures D2-1 and D2-2 contain box and whisker plots that present the distribution of
 dust-lead loadings from the first and passed clearance visits by component type. Figures D2-3 to
 D2-5 contain box and whisker plots that present the distribution of dust-lead loadings from the
 first visit for floor, window sill, and window trough, respectively; Figures D2-6 to D2-8 contain
box and whisker plots that present the distribution of dust-lead loadings from the passed
 clearance visits.

 D2-3.  Objective 3:  Characterization of the Correlation Between  Components Sampled in
       the Same Work Area
       The relationships in dust-lead loadings among floor, window sill, and window trough
wipe samples collected from within the same room are another important aspect of examining a
 clearance testing program. By estimating linear correlation coefficients, the strength of these
relationships  can be assessed. Table D2-3 displays the Pearson product-moment correlation
 coefficients and associated sample sizes for the log-transformed dust-lead loading measurements
 of individual  floor, window sill, and window trough samples taken within the same room.  For
the first site visit, dust-lead loadings for floor samples were significantly positively correlated at
the 0.01 level with loadings for both window sill and window trough samples, as was the
 correlation between window sills and window troughs.  The correlations from the second site
 visit were  again significantly positive for the relationships between floor and sill dust-lead
 loadings (at the 0.01 level) and between floor and trough dust-lead loadings (at the 0.05 level).
Data was not sufficient to estimate the correlation between window sill and window trough dust-
 lead loadings from the second site visit.
                                          D2-6
-------
             100.0
          CM
          <   10.0
          I   1JH
               0.1
                     Roots      WMowSBs    Wndw Trojghs     Bus       WMtwSfc    WHowtxtfs

                               Rnst Visit                         Passed Clearance

                                               Surface
Figure D2-1.  Comparison of In-Transformed Dust-Lead Loadings from the First Site Visit
              vs. Passed Clearance Results on an Expanded Scale.
                                            D2-7
-------
             TOO
          CM
          <  500
          3

          g 400
             300


             200


             100


               0
                  Floors
Window Sills
  Surface
Window Troughs
Figure D2-2.  Box and Whisker Plots of the Distribution of Dust-Lead Loadings from the
              Passed Clearance Data.
                                           D2-8
-------
             10000
              1000
               100
                10
Wood        Vnyl      Alumiun
               Substrate
Other
                                                                            Unknown
Figure D2-3.  Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the First Site Visit.
                                            D2-9
-------
             100000
              10000
               1000
           D




           C
          .Q
          a.
                100
                 10
                       Ai
Wood
Vnyl      Aljnjnum      Other      Unknown
Figure D2-4.  Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings

              by Substrate for the First Site Visit.
                                            D2-10
-------
             1000000
              100000
          CM   10000
                1000
          S.     100
                  10
                                 Wood
Vinyl
Aluminum
Other
Unknown
Figure D2-5.  Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the First Site Visit.
                                           D2-11
-------
             1000
          CM   100

          o>

          1
          Q.   10
                    M
Wbod        Vnyl       Aluminum      Other

                Substrate
Unknown
Figure D2-6.  Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the Passed Clearance Visits.
                                           D2-12
-------
             1000
          CM   100
              10
                    Al
Wood
Vinyl
Aluminum
Other
Unknown
Figure D2-7.  Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
              by Substrate for the Passed Clearance Visits.
                                          D2-13
-------
             1000
          CM  100
               10
                     Al
Wood       Vhyl       Aluminum      Other
               Substrate
Unknown
Figure D2-8.  Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the Passed Clearance Visits.
                                           D2-14
-------
Table D2-3.   Observed Within-Room Correlation Coefficients Between Dust-Lead Loading
              Measurements Collected From Floors, Window Sills and Window Troughs
              from the Grantees with Lower Floor Dust-lead Clearance Standard in the HUD
              Grantee Program.
                733S;vfl*SLBlft;

     First
0.48'
1687
0.40'
809
0.59a
20
   Second
0.20"
 21
0.53"
 16
       ' Statistically significantly different from zero at the 0.01 level.
       " Statistically significantly different from zero at the 0.05 level.
The conditional probabilities of a sample passing or failing a standard are given in Table D2-4.
These analyses were conducted on two different sets of data, the first set using all possible paired
observations from within the same room. The second subset of data restricted the
analyses to rooms in which floor, window sill, and window trough lead loadings were
simultaneously observed.
       As shown in Table D2-4, results from contingency table estimates and normal theory
estimates are roughly consistent.  For both types of estimates and both data sets included in the
analysis, the probability that one component passes clearance given that another component
passes clearance ranges from 80% to 97%. The probability that a floor sample is less than
200 ug/ft2, given the samples from the window sills are less than 500 ng/ft2, is the highest at
97.4%.
       For both types of estimates and both data sets included in the analysis, the probability that
one component fails clearance testing given that another component in the same room fails
clearance testing ranged from 13% to 67%.

D2-4.  Objective 4: Demonstration of the Impact of Composite Sampling on Pass/Fail
       Rates of Houses

       Table D2-5 provides, for each component type, the number of residential units having a
given number of samples collected from a given component type. Across component types, most
of the residential units had fewer than four clearance samples collected. Data for residential units
having more than four samples from a given component type were used to construct multiple
                                        D2-15
-------
~
i
i
IS
 o
z
•o
 (0
 0
S
 o
 §
 o>
•*3
 O
0
 X
CM
 O)
"55
CO
IU
O)

I
8
c s
S .2
IS
o S
CA ^
e IE
1 a)
o
O
Q
JB
JO

                                         D2-16
-------
Table D2-5.   Number of Residential Units that Contained (N) Individual Samples of Each
              Component Type Based on First Site Visit Clearance Testing Data from the
              Grantees with Lower Floor Dust-lead Clearance Standard in the HUD Grantee
              Program.
  ^IridividuaHSampIes?
  "•-1fft1Pjf-*£&-Jt>~,*Ui!:'*fl-'~
                                             •v--i^T--ii»:»w»T<"-^:Tri^' •- ^i^*5 I^Vi'llKf**-.":'™
                                             |tornpj^n^T^ej;$r~
                              129
                       120
                      495
                              124
                       587
                      302
                              159
                      214
                      60
                              346
                       40
                       13
                              201
          6
 50
          8
        9  +
        Total
1021
968
873
   Total with N
263
simulated composite samples. When there were four or fewer individual samples from a
component type within a housing unit, the simulated composite sample included all samples.
From Table D2-5, the number of homes with five or more samples was 263 for floors, 7 homes
for window sills, and 3 homes for window troughs.  Therefore, data for approximately 26%
(263/1,021), 1% (7/968), and less than 1% (3/873) of the residential units were used when
constructing multiple simulated composite samples from floors, window sills, and window
troughs, respectively.
       For each component type within a residential unit, the set of individual clearance sample
lead loading results for the first site visit were used to construct simulated composite samples for
the purpose of evaluating the three composite sample clearance criteria introduced in Section
5.4.2 (Standard, Standard/n, and 2*Standard/n). The construction of simulated composite
samples is discussed in Section 5.4.1. For each combination of component type and composite
clearance criterion, each residential unit either passed clearance,  failed clearance, or yielded
inconclusive results, according to whether the sets of simulated composite samples for the unit
                                         D2-17
-------
 either all resulted in a pass decision, all resulted in a fail decision, or had some combination of
 pass and fail decisions, respectively.
Table D2-6.   Numbers of Residential Units that Pass or Fail Clearance, Based on Individual
              Sample Clearance Results vs. Simulated Composite Clearance Results, Using
              Data from Grantees with Lower Floor Dust-lead Clearance Standard.
                                            lndividual;Sampl9iClearance;Resu
                Pass
    Standard
Inconclusive
                Fail
                    909
         39
897
34
         18
                             55
                            34
772
 20
                             81
                Pass
   Standard/n
Inconclusive
                Fail
                    858
                  883
14
                    37
        112
 14
70
                   757
 14
101
                Pass
  2xStandard/n
Inconclusive
                Fail
                   896
                  892
         13
                             92
                            65
                   762
                    10
                              0
                   101
       Values of the four performance characteristics defined in Section 5.4.3 (sensitivity,
specificity, positive predictive value, negative predictive value) are presented in Table D2-7 for
each combination of component type and composite clearance criterion.  By design, the
sensitivity for the Standard/n Rule is always 1.00 (as all sets of simulated composite samples
would fail if at least one individual sample result failed) while the specificity for this rule is
estimated at 0.94 for floors and 0.98 for both window sills and window troughs. In contrast, the
specificity of the Standard Rule is always 1.00 (as all sets of simulated composite samples would
pass if all individual samples passed) while the sensitivity for this rule is estimated at 0.49 for
both floors and window sills and 0.80 for window troughs. The 2 * Standard/n Rule attempts to
maximize both sensitivity and specificity.  For the 2* Standard/n Rule, the values of sensitivity
are higher than those calculated for the Standard Rule, while the values of specificity are higher
than those calculated for the Standard/n Rule. Estimates of sensitivity and specificity in these
                                         D2-18
-------
•examples are always conservative, because inconclusive composite test results factor into the
denominator for each estimate, but never factor into the numerator.
 Table D2-7.  Performance Characteristics of Composite Clearance Criteria Based on Data
              from Grantees with Lower Floor Dust-lead Clearance Standard.
                   Sensitivity
       Floors
                   Specificity
                   PPV
                   NPV
0.49
1.00
1.00
0.96
1.00
0.94
0.75
1.00
0.82
0.99
0.91
0.99
                   Sensitivity
    Window Sills
                   Specificity
                   PPV
                   NPV
0.49
1.00
1.00
0.96
1.00
0.98
0.83
1.00
0.93
0.99
0.93
1.00
                   Sensitivity
  Window Troughs
                   Specificity
                   PPV
                   NPV
0.80
1.00
1.00
0.97
1.00
0.98
0.88
1.00
1.00
0.99
0.91
1.00
       The above performance characteristics estimates illustrate that the three composite
clearance testing criteria have different specificity and sensitivity rates. These rates correspond
to the consistency between clearance decisions and the true lead hazards present in the various
locations sampled (assuming the individual sample lead-loading results are representative of
these lead hazards). To further characterize the performance of each of the three composite
clearance criteria, the following logistic regression model was fitted to clearance data for each
combination of component type and composite clearance criterion:
                                         D2-19
-------
where n^ is the estimated probability of clearance for component(/) in house(f) under composite
criterion (k), and Maxjy is the maximum individual sample lead loading result in house(i) for
component(/).
       For each component type, the estimated relationship between the probability of passing
clearance and the maximum lead-loading is presented graphically in Figure D2-9. In this figure,
the solid, long-dashed, and finely-dashed lines represent the estimated relationship for the
Standard,  Standard/n, and 2* Standard/n Rules, respectively.
       For each combination of component type and composite clearance criterion, Table D2-8
provides parameter estimates and associated standard errors from fitting the above logistic
regression model to data. Table D2-8 also presents estimates of the probability of passing
clearance (using composite samples) when the maximum lead loading among all locations
included in the composite sampling scheme is greater than or equal to  Vi, 1,2 and 4 times the
associated HUD interim standard for individual samples. Results from these logistic regression
model fits show that there is a better than 55% chance that a residential unit will pass clearance
for floors and window sills under the Standard Rule for composite sampling, when there is an
individual sample location which has a lead-loading level that is equal to twice the HUD
clearance standard. Estimates for the Standard/n Rule demonstrate that this rule's high sensitivity
(probability of passing is below 19% when the maximum individual sample lead loading is
greater than or equal to the HUD Standard) along with a loss in specificity (probability of passing
is 85% for floors, 96% for window sills, and 99% for window trough when the maximum
individual sample lead loading is equal to V* HUD Standard).  Once again, the 2xStandard/n
Rule is shown to be a  compromise between the Standard and Standard/n Rules. At 1A the HUD
Standard, the estimated probability of passing clearance testing under the 2xStandard/n Rule is
one or nearly one, while at 2>
-------
                          ProbobBV of Passha Clearance Ten**) LMng Composte Floor Samples
                              10O
                                    20O    300    400    600     60D
                                    Maxknum kvMdual Sample ROOT Pb Loading
                                                                    TOO
                                                                          BOO
                                                U*Q CQRfXMte Widow SB Samptee
                              250
                                    500    750   1000    1250    1500
                                  Maxknum MMdue) Swnpto WMow sn Pb Loadk«
                                                                   1750
                                                                         aooo
                                     CtBoronov ^stkio UvhQ CornpovftB Wtidow VMsM
                                                               Approach
                                                              men) I
                                                                     / n) R
                                                                     Ruto
                              400    flOO    lann   TGDO   2000    24OO
                                  MBxknum MMcLiel Sample WMow WM Pb
                                                                   2800
                                                                         3200
Figure D2-9.  Estimated Relationship Between the Probability of a Residential Unit Passing
               Clearance Testing versus the Maximum Individual Lead-Loading Results by
               Component Type Based on Simulated Composite Samples from the Grantees
               with Lower Floor Dust-lead Clearance Standard in the HUD Grantee Program.
                                             D2-21
-------

 = j? 5>
 w O OT
 « ^ o
* £ «
 0.3= J>
Z Q O

 c g-°
 o £ g
•-c f. £
£ »- *•
 ® i- 2
"* §
£ o as

 «««
2 a S
 g E 2
_ (0 c
 s > *
-ll E
 8-2
 2 e «^
••s « ro
.2 g c
      '
 c e .S  o>
 I s«l

I III

tt o Q a
CO

CNJ
                                  §
                                  o
                                  o
                                  00
                                  
                                  (O
                                  CO
    o
    o
    CD
                                         0)

                                         d
        (D

        CO
                00

                d
                                                     00
            CN

            8
            oi
            03
            «™
            en
CO
CO
en
cd
CM
o>
CD
                                         •<*
                                         in
                                         CO
                                         Ps
    in
    IX
            a>
    CO
    in
    cd
    o>
    in
    in
                       o
                       o
                                         CO
                                                 CO
                                                             in
                                                     CO
                             CO
                                 CO
                                 o
                                 o>
    oo
    in
    oo
    ro
    0)
                                             (D
                                             en
                                             03
    to
    CO
    CO
    CN
    CO
    r-
                       in
                                  ID
                                 T3
                                      re
                                      •a

                                      I
                                      CO
    •s
    CD
        (0
                                         (A
            nj
            T)
            X
            CM
        •o

        CO
                    CO
                    X
                   CN
            ^
            •S
            co
            •a

            I
            c/)
          CA


          O
          •a
                                            D2-22
                                                       o
                                                       •o
-------
  APPENDIX E

