Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Multilevel sociodemographic correlates of the health and healthcare utilization of childhood cancer survivors
(USC Thesis Other)
Multilevel sociodemographic correlates of the health and healthcare utilization of childhood cancer survivors
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Copyright 2020 Jessica L. Tobin
Multilevel Sociodemographic Correlates of the Health
and Healthcare Utilization of Childhood Cancer Survivors
by
Jessica L. Tobin
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
(PREVENTATIVE MEDICINE
(HEALTH BEHAVIOR RESEARCH))
August 2020
ii
Acknowledgments
This work would not have been possible without the broad support I have received over the past
six years. I am grateful for my committee, which comprises an immense pool of talent and
expertise I am fortunate to learn from. I thank Myles for his expertise in neighborhood data,
exposure assignment, and cancer epidemiology, Jennifer for her expertise and guidance on ethnic
disparities and acculturation, Brian for his multidisciplinary expertise in social epidemiology,
demography, and disparities research, and Ann for her depth of knowledge and insight on cancer
epidemiology and research methods. I am thankful to Joel for fostering an immensely supportive
research training experience from the start, which has allowed me to explore various interests
and develop not only my technical skills, but also an appreciation for the soul of this work.
Following his example, I have learned to maintain gratitude and respect for the real human lives
behind our data and aim to honor our participants through our research.
I am thankful every day for Lorenzo’s unconditional support. He has encouraged me to pursue
my professional goals, even when it required him to take on the bulk of parenting so that I could
attend conferences, courses, an internship, and complete this dissertation on time. I am also
grateful for my mother’s encouragement which has been a constant source of motivation, and for
Kim, Cynthia, Georgia, Afton, and Anuja, who have provided a safe sounding board to process
all of the challenges, joys, and opportunities in this process. Finally, I am thankful to Camilla, for
showing me new depths of love and for reminding me daily to live out my gratitude by being
fully present with those I cherish.
iii
Table of Contents
Acknowledgments ........................................................................................................................................................ ii
List of Tables ................................................................................................................................................................ iv
List of Figures .............................................................................................................................................................. vii
Abbreviations .............................................................................................................................................................. viii
Abstract ........................................................................................................................................................................ ix
Introduction .................................................................................................................................................................. 1
Overview of studies and implications ........................................................................................................................... 8
Chapter 1: Effects of neighborhood socioeconomic status and mobility on childhood cancer survivors’ (CCS) follow
up care and health as young adults ............................................................................................................................ 10
Overview and Specific Aims .............................................................................................................................. 10
Introduction ...................................................................................................................................................... 12
Methods ............................................................................................................................................................ 16
Results ............................................................................................................................................................... 25
Discussion .......................................................................................................................................................... 60
Limitations and strengths .................................................................................................................................. 66
Conclusion ......................................................................................................................................................... 68
Chapter 2: Multilevel demographic and cultural factors associated with mental health, physical health, and
healthcare utilization among childhood cancer survivors .......................................................................................... 69
Overview and Specific Aims .............................................................................................................................. 69
Introduction ...................................................................................................................................................... 71
Methods ............................................................................................................................................................ 75
Results ............................................................................................................................................................... 81
Discussion ........................................................................................................................................................ 103
Limitations and strengths ................................................................................................................................ 108
Conclusion ....................................................................................................................................................... 110
Conclusion ................................................................................................................................................................. 111
References ................................................................................................................................................................ 113
iv
List of Tables
Table 1. Bivariate associations between nSES at diagnosis and each outcome (separate models)
Table 2. Bivariate associations between nSES at survey and each outcome (separate models)
Table 3. Descriptive statistics, n=877
Table 4. Mobility table: dichotomized neighborhood SES at diagnosis by neighborhood SES at
survey. Total n=877
Table 5. Bivariate association between mobility and race/ethnicity
Table 6. Bivariate associations between mobility and each outcome (separate models)
Table 7. Bivariate associations between outcomes and covariates (each cell a separate model)
Table 8. Multivariable model of follow up care, with nSES at diagnosis
Table 9. Multivariable model of follow up care, with nSES at diagnosis and survey
Table 10. Multivariable model of follow up care, with mobility indicator
Table 11. Multivariable model of depressive symptoms, with nSES at diagnosis
Table 12. Multivariable model of depressive symptoms, with nSES at diagnosis and survey
Table 13. Multivariable model of depressive symptoms, with mobility indicator
Table 14. Multivariable model of wellbeing, with nSES at diagnosis
Table 15. Multivariable model of wellbeing, with nSES at diagnosis and survey
Table 16. Multivariable model of wellbeing, with mobility indicator
Table 17. Multivariable model of late effects, with nSES at diagnosis
Table 18. Multivariable model of late effects, with nSES at diagnosis and survey
Table 19. Multivariable model of late effects, with mobility indicator
Table 20. Diagonal reference model of late effects (any versus none)
Table 21. Multivariable model of self-rated health, with nSES at diagnosis
Table 22. Multivariable model of self-rated health, with nSES at diagnosis and survey
Table 23. Multivariable model of self-rated health, with mobility indicator
Table 24. Diagonal reference model of self-rated health (low (Poor/Fair/Good) versus high (Very
good-Excellent)
Table 25. Weights for diagonal reference model of self-rated health, with response level 3 coded
as low
Table 26. Weights for diagonal reference model of self-rated health, with response level 3 coded
as high
Table 27. Proportion of the sample providing at least a minimum threshold of residential history
(relative to time since diagnosis)
Table 28. Educational attainment by race/ethnicity
Table 29. Bivariate associations of all outcomes and number of addresses per year since
diagnosis
Table 30. Multivariable model of follow up care and number of addresses per year since diagnosis
Table 31. Multivariable model of depressive and number of addresses per year since diagnosis
Table 32. Multivariable model of wellbeing and number of addresses per year since diagnosis
Table 33. Multivariable model of late effects and number of addresses per year since diagnosis
Table 34. Multivariable model of self-rated health and number of addresses per year since
diagnosis
Table 35. Bivariate associations of all outcomes and average nSES
Table 36. Multivariable model of follow up care and average nSES
Table 37. Multivariable model of depressive symptoms and average nSES
Table 38. Multivariable model of wellbeing and average nSES
33
33
34
36
38
38
39
40
40
41
42
42
43
43
44
44
45
46
46
47
48
48
49
49
50
50
53
53
53
54
54
55
56
56
57
58
59
59
v
Table 39. Multivariable model of late effects and average nSES
Table 40. Multivariable model of self-rated health and average nSES
Table 41. Distribution of addresses per year, by time since diagnosis
Table 42. Comparison of the distribution of time-weighted average neighborhood SES with
missing years imputed by alternate methods
Table 43. Comparison of coefficients for time-weighted average nSES derived from alternative
imputation methods
Table 44. Comparison of coefficients for time-weighted average nSES with varied exclusion
criteria
Table 45. Sample descriptive statistics, n=989 overall
Table 46. Unadjusted associations with cancer-related follow-up care
Table 47. Multivariable model of follow-up care, n=931
Table 48. Multivariable model of follow-up care, including interaction between ethnicity and
enclave. Results not in OR form to provide estimates for interaction terms, n=931
Table 49. Multivariable model of follow-up care, assessing acculturation and enclave main
effects (restricted to Hispanics), n=481
Table 50. Multivariable model of follow-up care, assessing acculturation*enclave interaction
(restricted to Hispanics) Results not in OR form to provide estimates for interaction terms, n=481
Table 51. Stratified models of follow up care: Among high ethnic enclaves (top 2 quintiles)
n=372
Table 52. Stratified models of follow up care: Among low ethnic enclaves (bottom 3 quintiles)
n=111
Table 53. Unadjusted associations with depressive symptoms (CESD sum)
Table 54. Multivariable model of depressive symptoms, n=876
Table 55. Multivariable model of depressive symptoms, including interaction between ethnicity
and enclave, n=876
Table 56. Multivariable model of depressive symptoms, assessing acculturation and enclave
main effects (restricted to Hispanics), n=460
Table 57. Multivariable model of depressive symptoms, assessing acculturation*enclave
interaction (restricted to Hispanics), n=460
Table 58. Multivariable model of depressive symptoms among Hispanics, assessing acculturation
effect among low ethnic enclaves, n=104
Table 59. Multivariable model of depressive symptoms among Hispanics, assessing acculturation
effect among high ethnic enclaves, n=356
Table 60. Unadjusted associations with wellbeing (Mental Health Continuum total score)
Table 61. Multivariable model of wellbeing, n=896
Table 62. Multivariable model of wellbeing, with Hispanic enclave interaction, n=896
Table 63. Multivariable model of wellbeing, assessing acculturation and enclave main effects
(restricted to Hispanics), n=475
Table 64. Multivariable model of wellbeing, assessing acculturation*enclave interaction
(restricted to Hispanics), n=475
Table 65. Stratified models of wellbeing: Among high ethnic enclave (top 2 quintiles), n=369
Table 66. Stratified models of wellbeing: Among low ethnic enclave (bottom 3 quintiles) n=106
Table 67. Unadjusted associations with total late effects
Table 68. Multivariable negative binomial poisson regression model of late effects, n=940
60
61
62
62
63
64
85
88
88
89
89
90
90
90
91
92
92
92
92
93
93
95
95
95
96
96
96
96
98
98
vi
Table 69. Multivariable negative binomial poisson regression model of late effects, interaction
between Hispanic and ethnic enclave n=940
Table 70. Multivariable model of late effects, assessing acculturation and enclave main effects
(restricted to Hispanics), n=497
Table 71. Multivariable model of late effects, assessing acculturation*enclave interaction
(restricted to Hispanics) Results not in OR form to provide estimates for interaction terms, n=497
Table 72. Stratified models of late effects: Among high ethnic enclave (top 2 quintiles) n=386
Table 73. Stratified models of late effects: Among low ethnic enclave (bottom 3 quintiles) n=111
Table 74. Unadjusted associations with self-rated health
Table 75. Multivariable model of self-rated health, n=931
Table 76. Multivariable model of self-rated health, including interaction between ethnicity and
enclave, n=931
Table 77. Multivariable model of self-rated health, assessing acculturation and enclave main
effects (restricted to Hispanics), n=488
Table 78. Multivariable model of self-rated health, assessing acculturation*enclave interaction
(restricted to Hispanics), n=488
Table 79. Stratified models of self-rated health: Among high ethnic enclave (top 2 quintiles)
n=379
Table 80. Stratified models of self-rated health: Among low ethnic enclave (bottom 3 quintiles)
n=109
99
99
99
99
100
101
101
101
102
102
102
103
vii
List of Figures
Figure 1. Conceptual model of potential associations between multi-level factors and health
outcomes addressed in this dissertation
Figure 2. Study 1 conceptual model
Figure 3. Mobility estimation schema
Figure 4. Distribution of neighborhood SES at diagnosis and at survey
Figure 5. Mobility distribution
Figure 6. Distribution of nSES at diagnosis, by race/ethnicity
Figure 7. Distribution of nSES at survey, by race/ethnicity
Figure 8. Distribution of number of addresses provided per year, all versus within California
Figure 9. Distribution of proportion of time since diagnosis missing residential data
Figure 10. Comparison of the distribution of time-weighted average neighborhood SES derived
from three alternate methods of missing data imputation
Figure 10. Study 2 conceptual model
Figure 11. Interaction between ethnic enclave and Hispanic orientation on depressive symptoms
Figure 12. Interaction between ethnic enclave and Hispanic orientation on wellbeing
16
18
29
33
37
37
38
52
52
63
74
94
97
viii
Abbreviations
CCS- Childhood cancer survivors
SES- Socioeconomic status
nSES- Neighborhood socioeconomic status
ALL- Acute lymphoblastic leukemia
MAUP- Modifiable areal unit problem
ix
Abstract
Disparities persist in childhood cancer outcomes. Social determinants of health at the
individual and contextual level such as socioeconomic status (SES) and acculturation may be
important correlates of health in this population, though they remain under-addressed in most
prior childhood cancer research. This dissertation assessed associations of neighborhood SES,
residential and social mobility, race/ethnicity, and individual acculturation and neighborhood
ethnic enclave with measures of psychosocial and physical as well as healthcare utilization.
