University of Southern California Dissertations and Theses

Lead exposure and health outcomes in older adults

(USC Thesis Other)

Lead exposure and health outcomes in older adults

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Lead Exposure and Health Outcomes in Older Adults
by
Kate N. Vavra-Musser
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(POPULATION, HEALTH, AND PLACE)
August 2023
Copyright 2023 Kate N. Vavra-Musser
I dedicate this thesis to my parents, for their unwavering love and support.
And to my grandparents Geraldine, Janet, Lee, and Richard.
And my to nieces and nephew Lilla Mai, Freya, Rowan, Atticus, and Baby TBD.
ii
Acknowledgements
I would like to thank my co-advisors Dr. Jennifer Ailshire and Dr. An-Min Wu and my committee
members Dr. Meredith Franklin and Dr. Eileen Crimmins; my amazing PhD colleagues and
friends Avery, Dan, Emily, Johanna, Karina, Li, Lois, Margarita, Nathan, Rachel, Vania, William,
Xiaozhe, and Yan. I would like to give a very special thank you to Dr. Calley Fisk. Finally, I would
like to thank my family, most of all my amazing parents.
iii
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables viii
List of Figures x
Chapter 1: Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 History of Lead Use and Regulation in the United States . . . . . . . . . . 2
1.1.2 Lead Exposure and Human Health . . . . . . . . . . . . . . . . . . . . . . 3
1.1.3 Lead and Older Adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Dissertation Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Chapter 2: Quantification of Residential Lead-Based Paint Presence Using Household
Characteristics 9
2.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.1 Residential Lead-Based Paint Use and Risk in the United States . . . . . . 13
2.3.2 Lead Biokinetics and Toxicity . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.3 Lead and Older Adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.4 Project Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.1.1 Paint Chip Data from the California Department of Toxic Sub-
stances Control (DTSC) . . . . . . . . . . . . . . . . . . . . . . 16
2.4.1.2 American Community Survey (AHS) . . . . . . . . . . . . . . . 16
2.4.2 Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.2.1 Parcel Lead-Based Paint . . . . . . . . . . . . . . . . . . . . . . 17
2.4.2.2 Parcel Building Construction Year . . . . . . . . . . . . . . . . 17
2.4.2.3 Parcel Land Use . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.2.4 Parcel Housing Type . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.2.5 Household Composition . . . . . . . . . . . . . . . . . . . . . . 19
2.4.2.6 Householder Race . . . . . . . . . . . . . . . . . . . . . . . . . 19
iv
2.4.2.7 Peeling Paint . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.3 Data Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.3.1 Parcel Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.3.2 Household Data . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.4 Analysis Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.4.1 Multivariable Analysis . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.4.2 Exposure Risk Estimation for US Older Adults . . . . . . . . . . 21
2.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5.1 Parcel Analysis Sample Characteristics . . . . . . . . . . . . . . . . . . . 22
2.5.2 Lead-Based Paint Presence . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.3 Household Analysis Sample Characteristics . . . . . . . . . . . . . . . . . 24
2.5.4 Multivariate Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5.5 Exposure Risk Estimation for US Older Adults . . . . . . . . . . . . . . . 27
2.5.5.1 Exposure Risk and Construction Year . . . . . . . . . . . . . . . 27
2.5.5.2 Exposure Risk and Household Composition . . . . . . . . . . . 27
2.5.5.3 Exposure Risk and Householder Race . . . . . . . . . . . . . . 28
2.5.5.4 Exposure Risk and Peeling Paint . . . . . . . . . . . . . . . . . 28
2.5.6 Comparison of Model Results with Year-Threshold Risk Frameworks . . . 29
2.6 Conclusions and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.6.1 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Chapter 3: Demographic and Socioeconomic Drivers of Airborne Lead Exposure Risk
in Older Adults 41
3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.1.1 Individual and Neighborhood Characteristics as Determinants of Environ-
mental Exposure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.1.2 Project Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2.1.1 National Air Toxics Assessment (NATA) . . . . . . . . . . . . . 45
3.2.1.2 Health and Retirement Study (HRS) . . . . . . . . . . . . . . . 46
3.2.1.3 Decennial Census and American Community Survey (ACS) . . . 46
3.2.2 Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2.2.1 Tract of Residence Lead Exposure Risk . . . . . . . . . . . . . . 47
3.2.2.2 Covariates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2.3 Data Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.4 Analysis Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.4.1 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.4.2 Bivariate Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2.4.3 Multivariable Analysis . . . . . . . . . . . . . . . . . . . . . . . 53
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.1 Study Population Characteristics . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.2 Bivariate Analyses with Tract of Residence Lead Exposure Quartile . . . . 55
3.3.2.1 Individual Demographic Characteristics . . . . . . . . . . . . . 55
3.3.2.2 Individual Socioeconomic Characteristics . . . . . . . . . . . . 57
v
3.3.2.3 Tract Characteristics . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3.3 Ancillary Analysis
Individual Race/Ethnicity Stratified by Tract Urbanicity . . . . . . . . . . 59
3.3.4 Multivariable Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.4 Conclusions and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.4.1 Summary of Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.4.1.1 Individual Race/Ethnicity Interaction with Tract Urbanicity . . . 62
3.4.2 Comparison with Previous Literature . . . . . . . . . . . . . . . . . . . . 63
3.4.3 Implications and Significance . . . . . . . . . . . . . . . . . . . . . . . . 63
3.4.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Chapter 4: Estimated Neighborhood Airborne Lead Exposure Risk and Cognitive Func-
tioning in Older Adults 65
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2.1 Lead Biokinetics and Toxicity . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2.2 Lead and Older Adults Literature Review . . . . . . . . . . . . . . . . . . 68
4.2.3 Project Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3.1.1 National Air Toxics Assessment (NATA) . . . . . . . . . . . . . 74
4.3.1.2 Health and Retirement Study (HRS) . . . . . . . . . . . . . . . 75
4.3.1.3 Decennial Census and American Community Survey (ACS) . . . 75
4.3.2 Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.2.1 Cognitive Functioning . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.2.2 Tract of Residence Lead Exposure Risk . . . . . . . . . . . . . . 77
4.3.2.3 Covariates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.3.3 Data Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.3.4 Analysis Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.3.4.1 Descriptive Statistics and Bivariate Analyses . . . . . . . . . . . 82
4.3.5 Multivariable Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.3.5.1 Ancillary and Supplemental Analyses . . . . . . . . . . . . . . . 83
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4.1 Study Population Characteristics . . . . . . . . . . . . . . . . . . . . . . . 84
4.4.2 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4.2.1 Cognitive Functioning . . . . . . . . . . . . . . . . . . . . . . . 84
4.4.2.2 Tract of Residence Lead Exposure . . . . . . . . . . . . . . . . 84
4.4.3 Bivariate Analyses with Tract of Residence Lead Exposure Quartile . . . . 86
4.4.4 Multivariable Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.4.5 Ancillary Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.4.5.1 Race/Ethnicity . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.4.5.2 Urbanicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.5 Conclusions and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.5.1 Summary of Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.5.1.1 Race/Ethnicity Ancillary Analysis . . . . . . . . . . . . . . . . 97
vi
4.5.2 Comparison with Previous Literature . . . . . . . . . . . . . . . . . . . . 97
4.5.3 Implications and Significance . . . . . . . . . . . . . . . . . . . . . . . . 98
4.5.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Chapter 5: Conclusion 101
References 104
Appendices 119
A Housing Type Categories and Corresponding Categories used in the 2019 Ameri-
can Community Survey (AHS) and by Los Angeles County . . . . . . . . . . . . . 120
B Categorization of Household Composition Types from the 2019 American Com-
munity Survey (AHS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
C Lead-Based Paint (LBP) Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
D High Concentration Lead-Based Paint (H-LBP) Models . . . . . . . . . . . . . . . 123
E Very High Concentration Lead-Based Paint (VH-LBP) Models . . . . . . . . . . . 124
F Frequency of Parcel Building Construction Year 1960-2009 . . . . . . . . . . . . . 125
G Previous Work Linking NATA Exposure Data with Human Health Outcomes . . . 126
H Range of Log Lead Exposure for Lead Exposure Risk Quartile Groups . . . . . . . 127
I Race and Ethnicity Variable Recode Table . . . . . . . . . . . . . . . . . . . . . . 128
J Urbanicity Variable Recode Table . . . . . . . . . . . . . . . . . . . . . . . . . . 129
K Census Regions and Divisions of the United States . . . . . . . . . . . . . . . . . 130
L Smoking Status Variable Recode Table . . . . . . . . . . . . . . . . . . . . . . . . 131
M Model Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
N Chapter 3 Supplemental Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
O Chapter 3 Models with Alternative Explanatory Variable . . . . . . . . . . . . . . 136
vii
List of Tables
2.1 Variables Used in this Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Parcel Lead-Based Paint Presence Definitions . . . . . . . . . . . . . . . . . . . . 17
2.3 Parcel Sample Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Parcel Paint Presence Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Paint Presence by Parcel Characteristics . . . . . . . . . . . . . . . . . . . . . . . 24
2.6 Paint Presence by Household Characteristics . . . . . . . . . . . . . . . . . . . . . 25
2.7 Multivariate Marginal Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.8 Estimated Older Adult Population at Risk of Lead-Based Paint . . . . . . . . . . . 28
2.9 Risk Estimation by Householder Race . . . . . . . . . . . . . . . . . . . . . . . . 28
2.10 Risk Estimation by Peeling Paint . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1 Variables Used in this Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Summary Statistics for Lead Exposure Quartile Groups . . . . . . . . . . . . . . . 48
3.3 Demographic characteristics of the study population . . . . . . . . . . . . . . . . . 54
3.4 Demographic and tract characteristics by lead exposure quartile and results of tests
of bivariate association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.5 Logistic Regression Model Results . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1 Variables Used in this Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2 Summary Statistics for Lead Exposure Quartile Groups . . . . . . . . . . . . . . . 77
4.3 Characteristics of Project B Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.4 Bivariate Correlations for Project B . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5 Project B Model Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.6 Project B Race/Ethnicity Stratified Model Results . . . . . . . . . . . . . . . . . . 92
viii
4.7 Project B Urbanicity Stratified Model Results Part I . . . . . . . . . . . . . . . . . 94
4.8 Project B Urbanicity Stratified Model Results Part II . . . . . . . . . . . . . . . . 95
5.1 Housing Type Categorization Schema . . . . . . . . . . . . . . . . . . . . . . . . 120
5.2 Housing Type Categorization Schema . . . . . . . . . . . . . . . . . . . . . . . . 121
5.3 Lead-Based Paint (LBP) Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.4 High Concentration Lead-Based Paint (LBP) Models . . . . . . . . . . . . . . . . 123
5.5 Very High Concentration Lead-Based Paint (LBP) Models . . . . . . . . . . . . . 124
5.6 Race and Ethnicity Variable Recode Table . . . . . . . . . . . . . . . . . . . . . . 128
5.7 Urbanicity Variable Recode Table . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.8 Smoking Status Variable Recode Table . . . . . . . . . . . . . . . . . . . . . . . . 131
5.9 Chapter 3 Supplemental Models (Part I) . . . . . . . . . . . . . . . . . . . . . . . 134
5.10 Chapter 3 Supplemental Models (Part II) . . . . . . . . . . . . . . . . . . . . . . . 135
5.11 Chapter 3 Supplemental Models (Part III) . . . . . . . . . . . . . . . . . . . . . . 137
5.12 Chapter 3 Alternative M2 Models with Alternative Explanatory Variable . . . . . . 138
5.13 Chapter 3 Alternative M5 Models with Alternative Explanatory Variable . . . . . . 139
ix
List of Figures
1.1 History of Lead Use in the United States . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 History of Lead Use and Regulation in the United States and Older Adult Life
Trajectory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Histogram of Building Construction Year . . . . . . . . . . . . . . . . . . . . . . 35
2.2 Project A Parcel Data Restriction Process . . . . . . . . . . . . . . . . . . . . . . 36
2.3 Project A Household Data Restriction Process . . . . . . . . . . . . . . . . . . . . 37
2.4 Predicted Likelihood of Lead-Based Paint Presence by Parcel Characteristics . . . 38
2.5 Risk Estimation by Householder Race . . . . . . . . . . . . . . . . . . . . . . . . 39
2.6 Risk Estimation Framework Comparison . . . . . . . . . . . . . . . . . . . . . . . 40
3.1 Lead Emissions from All Sources Exposure Risk Quartile . . . . . . . . . . . . . . 48
3.2 Project C Data Restriction Process . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3 Observed vs Expected Likelihood of Living in a Q1-3 vs Q4 Lead Exposure Tract
by Individual Race/Ethnicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1 Lead Emissions from All Sources Exposure Risk Quartile . . . . . . . . . . . . . . 78
4.2 Project B Data Restriction Process . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.3 Overall Cognition Range for Age Groups . . . . . . . . . . . . . . . . . . . . . . 86
4.4 Distribution of Lead Exposure for Study Population . . . . . . . . . . . . . . . . . 87
4.5 Mean Overall Cognition Score by Tract of Residence Lead Exposure Quartile . . . 88
5.1 Frequency of Parcel Building Construction Year 1960-2009 . . . . . . . . . . . . . 125
5.2 Range of Log Lead Exposure for Lead Exposure Risk Quartile Groups . . . . . . . 127
5.3 Census Regions and Divisions of the United States . . . . . . . . . . . . . . . . . 130
x
Chapter 1
Introduction
Lead is a toxic heavy metal significant adverse health outcomes for those who are exposed. The
overall goal of this dissertation is to evaluate and quantify the risk of exposure to sources of lead
for older adults in the United States and assess potential health outcomes related to this exposure.
1.1 Background
Lead is a toxic heavy metal with significant negative health consequences associated with both
acute and chronic exposure. While lead is present in the Earth’s crust at an average concentration
of approximately 14 mg/kg, naturally occurring environmental lead rarely exceeds a concentration
of 100 mg/kg and is not considered to be a significant risk to human health [1].
Instead, lead hazards which do pose a significant threat to human health are almost exclusively
the result of human activity; high lead exposure hazard risk is usually associated with densely
populated urban areas, with older housing built prior to federal lead-based paint regulations, and
with industrial or mining areas. For example, the average soil lead concentration at United States
Environmental Protection Agency (EPA) Superfund sites was recently found to be 25.8 mg/kg [2],
and the overall average soil lead level in the highly populated Los Angeles metropolitan area to be
181 mg/kg [3], both far higher than the naturally occurring average soil lead concentration.
1
1.1.1 History of Lead Use and Regulation in the United States
Historically, lead has been used in many industrial and consumer products but its use has decreased
as the negative health consequences of lead exposure have become better characterized and more
well-known. During the second industrial revolution (1860-1914), the concentration of lead in
air, water, and soil in the United States significantly increased due to expanded use of lead in
industrial processes and incorporation of lead in products [1]. Throughout the first half of the 20th
century, lead was commonly used in the manufacture of gasoline, paint, plumbing, solder, and
other consumer products [4]. During the 20th century, lead was removed from many consumer
products in the United States, most notably from house paint by 1978 and gasoline by 1996.
Figure 1.1, reproduced from Laidlaw and Filippelli 2008, illustrates the extensive use of lead
in paint and gasoline in the United States during the 20th century as well as the rapid decline in
lead use in paint which occurred primarily from the 1950s to the minimization of lead in paint in
1978 and the rapid decline in lead use in gasoline during the 1970s and 1980s leading up to federal
regulation in 1996 [5]. While it should be noted that Figure 1.1 does not incorporate lead from
other sources besides paint and gasoline, it may be interpreted as a useful representation of the rise
and fall of lead use in the United States and, by extension, the rise and fall of new lead deposited
in the environment.
Although by the end of the 20th century lead had minimized or removed lead from many
consumer products, legacy concentrations of lead remain in the environment and continue to pose
a threat to public health [7, 8]. For example, although lead-based paint was banned from use in
residential homes in 1978, buildings built prior to 1978 are likely to contain remaining lead-based
paint and highly populated areas continue to be associated with high concentrations of lead in soil
and other media due to past leaded gasoline and industrial emissions.
In addition, many ongoing sources of lead remain un- or under-regulated; lead processing,
lead recycling, mining, and airplane fuel combustion continue to contribute to environmental lead
contamination and to both residential and occupational health risks [9, 10].
2
Figure 1.1: History of lead use in paint and gasoline in the United States during the 20th century,
showing the early dominance of lead-based paints followed by the boom in transportation resulting
in a high use of leaded gasoline [6]. The decline after the mid-1970s was due to controls put into
place to reduce the use of leaded gasoline [5].
1.1.2 Lead Exposure and Human Health
There are multiple pathways by which humans may be exposed to lead [8, 11]. Airborne lead, for
example from industrial facility emissions, re-suspended dust or dry soil containing lead, or leaded
gasoline emissions, may be be inhaled [9, 10]. Lead in food and water [12, 13], or in contaminated
soil or dust [14], may be ingested and absorbed into the body in the gastrointestinal tract. Lead can
also be absorbed directly through the skin [15]. In addition, lead is retained in the body in certain
soft tissues and bone and may be released back into the bloodstream due to deterioration of bone
even decades following initial exposure; osteoporosis has been found to re-expose individuals to
acute lead toxicity as lead retained in the bone earlier in life is released as the bone breaks down.
3
Although exposure to small doses of lead or the presence of small amounts of lead in the body
were considered “normal” or “safe” during the first half of the 20th century, it has since been
established that lead has no threshold of effect. No level of lead exposure or amount of lead in the
body is considered “safe” and negative health consequences can occur with any level of lead in the
body [16].
Lead exposure is most commonly associated with adverse neurological health outcomes in-
cluding neurocognitive [17–19], behavioral [20, 21], and motor outcomes [22, 23]. However non-
neurological health outcomes are also associated with lead including negative impacts on renal
function [22, 24, 25], cardiovascular function [26], and sexual development [27, 28].
1.1.3 Lead and Older Adults
As outlined in Section 1.1.1, lead use in the United States was at its peak during the 20th century
and major federal regulations limiting the amount of lead in consumer products were not estab-
lished until the final decades of the century. Individuals classified as “older adults” today (i.e.
adults aged 65 and older in 2023; born in the 1910s through 1950s) lived their childhood during
the peak years of lead use in the United States and had already reached adulthood by the time
federal lead regulations were being implemented. Figure 1.2 re-frames the information presented
in Figure 1.1 to show the relationship between lead use and major regulatory action in the United
States and the life course of current older adults. As a result, adults in this cohort have arguably
experienced a lifetime lead burden greater than any United States generation before or since.
As mentioned in Section 1.1.1, the detrimental effects of even very low levels of lead exposure
on the developmental, neurological, and behavioral health of young children is well-established
in the literature. In addition, the research on this topic has further determined that the impacts
of lead exposure early in life may have a permanent detrimental impact on neurocognitive health
which extends into adulthood and older age. Adults with a history of lead exposure during their
lifetime may be approaching older age, and the associated additional neurocognitive health issues
associated with the aging process, with a baseline of already poor neurocognitive health. For these
4
Figure 1.2: History of lead use in paint and gasoline in the United States during the 20th century
relative to life trajectories of older adults (aged 70-100 in 2010). Paint and gasoline use intensity
based on a figure from Mielke et al. 1999 [6].
adults, the likelihood of detrimental neurocognitive health outcomes, such as cognitive decline,
dementia, or Alzheimer’s disease, may be more likely, come earlier in adulthood, be more severe,
or some combination of the three.
In addition, research has also determined that exposure during adulthood, not only childhood,
results in adverse neurocognitive and other health outcomes. This research has historically fo-
cused on occupational lead exposure but has recently begun to explore the effect of environmental
exposures and found that even low-level, non-occupational adult exposure can result in adverse
neurocognitive outcomes such as cognitive decline. Therefore, current US older adults, who are
already likely to have experienced some of the highest levels of lead exposure as children and some
of the highest risk of adverse lead-related health outcomes both as children and in adulthood, may
5
continue to compound on the likelihood of adverse lead-related health outcomes by experiencing
additional lead exposure as adults. While adults with a history of lead exposure may already be
approaching older age with a baseline of poor neurocognitive health due to lead exposure, this
baseline may be further depressed by ongoing lead exposure from residual legacy and contempo-
rary sources. Again, this further depressed baseline may further impact the likelihood, timing, and
severity of serious neurocognitive health outcomes later in life.
Compounded across an entire generation of older adults in the United States who are approach-
ing end of life and the associated health care needs, the rippling impacts of lead lifetime lead burden
and related neurocognitive outcomes may have massive impacts on family member and other care-
givers, the healthcare system, and the older adult care industry. The overarching hypothesis of this
dissertation is that ongoing lead exposure, experienced during adulthood and older adulthood and
from contemporary sources of lead, has the potential to exacerbate the neurocognitive problems
caused by lifetime lead exposure and, at a population level, increase the incidence of serious health
issues and the older adult care burden in the United States.
This dissertation also asserts that, while past lead exposures are undeniably important and likely
remain the main contributor to lifetime lead burden for most older adults, these past exposures
cannot be undone nor the resulting health effects reversed. However, potential future lead exposure,
including from both residual environmental lead from legacy sources which have already been
regulated as well as from sources which have yet to be fully or effectively regulated, may still be
mitigated or eliminated. Eliminating these future exposures from the remaining sources of lead
may have the potential to minimize the extent of adverse neurocognitive health outcomes and care
burden for the older adult population in the United States.
6
1.2 Dissertation Structure
Chapter 2 uses a data set of paint chip samples collected in the Los Angeles metropolitan areas
to develop a series of models for predicting the likelihood of lead-based paint, high-concentration
lead-based paint, and very high-concentration lead-based paint in older adult residences based on
housing age and residential use characteristics. The resulting models had prediction accuracies of
89%, 63%, and 86%, respectively indicting that housing age and use may estimate lead-based paint
presence with a relatively high degree of predictive power. The models have the ability to predict
and estimate the presence of lead-based paint with more detail than a traditional approach that only
considers whether housing was built prior to 1978, when lead-based paint was federally banned
from residential use, an approach which implies no lead-based paint risk for housing built after the
late 1970s and no variation in risk for housing built while lead-based paint was federally legal. The
final models were applied to data from the 2019 United States American Housing Survey (AHS) to
determine that approximately 42% of older adults living in non-mobile housing in the United States
are at some risk of residential lead-based paint. The results of this chapter also show differential
risk based on householder race and household composition, suggesting that demographic and other
social characteristics are indicators of exposure risk.
Chapter 3 uses data from the Health and Retirement Study (HRS) and the EPA’s National Air
Toxics Assessment (NATA) to explore the relationship between demographic and socioeconomic
characteristics in the United States older adult population and the risk of neighborhood-level ex-
posure to contemporary sources of airborne lead emissions. The results of this chapter show that
individual race and ethnicity is strongly related to neighborhood exposure risk, with older adults
of color at higher risk compared to White non-Hispanic older adults, and that this relationship
is not modified by other individual socioeconomic characteristics such as wealth, income, or ed-
ucational attainment. These results support a growing body of literature highlighting differential
race-based exposure to environmental pollutants in the United States and the importance of placing
environmental health research, outreach, and policy within an environmental justice framework.
7
Chapter 4 explores the relationship between risk of neighborhood-level exposure to contem-
porary sources of airborne lead emissions and overall cognitive functioning in older adults, also
using data from the Health and Retirement Study (HRS) and the EPA’s National Air Toxics As-
sessment (NATA). The resulting models indicate that older adults living in areas with the highest
exposure risk have lower overall cognitive functioning and when controlling for age and longitu-
dinal change over time. This relationship, however, is attenuated after adjusting for individual and
neighborhood-level covariance, namely individual race/ethnicity, neighborhood racial demograph-
ics, and neighborhood educational attainment, again indicating the significance of demographic
and socioeconomic characteristics and further supporting the findings from chapters 1 and 2 which
emphasize differential exposure risk by racial group for older adults in the United States.
8
Chapter 2
Quantification of Residential Lead-Based Paint Presence Using
Household Characteristics
2.1 Abstract
Older adults are at risk of adverse health outcomes due to a lifetime of chronic lead exposure from
paint, gasoline, and other products. Residential lead-based paint is a contemporary source of lead
which may contribute to this lifetime lead burden, exacerbate health outcomes, and increase care
burden. Housing characteristics, such as housing age, may be used to determine the risk of lead-
based paint presence and, therefore, the number of the older adults with potential residential lead
exposure in the United States.
Residential lead-based paint presence likelihood models were developed using 1949 urban res-
idential property parcels, corresponding to 12,646 collected paint chip samples, from the Los An-
geles metropolitan area. Final models indicated that the likelihood of lead-based paint presence
(paint with lead loading of 1 mg/cm
2
or greater), high-concentration lead-based paint presence (5
mg/cm
2
or greater), and very high-concentration lead-based paint presence (10 mg/cm
2
or greater)
were all strongly associated with housing construction year. Lead-based paint presence was signif-
icantly lower for apartments built between 1950 and 1959 compared to single-family homes built
in 1950-1959, but housing type was not significantly associated with lead-based paint presence for
all other construction time periods and lead concentrations.
9
Model results were applied to the 2019 American Housing Survey (AHS) data and used to
estimate that approximately 19.7 million older adults live in households with some risk of lead-
based paint. Older adults living alone have a higher likelihood of residential lead-based paint
compared to older adults living another older adult and older adults living in households with black
or mixed-race householders have a higher likelihood of residential lead-based paint compared to
older adults living in households with white, Asian, or Native American race householders.
Housing construction year and housing type should be considered when implementing lead-
based paint remediation, health, and education. Residents in housing built prior to 1950 should
be prioritized for health outreach due to the very high likelihood of lead-based paint and higher
likelihood of lead-based paint with very high concentration.
2.2 Introduction
Lead exposure is a pervasive environmental health issue that has been linked to numerous adverse
health effects in humans, including cognitive impairment [25, 29–31]. While the health risks of
lead exposure have been extensively studied in children [32–35], less is known about the effects
of lead exposure on older adults [36–77], who may be more susceptible to the health effects of
lead due to age-related declines in organ function and physiological resilience as well as higher
prevalence of preexisting health conditions. Given that older adults are both the fastest growing
sub-population in the United States and have arguably been exposed to some of the highest chronic
lifetime lead exposure levels in United States history, it is essential to understand the causes and
sources of potential historical and ongoing lead exposure in the older adult population.
Lead-based paint, which was not federally regulated in the United States until the 1970s, is
a common pathway for household lead exposure and homes built prior to 1978 which have not
undergone lead-based paint remediation, and is generally considered to pose some health risk to
residents due to the possibility of lead in remaining paint. However, both the concentration of
lead in paint and the frequency of lead-based paint use declined during the 20th century prior to
10
regulations in the 1970s and therefore, homes built throughout the first seven decades of the 20th
century should not be expected to contain similar concentrations of lead-based paint.
The concentration and extent of residential lead-based paint risk may be compounded for older
adults in the United States who have lived in the same residence for decades and experienced a
lifetime of lead exposure from multiple sources. Lead accumulates in the body and chronic long-
term exposure compounds the adverse potential health outcomes. Many of these health outcomes
are neurocognitive and result in mental, motor, and neurologic decline, placing further burden on
elder care resources. Prior research has shown that lead exposure is significantly associated with
adverse changes in overall cognition as well as in specific cognitive domains in older adults. Due
to their lifetime of exposure and the potential adverse health impacts late in life, lead exposure for
older adults should be characterized and quantified.
This project has two objectives, the first is to create a residential lead-based paint exposure
estimation framework based on housing characteristics. This framework may be used to analyze
and assess property parcel-specific risk of presence of lead-based paint while taking housing con-
struction year and other characteristics into consideration. Housing in the United States is flagged
as a potential risk for lead-based paint exposure if constructed prior to 1978, when current federal
lead-based paint restrictions were put in place. However, this strategy does not consider changes
in lead-based paint use and changes in the concentration of lead in paint which occurred in the
decades leading up to major federal restrictions implemented in the 1970s. As a result, risk assess-
ment based solely on when restrictions were implemented may fail to capture important nuance in
the lead-based paint exposure risk landscape in the United States which may in turn have important
consequences when assessing health risk and designing remediation strategies.
The second objective is to use the exposure estimation framework to characterize residential
lead-based paint exposure risk for older adults in the United States. As mentioned, this population
is likely to have already experienced a lifetime of chronic lead exposure, is at risk for outcomes
related to lead-exposure including neurocognitive and motor decline, and is the fastest growing
subset of the United States population.. Characterizing and understanding the risk of ongoing lead
11
exposure for older adults can aid in mitigation of these health risks for older adults, the fastest
growing population in the United States.
This project uses data on lead content in residential paint chips collected by the California
Department of Toxic Substances Control (DTSC) between 2014 and 2019 for a region within the
Los Angeles metropolitan area and information on national housing characteristics from the 2019
American Housing Survey (AHS)
2.3 Background
Lead is a toxic heavy metal with well-established adverse health effects for humans associated with
both acute and chronic exposure. Although regulations implemented during the past few decades
have significantly reduced many sources of environmental lead in the United States, lead exposure
remains a significant public health concern. Legacy sources of lead, namely lead-based paint in
buildings constructed prior to 1978 [78] and leaded pipes in buildings constructed prior to 1986
[79], continue to be sources of lead exposure even decades following regulatory action. Previous
work has estimated that approximately 24 million homes in the United States have hazardous levels
of lead-based paint or dust [78]. However, although many sources of lead have been regulated and
mitigated, many contemporary sources of new lead exposure exist including air pollution from
industrial activities and exhaust from airplane fuel. Both legacy sources of lead and contemporary
primary sources of lead contribute to contaminated soil, dust, water, and other media.
Once deposited, lead does not degrade in the environment and can reside in the soil and other
media for multiple decades following dispersion [5, 6, 80, 81]. Residual lead particles have the
potential to re-suspend into the air, for example during weather events or construction activity,
which in turn may contribute to re-exposure or contamination of new media even years following
initial release into the environment [6, 80].
12
2.3.1 Residential Lead-Based Paint Use and Risk in the United States