  Atlantic City
Housing Authority
       E-1
-------
                                     APPENDIX E
                                   ATLANTIC CITY
                               HOUSING AUTHORITY
       The Atlantic City Housing Authority provided data from a comprehensive rehabilitation
project performed on public housing buildings containing 10 to 23 multi-family housing units
[18]. These two- to three-story brick buildings had lead-based paint on the doors, windows,
radiators, trim, and stairwells. The doors and windows, along with most of the trim, were
removed and replaced during abatement. Walls were enclosed by drywall, then painted, and steel
panels were placed on the walls in the stairwells. Atlantic City Housing Authority employed a
two-phased lead clearance process: worker entry clearance and re-occupancy clearance.  The first
set of lead clearance samples was collected after protected workers had finished the abatement
job (worker lead clearance testing).  The second set of lead clearance samples was taken after the
renovation job was completed, but before the unit was occupied (re-occupancy clearance).
Worker lead clearance testing results were used to ensure that the housing unit was safe for
renovation workers (e.g. carpenters) to enter to complete their work. Re-occupancy clearance
testing results were collected after all work on the housing unit was completed (including
abatement and renovation and remodeling) to ensure that the unit was not contaminated by
leaded dust and could be re-occupied by residents. This report only analyzed data collected in
the re-occupancy clearance testing phase from the Atlantic City Housing Authority project.
Wipe samples were collected in accordance with HUD protocol. Within each completed unit,
one dust sample was collected from each room or area where abatement occurred. Sample
locations were randomly distributed between floors and window troughs. Since the abatement
process included removal of many windows, few window sill samples were tested. The standard
interim HUD-established thresholds  of  200 ug/ft2 for floors, 500 jig/ft2 for window sills, and
800 ug/ft2 for window troughs were utilized in determining whether dust wipe samples passed or
failed. In all cases where results exceeded the clearance thresholds, the areas were re-cleaned and
re-tested until acceptable results were obtained.
                                          E-2
-------
E-1.    Objective1: Characterization of the Number of Individual Samples. Work Areas,
        and Housing Units That Pass or Fail Clearance Testing Standards
        Individual dust wipe samples were collected in Atlantic City from June 1994 through
May 1995 as part of their clearance testing program. In all, 923 individual dust wipe samples
were collected from floors, window sills, and window troughs within 779 rooms in 160
residential units. Table E-1 presents the number of individual samples, work areas, and
Table E-1.   Clearance Testing Results by Individual Sample, Room, and Residential Unit for
            Atlantic City.
                                       s&w
                                                         -
                                                      Sam
                                              feTotal
    First
              Floor
               Sill
             Trough
               All
            516
       38
554
479
37
516
127
            46
              51
        46
               51
           252
             258
       248
               254
               114
           814
       49
863
714
48
762
121
32
39
159
                             20
                      119
160
  Second
              Floor
Trough
               All
            29
       13
42
 25
12
37
12
14
 17
 14
        17
            43
       16
 59
31
15
46
13
       16
                      11
       20
   Third
             Trough
               All
Total
Floor
Sill
Trough
All
545
46
267
858
51
5
9
65
596
51
276
923
502
46
260
733
37
5
6
46
539
51
266
779
136
16
118
133
23
4
4
27
159
20
122
160
residential units that passed or failed clearance testing within each combination of component
type and site visit. Approximately 93% (863/923) of the dust samples were collected during the
first site visit to a residential unit. While 94% (814/863) of the dust samples fell below the
clearance standards of 200 ng/ft2 for floors, 500 ug/ft2 for window sills, and 800 n.g/ft2 for
window troughs, and 94% (714/762) of the rooms met clearance standards, only 76% (121/160)
of the residential units passed clearance on the first site visit.  This increase in the failure rate
from the percentage of individual samples and rooms that fail to the percentage of residential
units that fail is attributable to the fact that all individual samples must pass for a unit to pass.
                                           E-3
-------
That is, if any individual sample exceeds the standard and fails clearance, then the entire
residential unit also fails clearance. Of the 39 residential units that failed clearance on the first
site visit, 20 (51%) of these residential units are known to have been revisited for a second
clearance testing site visit. Eventually, 133 of the 160 residential units (83%) are known to have
passed clearance testing.
       Floor samples accounted for 64% (554/863) of the first site visit samples while window
troughs made up 30% (258/863) of the samples and window sills comprised the final 6%
(51/863) of the first site visit samples. The failure rate for individual samples during the first site
visit was highest for window sill samples (10% (5/51)) followed by floors (7% (38/554)) and
window troughs (2% (6/258)).

E-2.   Objective 2: Characterization of the Distribution of the Dust-Lead Loadings.
       Geometric Mean Dust-Lead Loadings. Variability Within a Housing Unit, and
       Variability Between Housing Units
       As seen in Table E-2, the geometric means for floor and window trough samples were
18.5 ug/ft2  and 19.4 ug/ft2, respectively.  The within-house variance components and the
between-house variance components are similar to those estimated for the other data sources.
       Figures E-l  and E-2 contain box and whisker plots that present the distribution of dust-
lead loadings from the first and passed clearance visits by component type. Figures E-3 to E-5
contain box and whisker plots that present the distribution of dust-lead loadings from the first
visit for floor, window sill, and window trough, respectively; Figures E-6 to E-8 contain box and
whisker plots that present the distribution of dust-lead loadings from the passed clearance visits.

E-3.   Objective 3: Characterization of the Correlation Between Components Sampled in
       the Same Work Area
       The relationships among floor, window sill, and window trough wipe samples are
another important aspect of examining a  clearance testing program. One method to assess the
relationships among individual floor, window sill, and window trough wipe samples collected
from the same room is to estimate linear  correlation coefficients. Table E-3 displays the Pearson
product-moment correlation coefficients  and the associated sample size  for the log lead loading
                                          E-4
-------
 c
 o

 o
 Q.
 E
 o
o
•a

 a
**
 w
2
35

JO

o
**
0



UJ


0

o
a


I


i
ja

a
 c  **
 m  V
 O  W
 oO
8*

O

Ul

 O

2
03

ID
                               O
                               o
U)

If)
               01
               oo
                                              cq
                                              ai
     o>
     CM
           ao
           CM
CM
          00
          C3
                                    CO
                                    CO
          CO
          03
          CM
          O
                               tf)

                               CO
                                              CM
                                                   CO
                                                   ID
                              O3
                              Ul
                              CO

                              in
          10

          CM
     ro
                              u>
                              10
     It)
00
in
cs
               o
               o
                a

                §
                  o
                  o
                  o
                 CO
               •g

               £
                                                      E-5
-------
             100.0
          <   10.0
           ?
          D
          &
               1.0
               0.1
                     Ftas
WndwSfc
 RtstVtsft
Whto* Troupe
Rons      Window Sfe    Wndw loupes
       Passed Qearance
Figure E-1.  Comparison of In-Transformed Dust-Lead Loadings from the First Site Visit vs.
             Passed Clearance Results on an Expanded Scale.
                                             E-6
-------
800
700
600

N
<. 500
3
0400
1
^300
CL
200
100
0








I





















i i i
Boors Window Sills Window Troughs
Surface
Figure E-2.  Box and Whisker Plots of the Distribution of Dust-Lead Loadings from the
            Passed Clearance Data.
                                          E-7
-------
              10000
              1000
               100
                10
                      All
Wood
Vryl
Aluminum
Other
Unknown
Figure E-3.  Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
             Substrate for the First Site Visit.
                                              E-8
-------
             100000
              10000
           CM
           <
               1000
           0)

           1

           e
100
                 10
                       All
                  Wood        Vinyl      Aluminum      Other       Unknown

                                 Substrate
Figure E-4.  Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
             by Substrate for the First Site Visit.
                                               E-9
-------
              1000000


              100000


          w   10000

          3
          I    1000


          £     100


                  10
                        All
Wood
Vinyl
ALrmim
Other
Unknown
Figure E-5.  Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
             Loadings by Substrate for the First Site Visit.
                                            E-10
-------
             1000
          CM  100
               10
                     Al
Wood       Vhyl       Aluntun
               Substrate
Otter
Unknown
Figure E-6. Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
            Substrate for the Passed Clearance Visits.
                                            E-11
-------
             1000
          N   100
          c
          •o
              10
                              Wood        Vinyl       Aluminum      Other       Unknown

                                             Substrate
Figure E-7.  Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings

            by Substrate for the Passed Clearance Visits.
                                           E-12
-------
             1000
          CM  100
              10
                    All
Wood        Vinyl      Akmnurn       Other       Unknown
                Substrate
Figure E-8.  Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
            Loadings by Substrate for the Passed Clearance Visits.
                                           E-13
-------
Table E-3.   Observed Within-Room Correlation Coefficients between Pb Loading
             Measurements Collected from Floors, Window Sills and Troughs for the
             Atlantic City Data.

              First
            Second
                         'sw)fStsSS<^^se^fni'>t>^itafa9lS>»t9a)ia/KS»*---t^ ••' ".-.-v.c.v.f;. ---.;,•..;.• a*<**K!yitfaiQ:iini*iiivv*iiHttt
0.32
26
0.65"
                              0.10
33
       8 Statistically significantly different from zero at the 0.01 level.
measurements of individual floor, window sill, and window trough samples taken within the
same room. The correlation between window sill samples and window trough samples was not
able to be estimated due to insufficient data. In fact, for this data set, there were few cases where
multiple samples were taken per room. The data from the first site visit show that correlation
between floors and troughs is positive and significant at the 0.01 level. The correlation between
floors and sills was estimated to be positive but was not statistically significantly different from
zero. The correlation from the second site visit is difficult to interpret because of the small
number of samples from which it was estimated.
       The conditional probabilities of a sample passing or failing a standard are given in
Table E-4. These analyses were intended to conduct on two different sets of data, the first set
using all possible paired observations from within the same room. The second subset of data
restricted the analyses to rooms in which floor, window sill, and window trough lead loadings
were simultaneously observed.  However, no such results were observed in the Atlantic City data
for the second subset of data. Window sill and window trough samples were not simultaneously
present in rooms in this data set.
       As shown in Table E-4, results from contingency table estimates and normal theory
estimates are consistent for the left side of the table, but less so for the right side. For both types
of estimates, the probability that one component passes clearance given that another component
passes clearance is quite high (over 88%).  The probability that one component fails clearance
testing given that another component in the same room fails clearance testing ranged from 0% to
51%.
                                           E-14
-------
JQ
 10
JQ
I
•o
 CO
 W
 o
 O>
V
 o
o
CM
 X
eg

f
 (0

T3
 0
+*
 a

I
ILI
 O>
 V

 O
 O
 c *r
 2 .2

II
O JC
O I-
t
UJ
Q)

           ..I
           f

           .
§
      8
     o
     CO
     V
     >-_
     8
                         o
                         o
                         rg
                         V
                         -
                         a.
                            E-15
-------
E-4.   Objective 4: Demonstration of the Impact of Composite Sampling on Pass/Fail Rates
       of Houses
        Simulated composite sample results were constructed from individual sample lead
loading results from each component type within a residential unit. These simulated composite
samples were then used to evaluate the three composite sample clearance criteria (Standard,
Standard/n, and 2xStandard/n).
       Table E-5 displays the number of residential units that were tested by the number of
individual samples that were collected and each component type. For example, there were 57
residential units for which two dust-wipe samples were collected from floors on the first site
visit.  Residential units containing more than four samples from a particular component type
resulted in the estimation of multiple simulated composite sample results in this analysis.
Therefore, summing all the units that had five or more individual samples, approximately 25%
(40/159), 5% (1/20), and 1% (1/119) of the residential units resulted in the estimation of
multiple simulated composite samples from floors, window sills, and window troughs,
respectively. When there were four or fewer individual samples from a component type within a
housing unit, the  simulated composite sample included all samples.
       For each component type within a residential unit, the set of individual clearance sample
lead loading results were used to construct simulated composite samples for the purpose of
evaluating the three composite sample clearance criteria For each combination of component
type and composite clearance criterion, Table E-6 gives the number of residential units that either
passed clearance, failed clearance, or yielded inconclusive results based on the composite
clearance samples.  Inconclusive results could only occur in a residential unit if more than four
individual samples were collected from that unit.  They occur due to the many possible ways of
combining five (or more) individual samples into composite samples.  For floor samples, the
percentages of residential units with greater than four individual clearance samples that actually
resulted in inconclusive results were 15% (6/40),  7.5% (3/40), and 17.5% (7/40) for the Standard,
Standard/n, and 2*Standard/n Rules, respectively. No window sill or window trough composite
sample yielded inconclusive results for any of the three criteria, and, of course, for each of these .
components, there was only on unit with more than four individual clearance samples.
                                         E-16
-------
Table E-5.   Number of Residential Units that Contained (N) Individual Samples of Each
            Component Type Based on First Site Visit Clearance Testing Data From the
            Atlantic City Housing Authority.
                      9 +
                     Total
                 Total with Ma 5
                                     57
                                     33
                                     26
                                     23
                                     10
                   159
                   40
     20
                                                            ^Window
                                                            •- .....
                                                                ughs
                                                               14
                                              79
                                              19
        119
Table E-6.  Individual Sample Clearance Results Versus Simulated Composite Clearance
           Results Based on Data from the Atlantic City Housing Authority.
                  Composites^.
                             ^^?¥^&^^s^*^SR«as
                             Individual Sample ^Clearancif
                                                                      esu
                  Pass
     Standard
Inconclusive
                  Fail
                    127
16
16
                            10
114
                  Pass
    Standard/n
Inconclusive
                  Fail
                   119
         15
                            32
                 109
                  Pass
   2 x Standard/n
Inconclusive

Fail
                   126
         15
                                              22
                 111
                                         E-17
-------
       The performance characteristics of each combination of component type and composite
clearance criteria are presented in Table E-7 in terms of sensitivity, specificity, positive
Table E-7.  Performance Characteristics of Composite Clearance Criteria Based on Data
            from the Atlantic City Housing Authority.

       Floors
                    Sensitivity
                    Specificity
                    PPV
                   NPV
0.31
1.00
1.00
0.89
1.00
0.94
0.87
1.00
0.69
0.99
1.00
0.97
      Window
        Sills
                    Sensitivity
                   Specificity
                   PPV
                   NPV
0.25
1.00
1.00
0.84
1.00
0.94
0.80
1.00
0.50
0.94
0.67
0.88
      Window
      Troughs
                    Sensitivity
                   Specificity
                   PPV
                   NPV
0.80
1.00
1.00
0.99
1.00
0.96
0.50
1.00
1.00
0.97
0.63
1.00
predictive value (PPV), and negative predictive value (NPV). By design, the specificity of the
Standard Rule is always 1.00 and the sensitivity of the Standard/n Rule is 1.00. So, in some
sense, the Standard Rule sacrifices sensitivity for specificity while the Standard/n Rule sacrifices
specificity for sensitivity.  The 2 * Standard/n Rule attempts to find a balance between these two
criteria. For this data, the Standard Rule's sensitivity is estimated to be 0.31 for floors, 0.25 for
window sills, and 0.80 for window troughs. The 2*Standard/n Rule seems to perform much
better in terms of sensitivity with estimates of 0.69, 0.50, and 1.00 for floors, window sills
and window troughs, respectively.  The estimates of sensitivities for window sills and window
troughs, however, are based on very few units which failed individual sample clearance testing.
The 2*Standard/n Rule seems to perform quite well in terms of specificity with estimates of 0.99
for floors, 0.94 for window sills, and 0.97 for window troughs. When comparing the overall
performance of the three criteria in this case, the Standard/n Rule seems to perform quite well
both in terms of sensitivity and specificity.
                                           E-18
-------
       It is evident that all three composite clearance testing criteria have different sensitivities
and specificities associated with their application to the simulated composite samples. In order
to further investigate the performance of each of the three composite clearance criteria, the
following logistic regression model was fitted for each combination of component type and
composite clearance criterion to describe the relationship between the probability of passing
clearance based on composite samples and the maximum lead loading present in the individual
samples collected:
where nijk is the estimated probability of clearance for component(/) in house(i) under composite
criterion (&), and Max,-,- is the maximum individual sample lead loading result in house(i') for
component(/).
       For each component type, the estimated relationship between the probability of passing
clearance and the maximum lead loading is presented in Figure E-9. Note that the relationship
could not be estimated for window troughs using the Standard Rule due to the nature of the data.
Under this rule, the observed data were such that at a given level the unit passed, and at a given
level the sample failed. Estimation problems arise when trying to estimate the relationship
between these two levels. An infinite number of curves could be fit to estimate the relationship
between the two levels, but there is no criterion to choose the optimal one. Therefore, the
relationship cannot be estimated for this rule.
       Table £-8 provides parameter estimates and associated standard errors from the logistic
regression models, as well as estimates of the probability of passing clearance (using composite
samples) when the maximum lead loading among all locations included in the composite
sampling scheme is greater than or equal to 1A, 1 , 2, and 4 times the associated interim HUD
standard for individual samples. The low sensitivity of the Standard Rule is illustrated by the
high (73%) estimated probability of a unit passing clearance testing based on floor composite
samples and a maximum individual sample lead-loading of 2*HUD standard.  Conversely, the
estimated probability of passing based on floor composite samples using the 2>
-------
to probabilities of 0.98 and 1.00 for the other two decision rules. The estimates for window
troughs at one-half the HUD standard are similar to those for floors, but the estimates of the
probabilities of passing with a maximum individual sample lead loading at the HUD standard are
much lower for the 2>
-------
                           Probability of Passing Clearance Testing Using Composite Floor Samples
                                                          	(2 x Standard /n)RJo
                                                          	(Standard /n)Rule
                               100
 200    300    400    500    600

Maximum Individual Sample Floor Pb Loading
                                                                                 800
                         Probability of Passing Clearance Testing Using Composite Window SiD Samples
                                                                    Approach

                                                          	StanttordRuto
                                                          	(2 x Standard/n)Riilo
                                                          	(Standard/n) Rule
                               250    500    750    1000    1250    1500   1750   2000

                                    Maximum Individual Sample Window SRI Pb Loading
                         ProbabHHy of Passing Clearance Testing Using Composite Window WeO Samples
                                                                    Approach

                                                          	Standard Rule
                                                          	(2 x Standard/n}Rute
                                                          	(Standard/n) Rote
                               400
                                      BOO
        1200    1600    2000    2400    2BOO    3200
                                    Maximum Individual Sample Window Well Pb Loading
Figure E-9.  Estimated Relationship Between the Probability of a Residential Unit Passing
              Clearance Testing versus the Maximum Individual Lead-Loading Result by
              Component Type Based on Simulated Composite Samples from the Atlantic
              City Housing Authority.
                                                 E-21
-------
>*• O)
 o c

II
_Q (0
 « w
•8 »
 s «
a g

i!
 3 c
 >
 «
 C

 o c
* .£
o  a> O
IE "a o
3  £f
—  * i
•SfS
O  (D <
•-g E
S - 2
2  E *
a> E Q
o  3 «S
oc  E s
o '5 Q

111
     .