Chapter one identified lower rates of follow up care among the most residentially unstable long-
term survivors, as well as highest self-rated health among those who resided in the highest SES
neighborhoods from diagnosis into young adulthood. Chapter two identified protective
associations among Hispanics between Hispanic orientation and follow up care, psychosocial
health, and late effects of treatment, as well as a protective association between Anglo
orientation and wellbeing, self-rated health. Anglo orientation was also associated with fewer
late effects among Hispanics residing in high ethnic enclaves. Findings provide novel insights
into unique factors that may promote or inhibit health outcomes among young adult survivors of
pediatric cancer, which may have implications for targeted support services aimed at retaining
survivors in care and promoting their long-term health.
1
Introduction
Childhood cancer survivorship
Improvements in treatment of childhood cancers in recent decades has resulted in overall 5-
year survival rates of approximately 80%. (Ward, DeSantis, Robbins, Kohler, & Jemal, 2014)
Disparities in longer-term survival are primarily explained by treatment exposure, as
chemotherapy and radiation dose are associated with an increased risk of late mortality among
CCS. (Tukenova et al., 2010) For example, over the period between 1970 and 1999, reduced
rates of cranial, abdominal, and chest radiotherapy for acute lymphoblastic leukemia, Wilms’
tumor, and Hodgkin’s lymphoma, respectively, contributed to improvements in overall
childhood cancer mortality. (Armstrong et al., 2016) Despite these reductions in toxic treatment
regimens, late mortality remains a concern. While deaths due to progression of primary disease
have decreased, deaths attributable to subsequent cancers has shown increases among childhood
cancer survivors (CCS). (Armstrong et al., 2009; Reulen et al., 2010)
In addition to late mortality, heightened morbidity is also common among CCS. Late effects,
or cancer treatment-related sequelae, may emerge months, years, or even decades after treatment
completion. (Kremer et al., 2013) In a large cohort study of over 10,000 CCS, the cumulative
incidence of a chronic health condition was over 73% by 30 years post-diagnosis, and was over
42% for a severe or life threatening condition. (Kevin C. Oeffinger et al., 2006) This may include
secondary cancers, organ dysfunction, metabolic disorders, cardiovascular disease, cognitive
issues, or a range of other conditions. (M. M. Hudson et al., 2003; Kremer et al., 2013; Nathan et
al., 2009) Compared to non-cancer controls, CCS are more likely to report adverse health in
general, activity limitations, and functional impairment. (M. M. Hudson et al., 2003)
2
Psychosocial issues are also commonly reported. In one study, CCS were 80% more likely to
report adverse mental health compared to sibling controls. (M. M. Hudson et al., 2003) A large
study in the UK found that CCS reported a significantly higher prevalence of mental health
dysfunction compared to the general population, and this excess dysfunction increased with age.
(Fidler et al., 2015) These issues may present as post-traumatic stress symptoms, depression, or
anxiety. Predictors of psychological distress among CCS may include age at diagnosis, cancer
type, and sex. (Jacobs & Pucci, 2013) Mental and emotional problems may then inhibit
functioning, ability to reach educational and occupational goals, likelihood of marriage, and
overall quality of life, (Gurney et al., 2009) so preventive efforts to manage psychological issues
are critical to support long-term wellbeing among CCS.
Follow up care
Behavioral strategies may attenuate the impact of late effects. Paramount among these
strategies is the maintenance of follow up care. It is recommended that all CCS receive regular
medical care based on the specific risks associated with their previous cancer and the treatment
they received, in addition to other predispositions or health conditions. (Landier et al., 2004) This
recommendation arose out of the recognition that CCS are a vulnerable group with unique
healthcare needs that were not adequately addressed by standard healthcare practices. (Council,
2003) However, previous studies have shown that between just 31% and 75% of CCS report
having received a recent cancer-related medical visit. (Milam et al., 2015; Nathan et al., 2009;
Nathan et al., 2008; Kevin C Oeffinger et al., 2004) Moreover, as the cumulative incidence of
late effects increases with time since diagnosis (and age), the likelihood of CCS maintaining
contact with the healthcare system for active surveillance decreases. (Nathan et al., 2009) Such a
discrepancy between recommended and actual frequency of care means that healthcare providers
3
may miss the opportunity for early detection of health problems, including potential secondary
malignancies, and timely intervention. (Kremer et al., 2013; Landier et al., 2004)
Disparities in childhood cancer survivorship
Racial/ethnic disparities in health status and survival have been explored in several studies. A
recent survival analysis of childhood acute lymphoblastic leukemia (ALL), acute myelogenous
leukemia, and Hodgkin lymphoma found that despite improved overall survival of these cancers
in recent decades, Hispanics suffered inferior survival of ALL compared to Whites, (Kahn et al.,
2016) a finding consistent with earlier work. (Kadan-Lottick, Ness, Bhatia, & Gurney, 2003)
Follow up care has also been found to vary by ethnicity, with multiple studies finding that
Hispanic CCS are less likely than Whites to remain engaged in care. (Milam et al., 2015; Rokitka
et al., 2017)
Cultural factors may potentially contribute to racial/ethnic disparities in health, although they
have not been widely studied in the context of childhood cancer survivorship. Acculturation,
referring to the multi-dimensional adaptation stemming from contact with a new culture, may be
associated with health status and healthcare use by various mediating pathways. Such potential
mechanisms linking acculturation and health include health literacy (the ability to understand
medical information), which itself may be impacted by culturally-derived beliefs about disease,
or linguistic barriers inhibiting effective patient-provider communication.(DuBard & Gizlice,
2008; S. J. Shaw, Huebner, Armin, Orzech, & Vivian, 2009) Heritage cultural values may be
protective of health, resulting in poorer health with increase acculturation to the US, whereas
decreased discrimination associated with increased acculturation may lead to improved health
outcomes.(Flores et al., 2008; Johnson, Carroll, Fulda, Cardarelli, & Cardarelli, 2010)
4
Associations between acculturation and health outcomes among CCS have yet to be thoroughly
explored and may lend novel insights into disparities in survivorship.
Socioeconomic disparities in health outcomes also persist. However, in population-based
studies of CCS, individual-level SES is typically unavailable (as this is not captured in cancer
registries). However, one study of adverse health status among CCS across multiple domains
including mental health, functional status, activity limitations, cancer-related pain, cancer-related
anxiety, and general health, found that both lower educational attainment and annual income
under $20,000 were associated with increased odds of reporting at least one adverse health status
domain. (M. M. Hudson et al., 2003) Another found that CCS with an annual household income
under $40,000 were more likely to report having had no medical visits in the past two years,
compared to those with higher incomes, even after adjusting for insurance status. (Nathan et al.,
2008) These findings suggest that individual income may be an important contributor to access
to care as well as health status.
Despite the evidence of individual-level drivers of disparities in cancer survivorship,
unexplained differences remain, suggesting the need to explore contextual factors that may
influence health outcomes such as local environmental or ecological neighborhood
characteristics.
Contribution of contextual factors to individual health disparities
The contribution of contextual factors to health disparities is an increasing focus of
epidemiological research. Two contextual factors relevant to long-term survivorship among CCS
include SES and acculturation.
5
Neighborhood socioeconomic status
Disparities in morbidity and mortality by neighborhood socioeconomic status (nSES)
have been consistently observed, with nSES positively associated with improved health
outcomes.(Pickett & Pearl, 2001) Numerous measures of nSES exist, reflecting mainly three
domains- income, education, and occupational status- which represent differential distributions
of resources and conditions across neighborhoods.(Diez Roux & Mair, 2010) For example, lower
SES neighborhoods typically lack health-promoting resources such as quality housing, protection
from harmful environmental exposures, healthy food options, and safe recreational spaces. (Diez
Roux & Mair, 2010) These factors then impact the likelihood of engaging in health-promoting
behaviors such as healthy eating and physical activity, contributing to downstream physical
health consequences. (Gordon-Larsen, Nelson, Page, & Popkin, 2006) For example, residents in
communities of lower SES are also more likely to experience chronic stress from issues such as
crime and violence. (Steptoe & Feldman, 2001) Chronic stress, particularly when accumulated
over years of exposure, then has significant consequences for downstream mental health and
wellbeing.
Neighborhood acculturation
Acculturation at the neighborhood level may be approximated by the proportion of the
community that is an ethnic minority, foreign-born, and proficient in the English language. These
and other similar variables have previously been combined into a single index reflecting
ethnicity-specific acculturation. (Keegan et al., 2010) Less acculturated neighborhoods are likely
to have widely held cultural norms and practices that are culturally or ethnically distinct from
surrounding areas. (S. L. Gomez et al., 2011) While neighborhood acculturation has been less
examined than nSES, there are several potential mechanisms by which it may impact health.
6
Less acculturated neighborhoods, or ‘ethnic enclaves,’ may be characterized by low English
language proficiency and possibly also cultural beliefs and norms related to disease and/or the
healthcare system, which may influence access to and utilization of care in both positive and
negative ways. (S. L. Gomez et al., 2015) For example, social cohesion may be greater in ethnic
enclaves, which may provide access to health-promoting resources generated by relationships
between people in well-connected communities. (Morenoff & Lynch, 2002) It is important to
note however that the effects of neighborhood acculturation are likely to differ by individual
ethnicity in determining individual health outcomes. Individuals may be more susceptible to the
health-promoting or health-inhibiting aspects of a community if his/her ethnicity and/or level of
acculturation is consistent with the ethnic or cultural majority. (Gaskin, Dinwiddie, Chan, &
McCleary, 2012; Haas et al., 2004)
Additional moderating factors
Levels of influence on individual health may be described as micro- individual-level
factors such as demographic characteristics, macro- social structural factors such as healthcare
systems and policies, and meso- the level between the two extremes that refers to the proximate
settings of an individual, such as neighborhoods and schools. (Short & Mollborn, 2015) The
contribution of meso level neighborhood factors to individual health may lie not only in direct
effects, but also in their capacity to interact with other exposures from all levels of influence.
(Morenoff & Lynch, 2002) Neighborhood conditions may pose a health risk only for certain
subsets of individuals. For example, the association between availability of nearby healthcare
facilities (meso level) and likelihood of receiving routine medical care may depend in part on a
cancer survivor’s perceived risk (micro level) and therefore his/her motivation to obtain care.
7
Thus, it is important to consider factors that may modify the effects of neighborhood
characteristics on health.
Time may also affect the association between neighborhoods and health. A life course
perspective considers the time and timing of links between exposures and outcomes, referring to
both the duration of exposure as well as the developmental period of life in which an exposure is
experienced. (Kuh, Ben-Shlomo, Lynch, Hallqvist, & Power, 2003; Kuh & Shlomo, 2004) This
approach considers both early life exposures and adult risk factors as complementary in
explaining adult health outcomes. From this perspective, neighborhood characteristics during
childhood and adolescence may have indirect, long-term effects on health, may co-evolve with
biological factors over the life course, and may impart cumulative effects over time. (Morenoff
& Lynch, 2002) In this way, the life course approach to researching disparities in adult health
expands the focus from individual-level risk factors acting instantaneously and supposedly
independently, to interwoven, multi-level exposures.(Morenoff & Lynch, 2002) That is to say,
assessing only the impact of adult risk factors contemporaneous with the outcomes under study,
or only considering early life exposures in association with adult health, is inadequate to uncover
pathways to health and disease as it would not capture long-term, latent, or cumulative effects of
exposures throughout the lifespan.