Although lead-based paint was regulated for residential use during the mid-20th century, lead
persists in houses built before regulations were put into place and continue to pose a potential
public health threat. The United States Department of Housing and Urban Development (US HUD)
associates a potential hazard of lead exposure to any house built prior 1978.
While lead-based paint regulations are often expressed in percent by weight while lead con-
centration in paint chips or in paint on surfaces are reported in area concentration using units such
as mg/cm
2
or weight of lead using units such as parts per million (ppm) or mg/kg [82]. Percent
by weight and cannot be directly converted to area concentration or weight or lead. However,
according to the United States Department of Housing and Urban Development, “[. . . ] in theory,
1 mg/cm
2
corresponds to [about 12% lead] in a single layer of paint [. . . ]. Because of the pres-
ence of many layers of paint in target housing, on average 1 mg/cm
2
is about equal to 1% lead.”
[82] Currently, lead-based paint is defined as paint with an area concentration of 1 mg/cm
2
lead, a
percent weight of 0.5% lead, or 5000 mg/kg ppm lead by weight [83].
2.3.2 Lead Biokinetics and Toxicity
Lead can enter the body through multiple pathways including ingestion, inhalation, skin contact, or
through the placenta [8–15]; the most common routes of exposure are through the gastrointestinal
or respiratory tract [65, 84]. Once absorbed, lead is transported throughout the body by the blood
and has the potential to impact multiple bodily organs and systems. While lead in the blood has a
half-life of approximately 25 to 35 days, lead may be absorbed by and accumulate in certain soft
tissues, bone, and teeth where it can remain for multiple decades [85, 86]. Multiple studies have
identified that lead retention in the bone is cumulative throughout the lifespan to at least age 50
with evidence of continued accumulation through the 70s [87–89]. The bones effectively become
a reservoir of lead in the body as lead may release back into the blood during bone remodeling or
deterioration. This process re-exposes the individual to the potential toxic effects of blood lead,
13
causing adverse health effects to occur even decades following exposure [85], a process which is
especially significant for older adults who are experiencing bone deterioration due to osteoporosis.
2.3.3 Lead and Older Adults
In the United States, older adults (age 65 years and older) have been exposed to more lead during
the course of their lifetime than perhaps any other generation in history. Older adults today lived
their childhood and young adulthood during the peak years of residential lead-based paint from
the beginning of the 20th century through the 1940s and peak years of leaded gasoline use from
the 1950s through 1980s. Members of this generation had already reached adulthood by the time
contemporary federal regulations were placed on lead in paint (1978; current older adults aged
20+) and gasoline (1996; current older adults aged 38+).
In older adults, lead exposure has been associated with faster declines in overall cognitive
functioning [57, 60–62, 90], adverse changes in the brain [41, 52, 65, 91–94], increased prevalence
of Alzheimer’s Disease and other dementias [90, 95, 96], hypertension [97], cardiovascular disease
[98], chronic kidney disease [98], osteoporosis [98–100], frailty [101, 102], and cataracts [103].
Notably, it has been hypothesized that, for adults who were exposed to lead in their lifetime, a
significant amount of the decline in cognitive functioning previously considered part of the natural
aging process may be attributed to lifetime lead exposure [54, 67, 104].
Despite this, there is limited research on environmental contextualization of lead exposure for
older adults and lead regulations and restrictions are generally focused on young children rather
than adults and older adults. The IEUBK model for lead in children is a complex model designed to
estimate the likely blood lead level based on environmental conditions [105, 106]. However, there
is no analog model designed for older adults who have experienced a lifetime of lead exposure and
may be experiencing lead re-exposure as lead in the body is released from the bone, as well as
potential continued contemporary lead exposure through contact with remaining lead-based paint
and other media.
14
The older population is the fastest growing population in the United States and health concerns
related to older adults have become and will continue to be an increasingly important focus. The
neurocognitive outcomes associated with chronic and lifetime lead exposure will continue to be-
come increasingly important as the population which grew up during peak lead-based paint use
reaches old age and experiences the health outcomes associated with chronic exposure in older
adults. Considering many of the outcomes are neurocognitive or motor related, such as increased
dementia incidence and increased fall risk, the increased burden of chronic lifetime exposure in
older populations will also result in an increased care burden and pressure on family members who
provide care to their older relatives and on the older adult care profession and industry.
In addition, residual lead-based paint presence in residential homes is also often more common
in low-income and vulnerable communities and has been associated with historical racist housing
practices such as redlining. Low-income and vulnerable communities may be less likely to have
the means to remediate homes for lead and address lead and other hazards as they become known.
Homes in low-income and vulnerable communities, as well as homes with known lead presence,
are also more likely to be devalued which may discourage homeowners from renovating or sell-
ing their homes and may contribute to suppressed generational wealth. As a result, the increased
burden of care associated with lead-related health issues in older populations is likely to also be
concentrated in communities and populations which are already under-served, economically vul-
nerable, and socially marginalized.
2.3.4 Project Objectives
The first objective of this research is to create a residential lead-based paint exposure risk scale
which considers the changes in both residential lead-based paint use and the lead concentration in
paint which occurred during the first half of the 20th century in the United States leading up to the
implementation of federal lead-based paint restrictions in the 1970s. This research will utilize a
previously collected dataset of paint chips collected in an urban residential neighborhood with a
high density of older housing stock and tested for lead concentration to estimate lead-based paint
15
presence for residential parcels based on building construction age. This risk estimation scale may
be implemented to determine lead-based paint risk levels for residential parcels at a greater level of
detail than simply indicating that a parcel may have a lead-based paint risk if construction occurred
on the parcel prior to 1978.
The second objective of this research is to use the residential lead-based paint estimation frame-
work mentioned in the previous section to estimate residential lead-based paint exposure risk for
older adults (aged 65+) in the United States. This information may be utilized to underscore the
importance of assessing lead exposure for older adults and to better situate continued lead exposure
risk in the scope of an older adult’s lifetime of chronic lead exposure.
2.4 Methods
2.4.1 Data Sources
2.4.1.1 Paint Chip Data from the California Department of Toxic Substances Control (DTSC)
Between January 2014 and June 2019, the California Department of Toxic Substances Control
(DTSC) collected over 340,000 soil and paint chip samples from communities located within a
region of the Los Angeles metropolitan area. These samples were tested for concentrations of
lead and other heavy metals. The May 31, 2020 release of the sampling data contains information
on 22,889 paint chip samples tested for lead content representing 6,117 parcels including 6,036
residential parcels, 62% of the 9,728 residential property parcels in the study area [107].
2.4.1.2 American Community Survey (AHS)
The 2019 AHS survey includes information on 63,185 residences, 54,455 (86%) of which were oc-
cupied at the time of interview representing 134,160 residents. Survey weight values are provided
with the data to generate estimates for the entire US population. With survey weights applied,
16
the 2019 AHS survey data represents a population of approximately 306 million including 50.3
million older adults (16.4% of total population).
2.4.2 Measures
Table 2.1 provides an overview of the variables used in this analysis including the data source and
release year, variable type, range (for numeric variables), and levels (for indicator and categorical
variables).
Table 2.1: Variables Used in this Analysis
Level Variable Source Type Range or Levels
Parcel
Lead-Based Paint Presence 2014-2019 DTSC categorical 3 categories
Building Construction Year City of Los Angeles categorical 7 categories
Land Use City of Los Angeles categorical 2 categories
Housing Type 2019 AHS and Los Angeles County Assessor’s Office categorical 2 categories
Household
Household Composition 2019 AHS categorical 7 categories
Householder Race 2019 AHS categorical 5 categories
Peeling Paint Presence 2019 AHS indicator yes (vs no)
2.4.2.1 Parcel Lead-Based Paint
DSTC-samples property parcels with at least 5 paint chip samples were categorized using three
lead-based paint presence distinctions indicating the presence of lead-based paint (LBP), the pres-
ence of high concentration lead-based paint (H-LBP), and the presence of very high concentration
lead-based paint (VH-LBP). The definitions of these three distinctions are given in Table 2.2.
Table 2.2: Parcel Lead-Based Paint Presence Definitions.
Category Description
Lead-Based Paint (LBP) one or more paint chips with a lead loading of 1 mg/cm2 or greater
High Concentration Lead-Based Paint (H-LBP) one or more paint chips with a lead loading of 5 mg/cm2 or greater
Very High Concentration Lead-Based Paint (VH-LBP) one or more paint chips with a lead loading of 10 mg/cm2 or greater
2.4.2.2 Parcel Building Construction Year
Parcel building construction year for the parcels located within the study area were derived from
data from the city of Los Angeles [108]. Construction year ranged from 1854 to 2014 for the study
17
area. Construction year for parcel buildings was categorized into ranges roughly based on the con-
struction year ranges used in the 2019 American Housing Survey (AHS) [109]. The AHS provides
individual construction year information for buildings in 2010 or later and construction decade for
buildings built between 1920 and 2009 while all buildings built prior to 1920 are categorized into
a single “1919 or earlier” category. Construction year for the study data were categorized using
the same categories as the AHS data for buildings built between 1900 and 1959. Buildings built
in 1960 or later were categorized into 1960-1979 and 1980 or later year ranges. The 1980 or later
year range approximately corresponds to buildings constructed after lead-based paint was fully
restricted from residential use in the United States in 1978. Buildings built between 1960-1979
correspond to a period of limited, but still federally legal, use of lead-based paint. The 1960s and
1970s were combined into a single category due to the relatively low number of buildings in the
study area which were constructed during these two decades.
Figure ‘2.1 shows the distribution of construction years for buildings on residential parcels
represented by a subset of 1946 parcels with at least 5 associated paint chip samples used for
analysis. The distribution of housing age in the study area is highly representative of the building
era associated with peak lead-based paint use in the United States with a mean and median building
age of 1938 and 1928, respectively.
2.4.2.3 Parcel Land Use
Parcel land use classification information was derived from parcel data from the city of Los Ange-
les [108] [76]. This variable was not re-coded for this analysis.
2.4.2.4 Parcel Housing Type
Both the 2019 AHS data and Los Angeles County assessor’s office data categorizes households
into multiple use categorizations. Both categorization schemes were simplified to two categories:
single-family home and apartments. As mentioned in the previous section, all households catego-
rized as mobile homes, trailers, boats, RVs, vans, or other mobile type homes in the 2019 AHS
18
data were not included in the analysis. Further information on the housing type categorization
scheme and corresponding categories from the 2019 AHS and Los Angeles County assessor’s data
is provided in Appendix A.
2.4.2.5 Household Composition
The 2019 AHS data includes information on the total number of people as well as the number
of older adults (age 65+) and children (age 0-17) in each surveyed occupied household. Based
on this information, each household was categorized into seven household composition categories
for this analysis. Households were also broadly classified as older adult households (households
with one or more residents aged 65 or older) and non-older adult households (all residents in the
household are 64 or younger). Further information on the household composition categorization
scheme using AHS data is provided in Appendix B.
2.4.2.6 Householder Race
Householder race information was derived from the 2019 AHS. The original householder race
variable has six race categories including White, Black, Asian, American Native, mixed race, and
other races. A five-category version of this variable, excluding the “other races” category, was used
for this analysis.
2.4.2.7 Peeling Paint
Information on the presence of peeling paint was derived from the 2019 AHS. This variable was
not re-coded for this analysis.
19
2.4.3 Data Management
2.4.3.1 Parcel Data
As mentioned above, the May 31, 2020 release of the DTSC’s parcel sampling data contains infor-
mation on 22,889 paint chip samples tested for lead content representing 6,117 parcels including
6,036 residential parcels. These data were merged with parcel building construction year and land
use data from the city of Los Angeles. Of the 6,117 sampled parcels, 81 non-residential parcels,
1 residential parcel classified as a group home, and 10 parcels with missing use code information
were removed from the sample leaving 6,025 parcels. Of these a further 15 residential parcels with
a corresponding building constructed before 1900 (range 1854 to 1899) and 17 parcels with miss-
ing building construction year information were also removed leaving 5,993 residential parcels
used for analysis.
The 5,993 residential property parcels were associated with an average and median of 3.73 and
3 paint chip samples per parcel, respectively. 1,946 (33%) parcels were associated with at least 5
paint samples and this subset was used when calculating parcel-level summary information and in
model development. Figure 2.2 provides information on the data restriction process for the parcel
data.
2.4.3.2 Household Data
Of the 54,455 occupied households represented in 2019 ACS, 2,462 households categorized as
mobile homes, trailers, recreational vehicles (RVs), boats, vans, and other mobile home types were
excluded from this analysis leaving 51,993 occupied households. Of these 16,391 (32%) house-
holds included at least one older adult (age 65 or older) representing a total of 21,943 surveyed
older adults corresponding to approximately 47.4 million older adults in the US population (94.3%
of total US older adult population) after survey weights are applied. Information from the 2019
AHS on housing construction year, housing type, household composition, householder race, and
presence of peeling paint for the subset of 51,993 occupied non-mobile households and 16,391
20
occupied non-mobile households with at least one older adult was utilized for analysis. House-
holds were weighted based on the weight values provided by the AHS metadata to generate lead
exposure risk estimates for the entire US population. Figure 2.3 below provides information on the
data restriction process for the household data.
2.4.4 Analysis Plan
2.4.4.1 Multivariable Analysis
Logistic regression was used to estimate the likelihood of the presence of lead-based paint (LPB),
high concentration lead-based paint (H-LBP), and very high concentration lead-based paint (VH-
LBP) at the parcel level based on parcel-level characteristics for the 1,946 parcels in the study
area with at least 5 paint chip samples. Construction year categories and residential use categories
(described above) were assessed as potential independent variables. Models were tested for sensi-
tivity to the inclusion of an interaction term between construction year category and residential use
category. Variables for final models were selected based on BIC minimization and selected models
were tested for predictive power and sensitivity to the subset of parcels used for model develop-
ment. Validation was carried out by carrying out the model development process on 500 randomly
selected subsets of 80% of the parcels and calculating the predictive power on the remaining 20%
of parcels.
2.4.4.2 Exposure Risk Estimation for US Older Adults
The final selected models were used to estimate construction year- and housing type-specific like-
lihood of lead-based paint (LBP), high concentration lead-based paint (H-LBP), and very high
concentration of lead-based paint (VH-LBP) for all 51,993 occupied households and the subset
of 16,391 occupied households with at least one older adult based on the 2019 AHS data. AHS
household survey weights were applied to determine the total US population living in housing
scenarios based on building construction year, housing type, household composition, householder
21
race, and presence of peeling paint. The total population in each scenario was then multiplied by
the LBP likelihood for the given scenario and the results used to calculate the overall estimated
population-risk of exposure to residential parcel LBP, H-LBP, and VH-LBP for the United States
population.
(population)(surveyweights)(likelihood of LBP presence)
(total population)(surveyweights)
=populationatriskof LBPpresence
(2.1)
2.5 Results
2.5.1 Parcel Analysis Sample Characteristics
Table 2.3 provides information on the number and percent of residential property parcels repre-
sented by the DTSC dataset in each of the building construction year and housing type categories
for all parcels (n = 5,993) and for the subset of parcels with at least 5 corresponding paint chip
samples (n = 1,946).
Table 2.3: Building Construction Year and Residential Use Categories for All Parcels and for
Parcels with at Least 5 Paint Samples.
Count (%) of All Parcels
Count (%) of Parcels
with at Least 5
Paint Chip Samples
N 5,993 1,946
Parcel Building Construction Year
1900-1919 972 (16.2%) 291 (15.0%)
1920-1929 2914 (48.6%) 911 (46.8%)
1930-1939 612 (10.2%) 252 (12.9%)
1940-1949 766 (12.8%) 304 (15.6%)
1950-1959 419 (7.0%) 132 (6.8%)
1960-1979 218 (3.6%) 31 (2.1%)
1980-2019 92 (1.5%) 15 (0.8%)
Parcel Use Code
Single-Family Home 5659 (94.4%) 1796 (92.3%)
Apartments 334 (5.6%) 150 (7.7%)
22
2.5.2 Lead-Based Paint Presence
Overall, most (70%) parcels in the study area contained at least one lead-based paint chip, re-
gardless of the number of samples collected in the parcel. Of the subset of parcels with at least
5 associated paint chip samples, approximately 87% contained at least one lead-based paint chip,
approximately 40% contained at least one high concentration lead-based paint chip, and approxi-
mately 14% contained at least one very-high concentration lead-based paint chip. Table 2.4 pro-
vides summary information on individual paint chip lead concentration as well as the number and
count of parcels which fall into each of the three lead-based paint presence distinctions for all
parcels and the subset of parcels with at least 5 associated paint chip samples.
Table 2.4: Summary of Lead-Based Paint Presence for All Parcels and for Parcels with at Least 5
Paint Samples.
Count (%) of All Parcels Count (%) of Parcels with at Least 5 Paint Chip Samples
N 5,993 1,946
Total Paint Chip Samples 22,899 12,646
Paint Lead (mg/cm2) mean, median, maximum, [IQR] 1.96; 0.70; 37.1; [0.10, 2.63] 2.12; 0.81; 37.1; [0.10, 2.90]
Mean Parcel Paint Lead (mg/cm2) mean, median, [IQR] 1.82; 1.16; [0.32, 2.54] 2.14; 1.51; [0.64, 2.88]
LBP Present 4181 (69.8%) 1689 (86.8%)
H-LBP Present 1549 (25.9%) 782 (40.2%)
VH-LBP Present 444 (7.4%) 270 (13.9%)
Parcel building construction year range was strongly associated with the presence of lead-
based paint chips for all three lead-based paint presence distinctions. Parcels with more recently
constructed buildings were less likely to contain lead-based paint chips. Approximately 27% of
buildings constructed following the minimization of lead from paint in the United States in 1978
were still associated with lead-based paint, albeit not with high concentration or very high concen-
tration lead-based paint.
Housing type was also associated with the presence of lead-based paint. Apartment parcels
tended to have significantly lower presence of lead-based paint chips compared to single-family
home parcels.
23
Table 2.5: Count and Percent of Parcels with Lead-Based Paint Chip Presence by Building Con-
struction Year Categories and Residential Use Categories for Parcels with at Least 5 Paint Samples.
Parcels with at Least Parcels with at Least Parcels with at Least Parcels with at Least
5 Paint Chip Samples 5 Paint Chip Samples 5 Paint Chip Samples 5 Paint Chip Samples
with LBP Present with H-LBP Present with VH-LBP Present
N 1,946 1689 (86.6%) 782 (40.2%) 270 (13.9%)
Parcel Building Construction Year
1900-1919 291 278 (95.5%) 173 (59.5%) 89 (30.6%)
1920-1929 911 811 (89.0%) 416 (45.7%) 149 (16.4%)
1930-1939 252 230 (91.3%) 105 (41.7%) 20 (7.9%)
1940-1949 304 274 (90.1%) 80 (26.3%) 11 (3.6%)
1950-1959 132 81 (61.4%) 8 (6.1%) 1 (¡ 1%)
1960-1979 31 11 (26.8%) 0 0
1980-2019 15 4 (26.7%) 0 0
Parcel Use Code
Single-Family Home 1796 5659 (94.4%) 740 (41.2%) 257 (14.3%)
Apartments 150 334 (5.6%) 42 (28.0%) 13 (8.7%)
2.5.3 Household Analysis Sample Characteristics
Table 2.6 provides information on the number and percent of households represented by the 2019
AHS dataset in each of the building construction year categories, housing type categories, house-
hold composition categories, householder race categories, and with or without the presence of
peeling paint for all households (n = 51,993) and for the subset of households with at least one
older adult (n = 16,391).
2.5.4 Multivariate Analysis
Final logistic regression model marginal effects and 95% confidence intervals for each of the three
final models are given in Table ?? and Figure 2.4. Further details on the models tested in the
selection process, as well as final models, are provided in Appendices C, D, and E.
The probability of lead-based paint presence is strongly associated with the construction year
of the building on the residential property parcel. Probability of lead-based paint is also associated
with residential use type, but this association was insignificant in most situations when construction
year is included in the model. The exception to this is for the LBP model for buildings built
between 1950-1959. For this model, for buildings constructed in 1950-1959, there is a significant
difference between the expected likelihood of lead-based paint presence for single-family homes
24
Table 2.6: Building Construction Year, Residential Use, Household Composition, Householder
Race, and Presence of Peeling Paint for All Surveyed Non-Mobile Households and for Surveyed
Non-Mobile Households with at Least One Older Adult.
All Households Households with at Least One Older Adult (65+)
N 51,993 16,391
Parcel Building Construction Year
1900-1919 2978 (5.8%) 913 (5.6%)
1920-1929 2092 (4.0%) 625 (3.8%)
1930-1939 1395 (2.7%) 474 (2.9%)
1940-1949 2495 (4.8%) 831 (5.1%)
1950-1959 5502 (10.6%) 1912 (11.7%)
1960-1979 13,245 (25.5%) 4713 (28.8%)
1980-2019 24,286 (46.7%) 6923 (42.2%)
Parcel Use Code
Single-Family Home 35,977 (69.2%) 12,230 (74.6%)
Apartments 16,016 (30.8%) 4161 (25.4%)
Household Composition
One Older Adult (65+) 6328 (12.2%) 6328 (38.6%)
Two Older Adults (65+) 4327 (8.3%) 4327 (26.4%)
One or More Older Adults (65+)
with One or More Adults (18-64),
with or without Children (0-17)
5573 (10.7%) 5573 (34.0%)
All Other Households
with at Least One Older Adult (65+)
163 (0.31%) 163 (1.0%)
One Adult (18-64) 8872 (17.1%) -
Two Adults (18-64) 9170 (17.6%) -
All Other Households 17,560 (74.9%) -
Householder Race
White 38,960 (74.9%) 13,143 (80.2%)
Black 8036 (15.5%) 2034 (12.4%)
Asian 3565 (6.9%) 919 (5.6%)
America Native 545 (1.1%) 110 (0.67%)
mixed race 700 (1.4%) 147 (0.90%)
other (not analyzed) 187 (0.36%) 38 (0.23%)
Peeling Paint
Yes 1020 (2.0%) 249 (1.5%)
No 50,973 (98.0%) 16,142 (98.5%)
25
relative to that for apartment buildings. An indicator term identifying apartment buildings built
between 1950 and 1959 was added to this model as an interaction term with the construction year
category to account for this specific scenario. Apart from this indicator term, housing use type was
not included in the models.
For all other scenarios, inclusion of the residential building use type or an interaction term be-
tween construction year and building use type in multiple linear regression analysis did not improve
the model based on BIC criteria and did not significantly change the coefficients or marginal effect
estimates for LBP, H-LBP, and VH-LBP. Therefore, apart from this indicator term for apartment
buildings built between 1950 and 1959, housing use type was not included in the final model.
Table 2.7: Marginal effects of likelihood of LBP, H-LBP, and VH-LBP presence for parcels based
on construction year range and housing type from logistic regression analysis.
Likelihood of LBP (95% CI) Likelihood of H-LBP (95% CI) Likelihood of VH-LBP (95% CI)
Parcel Building Construction Year
1900-1919 95% (92-97%) 59% (54-65% 31% (26-36%)
1920-1929 89% (87-91%) 46% (42-49%) 16% (14-19%)
1930-1939 91% (87-94%) 42% (36-48%) 8% (5-12%)
1940-1949 90% (86-93%) 26% (22-32%) 4% (2-6%)
1950-1959
SFH: 65% (59-76%)
6% (3-12%) 1% (0-5%)
Apt: 32% (17-52%)
1960-1979 27% (16-42%) 0% 0%
1980-2019 27% (10-53%) 0% 0%
Likelihood of LBP was approximately 90% for parcels with buildings constructed in 1920-
1949 and did not differ significantly within this construction year time range, only declining from
90% beginning with the 1950-1959 construction year category. Similarly, the likelihood of H-LBP
remained at approximately 45% for buildings constructed in the range of 1920-1939 and did not
differ significantly between the two subgroups within this time range.
Parcels built in 1960 or later were associated with approximately 27% likelihood of LBP al-
though this likelihood approached non-significance for parcels built in 1980 or later. Parcels built
in 1960 or later were associated with no significant likelihood of H-LBP and parcels built in 1950
or later were associated with no significant likelihood of VH-LBP.
26
Overall, the LBP and VH-LBP prediction models had good mean prediction accuracy of ap-
proximately 89% and 86%, respectively. The H-LBP model had a significantly lower mean predic-
tion accuracy of only approximately 63%. During model validation, results using test data subsets
did not produce substantially different prediction accuracy for any of the three models.
2.5.5 Exposure Risk Estimation for US Older Adults
Based on the lead-based paint presence risk models defined in the section above and the household
dataset based on 2019 AHS data, the United States older adult population (aged 65+) has approx-
imately 41.5% person-risk of residential LBP presence, 8.3% person-risk of residential H-LBP
presence, and 2.9% person-risk of residential VH-LBP presence. Using nationally representative
data from the 2019 AHS, the final model estimates 19.7 million older adults (aged 65+) in the
United States at some risk of residential LBP
2.5.5.1 Exposure Risk and Construction Year
Approximately 30.2% of the 41.5% LBP person-risk was associated with the 42% of the older
adult population living in residences built in 1980 or later while the remaining 11.3% of the 41.5%
LBP person-risk was associated with the 58% of the older adult population living in residences
built in 1980 or later. Overall, older adults had a slightly higher person-risk of residential lead-
based paint presence compared to adults (aged 18-64), but a slightly lower person-risk compared
to children (aged 0-17).
2.5.5.2 Exposure Risk and Household Composition
The analysis also indicated that older adults who are currently living alone have a greater risk of
residential LBP, H-LBP, and VH-LBP compared to older adults living in households with two older
adults (aged 65+), but a lower risk of residential LBP compared to older adults living in household
with at least one adult (aged 18-64). Table 2.8 provides additional detail on the number and percent
of the US population at risk of LBP, H-LBP, and VH-LBP by household composition.
27
Table 2.8: Estimated count and percent of United States older adult population at risk of residential
lead-based paint by household composition. Includes comparison with adult and child populations.
Percentages correspond to percent of population living in non-mobile housing.
Estimated Population (Millions) and Percent at Risk of LBP Presence by Household Composition
Older Adults (65+) Adults Children
All Older Adults
Older Adults
Living Alone
Older Adults
Living with
One Other
Older Adult
Older Adults
Living in
Households
with both Adults
and Older Adults
All Adults All Children
Presence of LBP 19.7 (41.5%) 5.4 (42.5%) 8.0 (39.9%) 5.9 (43.1%) 74.4 (41.6%) 25.5 (40.4%)
Presence of H-LBP 4.0 (8.3%) 1.2 (9.0%) 1.5 (7.4%) 0.5 (9.1%) 15.8 (8.8%) 5.6 (8.0%)
Presence of VH-LBP 1.4 (2.9%) 0.4 (3.0%) 0.5 (2.6%) 0.4 (3.1%) 5.0 (3.1%) 1.8 (2.8%)
Total Population
(in non-mobile housing)
47.4 million 12.8 million 20.1 million 13.8 million 179.4 million 63.1 million
2.5.5.3 Exposure Risk and Householder Race
When broken down by householder race, the analysis indicated a difference in the percent of the
older adult population at risk of residential lead-based paint based on householder race. Older
adults living in households with Native American or Asian householders were at the lowest risk
of residential lead-based paint while older adults living in households with Black or mixed-race
householders had the highest risk of residential lead-based paint based on the five race categories
analyzed.
Table 2.9: Estimated number and percent of United States older adult population at risk of resi-
dential lead-based paint by householder race.
Estimated Population (Millions) and Percent of Older Adults
at Risk of LBP Presence by Householder Race
White
Householder
Black
Householder
Asian
Householder
America Native
Householder
Mixed Race
Householder
Presence of LBP 15.0 (39.4%) 2.2 (44.6%) 0.80 (38.7%) 0.12 (36.4%) 0.17 (42.0%)
Presence of H-LBP 2.8 (7.5%) 0.50 (10.0%) 0.14 (6.8%) 0.017 (5.3%) 0.038 (9.6%)
Presence of VH-LBP 1.1 (2.8%) 0.16 (3.2%) 0.045 (2.2%) 0.0064 (1.9%) 0.013 (3.3%)
Total Population
(in non-mobile housing)
30.1 million 4.9 million 2.1 million 0.3 million 0.4 million
2.5.5.4 Exposure Risk and Peeling Paint
Based on the 2019 AHS data, approximately 1.5% of the United States older population resides
in a household with a reported presence of peeling paint. When considering reported presence of
28
peeling paint based on the 2019 AHS survey data, older adults in the United States who are living in
a residence with peeling paint are much more likely to also have a risk of LBP presence compared
to older adults living in a residence with no reported peeling paint. Approximately 58% of older
adults living in a house with peeling paint have some estimated LBP presence risk compared to
only 41% of older adults living in a house with no peeling paint.
Table 2.10: Estimated count and percent of United States older adult population at risk of resi-
dential lead-based paint by households with or without peeling paint. Percentages correspond to
percent of the total older adult population living in non-mobile housing.
Estimated Population (Millions) and Percent of Older Adults
at Risk of LBP Presence by Householder Race
Older Adults in Households
with No Peeling Paint
Older Adults in Households
with Peeling Paint
All Older Adults
No LBP 0.30 (42.2%) 27.5 (58.9%) 27.8 (58.5%)
LBP Present 0.41 (57.8%) 19.3 (41.3%) 19.7 (41.5%)
H-LBP Present 0.14 (19.7%) 3.8 (8.1%) 4.0 (8.3%)
VH-LBP Present 0.053 (7.5%) 1.3 (2.8%) 1.4 (2.9%)
All Older Adults
(in non-mobile housing)
0.71 million 46.7 million 47.4 million
2.5.6 Comparison of Model Results with Year-Threshold Risk Frameworks
Using the 2019 AHS survey data, approximately 27.4 million older adults reside in non-mobile
housing constructed prior to 1980. This is roughly the size of the population which would be con-
sidered to have some potential risk of residential lead-based paint using a risk estimation frame-
work which attributes “some risk” to housing built prior to 1978. In comparison, as mentioned
above, the LBP exposure risk model developed for this analysis estimated 19.7 million older adults
at some risk of residential LBP. Most of the discrepancy is associated with residences constructed
between 1950 and 1979. The LBP exposure risk model estimates approximately only 58.0% of
older adults residing in residences constructed between 1950 and 1959 and only 26.7% of older
adults residing in residences constructed between 1960 and 1979 are at risk of LBP presence.
In contrast, the LBP risk exposure model estimates approximately 91.7% of older adults resid-
ing in residences constructed between 1900 and 1949 are at risk of LBP presence and therefore
roughly agrees with a broad risk framework which assumes all residences constructed during this
29
time frame are associated with LBP risk. Figure 2.6 compares the total expected number of older
adults at risk of LBP comparing use of the LBP model developed in this analysis and an assumption
that all residences built prior to 1980 have some risk of LBP presence.
2.6 Conclusions and Discussion
The results of this study indicate that 1) the likelihood of residential lead-based paint presence is
highly associated with housing construction year and, to a lesser extent, with the type of residential
housing; 2) the concentration of lead in residual lead-based paint is also highly associated with
housing construction year; 3) a large percentage of the US older adult population is at risk of
residential lead-based paint; and 4) LBP presence risk is associated with household composition
and householder race.
The sample of residential lead-based paint chips has a large range of lead concentration; 12.0%
of all paint chip samples collected in the study area had a lead loading of 5 mg/cm
2
or more
and 2.9% with a lead loading of 10 mg/cm
2
or more, far more than the current EPA definition of
lead-based paint of 1 mg lead per cm2. This underscores the fact that simply identifying that a
residence has lead-based paint does not fully characterize the potential health and environmental
risk caused by the paint as higher-concentration lead-based paint poses a higher risk than lower-
concentration lead-based paint. The fact that paint lead concentration was highly dependent on
residential building construction year is indicative of the decline in lead concentration in paint
during the early half of the 20th century while lead-based paint use was also declining in the
United States. he wide range of potential lead-based paint concentration and the strong relationship
between paint lead concentration with building age indicate that, when estimating lead-based paint
risk from housing characteristics, it is essential to consider the housing construction era.
Applying the finalized models to the 2019 AHS data determined that approximately 19.7 mil-
lion older adults (42% of the total US older adult population living in non-mobile housing) is
estimated to have residential LBP presence. Of the 19.7 million older adults at risk of lead-based
30
paint presence, 2.8 million are estimated to have high-concentration residential lead-based paint
and 1.1 million estimated to have very high-concentration residential lead-based paint. A common
rule of thumb is that housing built prior to 1978 is associated with some risk of lead-based paint
presence. According to the 2019 AHS survey data, approximately 27.4 million older adults reside
in non-mobile housing constructed prior to 1980 and could be considered to have “some risk” of
lead-based paint. However, this method does not account for the lower risk of overall LBP presence
as well as lower risk of H-LBP and VH-LBP associated with residences constructed at the tail end
of this time range. While most (92%) of older adults living in non-mobile housing built prior to
1950 are at risk of LBP presence, a much smaller proportion living in housing built in 1950 or later