••s S
i  S.S


i;?
CO  C o
**  ffi »™
3  ™ .2

S — 9
EC O Q

 •
CO
111
a
      O
      O

      6
    o
    o
   o
   o
               o
               o
                                            o
                                            o
   o
   6
      o
      o
          o
          o
          6
            8
            0
                         o
                         q
                         d
                       oo
                       o>
      in
      q
      d
   in
   a>
      in


      d
                      (D
                      ^*

                      d
                   q
                   d
                       8
   00
   a>
oi
OD
      (O
      03
                      O)

                      d
                         0>
                         CO
                       S
                       o
      q
      d
         01
         o
         o
      n
      o
      o
      PS.
      o
      o
                      S
                      q
                      d
                                               o
                                               q
                                               d
                       q

                       9
   00
   PJ
   o
o
u>
o
CO
o
o
            (0
            o
            o
                      q
                      d
                                               (O

                                               q

                                               d

                         §
                         CO
                          8
                          EE
            =
            to




            I

            5
               It
                                  E-22
-------
             APPENDIX F
Cleveland Lead Hazard Abatement Center
                 F-1
-------
                                    APPENDIX F
                 CLEVELAND LEAD HAZARD ABATEMENT CENTER

       The lead abatement program managed by the Cleveland Lead Hazard Abatement Center
recruited houses where children with elevated blood lead levels lived [19]. Most of the units
were large single-family houses built before 1950.  The level of intervention depended on the age
and number of children in the house. The homes had a mix of interim controls and abatement
treatments. Chipped paint was scraped and repainted with a primer and a coat of paint. Carpets
were removed and the floors were either repaired, covered with plywood, or refinished.
Windows were replaced in some units. Vinyl siding was applied to homes with lead-based
painted wood siding. Porches were repaired, deteriorated sections replaced, and other surfaces
scraped and repainted. Lead contaminated soil was generally covered with sod, wood chips, or
some other form of landscape cover; if lead contamination was extremely high, soil was removed
and replaced.  This project is ongoing. Results for window trough, window sill, and floor wipe
samples collected from each room in the unit were analyzed.

F-1.   Objective 1: Characterization of the Number of Individual Samples. Work Areas, and
       Housing Units That Pass or Fail Clearance Testing Standards
       As part of clearance testing within the Cleveland Lead Hazard Abatement Center's
intervention program, 312 individual dust wipe samples were collected from floors, window sills,
and window troughs within  196 rooms in 38 residential units from December 1993 through May
1995. Table F-1 presents the number of individual  samples, work areas, and residential units that
passed or failed  clearance testing within each combination of component type and site visit.
Approximately 93% (290/312) of the dust samples were collected during the first site visit to a
residential unit.  Although 92% (267/290) of the dust samples fell below the clearance standards
of 200 ng/fi2 for floors, 500 ug/ft2 for window sills, and 800 ug/ft2 for window troughs, and 88%
(173/196) of the rooms passed clearance on the first site visit, only 61% (23/38) of the residential
units passed clearance on the first site visit. This increase in the failure rate from the percentage
of individual samples and rooms that fail to the percentage of residential units that fail is
                                         F-2
-------
 Table F-1.  Clearance Testing Results by Individual Sample, Room, and Residential Unit for
             the Cleveland Lead Hazard Abatement Center.
                               XBffi
    First
               Floor
                Sill
              Trough
                All
138
20
158
137
20
157
24
 92
       93
        92
               93
               32
 37
       39
        37
               39
               22
267
23
290
173
23
196
23
14
15
38
                      33
                      24
38
   Second
               Floor
                Sill
              Trough
                All
 15
       18
        15
               18
               10
 18
       21
        18
               21
               11
                      13
                      14
    Third
               Floor
                All
Total
Floor
Sill
Trough
All
154
93
39
286
23
1
2
26
177
94
41
312
153
93
39
192
4
0
0
4
157
93
39
196
35
33
24
35
3
0
0
3
38
33
24
38
attributable to the fact that if any individual sample exceeds the standard and fails clearance, then
the residential unit also fails clearance. Of the 15 residential units that failed clearance on the
first site visit, 14 (93%) of these residential units were revisited for a second clearance testing.
Seventy-nine percent (11/14) of the units passed clearance on the second site visit. Of the three
units that did not pass clearance on the second visit, one passed clearance after a third visit.
Thus, at the time the data for this report were collected, 35 of the 38 residential units (92%) had
passed clearance testing.
        The failure rate for individual samples during the first site visit was higher for floor
samples (13% (20/158)) than that for window troughs (5% (2/39)) or window sills (1% (1/93)).
This pattern is reflected in the failure rates for residential units based on results from individual
components: 37% (14/38) of the residential units failed based on the results of floor samples, 8%
                                            F-3
-------
(2/24) of the residential units failed based on the results of window trough samples, and 3%
(1/33) of the residential units failed based on the results of window sill samples.

F-2.   Objective 2: Characterization of the Distribution of the Dust-Lead Loadings.
       Geometric Mean Dust-Lead Loadings, Variability Within a Housing Unit, and
       Variability Between Housing Units
       The Cleveland program had the fewest number of residential units tested for clearance.
The number of samples collected during the second site visit was very small. The geometric
mean for the floor clearance samples was less than the window sills which was less than the
window troughs for the first site visit. However, they were not statistically different from one
another.  For the first site visit, the estimated within components of variances are greater than the
between components for floors and window sills; the opposite is true for window troughs.
       Figures F-l and F-2 contain box and whisker plots that present the distribution of dust-
lead loadings from the first and passed clearance visits by component type. Figures F-3 to F-5
contain box and whisker plots that present the distribution of dust-lead loadings from the first
visit for floor, window sill, and window trough, respectively; Figures F-6 to F-8 contain box and
whisker plots that present the distribution of dust-lead loadings from the passed clearance visits.

F-3.   Objective  3: Characterization of the Correlation Between  Components Sampled  in
      the Same  Work Area
      The relationships among floor, window sill, and window trough wipe samples are another
important aspect of examining a clearance testing program.  One method of assessing the
relationships among individual floor, window sill, and window trough wipe samples collected
from the same room is to estimate linear correlation coefficients. Table F-3 displays the Pearson
product-moment correlation coefficients and the associated sample size for the log lead loading
measurements of individual floor, window sill, and window trough samples taken within the
same room. The correlation between window sills and window troughs is the highest of any
observed correlation but it is not statistically significant. None of the observed correlations are
statistically significantly different from  zero at the 0.05 level.
                                          F-4
-------
 o
 a
 c
 u>

 o
 Q.
 E
 o
u
 «
 *<
w
 S
UJ
 o

 o
 Q.

 O
U

 8
 (0

•o
 g

 i
 O  0
 Ew

 O  (0
 0  +*
 oi
H  >

 SO

 i£
 o **
 O  k.
rsl
uL
          .••jJO' C'-vg*
          -      '
                   til
                   •pit
           -•->'-'i;r!*'Vi..5'ls"«
                            CO

                            0>
                            CO
                                <0

                                CO
                                in
                                    en
         oq
         ifi
         
                            in
                                CO
                                    o
                                    o
         CO
         CO
CO
o
                            to
                                CO
    in
    *t
                                        O)
                                        CM
                                                 <0

                                                 o
                            0)
                            CM
                                CD
                                CO
         O)
         01
                                                 CO
                           00
                           co
05
m
                                    
    en
    co
                                        eo
                                                CM
                                                      O
                                                      o
                                            TJ

                                             O
                                             U
                                                 F-5
-------
              100.0
          sr
          <   10.0
          I   to
               0.1
                     Hans       WndowSfe   Window Troughs


                                Fust Visit
Rows       Window Sfc    Window Houghs


       Passed Clearance
                                                 Surface
Figure F-1.   Comparison of In-Transformed Dust-Lead Loadings from the First Site Visit

              vs. Passed Clearance Results  on an Expanded Scale.
                                              F-6

-------
800
700

600
N
^ 500
3
"400
1
vu
^300
0.
200
100
0











i

















i









                   ROOTS
Window Sills
  Surface
Window Troughs
Figure F-2.   Box and Whisker Plots of the Distribution of Dust-Lead Loadings from the
              Passed Clearance Data.
                                            F-7
-------
              10000
               1000
           CM
           <
               100
          £1
          D.
                10
                       AS         Wood
Vty       Aluminum       Other        Unknown

    Substrate
Figure F-3.    Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
               Substrate for the First Site Visit.
                                              F-8
-------
             100000
              10000
          CM
          <
          0>
              1000
               100
          a
          CL
                10
                 1
                      fl        Vtad       Vty      Ahmhum       Other      Unknown

                                             Substrate
Figure F-4.    Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
              by Substrate for the First Site Visit.
                                          F-9
U.S. EPA Headquarters Library
       Mail code 3201
1200 Pennsylvania Avenue NW
   Washington DC 20460
-------
             1000000
              100000
          cvi    10000
                1000
          S.      100
                 10
                                 Wood        Vnyl      ALmwi
                                                Substrate
Other      Unknown
Figure F-5.    Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the First Site Visit.
                                             F-10
-------
             1000
          w  100
          CL   io
                                Wood
Vhyl
Aluminum
Other
Unknown
Figure F-6.    Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
              Substrate for the Passed Clearance Visits.
                                             F-11
-------
             1000
          CM  100
          ^
               10
                     All
Wood
      Aluminum       Other       Unknown
Substrate
Figure F-7.   Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
              by Substrate for the Passed Clearance Visits.
                                            F-12
-------
             1000
          CM   100
          S.    10
                    Alt
Wood        Vryl       Alumirun      Other
               Substrate
Unknown
Figure F-8.    Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the Passed Clearance Visits.
                                           F-13
-------
  Table F-3. Observed Within-Room Correlation Coefficients between Pb Loading
            Measurements Collected from Floors, Window Sills and Troughs for the
            Cleveland Data.
         Site
                     •-.- >bbservedWithiii
                  Floors and Window
                 Pnoor.SiU»
         First
0.14
64
0.30
24
0.47
13
       The conditional probabilities of a sample passing or failing a standard are given in
Table F-4. These analyses were conducted on two different sets of data, the first set using all
possible paired observations from within the same room. The second subset of data restricted the
analyses to rooms in which floor, window sill, and window trough lead loadings were
simultaneously observed.
       As shown in Table F-4, results from contingency table estimates and normal theory
estimates are consistent, although more so on the left side of the table than on the right. For both
types of estimates and both data sets included in the analysis, the probability that one component
passes clearance given that another component passes clearance is quite high (over 88%), while
the probability that one component fails clearance testing given that another component in the
same room fails clearance testing is low (mostly under 26%).

F-4.    Objective 4: Demonstration of the impact of Composite Sampling on Pass/Fail Rates
       of Houses
       Individual sample lead loading results from each component type within a residential unit
were combined to construct simulated composite sample results.
       Table F-5 provides the number of residential units that were investigated by the number
of individual samples that were collected for each component type. For example, there were 6
residential units which included four window sill dust-wipe samples on the first site visit.
Residential units containing more than four samples from a component type resulted in the
estimation of multiple simulated composite sample results in this analysis. Therefore,
approximately 37% (14/38) of the residential units had five or more samples and resulted in the
                                         F-14
-------
JO
 CO
.a

 s
O.
 <0

 w
 I

 0
 O)



 o
u

CM

 X
 O)

*«


•o

I


'+:
 (0
Ul

 o>
.£
 8
 «>
.2  u.
£  o


I
 g  S
 go

S  >

12  °
 c  v
_£

.a
a
                                                    F-15
-------
 estimation of multiple simulated composite samples from floors.  Conversely, only 9% (3/33) of
 the units which had window sill samples collected and none of the residential units which had
 window trough samples collected resulted in the estimation of multiple simulated composite
 samples. When there were four or fewer individual samples from a component type within a
 housing unit, the simulated composite sample included all samples.
Table F-5.   Number of Residential Units that Contained (N) Individual Samples of Each
            Component Type Based on First Site Visit Clearance Testing Data From the
            Cleveland Lead Hazard Abatement Center.
                    9 +
                    Total
               Total with N;>5
                                   13
                                   10
                                   10
38
14
                                                    '
             13
33
                                                        •^ti'>"™&'?- *iw^ss.^"r
                                                             14
24
      For each component type within a residential unit, the set of individual clearance sample
lead loading results was used to construct simulated composite samples for the purpose of
evaluating the three composite sample clearance criteria (Standard, Standard/n, and
2xStandard/n). For each combination of component type and composite clearance criterion,
Table F-6 indicates the number of residential units that either passed clearance, failed clearance,
or yielded inconclusive results. The percentage of residential units that had more than four
individual samples for a particular component which actually resulted in inconclusive results was
                                         F-16
-------
low (no more than 5.3% for any given combination of component type and clearance criterion)
for this data set.
Table F-6. Individual Sample Clearance Results Versus Simulated Composite Clearance
           Results Based on Data from the Cleveland Lead Hazard Abatement Center.

                     mposite Sample
                                              ihliiv^



                                                                          '
     Standard
                   Pass
Inconclusive
                   Fail
                    27
32
22
     Standard/n
                   Pass
Inconclusive
                   Fail
                    20
30
                            11
22
   2xStandard/n
                   Pass
Inconclusive
                   Fail
                    26
32
22
       The performance characteristics of each combination of component type and composite
clearance criteria are presented in Table F-7 in terms of sensitivity, specificity, positive
predictive value (PPV), and negative predictive value (NPV). These performance characteristics
attempt to estimate the false negative and false positive error rates for the different clearance
criteria.
       The Standard/n Rule is formulated such that specificity is always sacrificed for
sensitivity. The sensitivity for the Standard/n Rule is always 1.00 while the specificity for this
rule is estimated to be 0.74 for floors, 0.94 for window sills, and 1.00 for window troughs. On
the other hand, the Standard Rule is designed so that it always sacrifices sensitivity for
specificity. The specificity for the Standard Rule is always 1.00 while the sensitivity for this rule
is estimated at 0.18 for floors, 1.00 for window sills, and LOO for window troughs. For the
2 x Standard/n Rule, the values of sensitivity are higher than or equal to those calculated for the
Standard Rule, while the values of specificity are higher than or equal to those calculated for the
Standard/n Rule. Estimates of sensitivity and specificity in these examples are always
                                          F-17
-------
 Table F-7. Performance Characteristics of Composite Clearance Criteria Based on Data
           from the Cleveland Lead Hazard Abatement Center.