8
Overview of studies and implications
The life course approach highlights the importance of considering exposures throughout the
lifespan when assessing predictors of adult health and disease. This approach may be enhanced
by accounting for broader influences at not only the micro (individual) level, but also the meso
(proximate settings), and macro (societal, institutional) levels. This dissertation utilized data
from a range of sources to assess the impact of both micro- and meso-level sociodemographic
and cultural factors corresponding to various points across the life course, on health status and
healthcare utilization among CCS, to address two important gaps in the CCS literature: cultural
differences and the inclusion of both childhood (early) and adult (contemporaneous) risk factors
for health outcomes.
Study 1 focused on the physical and mental health status and healthcare utilization of a
population-based sample of CCS diagnosed in Los Angeles County who participated in a survey
study. Neighborhood SES is the primary predictor of interest and has been be explored in several
ways- statically at diagnosis, statically in adulthood (proximal to outcomes), to characterize
mobility (i.e. stability of neighborhood SES context from diagnosis to adulthood), and
cumulatively from diagnosis to adulthood. This approach allowed for a thorough examination of
the impact of timing of exposure to neighborhood deprivation on adult health outcomes, a central
focus of the life course approach and an understudied aspect of social epidemiological research
on the health of adult survivors of pediatric cancers. Findings improve our understanding of the
vulnerability of CCS to social risk factors and the potentially dynamic nature of the influence of
these factors over time, net of physiological vulnerability attributed to cancer treatment intensity.
This may inform the timing of interventions aimed at reducing socioeconomic disparities among
CCS.
9
Study 2 assessed the association of individual and neighborhood acculturation with physical
and mental health status and healthcare utilization among the same sample of CCS. This study
focused on cross-level interactions between micro-factors (individual ethnicity and acculturation)
and the meso-level neighborhood acculturation variable, a novel approach given the paucity of
research on cultural correlates of health among CCS. Findings contribute to the identification of
factors that are uniquely influential to the health of Hispanic CCS, a more vulnerable population
that is likely to represent an increasing proportion of childhood cancer cases given US population
demographic trends.
Figure 1. Conceptual model of potential associations between multi-level factors and health
outcomes addressed in this dissertation
Macro-level factors
Proximate settings-
neighborhood
characteristics
Micro-level factors
Demographic,
clinical,
socioeconomic,
cultural
Follow up care
Psychosocial health
Physical health
Length of exposure,
change in context
10
Chapter 1: Effects of neighborhood socioeconomic status and
mobility on childhood cancer survivors’ (CCS) follow up care and
health as young adults
Overview and Specific Aims
Obtaining recommended follow up care and managing physical symptoms is crucial for
survivors of childhood cancer. However greater residential mobility and poverty exposure, and
associated instability and material deprivation, may inhibit health-promoting behaviors among
survivors. Additionally, chronic deprivation contributes to chronic stress over the life course,
contributing to poor mental health and wellbeing in young adulthood. While many studies have
explored the effects of childhood contextual exposures on adult health, the impact of mobility
and cumulative exposure is less understood. This may have significant consequences for high
risk, vulnerable childhood cancer survivors (CCS). Study 1 describes the residential mobility and
cumulative exposure to neighborhood disadvantage (low neighborhood SES (nSES)) among a
sample of young adult CCS. Potential ethnic disparities in mobility and cumulative poverty
exposure will be explored, as well as the association between these exposures and indicators of
physical, psychological, and behavioral health. A conceptual model of these associations is
presented in Figure 3.
Aim 1. Explore whether mobility (both the number of moves and type of mobility) and
cumulative poverty differ by ethnicity (Hispanic, Non-Hispanic White, other)
Hypothesis 1. Non-Whites will report greater residential mobility and greater cumulative poverty
11
Aim 2. Test whether nSES at diagnosis is associated with current health and health behaviors
(late effects, self-rated health, cancer-related follow-up care, depressive symptoms, wellbeing).
Hypothesis 2. nSES at diagnosis will be positively associated with self-rated health, health care
use, and wellbeing and negatively associated with late effects, and depressive symptoms
Aim 3. Test whether cumulative poverty is associated with adult health outcomes
Hypothesis 3. Cumulative poverty will be negatively associated with self-rated health, health
care use, and wellbeing, and positively associated with late effects and depressive symptoms
Figure 2. Study 1 conceptual model
Macro-level factors
Proximate settings-
neighborhood
SES
Micro-level factors
Demographic,
clinical,
socioeconomic
Follow up care
Psychosocial health
Physical health
Length of exposure,
change in context
12
Introduction
Treatment advances in recent decades have improved 5-year survival rates of childhood
cancers considerably. While early mortality is a risk for some CCS (particularly those who
received more toxic treatment regimens), other adverse health effects impact a greater majority
of CCS. Up to 75% of CCS may experience a late effect resulting from their cancer.(Kremer et
al., 2013) CCS are also more likely than non-cancer controls to report psychological distress,
which has been reported across the cancer continuum among young adult cancer survivors.(Brad
J Zebrack et al., 2004; Brad J. Zebrack et al., 2002) These issues can significantly affect quality
of life and in some cases may inhibit achievement of key developmental milestones.(Langeveld,
Stam, Grootenhuis, & Last, 2002; Stam, Grootenhuis, & Last, 2005) Cancer-related follow up
care may help identify and manage these issues among young adult CCS, but many do not
receive the recommended frequency of care.(Milam et al., 2015; Nathan et al., 2008; Kevin C
Oeffinger et al., 2004) Thus, understanding correlates of health and potential sources of
disparities in adult survivors of childhood cancer is of continual interest.
Childhood SES and Adult Health in CCS
Early life exposures impact health and wellbeing across the life course.(Braveman &
Barclay, 2009; Hertzman & Power, 2003) Among childhood cancer cases specifically, recent
systematic reviews and meta-analyses indicate that SES at diagnosis is associated with survival
and that disparities between low and high SES children have widened over time in the US,
highlighting the need for continued investigation and deeper understanding of the underlying
causes of differential survival gains.(Gupta, Wilejto, Pole, Guttmann, & Sung, 2014; Petridou et
13
al., 2014) The effect of childhood SES on health disparities in adulthood may be explained by
one or more of several potential mechanisms.
Physical, psychological, and behavioral health may be impacted by neighborhood
characteristics. Health behaviors are also shaped by the physical environment, such as the
availability of healthy food and green spaces or recreational facilities that promote healthy eating
and physical activity.(Diez Roux & Mair, 2010; Gordon-Larsen et al., 2006) Various features of
the physical environment may also contribute to disparities in healthcare access by nSES,
including the unequal distribution of resources.(Diez Roux & Mair, 2010) The social
environment may affect health behaviors via group norms and behavior modeling.(Diez Roux &
Mair, 2010) Chronic stress and its physiological sequelae is another potential pathway by which
nSES impacts health. Exposure to violence, material deprivation, low social capital, or other
aspects of the physical and social environment can result in chronic stress, exacerbating
physiological vulnerability and leading to poor health.(Matheson et al., 2006; Steptoe &
Feldman, 2001) Research indicates that variation in adult health status is associated with social
class in childhood even after controlling for adult circumstances, suggesting that an effect
remains independent of the indirect effect via the association between childhood SES and adult
SES.(Conroy, Sandel, & Zuckerman, 2010) Moreover, the effects of residence in a low SES/high
deprivation area may be more salient when exposure occurs in childhood given the vulnerability
of that physical and psychological developmental stage.(Ben-Shlomo & Kuh, 2002; Hertzman &
Boyce, 2010)
14
Limitations of Static nSES mMeasurements, and Alternatives
While the overall association between SES and health seems relatively clear, most studies
utilize static measures of SES, which do not account for important moderating or cumulative
effects over time. Consideration of dynamic contextual changes and various exposure-outcome
relationship patterns may not only reduce bias in statistical estimates of effect,(Do, Wang, &
Elliott, 2013) but also further clarify the nature of the impact of SES on cancer survivors’ health
across the life span.
Mobility may moderate the effects of single-point-in-time measures of neighborhood
characteristics in childhood because it may reflect change in an individual’s context. Without
accounting for the dynamic nature of residence and context we may overlook important variation
by classifying individuals based on a neighborhood context from a limited period of childhood,
while their lifetime exposure to poverty and deprivation may actually be quite disparate
depending on residential and social mobility, which would result in inaccurate and/or biased
findings. Residential instability (number of residential moves) has been associated with medical
coverage and having a stable medical ‘home,’ both of which are particularly crucial for CCS.
(Busacker & Kasehagen, 2012) Previous research on economic hardship in childhood that
accounted for changes over time found that health outcomes differed between those that began
life and remained in, or moved into poverty, and those that had more advantaged early lives or
that moved out of poverty by adulthood.(Shuey & Willson, 2014) Another study found that
healthcare utilization (mammography screening) differed between the upwardly and downwardly
mobile, such that individuals tended to adapt to the screening practice norms of the social
position they ended up in rather than those of their social position of origin,(Missinne,
15
Daenekindt, & Bracke, 2015) highlighting the importance of accounting for variation in SES
over time.
In addition to mobility as a potential moderator of the effect of childhood SES, various
potential trajectories by which childhood exposure may influence health in adulthood have been
described, including latency, cumulative, and pathway. Latency refers to exposure at one point in
time with effects occurring years later (‘lagged’ effects), cumulative refers to the combined effect
of multiple exposures over time, and pathway refers to dependent sequences of exposures, where
exposure during one phase of life influences the likelihood of other exposures later in life (e.g.
lacking the necessary competencies to begin school in childhood, leading to school failure,
leading to unemployment). (Hertzman & Power, 2003) The life course approach to studying
health supports this notion, and emphasizes that early life exposures can have direct and indirect
long-term effects, and that effects of exposures can accumulate over time. (Morenoff & Lynch,
2002) Measures of nSES that account for the total amount of time spent in disadvantaged
neighborhoods throughout childhood and young adulthood provide the opportunity to explore
such potential pathways between SES and adult health. One study found that those who had
experienced chronic poverty only exhibited significantly worse health outcomes compared to the
never-poor,(Béatrice, Lise, Victoria, & Louise, 2012) while others have found a gradient effect
where increasing exposure to poverty or disadvantage was associated with increasing risk of
poor health.(Evans & Kim, 2007; D. L. Hudson, Puterman, Bibbins-Domingo, Matthews, &
Adler, 2013; McDonough, Sacker, & Wiggins, 2005; Stansfeld, Clark, Rodgers, Caldwell, &
Power, 2011; Walsemann, Geronimus, & Gee, 2008) Thus, cumulative pathways should to be
considered, particularly among CCS, for whom previous work has focused primarily on static
measures of SES.
16
The present study assessed the effects of residential mobility and exposure to
neighborhood disadvantage on indicators of physical, psychological, and behavioral health in
young adulthood among survivors of childhood cancer. Main effects as well as potential
interactions by age at diagnosis and time since diagnosis were explored.
Methods
Participants
Participants were recruited to Project Forward, a study of follow up care and health
among CCS. Eligible participants were recruited from the Los Angeles Cancer Surveillance
Program, the Surveillance, Epidemiology, and End Results (SEER) cancer registry for Los
Angeles County, and included CCS diagnosed at age 19 or younger who were between the ages
of 18 and 39 at the time of the study in 2015. Participants who received cancer treatment less
than two years ago were excluded from analyses, as well as those living outside California at the
time of the survey, because neighborhood-level variables are created from statewide census data.