is estimated to be at risk of residential LBP. In addition, older housing is associated with higher
concentrations of lead, amplifying the health risk for the population residing in these homes. When
analyzing and reporting LBP risk for population exposure risk estimation, household remediation
prioritization, or public education and outreach, the greater likelihood of LBP presence as well as
the amplified health and environmental risks associated with the greater likelihood of H-LBP and
VH-LBP presence for older residences should be considered. When prioritization is necessary, the
oldest residential buildings should be prioritized for remediation and residences of these buildings
prioritized for health and risk related outreach.
Estimated risk of LBP, H-LBP, and VH-LBP was associated with household composition and
householder race. Overall, older adults living alone were found to be more likely to be at risk of
LBP presence when compared to older adults living in a household with two older adults. This
finding indicates that older adults living alone are more likely to live in housing with characteris-
tics more strongly associated with LBP presence than older adults living together. It is possible
that the relationship between living alone and lead-based paint presence may be due to an un-
derlying relationship between resident age and housing age. The oldest adults in the older adult
generation, for example adults in their 80s and 90s, are more likely to have a deceased partner,
and therefore more likely to live alone, compared to younger adults in the older adult category
who may be as young as their mid- to late-60s. The oldest older adults may be also more likely
31
to live in older housing compared to younger older adults. Both these possibilities would result
in a spurious connection between living alone and lead-based paint presence. When carrying out
lead risk assessments and outreach for older adult populations, older adults living alone should be
prioritized due to their greater risk of residential LBP exposure. Further research should also be
carried out using more detailed information on resident age to better understand the relationships
between household composition, resident age, housing age, and lead-based paint presence.
Older adults living in households with black or mixed-race householders had the highest likeli-
hoods of residential LBP, H-LBP, and VH-LBP when compared to older adults living in households
with householders from other races. This finding may reflect historical race-related housing prac-
tices, such as redlining which restricted people of color to poorer quality housing compared to
white residents, or other social factors associated with race.
Overall, a large proportion of the United States’ older population is at risk of residential LBP
exposure contributing to a lifetime of chronic lead exposure from paint, gasoline, and household
products which this population has been more likely to have experienced than younger generations.
Chronic lifetime exposure is associated with neurocognitive and motor decline in older adults and
these issues may be exacerbated by ongoing lead exposure household LBP. As the older adult
population continues to increase in size and as their care needs increase as the population ages, it
is essential to characterize and quantify all potential risks which may exacerbate the older adult
care burden. While lead-based paint is a major source of potential lead exposure for older adults,
it is one of many potential ongoing exposures including lead emissions from industry and mining
activities, residual legacy lead in environmental media including soil, leaded pipes in use in the
water system which may leach lead into the water supply, and other sources. Consideration should
be made for all potential sources of lead exposure with the goal of mitigating all hazards for all
populations in the United States. Older adults living in residences built prior to the United States
federal lead-based paint legislation in 1978, and especially older adults living in residences built
prior to 1950, which are at very high risk of LBP presence, should be prioritized for lead outreach
education, lead remediation projects, and lead risk assessment targeted at older adult populations.
32
2.6.1 Limitations
Only 41 (2.1%) of parcels used for the model development process were associated with buildings
built between 1960 and 1979 and only 15 (0.77%) with buildings constructed between 1980 and
2019. A histogram of parcel building construction year for only the subset of parcels with buildings
constructed between 1960 and 2009 is provided in Appendix F for reference. Although the LBP
model produced significant likelihood estimates for parcels with buildings constructed in 1960 or
later, extra care should be taken when using these results due to the small sample size. In addition,
none of the parcels with buildings constructed in 1960 or later were associated with H-LBP and
VH-LBP presence and therefore the models estimated 0% likelihood of H-LBP and VH-LBP for
those parcels. While this result agrees with the fact that federal US legislation banned the use of
lead-based paint in 1977, care should be still used when assuming that all residential parcels with
buildings constructed in 1960 or later do not contain H-LBP or VH-LBP. It should be noted that
both the LBP model and empirical data indicated that a significant percent of parcels with build-
ings built in 1980 or later are associated with LBP presence, which does not agree with legislation
restricting LBP beginning in 1978. As discussed above, the LBP presence in parcels with recently
constructed buildings may be the result of residual lead from previous construction, cross contam-
ination from nearby properties with older buildings which do contain lead-based paint, or other
conditions. Due to the small available sample size of parcels with buildings constructed in 1960
or later, the possibility that H-LBP and VH-LBP may be present in these parcels should not be
discredited. Overall, more analysis should be carried out with a paint chip sample data set which
better represents buildings constructed during or after 1960.
Similarly, only 150 (7.7%) of parcels used in model development were associated with apart-
ment buildings and only 25 (1.3%) with apartment buildings built between 1950 and 1959. Al-
though the LBP model estimated a significant difference in likelihood of LBP between single-
family homes built between 1950 and 1959 vs apartment buildings built between 1950 and 1959,
these results should be utilized with caution due to the small sample size available for this compar-
ison. Further analysis should be carried out on a larger sample of apartment parcels.
33
An additional major limitation of this study is the fact that all parcels and residences in the
study population are located in the same neighborhood in metropolitan Los Angeles. While the
distribution of housing age may be considered to effectively represent at least the time period
when lead-based paint was in use in the United States - prior to the 1978 federal limitation - the
Los Angeles housing sample may not be representative of housing in other areas of the United
States. Although lead-based paint was fully regulated at the federal level in 1978, many cities and
areas had already implemented regulations by that time, with some of the earliest regulations put
in place in the 1950s in the city of Baltimore. Therefore, although the Los Angeles sample may
be representative of areas which did not impose local lead-based paint regulations until the federal
regulation was implemented, it is less likely to be representative of areas which implemented earlier
local bans. Care should be taken when applying the results of this study to local housing stock
and communities and any comparisons between the results of this study and estimates for other
locations should be contextualized within the local history of lead-based paint regualtions.
34
Figure 2.1: Histogram of building construction year for the 1946 residential parcels used for
analysis. Construction year is based on 2014 data from Los Angeles County [108].
35
Figure 2.2: Flow chart of the data restriction process for parcel data.
36
Figure 2.3: Flow chart of the data restriction process for parcel data.
37
Figure 2.4: Marginal effect plots for likelihood of lead-based paint presence based on construction
year and use code for three lead concentration categories of lead-based paint.
38
Figure 2.5: Estimated percent of older adult population at risk of residential lead-based paint by
householder race.
39
Figure 2.6: Comparison of count of older adults at risk of LBP based on the LBP exposure risk
model developed for this analysis with the count of older adults considered at some risk when
using a crude building construction year threshold for risk assessment.
40
Chapter 3
Demographic and Socioeconomic Drivers of Airborne Lead
Exposure Risk in Older Adults
3.1 Background
Lead is a toxic heavy metal with well-established adverse health effects for humans associated with
both acute and chronic exposure. Although regulations implemented during the past few decades
have significantly reduced many sources of environmental lead in the United States, lead exposure
remains a significant public health concern. Legacy sources of lead, namely lead-based paint in
buildings constructed prior to 1978 [78] and leaded pipes in buildings constructed prior to 1986
[79], continue to be sources of lead exposure even decades following regulatory action. Previous
work has estimated that approximately 24 million homes in the United States have hazardous levels
of lead-based paint or dust [78]. However, although many sources of lead have been regulated and
mitigated, many contemporary sources of new lead exposure exist including air pollution from
industrial activities and exhaust from airplane fuel. Both legacy sources of lead and contemporary
primary sources of lead contribute to contaminated soil, dust, water, and other media.
Once deposited, lead does not degrade in the environment and can reside in the soil and other
media for multiple decades [5, 6, 80, 81]. Deposited lead particles have the potential to re-suspend
41
into the air, for example during weather events or construction activity, which in turn may con-
tribute to re-exposure or contamination of new media even years following initial release into the
environment [6, 80].
Lead can enter the body through multiple pathways including ingestion, inhalation, skin con-
tact, or through the placenta [8–15]. Once absorbed, lead is transported throughout the body by
the blood and may be absorbed by and accumulate in certain soft tissues, bone, and teeth where it
can remain for multiple decades [85, 86].
In older adults, lead exposure has been associated with adverse neurocognitive outcomes [5,
38–41, 52, 54, 57, 60–62, 65, 90–94, 110], increased prevalence of Alzheimer’s Disease and other
dementias [90, 95, 96], hypertension [97], cardiovascular disease [98], chronic kidney disease [98],
osteoporosis [98–100], frailty [101, 102], cataracts [103], and other outcomes.
3.1.1 Individual and Neighborhood Characteristics as Determinants of Environmental
Exposure
Previous work has established that lead exposure disproportionately affects low-income commu-
nities and communities of color [111–115]. For example, Lynch et al. (2020) used blood lead
data with decennial census data to determine relationships between census tract sociodemographic
and socioeconomic characteristics and incidence of elevated blood lead in children in Milwaukee
county, Wisconsin. They found that elevated blood lead incidence was higher in areas character-
ized by high poverty, low home ownership, and predominantly non-White residents [113]. This
differential environmental exposure between communities has been attributed to multiple factors
such as discriminatory housing practices such as redlining. These discriminatory structural forces
have resulted in spatial clustering of populations by race/ethnicity and economic class.
Overall, the literature on disparities in lead exposure highlights the need for targeted inter-
ventions and policies that address the underlying social and structural determinants of health that
contribute to these inequities. Such interventions may include lead hazard remediation in older
42
housing stock, targeted public health campaigns, and policy changes that address systemic in-
equalities in housing, education, and healthcare.
While lead exposure has been studied extensively, there is a need to examine the social and
environmental factors that contribute to disparities in exposure and health outcomes among older
adults. Older adults, particularly those living in urban areas or in housing built before the 1970s,
are at increased risk of lead exposure due to environmental and structural factors. This paper will
examine the extent and distribution of airborne lead exposure among older adults in the United
States and highlight the need for targeted interventions and policies to promote environmental
justice and reduce the burden of lead exposure on vulnerable communities.
3.1.2 Project Objectives
The main objectives of this study are to characterize the relationship between demographic and so-
cioeconomic characteristics of older adults and likelihood of living in a neighborhood (census tract)
characterized by a high risk of exposure to airborne lead emissions, and to identify what neighbor-
hood level characteristics are associated with exposure risk for the neighborhoods in which older
adults reside. Specifically, this project will assess if the same relationships between individual and
neighborhood characteristics are associated with disproportionate environmental exposure in other
groups and for other pollutants, such as for children and lead exposure, are also present for older
adults and lead emissions exposure risk in the United States. This paper focuses on characteristics
which have been associated with disproportionate environmental exposure risk - namely individual
or neighborhood race/ethnicity and income - but also expands this exploration to other potential
indicators of vulnerability such as educational attainment, unemployment, and other variables.
In addition, this paper seeks to determine if differential exposure by demographic characteris-
tics, namely race and ethnicity, persists even after accounting for socioeconomic variables such as
income and wealth. As mentioned above, due to discriminatory practices, populations character-
ized by social and economic vulnerability and communities of color are often spatially clustered
43
making it difficult to disentangle the effects of race and ethnicity, i.e. historical and ongoing racist
housing and economic practices, from those of economic class and social mobility.
I hypothesize that, similar to previous literature on other populations and environment contam-
inants, older adults who are non-White, in poverty, or who are less well-educated will be more
likely to live in neighborhoods with high ambient lead levels. I further hypothesize that the high
risk neighborhoods in which older adults reside will be more likely to be characterized by a larger
proportion of non-White residents, higher poverty, older housing, lower income, and lower educa-
tional attainment compared to lower risk neighborhoods. When considered in concert, I hypothe-
size that the effects of race and ethnicity will remain a significant indicator of exposure risk, even
after accounting for individual economic characteristics, educational attainment, and other factors.
This paper aims to add to the literature by exploring the relationships between race, ethnicity,
income, and other demographic characteristics with airborne lead exposure in older adults, and the
implications for environmental justice. Specifically, this work contributes to the literature by 1)
exploring the relationship between airborne lead exposure risk and demographic or socioeconomic
characteristics for older adults and the neighborhoods in which they live and 2) using a nationally
representative population of older adults with an exposure dataset, also at the national scale, there-
fore producing results which may be applied to populations of older adults throughout the United
States.
3.2 Methods
3.2.1 Data Sources
Annual lead emissions data is derived from the 2011 EPA National Air Toxics Assessment (NATA)
[116]. Information on the study population, including overall cognition based on TICS assessment
performance, is derived from the 2012 Health and Retirement Study (HRS) survey wave [117].
Additional census tract characteristics are derived from the 2010 decennial census [118] and 2006-
2010 5-year ACS summary datasets [119].
44
3.2.1.1 National Air Toxics Assessment (NATA)
The National Air Toxics Assessment (NATA), established by the Environmental Protection Agency
(EPA), produces exposure and health risk estimates caused by the approximately 187 hazardous
pollutants
1
(HAPs) and diesel particulate matter (PM) [120]. An updated NATA dataset was re-
leased every 3-6 years between 1996 and 2014 and has been released annually beginning with 2017
data [116, 121–128].
Using a battery of several statistical models
2
processing daily input from air monitors across the
United States, NATA data products include information on both estimated total exposure for each
pollutant, measured in units per square meter, as well as estimated human exposure from specific
source categories (e.g., airports, major point sources) for each pollutant for all census tracts in the
contiguous United States. The NATA dataset also specifies a reference concentration (RfC) for
each analyzed pollutant corresponding to an “estimate of a continuous inhalation exposure that is
thought to be without an appreciable risk of adverse health effects over a lifetime [130].” For the
2011 NATA assessment, the RfC for lead is 0.00015 mg/m3 [130]. For a short review of previous
work linking NATA exposure data with human health outcomes, please see Appendix G.
This project uses data on exposure to lead or lead compounds at the census tract level from the
2011 NATA data release. Although more recent NATA assessments were carried out for the years
2014, 2017, 2018, and 2019, this study uses the NATA 2011 dataset as an indicator of exposure to
past lead exposure for older adults surveyed between 2012 and 2016. Because the NATA modeling
1
Hazardous air pollutants (HAPs) are a group of approximately 187 chemicals and compounds that are known or
suspected to cause cancer, reproductive problems, or other serious health effects. The EPA regulates the HAPs under
the Clean Air Act.
2
Beginning with the 2011 assessment, the NATA exposure evaluation process utilizes the Hazardous Air Pollutant
Exposure Model (HAPEM) and Assessment System for Population Exposure Nationwide (ASPEN) Model [61]. Be-
ginning with the 2011 assessment, the NATA assessment has also utilized a battery of additional models to contribute
to the final exposure estimate for each pollutant including AMS/EPA Regulatory Model (AERMOD), Community
Multi-Scale Air Quality (CMAQ) Modelling System, Consolidated Human Activity Database (CHAD), Emission In-
ventory System (EIS), Human Exposure Model (HEM), Integrated Risk Information System (IRIS), Motor Vehicle
Emission Simulator (MOVES), National Emissions Inventory (NEI), National Mobile Inventory Model, Sparse Matrix
Operator Kernel Emissions (SMOKE), and Weather Research and Forecasting (WRF) Model [129]
45
process underwent a major change beginning with the 2011 assessment and expanded to incorpo-
rate many additional models in the estimation process, data from earlier NATA assessments (1996,
1999, 2002, and 2005) were not utilized for this project.
3.2.1.2 Health and Retirement Study (HRS)
The Health and Retirement Study (HRS) is a longitudinal survey of older adults carried out by
the University of Michigan Institute for Social Research and funded by the United States National
Institute on Aging (NIA) and Social Security Administration (SSA) [117]. The study was created
in 1990 and survey waves have been administered approximately every two years since 1992. The
survey includes information on older adults aged 50 years and older and collects data on health,
retirement, aging, demographic, socioeconomic, and labor characteristics.
This project uses information from the 2012 HRS survey wave. To assess the relationship
between airborne lead exposure risk and individual characteristics from the HRS, I link census
tract of residence from 2012 HRS survey participants to census tract lead exposure estimates from
the 2011 NATA data.
3.2.1.3 Decennial Census and American Community Survey (ACS)
The United States has conducted a decennial census every ten years beginning in 1790. The US
Census Bureau agency was created to carry out the 1900 census and has overseen each subsequent
decennial census. The aim of the decennial census is to produce a complete count of the US
population as well as information on demographic and socioeconomic characteristics of the US
population.
The American Community Survey (ACS), also produced by the US Census Bureau, has been
carried out annually beginning in 2005. The survey consists of a long-form questionnaire of so-
cioeconomic and social questions administered to a subset of the US population. It replaced the
long-form component of the decennial census and was created with the goal of providing detailed
population information more frequently than every ten years. The US Census Bureau produces
46
multiple data products based on the ACS including summary estimates for different geographies
and time ranges. This project uses census tract level information from the 2010 decennial census
and 2006-2010 5-year ACS summary estimates.
3.2.2 Measures
Table 3.1 provides an overview of the variables used in this analysis including the data source and
release year, variable type, range (for numeric variables), and levels (for indicator and categorical
variables). All analyses and data management were conducted using Stata release 17 [131].
Table 3.1: Variables Used in this Analysis
Level Variable Source Type Range or Levels
Individual
Cognitive Functioning 2012 HRS discrete 0 to 27
Age (discrete) 2012 HRS (re-scaled) discrete 52+
Age (categorical) 2012 HRS (derived) categorical 4 categories
Female Sex 2012 HRS indicator female/male
Race and Ethnicity 2012 HRS (derived) categorical 4 categories
Nativity 2012 HRS indicator US/foreign born
Educational Attainment 2012 HRS (derived) categorical 3 categories
Poverty Status 2012 HRS (derived) indicator in/above poverty
Household Income Quartile 2012 HRS (derived) categorical 4 categories
Wealth Quartile 2012 HRS (derived) categorical 4 categories
Census Tract
Estimated Lead Exposure Risk Quartile 2011 NATA (derived) categorical 4 categories
Urbanicity 2012-2016 HRS (derived) categorical 6 categories
Census Division 2012-2016 HRS (derived) categorical 9 categories
% White Non-Hispanic/Latino 2010 Decennial Census continuous 0 to 100%
Percent of Adult Population without High School Degree 2006-2010 ACS 5-year Estimates continuous 0 to 100%
Percent of Housing Built in 1950 or Earlier 2006-2010 ACS 5-year Estimates continuous 0 to 100%
Median Household Income 2006-2010 ACS 5-year (re-scaled) continuous 0+
3.2.2.1 Tract of Residence Lead Exposure Risk
Annual airborne lead emissions exposure risk at the census tract of residence was used as the
main explanatory variable. To differentiate census tracts by lead emissions exposure risk, tracts
were categorized into quartiles based on estimated lead exposure values from the 2011 NATA
assessment. Census tracts in the United States were included in quartile categorization even if they
were not represented by individuals in the HRS data subset used for analysis. Table 3.2 provides
summary information on the four quartile groups in mg/m
3
estimated lead exposure. The table
47
also provides information on group interquartile range (IQR) relative to the lead RfC
3
of 0.00015
mg/m
3
and relative to the overall standard deviation (using log-transformed lead exposure values).
A figure showing the distribution of log estimated lead exposure values for each exposure quartile
is provided in Appendix H. Figure 3.1 depicts tract lead exposure risk quartile throughout for all
census tracts in the contiguous United States.
Table 3.2: Summary Statistics for Lead Exposure Quartile Groups
Quartile Mean (mg/m
3
) Median (mg/m
3
) IQR (mg/m
3
) IQR vs RfC
3
logIQR relative to logSD
Q1 (lowest exposure) 2.5x10
− 4 2.7x10
− 4 1.9x10− 3.3x10
− 4 1.2x to 2.2x RfC -1.45 to -0.80 SD
Q2 5.3x10
− 4 5.2x10
− 4 4.6− 5.9x10
− 4 3.1x to 4.0x RfC -0.42 to -0.12 SD
Q3 8.7x10
− 4 8.6x10
− 4 7.6− 9.8x10
− 4 5.2x to 6.5x RfC 0.16 to 0.45 SD
Q4 (highest exposure) 23.1x10
− 4 16.7x10
− 4 13.4− 23.0x10
− 4 8.9x to 15.3x RfC 0.81 to 1.43 SD
Figure 3.1: Lead Emissions from All Sources Exposure Risk Quartile
3
The EPA’s NATA dataset specifies a reference concentration (RfC) for each analyzed pollutant corresponding to
an imat of a continuous inhalation exposure that is thought to be without an appreciable risk of adverse health effects
over a life For the 2011 NATA assessment, the RfC for lead is 0.00015 mg/m
3
[130].
48
3.2.2.2 Covariates
Individual Covariates from the Health and Retirement Study (HRS)
Age Continuous age in years was centered at the mean age for the study population (64
years) and re-coded to 10-year increments. A four-part categorical age variable was also created
with categories representing age ranges 50-59, 60-69, 70-79, and 80+.
Female Sex The sex variable consists of two levels (female and male) and was not re-coded
for this analysis.
Race and Ethnicity A categorical race/ethnicity variable was created using self-reported
race (White, Black, and other) and Hispanic or Latino ethnicity. The final race/ethnicity variable
consists of four categories: White non-Hispanic/Latino, Black non-Hispanic/Latino, other non-
Hispanic/Latino, and Hispanic or Latino. A table showing the relationship between the original
race and ethnicity variables and the final race and ethnicity variable is provided in Appendix I.
Nativity The nativity variable consists of two levels (US born and foreign born) and was not
re-coded for this analysis.
Educational Attainment Information on individual highest degree obtained was re-coded to
a three-part educational attainment variable with levels for no degree; high school degree, GED, or
equivalent; and some college or more.
Below Poverty Threshold A binary poverty status variable, indicating if the individual is
above or below the poverty threshold was calculated using a ratio of household income to poverty
threshold level variable provided in the HRS RAND data set. Individuals were coded as below
poverty threshold if their household income to poverty ratio was less than 1 and above poverty
threshold if their household income to poverty ratio was equal to or greater than 1.
49
Household Income A four-part categorical household income variable was calculated based
on individual household income quartiles within the study population.
Wealth A four-part categorical wealth variable was calculated based on wealth income quar-
tiles within the study population.
Census Tract Variables from the Health and Retirement Study (HRS)
Urbanicity The original nine-category urbanicity variable was re-coded to six categories:
urban metro with 20,000 or more residents and urban non-metro with 20,000 or more residents
were collapsed into an urban with 20,000 or more residents category; urban metro with fewer than
20,000 residents and urban non-metro with fewer than 20,000 residents were collapsed into an
urban with fewer than 20,000 residents category; rural metro and rural non-metro were collapsed
into a rural category. A table showing the relationship between the original and final urbanicity
variables is provided in Appendix J.
Census Division The census division variable was created based on the Census Regions and
Divisions of the United States using state of residence from the HRS data. A map of the Census
Regions and Divisions of the United States is provided in Appendix K.
Census Tract Variables from the Decennial Census and American Community Survey (ACS)
Percent of population in race/ethnicity groups was derived from the 2010 Decennial Census. This
variable was not re-coded for this analysis.Percent of the adult (age 25 years and older) population
with less than a high school degree, percent of occupied housing units built in 1950 or earlier,
percent of households with female head, percent in poverty, percent unemployed, and median
household income (in 2010 adjusted dollars) were derived from the 2006-2010 ACS 5-year esti-
mates. Median household income was re-scaled to units of $10,000. The other two ACS-derived
variables (percent of adult population without a high school degree and percent of housing built in
1950 or earlier) were not re-coded for this analysis.
50
3.2.3 Data Management
The 2012 HRS waves includes information on 17,545 individuals ages 55 or older representing
6,380 unique census tracts. This data subset was merged with individual data from the HRS RAND
dataset containing information on health behaviors, income, and poverty status used in this anal-
ysis; this step did not remove any observations. The data subset was then merged with the 2011
NATA lead exposure data using survey respondents’ census tract of residence, which removed
299 individuals either living outside the contiguous United States (5 individuals living in Alaska,
Hawaii, or Puerto Rico), missing census tract of residence information (293 individuals), or asso-
ciated with an erroneous census tract code (1 individual). Next, 61 individuals with missing HRS
covariate information were dropped
4
. The data set was also restricted to individuals with non-zero
sample weight but this step did not remove any observations. The HRS sample was then merged
with census tract demographic information from the 2010 Decennial Census and the 2006-2010
American Community Survey (ACS) 5-year estimates using census tract of residence; this step
did not remove any observations. The final sample included 17,185 individuals representing 6,311
unique census tracts; the overall sample selection process eliminated 360 (2.1%) individuals and
69 (1.1%) census tracts. The figure below provides information on the analytic sample selection
process.
3.2.4 Analysis Plan
3.2.4.1 Descriptive Statistics
Following formation of the analysis subsample and merging with other datasets, I first produced
descriptive statistics for each variable including summary statistics for continuous and discrete
variables and frequency information for indicator and categorical variables. All analyses and data
management were conducted using Stata release 17 [131].
4
Of the 61 individuals dropped due to missing covariate information, 14 were missing tract census tract division
only, 20 missing race/ethnicity only, 12 missing nativity only, 3 missing tract percent of 65+ population living in
poverty only, 1 missing tract median household income only, and 1 missing race/ethnicity and nativity.
51
Figure 3.2: Flow chart of the data restriction process using data from the Health and Retirement
Study and National Air Toxics Assessment (NATA).
3.2.4.2 Bivariate Analysis
Next, I carried out correlation tests to determine the bivariate relationship between individual de-
mographic and socioeconomic characteristics with tract lead exposure risk quartile. χ
2
tests were
used to assess the relationship between tract lead exposure risk quartile and indicator or categorical
covariates (race/ethnicity, educational attainment, poverty, income quartile, wealth quartile, urban-
icity, and census division); one-way ANOV A was used to assess the relationship between tract lead
exposure risk quartile and continuous or discrete variables (individual age, tract census and ACS
variables). For continuous or discrete variables which showed a significant relationship with tract
lead exposure quartile based on the initial ANOV A test, I further carried out pairwise comparisons
between lead exposure quartiles using the Tukey’s Honestly Significant Difference (HSD) test. All
analyses and data management were conducted using Stata release 17 [131].
52
3.2.4.3 Multivariable Analysis
Following bivariate analysis, I used logistic regression models to assess the relationship between
individual demographic and socioeconomic characteristics in older adults and census tract of resi-
dence lead exposure risk quartile using a binary version of the lead exposure risk quartile variable
indicating residence in the quartile of tracts associated with the highest lead exposure risk (Q4) vs
residence in a low risk tract (Q1-3). Because individual race/ethnicity and socioeconomic charac-
teristics are most commonly cited in the literature as related to environmental risk and exposure,
model development and interpretation focused on individual race/ethnicity and an indicator of so-
cioeconomic status (poverty status, income quartile, or wealth quartile) as the primary exposure
variables of interest.
Interpretation was focused on comparing the likelihood of individuals from different race/ethnicity
groups to reside in a tract within the highest lead exposure quartile (referred to as “Q4 tract”) as
well as the likelihood of individuals characterized by poverty or lower income or wealth to reside
in a Q4 tract. Individual age category, sex, educational attainment; and census tract urbanicity
classification were included as covariates.
3.3 Results
3.3.1 Study Population Characteristics
The study population consisted of 17,188 individuals. The study population, with survey weights
applied, had a mean age of 67.5 and was 54% female and 79% non-Hispanic White. Study pop-
ulation individuals were distributed roughly evenly between the four lead exposure risk quartiles;
percent of the total number of observations in each quartile ranged from 22.5% (Q4) to 26.7%
(Q2). Survey-weighted information on individual and tract of residence characteristics is provided
in Table 3.3
53
Table 3.3: Demographic characteristics of the study population
Mean (SD) or Frequency (%)
N 17,188
Age 67.5 (9.8)
Gender
Male 45.5%
Female 54.5%
Race/Ethnicity
non-Hispanic White 78.7%
non-Hispanic Black 9.9%
non-Hispanic Other Race 3.2%
Hispanic 8.3%
Nativity
US Born 90.1%
Foreign Born 9.9%
Educational Attainment
No High School Degree 14.4%
High School Degree, GED, or Equivalent 51.8%
Some College or More 33.8%
Poverty Status
At or Above Poverty Threshold 89.3%
Below Poverty Threshold 10.7%
Urbanicity
Very Large Metropolitain (population 1 million or more) 49.7%
Large Metropolitain (population 250,000 to less than 1 million) 23.1%
Small Metropolitain (population less than 250,000) 7.8%
Large Urban (population 20,000 or more) 8.6%
Small Urban (population less than 20,000) 10.0%
Rural 0.8%
Census Division
New England 4.9%
Mid-Atlantic 12.6%
Northeast Central 16.6%
Northwest Central 8.1%
South Atlantic 21.0%
Southeast Central 6.6%
Southwest Central 9.7%
Mountain 6.9%
Pacific 13.8%
Population Density (population per square mile) 5,221 (10,701)
Percent 55+ 27.7% (10.3%)
Percent 65+ 15.0% (8.3%)
Percent 75+ 7.0% (5.0%)
Percent White 76.1% (23.8%)
Percent Black 11.9% (20.5%)
Percent Non-Hispanic White 68.1% (28.8%)
Percent Non-Hispanic Black 11.5% (20.2%)
Percent Hispanic 14.3% (20.8%)
Percent of Households with Female Head 12.9% (7.2%)
Percent 18+ in Poverty 11.4% (9.0%)
Percent 18-64 in Poverty 11.8% (9.7%)
Percent 65+ in Poverty 9.4% (9.3%)
Percent 25+ with No High School Degree 14.4% (11.2%)
Percent Unemployed 8.0% (4.7%)
Per-Capita Income 28,588(14,272)
Median Household Income 59,589(27,822)
Percent Vacant Housing 9.9% (8.2%)
Percent of Housing Build 1950 or Earlier 30.4% (25.9%)
54
3.3.2 Bivariate Analyses with Tract of Residence Lead Exposure Quartile
Bivariate analyses testing for significant differences indicated that all variables, with the exception
of gender, were significantly different between the four quartiles of lead exposure risk. Distribu-
tions of each of the tested variables by lead exposure risk quartile as well as significance results
for ANOV A or chi2 tests are provided in Table 3.4.
3.3.2.1 Individual Demographic Characteristics
Among individual demographic characteristics, race/ethnicity and nativity had strong relationships
with lead exposure risk quartile in bivariate analysis. Older adults characterized as non-Hispanic
Black, non-Hispanic other race, Hispanic, or foreign born were more likely to live in a Q4 tract
compared to lower-exposure risk tracts while non-Hispanic White and US-born older adults were
less likely to live in a Q4 tract compared to lower exposure risk tracts. Non-Hispanic Black,
Hispanic, and non-Hispanic other races had 1.7, 1.9, and 1.6 greater odds of residing in a Q4
tract compared to non-Hispanic Whites. Overall, non-Hispanic/Latino White individuals were less
likely than expected to live in a Q4 tract while non-Hispanic/Latino Black, Hispanic/Latino, and
non-Hispanic/Latino other race were more likely than expected to live in a Q4 tract.
Foreign born older adults had 1.7 greater odds of residing in a Q4 tract compared to US born
older adults. Figure 3.3 shows the relationship between lead exposure quartile and the ratio be-