Floors
Window
Sills
Window
Troughs


Sensitivity
Specificity
PPV
NPV
Sensitivity
Specificity
PPV
NPV
Sensitivity
Specificity
PPV
NPV


0.18
1.00
1.00
0.79
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00

1.00
0.74
0.69
1.00
1.00
0.94
0.50
1.00
1.00
1.00
1.00
1.00
S^2x^ln^^4|-fi
0.73
0.96
1.00
0.90
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
conservative, because inconclusive composite test results factor into the denominator for each
inconclusive composite test results factor into the denominator for each estimate, but never factor
into the numerator.
       It is evident that all three composite clearance testing criteria have different sensitivities
and specificities associated with their application to the simulated composite samples. To
evaluate the performance of each of the three composite clearance criterion in another manner,
the following logistic regression model was fitted for each combination of component type and
composite clearance criterion. The goal was to describe the relationship between the probability
of passing clearance and the maximum lead loading present in all of the individual samples
tested in a residential unit:
where nijk is the estimated probability of clearance for component(/) in house(i) under composite
criterion (&), and Max,-, is the maximum individual sample lead loading result in house(i) for
component(/).
                                          F-18
-------
       The estimated relationship between the probability of passing clearance and the
maximum lead loading for floors and window sills is presented graphically in Figure F-9. The
relationship was not able to be estimated for floors using the 2xStandard/n Rule and for window
sills using either the 2xStandard/n Rule or the Standard Rule because of the nature of the data.
The data were such that, under the 2* Standard/n Rule for floors and under the 2>
-------
                 Probability of Passing Clearance Testing Using Composite ROOT Samples
                       t
                       i
                       t
                        i
                        i
                        i
                        i
                        i
                        i
                        i
                         i
                         i
                         i
                         i
                         t
                                                                 Approach

                                                                 Standard Bute
                                                         	(Standard / n) Rute
                       100      200     300     400     500     600

                              Maximum individual Sample Floor Pb Loading
 700
 800
              Probability of Passing Clearance Testing Using Composite Window SOI Samples
                                                                Approach

                                                                - (Standard / n) Bute
                       250      500      750     1000    1250    1500

                            Maximum Individual Sample Window Sill Pb Loading
1750
2000
Figure F-9.  Estimated Relationship Between the Probability of a Residential Unit Passing

            Clearance Testing versus the Maximum Individual Lead-Loading Result by

            Component Type Based  on Simulated Composite Samples from Cleveland Lead

            Hazard Abatement Center.
                                            F-20
-------
**•* Q)
0 .E
>. "5.

i I

"§ w
 5 c
 CO

 M 01
 » c
   •D a
 O 0) "O

1 Is
 * 5 »
 • w |

 O « O
s = «

 §i25
I-S E
 10 £ O
 £ c^
 ro E (a
 o 2 5
cc £ ro
 o 'x Q
n "O 2
ESl
S  ««r
^  « c

ill

111
CC O O
u

«
J5
n
                            CM
                            m
                            o
                                s
CO
                                   8
    (D
   CM
o
^
u.
               O
               O
               O
               o
               CO


               C)
                                              CD
                                              co

                                              d
              en
              o
              o
               in

               o

               9
              in
              IN
              10

              (N
              es
              tv
              h.

              ID
                      CM
                         
                  CM
                  CM
    CM

    CM
CM

CM
               CO
              •o

               ID
TO
•a

-------
-------
     APPENDIX G

Dover Housing Authority,
    New Hampshire
          G-1
-------
                                    APPENDIX G
                           DOVER HOUSING AUTHORITY,
                                  NEW HAMPSHIRE

       One-hundred and eighty-four units in 49 multi-family buildings were abated and
renovated [20]. Forty-three of the buildings had four units per building and the other six
buildings had two units per building. The exterior siding and window sashes of the buildings
contained lead-based paint, so siding and windows including the sashes were removed and
replaced. Interior lead-based paint was found only on water radiators. Thirty-one of the units
had cast iron radiators. The radiators were removed from the apartments, sand blasted and
repainted offsite, and reinstalled.  Even though there was very limited abatement activity inside
the units, dust wipe samples from the window troughs and sills along with floor samples were
analyzed for lead clearance.

G-1.   Objective 1: Characterization of the Number of Individual Samples. Work Areas, and
       Housing Units That Pass or Fail Clearance Testing Standards
       The Dover Housing Authority tested all 184 units for clearance. Clearance data were
made available for this analysis from 704 rooms in 158 of the 184 residential units that were
abated and renovated. Nine-hundred and sixteen individual dust wipe samples were collected
over a six month period of time, spanning from May 1992 through October 1992.  Three
different components were sampled with 633 samples (69%) being collected from floors, 18
samples (2%) from window sills and 265 samples (29%) from window troughs. Table G-1
presents the number of individual samples, work areas and residential units that passed or failed
clearance testing within each combination of component type and site visit. All of the dust
samples were collected during the first site visit to a residential unit.  On the first site visit, 97%
(891/916) of the individual dust samples fell below the clearance standards of 200 ug/ft2 for
floors, 500 ug/ft2 for window sills and 800 ug/ft2 for window troughs,  96% (679/704) of the
rooms passed clearance and 87% (137/158) of the residential units met clearance standards.  This
increase in the failure rate from the percentage of individual samples and rooms that fail to the
percentage of residential units that fail is attributable to the fact that if any individual sample
exceeds the standard and fails clearance, then the residential unit from which it was taken also
                                         G-2
-------
Table G-1.  Clearance Testing Results by Individual Sample, Room, and Residential Unit for
            the Dover Housing Authority.
   visit, :>
         ^Component

                                   .
                                                          -petal?
            Floor
620
13
633
610
13
623
140
11
151
             Sill
 11
        18
        11
               18
   First
                              12
            Trough
260
       265
        256
               261
              141
                      146
             All
891
25
916
679
25
704
137
21
158
Total
Floor
Sill
Trough
All
620
11
260
891
13
7
5
25
633
18
265
916
610
11
256
679
13
7
5
25
623
18
261
704
140
7
141
137
11
5
5
21
151
12
146
158
fails clearance. The residential units that failed clearance on the first site visit
were revisited for a second clearance testing. The location of the failed test was recleaned and
retested.  All of the samples collected during retesting passed the clearance standards. The
retesting data were not available.  This report only analyzed the first site visit data.
       The failure rate for individual samples was highest for window sill samples (39%),
followed by floors (2.1%) and window troughs (1.9%).  Floors were also the most densely
sampled component with an average of 4.2 samples (633/151) taken per unit while 1.8 samples
(265/146) were taken from window troughs per unit. Among 158 housing units which were
included in the Dover data, only 12 housing units had window sill samples. While nine housing
units had only one window sill sample per house, 3 housing units had multiple window sill
samples collected within a house. Altogether there were 18 window sill samples collected in 12
housing units (an average  of 1.5 samples per unit). Failure rates of residential units were low
based on floors or window troughs individually. Only 7% (11/151) of the residential units would
have failed based on the results of floor samples and 3% (5/146) of the residential units would
have failed based on the results of window trough samples. Among 12 housing units which had
window sill samples, 5 housing units had at least one window sill dust-lead loading exceeding
                                          G-3
-------
the HUD Interim Guideline window sill standard, which resulted in a residential unit failure rate
of 42% (5/12) based on the results of window sill samples.

G-2.   Objective 2; Characterization of the Distribution of the Dust-Lead Loadings.
       Geometric Mean Dust-Lead Loadings. Variability Within a Housing Unit, and
       Variability Between Housing Units
       Table G-2 presents geometric means and estimated variance components for lead
loadings by component type. Dover data had a high window sill geometric mean, 462 ug/ft2, and
low floor and window trough clearance sample geometric means, 13.5 jig/ft2 and 15.8 ug/ft2,
respectively.  The interior renovation in Dover consisted of 1) removing lead painted radiators
generally located under a window, and 2) window replacement. Window replacement did not
necessarily include replacing the window sill.  Among 12 housing units which had window sill
samples, 4 housing units with 5 window sills had dust-lead loading results over 1,100 jig/ft2.
The fact that floor and window trough dust-lead loadings for those 4 housing units were all
below HUD Interim Guideline standards suggests that window sills might have failed to be
cleaned after renovation.
       Figures G-l and G-2 contain box and whisker plots that present the distribution of dust-
lead loadings from the first and passed clearance visits by component type. Figures G-3 to G-5
contain box and whisker plots that present the distribution of dust-lead loadings from the first
visit for floor, window sill, and window trough, respectively; Figures G-6 to G-8 contain box and
whisker plots that present the distribution of dust-lead loadings from the passed clearance visits.

G-3.   Objective 3: Characterization of  the Correlation Between Components Sampled in
       the Same Work Area
       The relationships among floor, window sill, and window trough wipe samples are another
important aspect of examining a clearance testing program.  One method of assessing the
relationships among individual floor, window sill, and window trough wipe samples collected
from the same room is to estimate linear correlation coefficients. Table G-3 displays the Pearson
product-moment correlation coefficients and the associated sample size for the log lead loading
measurements of individual floor, window sill,  and window trough samples taken within the
same room. The data shows that floor samples are positively correlated with both window sill
                                          G-4
-------
 o

I

I

 I
 o
o
*=
(0
LU
 o
 a
 i
 a
•e
 a

•o

 a
I"
=  a

.2  w
0>
 v .£
H  §

 §i

 at  i-
 (0  >
•S  O
O O

                            CD
                            in

                            co
in
co
CM
CO
n
m
co
                            o
                            o
                                 CO
     in
     ID
     CM
          to
          CM
          n
          CD
                                 CM
                                 
-------
             100.0
          sr
          <   10.0
          c
          n
          K   1.0

               0.1
                    Boos       WntawSfc    Wnfaw Troughs      Roos       Window Sis
                              RrstM                         Passed Clearance
                                               Surface
Figure G-l.   Comparison of In-Transformed Dust-Lead Loadings from the First Site Visit
              vs. Passed Clearance Results on an Expanded Scale.
                                            G-6
-------
             700
          CM
          <   500
             400


             300


             200


             100


              0
                  ROOTS
Window Sills
  Surface
Figure G-2.   Box and Whisker Plots of the Distribution of Dust-Lead Loadings from the
              Passed Clearance Data.
                                           G-7
-------
              10000
               1000
           g    100
                 10
                       All
Wood        Vnyl       Aluminum      Ofter      Unknown
               Substrate
Figure G-3.   Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
               Substrate for the First Site Visit.
                                              G-8
-------
             100000
              10000
          CM
          <
               1000
           0>
                100
                                Wood
Vnyl
Aiuroinum
Other
Unknown
Figure 6-4.   Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
              by Substrate for the First Site Visit.
                                             G-9
-------
             1000000
              100000
          oi   10000
           g    1000
          TJ
                 100
                  10
                        AS        Wood       Vnyl       Aluminum
                                                 Substrate
Other
Unknown
Figure G-5.   Box and Whisker Plots of the Distribution of Window Trough Dust-Lead
              Loadings by Substrate for the First Site Visit.
                                             G-10
-------
            1000
             100
         £   io
                   All
Wood       Vnyl       Aluminum

               Substrate
Other
Unknown
Figure G-6.   Box and Whisker Plots of the Distribution of Floor Dust-Lead Loadings by
             Substrate for the Passed Clearance Visits.
                                          G-11
                                  U S. EPA Headquarters Library
                                         Mail code 3201
                                  1200 Pennsylvania Avenue NW
                                     Washington  DC 20460
-------
              1000
           CM   100
           Q
           I
           3
           S.   10
                     Al
Wood        Vnyj      Akminun       Other       Unknown

               Substrate
Figure G-7.    Box and Whisker Plots of the Distribution of Window Sill Dust-Lead Loadings
               by Substrate for the Passed Clearance Visits.
                                             G-12
-------
             1000
          CM  100
          0)
              10
K
CL
                    AJ
                    Wood        Vinyl       Akmnum       Other

                                   Substrate
Unknown
Figure G-8.   Box and Whisker Plots of the Distribution of Window Trough Dust-Lead

             Loadings by Substrate for the Passed Clearance Visits.
                                           G-13
-------
 Table G-3. Observed Within-Room Correlation Coefficients Between Pb Loading
            Measurements Collected From Floors, Window Sills and Troughs for
            the Dover Housing Authority Data.
i
- 1. f
<-.
•v - *
'>>
*"• s .. _^
>
"* « ,
...Site Visit *
First
-v
'f
#
* City.'
Dover
" * * 'r'.T' ' ' " '* *
Observed*Within*RoornPCorrelation Coefficients Between!
V- ' L irffcsT
Floors and Window Sills-
?.
" r> **"*
Pptoor* suit
0.16
* *3- S* « * IS* ;T I j
i;n- _Jt *•
13
i, . J£§£*^f|
" Floors and Window Troughs
iff
Pploars Tioujli* k
0.12

184
and window trough samples. The correlations between floor and window trough results and
between floor and window sill results were not significant at the 0.05 level. A lack of data
prevented a correlation between window sills and window troughs from being estimated.
       The conditional probabilities of a sample passing or failing a standard are given in
Table G-4.  These analyses were conducted on two different sets of data, the first set using all
possible paired observations from within the same room. The second subset of data restricted the
analyses to rooms in which floor, window sill, and window trough lead loadings were
simultaneously observed. Note that the conditional probabilities results between window sills
and floors or window troughs may not be reliable due to lack of data.
       For both types of estimates and both data sets included in the analysis, between floors and
window sills and between floors and window troughs, the probability that one component passes
clearance given that another component passes clearance ranges from 60% to  100%, while the
probability that one component fails clearance testing given that another component in the same
room fails clearance testing ranges from 0% to 54%.

G-4.   Objective 4: Demonstration of the Impact of Composite Sampling on Pass/Fail Rates
       of Houses
       Individual sample lead loading results from each component type within a residential unit
were combined to construct simulated composite sample results.
       Table G-5 provides the number of residential units which were investigated by the
number of individual samples collected from each component type. For example, there were 17
residential units where five floor dust-wipe samples were taken. In this analysis, residential units
containing more than four samples from a component type resulted in the estimation of multiple
                                         G-14
-------
JD
 CO
12
 O
  o


 o
Z

•o

 (0
 (A
 I
 O

U

CM

 X

CM
 in
iu

 o>

"•i5
 0>
 0)
 o
 c  *i
 «  to

 a  >

    •
    .
 a>  iu
I  a>

•i  ^
 >o  *-
-°  n
 2
    "
 O  £
O  t-

 63
1
F
i
•s
S:
"5
g
S
i1;

i
s-
•*•'
• .
H.
>-
C

o
ri:


"-

s
•p
•»=

1
*5
£
I
i
i
K
i
"J"
I






• ."




ill

CA ""S C
IL* 5
•1
JB
5
B-.
;*- j E

'«J*g |
IL"? 5



. .
•^
II
,3
i*. '
1 " '
< i '
•« •
5 '
Condltt
if!


it»*-S
J
'' 8 8
•££

i- i 6
1|]
* ^ *

a •
§ "i
: «

•*• • =
2
I-
*• -
§
*g

t3







s
0
o









§
e


o
S
CO
| P(Fi200





in
in
en
0
o
8


o
o
o
V-



§
V
CO
o
o
CN
V
IL
oT





n
UD
e









.



o
o
CM
U.
| PIStBOO





en
en
to
o
e
o
9


-------
 simulated composite sample results. Therefore, summing all the units that had five or more
 individual samples, approximately 19% (29/151) of the residential units resulted in the
Table G-5.  Number of Residential Units that Contained (N) Individual Samples of Each
            Component Type Based on First Site Visit Clearance Testing Data From the
            Dover Housing Authority.
                      Total
                 Total with N*5

                                     111
                                     17
151
29
12
                                                                32
                                                                110
146

estimation of multiple simulated composite samples from floors. None of the residential units
tested resulted in the estimation of multiple simulated composite samples from window sills or
window troughs. This occurred since all of the units investigated had four or less individual
samples taken from window sills and four or less samples taken from window troughs. When
there were four or fewer individual samples from a component type within a housing unit, the
simulated composite sample included all samples.
       For each component type within a residential unit, the set of individual clearance sample
lead loading results were used to construct simulated composite samples for the purpose of
evaluating the three composite sample clearance criteria (Standard, Standard/n, and
2xStandard/n). For each combination of component type and composite clearance criterion, each
residential unit either passed clearance, failed clearance, or yielded inconclusive results based on
the simulated composite samples. Inconclusive results were only possible for those residential
units which contained multiple simulated composite samples. Uncertainty in the decision rule is
                                         G-16
-------
created by the multitude of ways in which multiple composite samples can be formed within
residential units which contained five or more individual clearance samples.  Some of the
possible multiple composite samples formed may pass while other possible composite samples
may fail.  As seen in Table G-6, the incidence of this type of uncertainty was very low for this
data set compared to the other data sets. The maximum observed percentage of residential units
with greater than four individual clearance samples of a given component type that resulted in
inconclusive results was 2.0% for floor samples using the Standard Rule.
Table G-6.  Individual Sample Clearance Results Versus Simulated Composite Clearance
            Results Based on Data from the Dover Housing Authority.
   IfiCbmpbsiteo*
   VKejxfusitfi&rtpgs
    iCearance^
                                                      Sample Clearance I
                                              Fail.:;
                                   SPass'-
                  Pass
      Standard
Inconclusive
                  Fail
                    140
                         141
                  Pass
     Standard/n
Inconclusive
                  Fail
                    124
                    14
11
                         141
                  Pass
   2 x Standard/n
Inconclusive
                  Fail
                    138
                         141
       The performance characteristics of each combination of component type and composite
clearance criteria are presented in Table G-7 in terms of sensitivity, specificity, positive
predictive value (PPV), and negative predictive value (NPV).  By design, the Standard/n Rule
sacrifices specificity for sensitivity. The sensitivity for the Standard/n Rule is always 1.00 while
the specificity for this rule is estimated to be 0.89 for floors, 1.00 for window sills, and 1.00 for
window troughs. On the other hand, the Standard Rule sacrifices sensitivity for specificity. The
specificity for the Standard Rule is always 1.00 while the sensitivity for this rule is estimated to
be 0.27 for floors, 1.00 for window sills, and 1.00 for window troughs. The 2xStandard/n Rule
attempts to maximize both sensitivity and specificity. The values of sensitivity for the
2*Standard/n Rule are always at least as high as those calculated for the Standard Rule. For
floors, the 2xStandard/n Rule, with an estimated sensitivity of 0.82, seems to do quite a bit better
than the Standard Rule.  Similarly, the values of specificity are at least as high as those calculated
                                          G-17
-------
for the Standard/ii Rule.  Estimates of sensitivity and specificity in these examples are always
conservative, because inconclusive composite test results (i.e., houses in which only a fraction of
the simulated composite samples passed or failed) are included in the denominator for each
estimate, but are never included in the numerator.
Table G-7. Performance Characteristics of Composite Clearance Criteria Based on Data
           from the Dover Housing Authority.