Procedures
Treating physicians of eligible CCS were informed of the study and given opportunity to
deny permission to contact their patient for any reason. Paper surveys were mailed to eligible
CCS either in English (or Spanish upon request or selection online), with the option to also
complete the survey online, over the phone, or in person if requested. Follow-up calls and
mailers were completed for non-respondents.(Hamilton et al., 2019) Participants received $20
cash and entry into a lottery for a $300 prize for participating in the survey. All procedures were
17
approved by the California Committee for the Protection of Human Subjects, California Cancer
Registry, and by the human subjects committee at the University of Southern California.
Measures
Demographic information including age, sex, race/ethnicity (Hispanic, non-Hispanic
White, and other), age at diagnosis were obtained from the cancer registry. Health insurance and
education were self-reported.
Residential Address History. Participants were asked to self-report residential address
history from diagnosis to the time of the survey. These addresses were geocoded using Texas
A&M’s geoservices.(Texas A&M University GeoInnovation Center, 2018) Census tract
determined in geocoding was then used to match participants with their neighborhood-level SES
derived from census data for the appropriate corresponding census year. Addresses between the
years of 1996 (earliest year of diagnosis in our sample) and 2005 corresponded to the 2000 5-
year average American Community Survey data, while addresses between 2006 and 2014 (latest
year of diagnosis included in the analytic sample) corresponded to the 2010 5-year average
American Community Survey data.
Due to incomplete reporting of residential history, missing data was assessed (see next
section).
Neighborhood SES. Neighborhood SES was calculated using seven census tract-level
indicator items from the US Census, including education, ratio of household income to poverty
line, employment, blue collar employment, median rental, median value of owner-occupied
housing, and median household income.(Yang, 2014; Yost, Perkins, Cohen, Morris, & Wright,
2001) For the earliest diagnosis years in our sample (1996-2005), nSES at diagnosis was
18
calculated using 2000 census data, while for later years of diagnosis (2006-2010), 2010 census
data were used. Quintiles were assigned to each tract based on the California statewide
distribution. Addresses outside of California were omitted. For mobility analyses, nSES was
dichotomized as the bottom two statewide quintiles versus the top three quintiles, consistent with
previous research on nSES and cancer outcomes.(N. Gomez, Guendelman, Harley, & Gomez,
2015)
Mobility. Mobility type was operationalized as change in nSES from diagnosis to survey.
Mobility type was coded as 0 for stable low (low nSES at diagnosis and at survey), 1 for
downwardly mobile (high nSES at diagnosis and low nSES at survey), 2 for upwardly mobile
(low nSES at diagnosis and high nSES at survey), and 3 for stable high (i.e. high nSES at
diagnosis and at survey), consistent with previous analyses of mobility effects relating childhood
disadvantage to adult health outcomes.(Poulton et al., 2002; Vable, Gilsanz, & Kawachi, 2019)
Raw mobility (“residential instability”) was calculated as the sum of addresses (regardless of
location) divided by the number of years since diagnosis to derive the average number of
residences per year for all cases who provided at least one residential address.
Cumulative SES. Cumulative SES was calculated as the time-weighted average nSES
per year since diagnosis (the sum of all nSES for each residence multiplied by the length of time
spent there, divided by total years since diagnosis). This approach allows for consideration of
both the magnitude of exposure as well as the variable lengths of time since diagnosis across
participants. For missing intervals (whether not reported or removed due to being outside the
California study area), the time-weighted average nSES across nonmissing years within each
individual was imputed prior to calculating to overall time-weighted average.(Cockburn et al.,
2011; Weinberg, Moledor, Umbach, & Sandler, 1996)
19
Treatment Intensity. The Intensity of Treatment Rating Scale 2.0 (ITR-2) uses clinical
and treatment characteristics obtained from a combination of cancer registry data and data
collected from medical charts to categorize cancer cases into four levels of treatment intensity,
where 1=least intensive (e.g. surgery only) 2=moderately intensive (e.g. chemotherapy or
radiation), 3=very intensive (e.g. 2+ treatment modalities), and 4=most intensive (e.g. relapse
protocols). (Kazak et al., 2012) However, large scale registry-based studies are often unable to
access the medical charts of every participant, so a novel method of calculating treatment
intensity was developed using exclusively cancer registry data as a proxy for chart data.
Using our pilot study sample, for which treatment intensity had been previously
determined using medical charts, concordance between treatment intensity estimated by our
method and treatment intensity estimated by the original chart-based method was assessed with
Cohen’s Kappa statistic to validate this approach, showing reasonable agreement between
methods. Full methods on our method of estimation for treatment intensity are described
elsewhere. (Tobin et al., under review)
Late effects. Participants self-reported in the survey which late effects they had already
experienced from a given list, including inability to have children, heart problems, cancer
recurrence, weight gain, liver damage, hearing problems, learning or memory difficulties, lung
problems/difficulty breathing, poor eyesight, problems with sexual functioning, early
menopause, or bone fractures. Late effects were analyzed as the sum of late effects reported.
20
Self-Rated Health. Self-rated health was measured by a single item from the SF-36
asking participants to rate their general health overall.(Ware & Gandek, 1998) Responses ranged
from 0–“Poor” to 4–“Excellent.” To avoid small cell sizes in multivariable models, self-rated
health was recoded to combine the lower categories “poor” and “fair” into a single category,
preserve the middle category “Good,” and combine the top categories “Very good” and
“Excellent.”
Cancer-related follow up care. Participants were coded as a 1 for cancer related follow
up care if they reported having received care within the past 2 years, and 0 if they reported
receiving care never or more than 2 years ago.
Depressive Symptoms. The Center for Epidemiologic Studies Depression Scale was
used to assess past week depressive symptoms. (Radloff, 1977) This scale includes 20 items
about how often participants experienced symptoms in the past week, such as depressed mood,
sleep disruption, and feelings of hopelessness. Response options range from 0-“Rarely/none of
the time” to 3-“Most or all of the time.” For analyses, scores were summed across items with a
possible range of 0-60. Chronbach’s alpha was 0.91.
Wellbeing. Wellbeing was assessed using the 14-item Mental Health Continuum- Short
Form.(Lamers, Westerhof, Bohlmeijer, ten Klooster, & Keyes, 2011) Participants were asked to
indicate how often in the past month their felt a certain way, such as “interested in life,” “that
you belonged to a community,” “good at managing the responsibilities of your daily life,” or
“that your life has a sense of direction or meaning to it.” Response options ranged from 0-
“Never” to 5-“Every day.” Chronbach’s alpha was 0.94.
21
Statistical Analysis
In descriptive analyses, racial/ethnic differences in mobility and cumulative SES were
assessed with Chi square and ANOVA tests (Aim 1). Initial multivariable regression models
were performed to estimate the association between nSES at diagnosis and each outcome as
follows, modeled separately: self-rated health, late effects, follow up care, depressive symptoms,
and wellbeing. Depressive symptoms and wellbeing were modeled using linear regression, self-
rated health was modeled using ordinal logistic regression, follow up care was modeled using
logistic regression, and late effects were modeled using negative binomial Poisson regression
due to overdispersion. The following variables were included as covariates based on previous
literature, (Jacobs & Pucci, 2013; Landier et al., 2004; Milam et al., 2015; Rokitka et al., 2017):
sex, age at diagnosis, age at survey, treatment intensity, race/ethnicity, and education (as a proxy
for individual-level SES), (Aim 2a). Health insurance was also controlled for in models of follow
up care. nSES at survey was then added to subsequent models to estimate associations with
childhood nSES (at diagnosis) net of the effects of adult nSES (at survey). To address Aim 2b,
preliminary mobility associations were assessed with the 4-category mobility indicator (stable
low SES, upwardly mobile, downwardly mobile, stable high SES). If significant mobility effects
were observed, diagonal reference models were run to obtain more accurate estimates, as
detailed below. Given software package limitations on model specification options for diagonal
reference models, outcomes were recoded as dichotomous for all diagonal reference models. To
address Aim 3, models were run using time-weighted average SES as the explanatory variable of
interest, controlling for the same covariates as in mobility analyses. Assumptions, fit statistics,
and collinearity were checked for all models to ensure appropriate model selection and
specification. Statistical analyses were run in R (R Core Team, http://www.R-project.org/).
22
Analysis of Mobility Effects. The use of standard regression models to estimate mobility
effects while simultaneously accounting for the distinct effects of ‘origin’ status (at diagnosis)
and ‘destination’ status (in adulthood) will produce invalid estimates due to the linear
dependency of mobility on both SES at diagnosis and SES in adulthood.(Blalock Jr, 1966; Sobel,
1981) Diagonal reference models, in contrast, estimate mobility effects as the independent effect
of the status change itself, after accounting for the relative contribution of origin and destination
class as well as average values of the outcome for each class. (Houle & Martin, 2011) The
underlying concept is that socioeconomically immobile people represent the true core of a
socioeconomic stratum, and that we might define the characteristics of a given stratum by those
who were born and raised within that stratum.(Missinne et al., 2015) These individuals would lie
in the diagonal cells in Figure 3 labelled µ11 and µ22.
Figure 3.
Mobility estimation schema
The diagonal reference model is estimated as follows:
Yijk = w ´ µii + (1 - w) ´ µjj + S b xijkl + eijk
23
Where:
• Yijk = dependent variable in cell ij of the mobility table, which has k observations
• The portion in blue specifies the influence of the position of origin (i) and destination (j)
• The portion in red specifies the effect of mobility in addition to the effects of origin and
destination
• w estimates the strength of the effect of position of origin relative to the effect of the
position of destination, and lies between 0 and 1. The intercepts estimated for the
diagonal cells, combined with this w parameter enable us to estimate cell-specific
intercepts for all other off-diagonal cells. For example, a w of 1 would indicate that the
position of destination has no effect. In this case, the blue portion of the equation would
be the same for all cells that shared the same origin status.
• µii reflects the estimated mean of Y in the diagonal cell in the row associated with
position of origin
• µjj reflects the estimated mean of Y in the diagonal cell in the column associated with
position of destination
• xijkl is a vector of l covariates (including the mobility predictor of interest) for each of k
observations in the ij
th
cell
• Parameters are interpreted in the same way as in regular regression models
If we were estimating Y for an individual located in the gray box in Figure 3 (who had high
nSES at diagnosis but low nSES in adulthood), µii would refer to the estimated mean of Y in the
bottom right cell, which would be used as the ‘origin effect,’ and µjj would refer to the estimated
mean of Y in the top left cell, used as the ‘destination effect.’ By combining the origin and
24
destination effects in intercepts specific to each cell (and thus, each ‘mobility type’), this model
allows for the specification of the effect of mobility in addition to the contributing origin and
destination effects. (Sobel, 1985; van der Waal, Daenekindt, & de Koster, 2017) Diagonal
reference models were run using the Diagonal Reference (“Dref”) subcommand in the
Generalized Nonlinear Models (“gnm”) R package. (Turner et al., 2007)
Exploratory analysis. While the variability in age at diagnosis and time since diagnosis in
our sample was addressed by creating time-weighted averages per participant, exploratory
analyses were pursued to further explore differences in each outcome by these factors. Critical
periods of child development may amplify negative effects of neighborhood disadvantage on
health outcomes, so interactions with age at diagnosis were considered to uncover potential
differences in vulnerability to the risks of neighborhood disadvantage across childhood.