tween the observed proportion of race/ethnicity in each tract quartile compared to the expected
proportion of race/ethnicity in that quartile for each of the four individual race/ethnicity groups.
Expected count of individuals of a given race/ethnicity in a given tract exposure quartile was cal-
culated using Equation 3.1. Age had a small, but significant, negative relationship with higher
exposure risk tracts. Gender did not show a significant relationship with exposure risk.
n
xy
N∗ n
x
∗ n
y
=observed overexpected forrace/ethnicityxinexposurequartilex (3.1)
55
Table 3.4: Demographic and tract characteristics by lead exposure quartile and results of tests of
bivariate association
Tract Lead Exposure Risk Quartile
P-Value
Q1
(Lowest Lead)
Q2 Q3
Q4
(Highest Lead)
N 4,145 4,595 4,583 3,865
Age 68.1 (9.5) 67.3 (9.8) 67.4 (9.9) 67.3 (9.9) 0.0002
Gender 0.721
Male 44.0% 46.1% 45.6% 46.1%
Female 56.0% 53.9% 54.4% 53.9%
Race/Ethnicity < 0.0005
non-Hispanic White 80.5% 83.6% 78.8% 70.6%
non-Hispanic Black 5.4% 9.5% 11.5% 13.3%
non-Hispanic Other Race 3.2% 2.0% 3.5% 4.1%
Hispanic 10.8% 4.9% 6.2% 12.0%
Nativity < 0.0005
US Born 92.5% 92.3% 89.4% 85.9%
Foreign Born 7.5% 7.7% 10.6% 14.1%
Educational Attainment < 0.0005
No High School Degree 16.8% 12.1% 11.8% 17.7%
High School Degree, GED, or Equivalent 52.9% 52.1% 48.7% 54.1%
Some College or More 30.3% 35.8% 39.5% 28.2%
Poverty Status < 0.0005
At or Above Poverty Threshold 88.4% 91.3% 90.2% 86.7%
Below Poverty Threshold 11.5% 9.0% 9.9% 12.8%
Urbanicity < 0.0005
Very Large Metropolitan (1 million +) 21.7% 46.2% 61.8% 69.6%
Large Metropolitan (250,000 to < 1 million) 30.0% 28.8% 21.0% 11.4%
Small Metropolitan (< 250,000) 7.9% 9.1% 8.1% 5.8%
Large Urban (20,000 +) 14.7% 6.8% 5.9% 7.3%
Small Urban (< 20,000) 23.0% 8.7% 3.1% 6.0%
Rural 2.7% < 1% < 1% 0%
Census Division < 0.0005
New England 3.9% 9.3% 4.5% 1.0%
Mid-Atlantic 8.0% 6.5% 18.3% 17.4%
Northeast Central 6.5% 19.9% 21.3% 17.5%
Northwest Central 11.2% 10.8% 4.8% 5.5%
South Atlantic 26.1% 23.6% 21.4% 13.7%
Southeast Central 5.0% 8.5% 7.0% 5.4%
Southwest Central 17.6% 9.4% 5.0% 6.9%
Mountain 11.1% 2.9% 5.8% 8.1%
Pacific 10.7% 9.1% 11.9% 24.4%
Tract Demographic Characteristics
Percent White 83.5% (17.9%) 79.3% (21.4%) 73.3% (24.9%) 67.5% (27.1%) < 0.00005
Percent Black 7.0% (13.7%) 11.2% (19.0%) 14.0% (22.4%) 15.3% (24.5%) < 0.00005
Percent Non-Hispanic White 72.9% (27.3%) 73.7% (24.9%) 66.7% (28.2%) 57.7% (32.4%) < 0.00005
Percent Non-Hispanic Black 6.8% (13.6%) 10.9% (18.8%) 13.6% (22.0%) 14.8% (24.2%) < 0.00005
Percent Hispanic 15.3% (24.3%) 10.1% (15.2%) 12.6% (17.1%) 20.0% (24.6%) < 0.00005
Tract Socioeconomic Characteristics
Percent Households with Female Head 11.6% (5.9%) 12.1% (6.5%) 13.0% (7.4%) 15.2% (8.3%) < 0.00005
Percent 18+ in Poverty 12.9% (9.1%) 10.0% (8.1%) 10.1% (8.3%) 13.0% (10.2%) < 0.00005
Percent 18-64 in Poverty 13.4% (9.5%) 10.3% (8.7%) 10.4% (9.0%) 13.6% (10.9%) < 0.00005
Percent 65+ in Poverty 10.9% (10.3%) 8.7% (8.5%) 8.1% (8.0%) 10.2% (10.1%) < 0.00005
Percent 25+ without High School Degree 15.5% (11.4%) 13.0% (9.9%) 12.5% (9.9%) 17.2% (13.1%) < 0.00005
Percent Unemployed 7.9% (4.5%) 7.8% (4.6%) 7.8% (4.5%) 8.7% (5.2%) < 0.00005
Per-Capita Income $26,001 ($12,452) $29,289 ($12,452) $32,045 ($16,677) $26,428 ($14,106) < 0.00005
Median Household Income $52,377 ($22,674) $63,367 ($27,698) $66,693 ($32,350) $54,406 ($24,065) < 0.00005
Percent Vacant Housing 14.5% (11.9%) 8.9% (7.2%) 7.9% (4.9%) 8.5% (5.3%) < 0.00005
Percent Housing Build≤ 1950 24.3% (20.5%) 25.6% (23.1%) 34.6% (28.5%) 37.9% (28.2%) < 0.00005
56
Figure 3.3: Figure: The ratio of observed over expected number of individuals in the study popu-
lation residing in a Q1, Q2, or Q3 tract compared to a Q4 tract by individual race/ethnicity. The
expected population in each tract group by race/ethnicity was calculated by multiplying the percent
of the entire study population in a given lead quartile exposure tract group by the percent of the
entire study population in a given race/ethnicity group. If no relationship between race/ethnicity
and lead exposure quartile existed, all ratios would be 0%.
where:
N = total sample population
n
x
= count of sample population of race/ethnicityx
n
y
= count of sample population living in lead exposure quartiley
n
xy
= count of sample population of race/ethnicityx living in lead exposure quartiley
3.3.2.2 Individual Socioeconomic Characteristics
Socioeconomic variables including the educational attainment and poverty status variables tended
to have a U-shaped relationship with lead exposure tract quartile. Older adults living in Q1 and
57
Q4 tracts were more likely to have less educational attainment and more likely to be in poverty
compared with older adults living in Q2 and Q3 tracts. Older adults who had not completed high
school had 1.4 greater odds of living in a Q4 tract compared to older adults with more education;
older adults who had completed high school had a 1.3 greater odds of living in a Q4 tract compared
to older adults with some college or more. Older adults in poverty had 1.4 greater odds of living in
a Q4 tract compared to older adults living at or above the poverty line.
3.3.2.3 Tract Characteristics
Lead exposure risk quartile was also strongly related to urbanicity; older adults living in more
metropolitan and more populated areas were much more likely to live in a higher exposure lead
tract. This is expected as many lead-producing institutions and industries, such as airports and
factories, are more likely to be located in densely populated urban areas.
Tract-level demographic and socioeconomic bivariate relationships with tract exposure risk
were broadly similar to relationships between individual-level demographic characteristics and
lead exposure risk. Tracts with higher lead exposure risk were more likely to have a larger percent
of the population characterized as non-Hispanic Black, a smaller percent of the population charac-
terized as non-Hispanic White, and a larger percent of the adult (age 25+) population with no high
school degree. Again, the percent of the population characterized as Hispanic and tract variables
related to poverty, including the percent of the population in poverty median household income,
and per-capita income had roughly U-shaped relationships with lead exposure risk with both Q1
and Q4 exposure tracts associated with indicators of greater poverty and lower income. Finally, the
percent of occupied housing built in 1950 or earlier had a positive relationship with lead exposure
risk quartile.
58
3.3.3 Ancillary Analysis
Individual Race/Ethnicity Stratified by Tract Urbanicity
Because lead exposure risk and urbanicity were highly related, the this relationship is likely to be
an underlying component of the relationship between lead exposure risk and certain demographic
and socioeconomic characteristics which are also associated with urbanicity, such as the individual
and tract race/ethnicity, nativity, income, and other variables, as certain racial and ethnic groups,
as well as foreign-born individuals, are more likely to live in urban areas than non-urban areas.
This may mean that the underlying exposure-urbanity relationship may have resulted in a spurious
connection between certain demographic and socioeconomic variables and lead exposure risk.
After stratifying by urbanicity, the relationships between lead exposure risk quartile and indi-
vidual/race ethnicity were broadly consistent between each of the urbanicity groups. Non-Hispanic
Black individuals demonstrated a consistently positive relationship with higher lead exposure quar-
tiles. This relationship appears to be more pronounced for smaller urban areas compared to metro
areas although, due to limited data for higher lead quartiles for smaller urbanicities, it is difficult to
determine the relationship for all lead exposure quartiles. Similarly, the relationship between His-
panic individuals and lead exposure quartile remained roughly U-shaped for the two major metro
urbanicities with the Hispanic population being greater than expected for Q1 and Q4 but less than
expected for Q2 and Q3. Finally, non-Hispanic white individuals showed roughly the expected per-
cent of the population regardless of urbanicity and lead exposure quartile but with slightly lower
than expected percent of the population for Q4. There was insufficient information on individuals
characterized as non-Hispanic other race to perform an urbanicity stratified analysis for this group.
3.3.4 Multivariable Analysis
Individual race/ethnicity and educational attainment were significantly associated with likelihood
of living in a Q4 (highest lead exposure) tract in final full models. Black non-Hispanic/Latino,
Hispanic/Latino, and other non-Hispanic/Latino older adults had 1.29 (95% CI: 1.17-1.42), 1.62
59
(1.44-182), and 1.41 (1.15-1.73) greater odds of living in a Q4 tract compared to White non-
Hispanic/Latino older adults. Older adults with only a high school degree or equivalent, but no
college, had 1.28 greater odds of living in a Q4 tract compared to older adults with some college
or more while older adults with no high school degree had a 1.41 greater odds of living in a Q4
tract compared to older adults with some college or more.
Adjusting for educational attainment reduced the effect of race/ethnicity for Black non-Hispanic
/Latino and Hispanic/Latino older adults but not other race non-Hispanic Latino older adults.
There was no statistically significant associations between individual socioeconomic character-
istics (poverty, income quartile, or wealth quartile) and neighborhood lead exposure risk. As all
three socioeconomic status variables had a similar effect on model results.
As expected, tract-level urbanicity classification had a significant effect on likelihood of living
in a Q4 tract. Older adults living in a metropolitan area with 1 million population or more had 2.08
to 3.13 greater odds of living in a Q4 tract compared to smaller metropolitan areas or urban areas.
Non tracts classified as “rural” were also classified as a Q4 tract. Adjusting for urbanicity reduced
the effect of race/ethnicity and magnified the effect of educational attainment across all groups.
Table 3.5 below provides results for these selected models. Additional tested models are provided
in Supplement.
3.4 Conclusions and Discussion
3.4.1 Summary of Findings
Overall, older adults living in a Q4 exposure risk tract tended to also be characterized by other
indicators of socioeconomic vulnerability including living in poverty and having less educational
attainment as well as demographic characteristics associated with vulnerable populations such as
foreign born individuals and persons of color. This is in agreement with the literature which has
consistently found a disproportionate burden of environmental contaminants on communities of
60
Table 3.5: Results of dichotomous logistic regression models examining the relationship between
demographic and socioeconomic characteristics of older adults and likelihood of living in a high
lead exposure risk tract, adjusting for relevant covariates
Base
Race/Ethnicity
Only
Race/Ethnicity
and Age Only
Race/Ethnicity,
Age, and
Education Only
Full Model
M0 M1 M2 M3 M4
N 17188 17188 17188 17188 17188
Residence in a Q4 Tract (vs Q1-3) 0.31 *** 0.25 *** 0.23 *** 0.20 *** 0.33 ***
White Non-Hispanic/Latino - - - -
Black Non-Hispanic/Latino 1.67 *** 1.63 *** 1.58 *** 1.29 ***
Hispanic 2.05 *** 1.99 *** 1.88 *** 1.62 ***
Other Race Non-Hispanic/Latino 1.69 *** 1.64 *** 1.63 *** 1.41 **
Age 50-59 1.26 *** 1.30 *** 1.23 ***
Age 60-69 1.07 1.10 + 1.09 +
Age 70-79 - - -
Age 80+ 1.10 + 1.10 1.05
No High School Degree 1.25 *** 1.41 ***
High School Degree or Equivalent 1.21 *** 1.28 ***
Some College or More - -
Household Income Quartile 1 (lowest income) -
Household Income Quartile 2 0.95
Household Income Quartile 3 0.92
Household Income Quartile 4 (highest income) 0.89 +
Metro 1mill + -
Metro 250k to < 1 mill 0.33 ***
Metro < 250k 0.39 ***
Urban 20k + 0.48 ***
Urban < 20k 0.32 ***
Rural 1
+p <0.10 *p <0.05 **p <0.01 ***p <0.001
61
color as well as communities characterized by vulnerabilities such as poverty and low social or
socioeconomic mobility.
Tract exposure quartile was also highly correlated with urbanicity class, most notably older
adults living in metropolitan areas with 1 million population or more were much more likely to
live in a Q4 tract compared to older adults in other urbanicities. This is not surprising considering
environmental exposure risk is often associated with large metropolitan areas but it does potentially
contribute to the relationships between socioeconomic and demographic characteristics and lead
exposure mentioned in the previous paragraph.
Logistic regression model results broadly agreed with the results and conclusions from bivari-
ate analysis and indicated that persons of color, and individuals with lower educational attainment
are more likely to live in high exposure tracts. Interestingly, however, in contrast to bivariate anal-
ysis outcomes, logistic regression analysis indicated that individuals in poverty or in lower income
or wealth brackets are no more likely to live high exposure tracts than more economically advan-
taged individuals if other demographic characteristics are taken into account. In other words, an
individual’s race/ethnicity and educational attainment remain the key indicators of the likelihood
of neighborhood exposure risk tract regardless of individual socioeconomic characteristics. In ad-
dition, although educational attainment was significantly associated with tract quartile, adding the
educational attainment variable only slightly reduced the magnitude of the effect of race/ethnicity
on tract exposure risk quartile for non-Hispanic/Latino Black and Hispanic/Latino individuals and
did not reduce the effect for non-Hispanic/Latino other race individuals. This finding indicates
that, although educational attainment is related to likelihood of living in a high risk tract, it, at
most, only slightly alters the effect of race/ethnicity on that likelihood.
3.4.1.1 Individual Race/Ethnicity Interaction with Tract Urbanicity
After stratifying for urbanicity classification, the strong relationship between being non-Hispanic
Black or Hispanic and odds of living in a Q4 tract relative to non-Hispanic Whites was retained,
62
indicating that the relationship between exposure risk and race/ethnicity is consistent across ur-
banicities. However, it should be noted that due to small sample sizes, this analysis could not be
carried out for all race/ethnicity categories for all urbanicity classifications.
3.4.2 Comparison with Previous Literature
As mentioned above, the relationships described in this analysis are broadly in agreement with
previous work exploring the relationship between population characteristics and risk of exposure
to environmental contaminants. The literature has established that communities of color and com-
munities characterized by socioeconomic vulnerabilities are disproportionately exposed to envi-
ronmental contaminants while this study indicated that older adults who self identified as persons
of color or who were foreign born, in poverty, or characterized by lower educational attainment
were more likely to live in the highest risk areas.
3.4.3 Implications and Significance
This paper represents the first analysis to explore the relationship between risk of exposure to air-
borne lead emissions and individual characteristics in older adults in the United States. In addition,
the sample of older adults used in this analysis is nationally representative and therefore the results
may be applicable to the entire population of older adults in the United States, rather than a geo-
graphically local or demographically limited sample which may have limited applicability to other
populations.
3.4.4 Limitations
While the HRS sample used in this analysis is one of the largest samples used in studies explor-
ing the relationship between lead exposure and cognition and is also nationally representative,
the race/ethnicity subsamples have less statistical power, especially for non White non-Hispanic
individuals. More work should be carried out focusing on persons of color with larger national
63
samples to further explore the confounding effect of race and ethnicity on the relationship between
neighborhood lead exposure risk and cognitive changes in older adults.
As with the race/ethnicity subgroup analysis, the HRS sample used for this study had lim-
ited statistical power for certain urbanicity groups and too few individuals were represented in
the metropolitan with less than 250,000 population; urban with 20,000 population or more; urban
with less than 20,000 population; and rural urbanicity categories. The fact that the two groups
which had a sufficient sample to be studied produced different results indicates that different ur-
banicity regions may moderate the relationship between lead exposure risk and cognitive outcomes
for older adults. More research should be carried out with additional study subjects with greater
representation for less-populated urbanicity categories.
An additional major limitation of this work is the inability to include variables correspond-
ing to individual or community residential housing age, an indicator of potential lead-based paint
presence. Historical lead-based paint use, and the residual concentrations of lead-based paint still
present in many homes in the United States, remains one of the most significant sources of lead
exposure. Unfortunately, the HRS dataset does not include information on age or building year
of participant residences and therefore this control variable could not be included in the analysis.
Future work should be carried out accounting for potential residential lead-based paint or other
potential major sources of lead exposure.
In addition, because the two sub-analyses, on race/ethnicity and urbanicity, both produced
interesting results, additional work should be carried out on a study population which allows for
further dissection by both race/ethnicity and urbanicity. The sample used for this analysis did
not have a sufficiently large study population to be able to carry out this subsequent analysis. In
addition, race/ethnicity and urbanicity were somewhat correlated and some intersections of these
two categorical variables included very few individuals.
64
Chapter 4
Estimated Neighborhood Airborne Lead Exposure Risk and
Cognitive Functioning in Older Adults
4.1 Introduction
Lead exposure is a pervasive environmental health issue that has been linked to numerous adverse
health effects in humans, including cognitive impairment [25, 29–31]. While the health risks of
lead exposure have been extensively studied in children [32–35], less is known about the effects
of lead exposure on older adults [36–77], who may be more susceptible to the health effects of
lead due to age-related declines in organ function and physiological resilience as well as higher
prevalence of preexisting health conditions. Cognitive health may be especially important for
older adults exposed to lead as it is hypothesized that a significant portion of the cognitive decline
associated with the “natural” aging process may be due to chronic lifetime lead burden accumulated
over time through various environmental exposures [54, 67, 104]. Given that older adults are
both the fastest growing sub-population in the United States and have arguably been exposed to
some of the highest chronic lifetime lead exposure levels in United States history, it is essential to
understand the causes and sources of potential historical and ongoing lead exposure in the older
adult population.
Prior research has shown that lead exposure is significantly associated with adverse changes in
overall cognition as well as in specific cognitive domains in older adults. Despite this, there remain
65
significant gaps in our understanding of the relationship between lead exposure and cognition in
older adults which this paper seeks to address such as the effect of contemporary airborne lead
exposures, the impact of neighborhood-level lead exposures, and applicability to the general US
population of older adults. The overall goal of this study is to address these gaps by determining
the relationship between neighborhood-level exposure to airborne lead emissions and cognitive
decline in a nationally-representative sample of older adults in the United States.
This project’s objective is to identify the association between neighborhood lead emission ex-
posure risk and performance on a test of overall cognitive status. Specifically, I hypothesize that
older adults living in census tracts with high airborne lead exposure risk will perform more poorly
on cognition tests and will experience a more rapid decline in cognition over time compared to
older adults with lower neighborhood lead exposure risk. This work contributes to the literature in
multiple ways: 1) uses both a nationally representative population of older adults and a national
lead emissions exposure risk data set, which have not previously been used together to study the
impact of a specific lead exposure on cognition outcomes in older adults, 2) relates a source of
exposure directly to cognitive outcomes, and 3) explores the effect of exposure to airborne lead,
which is an under-studied source of lead exposure and, to the knowledge of the author, has not
been studied in the context of cognitive outcomes in older adults. In addressing these gaps, this
paper seeks to develop a more nuanced understanding of the risks associated with lead exposure in
older adults and develop evidence-based interventions to reduce these risks and improve the health
outcomes of this vulnerable population.
This project uses data from three waves spanning six years of the Health and Retirement Study
(HRS), a national population-based sample of older adults, and emissions exposure data from the
2011 EPA National Air Toxics Assessment (NATA) as well as census tract demographic variables
from the 2010 United States decennial census and the 2006-2010 American Community Survey
(ACS) 5-year estimates.
66
4.2 Background
Lead is a toxic heavy metal with well-established adverse health effects for humans associated with
both acute and chronic exposure. Although regulations implemented during the past few decades
have significantly reduced many sources of environmental lead in the United States, lead exposure
remains a significant public health concern. Legacy sources of lead, namely lead-based paint in
buildings constructed prior to 1978 [78] and leaded pipes in buildings constructed prior to 1986
[79], continue to be sources of lead exposure even decades following regulatory action. Previous
work has estimated that approximately 24 million homes in the United States have hazardous levels
of lead-based paint or dust [78]. However, although many sources of lead have been regulated and
mitigated, many contemporary sources of new lead exposure exist including air pollution from
industrial activities and exhaust from airplane fuel. Both legacy sources of lead and contemporary
primary sources of lead contribute to contaminated soil, dust, water, and other media.
Once deposited, lead does not degrade in the environment and can reside in the soil and other
media for multiple decades following dispersion [5, 6, 80, 81]. Residual lead particles have the
potential to re-suspend into the air, for example during weather events or construction activity,
which in turn may contribute to re-exposure or contamination of new media even years following
initial release into the environment [6, 80].
4.2.1 Lead Biokinetics and Toxicity
Lead can enter the body through multiple pathways including ingestion, inhalation, skin contact, or
through the placenta [8–15]; the most common routes of exposure are through the gastrointestinal
or respiratory tract [65, 84]. Once absorbed, lead is transported throughout the body by the blood
and has the potential to impact multiple bodily organs and systems. While lead in the blood has a
half-life of approximately 25 to 35 days, lead may be absorbed by and accumulate in certain soft
tissues, bone, and teeth where it can remain for multiple decades [85, 86]. Multiple studies have
identified that lead retention in the bone is cumulative throughout the lifespan to at least age 50
67
with evidence of continued accumulation through the 70s [87–89]. The bones effectively become
a reservoir of lead in the body as lead may release back into the blood during bone remodeling or
deterioration. This process re-exposes the individual to the potential toxic effects of blood lead,
causing adverse health effects to occur even decades following exposure [85], a process which is
especially significant for older adults who are experiencing bone deterioration due to osteoporosis.
As the brain is the principal target organ of lead exposure [84], neurotoxicological effects
are the most commonly cited and studied outcomes of lead in the body. Most importantly the
disruption of neuronal signaling (calcium signaling) which alters neurotransmitter release and im-
pairs synaptic plasticity [65, 132] can adversely impact cognitive function. Lead can also lead to
oxidative stress, which can damage cellular components or cause neuronal apoptosis, as well as
inflammation, which can further exacerbate neuronal damage in the brain, the effects of which may
also negatively impact cognitive function [61, 65, 132].
4.2.2 Lead and Older Adults Literature Review
For older adults, lead exposure has been associated with faster declines in overall cognitive func-
tioning [57, 60–62, 90], adverse changes in the brain [41, 52, 65, 91–93], and increased prevalence
of Alzheimer’s Disease and other dementias [90, 95, 96]. Outcomes associated with specific cog-
nitive domains include declines in executive function [38, 40, 54, 110], manual dexterity [40, 110],
verbal memory and learning [38, 54], visual memory [40, 54], hand-eye coordination [56], pro-
cessing speed, attention, and others [39, 54, 110]. Notably, as mentioned in the introduction, it has
been hypothesized that, for adults who were exposed to lead in their lifetime, a significant amount
of the decline in cognitive functioning previously considered part of the natural aging process may
be attributed to lifetime lead exposure [54, 67, 104]. In addition to neurocognitive outcomes, lead
exposure is associated with many other health outcomes in older adults including hypertension
[97] cardiovascular disease [98], chronic kidney disease [98], osteoporosis [98–100], frailty [101,
102], and cataracts [103].
68
Contemporary cross-sectional and longitudinal work focused on the relationship between lead
exposure and health in adults and older adults began in the mid-1980s. Prior to this point, almost all
work studying the health effects of lead exposure focused on childhood exposure, especially early
childhood exposure, and childhood health outcomes. However, the current cohort of adults in the
United States are likely to have been exposed to multiple sources of lead exposure throughout their
lives and have been exposed to more lead during the course of their lifetime than perhaps any other
generation in history. Older adults today lived their childhood and young adulthood during the
peak years of residential lead-based paint from the beginning of the 20th century through the 1940s
and peak years of leaded gasoline use from the 1950s through 1980s. By the time contemporary
federal regulations were placed on lead in paint in 1978 and gasoline in 1996, today’s older adults
generation were already in their mid- to late-adulthood.
Much of the work which has been carried out on the relationship between lead exposure and
cognitive outcomes in older adulthood has focused on occupational exposure. The earliest work
studying lead exposure and cognitive health among older adults focused on occupationally-exposed
adults and established that occupational lead exposure is associated with adverse cognitive out-
comes and other health issues [36–40, 47, 58, 66–77]. In addition, many of these studied compared
blood lead or bone lead levels between a cohort of occupationally exposed adults or older adults
and a comparison cohort of adults or older adults who had not experienced occupational expo-
sure but were similar in other demographic characteristics such as age, gender, race and ethnicity,
and geographic location and may be considered to have had similar levels of lead exposure from
non-occupational sources during their lifetimes. The findings from these studies further indicate
that a lead exposure experienced as an adult may compound on the biological and health effects
of any previously experienced lead exposure from other sources and, furthermore, that remaining
sources of lead may continue to exacerbate health outcomes in older adults who are exposed to
those sources today. Although lead exposure experienced in the past cannot be undone, remaining
sources of lead may be mitigated and better regulated. The past research indicates that this mitiga-
tion and regulation may not only protect future generations from the health outcomes from lifetime
69
exposure, but, by eliminating any additional exposure sources, may also stymie health decline in
today’s older adults.
Although the findings from work focused on occupationally exposed adults may not be directly
applicable to adults who did not experience occupational lead exposure during adulthood, these
studies establish that adulthood exposure to lead, not only childhood exposure, is associated with
adverse health issues, including adverse impacts on cognitive function, in older adulthood.
Although much previous work has focused on individual-level sources of lead exposure, such as
residential lead-based paint or occupational exposure, recent work using data from the Health and
Retirement (HRS) study established a relationship between living in a city with lead pipes as a child
- a neighborhood-level exposure source - and worse cognitive outcomes in older adulthood [133,
134]. In addition, the authors found that the relationship between neighborhood lead exposure
and cognitive outcomes was partially explained by individual educational attainment, indicating
an inter-relationship between neighborhood lead exposure, individual social characteristics, and
cognitive outcomes in older age.
the importance of lifetime environmental lead exposure on cognition in older adults. For ex-
ample, Lee, Lee, Warren, and colleagues found a significant relationship between living in a city
associated with lead plumbing as a child and worse cognitive function outcomes in older adulthood
[133, 134]. However, most existing work has focused on the impact of occupational exposure or
childhood exposure on cognitive outcomes in older age. In addition, most research has been framed
in terms of exposure to
For example, between 1993 and 2007, Stewart, Schwartz, and colleagues conducted multiple
studies following cohorts of adults who were or had been occupationally exposed to lead. They
collected information on bone and blood lead as well as performance on a battery of tests on
multiple specific cognitive domains. The authors found that, both before and after adjusting for
demographic and other covariates, bone lead, indicative of lifetime or cumulative lead exposure,
was negatively associated with all cognitive domains assessed in the study. Based on their results,
the authors hypothesized that the difference in bone level equal to the IQR of bone lead values
70
observed in their work corresponded to cognitive declines equal to roughly 5 years of aging. They
further hypothesized that health outcomes related to lead exposure are likely a major component of
the cognitive decline usually considered part of the “natural” aging process and that lead’s impacts
on the brain are at least persistent if not progressive over time [40, 52, 67, 104].
While the findings from these studies were significant, the results had limited potential to be
applied to members of the general population as the majority of older adults did not experience
occupational lead exposure and older adults aged 65 or older that were occupationally exposed in
the past are not likely to still be employed.
However, the majority of older adults did not experience occupational lead exposure and the
general non-occupationally exposed US adult population is likely to have experienced lead ex-
posure from multiple past and current sources. Recent work on health outcomes associated with
lifetime chronic exposure representative of the general population of non-occupationally exposed
adults [40–44, 46, 48–57, 59–65, 92] has determined that the chronic, lifelong, but relatively low-
level exposure characteristic of the general adult population, heretofore referred to as “environ-
mental exposure” in this document, is also associated with cognitive impairment, reduced IQ,
behavioral disorders, and many other health issues [40–44, 46, 48–57, 62–65, 92].
For example Stewart, Schwartz, and colleagues conducted additional work focused on the ef-
fects of environmental lead exposure, again as measured by bone lead, and changes in cognitive
domains. They again found that, even after adjusting for covariates, bone lead was consistently
inversely associated with worse outcomes for all measured cognitive domains [72, 104]. Addition-
ally, Weisskopf et al. 2004 determined that bone lead was inversely associated with performance
on a test of general cognitive functioning for a cohort of older men [48].
Epidemiological work on this topic has also generally utilized blood lead (representing recent
exposure within approximately one month) and/or bone lead content (representing long-term ex-
posure up to multiple decades) as biomarkers of past lead exposure. While previous work has
established strong connections between lead biomarkers and cognitive outcomes in older adults,
71
bone or blood lead lacks specificity of source and therefore has limited potential to inform an un-
derstanding of the relationship between specific sources of lead exposure and exposure risk. The
differential impact of different sources may be especially important when estimating outcomes and
determining mitigation strategies for populations such as current older adults in the United States
who have and continue to experience lead exposure from multiple sources.
In addition, the discussion in the literature on environmental lead exposure and cognitive
changes in older adults has often been framed in terms of legacy sources, such as lead-based paint
or leaded pipes, and as exposures experienced at the individual level, such as in the home or in an
occupational setting.
However, although major sources of lead were phased out or minimized during the second
half of the 20th century, ongoing sources of lead new exposure remain, such as lead emissions
from industrial practices, mining, or airplane fuel combustion. In the absence of robust regulations
these sources continue to contribute to environmental lead contamination, in turn contributing to
increased exposure risk, unlike more high-profile sources of lead, such as lead-based paint and
leaded gasoline, which have been regulated and are no longer adding to total contamination.
These examples, and most contemporary sources of new lead exposure, are experienced at the
neighborhood level, rather than the individual level, impacting entire neighborhoods populations
regardless of individual behavior or choices. Previous work has shown that neighborhoods mat-
ter for health outcomes [135–137] and, while past research has examined other pollutants at the
neighborhood level, little is known about how exposure to ongoing neighborhood lead emissions
impact cognitive decline during the aging process.
Both the lack of work focused on neighborhood-level exposures and the lack of source speci-
ficity in existing work on lead exposure and cognition in older adults limit the ability of policy
makers and other stakeholders to produce effective mitigation strategies for remaining sources of
lead which are targeted at specific sources or which identify neighborhoods with the highest risk.
In addition, previous work studying older adults has focused on relatively small, local cohorts
of older adults or national samples which are not representative of the general population, such as
72
veterans (all male) in the V A Aging Study. In addition, previous work has been criticized for under-
representing women and people of color. These limitations of the literature limit the applicability
of findings to the general population of older adults.
4.2.3 Project Objectives
The main objective of this study is to characterize the relationship between living in a neighbor-
hood (census tract) characterized by a high risk of exposure to airborne lead emissions, heretofore
referred to as “high exposure neighborhood” or “high exposure tract”, and change in overall cog-
nitive function over time in older adults. Specifically, this project will assess if older adults living
in census tracts in the highest risk quartile (“high risk tract”) experienced a more rapid decline in
overall cognition between 2012 and 2016 compared to older adults living in lower risk tracts.