                   Sensitivity
0.27
1.00
0.82
       Floors
                   Specificity
1.00
0.89
0.99
                   PPV
1.00
0.44
0.82
                   NPV
0.97
1.00
0.99
                   Sensitivity
1.00
1.00
1.00
      Window
        Sills
                   Specificity
1.00
1.00
1.00
                   PPV
1.00
1.00
1.00
                   NPV
1.00
1.00
1.00
                   Sensitivity
0.60
1.00
1.00
      Window
      Troughs
                   Specificity
1.00
1.00
1.00
                   PPV
1.00
1.00
1.00
                   NPV
0.99
1.00
1.00
       Table G-7 shows that each of the three composite clearance testing criteria can have
different specificity and sensitivity rates.  These rates correspond to the consistency between
clearance decisions based on individual clearance samples and clearance decisions based on
simulated composite samples.  To further characterize the performance of each of the three
composite clearance criterion, the following logistic regression model was fitted for each
combination of component type and composite clearance criteria to describe the relationship
between the probability of passing clearance and the maximum lead loading present in all of the
sampling locations tested in  a residential unit:
where nijk is the estimated probability of clearance for component(/) in house(j) under composite
criterion (£), and Max,j is the maximum individual sample lead loading result in house(0 for
component(/).
                                           G-18
-------
       The estimated relationship between the probability of passing clearance based on
simulated composite samples and the maximum lead loading in individual floor samples is
presented graphically in Figure G-9. In this figure, the solid, long-dashed, and finely-dashed
lines represent the estimated relationship for the Standard, Standard/n, and 2xStandard/n Rules,
respectively. These same relationships for window sills and window troughs were inestimable
because of the nature of the data. Essentially, an infinite number of curves could be fit to
estimate the relationship, but there is no criterion for choosing the optimal one.
       Table G-8 provides parameter estimates and associated standard errors from the logistic
regression models, as well as estimates of the probability of passing clearance (using composite
samples) when the maximum lead loading among all locations included in the composite
sampling scheme is greater than or equal to 1A, 1,2,  and 4 times the associated interim HUD
standard for individual samples.  The estimates for the probability of passing based on composite
samples formed from window sill samples or window trough samples do not appear in the table
because the related parameters are not estimable due to the reasons stated in the previous
paragraph. The low sensitivity of the Standard Rule is demonstrated by the high estimated
probability (0.83) of passing when the maximum individual floor sample lead loading is equal to
the 2 times the HUD Standard. The estimated relationship for the Standard/n Rule demonstrates
this rule's high sensitivity (probability of passing is very low when the maximum individual
sample lead loading is greater than or equal to the HUD Standard) along with the loss in
specificity for this rule (probability of passing is only 0.73 when the maximum individual sample
lead loading is equal  to !/2 HUD Standard). Once again, the 2xStandard/n Rule is shown to be a
compromise between the Standard and Standard/n Rules. At 1A HUD Standard, the estimated
probability of passing clearance testing under the 2xStandard/n Rule is 0.99, and at 2xHUD
Standard, the estimated probability of passing clearance testing under the 2 x Standard/n Rule is
0.00.
                                          G-19
-------
              Probability of Passing Clearance Testing Using Composite ROOT Samples
                                                            Standard Rule
                                                    	(2 x Standard / n) Rule
                                                    	(Standard / n) Bute
                            200     300     400     500      600
                            Maximum Individual Sample Floor Pb Loading
Figure G-9.   Estimated Relationship Between the Probability of a Residential Unit Passing
             Clearance Testing versus the Maximum Individual Lead-Loading Result by
             Component Type Based on Simulated Composite Samples from the Dover
             Housing Authority.

                                          G-20
-------
£ a


li
2 w

2 S
°- '«
Q) O
£ &

C E
g o
   *
   «

II
CC O
      .
W    =
e -1 <



I « ?
   0) .=
   TO Q

111
.2 .£ **
« ^ £
g = o
.   x
 0 ^ •»


 2^1
•^ o

•S c §
 3 S «
   O Q
CO

6
            •lStiLl-_?S *K*
            SSJfrV.i

            •'->?* 1-1
.lit

Blls;
-;-S*»5',««
^524jg;
              y»:-c
               ^ll
               5J«0.
            S^i?;
            :^T"\.J «;*
            ",;«:.4^:'.'-.-i;Si
            iii
            ilil
            " ^p  >«-^



                    ::;*ib£

                              g
        8
        d
                           P)
                           CO
        o
        q

        d
                           o>
                           a>
                              I";

                              d
       CM
       o
o
o
                              03

                              0}
                      CO


                      d
                           CD

                           O

                           O
                              q

                              d
       q

       d
CD
^
q

d
                              in
                              n
                              o
                                  to
                                  q

                                  d
                          s
                          CN
       O

       5
                           ao
                              en
       s
       q
       (D
       U>

       CM
                               a

                              ?
                               Q
                               X

                               CM
       I
        ID
        **

        CO
                                      I
                                      CO
                                          0
                                          •o
                              (O

                              X

                              (M
-a


I
                                                  a
                                                 •o
                                                  c
                                                  a
                                         a

                                         ?
                                         a
X

CM
    .C

    s

    <5

    ?
    
-------
-------
                              APPENDIX H

    Additional Analysis Results on the Percentage of Housing Units That
Passed Clearance on the First Site Visit, by the Number of Individual Samples
                     Collected Within a Housing Unit
                                  H-1
-------
                                     APPENDIX H
        Additional Analysis Results on the Percentage of Housing Units That
    Passed Clearance on the First Site Visit, by the Number of Individual Samples
                           Collected Within a  Housing  Unit
       Appendix H presents additional analysis results on the percentage of housing units that
passed clearance on the first site visit, by the number of individual samples collected within a
housing unit. Clearance standards used in this appendix were based on HUD Interim Guidelines
(200 ug/ft* for floors, 500 ng/ft2 for window sills, and 800 fig/ft2 for window troughs). Table H-
1 uses a format similar to that of Table 6-13 in the text, but Table H-l presents the percentage of
units that passed clearance on the first site visit for the applicable component in each part of the
table (floor, window sill, and window trough).
       Tables H-2a, H-3a, and H-4a present the number of residential units that contained
individual samples for the first site visit clearance testing data from the HUD Grantee Program
(both High and Low data groups), by interior intervention strategy code, for floors, window sills,
and window troughs, respectively. Tables H-2b, H-3b, and H-4b, companion tables for Tables
H-2a, H-3a, and H-4a, present the percentage of residential units that passed clearance for the
first site visit clearance testing data from the HUD Grantee Program (both High and Low data
groups), by interior intervention strategy code, for floors, window sills, and window troughs,
respectively. HUD Grantees reported the intensity of the interior intervention as a strategy code
(level 01 to 07).  Higher strategy levels reflect more intensive treatment. Each dwelling unit was
assigned only one interior intervention strategy. The interior intervention strategies were
summarized in Table 4-1.
       Table H-2a, H-3a,  and H-4a show that, regardless of component type, level 05 of the
interior intervention strategy was used in housing units more  than any other intervention strategy
for both HUD Grantee High and Low data groups. Level  05 of the interior intervention strategy
includes window replacement and wall enclosure/encapsulation, and other lower levels of
intervention activities.
       Tables H-2b, H-3b, and H-4b show that, generally, the percentage of housing units
passing clearance does not increase as the intensity of the  interior intervention increases. This is
                                          H-2
-------
true across all 3 components and for both HUD Grantee High and Low data groups. The most
popular interior intervention strategy, level 05, has residential unit passing rates of 86%, 95%,
and 94% based on floors, window sills, and window troughs, respectively, for the HUD Grantee
High data group.  For the HUD Grantee Low data group, interior intervention strategy level 05
has residential unit passing rates of 83%, 91%, and 91% based on floors, window sills, and
window troughs, respectively.
                                          H-3
-------
Table H-1. Percentage of Housing Units that Passed Clearance for Each Data Source That
            Contained (N) Individual Clearance Samples of Each Component Type
            Based on the First Site Visit.
-fiSJ^Data!i^rui!c?|S^f|
'^^^^^^^^^M^^^^^^^^^Utt^
M-SiSi
HlUt
•,-s.v. £"amt.
.tei^sss
?Klfe#
~!V>';L'Wi.^:i,M
^;w3-.;X'-r
••*;""-•**?#
••'. -.f^i-SS
*s-fci






... Floor Standard 200 t/g/ft2 •.-•_.' :
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)'
HUD Grantee (Low)"
Atlantic City
Cleveland
Dover
89
50
100
87
93
100

67
87
71
92
86
92
86
100
100
80
33
75
89
82
70
62
83
81
63
69
86
92
85
50
96
73
58
71
82
90
70
60
82
70
53
63
85
78
80
100
88
67
47
76
79
57
100

100
74
65
38
81
50
50
.
67
64
41
80
60
67
100
100
•
79
54
74
85
89
80
63
93
"— ~. -, " " '-- •. . Window Sill Standard 500 j/g/ft2 I - ""•';--*•"•- "- -
Maryland
HUD FHA
HUD PHA
HUD Grantee (High)'
HUD Grantee (Low)"
Atlantic City
Cleveland
Dover
79
88
100
90
94
100
100
78
79
60
93
94
95
89
100
0
74
71
87
89
91
67
92
0
73
88
96
88
73
50
100
0
88
57
86
90
75
100
100

80
70
96
65
50

.
.
71
55
.
100
.
.

.
78
33
100
100
.
.


69
57
.
0
0

.
-
77
63
93
91
93
80
97
58
Window Trough Standard 800 //g/ft2 '• • • - ...-,, "•'
Maryland
HUD FHA
HUD PHA
HUD Grantee (High}'
HUD Grantee (Low)"
Atlantic City
Cleveland
Dover
72
92
100
92
91
100
93
94
65
50
87
89
86
95
83
97
62
26
100
94
85
100
100
100
65
38
100
96
62
83
100
100
63
50
91
92
67
100

.
68
21
100
71




36
50

100
.
.


46
0
100
100
.
.
.

47
0


.
,
.
.
63
44
95
91
88
96
92
97
*  Grantees that used
* * Grantees that used
200 jug/ftz as clearance standard for floor.
100 //g/ft2 or 80 //g/ft2 as clearance standard for floor.

                                            H-4
-------
Table H-2a.  Number of Residential Units that Contained (N) Individual Floor Samples for
             the First Site Visit Clearance Testing Data from the HUD Grantee Program,
             by Interior Intervention Strategy Code.
.i»Jnteiy«ntion,-jW:i?'
lsSt^egy;Co^l|!

prill
•- • f-
01
02
03
04
05
06
07
Missing
Total
14
30
24
29
83
1
0
55
236


10
70
31
80
113
0
2
107
413
• rPi.JSis&pigS-
^tttf.-'WtaElW
fe4W
$^~|||Ki
•^""SS^i^'r!;
'3f,»V>ivirji
^JJ6*rr»
••»*:S^K'ii^!<'t;
•^-i^T'-i-j,^
>'vr.v7?r,**..»v



, HUD Grantee (High)* -
7
28
15
64
123
1
2
94
334
4
24
9
61
128
2
1
114
343
2
15
7
45
108
5
0
76
258
2
9
5
38
161
4
0
69
288
0
1
4
11
93
2
0
31
142
0
0
0
19
14
0
0
3
36
- . ... - ' : : HUD Grantee (Low)"
01
02
03
04
05
06
07
Missing
Total
2
21
15
9
40
0
0
42
129
4
5
21
15
34
4
0
41
124
1
4
33
19
48
10
1
43
159
1
10
40
19
59
47
0
170
346
0
4
44
9
53
15
0
76
201
0
2
15
6
8
1
0
18
50
0
0
1
2
3
0
0
1
7
0
0
1
0
0
0
0
1
2
0
0
0
18
21
0
0
3
42

0
0
0
0
1
1
0
1
3
39
177
95
365
844
15
5
552
2092
' ."•*•• '
8
46
170
79
246
78
1
393
1021
  Grantees that used
  Grantees that used
200 /sg/ft2 as clearance standard for floor.
100 pg/ft2 or 80 jvg/ft2 as clearance standard for floor.
                                             H-5
-------
Table H-2b.  Percentage of Residential Units that Passed Clearance for Floors (Based on
             Standard at 200 //g/ft2) for the First Site Visit Clearance Testing Data from
             the HUD Grantee Program, by Interior Intervention Strategy Code.

^.liit«fv«*itioii.|fels;'l|i«^"'i-'»*-'."'-iv
Sstfet^c'odiftllisa'fe
^"H-^lIiT"' ,: *•£•"•'
i£i2' £••'
;^!p
'.sjsssf&sr
MS&





- " . HUD Grantee (High!" '/ • •-•••
01
02
03
04
05
06
07
Missing
Total
* - '*""»•
. . , ^ * , ' ~ ' '
01
02
03
04
05
06
07
Missing
Total
93
87
83
72
89
0
.
93
87
.
50
100
87
89
95
.
.
93
93
90
89
65
88
86
.
100
90
86
.'. ' .
100
80
90
87
91
100
.
95
92
71
86
80
94
89
100
100
89
89
100
88
78
82
85
100
0
89
86
50
87
43
84
89
60

76
82
100
89
100
82
86
75
.
81
85

100
75
73
80
100
.
77
79



84
71

.
100
81



50
67
.
.
67
60
87
88
74
83
86
73
80
86
85
HUD Grantee (Low)" ' ' .
100
100
94
74
77
60
100
86
82
100
80
93
74
81
96

97
92
.
75
91
78
81
93
.
96
90
.
50
93
67
63
100
.
78
78
.

0
50
100
.
.
0
57

.
0
.
.
.
.
100
50

.