Likewise, health risk factors may be more influential for those recently diagnosis as compared to
long-term survivors, so exploring interactions with time since diagnosis can uncover such
variations. Potential interactions were explored between age at diagnosis and time since
diagnosis and each predictor related to neighborhood SES (mobility, addresses per year, and
average nSES). Where significant interactions were observed, stratified models were
subsequently run to estimate associations by stratum of age at diagnosis or time since diagnosis.
Sensitivity Analysis. Cumulative SES analyses included all cases who provided at least one
residential address. For sensitivity analyses, the sample was restricted using alternative exclusion
criteria, including excluding those who were missing greater than 75%, 50%, 25%, or any of
their residential history, as well as excluding all addresses that did not provide sufficient street-
level information (i.e. geocoding was performed only on the basis of city name or zip code).
Further, results were also compared based on method of imputing missing residential history.
25
Two other methods were explored as alternatives to the time-weighted average method: the carry
back approach filled in gaps in residential history with the next known address (i.e. extending the
move-in date for the next address back to the move-out date of the prior known address), while
the carry forward approach filled in gaps with the prior known address. Coefficients for
cumulative SES in adjusted models for each outcome were compared across each of these
methods to assess the impact of imputation method on regression estimates.
Sensitivity to variable coding was also examined. Mobility was also re-coded grouping the
3
rd
quintile into the low category rather than high to determine if this coding impacted
conclusions. Self-rated health was dichotomized for diagonal reference models with the middle
response “Good” grouped into the low category. Analyses were also run including the middle
response into the high category to assess whether results were sensitive to this categorization.
Results
962 (87%) participants provided at least one address for residential history. Years of
available/eligible residential history data in the analytic sample ranged from 0.58-21,
corresponding to a range of 0-12 addresses per participant.
1
Neighborhood SES at Diagnosis and Survey
The sample was skewed toward lower nSES at diagnosis, but a greater proportion were
higher SES at survey (Figure 4). nSES at diagnosis and at survey were assessed in relation to all
outcomes. In bivariate analysis, nSES at diagnosis was significantly positively associated with
self-rated health, follow-up care, and wellbeing, while significantly negatively associated with
depressive symptoms (Table 1). nSES at survey was significantly positively associated with self-
1
Eligibility includes being in California and containing sufficient information for geocoding, which requires at least
one of street, city, or zip code.
26
rated health and significantly negatively associated with depressive symptoms (Table 2).
Associations with nSES alone and in conjunction with nSES at survey in fully adjusted models
are presented in Tables 8-9 (follow up care), Tables 11-12 (depressive symptoms), Tables 14-15
(wellbeing), Tables 17-18 (late effects), and Tables 21-22 (self-rated health), and are discussed
below by outcome.
Figure 4.
Distribution of Neighborhood Socioeconomic Status (SES) at Diagnosis and at Survey
Table 1.
Bivariate Associations Between nSES at Diagnosis and Each Outcome (Separate Models)
Dependent variable:
CESD MHC Late effects Fcare Self health
Model type linear linear
Negative
binomial
logistic linear
Estimate(standard error)
27
nSES at
diagnosis
-0.59
**
(0.26) 0.85
**
(0.35) -0.05(0.04) 0.12
**
(0.05) 0.14
***
(0.02)
Observations
811 826 872 873 862
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 2.
Bivariate Associations Between nSES at Survey and Each Outcome (Separate Models)
Dependent variable:
CESD MHC Late effects Fcare Self health
Model type linear linear
Negative
binomial
logistic linear
Estimate(standard error)
nSES at survey -0.58
**
(0.26) 0.46(0.36) -0.02(0.04) 0.06(0.05) 0.12
***
(0.02)
Observations 811 826 872 873 862
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Mobility type
877 (79% of all respondents) were included in mobility analyses based on address at
diagnosis and address at survey, and restricted to California residences. Included cases in
mobility analyses differed from those not included on gender (females overrepresented,
p<0.0001; 51.2% of those included were female compared to 44.1% of those not included), age
at diagnosis (mean age at diagnosis in respondents was slightly younger, 11.71 versus 12.52,
p<0.001), and age at survey (mean age at survey in included cases was slightly younger, 25.00
versus 25.68, p<0.001). Cases included in the mobility analyses did not differ significantly on
nSES at diagnosis nor on race/ethnicity from those not included.
The analytic sample was roughly half Hispanic and half female. Complete descriptive
statistics are provided in Table 3. The majority were stable in their nSES from diagnosis to
survey; 21.9% were mobile- 14.7% upwardly mobile and 7.2% downwardly mobile (Figure 5;
28
Table 4). Mobility type differed significantly by race/ethnicity, where the majority of non-
Hispanic Whites were classified as stable high nSES and the majority of Hispanics were
classified as stable low nSES (Figures 6 and 7; Table 5). In bivariate analyses, the stable high
SES group was associated with significantly fewer depressive symptoms and greater self-rated
health, follow-up care, and wellbeing (Table 6).
Table 3.
Descriptive Statistics, n=877
N
missing(%)
N(%)
Overall
Ethnicity
Hispanic
Non-Hispanic White
Other
0(0.0)
458(52.2)
250(28.5)
169(19.3)
Sex
Female
Male
0(0.0)
449(51.2)
428(48.8)
Treatment Intensity 0(0.0)
Least Intensive
Moderately Intensive
Very Intensive
Most Intensive
62(7.1)
291(33.2)
396(45.2)
128(14.6)
Last cancer treatment 0(0.0)
<2 years ago
>2 years ago
Don’t know
7(0.8)
1
841(95.9)
29(3.3)
Cancer site group 0(0.0)
Leukemia
Lymphoma
Brain & Other Nervous System
Endocrine
Skin
Other
325(37.1)
176(20.1)
121(13.8)
54(6.2)
34(3.9)
167(19)
Follow up care in past 2 years 4(0.5)
Never/2+ years ago
<2 years ago
341(38.9)
532(60.7)
Self-rated Health
Poor
Fair
15(1.7)
26(3)
29
Good
Very Good
Excellent
160(18.2)
308(35.1)
248(28.3)
120(13.7)
Late effects
0
1
2
3+
5(0.6)
532(60.7)
161(18.4)
92(10.5)
87(9.9)
Insured
No
Yes
4(0.5)
72(8.2)
801(91.3)
Insurance type
Uninsured
Private
Public
Other
4(0.5)
72(8.2)
508(57.9)
269(30.7)
15(1.7)
nSES at Dx
1
2
3
4
5
0(0.0)
269(30.7)
195(22.2)
144(16.4)
114(13.0)
155(17.7)
nSES at Survey
1
2
3
4
5
0(0.0)
219(25.0)
179(20.4)
155(17.7)
160(18.2)
164(18.7)
Mobility
Stable low SES
Downwardly mobile
Upwardly mobile
Stable high SES
0(0.0)
335(38.2)
63(7.2)
129(14.7)
350(39.9)
Education
Grade school
Some high school
High school grad/GED
Some college/training
Associate degree
College graduate
Post graduate degree
2(0.2)
4(0.5)
36(4.1)
163(18.6)
344(39.2)
80(9.1)
195(22.2)
53(6.0)
N
missing(%)
Overall
mean(SD)
Range
Age at diagnosis 0(0.0) 11.7(5.3) 0-19
Age at survey 0(0.0) 26.4(4.8) 18-39
30
Years since diagnosis 0(0.0) 14.4(4.4) 5-22
MHC
4
51(5.8) 47.4(15.0) 4-70
CESD Summary Score 66(7.5) 13.9(10.9) 0-58
Total Late Effects 5(0.6) 0.8(1.3) 0-10
1
Only chronic myeloid leukemia cases currently on treatment. All other cases treated less than 2 years prior were
excluded from analyses
Table 4.
Mobility Table: Dichotomized Neighborhood SES at Diagnosis by Neighborhood SES at Survey,
Total n=877
Origin
Destination
Low High
N(% overall)
Low
a
335(38.2) 129(14.7)
High 63(7.2) 350(39.9)
a
Low = bottom 2 nSES quintiles
Figure 5.
Mobility Distribution
Figure 6.
31
Distribution of nSES at Diagnosis, by Race/Ethnicity
Figure 7.
Distribution of nSES at Survey, by Race/Ethnicity
Table 5.
32
Bivariate Association Between Mobility and Race/Ethnicity
Total
Stable low
SES
Downwardly
mobile
Upwardly
mobile
Stable high
SES
n(row %)
Non-
Hispanic
White
16(6.4) 21(8.4) 27(10.8) 186(74.4) 250
Hispanic 285(62.2) 30(6.6) 73(15.9) 70(15.3) 458
Other 34(20.1) 12(7.1) 29(17.2) 94(55.6) 169
Total 335(38.2) 63(7.2) 129(14.7) 350(39.9) 877
χ
2
=308.965 · df=6 · p<0.0001
Table 6.
Bivariate Associations Between Mobility and Each Outcome (Separate Models)
Dependent variable:
CESD MHC Late effects Fcare Self health
Model type linear linear
Negative
binomial
logistic linear
Estimate(standard error)
Downwardly mobile -1.08 (1.56) 0.08 (2.08) 0.24 (0.23) 0.10 (0.28) 0.09 (0.14)
Upwardly mobile -0.09 (1.17) -0.69 (1.59) 0.12 (0.17) -0.19 (0.21) 0.08 (0.10)
Stable high SES -1.72
**
(0.87) 2.20
*
(1.18) -0.01 (0.13) 0.39
**
(0.16)
0.45
***
(0.08)
Observations 811 826 872 873 862
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Reference group = Stable low SES
Table 7.
Bivariate Associations Between Outcomes and Covariates (Each Cell a Separate Model)
Dependent variable:
CESD MHC Late effects Fcare Self health
33
Model type linear linear
Negative
binomial
logistic linear
Estimate(standard error)
Age_dx -0.07 (0.07)
0.13 (0.10) 0.01 (0.01) 0.01 (0.01) -0.01 (0.01)
Education -1.65
***
(0.29)
2.33
***
(0.39) -0.12
***
(0.04) -0.06 (0.05) 0.09
***
(0.03)
Treat_intensity 0.53 (0.47) -1.26
*
(0.64) 0.39
***
(0.07) 0.22
***
(0.09) -0.14
***
(0.04)
Female 0.36 (0.77) 2.11
**
(1.04) 0.32
***
(0.12) 0.26
*
(0.14) -0.07 (0.07)
White (Ref) (Ref) (Ref) (Ref) (Ref)
Hispanic 2.33
***
(0.89)
-3.27
***
(1.21) 0.08 (0.13) -0.27
*
(0.16) -0.40
***
(0.08)
Other 2.39
**
(1.12)
-3.18
**
(1.53) -0.14 (0.17) -0.22 (0.21) -0.17
*
(0.10)
Age_survey -0.08 (0.08) 0.05 (0.11) 0.03
**
(0.01) -0.10
***
(0.01) -0.02
***
(0.01)
Private (Ref)
Public -0.04 (0.16)
Other 0.44 (0.59)
Uninsured -1.32
***
(0.27)
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Follow up care. In bivariate analyses, nSES at diagnosis was positively associated with
follow up care (p<0.05) while nSES at survey was not. Bivariate analysis of follow up care and
mobility indicated that stable high nSES was significantly more likely to report recent follow up
care than the stable low nSES group (p<0.05). However, these significant associations did not
remain after accounting for covariates. In the multivariable model of mobility, age at diagnosis,
treatment intensity, and female gender were positively associated while current age and lack of
health insurance were significantly negatively associated with recent follow up care. Model
coefficients are presented in Tables 8-10.
Table 8.