I hypothesize that older adults living in census tracts in the highest risk group for airborne lead
exposure will perform more poorly on cognition tests and will experience a more rapid decline
in cognition over time, compared to older adults living in neighborhoods characterized by lower
exposure risk. This work contributes to the literature in three key ways: 1) uses airborne lead as the
main effect, which is an under-studied source of lead exposure experienced at the neighborhood
level, 2) uses a nationally representative population of older adults with an exposure data set also
at the national scale, which has not been used for lead exposure and cognition outcomes in adults,
3) focuses on a specific source, rather than using bone lead which lacks source specificity.
4.3 Methods
4.3.1 Data Sources
Annual lead emissions data is derived from the 2011 EPA National Air Toxics Assessment (NATA)
[116]. Information on the study population, including overall cognition based on TICS assessment
performance, is derived from the 2012, 2014, and 2016 Health and Retirement Study (HRS) survey
73
waves [117]. Additional census tract characteristics are derived from the 2010 decennial census
[119] and 2006-2010 5-year ACS summary data sets [118].
4.3.1.1 National Air Toxics Assessment (NATA)
The National Air Toxics Assessment (NATA), established by the Environmental Protection Agency
(EPA), produces exposure and health risk estimates caused by the approximately 187 hazardous
pollutants (HAPs) and diesel particulate matter (PM) [120]. An updated NATA data set was re-
leased every 3-6 years between 1996 and 2014 and has been released annually beginning with 2017
data [121–128].
Using a battery of several statistical models processing daily input from air monitors across
the United States, NATA data products include information on both estimated total exposure for
each pollutant, measured in units per square meter, as well as estimated human exposure from
specific source categories (e.g., airports, major point sources) for each pollutant for all census
tracts in the contiguous United States (excludes Alaska and Hawaii). The NATA dataset also
specifies a reference concentration (RfC) for each analyzed pollutant corresponding to an “estimate
of a continuous inhalation exposure that is thought to be without an appreciable risk of adverse
health effects over a lifetime [130].” For the 2011 NATA assessment, the RfC for lead is 0.00015
mg/m
3
[130]. For a short review of previous work linking NATA exposure data with human health
outcomes, please see Appendix G.
This project uses data on exposure to lead or lead compounds at the census tract level from the
2011 NATA data release. Although more recent NATA assessments were carried out for the years
2014, 2017, 2018, and 2019, this study uses the NATA 2011 data set as an indicator of exposure to
past lead exposure for older adults surveyed between 2012 and 2016. Because the NATA modeling
process underwent a major change beginning with the 2011 assessment and expanded to incorpo-
rate many additional models in the estimation process, data from earlier NATA assessments (1996,
1999, 2002, and 2005) were not utilized for this project.
74
4.3.1.2 Health and Retirement Study (HRS)
The Health and Retirement Study (HRS) is a nationally representative, longitudinal study of U.S.
adults over age 50 carried out by the University of Michigan Institute for Social Research and
funded by the United States National Institute on Aging (NIA) and Social Security Administra-
tion (SSA) [117]. The study began in 1992 and surviving respondents have been re-interviewed
approximately every two years since 1992. Every six years a new cohort of adults ages 51-56 is
added to maintain a steady state design and so the survey continues to be representative of the
population over age 50. The survey collects data on many topics related to health and aging, in-
cluding cognitive functioning, as well as demographic and economic information. Census tract of
residence is determined based on the respondent’s residential address at the time of interview and
is available to users with approved restricted data access.
This project uses information from the 2012, 2014, and 2016 HRS survey waves. To assess
level and change in cognitive performance following airborne lead exposure, I link census tract of
residence for HRS survey participants to census tract lead exposure estimates from the 2011 NATA
data.
4.3.1.3 Decennial Census and American Community Survey (ACS)
The United States has conducted a decennial census every ten years beginning in 1790. The US
Census Bureau agency was created to carry out the 1900 census and has overseen each subsequent
decennial census. The aim of the decennial census is to produce a complete count of the US pop-
ulation as well as information on demographic and economic characteristics of the US population.
The American Community Survey (ACS), also produced by the US Census Bureau, has been
carried out annually beginning in 2005. The survey consists of a long-form questionnaire of eco-
nomic and social questions administered to a subset of the US population. It replaced the long-form
component of the decennial census and was created with the goal of providing detailed population
information more frequently than every ten years. The US Census Bureau produces multiple data
products based on the ACS including summary estimates for different geographies and time ranges.
75
This project uses census tract level information from the 2010 decennial census and 2006-2010
5-year ACS summary estimates.
4.3.2 Measures
Table 4.1 provides an overview of the variables used in this analysis including the data source and
release year, variable type, range (for numeric variables), and levels (for indicator and categorical
variables).
Table 4.1: Variables Used in this Analysis
Level Variable Source Type Range or Levels
Individual
Cognitive Functioning 2012-2016 HRS discrete 0 to 27
Age 2012-2016 HRS (re-scaled) discrete 52+
Female Sex 2012-2016 HRS indicator female/male
Race and Ethnicity 2012-2016 HRS (derived) categorical 4 categories
Educational Attainment 2012-2016 HRS (derived) categorical 3 categories
Poverty Status 2012-2016 HRS (derived) indicator in/above poverty
Smoking History 2012-2016 HRS (derived) categorical 3 categories
Census Tract
Estimated Lead Exposure Risk Quartile 2011 NATA (derived) categorical 4 categories
Urbanicity 2012-2016 HRS (derived) categorical 6 categories
Census Division 2012-2016 HRS (derived) categorical 9 categories
% White Non-Hispanic/Latino 2010 Decennial Census continuous 0 to 100%
Percent of Adults without High School Degree 2006-2010 ACS 5-year continuous 0 to 100%
Percent of Housing Built in 1950 or Earlier 2006-2010 ACS 5-year continuous 0 to 100%
Median Household Income 2006-2010 ACS 5-year (re-scaled) continuous 0+
4.3.2.1 Cognitive Functioning
Overall cognitive functioning, based on individual performance on the Telephone Interview for
Cognitive Status (TICS) assessment, was used as the main dependent variable. Cognitive function
is assessed in HRS survey participants using the Telephone Interview for Cognitive Status (TICS)
assessment for participants [138, 139]. The TICS assessment is a modified version of the Mini-
Mental State Examination (MMSE) and assesses multiple cognitive domains including attention,
memory, language, and visuospatial ability [140]. Assessment performance ranged from 0 to 27
with lower values indicative of poorer cognitive performance. Previous work has classified the
TICS-27 point scale into ranges corresponding to normal cognition (12-27), cognitive impairment
with no dementia (CIND; 7-11), and dementia (0-6) [138].
76
4.3.2.2 Tract of Residence Lead Exposure Risk
Lead emissions annual exposure risk at the census tract of residence was used as the main explana-
tory variable. To differentiate census tracts by lead emissions exposure risk, tracts were categorized
into quartiles based on estimated lead exposure values from the 2011 NATA assessment. Census
tracts were included in quartile categorization even if they were not represented by individuals
in the HRS data subset used for analysis. Table 4.2 provides summary information on the four
quartile groups in mg/m
3
estimated lead exposure. The table also provides information on group
interquartile range (IQR) relative to the lead RfC of 0.00015 mg/m
3
and relative to the overall
standard deviation (using log-transformed lead exposure values). A figure showing the distribution
of log estimated lead exposure values for each exposure quartile is provided in Appendix H. Fig-
ure 4.1 depicts tract lead exposure risk quartile throughout for all census tracts in the contiguous
United States.
Table 4.2: Summary Statistics for Lead Exposure Quartile Groups
Quartile Mean (mg/m
3
) Median (mg/m
3
) IQR (mg/m
3
) IQR vs RfC logIQR vs logSD
Q1 (lowest exposure) 2.5× 10
− 4
2.7× 10
− 4
1.9− 3.3× 10
− 4
1.2x to 2.2x RfC -1.45 to -0.80 SD
Q2 5.3× 10
− 4
5.2× 10
− 4
4.6− 5.9× 10
− 4
3.1x to 4.0x RfC -0.42 to -0.12 SD
Q3 8.7× 10
− 4
8.6× 10
− 4
7.6− 9.8× 10
− 4
5.2x to 6.5x RfC 0.16 to 0.45 SD
Q4 (highest exposure) 23.1× 10
− 4
16.7× 10
− 4
13.4− 23.0× 10
− 4
8.9x to 15.3x RfC 0.81 to 1.43 SD
4.3.2.3 Covariates
Individual Covariates from the Health and Retirement Study (HRS)
Age Continuous age in years was centered at the mean age for the study population (64 years)
and re-coded to 10-year increments.
Female Sex The sex variable consists of two levels (female and male) and was not re-coded
for this analysis.
77
Figure 4.1: Lead Emissions from All Sources Exposure Risk Quartile
Race and Ethnicity A categorical race/ethnicity variable was created using self-reported
race (White, Black, and other) and Hispanic or Latino ethnicity. The final race/ethnicity variable
consists of four categories: White non-Hispanic/Latino, Black non-Hispanic/Latino, other non-
Hispanic/Latino, and Hispanic or Latino. A table showing the relationship between the original
race and ethnicity variables and the final race and ethnicity variable is provided in Appendix I.
Educational Attainment Information on individual highest degree obtained was re-coded to
a three-part educational attainment variable with levels for no degree; high school degree, GED, or
equivalent; and some college or more.
Below Poverty Threshold A binary poverty status variable, indicating if the individual is
above or below the poverty threshold was calculated using a ratio of household income to poverty
78
threshold level variable provided in the HRS RAND data set. Individuals were coded as below
poverty threshold if their household income to poverty ratio was less than 1 and above poverty
threshold if their household income to poverty ratio was equal to or greater than 1.
Smoking Status The smoking variable history consists of three categories: current smoker,
former smoker, and never smoker. A table showing the relationship between the original current
smoker and ever smoker variables and the smoking history variable is provided in Appendix L.
Census Tract Characteristics from the Health and Retirement Study (HRS)
Urbanicity The original nine-category urbanicity variable was re-coded to six categories:
urban metro with 20,000 or more residents and urban non-metro with 20,000 or more residents
were collapsed into an urban with 20,000 or more residents category; urban metro with fewer than
20,000 residents and urban non-metro with fewer than 20,000 residents were collapsed into an
urban with fewer than 20,000 residents category; rural metro and rural non-metro were collapsed
into a rural category. A table showing the relationship between the original and final urbanicity
variables is provided in Appendix J.
Census Division The census division variable was created based on the Census Regions and
Divisions of the United States using state of residence from the HRS data. A map of the Census
Regions and Divisions of the United States is provided in Appendix K.
Census Tract Characteristics from the Decennial Census and American Community Survey
(ACS)
Demographic Percent of tract population which are White Non-Hispanic/Latino was derived
from the 2010 Decennial Census. This variable was not re-coded for this analysis.
79
Socioeconomic Percent of tract adult (age 25 years and older) population with less than a
high school degree, percent of tract occupied housing built in 1950 or earlier, and tract median
household income (in 2010 adjusted dollars) were derived from the 2006-2010 ACS 5-year es-
timates. Median household income was rescaled to units $10,000. The other two ACS-derived
variables (percent of adult population without a high school degree and percent of housing built in
1950 or earlier) were not re-coded for this analysis.
4.3.3 Data Management
The 2012, 2014, and 2016 HRS waves include information on 21,090 unique individuals aged 52
or older in 2012 with non-missing age information. These 21,090 unique individuals are repre-
sented by 53,466 observations across the three survey years 2012-2016 and represent 9,117 unique
census tracts. Of these, 1,062 (5%) individuals missing the 27-point overall cognitive functioning
variable were excluded. Information on the remaining 20,028 individuals from the HRS survey
were then merged with individual data from the HRS RAND dataset containing information on
health behaviors, income, and poverty status used in this analysis; this step did not eliminate any
observations. Next the data was merged with the 2011 NATA lead exposure data using survey
respondents’ census tract of residence; this step removed an additional 244 (1.2%) individuals ei-
ther living outside the contiguous United States (4 individuals living in Alaska, Hawaii, or Puerto
Rico), missing census tract of residence information (236 individuals), or associated with an erro-
neous census tract code (1 individual). Next, 63 (¡1%) individuals with zero sample weight and
an additional 163 (¡1%) individuals with missing HRS covariate information were dropped
1
. The
HRS sample was then merged with census tract demographic information from the 2010 Decen-
nial Census and the 2006-2010 American Community Survey (ACS) 5-year estimates using census
tract of residence; this step excluded an additional 3 observations with missing census tract covari-
ate information but did not reduce the number of unique individuals. The final sample included
1
Of the 163 individuals dropped due to missing covariate information, 98 were missing smoking status only, 32
missing race/ethnicity only, 21 missing tract census division only, 7 missing nativity only, 3 missing nativity and
smoking status, 1 missing race/ethnicity and smoking, and 1 missing race/ethnicity and nativity.
80
19,558 individuals represented by 48,541 observations and residing in 8,595 unique census tracts
(11.8% of all 72,539 tracts in the contiguous United States based on the 2010 US Census; excludes
Alaska, Hawaii, and Puerto Rico); the overall sample selection process eliminated 1,532 (7.3%)
individuals, 4,925 (9.2%) observations, and 522 (5.7%) census tracts. The figure below provides
information on the analytic sample selection process.
Figure 4.2: Flow chart of the data restriction process using data from the Health and Retirement
Study and National Air Toxics Assessment (NATA).
81
4.3.4 Analysis Plan
Data analysis proceeded in several steps beginning with exploratory descriptive analysis and bi-
variate analysis. The main analysis used multilevel mixed-effects models to assess the effect of
lead exposure risk on cognition in older adults. All analyses and data management were conducted
using Stata release 17 [131].
4.3.4.1 Descriptive Statistics and Bivariate Analyses
Following formation of the analysis sub-sample and merging with other data sets, I first produced
descriptive statistics for each variable including summary statistics for continuous and discrete
variables and frequency information for indicator and categorical variables.
Next, I carried out correlation tests to determine the bivariate relationship between tract lead
exposure risk quartile and overall cognition as well as tract lead exposure risk quartile and all in-
dividual and tract level covariates. One-way ANOV A was used to assess the relationship between
tract lead exposure risk quartile and continuous or discrete covariates and χ
2
tests were used to
assess the relationship between tract lead exposure risk quartile and indicator or categorical co-
variates. In some cases, when the bivariate relationship between tract lead quartile an indicator or
categorical covariate was of particular interest, and a I further carried out pairwise comparisons
between lead quartiles using the Tukey’s Honestly Significant Difference (HSD) test.
4.3.5 Multivariable Analysis
For the main analysis, I used mixed-effects models to assess the effect of lead emissions exposure
risk on overall cognition in older adults. The primary exposure variable of interest was census
tract of residence lead exposure risk quartile, which was included as a four-part categorical vari-
able. Interpretation was focused on comparing living in a tract within the highest lead exposure
quartile (heretofore referred to as “Q4 tract”) on overall cognitive performance relative to living in
a tract within the lowest lead exposure quartile (heretofore referred to as a “Q1 tract”). I specified
82
a random intercept model, with individual age, sex, race/ethnicity, educational attainment, poverty
status, and smoking history; and census tract lead exposure risk quartile, urbanicity, census divi-
sion, percent population not non-Hispanic White, percent adult population without a high school
degree, median household income, and percent of housing built prior to 1950 included as fixed
effects. The correlation between repeated measurements from the same survey participant was
accounted for by running a series of multilevel linear regression models using themixed command
in Stata release 17 [131]. Full models used time, individual, and census tract of residence as model
three levels. Age was used as the indicator of time and centered at 64 years of age, corresponding
to the mean age in the sample. Age was also included as an interaction term for all covariates to
test for effect modification by age. I used maximum likelihood estimation to estimate model pa-
rameters and calculated adjusted mean differences in cognitive performance between individuals
living in different exposure risk quartiles, focusing on comparing the highest and lowest quartiles,
adjusting for covariates.
4.3.5.1 Ancillary and Supplemental Analyses
Following the main multivariate analysis, I carried out two ancillary analyses stratifying the sample
population by 1) individual race/ethnicity and 2) tract of residence urbanicity category.
In addition, I carried out a supplemental analysis testing different versions of the main explana-
tory variable, tract of residence lead exposure. These versions included 1) tract of residence log
continuous lead exposure estimate as well as two versions of the main effect variable which take
into account the lead exposure level of neighboring census tracts: 2) mean of tract of residence
and all neighboring tracts log continuous lead exposure estimate, and 3) categorical tract of res-
idence versus neighboring tracts lead exposure quartile. The categorial tract of residence versus
neighboring tracts lead exposure quartile variable was defined based on the exposure quartile of the
residence tract and the mean exposure quartile of all neighboring tracts using the schema provided
in Table x below.
83
4.4 Results
4.4.1 Study Population Characteristics
The study population consisted of 48,541 observations on 19,558 individuals. The mean age at first
survey assessment was 63.3 and the population was 56% female and 58% non-Hispanic White. In-
dividuals were distributed roughly evenly between the four lead exposure risk quartiles; percent of
the total number of observations in each quartile ranged from 21.1% (Q1) to 27.2% (Q3). Survey-
weighted information on individual and tract of residence characteristics is provided in Table 4.3.
4.4.2 Descriptive Statistics
4.4.2.1 Cognitive Functioning
Overall cognitive functioning test scores for the study population ranged from 0 to 27 with a
survey-weighted mean, median, and standard deviation of 15.7, 16, and 4.4, respectively. The
distribution of overall cognitive functioning scores was approximately normal.
As expected, overall cognition was lower for older study participants and mean cognition sig-
nificantly declined with increasing age. Figure ?? shows the distribution of cognition by age range
for study participants based on the first year they were surveyed between 2012 and 2016 with
survey weights applied.
4.4.2.2 Tract of Residence Lead Exposure
Figure 4.4 provides information on tract-level 2011 estimated log lead exposure in mg/m3 for
the 19,558 individuals in the study sample using the earliest recorded tract of residence for each
participant between 2012 and 2016. The distribution of participant tract of residence lead expo-
sure is highly right-skewed but approximately log-normal. Tract-level lead emissions range from
0.0000231 to 0.0402002 mg/m
3
with a mean and median value of 0.0010242 and 0.0007078, re-
spectively. Based on the 2011 NATA EPA’s reference concentration (RfC) of 0.00015 mg/m
3
for
84
Table 4.3: Demographic Characteristics of Study Population and Tracts of Residence
Observations
Tract Lead Exposure Risk Quartile
Q1 (Lowest Lead) Q2 Q3 Q4 (Highest Lead)
N 48,541 11,449 13,005 13,211 10,876
Overall Cognition (27-Point Scale) 15.7 (4.4) 15.4 (4.4) 15.8 (4.4) 16.0 (4.3) 15.3 (4.5)
Age 66.7 (9.6) 67.4 (9.4) 66.5 (9.7) 66.3 (9.6) 66.6 (9.8)
Gender
Male 22,024 (45.4%) 5,051 (44.1%) 5,955 (45.8%) 6,059 (45.9%) 4,958 (45.6%)
Female 26,517 (54.6%) 6,398 (55.9%) 7,050 (54.2%) 7,152 (54.1%) 5,918 (54.4%)
Race/Ethnicity
non-Hispanic White 37,743 (77.8%) 9,121 (79.7%) 10,667 (82.0%) 10,331 (78.2%) 7,623 (70.1%)
non-Hispanic Black 4,878 (10.0%) 605 (5.3%) 1,320 (10.1%) 1,529 (11.6%) 1,424 (13.1%)
non-Hispanic Other Race 1,739 (3.6%) 405 (3.5%) 386 (3.0%) 497 (3.8%) 452 (4.1%)
Hispanic 4,181 (8.6%) 1,318 (11.5%) 631 (4.9%) 854 (6.5%) 1,378 (12.7%)
Smoking Status
Never Smoker 21,653 (44.6%) 5,125 (44.8%) 5,938 (45.6%) 5,856 (44.3%) 4,735 (43.5%)
Former Smoker 20,562 (42.4%) 4,896 (42.7%) 5,483 (42.2%) 5,620 (42.6%) 4,564 (42.0%)
Current Smoker 6,326 (13.0%) 1,428 (12.5%) 1,585 (12.2%) 1,735 (13.1%) 1,578 (14.5%)
Educational Attainment
No High School Degree 6,097 (12.6%) 1,736 (15.2%) 1,294 (10.0%) 1,312 (19.9%) 1,755 (16.1%)
High School Degree, GED, or Equivalent 24,857 (51.2%) 5,976 (52.2%) 6,652 (51.1%) 6,477 (49.0%) 5,752 (52.9%)
Some College or More 17,587 (36.2%) 3,737 (32.6%) 5,059 (38.9%) 5,422 (41.1%) 3,369 (31.0%)
Poverty Status
Below Poverty Threshold 43,524 (89.7%) 10,128 (88.5%) 11,912 (91.6%) 11,912 (90.8%) 9,492 (87.3%)
At or Above Poverty Threshold 5,017 (10.3%) 1,321 (11.5%) 1,093 (8.4%) 1,220 (9.2%) 1,384 (12.7%)
Urbanicity
Metropolitan (1 million +) 24,563 (50.6%) 2,505 (21.9%) 6,105 (46.9%) 8,271 (62.6%) 7,681 (70.6%)
Metropolitan (250,000 to < 1 million) 11,331 (23.3%) 3,541 (30.9%) 3,778 (29.0%) 2,778 (21.0%) 1,234 (11.4%)
Metropolitan (< 250,000) 3,788 (7.8%) 1,047 (9.2%) 1,113 (8.6%) 996 (7.5%) 632 (5.8%)
Urban (20,000 +) 3,907 (8.1%) 1,627 (14.2%) 857 (9.6%) 734 (5.6%) 689 (6.3%)
Urban (< 20,000) 4,619 (9.5%) 2,451 (21.4%) 1,104 (8.5%) 424 (3.2%) 639 (5.9%)
Rural 333 (0.7%) 278 (2.4%) 47 (0.4%) 8 (0.1%) 0 (0%)
Census Division
New England 2,329 (4.8%) 469 (4.1%) 1,111 (8.5%) 626 (4.7%) 123 (1.1%)
Mid-Atlantic 6,088 (12.6%) 874 (7.6%) 918 (7.1%) 2,426 (18.4%) 1,870 (17.2%)
Northeast Central 8,060 (16.6%) 692 (6.1%) 2,640 (20.3%) 2,809 (21.3%) 1,919 (17.7%)
Northwest Central 3,893 (8.0%) 1,214 (10.6%) 1,382 (10.6%) 608 (4.6%) 688 (6.3%)
South Atlantic 10,223 (21.1%) 3,084 (26.9%) 3,053 (23.5%) 2,886 (21.9%) 1,200 (11.0%)
Southeast Central 3,116 (6.4%) 563 (4.9%) 1,116 (8.6%) 845 (6.4%) 592 (5.4%)
Southwest Central 4,679 (9.6%) 2,042 (17.8%) 1,241 (9.5%) 6256 (4.7%) 770 (7.1%)
Mountain 3,514 (7.2%) 1,302 (11.4%) 384 (3.0%) 862 (6.5%) 965 (8.9%)
Pacific 6,639 (13.7%) 1,209 (10.6%) 1,159 (8.9%) 1,522 (11.5%) 2,749 (25.3%)
% White Non-Hispanic 67.7% (28.9%) 72.5% (27.7%) 73.0% (25.3%) 66.4% (28.2%) 58.0% (32.1%)
% With No High School Degree 14.3% (11.2%) 15.5% (11.8%) 12.9% (9.8%) 12.5% (9.8%) 17.0% (13.0%)
Median Household Income (2010 Dollars) $59,767 ($27,538) $52,888 ($23,256) $63,336 ($27,515) $66,426 ($31,354) $54,652 (23,894)
% Housing Built < 1950 29.7% (26.0%) 23.5% (20.4%) 24.9% (23.1%) 33.5% (28.5%) 37.5% (28.4%)
85
Figure 4.3: Overall Cognition Range for Age Groups
lead, 96.4% of participants in the study sample lived in a census tract associated with estimated
lead exposure greater than the RfC in 2011. Note that numeric lead exposure values in mg/m
3
were
not used in the main analysis and are provided here for reference only.
4.4.3 Bivariate Analyses with Tract of Residence Lead Exposure Quartile
Overall cognition was statistically significantly associated with residential census tract lead expo-
sure quartile in bivariate analysis (p = 0.016). Mean overall cognition was 15.4, 15.8, 16.0, and
15.3 for lead exposure Q1 (lowest lead), Q2, Q3, and Q4 (highest lead), respectively. Further
pairwise comparisons of means between lead exposure quartiles indicated significant differences
between all pairs of lead exposure quartiles (p ¡ 0.02) with the exception of quartile 2 vs 3 (p
86
Figure 4.4: Distribution of Lead Exposure for Study Population
= 0.182). Figure 4.5 provides information on the survey-weighted mean and standard deviation
overall cognition score by tract of residence lead exposure quartile.
All other individual covariates, with the exception of gender, were also statistically significantly
associated with tract lead exposure quartile in bivariate analysis. Residence in a Q4 (highest lead)
tract was positively associated with being non-Hispanic Black, Hispanic, or non-Hispanic other
race (vs non-Hispanic White); foreign born (vs US born); a current smoker (vs former smoker or
never smoker); not having a high school degree or equivalent (vs having a high school degree or
more), and being in poverty (vs not in poverty).
Older adults living in metropolitan areas with 1 million population or more were also signif-
icantly more likely to live in a Q4 tract (33% in Q4 tracts) compared to older adults in smaller
87
Figure 4.5: Survey-weighted mean overall cognition score by tract of residence lead exposure
quartile. Standard deviations for each quartile are provided as error bars.
metropolitan and urban areas (13-17% in Q4 tracts). None of the members of the sample popula-
tion living in a rural area were in a Q4 tract. None of the members of the sample population living
in a rural area resided in a Q4 tract.
Based on bivariate analysis with tract-level sociodemographic covariates, living in a Q4 tract
was also significantly associated with lower proposition of non-Hispanic White, higher proportion
of adults without a high school degree, lower median income, and higher proportion of occupied
housing built in 1950 or earlier.
Table 4.4 provides p-values for ANOV A orx
2
tests for bivariate associations between tract lead
exposure quartile and all individual and tract-level variables used in this analysis.
88
Table 4.4: Bivariate correlations between tract of residence lead exposure and all covariates
Individual Covariates ANOV A orx
2
p-value
Overall Cognition (27-Point Scale) 0.016
Age < 0.00005
Gender 0.357
Race/Ethnicity < 0.0005
Nativity < 0.0005
Smoking History < 0.0005
Educational Attainment < 0.0005
Poverty Status < 0.0005
Tract-Level Covariates ANOV A orx
2
p-value
Urbanicity < 0.0005
Census Division < 0.0005
Percent of Population White Non-Hispanic < 0.00005
Percent of Population 25+ Without High School Degree < 0.00005
Median Household Income (2010 Inflation-Adjusted Dollars) < 0.00005
Percent of Occupied Housing Units Built ¡ 1950 < 0.00005
4.4.4 Multivariable Analyses
Main model results are provided in Table 4.5 (models M0-M5). Appendix N includes additional
models used in the model development process (models A1-A5, B1-B3, C1-C6, and D1-D5) as
well as the main models (models M0-M5) and provides additional details of the effects of individ-
ual covariates and covariate groups on the relationship between residential tract lead quartile and
cognition outcomes in older adults.
Living in a Q4 (highest lead exposure) tract was significantly associated with -0.56 lower cog-
nitive score on the 27-point overall cognitive scale when compared with living in a Q1 (lowest
lead exposure) tract; the model did not show any significant impact of tract residence on cognitive
change over time (model M2). However, after controlling for all individual demographic vari-
ables and tract-level characteristics used in analysis, the effect of residential tract lead quartile was
no longer significant (model M5). Specifically, the significance of the effect of residential tract
lead quartile on cognition was removed after the inclusion of the individual race/ethnicity category
(model A2 in Appendix N) and/or the tract variable indicating percent of population White Non-
Hispanic (model C1 in Appendix N). The individual education variable (model A4), tract variable
for percent of adult population with less than a high school education (model C2), and tract vari-
able for percent of occupied housing built prior to 1950 (model C4) reduced the effect of residential
tract lead quartile on cognition by at least 10% but did not remove the effect significance below
89
an alpha of 10%. The tract census division control variable increased the effect of the tract lead
quartile on cognition by ¿ 10% but also did not remove the effect significance (model C6).
4.4.5 Ancillary Analyses
Following the main analysis, and based on the results from the main analysis, I conducted two
stratified analyses. The first ancillary analysis focuses on the individual race/ethnicity categorical
variable, the individual-level variable with the largest impact on the relationship between tract of
residence lead exposure quartile and individual cognition. The second ancillary analysis focuses
on the tract urbanicity control variable which is highly correlated with the likelihood of living in a
Q4 tract. The same set of individual and tract covariates used in the main analysis were also used
in the two ancillary analyses.
4.4.5.1 Race/Ethnicity
For the race/ethnicity stratified analysis, the Other non-Hispanic/Latino race/ethnicity category was
not analyzed due to small sample size and lack of specificity in this category’s definition. Table
4.6 provides main model results (models M2 and M5 for each race/ethnicity category).
Stratifying the main analysis by race/ethnicity categories rendered a similar result to the main
analysis. For all final race/ethnicity stratified models, after controlling for demographic and tract
characteristics, Q4 residential tract was not significantly associated with a change in cognition.
4.4.5.2 Urbanicity
Tables 4.7 and 4.8 provide model results (models M0b-M5b) for the urbanicity stratified analysis.
The models subsetting to only individuals living in a metropolitan area with 1 million residents
or more broadly mirror the overall analysis: the unadjusted model showed a significant -0.54 point
relationship between residing in a Q4 tract and overall cognition and no significant relationship
after adjusting for covariates.
90
Table 4.5: Results of mixed effects models examining the relationship between lead exposure
measures and cognitive test scores, adjusting for relevant covariates
Base Models Main Analysis
Base Age Only
Age and
Lead Risk
Quartile Only
Individual
Covariates
Only
Tract
Covariates
Only
Full Model
M0 M1 M2 M3 M4 M5
N 48541 48541 48541 48541 48541 48541
Total Cognitive Function (27-point scale) 14.44 *** 14.95 *** 15.13 *** 17.05 *** 13.82 *** 16.93 ***
Age (10 year increments; centered at 64) -1.12 *** -1.20 *** -0.41 -1.21 * -0.63
Lead Exposure Q1 (lowest lead) - - - -
Lead Exposure Q2 -0.07 -0.02 -0.09 -0.05
Lead Exposure Q3 -0.08 0.23 + 0.19 0.19
Lead Exposure Q4 (highest lead) -0.56 *** 0.04 -0.03 0.02
Lead Exposure Q1 x Age - - - -
Lead Exposure Q2 x Age 0.26 0.19 0.03 0.01
Lead Exposure Q3 x Age -0.01 -0.11 -0.27 -0.28
Lead Exposure Q4 x Age 0.06 0 -0.05 -0.05
Male - -
Female 0.76 *** 0.75 ***
White Non-Hispanic/Latino - -
Black Non-Hispanic/Latino -2.46 *** -2.17 ***
Other Race Non-Hispanic/Latino -1.53 *** -1.45 ***
Hispanic -1.52 *** -1.31 ***
Never Smoker - -
Former Smoker -0.14 -0.12
Current Smoker -0.16 -0.14
No High School Degree -4.24 *** -4.03 ***
High School Degree or Equivalent -1.63 *** -1.52 ***
Some College or More - -
At or Above Poverty Threshold - -
Below Poverty Threshold -0.07 -0.06
Metro; 1mill + - -
Metro; 250k to ¡ 1mill -0.14 -0.08
Metro; ¡ 250k -0.21 -0.19
Urban; 20k + -0.28 -0.21
Urban; ¡ 20k -0.17 -0.13
Rural -0.51 -0.39
New England -0.14 -0.11
Mid-Atlantic -0.13 -0.04
East-North Central -0.1 -0.07
West-North Central 0.04 -0.02
South Atlantic - -
East-South Central -0.74 *** -0.61 **
West-South Central 0.1 0.1
Mountain 0.33 0.1
Pacific 0.61 *** 0.23
Percent of Population White Non-Hispanic/Latino
(10% increments)
0.25 *** 0.03
Percent of Adults 25+ with No High School Degree
(10% increments)
-0.39 *** -0.16 ***
Median Household Income
(2010-adjusted dollars; $10k increments)
0.08 *** 0.03
Percent of Occupied Housing Built in 1950 or Earlier
(10% increments)
-0.03 -0.02
+p <0.10 *p <0.05 **p <0.01 ***p <0.001
91
Table 4.6: Results of mixed effects models examining the relationship between lead exposure
measures and cognitive test scores stratified by race/ethnicity categories
non-Hispanic White non-Hispanic Black Hispanic
Unadjusted Full Model Unadjusted Full Model Unadjusted Full Model
N 30984 30984 9435 9435 6529 6529
Total Cognitive Function (27-point scale) 15.89 *** 16.88 *** 12.52 *** 15.31 *** 13.32 *** 15.78 ***
Age (10 year increments; centered at 64) -1.18 *** -0.58 -2.05 ** -1.72 -0.7 2.59
Lead Exposure Q1 (lowest lead) - - - - - -
Lead Exposure Q2 -0.03 0.08 0.52 * -0.08 0.28 -0.32
Lead Exposure Q3 0.29 0.39 + 0.79 ** 0.08 0.07 -0.2
Lead Exposure Q4 (highest lead) 0.14 0.18 0.50 * 0.09 -0.47 + -0.13
Lead Exposure Q1 x Age - - - - - -
Lead Exposure Q2 x Age 0.2 -0.06 1.26 0.88 -0.55 -0.55
Lead Exposure Q3 x Age -0.15 -0.44 1.03 0.47 0.04 -0.24
Lead Exposure Q4 x Age -0.17 -0.23 1.29 0.75 -0.33 -0.5
Male - - -
Female 1.09 *** 0.46 *** -0.18
Never Smoker - - -
Former Smoker -0.17 -0.33 * 0.11
Current Smoker -0.09 -0.44 -0.07
No High School Degree -3.64 *** -4.62 *** -3.79 ***
High School Degree or Equivalent -1.44 *** -2.11 *** -1.39 ***
Some College or More - - -
At or Above Poverty Threshold - - -
Below Poverty Threshold 0.02 -0.05 -0.03
Metro; 1mill + - - -
Metro; 250k to ¡ 1mill 0.05 -0.38 * -0.42
Metro; ¡ 250k -0.1 -0.16 -2.35 *
Urban; 20k + 0.06 -0.85 ** -0.02
Urban; ¡ 20k 0.16 -0.17 -1.66 **
Rural -0.91 2.27 * -0.21
New England 0.19 0.58 -0.61
Mid-Atlantic 0.17 0.08 -0.92 *
East-North Central -0.03 0.01 0.07
West-North Central 0.15 -0.45 0.42
South Atlantic - - -
East-South Central -0.68 * -0.17 -1.67
West-South Central 0.07 -0.28 0.27
Mountain 0.1 0.66 0.61 +
Pacific 0.58 * 0.11 -0.15
Percent of Population White Non-Hispanic/Latino
(10% increments)
0 0.05 0.06
Percent of Adults 25+ with No High School Degree
(10% increments)
-0.28 * -0.16 * -0.09
Median Household Income
(2010-adjusted dollars; $10k increments)
0 0.06 0.08 +
Percent of Occupied Housing Built in 1950 or Earlier
(10% increments)
-0.02 0.03 0
+p <0.10 *p <0.05 **p <0.01 ***p <0.001
92
The models focused on metropolitan areas with 250,000 to less than 1 million population
showed a significant -1.3 point decrease in cognition associated with living in a Q4 tract, more
than double the effect seen in the overall analysis. Interestingly, unlike the main analysis, the
model retained a significant effect of -0.82 decline in cognition associated with living in a Q4
tract even after adjusting for individual covariates including race/ethnicity which removed the sig-
nificance of the effect in the main analysis. After adjusting for all individual and all tract-level
covariates in the same model, the effect is non-significant.
The analyses for the other three analyzed urbanicity classes modeled in this ancillary analysis
(metropolitan areas with less than 250,000 population, urban areas with 20,000 population or more,
and urban areas with less than 20,000 population) did not show a significant relationship between
lead exposure quartile and overall cognitive functioning in both unadjusted and adjusted models.
In addition, the urban areas with 20,000 population or more urbanicity class could not be fully
analyzed due to a lack of variation in the population subset which prevented some of the final
models from converging.
For the urbanicity stratified analysis, the rural urbanicity class was not analyzed because the Q4
exposure quartile was not represented in the rural sub-sample. The urban areas with 20,000 popu-
lation or more urbanicity class was not fully analyzed as lack of variation in the study population
in this sub-sample prevented the final models from converging.
4.5 Conclusions and Discussion
4.5.1 Summary of Findings
The base model in the main analysis indicated that older adults living in Q4 tracts have roughly half