0
100
.
100
67
88
89
91
77
83
91
100
94
89
* Grantees that used 200 pgltt? as clearance standard for floor.
** Grantees that used 100 /jg/ft2 or 80 //g/ft2 as clearance standard for floor.
                                             H-6
-------
Table H-3a. Number of Residential Units that Contained (N) Individual Window Sill Samples
             for the First Site Visit Clearance Testing Data from the HUD Grantee Program,
             by Interior Intervention Strategy Code.
-^^iSsSPSiPil
j'i sv- s?; UtUnOfnatSlSlar
fyt> j=>.i-ir«.-S«s>p;^!tMS»fe
^ lntwyttilMn%&
•t'S'tratelyttCodell?-:'

I?,-:;*:- -3i?,!-i^isilsp
-S^l»^5|*-s«2^|'

01
02
03
04
05
06
07
Missing
Total
•• ,-"'•'*.'•' ''.'. • " -'
01
02
03
04
05
06
07
Missing
Total
9
36
24
61
80
1
0
52
263
r*S.3SSR~i«f
PS33&
•rwaspQji
..,1:^.4^^
'Sifs^illfilefe-
;- 7 :.•;.?•

HUD Grantee (High)*
10
76
22
123
520
8
1
308
1068
11
37
20
107
155
3
3
165
501
3
13
7
51
42
2
1
14
133
0
3
2
10
24
0
0
2
41
0
0
0
8
9
0
0
3
20
0
0
0
1
1
0
0
0
2
1
0
0
0
0
0
0
1
2
' ;.:-' HUD Grantee (Low)"
2
13
11
7
30
9
0
48
120
2
28
128
31
108
49
1
240
587
2
3
20
30
80
9
0
70
214
0
0
3
7
15
0
0
15
40
0
0
1
1
0
0
0
2
4
0
0
0
0
1
0
0
1
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
1
34
165
75
361
832
14
5
545
2031

0
0
0
0
1
0
0
0
1
6
44
163
76
235
67
1
376
968
  Grantees that used
  Grantees that used
200 i/g/ft2 as clearance standard for floor.
100 //g/ft2 or 80 pg/ft2 as clearance standard for floor.
                                             H-7
-------
Table H-3b. Percentage of Residential Units that Passed Clearance for Window Sills (Based
             on Standard at 500 //g/ft2) for the First Site Visit Clearance Testing Data from
             the HUD Grantee Program, by Interior Intervention Strategy Code.
iiiSciiSiSysSi
•«»fSjS«(»«»".\«Hsp>
,-1*;£,liiieiv6!iti6ii?c<^^
*s'SaitwfCo" ...','?-. ••>:-• W..-T>>^-:. .r* :•»•-.,••!.? :'"W- s,v *»™i rr^iiiaetpfr'*-;', v-iis'j,-- siiii.^^wjsijT:^:'5.1 .I1 * ™''v,;«?4?f»HWi:**ip»s.M^j*^5^^
<$*|S*4l$S^


• .
67
78
96
87
93
100

98
90
90
80
86
93
97
88
100
93
94
100
73
90
94
93
100
100
85
89
^^^Mlifllt

HUD Grantee (High)* "
100
85
86
86
90
100
100
86
88

33
100
90
96

.
100
90
.

.
38
78
.
.
100
65
J.3.:~te~l.'$
&«57.>K$

.

.
100
100



100

'£M*i$eis IflWPSSS?
Ss^^li^rotaii!

100
.
.
.


.
100
100
.
.
.
.
.
.
,

.
88
78
91
90
95
93
100
91
91
-. •'.' .-!•'.-.•' •'•• "'•• . • -. '^ :,''<• ~'\ •• •-; .. 'HUD Grantee iUwj't; .:/!.;; , •••;••'• -..•.i."^'^ ••',../'• '--..'..;
01
02
03
04
OS
06
07
Missing
Total
50
100
91
86
97
89
.
96
94
100
96
95
87
93
98
100
95
95
100
33
95
80
90
100
.
97
91


67
29
80
.

87
73
.
.
0
100
.

.
100
75


.
.
0


100
50
.
.
.
.
.



-


.
.
.
.

.
.

.
.
.
.
.
.
.
-
83
93
94
79
91
97
100
95
93
  Grantees that used
  Grantees that used
200 pg/ft2 as clearance standard for floor.
100 //g/ft2 or 80 Arg/ft2 as clearance standard for floor.
                                             H-8
-------
Table H-4a. Number of Residential Units that Contained (N) Individual Window Trough
             Samples for the First Site Visit Clearance Testing Data from the HUD Grantee
             Program, by  Interior Intervention Strategy Code.
*|fe Interior;:*^
li'Strategy3 Codelif
^^^'i^^&^-i^^^^^^^^^^^^^*'y^^^^^^^^^l^^^^S^&
•iifciiif/jwi-ji"
•••MS*?-'— Ail- ;,'*
•sS??1;M*
••-."Vi'.'"-i ••**,'
'•^••.zS'^if
,£?!»! t"--'i'
'••^'^•'•i'-T"''' ••
•8*1*3 JX-;
-^!gSP|:|i»l;i^
lf:j'|4;35*j:|.^,!i|.5_ii^i

©^•gSpii'h^Si^MJ^irdtali
HUD Grantee (High)* "
01
02
03
04
05
06
07
Missing
Total
11
99
24
67
260
0
5
246
732
4
17
14
109
350
12
0
221
727
0
1
1
57
78
2
0
31
170
0
0
0
21
21
0
0
4
46
0
0
0
4
9
0
0
0
13
0
0
0
5
2
0
0
0
7
0
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
0
1
.. ••;: " '-
0
0
0
0
0
0
0
0
0
• •
15
117
39
284
721
14
5
502
1697
'L.:... . : '. . ^ . . - HUD Grantee (Low)" 5v : . ..•'>•.:-..••'
01
02
03
04
05
06
07
Missing
Total
4
13
59
32
148
51
1
187
495
2
27
72
27
52
14
0
108
302
0
2
21
10
9
0
0
18
60
0
0
0
4
3
1
0
5
13
0
0
1
0
2
0
0
0
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
6
42
153
73
214
66
1
318
873
* Grantees that used
* * Grantees that used
                 200 //g/ft2 as clearance standard for floor.
                 100 //g/ft2 or 80 //g/ft2 as clearance standard for floor.
                                            H-9
-------
Table H-4b.  Percentage of Residential Units that Passed Clearance for Window Troughs
             (Based on Standard at 800 /ig/ft2} for the First Site Visit Clearance Testing
             Data from the HUD Grantee Program, by Interior Intervention Strategy Code.
pSirP^Hi
^Strategy Code%

«t3f»? >v
•stSSJM
*£.*•'*(£*%:
&iS2-3;*a
lKs%.
g^'j-iO^.^.rr-

^ffs-^SiiS*
felit


•?.:'.•.•••-•.' "••I'.'i • : HUD Grantee (High)* :. l
01
02
03
04
.05
06
07
Missing
Total
100
88
96
95
96
.
100
88
92
.''-..-. • ' .'.'• " " ' r
01
02
03
04
05
06
07
Missing
Total
75
77
93
88
94
88
100
91
91
75
94
93
93
91
92
.
84
89

100
100
91
96
50
.
97
94



100
95

.
75
96
.


100
89
.
.
.
92
.
.

60
100


.
71
• ••;"•*.. HUD Grantee (Low)"
100
59
93
70
85
86
.
92
86

50
86
100
89

.
78
85
.
.
.
100
67
100
.
20
62
.

0

100
•
.
.
67

.
.
.

.
.


.
.
.
.
100
.
.
.
100
• -

.
.

.
.
.
.
.
.

.
100

.
.
.
100

.
.
.
.
.



93
89
95
93
94
86
100
86
91
.; ;- .. . '- - • ': '
,
.
.



.
.
.

.

.
.



.
83
64
92
84
91
88
100
90
88
  Grantees that used 200 //g/ft2 as clearance standard for floor.
  Grantees that used 100 jjg/tt1 or 80 fjg/ft* as clearance standard for floor.
                                            H-10
-------
       APPENDIX I
Additional Analysis Results
           1-1
-------
                                      APPENDIX I
                          ADDITIONAL ANALYSIS RESULTS

       Appendix I presents additional analysis results on the distributions of the clearance data.
Figures 1-1 to 1-6 display box and whisker plots for the distributions of floor, window sill, and
window well dust-lead loadings (ug/ft2) and log-transformed dust-lead loadings
(log ug/ft2) from the first site visit by substrates for Dover and Maryland data. Detailed values
for different percentiles for the untransformed dust-lead loadings (ng/ft2) are presented in Table
1-1. Figures 1-7,1-8, and 1-9 display box and whisker plots for the distributions of dust-lead
loadings (fig/ft2) and log-transformed dust-lead loadings (log ug/ft2) from the first site visit by
data source, for floors, window sills, and window wells, respectively. Similar detailed values for
different percentiles for the untransformed dust-lead loadings (ug/ft2) are presented in Table 1-2.
       Box and whisker plots of the percentiles by surface and substrate for Dover and Maryland
on the "Passed Clearance" data sets are provided in Figures 1-10 through 1-15. Specific upper tail
percentiles are given in Table 1-3. Figures 1-16 through 1-18 display the distribution of clearance
results for data sources that "Passed Clearance."  Additional percentile information for these data
sources is presented in Table 1-4.
       In Tables 1-2 and 1-4, Figures 1-7 to 1-9, and Figures 1-16 to 1-18, "Grantee 1" refers to
data from a group of nine grantees (Alameda County, Baltimore, Boston, California,
Massachusetts, Milwaukee, Rhode Island, Vermont, and Wisconsin) in the HUD Grantee
Program that used the  HUD Interim Guidelines clearance standards, i.e., 200,500, and 800 ug/ft2
for floors, window sills, and window troughs, respectively; "Grantee 2" refers to data from the
other group of five grantees (Cleveland, Chicago, New Jersey,  New York City, and Minnesota)
that used a lower floor dust-lead clearance standard (i.e., 100 ug/ft2 or 80 ug/ft2).  These were
referred as "HUD Grantee (High)" and "HUD Grantee (Low)" in the main body of the report.
                                           1-2
-------

        10000
    Of
    k
    e
woo
         wo-
                                                         Vlnyt
    *»
    I'
                                                        Vinyl
Figure 1-1.     Box and Whisker Plots of the Distributions of Floors Dust-Lead Loadings and
              In-Transformed Dust-Lead Loadings on Dover Data by  Substrate for the First
              Site Visit
                                           I-3
-------
    fi
       woo
        100
                  Al
                                              Vinyl
                                             Vta*
                                                          Mumtoum
Figure I-2.    Box and Whisker Plots of the Distributions of Window Sills Dust-Lead
             Loadings and In-Transformed Dust-Lead Loadings on Dover Data by
             Substrate for the First Site Visit
                                          I-4
-------

        K3000.0
        1000.0
    e
          KXO
          1.0
          0.1-
                                  Vinyl
                                               Aluminum
                                                               CXhtr
ID-
S'

w 7
1 e
j:
I 3
2'
1
0
-1




1






A










1 Vfe




-
I
U 5
t



«l Mumtom Qttm IMmmn
SubetratB
Figure [-3.    Box and Whisker Plots of the Distributions of Window Troughs Dust-Lead
             Loadings and In-Transformed Dust-Lead Loadings on Dover Data by Substrate
             for the First Site Visit
                                            1-5
-------
    £
       10000'
        1000-
         ttO-
         TOr
                                                             Other
       10
    f

                               Wood
                                                             Ottwr
Figure I-4.    Box and Whisker Plots of the Distributions of Floors Dust-Lead Loadings and
              In-Transformed Dust-Lead Loadings on Maryland Data by Substrate for the
              First Site Visit
                                            I-6

-------
    £
       100000'
        10000'
        woo
          1-
                              wsx*
                                                   Ahvrinum
                                                                Otar
11
10
9'
a
tlMinauot
-------
       1000000

    fi
        10000
         1000
         100
           1-
                                 Wood
                                               Vinyl
                                                            Mun*wn
13-
12
tl
TO
W 9
% *'
f7-
6
fs
4-
•
2
1-
0

1



















^

i !
_






'

























M Waod Vinyl AUrtntm IMnewi
Figure 1-6.    Box and Whisker Plots of the Distributions of Window Troughs Dust-Lead
             Loadings and In-Transformed Dust-Lead Loadings on Maryland Data by
             Substrate for the First Site Visit
                                           1-8
-------
Table 1-1. Percentiles (//g/ftz) of the Clearance Data by Surface and Substrate for the First
          Site Visit for Dover and Maryland.
' .,t'Data;~^
": Source "^
Dover
Maryland
• i i %
K \ * •
*• H
'Surface
Floors
Sitls
Troughs
Floors
Sills
Troughs
••* O ' -,
t \ •, *
Substrate "
All
Wood
Vinyl
Unknown
All
Wood
Vinyl
Aluminum
Unknown
All
Vinyl
Aluminum
Other
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Aluminum
Other
Unknown
All
Wood
Vinyl
Aluminum
Unknown
. Sample „
.' SiV'
633
3
625
5
18
9
1
4
4
265
185
11
2
67
27SO
1082
660
19
989
2465
1687
97
40
18
623
1985
458
551
286
690
l*50th .•
'Percentile •
5
31
5
9
443
718
212
443
993
7
7
10
136
7
19
18
21
84
17
28
24
12
10
12
48
92
283
43
51
133
80th ;
Percentile
44
61
44
72
1260
1260
212
467
2280
65
77
16
239
12
85
77
99
116
82
181
144
47
84
55
439
844
2160
239
337
1370
/•j-i-JBOth:^:^.
Percentile I
68
61
68
96
1620
1620
212
467
2280
137
152
16
239
68
181
159
200
725
182
511
376
116
186
327
1350
3120
7300
684
983
5700
|::|?J5thS|
fPercentiie*
107
61
107
96
2280
1620
212
467
2280
202
222
106
239
89
330
297
360
1110
347
1370
864
420
280
406
2790
9480
16600
1530
2580
15100
i!>>P.iereehtHe"v
: ?^::.-*- i™-.'.iaf.!.*iS-..rt •.
317
61
317
96
2280
1620
212
467
2280
1830
2040
106
239
643
1130
892
1400
1110
2580
5650
3590
6620
4080
406
9500
49800
59600
10200
38900
63800
                                           I-9
-------
           fi
100000.00


 10000.00


  1000.00


   100.00


    10.00


    1.00


    0.10


    0.01
                         PHA      Grantee 1   Grantee 2      FHA     Atlantic City   Cleveland
                                                 Data Source
11
10
9
8
* 7
< 6
H 5
E? 4
1 3
fi 2
* ;
-1
-2
-3
-4































,









j
4.
T
I
1
u
T
I
1






























                    PHA       Grantee 1    Grantee 2       FHA
                                              Data Source
                                                   Atlantic City    Cleveland
Figure I-7.   Box and Whisker Plots of the Distributions of Floors Dust-Lead Loadings and In-
             Transformed Dust-Lead Loadings by Data Source for the First Site Visit
                                              1-10
-------
              100000.00

               10000.00

               1000.00

                100.00

                 10.00

                  1.00

                  0.10

                  0.01
                         PHA     Grantee 1   Grantee 2      FHA     AttenBc City   Cleveland
                                                Data Source
11
10
9
8
sr 7
k 6
I :
.§ 3
§2
1
0
-1
-2
-3




























•









T
[

u
I

1



                    PHA      Grantee 1    Grantee 2       FHA
                                              Data Source
Atlantic City    Cleveland
Figure 1-8.   Box and Whisker Plots of the Distributions of Window Sills Dust-Lead Loadings
             and In-Transformed Dust-Lead Loadings by Data Source for the First Site Visit
                                             1-11
-------
             1000000.0

              100000.0

               10000.0

                1000.0

                 100.0:

                  10.0]