Multivariable Model of Follow Up Care, With nSES at Diagnosis
OR (95% CI)
Constant 32.82
***
(9.79,113.04)
34
nSES_dx 10.8 (-0.95, 1.23)
Age_dx 1.12
***
(1.08, 1.17)
Education 0.96 (0.85, 1.09)
Treat_intensity 1.34
***
(1.11, 1.63)
Female 1.34
*
(0.99, 1.80)
White (Ref)
Hispanic 0.78 (0.50, 1.20)
Other 0.68
*
(0.43, 1.06)
Age_survey 0.83
***
(0.79, 0.86)
Private (Ref)
Public 0.88 (0.62, 1.25)
Other 1.04
***
(0.33, 3.92)
Uninsured 0.28 (0.16, 0.49)
Observations -0.25
***
(-0.69, 0.19)
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 9.
Multivariable Model of Follow Up Care, with nSES at Diagnosis and Survey
OR (95% CI)
Constant 33.88(9.81,120.19)
nSES_dx 1.08(0.94,1.25)
nSES_surv 0.98(0.86,1.13)
Age_dx 1.12
***
(1.08,1.16)
Education 0.96(0.85,1.09)
Treat_intensity 1.34
***
(1.11,1.62)
Female 1.34
*
(0.99,1.81)
White (Ref)
Hispanic 0.77(0.49,1.2)
Other 0.67
*
(0.43,1.06)
Age_survey 0.83
***
(0.79,0.86)
Private (Ref)
Public 0.87(0.61,1.25)
Other 1.03(0.33,3.9)
Uninsured 0.28
***
(0.16,0.49)
Observations 858
Note:
*
p<0.10
**
p<0.05
***
p<0.01
35
Table 10.
Multivariable Model of Follow Up Care, with Mobility Indicator
OR (95% CI)
Constant 34.17(10.8,111.11)
Stable low SES (Ref)
Downwardly mobile 1.34(0.72,2.54)
Upwardly mobile 0.94(0.59,1.49)
Stable high SES 1.36(0.89,2.1)
Age_dx 1.12
***
(1.08,1.17)
Education 0.96(0.84,1.09)
Treat_intensity 1.34
***
(1.11,1.62)
Female 1.34
*
(0.99,1.81)
White (Ref)
Hispanic 0.82(0.53,1.26)
Other 0.69(0.44,1.08)
Age_survey 0.83
***
(0.79,0.86)
Private (Ref)
Public 0.87(0.61,1.24)
Other 1.05(0.33,3.97)
Uninsured 0.28
***
(0.16,0.49)
Observations 858
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Depressive symptoms. In bivariate analyses, both nSES at diagnosis and at survey were
negatively associated with depressive symptoms (p<0.05). Bivariate analysis of depressive
symptoms and mobility indicated that those with stable high nSES reported significantly fewer
depressive symptoms than the stable low nSES group (p<0.05). However, these significant
associations remained after accounting for covariates. In the multivariable model of mobility,
education was negatively associated while other ethnicity (compared to non-Hispanic White)
was positively associated with depressive symptoms. Model coefficients are presented in Tables
11-13.
Table 11.
36
Multivariable Model of Depressive Symptoms, with nSES at Diagnosis
b (95% CI)
Constant 17.54
***
(11.67, 23.41)
nSES_dx -0.11 (-0.74, 0.53)
Age_dx -0.003 (-0.18, 0.18)
Education -1.69
***
(-2.32, -1.06)
Treat_intensity 0.36 (-0.56, 1.29)
Female 0.48 (-1.00, 1.96)
White (Ref)
Hispanic 0.78 (-1.37, 2.94)
Other 1.86 (-0.36, 4.08)
Age_survey 0.09 (-0.12, 0.29)
Observations 809
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 12.
Multivariable Model of Depressive Symptoms, with nSES at Diagnosis and Survey
b (95% CI)
Constant 17.77
***
(11.76, 23.77)
nSES_dx -0.05 (-0.77, 0.67)
nSES_surv -0.13 (-0.82, 0.57)
Age_dx -0.004 (-0.18, 0.18)
Education -1.68
***
(-2.32, -1.04)
Treat_intensity 0.36 (-0.57, 1.29)
Female 0.49 (-0.99, 1.98)
White (Ref)
Hispanic 0.69 (-1.53, 2.91)
Other 1.84 (-0.38, 4.07)
Age_survey 0.09 (-0.12, 0.29)
Observations 809
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 13.
Multivariable Model of Depressive Symptoms, with Mobility Indicator
37
b (95% CI)
Constant 17.06
***
(11.48, 22.64)
Stable low SES (Ref)
Downwardly mobile 0.25 (-2.84, 3.35)
Upwardly mobile 0.84 (-1.51, 3.19)
Stable high SES 0.07 (-2.05, 2.19)
Age_dx 0.001 (-0.18, 0.18)
Education -1.72
***
(-2.35, -1.08)
Treat_intensity 0.38 (-0.55, 1.30)
Female 0.46 (-1.02, 1.95)
White (Ref)
Hispanic 0.98 (-1.17, 3.12)
Other 1.88
*
(-0.34, 4.10)
Age_survey 0.08 (-0.12, 0.29)
Observations 809
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Wellbeing. In bivariate analyses, nSES at diagnosis was positively associated with
wellbeing (p<0.05) while nSES at survey was not. Bivariate analysis of follow up care and
mobility indicated that stable high nSES was marginally more likely to report greater wellbeing
than the stable low nSES group (p<0.10). However, these effects did not remain after accounting
for covariates. In the multivariable model of mobility, education and female gender were
significantly positively associated while other ethnicity (compared to non-Hispanic White) was
negatively associated with wellbeing. Model coefficients are presented in Tables 14-16.
Table 14.
Multivariable Model of Wellbeing, with nSES at Diagnosis
b (95% CI)
Constant 44.66
***
(36.60, 52.72)
nSES_dx 0.07 (-0.79, 0.94)
Age_dx 0.13 (-0.12, 0.37)
38
Education 2.36
***
(1.50, 3.22)
Treat_intensity -0.91 (-2.16, 0.34)
Female 1.99
*
(-0.02, 3.99)
White (Ref)
Hispanic -1.38 (-4.30, 1.54)
Other -2.67
*
(-5.69, 0.36)
Age_survey -0.26
*
(-0.54, 0.01)
Observations 824
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 15.
Multivariable Model of Wellbeing, with nSES at Diagnosis and Survey
b (95% CI)
Constant 45.72
***
(37.49, 53.94)
nSES_dx 0.35 (-0.62, 1.32)
nSES_surv -0.61 (-1.54, 0.33)
Age_dx 0.12 (-0.13, 0.36)
Education 2.42
***
(1.55, 3.28)
Treat_intensity -0.92 (-2.17, 0.33)
Female 2.08
**
(0.07, 4.09)
White (Ref)
Hispanic -1.84 (-4.85, 1.16)
Other -2.74
*
(-5.76, 0.28)
Age_survey -0.26
*
(-0.54, 0.02)
Observations 824
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 16.
Multivariable Model of Wellbeing, with Mobility Indicator
b (95% CI)
Constant 45.67
***
(38.04, 53.31)
Stable low SES (Ref)
Downwardly mobile -1.85 (-5.98, 2.27)
Upwardly mobile -2.34 (-5.52, 0.84)
Stable high SES -0.90 (-3.79, 1.99)
39
Age_dx 0.12 (-0.13, 0.36)
Education 2.45
***
(1.59, 3.32)
Treat_intensity -0.95 (-2.20, 0.30)
Female 2.04
**
(0.03, 4.05)
White (Ref)
Hispanic -1.89 (-4.79, 1.01)
Other -2.70
*
(-5.72, 0.31)
Age_survey -0.26
*
(-0.53, 0.02)
Observations 824
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Late Effects. The conditional variance of late effects was greater than its conditional
mean, so negative binomial models were used (test for overdispersion z=5.6, p<0.0001, where
the alternative hypothesis is overdispersion). In bivariate analyses, neither nSES at diagnosis or
at survey, nor mobility were significantly associated with late effects. In the multivariable model
of mobility, both the downwardly mobile and stable high SES groups reported greater late effects
compared to the stable low SES group (IRR = 1.67 and 1.41, respectively). Additionally, age at
survey, treatment intensity, and female gender were positively associated with late effects.
However, in the diagonal reference model which simultaneously controls for SES at diagnosis
and at survey, no significant association is observed between mobility and late effects. Model
coefficients are presented in Tables 17-20.
Table 17.
Multivariable Model of Late Effects, with nSES at Diagnosis
IRR (95% CI)
Constant 0.14
***
(0.06, 0.33)
nSES_dx 1.04 (0.95, 1.14)
Age_dx 1.00 (0.97, 1.02)
Education 0.82
***
(0.75, 0.90)
Treat_intensity 1.50
***
(1.31, 1.73)
Female 1.41
***
(1.14, 1.75)
40
White (Ref)
Hispanic 0.99 (0.71, 1.38)
Other 0.81 (0.59, 1,13)
Age_survey 1.05
***
(1.02, 1.08)
Observations 870
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 18.
Multivariable Model of Late Effects, with nSES at Diagnosis and Survey
IRR (95% CI)
Constant 0.13 (0.06, 0.29)
nSES_dx 1.02 (0.91, 1.13)
nSES_surv 1.05 (0.95, 1.16)
Age_dx 1.00 (0.97, 1.02)
Education 0.82
***
(0.75, 0.90)
Treat_intensity 1.51
***
(1.31, 1.73)
Female 1.41
***
(1.14, 1.74)
White (Ref)
Hispanic 1.03 (0.74, 1.43)
Other 0.82 (0.59, 1.14)
Age_survey 1.05
***
(1.02, 1.08)
Observations 870
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 19.
Multivariable Model of Late Effects, with Mobility Indicator
IRR (95% CI)
Constant 0.12 (0.06, 0.25)
Stable low SES (Ref)
Downwardly mobile 1.67
**
(1.06, 2.62)
Upwardly mobile 1.34 (0.92, 1.95)
Stable high SES 1.41
**
(1.06, 1.88)
Age_dx 1.00 (0.97, 1.02)
Education 0.81
***
(0.74, 0.88)
Treat_intensity 1.53
***
(1.33, 1.75)
Female 1.41
***
(1.14, 1.73)
41
White (Ref)
Hispanic 1.10 (0.81, 1.50)
Other 0.85 (0.61, 1.17)
Age_survey 1.05
***
(1.02, 1.08)
Observations 870
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 20.
Diagonal Reference Model of Late Effects (Any Versus None)
OR (95% CI)
Stable low SES (Ref)
Downwardly mobile 2.34(1.51,40.13)
Upwardly mobile 0.84(0.43,1.65)
Stable high SES 1.18(0.57,2.44)
Age_dx 2.66(2.59,19.67)
Education 2.29
***
(2.08,18.68)
Treat_intensity 5.0
***
(3.86,43.21)
Female 3.82
**
(2.75,36.38)
White (Ref)
Hispanic 3.29(2.18,37.35)
Other 2.32(1.72,26.84)
Age_survey 2.86
**
(2.75,21.1)
Observations 870
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Self-Rated Health. In bivariate analyses, nSES at diagnosis and at survey were
positively associated with self-rated health (p<0.01). Bivariate analysis of self-rated health and
mobility indicated that stable high nSES was significantly more likely to report greater self-rated
health than the stable low nSES group (p<0.01). This association remained in the multivariable
model of mobility, while also controlling for current educational attainment. In the adjusted
42
mobility model, the odds of reporting greater self-rated health among those with stable high SES
were 1.64 times that of those with stable low SES (95% CI 1.15-2.35). There were no significant
differences between the upwardly mobile or downwardly mobile individuals compared to the
stable low SES group. Education was also significantly associated with greater odds of reporting
greater self-rated health, while treatment intensity and age at survey were significantly associated
with lower odds of reporting greater self-rated health. In the diagonal reference model which
simultaneously controls for SES at diagnosis and at survey, the significant difference between
the stable high group and the stable low group remained (OR 1.69, 95% CI 1.08-2.64). The
direction and significance of this association did not differ according to whether “good” self-
rated health was coded as low or high. Model coefficients are presented in Tables 21-24.