point lower overall cognition than older adults living in Q1 tracts based on the 27-point TICS scale,
but do not experience a significantly different rate of cognitive decline over time. This is consistent
with the first part of the original hypothesis that greater environmental exposure to lead emissions
would be associated with lower cognitive functioning in older adults due to the fact that other
93
Table 4.7: Results of mixed effects models examining the relationship between lead exposure
measures and cognitive test scores stratified by urbanicity categories
Metropolitain 1 Million + Metropolitain 250k to 1 Million
Unadjusted Full Model Unadjusted
Individual
Covariates Only
Full Model
N 25662 25662 11028 11028 11028
Total Cognitive Function (27-point scale) 15.16 *** 16.58 *** 15.24 *** 17.28 *** 17.43 ***
Age (10 year increments; centered at 64) -0.99 * -0.49 -1.54 *** -0.7 -1.51
Lead Exposure Q1 (lowest lead) - - - - -
Lead Exposure Q2 -0.21 -0.05 -0.08 -0.3 -0.22
Lead Exposure Q3 -0.1 0.27 -0.54 -0.41 -0.28
Lead Exposure Q4 (highest lead) -0.54 * 0.13 -1.30 ** -0.82 * -0.64
Lead Exposure Q1 x Age - - - - -
Lead Exposure Q2 x Age 0.4 0.22 0.34 0.4 0.22
Lead Exposure Q3 x Age -0.19 -0.32 0.52 0.59 0.33
Lead Exposure Q4 x Age -0.22 -0.22 0.72 0.86 0.71
Male - - -
Female 0.61 *** 0.73 *** 0.74 ***
White Non-Hispanic/Latino - - -
Black Non-Hispanic/Latino -2.02 *** -2.52 *** -2.11 ***
Other Race Non-Hispanic/Latino -1.43 *** -1.59 *** -1.51 **
Hispanic -1.27 *** -1.86 *** -1.49 ***
Never Smoker - - -
Former Smoker -0.13 -0.01 0
Current Smoker -0.13 -0.23 -0.21
No High School Degree -3.85 *** -4.05 *** -3.72 ***
High School Degree or Equivalent -1.43 *** -1.48 *** -1.33 ***
Some College or More - - -
At or Above Poverty Threshold - - -
Below Poverty Threshold -0.02 -0.15 -0.15
New England -0.66 0.77
Mid-Atlantic -0.24 -0.3
East-North Central -0.24 0.11
West-North Central -0.3 0.37
South Atlantic - - -
East-South Central -0.13 -0.85 *
West-South Central -0.01 0.09
Mountain 0.22 0.39
Pacific 0.16 0.36
Percent of Population White Non-Hispanic/Latino
(10% increments)
0.03 0.04
Percent of Adults 25+ with No High School Degree
(10% increments)
-0.15 ** -0.22 +
Median Household Income
(2010-adjusted dollars; $10k increments)
0.06 * -0.03
Percent of Occupied Housing Built in 1950 or Earlier
(10% increments)
0.01 -0.10 +
+p <0.10 *p <0.05 **p <0.01 ***p <0.001
94
Table 4.8: Results of mixed effects models examining the relationship between lead exposure
measures and cognitive test scores stratified by urbanicity categories
Metropolitain < 250k Urban 20k + Urban < 20k
Unadjusted Full Model Unadjusted Unadjusted Full Model
N 3327 3327 3905 4263 4263
Total Cognitive Function (27-point scale) 15.33 *** 16.27 *** 14.31 *** 15.11 *** 15.63 ***
Age (10 year increments; centered at 64) -0.38 -2.44 -1.37 ** -1.32 *** 3.67
Lead Exposure Q1 (lowest lead) - - - - -
Lead Exposure Q2 -0.23 0.21 0.01 0.26 -0.05
Lead Exposure Q3 0.28 1.18 + 0.98 -0.11 -0.54
Lead Exposure Q4 (highest lead) -0.12 0.96 0.91 -0.55 -0.08
Lead Exposure Q1 x Age - - - - -
Lead Exposure Q2 x Age -0.13 -0.28 0.02 -0.77 -0.65
Lead Exposure Q3 x Age -0.98 -1.28 -0.12 -0.74 -0.42
Lead Exposure Q4 x Age -0.3 -0.55 -0.17 0.53 0.78
Male - -
Female 0.46 1.22 **
White Non-Hispanic/Latino - -
Black Non-Hispanic/Latino -1.34 * -2.60 ***
Other Race Non-Hispanic/Latino 0.2 -1.17 +
Hispanic -1.02 -1.64 ***
Never Smoker - -
Former Smoker 0.56 -0.38
Current Smoker -0.12 0.16
No High School Degree -4.78 *** -3.74 ***
High School Degree or Equivalent -2.27 *** -1.74 ***
Some College or More - -
At or Above Poverty Threshold - -
Below Poverty Threshold -0.29 0.16
New England -7.66 -0.91
Mid-Atlantic 2.17 -0.58
East-North Central -0.1 -0.48
West-North Central 0.53 -0.71
South Atlantic - -
East-South Central 0.23 -2.49 *
West-South Central 1.37 -0.16
Mountain 0.6 -0.34
Pacific 0.68 -4.95
Percent of Population White Non-Hispanic/Latino
(10% increments)
0.21 0.24
Percent of Adults 25+ with No High School Degree
(10% increments)
-0.69 0.07
Median Household Income
(2010-adjusted dollars; $10k increments)
-0.13 -0.08
Percent of Occupied Housing Built in 1950 or Earlier
(10% increments)
-0.08 0.01
+p <0.10 *p <0.05 **p <0.01 ***p <0.001
95
sources of lead environmental lead exposure, such as occupational exposure, has been associated
with worse cognitive outcomes in older adults due to lead’s adverse neurocognitive effects.
In the main analysis, individual race/ethnicity had the largest impact on removing the statisti-
cal significance of the effect of tract of residence on overall cognition indicating that the effect of
tract lead exposure quartile on cognition was confounded by the effect of race/ethnicity on cogni-
tion. This is in agreement with previous literature which has found strong associations between
race/ethnicity and cognition, which are often attributed to differences in socioeconomic status or
educational attainment for different racial and ethnic groups in the United States, as well the com-
plex interrelationships between individual race/ethnicity, neighborhood racial-ethnic composition,
and neighborhood environmental health and exposure characteristics. This outcome indicates that
individual race and ethnicity, or other characteristics associated with race and ethnicity, have a
greater impact on cognitive outcomes in older adults than lead exposure from airborne sources as
measured by the EPA NATA program.
In the base model, including only lead exposure risk quartile and age, overall cognition was
statistically significantly lower, by roughly half a point for individuals living in the highest ex-
posure risk tracts compared to those living in the lowest exposure tracts. However, there was no
association after accounting for individual- and tract-level covariates. Among the tested covari-
ates, individual race/ethnicity and tract-level [percent of population not white or percent of adult
population with less than a high school education] were the variables which removed the signifi-
cance of the effect of tract risk on cognition when included in the model. This finding indicates
that this relationship may be confounded by individual race and ethnicity as well as neighborhood
characteristics related to educational attainment.
Interestingly, while individual educational attainment had a strong and significant impact on
overall cognition, this variable did not impact the relationship between tract exposure risk and
cognition despite the fact that the tract-level educational attainment variable did impact this rela-
tionship.
96
4.5.1.1 Race/Ethnicity Ancillary Analysis
After subsetting the analysis by race, the effect of tract risk exposure level was even further min-
imized for the White non-Hispanic race/ethnicity group and the effect was negligible and non-
significant for this group. Similarly, the model restricting to only Black non-Hispanic individuals
also showed a small and non-significant effect after controlling for covariates. However, interest-
ingly, the model for Hispanic individuals showed a -0.87 point decrease in cognition when compar-
ing the highest risk tracts with the lowest risk tracts, before controlling for covariates. Similar to
other models, after controlling for covariates, this effect was reduced and rendered non-significant.
The model restricted to only White non-Hispanic individuals was the only model that produced
a non-significant effect of tract risk quartile on cognition in the absence of covariates. Both the
models for Black non-Hispanic and Hispanic populations showed significant effects in the model
without covariates. This finding may point to a differential influence of the neighborhood effect of
lead exposure risk on White non-Hispanic individuals compared to persons of color. This finding
also indicates a potential limitation of data which has majority White non-Hispanic individuals. As
mentioned in the background section, the previous work on lead exposure and cognitive decline has
under-represented persons of color. The studies which did not exclusively or predominantly study
White non-Hispanic individuals found differential results by race/ethnicity finding that persons of
color had, on average, higher lead biomarkers when compared to White persons after controlling
for covariates.
4.5.2 Comparison with Previous Literature
Overall, the analysis indicated a relationship between neighborhood lead exposure risk and overall
cognition in older adults but further indicated that this relationship may be highly confounded or
modified by both individual and neighborhood factors, namely individual race and ethnicity and
neighborhood educational attainment and income demographics. Further subanalysis indicated
differential effect sizes and significance for different race and ethnicity groups as well as different
urbanicity categories.
97
Interestingly, this study found an effect on cognition of living in a high exposure (Q4) tract
compared to a Q1 tract to be approximately equal to 5 years of aging. Similarly, previous work
hypothesized that the difference in the effect on cognition between
4.5.3 Implications and Significance
The model restricted to only White non-Hispanic individuals was the only model that produced
a non-significant effect of tract risk quartile on cognition in the absence of covariates. Both the
models for Black non-Hispanic and Hispanic populations showed significant effects in the model
without covariates. This finding may point to a differential influence of the neighborhood effect of
lead exposure risk on White non-Hispanic individuals compared to persons of color. This finding
also indicates a potential limitation of data which has majority White non-Hispanic individuals. As
mentioned in the background section, the previous work on lead exposure and cognitive decline has
under-represented persons of color. The studies which did not exclusively or predominantly study
White non-Hispanic individuals found differential results by race/ethnicity finding that persons of
color had, on average, higher lead biomarkers when compared to White persons after controlling
for covariates. Both this study and previous work points to differential experiences of lead ex-
posure and of the relationship between lead exposure and health outcomes for different race and
ethnic groups. The race/ethnicity subanalysis for this study indicates that, for White non-Hispanic
individuals, the impact of living in a high risk neighborhood may be negligible but not for other
race groups. This indicates that other factors at the neighborhood level may be at play which in-
fluence exposure, cognition, or relationship between exposure and cognition in different ways for
different race groups.
both to reduce current and ongoing sources with the purpose of mitigating continued cognition-
related outcomes in older adults and to reduce potential adverse lead-related cognitive outcomes
for future generations of older adults.
Understanding how specific sources of lead, such as airborne lead, may contribute to cognitive
outcomes in older adults may support policy action by allowing policy to specifically focus on the
98
source in question. In addition, while much of the discussion surrounding lead exposure focused on
exposure to legacy sources of lead-based paint, occupational exposure, or past exposure such as to
leaded gasoline emissions, it is less common for work to focus on neighborhood-level exposures.
Unlike lead-based paint and occupational exposure, airborne emissions are neighborhood-level
exposure and are not based on individual characteristics or behaviors and, unlike past sources of
neighborhood-level exposure which have science been regulated (e.g. leaded gasoline emissions),
ongoing sources such as airborne emissions have a different exposure timeline.
These cognitive domains are essential for everyday activities such as driving, managing fi-
nances, and engaging in social activities and cognitive decline in these domains can have a signifi-
cant impact on older adult’s ability to carry out activities of daily living and live independently as
well as their quality of life more generally.
The increased need for care and support later in life due to deteriorating cognitive ability and
capacity for independence exacerbates the care burden placed on family members, caregivers, the
healthcare system, and the adult care industry. Populations characterized by financial or other
social vulnerabilities, who are also at greater risk of lead exposure to begin with, are more likely
than more advantaged populations to experience the increased care needs as a significant burden,
therefore deepening existing social inequities.
The pathways and moderators connecting lead exposure and health outcomes for neighborhood-
level exposure may differ from those for individual-level exposures and, unlike individual-level
exposures, individuals often have minimal agency to mitigate neighborhood-level exposures and
may only be able to avoid exposure risk by moving away from a high-risk neighborhood.
4.5.4 Limitations
While the HRS sample used in this analysis is one of the largest samples used in studies explor-
ing the relationship between lead exposure and cognition and is also nationally representative,
the race/ethnicity sub-samples have less statistical power, especially for non White non-Hispanic
individuals. More work should be carried out focusing on persons of color with larger national
99
samples to further explore the confounding effect of race and ethnicity on the relationship between
neighborhood lead exposure risk and cognitive changes in older adults.
As with the race/ethnicity subgroup analysis, the HRS sample used for this study had lim-
ited statistical power for certain urbanicity groups and too few individuals were represented in
the metropolitan with less than 250,000 population; urban with 20,000 population or more; urban
with less than 20,000 population; and rural urbanicity categories. The fact that the two groups
which had a sufficient sample to be studied produced different results indicates that different ur-
banicity regions may moderate the relationship between lead exposure risk and cognitive outcomes
for older adults. More research should be carried out with additional study subjects with greater
representation for less-populated urbanicity categories.
In addition, because the two sub-analyses, on race/ethnicity and urbanicity, both produced
interesting results, additional work should be carried out on a study population which allows for
further dissection by both race/ethnicity and urbanicity. The sample used for this analysis did
not have a sufficiently large study population to be able to carry out this subsequent analysis. In
addition, race/ethnicity and urbanicity were somewhat correlated and some intersections of these
two categorical variables included very few individuals.
An additional major limitation of this work is the inability to include variables correspond-
ing to individual or community residential housing age, an indicator of potential lead-based paint
presence. Historical lead-based paint use, and the residual concentrations of lead-based paint still
present in many homes in the United States, remains one of the most significant sources of lead
exposure. Unfortunately, the HRS dataset does not include information on age or building year
of participant residences and therefore this control variable could not be included in the analysis.
Future work should be carried out accounting for potential residential lead-based paint or other
potential major sources of lead exposure.
100
Chapter 5
Conclusion
The introduction of this dissertation indicated that the primary hypothesis of this work is that
ongoing contemporary sources of lead exposure have the potential to further exacerbate the related
health effects experienced by older adults, most significantly the effects on neurocognitive health,
and in turn exacerbate the care burden of the older adult population.
Chapter 2 underscores the fact that residential lead-based paint remains an important potential
source of lead exposure for the United States population including millions of older adults. Lead-
based paint has previously been the most commonly studied source of lead exposure and has been
strongly associated with adverse health outcomes in previous work.
All three chapters found either an association between lead exposure risk and race/ethnicity
(Chapters 2 and 3) or determined race/ethnicity to have a mediating or confounding effect on the
relationship between lead exposure and health outcomes (Chapter 4). Previous work has estab-
lished that, in the United States, environmental exposure risk is strongly associated with racioeth-
nic and socioeconomic population characteristics, with communities of color and communities
characterized by poverty or other socioeconomic vulnerabilities experiencing both a dispropor-
tionately high burden of environmental exposures and a disproportionately low rate and extent of
public health intervention efforts. In the United States, due to historical racist housing exclusion
and other social practices as well as the disproportionate inability of impoverished populations to
leave environmentally or otherwise hazardous areas, population-level environmental health issues
are often impossible to disentangle from racial and socioeconomic population characteristics and
101
other characteristics indicative of historical or ongoing social pressures and restrictions. The prob-
lem of disentangling these characteristics has been extensively discussed in the literature and the
work carried out in this dissertation was not done with the intent to solve this question. Instead,
however, the results described here contribute to what is already known by identifying that legacy
residential lead-based paint presence as well as contemporary lead emissions are two additional
sources of environmental hazard exposure which disproportionately affect communities of color
and, for lead emissions and possibly to a lesser extent, communities characterized by low educa-
tional attainment.
Chapter 2 contextualizes the importance of urban residential lead contamination for older
adults, who may have experienced a lifetime of lead exposure in addition to continued lead ex-
posure from residential sources such as lead-based paint. In addition, Chapter 2 highlights the
importance of using built environment characteristics to identify and characterize the extent of en-
vironmental lead and lead-related risk by determining that lead-based paint presence can be char-
acterized to a detailed extent using information on housing age, use type, household composition,
and householder race. These two combined approaches underscore the importance of framing lead
contamination, persistence, risk, exposure, and health concerns in a whole environment, lifelong
framework which also incorporates a sociological contextualization.
Chapter 3 explores the relationship between individual and neighborhood demographic and
socioeconomic characteristics and risk of airborne lead exposure for older adults. The results
of Chapter 3 reinforce what is already known about the environmental justice contextualization
of environmental health exposure and adds to this knowledge by providing information on this
relationship for older adults and exposure to airborne lead emissions. The results of this chapter
underscore the importance of approaching environmental exposure and other public health issues
through an environmental justice lens and with consideration for historical and ongoing structural
pressures which have results in disproportionate exposure for certain groups and populations in the
United States. 97
102
Finally, Chapter 4 expands the airborne lead exposure focus of Chapter 3 and studies the rela-
tionship between this exposure and cognitive health outcomes for older adults in the United States.
The results of this chapter highlight airborne lead emissions as an important potential source of
adverse outcomes in older adult cognition and change in cognition over time but also further sup-
port the conclusions from Chapter 3 which show the importance of contextualizing lead exposure
within environmental justice frameworks.
The older population is the fastest growing population in the United States and health concerns
related to older adults have become and will continue to be an increasingly important focus. The
neurocognitive outcomes associated with chronic and lifetime lead exposure will continue to be-
come increasingly important as the population which grew up during peak lead use reaches old age
and experiences the health outcomes associated with chronic exposure in older adults. Considering
many of the outcomes are neurocognitive or motor related, such as increased dementia incidence
and increased fall risk, the increased burden of chronic lifetime exposure in older populations will
also result in an increased care burden and pressure on family members who provide care to their
older relatives and on the older adult care profession and industry.
103
References
1. Needleman, H. L. History of Lead poisoning in the world. Climate Change 2013 - The
PhysicalScienceBasis 53, 1–30. doi:10.1017/CBO9781107415324.004 (9 2013).
2. EPA), U. S. E. P. A. ( Lead at Superfund Sites: Risk Assessment https://www.epa.gov/
superfund/lead-superfund-sites-risk-assessment.
3. Wu, J., Edwards, R., He, X. (, Liu, Z. & Kleinman, M. Spatial analysis of bioavailable
soil lead concentrations in Los Angeles, California.EnvironmentalResearch 110, 309–317.
doi:10.1016/j.envres.2010.02.004 (4 2010).
4. Of Environmental Health Hazard Assessment, C. O. Development of health criteria for
school site risk assessment pursuant to health and safety code section 901(g): child-specific
benchmark change in blood lead concentration for school site risk assessment. 901, 106
(April 2007).
5. Laidlaw, M. A. S. & Filippelli, G. M. Resuspension of urban soils as a persistent source of
lead poisoning in children: A review and new directions. Applied Geochemistry 23, 2021–
2039. doi:10.1016/j.apgeochem.2008.05.009 (2008).
6. Mielke, H. W., Gonzales, C. R., Smith, M. K. & Mielke, P. W. The Urban Environment
and Children’s Health: Soils as an Integrator of Lead, Zinc, and Cadmium in New Orleans,
Louisiana, U.S.A. Environmental Research 81, 117–129. doi:10.1006/enrs.1999.3966
(1999).
7. Laidlaw, M. A., Zahran, S., Mielke, H. W., Taylor, M. P. & Filippelli, G. M. Re-suspension
of lead contaminated urban soil as a dominant source of atmospheric lead in Birmingham,
Chicago, Detroit and Pittsburgh, USA. Atmospheric Environment 49, 302–310. doi:10.
1016/j.atmosenv.2011.11.030 (2012).
8. Lanphear, B. P. & Roghmann, K. J. Pathways of Lead Exposure in Urban Children. Envi-
ronmentalResearch 74, 67–73. doi:10.1006/enrs.1997.3726 (1997).
104
9. Chen, L., Xu, Z., Liu, M., Huang, Y ., Fan, R., Su, Y .,etal. Lead exposure assessment from
study near a lead-acid battery factory in China. Science of The Total Environment. Special
Section - Arsenic in Latin America, An Unrevealed Continent: Occurence, Health Effects
andMitigation 429, 191–198. doi:10.1016/j.scitotenv.2012.04.015 (2012).
10. Qu, C.-S., Ma, Z.-W., Yang, J., Liu, Y ., Bi, J. & Huang, L. Human Exposure Pathways
of Heavy Metals in a Lead-Zinc Mining Area, Jiangsu Province, China. PLOS ONE 7.
Publisher: Public Library of Science, e46793. doi:10.1371/journal.pone.0046793
(2012).
11. Pizzol, M., Thomsen, M. & Andersen, M. S. Long-term human exposure to lead from
different media and intake pathways. Science of The Total Environment 408, 5478–5488.
doi:10.1016/j.scitotenv.2010.07.077 (2010).
12. Triantafyllidou, S. & Edwards, M. Lead (Pb) in Tap Water and in Blood: Implications for
Lead Exposure in the United States.CriticalReviewsinEnvironmentalScienceandTechnol-
ogy 42. Publisher: Taylor & Francis eprint: https://doi.org/10.1080/10643389.2011.556556,
1297–1352. doi:10.1080/10643389.2011.556556 (2012).
13. Maas, R. P., Patch, S. C., Morgan, D. M. & Pandolfo, T. J. Reducing Lead Exposure from
Drinking Water: Recent History and Current Status. Public Health Reports (1974-) 120.
Publisher: Sage Publications, Inc., 316–321 (2005).
14. Mielke, H. W. & Reagan, P. L. Soil is an important pathway of human lead exposure. En-
vironmental Health Perspectives 106. Publisher: Environmental Health Perspectives, 217–
229. doi:10.1289/ehp.98106s1217 (suppl 1 1998).
15. Filon, F. L., Boeniger, M., Maina, G., Adami, G., Spinelli, P. & Damian, A. Skin Absorption
of Inorganic Lead (PbO) and the Effect of Skin Cleansers. Journal of Occupational and
EnvironmentalMedicine 48. Publisher: Lippincott Williams & Wilkins, 692–699 (2006).
16. For Disease Control (CDC), C. Response to Advisory Committee on Childhood Lead Poi-
soning Prevention Recommendations in “Low Level Lead Exposure Harms Children: A
Renewed Call of Primary Prevention”.MorbidityMortalityWeekly 61, 383 (20 2012).
17. Bellinger, D. C. Very low lead exposures and children’s neurodevelopment.CurrentOpinion
inPediatrics 20, 172. doi:10.1097/MOP.0b013e3282f4f97b (2008).
18. Bruce P Lanphear, Richard Hornung, Jane Khoury, Kimberly Yolton, Peter Baghurst, David
C Bellinger, et al. Low-Level Environmental Lead Exposure and Children’s Intellectual
Function: An International Pooled Analysis.EnvironmentalHealthPerspectives 113 (2005).
105
19. Lidsky, T. I. & Schneider, J. S. Lead neurotoxicity in children: basic mechanisms and clini-
cal correlates.Brain 126, 5–19. doi:10.1093/brain/awg014 (2003).
20. Chiodo, L. M., Jacobson, S. W. & Jacobson, J. L. Neurodevelopmental effects of postnatal
lead exposure at very low levels. Neurotoxicology and Teratology 26, 359–371. doi:10.
1016/j.ntt.2004.01.010 (2004).
21. Sciarillo, W. G., Alexander, G. & Farrell, K. P. Lead exposure and child behavior.American
Journal of Public Health 82. Publisher: American Public Health Association, 1356–1360.
doi:10.2105/AJPH.82.10.1356 (1992).
22. Zahran, S., Laidlaw, M. A. S., Rowe, D. B., Ball, A. S. & Mielke, H. W. Motor neuron
disease mortality and lifetime petrol lead exposure: Evidence from national age-specific and
state-level age-standardized death rates in Australia.EnvironmentalResearch 153, 181–190.
doi:10.1016/j.envres.2016.11.023 (2017).
23. Dietrich, K. N., Succop, P. A. & Berger, O. G. Lead Exposure and the Motor Developmental
Status of Urban Six-Year-Old Children in the Cincinnati Prospective Study. Pediatrics 91,
301–307. doi:10.1542/peds.91.2.301 (1993).
24. Fels, L. M., W¨ unsch, M., Baranowski, J., Norska-Bor´ owka, I., Price, R. G., Taylor, S. A.,
et al. Adverse effects of chronic low level lead exposure on kidney function - A risk group
study in children.NephrologyDialysisTransplantation 13, 2248–2256. doi:10.1093/ndt/
13.9.2248 (9 1998).
25. Banks, E. C., Ferretti, L. E. & Shucard, D. W. Effects of low level lead exposure on cognitive
function in children: a review of behavioral, neuropsychological and biological evidence.
Neurotoxicology 18, 237–281 (1997).
26. Navas-Acien, A., Guallar, E., Silbergeld, E. K. & Rothenberg, S. J. Lead exposure and
cardiovascular disease - A systematic review.EnvironmentalHealthPerspectives 115, 472–
482. doi:10.1289/ehp.9785 (3 2007).
27. Wu, T., Buck, G. M. & Mendola, P. Blood lead and sexual maturation in U.S. girls: The
third national health and nutrition examination survey, 1988-1994. Environmental Health
Perspectives 111, 737–741. doi:10.1289/ehp.6008 (5 2003).
28. Selevan, S. G., Rice, D. C., Hogan, K. A., Euling, S. Y ., Pfahles-Hutchens, A. & Bethel, J.
Blood lead concentration and delayed puberty in girls. New England Journal of Medicine
348, 1527–1536. doi:10.1056/NEJMoa020880 (16 2003).
106
29. Bellinger, D., Leviton, A., Waternaux, C., Needleman, H. & Rabinowitz, M. Longitudi-
nal Analyses of Prenatal and Postnatal Lead Exposure and Early Cognitive Development.
New England Journal of Medicine 316. Publisher: Massachusetts Medical Society eprint:
https://doi.org/10.1056/NEJM198704233161701, 1037–1043. doi:10.1056/NEJM198704233161701
(1987).
30. Bellinger, D., Leviton, A., Sloman, J., Rabinowitz, M., Needleman, H. L. & Waternaux,
C. Low-Level Lead Exposure and Children’s Cognitive Function in the Preschool Years.
Pediatrics 87, 219–227. doi:10.1542/peds.87.2.219 (1991).
31. Bellinger, D. & Dietrich, K. N. Low-Level Lead Exposure and Cognitive Function in Chil-
dren.PediatricAnnals 23. Publisher: SLACK Incorporated, 600–605. doi:10.3928/0090-
4481-19941101-08 (1994).
32. Needleman, H. L. & Gatsonis, C. A. Low-Level Lead Exposure and the IQ of Children:
A Meta-analysis of Modern Studies. JAMA 263, 673–678. doi:10.1001/jama.1990.
03440050067035 (1990).
33. Schwartz, J. Low-Level Lead Exposure and Childrens IQ: A Metaanalysis and Search for a
Threshold.EnvironmentalResearch 65, 42–55. doi:10.1006/enrs.1994.1020 (1994).
34. Needleman, H. L., McFarland, C., Ness, R. B., Fienberg, S. E. & Tobin, M. J. Bone lead
levels in adjudicated delinquents: A case control study.NeurotoxicologyandTeratology 24,
711–717. doi:10.1016/S0892-0362(02)00269-6 (2002).
35. Koller, K., Brown, T., Spurgeon, A. & Levy, L. Recent Developments in Low-Level Lead
Exposure and Intellectual Impairment in Children.EnvironmentalHealthPerspectives 112.
Publisher: Environmental Health Perspectives, 987–994. doi:10.1289/ehp.6941 (2004).
36. NORRIS, C. & GETTLER, A. O. POISONING BY TETRA-ETHYL LEAD: POSTMORTEM
AND CHEMICAL FINDINGS.JournaloftheAmericanMedicalAssociation 85, 818–820.
doi:10.1001/jama.1925.02670110032009 (1925).
37. Stollery, B. T. Reaction time changes in workers exposed to lead. Neurotoxicology and
Teratology.SymposiumonComputerizedBehavioralTestingofHumansinNeurotoxicology
Research 18, 477–483. doi:10.1016/0892-0362(96)00080-3 (1996).
38. Bleecker, M. L., Lindgren, K. N. & Ford, P. D. Differential contribution of current and
cumulative indices of lead dose to neuropsychological performance by age. Neurology 48.
Publisher: Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology
Section: Article, 639–645. doi:10.1212/WNL.48.3.639 (1997).
107
39. Payton, M., Riggs, K. M., Spiro, A., Weiss, S. T. & Hu, H. Relations of Bone and Blood
Lead to Cognitive Function: The V A Normative Aging Study. Neurotoxicology and Tera-
tology 20, 19–27. doi:10.1016/S0892-0362(97)00075-5 (1998).
40. Stewart, W. F., Schwartz, B. S., Simon, D., Bolla, K. I., Todd, A. C. & Links, J. Neurobe-
havioral function and tibial and chelatable lead levels in 543 former organolead workers.
Neurology 52. Publisher: Wolters Kluwer Health, Inc. on behalf of the American Academy
of Neurology Section: Article, 1610–1610. doi:10.1212/WNL.52.8.1610 (1999).
41. Benes, F. M., Turtle, M., Khan, Y . & Farol, P. Myelination of a Key Relay Zone in the
Hippocampal Formation Occurs in the Human Brain During Childhood, Adolescence, and
Adulthood. Archives of General Psychiatry 51, 477–484. doi:10.1001/archpsyc.1994.
03950060041004 (1994).
42. Yokoyama, K., Araki, S. & Aono, H. Reversibility of psychological performance in sub-
clinical lead absorption.Neurotoxicology 9, 405–410 (1988).
43. Muldoon, S. B., Cauley, J. A., Kuller, L. H., Morrow, L., Needleman, H. L., Scott, J., et al.
Effects of Blood Lead Levels on Cognitive Function of Older Women. Neuroepidemiology
15. Publisher: Karger Publishers, 62–72. doi:10.1159/000109891 (1996).
44. Stokes, L., Letz, R., Gerr, F., Kolczak, M., McNeill, F. E., Chettle, D. R., et al. Neurotoxi-
city in young adults 20 years after childhood exposure to lead: the Bunker Hill experience.
OccupationalandEnvironmentalMedicine 55. Publisher: BMJ Publishing Group Ltd, 507–
516. doi:10.1136/oem.55.8.507 (1998).
45. Links, J. M., Schwartz, B. S., Simon, D. & Bandeen-Roche, K. The Influence of Toxicant
”Residence Time” and Bioavailability from Body Stores in Estimation of Cumulative Tar-
get Organ Dose: Application to Lead-Associated Neurocognitive Decline. Environmental
HealthPerspectives (2001).
46. Wright, R. O., Tsaih, S. W., Schwartz, J., Spiro, A., McDonald, K., Weiss, S. T.,etal. Lead
Exposure Biomarkers and Mini-Mental Status Exam Scores in Older Men. Epidemiology
14. Publisher: Lippincott Williams & Wilkins, 713–718 (2003).
47. Beattie, A. D., Moore, M. R. & Goldberg, A. TETRAETHYL-LEAD POISONING. The
Lancet. Originally published as Volume 2, Issue 7766 300, 12–15. doi:10.1016/S0140-
6736(72)91276-7 (1972).
48. Weisskopf, M. G., Wright, R. O., Schwartz, J., Spiro III, A., Sparrow, D., Aro, A., et al.
Cumulative Lead Exposure and Prospective Change in Cognition among Elderly Men: The
108
V A Normative Aging Study. American Journal of Epidemiology 160, 1184–1193. doi:10.
1093/aje/kwh333 (2004).
49. Bleecker, M. L., Ford, D. P., Lindgren, K. N., Hoese, V . M., Walsh, K. S. & Vaughan, C. G.
Differential effects of lead exposure on components of verbal memory. Occupational and
EnvironmentalMedicine 62. Publisher: BMJ Publishing Group Ltd Section: Original article,
181–187. doi:10.1136/oem.2003.011346 (2005).
50. Schafer, J. H., Glass, T. A., Bolla, K. I., Mintz, M., Jedlicka, A. E. & Schwartz, B. S. Homo-
cysteine and Cognitive Function in a Population-Based Study of Older Adults.Journalofthe
AmericanGeriatricsSociety 53. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1532-
5415.2005.53153.x, 381–388. doi:10.1111/j.1532-5415.2005.53153.x (2005).
51. Winker, R., Barth, A., Ponocny-Seliger, E., Pilger, A., Osterode, W. & R¨ udiger, H. W. No
cognitive deficits in men formerly exposed to lead. Wiener Klinische Wochenschrift 117,
755–760. doi:10.1007/s00508-005-0466-0 (2005).
52. Stewart, W. F., Schwartz, B. S., Davatzikos, C., Shen, D., Liu, D., Wu, X., et al. Past adult
lead exposure is linked to neurodegeneration measured by brain MRI. Neurology 66. Pub-
lisher: Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology Sec-
tion: Articles, 1476–1484. doi:10.1212/01.wnl.0000216138.69777.15 (2006).
53. Winker, R., Ponocny-Seliger, E., R¨ udiger, H. W. & Barth, A. Lead exposure levels and
duration of exposure absence predict neurobehavioral performance. International Archives
of Occupational and Environmental Health 79, 123–127. doi:10.1007/s00420-005-
0031-1 (2006).
54. Shih, R. A., Glass, T. A., Bandeen-Roche, K., Carlson, M. C., Bolla, K. I., Todd, A. C.,etal.
Environmental lead exposure and cognitive function in community-dwelling older adults.
Neurology 67. ISBN: 9780000239839 Publisher: Wolters Kluwer Health, Inc. on behalf of
the American Academy of Neurology Section: Articles, 1556–1562. doi:10.1212/01.
wnl.0000239836.26142.c5 (2006).
55. Weisskopf, M. G., Proctor, S. P., Wright, R. O., Schwartz, J., Spiro, A., Sparrow, D., et al.
Cumulative Lead Exposure and Cognitive Performance among Elderly Men.Epidemiology
18. Publisher: Lippincott Williams & Wilkins, 59–66 (2007).
56. Bandeen-Roche, K., Glass, T. A., Bolla, K. I., Todd, A. C. & Schwartz, B. S. Cumulative
Lead Dose and Cognitive Function in Older Adults.Epidemiology 20. Publisher: Lippincott
Williams & Wilkins, 831–839 (2009).
109
57. Glass, T. A., Bandeen-Roche, K., McAtee, M., Bolla, K., Todd, A. C. & Schwartz, B. S.
Neighborhood Psychosocial Hazards and the Association of Cumulative Lead Dose With
Cognitive Function in Older Adults. American Journal of Epidemiology 169, 683–692.
doi:10.1093/aje/kwn390 (2009).
58. Mantere, P., H¨ anninen, H., Hernberg, S. & Luukkonen, R. A prospective follow-up study
on psychological effects in workers exposed to low levels of lead. Scandinavian Journal of
Work,Environment&Health 10. Publisher: Scandinavian Journal of Work, Environment &
Health, 43–50 (1984).
59. Weuve, J., Korrick, S. A., Weisskopf, M. A., Ryan, L. M., Schwartz, J., Nie, H., et al.
Cumulative Exposure to Lead in Relation to Cognitive Function in Older Women.Environ-
mental Health Perspectives 117. Publisher: Environmental Health Perspectives, 574–580.
doi:10.1289/ehp.11846 (2009).
60. Power, M. C., Korrick, S., Tchetgen Tchetgen, E. J., Nie, L. H., Grodstein, F., Hu, H., et
al. Lead exposure and rate of change in cognitive function in older women. Environmental
Research 129, 69–75. doi:10.1016/j.envres.2013.12.010 (2014).
61. FENGA, C., GANGEMI, S., ALIBRANDI, A., COSTA, C. & MICALI, E. Relationship
between lead exposure and mild cognitive impairment.JournalofPreventiveMedicineand
Hygiene 57, E205–E210 (2016).
62. Farooqui, Z., Bakulski, K. M., Power, M. C., Weisskopf, M. G., Sparrow, D., Spiro, A., et
al. Associations of cumulative Pb exposure and longitudinal changes in Mini-Mental Status
Exam scores, global cognition and domains of cognition: The V A Normative Aging Study.
EnvironmentalResearch 152, 102–108. doi:10.1016/j.envres.2016.10.007 (2017).
63. Fuller-Thomson, E. & Jopling, S. A. Could the Adoption of Unleaded Gasoline in the 1970s
Play a Role in the Precipitous Decline in Dementia Prevalence? JAMA Internal Medicine
177, 892–893. doi:10.1001/jamainternmed.2017.1074 (2017).
64. Mansouri, M. T., Mu˜ noz-Fambuena, I. & Cauli, O. Cognitive impairment associated with
chronic lead exposure in adults.Neurology,PsychiatryandBrainResearch 30, 5–8. doi:10.
1016/j.npbr.2018.04.001 (2018).
65. Ram´ ırez Ortega, D., Gonz´ alez Esquivel, D. F., Blanco Ayala, T., Pineda, B., G´ omez Manzo,
S., Marcial Quino, J.,etal. Cognitive Impairment Induced by Lead Exposure during Lifes-
pan: Mechanisms of Lead Neurotoxicity. Toxics 9, 23. doi:10.3390/toxics9020023
(2021).
110
66. Meyer-Baron, M. & Seeber, A. A meta-analysis for neurobehavioural results due to occu-