                   1.0

                   0.1]
                        PHA     Grantee 1    Grantee 2     FHA     Atlantic City   Cleveland
                                                Data Source
              13
              12
              11
              10
               9
               8
               7-
               6
                   PHA      Grantee 1     Grantee 2       FHA
                                              Data Source
Atlantic City    Cleveland
Figure 1-9.  Box and Whisker Plots of the Distributions of Window Troughs Dust-Lead
            Loadings and In-Transformed Dust-Lead Loadings by Data Source for the First
            Site Visit
                                             1-12
-------
Table 1-2.  Percentiles (j/g/ft2) of the Clearance Data by Surface and Data Sources for the
           First Site Visit.
f' Surf ace'; ^.
Floors
Sills
Troughs
; Data'Source '
PHA
Grantee 1
Grantee 2
FHA
Atlantic City
Cleveland
PHA
Grantee 1
Grantee 2
FHA
Atlantic City
Cleveland
PHA
Grantee 1
Grantee 2
FHA
Atlantic City
Cleveland
Sample *
Size :-~ •
558
8179
3642
967
554
158
411
4798
2147
668
51
93
371
3002
1346
442
258
39
:""••'. - • •' ,, • •-
s^'SOth-i^
;Percentite:',
48
12
16
48
10
27
28
17
38
52
31
33
42
30
81
375
10
33.3
:^80th'i^
•iPercentileft
111
47
39
187
55
87
80
84
106
271
150
125
85
150
285
2569
70
170
; :•''•' ;9oth Kl:
Percentile .r
200
110
76
418
130
190
175
202
171
714
380
190
167
396
692
5810
190
530
^'PercehtOe^?
312
227
168
871
230
320
281
418
356
1678
660
250
268
937
1527
9150
450
850
1-PercentBeF
600
1030
664
3119
700
440
1175
1870
1568
9054
790
1700
1089
7600
7088
50534
2130
2900
                                           1-13
-------
       1000-
    e
        «0
         1-
                 Al
                                   Wootf
    f
               Al
Figure 1-10. Box and Whisker Plots of the Distributions of Floors Dust-Lead Loadings and In-
            Transformed Dust-Lead Loadings on Dover Data by Substrate for the Passed
            Clearance Visits
                                         1-14
-------
       1000
    I
    £
       100
                                              Vinyl
                                                           Mun*wn
                             Wood
                                                           Aluminum
Figure 1-11. Box and Whisker Plots of the Distributions of Window Sills Dust-Lead Loadings
           and In-Transformed Dust-Lead Loadings on Dover Data by Substrate for the
           Passed Clearance Visits
                                          1-15
-------
    CM


    f
       noao
        101X0
        100
                                             MumkMn
                                                            Otttar
7'
6
5
C 4
S 3
ffc 8'
1-
0
-1-

4














•
i
•
*

I

•

Al Vinyl Abmhun OOMT Unknown
Figure 1-12. Box and Whisker Plots of the Distributions of Window Troughs Dust-Lead
            Loadings and In-Transformed Dust-Lead Loadings on Dover Data by Substrate
            for the Passed Clearance Visits
                                          1-16
-------
       1000-
    £
        100-
         1-
                                              Vh*
                                                            Ottwr
    f  4*
                                             Vlnyt
                                                            OOur
Figure 1-13. Box and Whisker Plots of the Distributions of Floors Dust-Lead Loadings and In-
            Transformed Dust-Lead Loadings on Maryland Data by Substrate for the
            Passed Clearance Visits
                                          1-17
-------

    £
        10
                            Wfcod
                                                               Othsr
                                                 Muntinum
                                                              Otfw
Figure i-14. Box and Whisker Plots of the Distributions of Window Sills Dust-Lead Loadings
            and In-Transformed Dust-Lead Loadings on Maryland  Data by Substrate for the
            Passed Clearance Visits
                                          1-18
-------
       1000
       100

    £
                              Wood
                                             Vinyl
                                                          Akmfemm
                                                         Aunvnum
Figure 1-15. Box and Whisker Plots of the Distributions of Window Troughs Dust-Lead
           Loadings and In-Transformed Dust-Lead Loadings on Maryland Data by
           Substrate for the Passed Clearance Visits
                                         1-19
-------
Table 1-3.
Percentiles (//g/ft2) of the Passed Clearance Data by Surface and Substrate for
Lead Loading Results for Dover and Maryland.
. Source •
Dover
Maryland
Surface
Floors
Sills
Troughs
Floors
Sills
Troughs
Substrate
All
Wood
Vinyl
Unknown
Ail
Wood
Vinyl
Aluminum
Unknown
All
Vinyl
Aluminum
Other
Unknown
All
Wood
Vinyl
Other
Unknown
All
Wood
Vinyl
Aluminum
Other
Unknown
All
Wood
Vinyl
Aluminum
Unknown
>'.- Sample >'s.
- ; Size' "•'•''
620
3
612
5
11
4
1
4
2
260
180
11
2
67
2712
1052
651
20
989
2411
1651
106
44
18
592
1827
373
575
282
597
" ' ,',/• - *•'•'* •„"•! '•%";
.^(Vj-hSOthv-iv- '.
• Percentile "\
5
31
5
9
361
262
212
443
398
7
7
10'
136
7
16
15
18
67
15
21
20
12
11
12
32
50
88
38
38
53
•r.";", ->:-'-'vv-:., .*%:.•
-' '•'•'• 80th ,.'.•.::: /.
rPercentile '•".
43
61
43
72
443
377
212
467
434
59
67
16
239
12
60
58
69
105
56
101
99
37
83
55
119
231
338
163
168
257
; Percentfl*%
61
61
60
96
443
377
212
467
434
111
140
16
239
68
98
88
106
116
98
201
193
66
149
327
238
400
530
308
322
412
f Percentile y
82
61
81
96
467
377
212
467
434
157
169
106
239
89
128
121
135
118
134
295
295
167
234
406
315
571
655
515
456
604

                                         1-20
-------
           £
              1000.00
               100.00
                10.00
                 1.00
                 0.10
                 0.01
                       PHA     Grantee 1    Grantee 2      FHA
                                                Data Source
 Atlantic City   Cleveland
           5  -1

              -2

              -3

              -4
                    PHA       Grantee 1    Grantee 2      FHA
                                              Data Source
Atlantic Cily   Cleveland
Figure 1-16.   Box and Whisker Plots of the Distributions of Floors Dust-Lead Loadings and
               In-Transformed Dust-Lead Loadings by Data Source for the Passed Clearance
               Visits
                                              1-21
-------
              1000.00
               100.00
           CM
           <
               10.00
           £
                1.00
                0.10
                0.01
                       PHA      Grantee 1   Grantee 2      FHA
                                               Data Source
                                                      Atlantic City   Cleveland
               7

               6

               5

          ST   4

          §   3

          •5   2

          n   1

               0

              -1

              -2

              —3-
*»•
I
                    PHA      Grantee 1    Grantee 2      FHA
                                             Data Source
                                                     Atlantic City    Cleveland
Figure 1-17.   Box and Whisker Plots of the Distributions of Window Sills Dust-Lead
              Loadings and In-Transformed Dust-Lead Loadings by Data Source for the
              Passed Clearance Visits

                                             I-22
-------
          CM
          <
          £
             1000.0
              100.0
               10.0
                1.0
                0.1
                     PHA      Grantee 1   Grantee 2      FHA
                                             Data Source
 Aflantte City   Cleveland
               7

               6

               51

          sr   4
          <


          I*   2

          £   1


              -1

             -2-

             -3-
                   PHA      Grantee 1    Grantee 2      FHA
                                            Data Source
Atlantic CIV   Cleveland
Figure 1-18.   Box and Whisker Plots of the Distributions of Window Troughs Dust-Lead
              Loadings and in-Transformed Dust-Lead Loadings by Data Source for the
              Passed Clearance Visits
                                            I-23
-------
Table 1-4.   Percentiles (/ig/ft2) of the Passed Clearance Visits by Surface and Data Sources
            for Lead Loading Results.
Surface
Floors
Sills
Troughs
Data Source
PHA
Grantee 1
Grantee 2
FHA
Atlantic City
Cleveland
PHA
Grantee 1
Grantee 2
FHA
Atlantic City
Cleveland
PHA
Grantee 1
Grantee 2
FHA
Atlantic City
Cleveland
Sample
Size
554
8345
3749
997
545
161
412
4906
2170
689
46
93
371
3036
1341
423
267
39
50th
Percentile
43
12
15
40
10
22
27
17
36
47
30
32
41
27
68
169
10
33
80th
Percentile
93
39
34
102
46
64
68
72
99
164
90
120
81
116
198
431
66
170
90th
Percentile
118
70
57
141
80
110
150
141
145
263
180
170
138
237
346
598
170
490
95th
Percentile
148
108
90
165
120
134
222
236
215
358
270
230
213
389
512
678
320
540
99th
. Percentile
185
173
163
191
170
182
378
410
372
481
380
388
399
691
725
760
650
648
                                          (-24
-------
         APPENDIX J
Comparison of Clearance Guidance
             J-1
-------
                                    APPENDIX J

                    COMPARISON OF CLEARANCE GUIDANCE


      The following is a comparison of the guidance on clearance in the 1990 HUD Interim

Guidelines [2], the 1995 EPA 403 Guidance [5], the 1995 HUD Guidelines [3], the 1996 EPA

402/404 Rule [4], the 1999 HUD 1012/1013 Final Rule [22], and the 2001 EPA 403 Final Rule

[1]. Table J-l presents a comparison of the number and location of samples for single sample

and composite sample testing in contained and non-contained areas. Clearance testing

procedures for visual inspections are shown in Table J-2, procedures for sealants are in Table J-3,

and waiting times for conducting clearance testing are compared in Table J-4.  Clearance

standards are outlined in Table J-5, failure procedures in Table J-6, a comparison of the units in

which to report samples is presented in Table J-7, compositing rules given in Table J-8, and

conflict of interest issues are compared in Table J-9.
      Below are some of the major changes that have taken place in clearance standards over

time.
      •   1990 HUD Interim Guidelines [2] clearance standards are 200 ng/ft2 for bare floors,
          500 ug/ft2 for window sills, and 800 ug/ft2 for window wells (troughs).

      •   1994 EPA 403 Guidance [5] clearance standards are 100 jig/ft2 for bare floors, 500
          ug/ft2 for interior window sills, and 800 ng/ft2 for exterior window sills (troughs)-and
          exterior horizontal surfaces.

      •   1995 EPA 403 Guidance was the same as  1994 EPA 403 Guidance, but was
          disseminated as a Federal Register Notice [5].

      •   1995 HUD Guidelines [3] clearance standards are 100 ug/ft2 for floors (including
          carpeted and uncarpeted floors), 500 ug/ft2 for window sills, and 800  ug/ft2 for
          window wells (troughs) and exterior concrete or other rough surfaces.

      •   1996 EPA 402/404 Rule [4] refers to the clearance levels in the EPA Guidance on
          Residential Lead-Based Paint, Lead-Contaminated Dust and Lead Contaminated Soil
          or other equivalent guidelines.

      •   1999 HUD 1012/1013 Final Rule [22] dust-lead clearance standards are 40 ^ig/ft2 for
          floors (including carpeted and uncarpeted  interior floors), 250 ^ig/ft2 for interior
         window sills, and  800 ug/ft2 for window troughs.
                                         J-2
-------
•   2001 EPA 403 Final Rule [1] dust-lead clearance standards are 40 ug/ft2 for floors
    (including carpeted and uncarpeted floors), 250 ug/ft2 for interior window sills, and
    400 ug/ft2 for window troughs.
                                    J-3
-------
£
 Q.
 E
 (9
V)
 (0

 £

 o
O
•o

 CO
 0)
J»
 Q.

 CO
_
o>
c
CO
**

CO
•o

CO
CO  m
5  «
o
flu
  o
  «=
   CO
•^ 2
0 c
c o
1 =
i§
Is
       CO
3
         u
         it

   «s-S-
   I 2 S £
   E 
-------
rlt
;-"-!V5^?(
-$$$!

Jilt
~ '^ySfeifl

vj'j-IS^IB
ijyploa
>fl
•i«j'<*jjc-.'.
18
i.jili;
'$£1 ip^
Mm
s, aj
Its
i*6 is
'If «S
•Ui
Hi
ft;*
I
If
IS
:« 3o
w ;:z
i S-
•'*<, '?£•'
11
if
5 St
III
J-> *-i- f"
Hi
'^ft;fe
t^i^-^"
H
c
a
•& S
a> I
3 jg
» a.
°> c _
— C T3
Q. CO 03
£ W £
Sol
0 .E -2
Z W -D

^
c

"a «
15
e ~
'» s S
lil
li|
0 =6 -D

J£
O
(0 o
°T3 1
CAO a>
|| |
ia »
S-D "s
.£•3 i
= | |
viost of the clearance
sampling should be
:onducted within the
;ontainment area.

Sj <»
£§
Sf
ME





















o .£ -i a ** o .S
SI IE « SI
s| 5* 1=1
*& I'l 5i5
2 ? Is -i S ?
I to •£ c.l =•-£ « ^
+-'55 w w S »*-'» w
c a? w o ,2 ccv
|l|Jii ||i|l
.Is »Il
o o £ .5 (- £ "K c
£ .ti 5 -— o CB B1 "C
"|| 2 llfS 5
S-8-o-g « a 8 | S 'g
= 2 § 5 I •§ ? ? 1 e
**

























"5 >^ *
Q> ^ <-
t E <»»;•£
1 Si tllil
2 IS SjJ|*-s
m *2 ^ScSw
Q.^^ (Oq> "colOirjO
§ o fi * |« «£ •
i! 11 ?i!«|
|5 .08 o^S.f»
!!
"5.-o -c-c CBCBI
Ibl £ I e 8 S -8
I I!|i f! 11,1 ififl
^ -55^0 S° vs.2'0 £„»£«
^ o.*"^eo(B *- ^^=c n"*Offl
•i SiSJt'l as8'" l^t!
.2 '1§^|S1.£ .2.11.0 fi-S-l-i
fit 17! ill Ullillllii
g^g •«-2£5^'-'(B ^oa^S^cSSioi





O
Z
a
                                                J-5
-------
g
 to



 o
U
•o

 re
 I
 (0
w
35
 0  •
 E=B
 3 S>
 » 2
 o .£
*., »
12 «
C 0
0
   «
3 |

O §
£ o

=
   -o

« §
c -
« g
Q o
ll
0

3
(B
        .1:1*
              9
£ a. ..  -a
        o> 3

        ^ O
                      o>  _
                      o> *-  o .E
                                       J-6
-------
 w
 o
 a.