The weights assigned to nSES at diagnosis and at survey suggest that, regardless of how
the outcome was coded, residence in a low SES neighborhood appears to be more influential on
adult self-rated health than residence in a high SES neighborhood, regardless of whether the
exposure was in childhood or adulthood. This is evidenced by the fact that for upwardly mobile
individuals (low nSES in childhood), their nSES at diagnosis was more influential, while for
downwardly mobile individuals (low nSES in adulthood), their nSES at survey was more
influential (Tables 25 and 26).
Table 21.
Multivariable Model of Self-Rated Health, With nSES at Diagnosis
OR(95% CI)
nSES_dx 1.17
***
(1.05,1.31)
Age_dx 1.01(0.98,1.04)
Education 1.23
***
(1.11,1.38)
Treat_intensity 0.82
**
(0.70,0.97)
Female 0.84(0.65,1.09)
White (Ref)
43
Hispanic 0.74(0.51,1.07)
Other 0.87(0.59,1.28)
Age_survey 0.92
***
(0.89,0.96)
Observations 860
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 22.
Multivariable Model of Self-Rated Health, With nSES at Diagnosis and survey
OR (95% CI)
nSES_dx 1.15
**
(1.0,1.31)
nSES_surv 1.03(0.91,1.16)
Age_dx 1.01(0.98,1.04)
Education 1.23
***
(1.1,1.3)
Treat_intensity 0.82
**
(0.70,0.97)
Female 0.84(0.65,1.08)
White (Ref)
Hispanic 0.76(0.52,1.10)
Other 0.87(0.59,1.29)
Age_survey 0.92
***
(0.89,0.96)
Observations 860
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 23.
Multivariable Model of Self-Rated Health, With Mobility Indicator
OR (95% CI)
Stable low SES (Ref)
Downwardly mobile 1.01(0.60,1.72)
Upwardly mobile 0.99(0.67,1.46)
Stable high SES 1.64
***
(1.15,2.35)
Age_dx 1.01(0.98,1.04)
Education 1.24
***
(1.11,1.38)
Treat_intensity 0.81
**
(0.69,0.96)
Female 0.83(0.64,1.08)
White (Ref)
44
Hispanic 0.73
*
(0.50,1.05)
Other 0.86(0.58,1.27)
Age_survey 0.92
***
(0.89,0.96)
Observations 860
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 24.
Diagonal Reference Model of Self-Rated Health (Low (Poor/Fair/Good) Versus High (Very
good-Excellent)
OR (95% CI)
Stable low SES (Ref)
Downwardly mobile 0.72(0.41,1.26)
Upwardly mobile 0.65(0.32,1.33)
Stable high SES 1.69
**
(1.08,2.64)
Age_dx 1.01(0.98,1.04)
Education 1.21
***
(1.07,1.37)
Treat_intensity 0.83
**
(0.7,0.99)
Female 0.87(0.65,1.15)
White (Ref)
Hispanic 0.7(0.47,1.03)
Other 0.99
*
(0.66,1.49)
Age_survey 0.93
***
(0.89,0.96)
Observations 860
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 25.
Weights for Diagonal Reference Model of Self-Rated Health, With Response Level 3 Coded as
Low
nSES at diagnosis nSES at survey
Weight
Immobile 0.55 0.45
Upward 0.91 0.09
Downward 0.38 0.62
Note: Weights for the two immobile categories were similar so they were grouped
Table 26.
45
Weights for Diagonal Reference Model of Self-Rated Health, With Response Level 3 Coded as
High
nSES at diagnosis nSES at survey
Weight
Immobile 0.55 0.45
Upward 0.76 0.24
Downward 0.07 0.93
Note: Weights for the two immobile categories were similar so they were grouped
Addresses Per Year
Of 1,106 Project Forward Cohort respondents, 962 (87%) provided at least one address
for residential history and were included in analyses assessing associations with number of
addresses per year since diagnosis. Included cases differed from those not included (all cases
eligible for the PFC study) on gender (females more likely to respond, p<0.0001; females
represented 43.6% of those not included and 51.6% of the sample that provided at least one
address), age at diagnosis (mean age at diagnosis in respondents was slightly younger, 11.68
versus 12.58, p<0.0001), and age at survey (mean age at survey in respondents was slightly
younger, 25.04 versus 25.68, p=0.001). Cases included in these analyses did not differ on nSES
at diagnosis nor on race/ethnicity from those not included.
Number of addresses (regardless of quality or location of residence) per year since
diagnosis ranged from 0.05-1.17. Number of addresses per year differed significantly by
race/ethnicity (F=28.3, p<0.0001), education (F=57.27(2,1099), p<0.0001) and age
(F=15.83(2,1100), p<0.0001) such that non-Hispanic Whites, those with higher education levels,
and older ages reported a greater number of addresses per year. To assess the impact of
restricting to California residences, the distribution of number of addresses per year was plotted
for all addresses and for only those in California (Figure 8), suggesting that restricting to
California residences did not markedly affect the distribution of the primary predictor in this
46
study. In other words, this suggests that reporting any residence outside California does not
appear to be associated with residential instability.
Missingness in residential history across observations ranged between 0-19.42 years. In
terms of the proportion of time since diagnosis, the range was 0-97% (Figure 9). 90% of the
sample had at least 75% of their residential history in California, while 64.8% of the sample
reported complete residential history within California (Table 27). When assessing differences in
missingness by demographic factors, no differences in any missingness (versus non) were
observed, however missing over 25% versus less than 25% significantly differed by
race/ethnicity. This is largely explained by the difference in education, as education is also
positively associated with a greater number of residences and likelihood of missing at least 25%
of residential history, and is higher among non-Hispanic whites than among Hispanics or other
races/ethnicities (Table 28). Among those with a high school education or less, those with some
college/vocational training, and those with a bachelors or graduate degree, mean addresses per
year were 0.17, 0.20, and 0.30 (p<0.0001), and percent missing at least 25% of residential history
was 7.1, 8.6, and 12.2 (p=0.02), respectively.
Figure 8.
Distribution of Number of Addresses Reported Per Year, All Versus Within California
47
Figure 9.
Distribution of Proportion of Time Since Diagnosis Missing Residential Data
Table 27.
Proportion of the Sample Providing at Least a Minimum Threshold of Residential History
48
(Relative to Time Since Diagnosis)
% time since
Dx
% of sample providing at
least this threshold of
residential history
100 64.8
75 90.4
50 95.2
25 98.6
Table 28.
Educational Attainment by Race/Ethnicity
Non-Hispanic
White
Hispanic Other
Column %
High school or less 12.23 28.51 18.07
Some college,
vocational, or AA
35.78 54.21 48.80
Bachelors or
graduate degree
51.99 17.28 33.13
Follow up care. In neither bivariate nor multivariable analyses, addresses per year were
not significantly associated with follow up care (Table 29). Age at diagnosis, treatment intensity,
and female gender were positively associated while Hispanic ethnicity, age at survey, and lack of
health insurance were significantly negatively associated with recent follow up care. Adjusted
model coefficients are presented in Table 30.
Table 29.
Bivariate Associations of All Outcomes and Number of Addresses Per Year Since Diagnosis
Dependent variable:
Fcare CESD Wellbeing Late effects Self-health
49
Model type logistic Linear Linear
Negative
binomial
Ordinal
logistic
Estimate (Standard error)
Adds p/year 0.31 (0.44) 0.81 (2.40)
1.91 (3.28) 0.10 (0.36) 0.26 (0.38)
Observations 958 891 910 956 946
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 30.
Multivariable Model of Follow Up Care and Number of Addresses Per Year Since Diagnosis
OR (95% CI)
Constant 28.75(10.05,84.3)
Adds p/year 0.55(0.19,1.56)
Age_dx 1.13
***
(1.09,1.17)
Education 0.98(0.87,1.1)
Treat_intensity 1.39
***
(1.16,1.68)
Female 1.5
***
(1.13,2)
White (Ref)
Hispanic 0.72
*
(0.5,1.02)
Other 0.69
*
(0.45,1.04)
Age_survey 0.83
***
(0.8,0.87)
Private (Ref)
Public 0.88(0.63,1.24)
Other 1.06(0.34,3.98)
Uninsured 0.30
***
(0.17,0.51)
Observations 942
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Depressive symptoms. In bivariate analyses, number of addresses per year was not
significantly associated with depressive symptoms. In adjusted models, it was marginally
associated with greater depressives symptoms (p=0.06). Education was significantly negatively
associated while Hispanic and other ethnicities (compared to non-Hispanic White) were
positively associated with depressive symptoms. Adjusted model coefficients are presented in
Table 31.
50
Table 31.
Multivariable Model of Depressive Symptoms and Number of Addresses Per Year Since Diagnosis
b (95% CI)
Constant 14.66
***
(9.54, 19.79)
Adds p/year 4.84
*
(-0.23, 9.91)
Age_dx -0.05 (-0.23, 0.13)
Education -1.72
***
(-2.31, -1.13)
Treat_intensity 0.50 (-0.40, 1.40)
Female 0.56 (-0.86, 1.97)
White (Ref)
Hispanic 1.61
*
(-0.10, 3.31)
Other 2.68
**
(0.61, 4.75)
Age_survey 0.13 (-0.06, 0.33)
Observations 889
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Wellbeing. In neither bivariate nor adjusted analyses, number of addresses per year was
not significantly associated with wellbeing. Education was significantly positively associated
while other ethnicity(compared to non-Hispanic White) and age at survey were negatively
associated with wellbeing. Adjusted model coefficients are presented in Table 32.
Table 32.
Multivariable Model of Wellbeing and Number of Addresses Per Year Since Diagnosis
b (95% CI)
Constant 46.85
***
(39.80, 53.90)
Adds p/year -5.13 (-12.02, 1.76)
Age_dx 0.17 (-0.07, 0.42)
Education 2.55
***
(1.75, 3.36)
Treat_intensity -1.07
*
(-2.29, 0.15)
Female 1.79
*
(-0.13, 3.70)
51
White (Ref)
Hispanic -1.95
*
(-4.25, 0.35)
Other -3.48
**
(-6.30, -0.66)
Age_survey -0.32
**
(-0.58, -0.05)
Observations 908
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Late effects. In neither bivariate nor adjusted analyses, number of addresses per year was
not significantly associated with late effects. Treatment intensity, female gender, and age at
survey were significantly positively associated while education was negatively associated with
late effects. Adjusted model coefficients are presented in Table 33.
Table 33.
Multivariable Model of Late Effects and Number of Addresses Per Year Since Diagnosis
IRR (95% CI)
Constant 0.16(0.08,0.33)
Adds p/year 1.53(0.79,2.95)
Age_dx 0.99(0.96,1.01)
Education 0.82
***
(0.75,0.89)
Treat_intensity 1.51
***
(1.32,1.74)
Female 1.41
***
(1.15,1.72)
White (Ref)
Hispanic 0.91(0.71,1.17)
Other 0.79(0.59,1.07)
Age_survey 1.05
***
(1.02,1.08)
Observations 953
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Self-Rated Health. In neither bivariate nor adjusted analyses, number of addresses per
year was not significantly associated with self-rated health. Treatment intensity, Hispanic
52
ethnicity, and age at survey were significantly negatively associated while education was
positively associated with self-rated health. Adjusted model coefficients are presented in Table
34.