pational lead exposure with blood lead concentrations ¡70 g/100 ml.ArchivesofToxicology
73, 510–518. doi:10.1007/s002040050002 (2000).
67. Schwartz, B. S., Lee, B. K., Lee, G. S., Stewart, W. F., Simon, D., Kelsey, K.,etal. Associ-
ations of blood lead, dimercaptosuccinic acid-chelatable lead, and tibia lead with polymor-
phisms in the vitamin D receptor and [delta]-aminolevulinic acid dehydratase genes. En-
vironmental Health Perspectives 108. Publisher: Environmental Health Perspectives, 949–
954. doi:10.1289/ehp.00108949 (2000).
68. Goodman, M., LaVerda, N., Clarke, C., Foster, E. D., Iannuzzi, J. & Mandel, J. Neurobe-
havioural testing in workers occupationally exposed to lead: systematic review and meta-
analysis of publications. Occupational and Environmental Medicine 59. Publisher: BMJ
Publishing Group Ltd Section: Review, 217–223. doi:10.1136/oem.59.4.217 (2002).
69. Baker, E. L., White, R. F., Pothier, L. J., Berkey, C. S., Dinse, G. E., Travers, P. H., et
al. Occupational lead neurotoxicity: improvement in behavioural effects after reduction of
exposure.OccupationalandEnvironmentalMedicine 42. Publisher: BMJ Publishing Group
Ltd Section: Research Article, 507–516. doi:10.1136/oem.42.8.507 (1985).
70. Wilson, R. S., Barnes, L. L., Leon, C. F. M. d., Aggarwal, N. T., Schneider, J. S., Bach, J.,et
al. Depressive symptoms, cognitive decline, and risk of AD in older persons.Neurology 59.
Publisher: Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology
Section: Articles, 364–370. doi:10.1212/WNL.59.3.364 (2002).
71. Fiedler, N., Weisel, C., Lynch, R., Kelly-McNeil, K., Wedeen, R., Jones, K.,etal. Cognitive
effects of chronic exposure to lead and solvents. American Journal of Industrial Medicine
44. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/ajim.10287, 413–423. doi:10.
1002/ajim.10287 (2003).
72. Hu, H., Shih, R., Rothenberg, S. & Schwartz, B. S. The Epidemiology of Lead Toxicity in
Adults: Measuring Dose and Consideration of Other Methodologic Issues. Environmental
Health Perspectives 115. Publisher: Environmental Health Perspectives, 455–462. doi:10.
1289/ehp.9783 (2007).
73. Walsh, T. J., McLamb, R. L., Bondy, S. C., Tilson, H. A. & Chang, L. W. Triethyl and
trimethyl lead: effects on behavior, CNS morphology and concentrations of lead in blood
and brain of rat.Neurotoxicology 7, 21–33 (1986).
74. Stollery, B. T., Broadbent, D. E., Banks, H. A. & Lee, W. R. Short term prospective study
of cognitive functioning in lead workers. Occupational and Environmental Medicine 48.
111
Publisher: BMJ Publishing Group Ltd Section: Research Article, 739–749. doi:10.1136/
oem.48.11.739 (1991).
75. Schwartz, B. S., Bolla, K. I., Stewart, W., Ford, D. P., Agnew, J. & Frumkin, H. Decrements
in Neurobehavioral Performance Associated with Mixed Exposure to Organic and Inorganic
Lead.AmericanJournalofEpidemiology 137, 1006–1021. doi:10.1093/oxfordjournals.
aje.a116757 (1993).
76. Balbus-Kornfeld, J. M., Stewart, W., Bolla, K. I. & Schwartz, B. S. Cumulative exposure
to inorganic lead and neurobehavioural test performance in adults: an epidemiological re-
view.OccupationalandEnvironmentalMedicine 52. Publisher: BMJ Publishing Group Ltd
Section: Research Article, 2–12. doi:10.1136/oem.52.1.2 (1995).
77. Lindgren, K. N., Masten, V . L., Ford, D. P. & Bleecker, M. L. Relation of cumulative ex-
posure to inorganic lead and neuropsychological test performance.OccupationalandEnvi-
ronmental Medicine 53. Publisher: BMJ Publishing Group Ltd Section: Research Article,
472–477. doi:10.1136/oem.53.7.472 (1996).
78. Legislative History of Lead-Based Paint.
79. Humphreys, E. H. & Tiemann, M. Safe Drinking Water Act (SDWA): A Summary of the
Act and Its Major Requirements.
80. Canfield, R. L., Henderson, C. R., Cory-Slechta, D. A., Cox, C., Jusko, T. A. & Lanphear,
B. P. Intellectual Impairment in Children with Blood Lead Concentrations below 10 g per
Deciliter.NewEnglandJournalofMedicine 348. Publisher: Massachusetts Medical Society
eprint: https://doi.org/10.1056/NEJMoa022848, 1517–1526. doi:10.1056/NEJMoa022848
(2003).
81. Chaney, R., Mielke, H. & Sterrett, S. Speciation, Mobility and Bioavailability of Soil Lead.
LeadinSoil:IssuesandGuidelines 11, 105–129 (1989).
82. United States Department of Housing and Urban Development. Units of Measure Used in
theLead-BasedPaintField ().
83. Lead-Based Paint Environment, Health and Safety. https://ehs.unc.edu/topics/
lead-based-paint/ (2023).
84. National Research Council (US) Committee on Measuring Lead in Critical Populations. in
Measuring Lead Exposure in Infants, Children, and Other Sensitive Populations (National
Academies Press (US), 1993).
112
85. White, R. F., Diamond, R., Proctor, S., Morey, C. & Hu, H. Residual cognitive deficits 50
years after lead poisoning during childhood.BritishJournalofIndustrialMedicine 50, 613–
622 (1993).
86. Papanikolaou, N. C., Hatzidaki, E. G., Belivanis, S., Tzanakakis, G. N. & Tsatsakis, A. M.
Lead toxicity update. A brief review.MedicalScienceMonitor:InternationalMedicalJour-
nalofExperimentalandClinicalResearch 11, RA329–336 (2005).
87. Drasch, G. A. & Ott, J. Lead in human bones. Investigations on an occupationally non-
exposed population in Southern Bavaria (F.R.G) II. Children.ScienceofTheTotalEnviron-
ment 68, 61–69. doi:10.1016/0048-9697(88)90361-0 (1988).
88. Drasch, G. A., B¨ ohm, J. & Baur, C. Lead in human bones. Investigations on an occupation-
ally non-exposed population in Southern Bavaria (F.R.G.) I. Adults. Science of The Total
Environment 64, 303–315. doi:10.1016/0048-9697(87)90252-X (1987).
89. Wittmers, L. E., Wallgren, J., Alich, A., Aufderheide, A. C. & Rapp, G. Lead in Bone. IV.
Distribution of Lead in the Human Skeleton.ArchivesofEnvironmentalHealth 43.
90. Bakulski, K. M., Seo, Y . A., Hickman, R. C., Brandt, D., Vadari, H. S., Hu, H.,etal. Heavy
Metals Exposure and Alzheimer’s Disease and Related Dementias. Journal of Alzheimer’s
Disease 76. Publisher: IOS Press, 1215–1242. doi:10.3233/JAD-200282 (2020).
91. H¨ anninen, H., Aitio, A., Kovala, T., Luukkonen, R., Matikainen, E., Mannelin, T., et al.
Occupational exposure to lead and neuropsychological dysfunction. Occupational and En-
vironmental Medicine 55. Publisher: BMJ Publishing Group Ltd, 202–209. doi:10.1136/
oem.55.3.202 (1998).
92. Links, J. M., Schwartz, B. S., Simon, D., Bandeen, .-. K. & Stewart, W. F. Characteriza-
tion of toxicokinetics and toxicodynamics with linear systems theory: application to lead-
associated cognitive decline. Environmental Health Perspectives 109. Publisher: Environ-
mental Health Perspectives, 361–368. doi:10.1289/ehp.01109361 (2001).
93. Cecil, K. M., Brubaker, C. J., Adler, C. M., Dietrich, K. N., Altaye, M., Egelhoff, J. C., et
al. Decreased Brain V olume in Adults with Childhood Lead Exposure. PLOS Medicine 5.
Publisher: Public Library of Science, e112. doi:10.1371/journal.pmed.0050112 (2008).
94. Scahill, R. I., Frost, C., Jenkins, R., Whitwell, J. L., Rossor, M. N. & Fox, N. C. A Longitu-
dinal Study of Brain V olume Changes in Normal Aging Using Serial Registered Magnetic
Resonance Imaging. Archives of Neurology 60, 989–994. doi:10.1001/archneur.60.7.
989 (2003).
113
95. Sun, L., Zhou, X.-L., Yi, H.-P., Jiang, S.-J. & Yuan, H. Lead-induced morphological changes
and amyloid precursor protein accumulation in adult rat hippocampus.Biotechnic&Histo-
chemistry 89. Publisher: Taylor & Francis eprint: https://doi.org/10.3109/10520295.2014.904926,
513–517. doi:10.3109/10520295.2014.904926 (2014).
96. Fathabadi, B., Dehghanifiroozabadi, M., Aaseth, J., Sharifzadeh, G., Nakhaee, S., Rajabpour-
Sanati, A., et al. Comparison of Blood Lead Levels in Patients With Alzheimer’s Disease
and Healthy People. American Journal of Alzheimer’s Disease & Other Dementias® 33.
Publisher: SAGE Publications Inc, 541–547. doi:10.1177/1533317518794032 (2018).
97. Martin, D., Glass, T. A., Bandeen-Roche, K., Todd, A. C., Shi, W. & Schwartz, B. S. Associ-
ation of Blood Lead and Tibia Lead with Blood Pressure and Hypertension in a Community
Sample of Older Adults. American Journal of Epidemiology 163, 467–478. doi:10.1093/
aje/kwj060 (2006).
98. Garc´ ıa-Esquinas, E., Navas-Acien, A., P´ erez-G´ omez, B. & Artalejo, F. R. Association of
lead and cadmium exposure with frailty in US older adults. Environmental Research 137,
424–431. doi:10.1016/j.envres.2015.01.013 (2015).
99. Khalil, N., Cauley, J. A., Wilson, J. W., Talbott, E. O., Morrow, L., Hochberg, M. C., et
al. Relationship of Blood Lead Levels to Incident Nonspine Fractures and Falls in Older
Women: The Study of Osteoporotic Fractures. Journal of Bone and Mineral Research 23.
eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1359/jbmr.080404, 1417–1425. doi:10.
1359/jbmr.080404 (2008).
100. Khalil, N., Faulkner, K. A., Greenspan, S. L., Cauley, J. A. & Group, T. O. F. i. M. R. Associ-
ations Between Bone Mineral Density, Grip Strength, and Lead Body Burden in Older Men.
JournaloftheAmericanGeriatricsSociety 62. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jgs.12603,
141–146. doi:10.1111/jgs.12603 (2014).
101. Garc´ ıa-Esquinas, E. & Rodr´ ıguez-Artalejo, F. Environmental Pollutants, Limitations in Phys-
ical Functioning, and Frailty in Older Adults.CurrentEnvironmentalHealthReports 4, 12–
20. doi:10.1007/s40572-017-0128-1 (2017).
102. Garc´ ıa-Esquinas, E. G., P´ erez-G´ omez, B., Junghoon, K., Navas-Acien, A., Choi, Y .-H. &
Rodr´ ıguez-Artalejo, F. P II – 3–6 Exposure to environmental pollutants and frailty in older
adults. Occupational and Environmental Medicine 75. Publisher: BMJ Publishing Group
Ltd Section: Big data, bio allergens and metal exposure and its health effects, A45–A45.
doi:10.1136/oemed-2018-ISEEabstracts.113 (Suppl 1 2018).
114
103. Wang*, W., Schaumberg, D. & Park, S. K. Cumulative Lead Exposure and Incidence of
Age-related Cataract in Older Men.ISEEConferenceAbstracts. doi:10.1289/isee.2016.
4135 (2016).
104. Stewart, W. F. & Schwartz, B. S. Effects of lead on the adult brain: A 15-year exploration.
AmericanJournalofIndustrialMedicine 50. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/ajim.20434,
729–739. doi:10.1002/ajim.20434 (2007).
105. Lee, R. C., Fricke, J. R., Wright, W. E. & Haerer, W. Development of a probabilistic
blood lead prediction model.EnvironmentalGeochemistryandHealth 17, 169–181. doi:10.
1007/BF00661329 (1995).
106. United States Environmental Protection Agency. Integrated Exposure Uptake Biokinetic
(IEUBK)ModelforLeadinChildren 2015.
107. California, S. o.SoilSamplingDatafortheExidePreliminaryInvestigationArea/Datosde
muestreo de suelo para el ´ area de investigaci´ on preliminar de Exide Department of Toxic
Substances Control. https://dtsc.ca.gov/soil-sampling-data-for-the-exide-
preliminary-investigation-area/ (2023).
108. County of Los Angeles.LosAngelesCountyParcels(2014) 2019.
109. United States Department of Housing and Urban Development. 2019 American Housing
Survey 2019.
110. Schwartz, B. S., Lee, B.-K., Bandeen-Roche, K., Stewart, W., Bolla, K., Links, J.,etal. Oc-
cupational lead exposure and longitudinal decline in neurobehavioral test scores. Epidemi-
ology (Cambridge, Mass.) 16, 106–113. doi:10.1097/01.ede.0000147109.62324.51
(2005).
111. Whitehead, L. S. & Buchanan, S. D. Childhood Lead Poisoning: A Perpetual Environmental
Justice Issue?JournalofPublicHealthManagementandPractice 25, S115. doi:10.1097/
PHH.0000000000000891 (2019).
112. Ruckart, P. Z., Ettinger, A. S., Hanna-Attisha, M., Jones, N., Davis, S. I. & Breysse, P. N.
The Flint Water Crisis: A Coordinated Public Health Emergency Response and Recov-
ery Initiative. Journal of public health management and practice : JPHMP 25, S84–S90.
doi:10.1097/PHH.0000000000000871 (Suppl 1 LEAD POISONING PREVENTION
2019).
115
113. Lynch, E. E. & Meier, H. C. S. The intersectional effect of poverty, home ownership, and
racial/ethnic composition on mean childhood blood lead levels in Milwaukee County neigh-
borhoods. PLOS ONE 15 (ed Navaneetham, K.) e0234995. doi:10.1371/journal.pone.
0234995 (2020).
114. Kraft, M. E. & Scheberle, D. Environmental Justice and the Allocation Of Risk: The Case of
Lead and Public Health.PolicyStudiesJournal 23. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1541-
0072.1995.tb00511.x, 113–122. doi:10.1111/j.1541-0072.1995.tb00511.x (1995).
115. Sargent, J. D., Brown, M. J., Freeman, J. L., Bailey, A., Goodman, D. & Freeman, D. H.
Childhood lead poisoning in Massachusetts communities: its association with sociodemo-
graphic and housing characteristics.AmericanJournalofPublicHealth 85. Publisher: Amer-
ican Public Health Association, 528–534. doi:10.2105/AJPH.85.4.528 (1995).
116. US EPA, O. 2011 NATA: Assessment Results https://www.epa.gov/national-air-
toxics-assessment/2011-nata-assessment-results (2023).
117. Welcome—HealthandRetirementStudyhttps://hrs.isr.umich.edu/about (2023).
118. United States Census Bureau.2010DecennialCensus
119. United States Census Bureau.2006-2010AmericanCommunitySurvey5-YearEstimates
120. US EPA, O. Air Toxics Screening Assessment https://www.epa.gov/AirToxScreen
(2023).
121. 1996 National Air Toxics Assessment Activities— Technology Transfer Network Air Tech-
nical Web site — US EPA https://archive.epa.gov/airtoxics/nata/web/html/
index-2.html (2023).
122. 1999National-ScaleAirToxicsAssessment—TechnologyTransferNetworkAirToxicsWeb
site — US EPA https://archive.epa.gov/airtoxics/nata1999/web/html/index.
html (2023).
123. 2002National-ScaleAirToxicsAssessment—TechnologyTransferNetworkAirToxicsWeb
site—USEPAhttps://archive.epa.gov/nata2002/web/html/index.html (2023).
124. US EPA, O. 2005 NATA: Assessment Results https://www.epa.gov/national-air-
toxics-assessment/2005-nata-assessment-results (2023).
116
125. US EPA, O. 2014 NATA: Assessment Results https://www.epa.gov/national-air-
toxics-assessment/2014-nata-assessment-results (2023).
126. US EPA, O.2017AirToxScreen:AssessmentResultshttps://www.epa.gov/AirToxScreen/
2017-airtoxscreen-assessment-results (2023).
127. US EPA, O.2018AirToxScreen:AssessmentResultshttps://www.epa.gov/AirToxScreen/
2018-airtoxscreen-assessment-results (2023).
128. US EPA, O.2019AirToxScreen:AssessmentResultshttps://www.epa.gov/AirToxScreen/
2019-airtoxscreen-assessment-results (2023).
129. Environmental Protection Agency (EPA). Technical Support Document EPA’s 2011 National-
scale Air Toxics Assessment.
130. Office of Air Quality Planning,
bibinitperiod S.TechnicalSupportDocumentEPA’s2011National-scaleAirToxicsAssess-
ment (2015).
131. StataCorp.StataStatisticalSoftware:Release17 College Station, TX, 2021.
132. Virgolini, M. B. & Aschner, M. inAdvancesinNeurotoxicology (eds Aschner, M. & Costa,
L. G.) 159–213 (Academic Press, 2021). doi:10.1016/bs.ant.2020.11.002.
133. Lee, H., Lee, M. & Warren, J. R. Childhood Lead Exposure and Cognitive Functioning
Among Older Adults: Evidence From the Health and Retirement Study.InnovationinAging
5, 311. doi:10.1093/geroni/igab046.1212 (Suppl 1 2021).
134. Lee, H., Lee, M. W., Warren, J. R. & Ferrie, J. Childhood lead exposure is associated with
lower cognitive functioning at older ages.ScienceAdvances 8. Publisher: American Associ-
ation for the Advancement of Science, eabn5164. doi:10.1126/sciadv.abn5164 (2022).
135. Ellen, I. G., Mijanovich, T. & Dillman, K.-N. Neighborhood Effects on Health: Exploring
the Links and Assessing the Evidence. Journal of Urban Affairs 23. Publisher: Routledge
eprint: https://doi.org/10.1111/0735-2166.00096, 391–408. doi:10.1111/0735-2166.
00096 (2001).
117
136. Arcaya, M. C., Tucker-Seeley, R. D., Kim, R., Schnake-Mahl, A., So, M. & Subramanian,
S. V . Research on neighborhood effects on health in the United States: A systematic re-
view of study characteristics. Social Science & Medicine 168, 16–29. doi:10.1016/j.
socscimed.2016.08.047 (2016).
137. Chyn, E. & Katz, L. F. Neighborhoods Matter: Assessing the Evidence for Place Effects.
JournalofEconomicPerspectives 35, 197–222. doi:10.1257/jep.35.4.197 (2021).
138. Crimmins, E. M., Kim, J. K., Langa, K. M. & Weir, D. R. Assessment of Cognition Us-
ing Surveys and Neuropsychological Assessment: The Health and Retirement Study and
the Aging, Demographics, and Memory Study.TheJournalsofGerontologySeriesB:Psy-
chological Sciences and Social Sciences 66B, i162–i171. doi:10.1093/geronb/gbr048
(Suppl 1 2011).
139. Herzog, A. R. & Wallace, R. B. Measures of Cognitive Functioning in the AHEAD Study.
The Journals of Gerontology: Series B 52B, 37–48. doi:10.1093/geronb/52B.Special_
Issue.37 (Special Issue 1997).
140. Brandt, J., Spencer, M. & Folstein, M. The Telephone Interview for Cognitive Status. Neu-
ropsychiatry,Neuropsychology,&BehavioralNeurology 1. Place: US Publisher: Lippincott
Williams & Wilkins, 111–117 (1988).
141. Benson, S. M., Talbott, E. O., Brink, L. L., Wu, C., Sharma, R. K. & Marsh, G. M. En-
vironmental lead and childhood blood lead levels in US children: NHANES, 1999–2006.
ArchivesofEnvironmental&OccupationalHealth 72. Publisher: Routledge eprint: https://doi.org/10.1080/19338244.2016.1157454,
70–78. doi:10.1080/19338244.2016.1157454 (2017).
142. Brink, L. L., Talbott, E. O., Sharma, R. K., Marsh, G. M., Wu, W. C., Rager, J. R.,etal. Do
US Ambient Air Lead Levels Have a Significant Impact on Childhood Blood Lead Levels:
Results of a National Study. Journal of Environmental and Public Health 2013. Publisher:
Hindawi, e278042. doi:10.1155/2013/278042 (2013).
143. Brink, L. A., Talbott, E. O., Marsh, G. M., Sharma, R., Benson, S., Wu, W. C., et al. Re-
visiting Nonresidential Environmental Exposures and Childhood Lead Poisoning in the US:
Findings from Kansas, 2000–2005.JournalofEnvironmentalandPublicHealth 2016. Pub-
lisher: Hindawi, e8791686. doi:10.1155/2016/8791686 (2016).
144. Palacios, N., Fitzgerald, K., Roberts, A. L., Hart, J. E., Weisskopf, M. G., Schwarzschild,
M. A., et al. A Prospective Analysis of Airborne Metal Exposures and Risk of Parkinson
Disease in the Nurses’ Health Study Cohort. Environmental Health Perspectives 122. Pub-
lisher: Environmental Health Perspectives, 933–938. doi:10.1289/ehp.1307218 (2014).
118
145. deCastro, B. R. Acrolein and Asthma Attack Prevalence in a Representative Sample of the
United States Adult Population 2000 – 2009. PLOS ONE 9. Publisher: Public Library of
Science, e96926. doi:10.1371/journal.pone.0096926 (2014).
146. Grineski, S. E., Collins, T. W., Chakraborty, J. & McDonald, Y . J. Environmental Health In-
justice: Exposure to Air Toxics and Children’s Respiratory Hospital Admissions in El Paso,
Texas.TheProfessionalGeographer 65. Publisher: Routledge eprint: https://doi.org/10.1080/00330124.2011.639625,
31–46. doi:10.1080/00330124.2011.639625 (2013).
147. Su, L. J., Young, S. G., Collins, J., Matich, E., Hsu, P.-C. & Chiang, T.-C. Geospatial As-
sessment of Pesticide Concentration in Ambient Air and Colorectal Cancer Incidence in
Arkansas, 2013–2017.InternationalJournalofEnvironmentalResearchandPublicHealth
19. Number: 6 Publisher: Multidisciplinary Digital Publishing Institute, 3258. doi:10.3390/
ijerph19063258 (2022).
148. Symanski, E., Tee Lewis, P. G., Chen, T.-Y ., Chan, W., Lai, D. & Ma, X. Air toxics and
early childhood acute lymphocytic leukemia in Texas, a population based case control study.
EnvironmentalHealth 15, 70. doi:10.1186/s12940-016-0154-8 (2016).
119
Appendices
A Housing Type Categories and Corresponding Categories used
in the 2019 American Community Survey (AHS) and by Los
Angeles County
Table 5.1: Housing Type Categories and Corresponding Categories used in the 2019 American
Community Survey (AHS) and by Los Angeles County
Housing Category AHS Housing Categories Los Angeles County Housing Categories
single-family home
single-family home, detached single-family residence
single-family home, attached double, duplex, or two units
three units (any combination)
four units (any combination)
apartments
2 apartments five or more units or apartments (any combination)
3-4 apartments
5-9 apartments
10-19 apartments
20-49 apartments
50 or more apartments
not analyzed
mobile home or trailer
boat, RV , van, etc.
120
B Categorization of Household Composition Types from the
2019 American Community Survey (AHS)
Table 5.2: Housing Type Categories and Corresponding Categories used in the 2019 American
Community Survey (AHS) and by Los Angeles County.
Composition Category Description
Older Adults
(aged 65+)
Adults
(aged 18-64)
Children
(aged 0-17)
One Older Adult Alone One Older Adult (65+) 1 0 0
Two Older Adults Two Older Adults (65+) 2 0 0
Older Adults with Adults
(with or without children)
One or More Older Adults (65+)
and One or More Adults (18-64)
1+ 1+ 0
One or More Older Adults (65+),
One or More Adults (18-64),
and One or More Children (0-17)
1+ 1+ 1+
All Other Older Adult Households
Three or More Older Adults (65+) 3+ 0 0
One or More Older Adults (65+)
and One or More Children (0-17)
1+ 0 1+
One Adult Alone One Adult (18-64) 0 1 0
Two Adults Two Adults (18-64) 0 2 0
All Other Households
Three or More Adults (18-64) 0 3+ 0
One or More Adults (18-64)
and One or More Children (0-17)
0 1+ 1+
121
C Lead-Based Paint (LBP) Models
Table 5.3: Odds of Lead-Based Paint (LBP) Presence
LBP 1 LBP 2 LBP 3 LBP 4
(Intercept) 21.38 *** 16.36 *** 12.00 * 21.38 ***
Parcel Building Construction Year
1900-1919 - - - -
1920-1929 0.38 ** 0.38 ** 0.69 0.38 ***
1930-1939 0.49 * 0.49 479,817.74 0.49 *
1940-1949 0.43 * 0.43 * 0.96 0.43 *
1950-1959 0.07 *** 0.08 *** 0.04 ** 0.02 ***
1960-1979 0.02 *** 0.02 *** 0.02 ** 0.02 ***
1980-2019 0.02 *** 0.02 *** 479,817.74 0.02 ***
Parcel Use Code
Single-Family Home 1.33 1.85
Apartments - -
Construction Year
Single-Family Home Interaction
1900-1919 single-family home -
1920-1929 single-family home 0.53
1930-1939 single-family home 0.00
1940-1949 single-family home 0.42
1950-1959 single-family home 2.47
1960-1979 single-family home 0.89
1980-2019 single-family home 0.00
1950-1959 Apartments 4.56 ***
Observations 1,946 1,946 1,946 1,946
Tjur’s R
2
0.144 0.146 0.163 0.156
AIC 1337.095 1337.742 1329.537 1328.111
*p<0.05 **p<0.01 ***p<0.0001
122
D High Concentration Lead-Based Paint (H-LBP) Models
Table 5.4: Odds of High Concentration Lead-Based Paint (H-LBP) Presence
H-LBP 1 H-LBP 2 H-LBP 3
(Intercept) 1.47 ** 1.11 1.17
Parcel Building Construction Year
1900-1919 - - -
1920-1929 0.57 *** 0.58 *** 0.41
1930-1939 0.49 *** 0.49 *** 0.95
1940-1949 0.24 *** 0.25 *** 0.21 *
1950-1959 0.04 *** 0.05 *** 0.07 **
1960-1979 0.00 0.00 0.00
1980-2019 0.00 0.00 0.00
Parcel Use Code
Single-Family Home 1.34 1.27
Apartments - -
Construction Year
Single-Family Home Interaction
1900-1919 single-family home -
1920-1929 single-family home 1.43
1930-1939 single-family home 0.49
1940-1949 single-family home 1.16
1950-1959 single-family home 0.54
1960-1979 single-family home 0.79
1980-2019 single-family home 0.79
Observations 1,946 1,946 1,946
Tjur’s R
2
0.094 0.095 0.097
AIC 2416.092 2415.997 2423.616
*p<0.05 **p<0.01 ***p<0.0001
123
E Very High Concentration Lead-Based Paint (VH-LBP) Models
Table 5.5: Odds of Very High-Concentration Lead-Based Paint (H-LBP) Presence
VH-LBP 1 VH-LBP 2 VH-LBP 3
(Intercept) 0.44 *** 0.35 ** 0.44
Parcel Building Construction Year
1900-1919 - - -
1920-1929 0.44 *** 0.58 *** 0.41
1930-1939 0.20 *** 0.49 *** 0.95
1940-1949 0.09 *** 0.25 *** 0.21 *
1950-1959 0.02 *** 0.05 *** 0.07 **
1960-1979 0.00 0.00 0.00
1980-2019 0.00 0.00 0.00
Parcel Use Code
Single-Family Home 1.28 0.99
Apartments - -
Construction Year
Single-Family Home Interaction
1900-1919 single-family home -
1920-1929 single-family home 1.03
1930-1939 single-family home 4,031,371.89
1940-1949 single-family home 1,762,220.97
1950-1959 single-family home 405,038.79
1960-1979 single-family home 1.01
1980-2019 single-family home 1.01
Observations 1,946 1,946 1,946
Tjur’s R
2
0.069 0.069 0.070
AIC 1430.188 1431.538 1438.565
*p<0.05 **p<0.01 ***p<0.0001
124
F Frequency of Parcel Building Construction Year 1960-2009
Figure 5.1: A histogram of parcel building construction year for only the subset of parcels with
buildings constructed between 1960 and 2009.
125
G Previous Work Linking NATA Exposure Data with Human
Health Outcomes
Data from the NATA assessments has previously been used to explore relationships between lead
emissions and individual or population health outcomes. Significantly, for this project, Benson
et al. (2017), Brink et al. (2016), and Brink et al. (2013) found relationships between NATA
lead emissions by tract and risk of elevated blood lead in children, even after controlling for other
ecological variables associated with lead exposure risk, such as proportion of older housing and
socioeconomic variables [141–143]. Although these studies did not explore the relationship be-
tween lead emissions and a specific health outcome they do establish that airborne lead emissions,
as reported in the NATA datasets, can be significantly linked to blood lead levels indicating that
airborne lead emissions may be a significant source of absorbed lead for exposed individuals. Pala-
cios et al. (2014) found a relationship between heavy metal emissions and Parkinson’s disease in
adults although this research was not specific to lead alone [144].
Other epidemiological work carried out relating NATA air pollutant exposure data to other
health outcomes, most commonly asthma [145, 146], cancers [147, 148], further support the ap-
plicability of NATA emissions data in studies exploring the relationship between environmental
pollutants and human health outcomes.
126
H Range of Log Lead Exposure for Lead Exposure Risk Quartile
Groups
Figure 5.2: Range of Log Lead Exposure for Lead Exposure Risk Quartile Groups
127
I Race and Ethnicity Variable Recode Table
Table 5.6: Race and Ethnicity Variable Recode Table
Original Race Variable Original Hispanic / Latino Indicator Variable Final Race/Ethnicity Variable
White Not Hispanic or Latino White Non-Hispanic/Latino
Black Not Hispanic or Latino Black Non-Hispanic/Latino
Other Not Hispanic or Latino Other Non-Hispanic/Latino
White OR Black OR Other Hispanic or Latino Hispanic or Latino
128
J Urbanicity Variable Recode Table
Table 5.7: Urbanicity Variable Recode Table
Original Urbanicity Variable Final Urbanicity Variable
Metropolitan with 1 Million Population or More Metropolitan with 1 Million Population or More
Metropolitan with 250,000 to Less Than 1 Million Population Metropolitan with 250,000 to Less Than 1 Million Population
Metropolitan Less Than 250,000 Population Metropolitan Less Than 250,000 Population
Urban Metro Area with 20,000 Population or More
Urban Area with 20,000 Population or More
Urban Non-Metro Area with 20,000 Population or More
Urban Metro Area with Less Than 20,000 Population
Urban Area with Less than 20,000 Population
Urban Non-Metro Area with Less Than 20,000 Population
Rural Metro Area
Rural Area
Rural Non-Metro Area
129
K Census Regions and Divisions of the United States
Figure 5.3: Census Regions and Divisions of the United States. Diagram retrieved from the United
States Census Bureau.
130
L Smoking Status Variable Recode Table
Table 5.8: Smoking Status Variable Recode Table
Original Current Smoker Indicator Original Ever Smoker Indicator Final Smoking Status Variable
Yes Yes Current Smoker
No Yes Former Smoker
No No Never Smoker
131
M Model Equations
132
N Chapter 3 Supplemental Models
133
Table 5.9: Chapter 3 Supplemental Models (Part I)
Base Models Single Individual Covariate Individual Covariate Groups
Base Age Only
Age and
Lead Quartile
Gender
Race
Ethnicity
Smoking
Status
Education
Poverty
Status
Gender
& Race
Ethnicity
Gender
Race/Ethnicity
& Smoking
Education
& Poverty
M0 M1 M2 A1 A2 A3 A4 A5 B1 B2 B3
N 48541 48541 48541 48541 48541 48541 48541 48541 48541 48541 48541
Total Cognitive Function (27-point scale) 14.44 *** 14.95 *** 15.13 *** 14.74 *** 15.89 *** 15.36 *** 16.97 *** 15.14 *** 15.50 *** 15.75 *** 16.97 ***
Age (10 year increments; centered at 64) -1.12 *** -1.20 *** -0.90 *** -1.28 *** -1.07 *** -1.54 *** -1.17 *** -0.98 *** -0.78 ** -0.72 **
Lead Exposure Q1 (lowest lead) - - - - - - - - -
Lead Exposure Q2 -0.07 -0.06 0.02 -0.06 -0.18 -0.07 0.03 0.04 -0.18
Lead Exposure Q3 -0.08 -0.06 0.28 + -0.06 -0.12 -0.07 0.30 * 0.31 * -0.11
Lead Exposure Q4 (highest lead) -0.56 *** -0.55 *** -0.01 -0.54 *** -0.39 ** -0.56 *** 0.01 0.03 -0.39 **
Lead Exposure Q1 x Age - - - - - - - - -
Lead Exposure Q2 x Age 0.26 0.25 0.27 0.25 0.2 0.24 0.26 0.25 0.19
Lead Exposure Q3 x Age -0.01 -0.02 -0.01 -0.01 -0.09 -0.02 -0.02 -0.02 -0.09
Lead Exposure Q4 x Age 0.06 0.05 0.03 0.05 0.06 0.06 0.02 0.01 0.06
Male - - -
Female 0.68 *** 0.69 *** 0.61 ***
Female x Age -0.51 ** -0.52 ** -0.57 **
White Non-Hispanic/Latino - - -
Black Non-Hispanic/Latino -2.90 *** -2.93 *** -2.93 ***
Other Race Non-Hispanic/Latino -1.55 *** -1.56 *** -1.60 ***
Hispanic -2.70 *** -2.71 *** -2.75 ***
Black Non-Hispanic/Latino x Age 0.19 0.22 0.21
Other Race Non-Hispanic/Latino x Age 0.27 0.25 0.24
Hispanic x Age 0.35 0.35 0.34
Never Smoker - -
Former Smoker -0.35 *** -0.32 ***
Current Smoker -0.38 ** -0.35 **
Former Smoker x Age -0.31 + -0.39 *
Current Smoker x Age 0.05 -0.02
No High School Degree -4.81 *** 4.80 ***
High School Degree or Equivalent -1.74 *** -1.74 ***
Some College or More - -
No High School Degree x Age -0.81 ** -0.76 **
High School Degree or Equivalent x Age -0.62 ** -0.61 **
Some College or More x Age - -
At or Above Poverty Threshold - -
Below Poverty Threshold -0.08 -0.08
Below Poverty Threshold x Age -0.20 * -0.18 *
*p<0.05 **p<0.01 ***p<0.0001
134
Table 5.10: Chapter 3 Supplemental Models (Part II)
Single Tract Covariate Tract Covariates with Tract Controls
% WNH % No HS
Median
HH Income
% Housing
Built < 1950
Urbanicity
Census
Division
Urbanicity
& Division
% WNH
Urbanicity
& Division
% No HS
Urbanicity
& Division
Med HHI
Urbanicity
& Division
% Old Housing
Urbanicity
& Division
C1 C2 C3 C4 C5 C6 D1 D2 D3 D4 D5
N 48541 48541 48541 48541 48541 48541 48541 48541 48541 48541 48541
Total Cognitive Function (27-point scale) 12.74 *** 16.60 *** 13.46 *** 15.45 *** 15.15 *** 15.07 *** 15.02 *** 12.87 *** 16.35 *** 13.17 *** 15.34 ***
Age (10 year increments; centered at 64) -1.08 *** -1.17 *** -1.53 *** -1.10 *** -1.01 *** -1.12 *** -0.85 ** -0.70 + -0.86 * -1.29 *** -0.80 *
Lead Exposure Q1 (lowest lead) - - - - - - - - - - -
Lead Exposure Q2 -0.01 -0.24 -0.28 + 0 -0.08 -0.1 -0.06 -0.06 -0.11 -0.16 -0.05
Lead Exposure Q3 0.37 * -0.14 -0.31 + 0.15 -0.08 -0.14 -0.1 0.25 -0.01 -0.15 0.06
Lead Exposure Q4 (highest lead) 0.18 -0.28 + -0.57 *** -0.28 + -0.55 ** -0.72 *** -0.67 *** -0.07 -0.27 -0.44 * -0.42 *
Lead Exposure Q1 x Age - - - - - - - - - - -
Lead Exposure Q2 x Age 0.25 0.24 0.19 0.26 0.19 0.23 0.1 0.09 0.1 0.05 0.09
Lead Exposure Q3 x Age -0.03 -0.02 -0.1 0.03 -0.13 -0.01 -0.19 -0.21 -0.19 -0.23 -0.17
Lead Exposure Q4 x Age 0.03 0.05 0.04 0.11 -0.06 0.12 -0.06 -0.1 -0.06 -0.01 -0.03
Metro; 1mill + - - - - - -
Metro; 250k to < 1mill -0.12 -0.02 -0.37 * 0.05 0.27 + -0.07
Metro; < 250k 0.28 0.24 -0.60 * 0.22 0.64 ** 0.26
Urban; 20k + -0.34 -0.11 -0.73 ** 0.16 0.29 -0.05
Urban; < 20k 0.13 0.19 -0.70 ** 0.39 + 0.59 * 0.32
Rural -0.5 -0.5 -1.17 -0.01 0.15 -0.32
New England 0.44 0.46 -0.24 0.14 0.04 1.06 *
Mid-Atlantic -0.05 -0.05 -0.24 -0.08 -0.16 0.65 **
East-North Central 0.3 0.25 -0.2 -0.01 0.19 0.69 ***
West-North Central 0.67 * 0.63 * -0.05 0.23 0.45 0.97 ***
South Atlantic - - - - - -
East-South Central -0.73 ** -0.76 ** -0.92 *** -0.60 ** -0.60 ** -0.66 **
West-South Central -0.57 ** -0.56 ** 0.02 0.05 -0.48 ** -0.53 **
Mountain 0.67 ** 0.67 ** 0.35 + 0.43 * 0.49 * 0.55 *
Pacific 0.51 ** 0.52 ** 0.62 *** 0.70 *** 0.17 0.68 ***
% White Non-Hispanic/Latino 0.35 *** 0.39 ***
% Less Than High School Degree -0.90 *** -0.90 ***
Median Household Income 0.33 *** 0.32 ***
% Housing Built < 1950 -0.16 *** -0.21 ***
*p<0.05 **p<0.01 ***p<0.0001
135
O Chapter 3 Models with Alternative Explanatory Variable
136
Table 5.11: Chapter 3 Supplemental Models (Part III)
Main Analysis
All Individual
Covariates
All Tract
Covariates
Full Model
M3 M4 M5
N 48541 48541 48541
Total Cognitive Function (27-point scale) 17.05 *** 13.82 *** 16.93 ***
Age (10 year increments; centered at 64) -0.41 -1.21 * -0.63
Lead Exposure Q1 (lowest lead) - - -
Lead Exposure Q2 -0.02 -0.09 -0.05
Lead Exposure Q3 0.23 + 0.19 0.19
Lead Exposure Q4 (highest lead) 0.04 -0.03 0.02
Lead Exposure Q1 x Age - - -
Lead Exposure Q2 x Age 0.19 0.03 0.01
Lead Exposure Q3 x Age -0.11 -0.27 -0.28
Lead Exposure Q4 x Age 0 -0.05 -0.05
Male - -
Female 0.76 *** 0.75 ***
Female x Age -0.49 ** -0.49 **
White Non-Hispanic/Latino - -
Black Non-Hispanic/Latino -2.46 *** -2.17 ***
Other Race Non-Hispanic/Latino -1.53 *** -1.45 ***
Hispanic -1.52 *** -1.31 ***
Black Non-Hispanic/Latino x Age 0.44 + 0.41
Other Race Non-Hispanic/Latino x Age 0.28 0.35
Hispanic x Age 0.70 * 0.77 *
Never Smoker - -
Former Smoker -0.14 -0.12
Current Smoker -0.16 -0.14
Former Smoker x Age -0.31 + -0.32 +
Current Smoker x Age 0.09 0.07
No High School Degree -4.24 *** -4.03 ***
High School Degree or Equivalent -1.63 *** -1.52 ***
Some College or More - -
No High School Degree x Age -1.01 *** -0.98 **
High School Degree or Equivalent x Age -0.62 ** -0.58 **
Some College or More x Age - -
At or Above Poverty Threshold - -
Below Poverty Threshold -0.07 -0.06
Below Poverty Threshold x Age -0.19 * -0.19 *
Metro; 1mill + - -
Metro; 250k to < 1mill -0.14 -0.08
Metro; < 250k -0.21 -0.19
Urban; 20k + -0.28 -0.21
Urban; < 20k -0.17 -0.13
Rural -0.51 -0.39
New England -0.14 -0.11
Mid-Atlantic -0.13 -0.04
East-North Central -0.1 -0.07
West-North Central 0.04 -0.02
South Atlantic - -
East-South Central -0.74 *** -0.61 **
West-South Central 0.1 0.1
Mountain 0.33 0.1
Pacific 0.61 *** 0.23
% White Non-Hispanic/Latino 0.25 *** 0.03
% Less Than High School Degree -0.39 *** -0.16 ***
Median Household Income 0.08 *** 0.03
% Housing Built < 1950 -0.03 -0.02
*p<0.05 **p<0.01 ***p<0.0001
137
Table 5.12: Chapter 3 Alternative M2 Models with Alternative Explanatory Variable
Model Comparison for M2
(Age and Lead Quartile Risk Only)
Tract
Category
Tract +
Neighbors
Category
Tract
Log Lead
Tract +
Neighbors
Log Lead
(mean)
M2 M2i M2ii M2iii
N 48541 48541 48541 48541
Total Cognitive Function (27-point scale) 15.13 *** 15.07 *** 13.25 *** 13.44 ***
Age (10 year increments; centered at 64) -1.20 *** -1.08 *** -0.85 -1.23
Lead Exposure Q1 (lowest lead) -
Lead Exposure Q2 -0.07
Lead Exposure Q3 -0.08
Lead Exposure Q4 (highest lead) -0.56 ***
Lead Exposure Q1 x Age -
Lead Exposure Q2 x Age 0.26
Lead Exposure Q3 x Age -0.01
Lead Exposure Q4 x Age 0.06
Low Tract Surrounded by Low Tracts -
Low Tract Surrounded by High Tracts 0.02
High Tract Surrounded by Low Tracts -0.4
High Tract Surrounded by High Tracts -0.51 ***
Low Tract Surrounded by Low Tracts x Age -
Low Tract Surrounded by High Tracts x Age -0.62
High Tract Surrounded by Low Tracts x Age -0.29
High Tract Surrounded by High Tracts x Age -0.03
Lead Exposure ln(mg/m3) -0.23 ***
Lead Exposure ln(mg/m3) x Age 0.04
Mean Tract and Neighbors Lead Exposure ln(mg/m3) -0.21 **
Mean Tract and Neighbors Lead Exposure ln(mg/m3) x Age -0.02
*p<0.05 **p<0.01 ***p<0.0001
138
Table 5.13: Chapter 3 Alternative M5 Models with Alternative Explanatory Variable
Model Comparison for M5
(Full Model)
Tract
Category
Tract +
Neighbors
Category
Tract
Log Lead
Tract +
Neighbors
Log Lead
(mean)
M5 M5i M5ii M5iii
N 48541 48541 48541 48541
Total Cognitive Function (27-point scale) 16.93 *** 16.99 *** 16.73 *** 16.82 ***
Age (10 year increments; centered at 64) -0.63 -0.76 -0.42 -0.85
Lead Exposure Q1 (lowest lead) -
Lead Exposure Q2 -0.05
Lead Exposure Q3 0.19
Lead Exposure Q4 (highest lead) 0.02
Low Tract Surrounded by Low Tracts -
Low Tract Surrounded by High Tracts 0.01
High Tract Surrounded by Low Tracts -0.04
High Tract Surrounded by High Tracts -0.04
Lead Exposure ln(mg/m3) -0.03
Mean Tract and Neighbors Lead Exposure ln(mg/m3) -0.02
Male - - - -
Female 0.75 *** 0.75 *** 0.75 *** 0.75 ***
White Non-Hispanic/Latino - - - -
Black Non-Hispanic/Latino -2.17 *** -2.17 *** -2.17 *** -2.17 ***
Other Race Non-Hispanic/Latino -1.45 *** -1.44 *** -1.45 *** -1.45 ***
Hispanic -1.31 *** -1.32 *** -1.32 *** -1.32 ***
Never Smoker - - - -
Former Smoker -0.12 -0.12 -0.12 -0.12
Current Smoker -0.14 -0.14 -0.14 -0.14
No High School Degree -4.03 *** -4.03 *** -4.03 *** -4.03 ***
High School Degree or Equivalent -1.52 *** -1.52 *** -1.52 *** -1.52 ***
Some College or More - - - -
At or Above Poverty Threshold - - - -
Below Poverty Threshold -0.06 -0.06 -0.06 -0.06
Metro; 1mill + - - - -
Metro; 250k to < 1mill -0.08 -0.09 -0.1 -0.09
Metro; < 250k -0.19 -0.2 -0.2 -0.2
Urban; 20k + -0.21 -0.25 -0.25 -0.24
Urban; < 20k -0.13 -0.17 -0.17 -0.17
Rural -0.39 -0.44 -0.5 -0.45
New England -0.11 -0.1 -0.11 -0.12
Mid-Atlantic -0.04 0 -0.01 -0.01
East-North Central -0.07 -0.06 -0.05 -0.06
West-North Central -0.02 -0.02 -0.04 -0.03
South Atlantic - - - -
East-South Central -0.61 ** -0.59 ** -0.60 ** -0.60 **
West-South Central 0.1 0.09 0.08 0.08
Mountain 0.1 0.13 0.13 0.12
Pacific 0.23 0.24 0.24 0.23
Percent of Total Population White Non-Hispanic/Latino (10% increments) 0.03 0.02 0.02 0.02
Percent of Adult Population with Less Than High School Degree (10% increments) -0.16 *** -0.17 *** -0.17 *** -0.17 ***
Median Household Income (2010-adjusted dollars; $10k increments) 0.03 0.03 0.04 0.04 +
Percent Occupied Housing Built in 1950 or Earlier (10% increments) -0.02 -0.01 -0.01 -0.01
*p<0.05 **p<0.01 ***p<0.0001
139