 E
 o
O
 a
  ° **
f -n ™ O> OS

i  § £1.1.
g  ° £E £
o S § S S
S  > 3 ~ £•£   s 5 | «•=

*l«Sl-s  . sS'i|£
.g  ? .g — J •£ 10^ oc-  |
o> < i
O> DL 3
•- ui tr
                                                      (O ^ >
                                                      o •& —
                                                      05 — co
                                       c * CN
                                       .£ o    *-

                                       g I i ? ?
                                       E §  o '•& =5.
                                       g «  g-E E
                                                       E

                                                       ?
                                                      '•=>-«
                                                       9--Q 7i
                                                       111   I
                                                       o « w° "
                                                         o ag
                                                         oca
                                         ""S?^-8S

                                       S€sfslls
                                       •g g -a  I -5 -S I a
                                       _. S *-  * an   * ?t
                                      •S w '«
                                       W oj AV)0Ca>cO'Q

                                                                      isT-cis w -"IE  s>- =
                                                                   	w^^w^gWTSWWflj

                                                                 5 5£ 9-®^ nia^J-S
                                                           Oe-—  'Cc"-Sco)i3c5
                                                           S E -^ •;:  .wES«i--r-*m —
                                                               x:
                                                                                ««
                                                                      c ~ g c S  .o>c -og

                                                                 1111 511| 1111 n
                                                                 O (0 A "O ic x   O CQ £ O CQ
                                                                 O *-* O   JJQ*^ r>*j Ort**
                                                                 c wu*-%=o=.£:«i>uc »
                                      ^1
                                      UJ|

                                      o "«
                                      o .S
                                      CM u.
                                               J-7
-------
Table J-2. Comparison of Procedures for Visual Inspections from HUD and EPA Clearance
          Guidance and Rule Documents.
Clearance Guidance Document 'i
-. *7_--l~ .a* .. ^""V/JTr •>.'*'*' "-." '-"-'-I'- 'VJ" > •" ""
1 990 HUD Interim Guidelines [2]
1995 EPA 403 Guidance [5]
1995 HUD Guidelines [3]
1999 HUD 101 2/1 01 3 Final
Rule [22]
1996 EPA 402/404 Rule [4]
2001 EPA 403 Rnal Rule [1]
KMi^ftfC^
Perform visual inspection after abatement but before repainting.
Surface dust sampling should not be conducted if there is a visible
accumulation of dust or debris
All interior rooms or areas and exterior areas should be visually clean
before collecting dust samples. If this is not the case, clean the rooms
and areas before starting dust collection for clearance testing
Room by room while environmental samples taken.
Prior to repainting
No evidence of settled dust
The visual assessment shall be performed to determine if deteriorated
paint surfaces and/or visible amounts of dust, debris, paint chips or other
residue are still present. Both exterior and interior painted surfaces shall
be examined for the presence of deteriorated paint. If deteriorated paint
or visible dust, debris or residue are present in areas subject to dust
sampling, they must be eliminated prior to the continuation of the
clearance examination, except elimination of deteriorated paint is not
required if it has been determined, through paint testing or a lead-based
paint inspection, that the deteriorated paint is not lead-based paint. If
exterior painted surfaces have been disturbed by the hazard reduction,
maintenance or rehabilitation activity, the visual assessment shall include
an assessment of the ground and any outdoor living areas close to the
affected exterior painted surfaces. Visible dust or debris in living areas
shall be cleaned up and visible paint chips on the ground shall be
removed.
A visual inspection shall be performed to determine if deteriorated
painted surfaces and/or visible amounts of dust, debris or residue are still
present. If deteriorated painted surfaces or visible amounts of dust,
debris or residue are present, these conditions must be eliminated prior to
the continuation of the clearance procedures.
Addressed in the 1996 EPA 402/404 Rule [4], 40 CFR §745.227 (e)(8)(i).
                                      J-8
-------
Table J-3.  Comparison of Procedures for Sealants from HUD and EPA Clearance Guidance
            and Rule Documents.
•:-Ji*W«i; J..-. * • "-. ••*V^!-3:.'."^I->>-'_V''V^^*ii"_v'. »•"• •«-• ; • -' 'Vi jV^-i^V"
|^C(eara^;-Gu^iweDpaimentjg^
1990 HUD Interim Guidelines [2]
1995 EPA 403 Guidance [5]
1995 HUD Guidelines [3]
1999 HUD 1012/1013 Rnal Rule [22]
1996 EPA 402/404 Rule [4]
2001 EPA 403 Final Rule [1]

Painting or otherwise sealing abated surfaces and all interior
floors is the next step of cleaning process.
All abated surfaces, including walls, ceilings, and woodwork,
should be primed with an appropriate primer. All applicable areas
may then be repainted. Wooden floors, vinyl tile, linoleum, and
concrete floors should be sealed.
Document does not address the use of sealants for clearance.
Seal floors before clearance testing
Document does not address the use of sealants for clearance.
Document does not address the use of sealants for clearance.
Document does not address the use of sealants for clearance.
Table J-4.  Comparison of Procedures for Waiting Time from HUD and EPA Clearance
            Guidance and Rule Documents.
  1990 HUD Interim Guidelines 12}
Dust sampling should take place no sooner than 24 hr. after
completion of post-abatement cleanup activities.
  1995 EPA 403 Guidance [5]
Sampling of dust should take place at least one hour after
completion of all abatement and interim control work, including
cleanup.
  1995 HUD Guidelines [3]
Wait one hour after cleaning.
  1999 HUD 1012/1013 Final Rule
  [22]
In accordance with 40 CFR 745.227(e)(8). This is part of the 1996
EPA 402/404 Rule (41.
  1996 EPA 402/404 Rule [4]
Dust samples for clearance purposes shall be taken a minimum of 1
hour after completion of final post-abatement cleanup activities.
  2001 EPA 403 Final Ruled]
Addressed in the 1996 EPA 402/404 Rule [4], 40 CFR §745.227
                                            J-9
-------
Table J-5.  Comparison of Clearance Standards from HUD and EPA Clearance Guidance
            and Rule Documents.
^-Clearance Guidance Document ^|e
1 990 HUD Interim Guidelines [2]
1995 EPA 403 Guidance [5]
1995 HUD Guidelines (3}
1996 EPA 402/404 Rule [4]«
1999 HUD 1012/1013 Final Rule
[221
2001 EPA 403 Rnal Rule [1]

200 //g/ft2 for bare floors
500 jt/g/ft2 for window sills
800 //g/ft* for window wells (troughs)
100 j/g/ft2 for bare floors
500 //g/ft2 for interior window sills
800 jug/ft2 for exterior window sills (troughs) and exterior horizontal
surfaces
1 00 //g/ft2 for floors (including carpeted and uncarpeted floors)
500 //g/ft2 for window sills,
800 j/g/ft* for window wells (troughs) and exterior concrete or
other rough surfaces.
Clearance levels which are appropriate for the purposes of this
section may be found in the EPA Guidance on Residential Lead-
Based Paint, Lead-Contaminated Dust and Lead Contaminated Soil
or other equivalent guidelines. (5, 23]
40 jug/ft2 for floors (carpeted or uncarpeted interior floors),
250 fjg/it2 for interior window sills,
800 #g/ft2 for window troughs
40 A/g/ft2 for floors (including carpeted and uncarpeted floors),
250 //g/ft2 for interior window sills,
400 //g/ft2 for window troughs.
   1998 EPA 403 Proposed Rule [24] included proposed amendments to the final Section 402 rules for post-
   abatement dust clearance standards: 50 i/g/ft2 for uncarpeted floors, 250 //g/ft2 for interior window sills,
   and 800 /sg/ft2 for window troughs.
                                            J-10
-------
Table J-6.   Comparison of Failure Procedures from HUD and EPA Clearance  Guidance and
               Rule Documents.
   1990 HUD Interim
   Guidelines [2]
If any of the residual lead dust level results exceeds the clearance criteria, the area must be
cleaned again and re-tested until the criteria are met.  All areas must pass clearance to be re-
occupied.
   1995 EPA 403
   Guidance [5]
Samples above or equal to the appropriate standard have failed clearance, and all rooms or areas
represented by those samples are said to have failed.  For samples that have failed, the
components represented by those samples (floors, interior window sills, exterior window sills,
exterior horizontal surfaces, or interior areas outside a containment area) must be recleaned and
retested. The process continues until clearance is obtained for all components.
   1995 HUD Guidelines
   I3J
If the dust sample for any surface is above the standard, all similar surfaces in the dwelling that
sample represents (e.g., all interior sills or floor) should be re-cleaned and re-tested. Only the
similar component needs to be re-cleaned. If a surface fails twice additional LHC measures
and/or further sealing should be considered.
   1999 HUD
   101 2/1 01 3 Final Rule
   [22]
All surfaces represented by a failed clearance sample shall be recleaned or treated by hazard
reduction, and retested, until the applicable clearance level is met.
   1996 EPA 402/404
   Rule (4]
If the residual lead levels in a dust sample exceed the clearance levels, all the components
represented by the failed sample shall be re-cleaned or treated by lead hazard reduction and re-
tested until clearance levels are met.
  2001 EPA 403 Final
  Ruled]
If a property fails clearance, it must be recleaned until it passes, although it is not automatically
necessary to reclean the entire property when clearance fails, such as when some of the visual
and dust-testing clearance results have indicated that portions of the property are already
clearned.
                                                      J-11
-------
Table J-7.   Comparison of Concentrations versus Loadings from HUD and EPA Clearance
               Guidance and Rule Documents.
                    nents.


£||fe£|||^

~     -     •  .      ling method which is currently recommended is surface wipe sampling
   iCtoarance Guidance
  '     '
  1990 HUD Interim
  Guidelines 1 2]
The surface dust sampling method which is currently recommended is surface wipe sampling
(i.e., loadings).
  1995 EPA 403
  Guidance [5]
The basic method for collecting dust clearance samples is the wipe method (i.e., loadings).
Other dust collection methods may be used provided the user establishes comparability to the
wipe method and is responsible for providing comparable standards for clearance.
Note that when assessing multiple sources of lead, dust lead concentration may be a more
appropriate measurement.  The utility of concentration measurements for identifying hazardous
room dust will be further considered in the development of section 403 rule making
  1995 HUD Guidelines
  [3]
Until the EPA standards and protocols are established, wipe sampling (loadings) should be
performed on all surfaces.  While vacuum samples (concentrations) can be collected, neither
HUD nor EPA can provide standards to interpret vacuum sampling results at this time.  Until
vacuum sampling standards have been established, wipe sampling is the preferred method.
  1999 HUD 1012/1013
  Final Rule [22].
Standards are reported as a loading for interior samples.
  1996 EPA 402/404
  Rule [4]
The type of dust samples to be collected, wipe (loading) or vacuum (concentration) are not
explicitly defined in this document.
  2001 EPA 403 Final
  Ruled]
Dust-lead hazard and clearance standards are set as in loadings. In 40 CFR 5745.65, it states
"A dust-lead hazard is surface dust in a residential dwelling or child-occupied facility that
contains a rnass-per-area concentration of lead equal to or exceeding 40 //g/ft2 on floors or
250 /ig/ft2 on interior window sills based on wipe samples."  Dust-lead clearance standards are
specified in 40  CFR §745.227 (e)(8)(viii) which were shown  in Table J-5 [41.
                                                    J-12
-------
Table J-8.   Comparison of Compositing Rules from HUD and EPA Clearance Guidance and
             Rule Documents.
^Clearance Guidance a
-r*3*3ii-'£--'*!*--- • -'"'< - -•: •-, •**:*•
ijsyKDocument^; -:•-<**
1990 HUD Interim
Guidelines [2!
1994/1 995 EPA 403
Guidance [5]
1995 HUD Guidelines
[31
1999 HUD
101 2/1 01 3 Rnal Rule
[221
1996 EPA 402/404
Rule [41*
2001 EPA 403 Final
Rule [11
||l^^!^a^^
Compositing was not addressed in this document.
Compositing was not addressed in this document.
Samples are composited in the field. Separate samples are required from carpeted and hard
surfaces; and from component sampled; and from each dwelling floor surface areas sampled in
each room should be the same size. Interior sill and well sampling sizes are dependent on
window characteristics, but should be similar from room to room; all subsamples should be
inserted into the same tube. No more than four different wipes should be inserted into a single
container.
Dust samples shall be collected and analyzed in accordance with standards established either by
a State or Indian tribe under a program authorized by EPA in accordance with 40 CFR part 745,
subpart Q, or by the EPA in accordance with 40 CFR 745.227 (both are part of the 1996 EPA
402/404 Rule 14]).
Composites expressly permitted for clearance testing. Composite dust samples consist of at
least two subsamples. Every component that is being tested shall be included in the sampling.
Composite dust samples shall not consist of subsamples from more than one type of component.
A composite sample may contain from two to four subsamples of the same area as each other
and of each single surface sample in the composite [40 CFR §745.63]
    1998 EPA 403 Proposed Rule [24] included proposed amendments to the final Section 402 rules includes requiring the
    risk assessor to compare the composite sample to the clearance standard divided by the number of subsamples in the
    composite.
                                            J-13
-------
Table J-9.  Comparison of Conflict of Interest Issues from HUD and EPA Clearance
          Guidance and Rule Documents.
-[Clearance Guidance •:;;
f^i^*hte::f
1990 HUD Interim
Guidelines [2]
1994/1 995 EPA 403
Guidance [51
1995 HUD Guidelines
[3]
1999 HUD 101 2/1 01 3
Final Rule [22]
1996 EPA 402/404
Rule (41
2001 EPA 403 Final
Ruled)
. ffi.\ :r> V * ; - ^r'> ':,-*-. (•:• i^f. . .* ., '"'' .'•., ; fc; - „• • "„• •• — ;1: • ?5a.J^ *" ' "V'i *•-.' '-; •" - •.•>".,--' -•'~.:r .'V - '.', .•„•>•'•. -^!t' '>:• . ' * '-,--' ™%V?*<*jB2s:?V3'*'0"*'^!''t ;«*»''
To avoid potential conflict of interest, the abatement contractor should not conduct the final
inspection. This should be done by a qualified inspector, industrial hygienist, or local public
health official.
Clearance testing should be conducted by an organization that is independent of the
organization that completed the abatement or interim controls.
Inspectors should be independent of abatement contractor
Clearance examinations shall be performed by persons or entities independent of those
performing hazard reduction or maintenance activities, unless the designated party uses
qualified in-house employees to conduct clearance. An in-house employee shall not conduct
both a hazard reduction or maintenance activity and its clearance examination.
EPA requested comment on whether to preclude individuals or firms conducting abatement
activities from performing inspection and risk assessment activities, and from performing
clearance procedures following an abatement. Although many public commentators supported
a requirement that inspection, risk assessment and clearance procedures be conducted by
individuals and firms independent of the individuals and firms conducting abatements, the final
rule does not include such a requirement. Some of the reasons for not supporting a conflict-of-
interest requirement were that the potential convenience and cost savings of hiring one firm, as
opposed to two or three firms should not be denied a property owner. The Agency also noted
that there may be instances in which, due to a regional scarcity of lead-based paint
professionals, it may be cost prohibitive or logistically difficult for a building owner to hire two
different companies. Nonetheless, the Agency believes that parties involved in lead-based paint
activities should avoid situations of potential conflict of interest.
Addressed in the 1996 EPA 402/404 Rule [4].
                                      J-14
-------
                     REPORT DOCUMENTATION PAGE
                                                     Form Approved
                                                     OMB No 0704-0188
 Sources, gathering and maintaining die data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other
 aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for information Operations and
 Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (07044188),
 Washington, DC 20S03.
  1.  AGENCY USE ONLY (Leave blank)
   2. REPORT.DATE
     December 2001
3. REPORT TYPE AND DATES COVERED
  Final Report
 4. TITLE AND SUBTITLE

  .  Analysis of Lead Dust Clearance Testing
  6. AUTHOR(s)
    Bradley Skarpness, Ying-Liang Chou, and Warren Strauss
                                          5. FUNDING NUMBERS
                                                                             C: 68-D5-0008,68-W-99-033
 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 ADDRESSfES)

   U.S. Environmental Protection Agency
   Office of Pollution Prevention and Toxics
   1200 Pennsylvania Avenue NW (7401)
   Washington, D.C. 20460
                                          10. SPONSORING/MONITORING AGENCY
                                             REPORT NUMBER

                                              EPA 747-R-01-005
  11. SUPPLEMENTARY NOTES
    Other Battelle staff involved in the production of this report included Jennifer Holdcraft, Pamela Hartford, and Matt
    Palmgren.
  12.a DISTRIBUTON/AVAILABILITY STATEMENT
                                          12b. DISTRIBUTION CODE
  13. ABSTRACT (Maximum 200 words)

  This report presented results from lead clearance testing activities that occurred between 1989 and 1999 under the 1990
  HUD Interim Guidelines, which set dust lead clearance standards at 200, 500, and 800 ng/ft2 for floors, window sills, and
  window troughs, respectively.  Dust lead clearance results were obtained from the HUD's FHA and PHA Demo Studies,
  and Grantee Program; the Maryland Department of the Environment; the Atlantic City and Dover Housing Authorities; and
  the Cleveland Lead Hazard Abatement Center. Over 90% of individual clearance samples passed clearance. Only 67% of
  the housing units passed  clearance on the first site visit. Eventually 87% of the housing units were known to have passed
  clearance. Floor samples had a smaller geometric mean lead loadings than sills, which in turn had a smaller geometric
  mean than troughs. Geometric mean lead loadings generally increased from the first site visit to the third site visit. The
  pairwise correlations between components during the first site visit were positive and significant.  The simulated composite
  samples analysis results indicated mat composite sampling is associated with a decrease of sensitivity if compared directly
  to the standards. Comparison of composite samples  to two lower standards resulted in an increase in sensitivity.
  14. SUBJECT TERMS
 Clearance Testing, Clearance Standards, Lead Loadings, Floors, Window Sills, Window
 Troughs, Composite Samples, Geometric Mean, Performance Characteristics (Sensitivity,
 Specificity, PPV, NPV), Correlation, Conditional Probability, Logistic Regression.
                                                      15. NUMBER OF PAGES
                                                         374
                                                      16. PRICE CODE
  17. 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
-------
-------