Abstract (if available)

Abstract Lead is a toxic heavy metal significant adverse health outcomes for those who are exposed. The overall goal of this dissertation is to evaluate and quantify the risk of exposure to sources of lead for older adults in the United States and assess potential health outcomes related to this exposure.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Biomarkers of age-related health changes: associations with health outcomes and disparities

PDF

Understanding the role of population and place in the dynamics of seasonal influenza outbreaks in the United States

PDF

Estimating at-risk population for lead service lines induced lead exposure and their correlation to socioeconomically disadvantaged neighborhoods in Milwaukee, Wisconsin

PDF

Personal PM2.5 exposure during pregnancy in an environmental health disparities population

PDF

Personal exposure to particulate matter PM2.5 sources during pregnancy and birthweight

PDF

Is stress exposure enough? Race/ethnic differences in the exposure and appraisal of chronic stressors among older adults

PDF

Optimizing small compounds to better understand tau fibril inhibition

PDF

Spatiotemporal studies of out-of-hospital cardiac arrests and bystander cardiopulmonary resuscitation in Los Angeles

PDF

Characterizing self-reported spatial and provider barriers to maternal health care utilization in Malawi

PDF

Alzheimer’s disease and related dementias research communication practices with minoritized racial and ethnic groups: a multiple case study

PDF

Analysis of ergodic and mixing dynamical systems

PDF

Children’s environmental and developmental equity within Los Angeles: a spatio-temporal exploration of varying conditions, perspectives, and outcomes

PDF

Urban air pollution and environmental justice: three essays

PDF

Investigating daily effects of activity space-based built environment exposures on physical activity behaviors in Hispanic women during pregnancy and early postpartum

PDF

Palliative care: breaking the knowledge barrier among community-based older adults

PDF

Tactile object localization: behavioral correlates, neural representations, and a deep learning hybrid model to classify touch

PDF

Effect of biomass fuel exposure on infant respiratory health outcomes in Bangladesh

PDF

Is healthcare for all? A qualitative questionnaire study of systemic racism in healthcare and the perception among stakeholders of U.S. healthcare corporations: an exploration of moral disengagement..

PDF

Smoke-free housing policies and secondhand smoke exposure in low income multiunit housing in Los Angeles County

PDF

Prenatal exposure to organophosphate esters and early neurodevelopment

Asset Metadata

Creator Vavra-Musser, Kate (author)

Core Title Lead exposure and health outcomes in older adults

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Health and Place,Population

Degree Conferral Date 2023-08

Publication Date 08/10/2023

Defense Date 06/19/2023

Publisher University of Southern California. Libraries (digital)

Tag aging,cognitive health,Environmental Health,environmental justice,lead exposure,neighborhood effects,OAI-PMH Harvest,population health

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ailshire, Jennifer (committee chair), Crimmins, Eileen (committee member), Franklin, Meredith (committee member), Wu, An-Min (committee member)

Creator Email vavramusser@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113298024

Unique identifier UC113298024

Identifier etd-VavraMusse-12241.pdf (filename)

Legacy Identifier etd-VavraMusse-12241

Document Type Dissertation

Rights Vavra-Musser, Kate

Internet Media Type application/pdf

Type texts

Source 20230810-usctheses-batch-1083 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA

Repository Email cisadmin@lib.usc.edu