Table 34.
Multivariable Model of Self-Rated Health and Number of Addresses Per Year Since Diagnosis
OR (95% CI)
Adds p/year 0.71(0.28,1.79)
Age_dx 1.01(0.97,1.04)
Education 1.26
***
(1.13,1.41)
Treat_intensity 0.80
***
(0.67,0.94)
Female 0.88(0.68,1.15)
White (Ref)
Hispanic 0.59
***
(0.43,0.8)
Other 0.84(0.59,1.21)
Age_survey 0.92
***
(0.88,0.95)
Observations 917
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Average nSES per year
833 participants were included in analyses of average nSES since diagnosis. Included
cases differed from those not included on nSES at diagnosis (a higher proportion of included
cases were from high SES neighborhoods compared to excluded cases, p=0.01). Cases included
in the mobility analyses did not differ significantly from those excluded on gender,
race/ethnicity, age at diagnosis, or age at survey.
Follow up care. In bivariate analysis, average SES was marginally positively associated
with follow up care (p<0.10, Table 35). In the adjusted model however this association did not
53
remain. Age at diagnosis, treatment intensity, and female gender were positively associated
while age at survey and lack of health insurance were significantly negatively associated with
recent follow up care. Adjusted model coefficients are presented in Table 36.
Table 35.
Bivariate Associations of All Outcomes and Average nSES
Dependent variable:
Fcare CESD Wellbeing Late effects Self-health
Model type logistic Linear Linear
Negative
binomial
Ordinal
logistic
Estimate (Standard error)
Avg. nSES 0.10
*
(0.05) -0.69
**
(0.30)
1.22
***
(0.40) -0.01 (0.04) 0.29
***
(0.05)
Observations 830 774 790 828 818
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Table 36.
Multivariable Model of Follow Up Care and Average nSES
OR (95% CI)
Constant 20.4(5.79,73.7)
Avg. nSES 1.07(0.92,1.24)
Age_dx 1.12
***
(1.08,1.16)
Education 0.96(0.84,1.1)
Treat_intensity 1.31
***
(1.08,1.6)
Female 1.51
***
(1.12,2.05)
White (Ref)
Hispanic 0.79(0.51,1.21)
Other 0.85(0.54,1.34)
Age_survey 0.84
***
(0.8,0.87)
Private (Ref)
Public 1.1(0.76,1.59)
Other 1.25(0.34,5.97)
Uninsured 0.26
***
(0.14,0.48)
54
Observations 818
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Depressive Symptoms. In bivariate analysis, average SES was significantly negatively
associated with depressive symptoms (p<0.05). In the adjusted model however this association
did not remain. Other ethnicity was significantly positively associated while education was
negatively associated with depressive symptoms. Adjusted model coefficients are presented in
Table 37.
Table 37.
Multivariable Model of Depressive Symptoms and Average nSES
b (95% CI)
Constant 14.33(8.06, 20.61)
Avg. nSES -0.04 (-0.78, 0.70)
Age_dx 0.02 (-0.16, 0.21)
Education -1.62
***
(-2.28, -0.95)
Treat_intensity 0.44 (-0.55, 1.42)
Female 0.71 (-0.83, 2.26)
White (Ref)
Hispanic 1.60 (-0.64, 3.84)
Other 2.97
**
(0.67, 5.26)
Age_survey 0.15 (-0.06, 0.36)
Observations 772
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Wellbeing. In bivariate analysis, average SES was significantly positively associated
with wellbeing (p<0.01). In the adjusted model however this association did not remain.
Treatment intensity and other ethnicity were significantly negatively associated while education
was positively associated with wellbeing. Adjusted model coefficients are presented in Table 38.
55
Table 38.
Multivariable Model of Wellbeing and Average nSES
b (95% CI)
Constant 45.06 (36.83, 53.29)
Avg. nSES 0.29 (-0.69, 1.27)
Age_dx 0.02 (-0.23, 0.27)
Education 2.55
***
(1.68, 3.42)
Treat_intensity -1.33
**
(-2.63, -0.04)
Female 1.59 (-0.44, 3.62)
White (Ref)
Hispanic -1.50 (-4.41, 1.42)
Other -3.27
**
(-6.29, -0.24)
Age_survey -0.23 (-0.51, 0.05)
Observations 788
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Late Effects. In neither bivariate nor adjusted analyses, average SES was not
significantly associated with late effects. Treatment intensity, female gender, and age at survey
were significantly positively associated while education was negatively associated with late
effects. Adjusted model coefficients are presented in Table 39.
Table 39.
Multivariable Model of Late Effects and Average nSES
IRR (95% CI)
Constant 0.13(0.06,0.29)
Avg. nSES 1.06(0.97,1.17)
Age_dx 0.99(0.96,1.01)
Education 0.84
***
(0.76,0.91)
Treat_intensity 1.50
***
(1.3,1.73)
Female 1.39
***
(1.12,1.71)
56
White (Ref)
Hispanic 0.96(0.72,1.3)
Other 0.80(0.58,1.09)
Age_survey 1.05
***
(1.02,1.08)
Observations 826
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Self-Rated Health. In bivariate analysis, average SES was significantly positively
associated with self-rated health (p<0.01). However when adjusting for covariates, this
association was attenuated. In the adjusted model, an increase of 1 in average SES was
associated with a 14% increase in the odds of reporting greater self-rated health (p=0.05).
Education was also significantly positively associated while treatment intensity was negatively
associated with self-rated health. Adjusted model coefficients are presented in Table 40.
Table 40.
Multivariable Model of Self-Rated Health and Average nSES
OR (95% CI)
Avg. nSES 1.14
*
(1.00,1.31)
Age_dx 1.00(0.96,1.03)
Education 1.26
***
(1.11,1.43)
Treat_intensity 0.75
***
(0.62,0.89)
Female 0.86(0.65,1.13)
White (Ref)
Hispanic 0.77(0.52,1.15)
Other 0.92(0.62,1.37)
Age_survey 0.92
***
(0.88,0.95)
Observations 792
Note:
*
p<0.10
**
p<0.05
***
p<0.01
57
Exploratory analysis
Interactions were explored between age at diagnosis (age <14 vs. >= 14) and the primary
predictors- mobility, addresses per year, and average nSES, for all outcomes. None of these
associations differed by age at diagnosis for any outcome.
Interactions were also explored between time since diagnosis (<16 vs. 16+ years) and the
primary predictors- mobility, addresses per year, and average nSES, for all outcomes. The only
association which differed significantly by time since diagnosis was addresses per year on follow
up care (interaction term beta= -0.29, p=0.02). The distribution of addresses per year by time
since diagnosis is presented in Table 41, showing that the maximum and standard deviation are
greater among those more recently diagnosed. Upon subsequent stratification of the sample by
time since diagnosis, no association was observed between addresses per year and follow up care
among those more recently diagnosed, whereas among those diagnosed greater than 15 years
prior, number of residences is negatively associated with the probability of having recent cancer-
related follow up care (b=-2.2, p=0.04).
Table 41.
Distribution of Addresses Per Year, by Time Since Diagnosis
Years since diagnosis Range Mean(SD)
<16 0.07-1 0.23(0.17)
16+ 0.05-0.63 0.17(0.12)
Sensitivity analyses
To assess whether method of imputation affected the distribution of time-weighted
average nSES, their distributions were compared (Table 42, and depicted in Figure 10), showing
consistency. Time-weighted mean imputation resulted in slightly lower variance, as expected.
Table 42.
58
Comparison of the Distribution of Time-Weighted Average Neighborhood SES with Missing
Years Imputed by Alternate Methods
Range Mean(SD)
Mean imputation 0.89-5 2.82(1.31)
Carry forward 0.03-5.19 2.70(1.33)
Carry back 0.03-5.49 2.69(1.33)
Note: Carry back method: Gaps filled in by extending the later address back to the end date of prior address;
Carry forward method: Gaps filled in by extending the prior address forward to the start date of the subsequent
address; Mean imputation method: Gaps filled in with the time-weighted average nSES of all addresses provided by
a given participant
Figure 10.
Comparison of the Distribution of Time-Weighted Average Neighborhood SES Derived From
Three Alternate Methods of Missing Data Imputation
Note: Carry back method: Gaps filled in by extending the later address back to the end date of prior address;
59
Carry forward method: Gaps filled in by extending the prior address forward to the start date of the subsequent
address; Mean imputation method: Gaps filled in with the time-weighted average nSES of all addresses provided by
a given participant
The estimated association between average nSES and all outcomes in adjusted models
was compared across imputation methods (Table 43) as well as across varying exclusion criteria
(Table 44). Results demonstrate that conclusions do not differ based on the method of imputation
nor by exclusion criteria.
Table 43.
Comparison of Coefficients for Time-Weighted Average nSES Derived from Alternative
Imputation Methods
Dependent variable:
CESD MHC Late effects Fcare Self-health
Model type linear linear
Negative
binomial
logistic linear
Estimate(standard error)
Weighted mean -0.04(0.37) 0.07(0.51) 0.06(0.06) 0.07(0.08) 0.13
**
(0.07)
Carry forward -0.15(0.36) 0.05(0.48) 0.04(0.05) 0.10(0.07) 0.14
***
(0.06)
Carry back -0.20(0.36) 0.13(0.48) 0.04(0.05) 0.12(0.07) 0.14
***
(0.06)
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Estimates are from multivariable adjusted models, covariate effects not shown
Table 44.
Comparison of Coefficients for Time-Weighted Average nSES with Varied Exclusion Criteria
Dependent variable:
CESD MHC Late effects Fcare Self health
Model type
linear linear
Negative
binomial
logistic linear
60
N(% of
full
sample)
Estimate(standard error)
All cases
1
958(100) -0.04(0.38) 0.29(0.50) 0.06(0.05) 0.07(0.08) 0.13
*
(0.07)
< 75%
2
821(98.6) -0.11(0.39) 0.35(0.51) 0.06(0.06) 0.08(0.08) 0.15
**
(0.07)
< 50%
3
793(95.2) -0.16(0.39) 0.39(0.51) 0.08(0.06) 0.07(0.08) 0.15
**
(0.07)
< 25%
4
753(90.4) -0.17(0.41) 0.43(0.54) 0.07(0.06) 0.10(0.08) 0.17
***
(0.07)
No missingness
5
540(64.8) -0.05(0.47) 0.02(0.65) 0.05(0.07) 0.04(0.10) 0.18
**
(0.09)
< NAACCR code
6
6
815(97.8)
-0.20(0.39) 0.60(0.51) 0.03(0.06) 0.05(0.08) 0.18
**
(0.07)
Note:
*
p<0.10
**
p<0.05
***
p<0.01
Estimates are from multivariable adjusted models, covariate effects not shown
1
Retained if at least one prior address was reported
2
nSES was imputed for less than 75% of years since diagnosis
3
nSES was imputed for less than 50% of years since diagnosis
4
nSES was imputed for less than 25% of years since diagnosis
5
Only cases who provided complete residential history included
6
Includes only addresses whose NAACCR geocoding quality code was less than
5- geocoded with greater precision than zip code or city centroid (i.e. at least
street name was known)
Discussion
While the collection of residential history data is intensive for participants and poses
numerous data management challenges, it provides a unique and valuable insight into mobility
patterns and socioeconomic trajectories of young adult survivors of childhood cancer. This study
provides insight into unique indicators of mobility and socioeconomic status across critical
developmental periods for young adult CCS.
Psychosocial health
In this study, none of the neighborhood SES exposure variables were significantly
associated with depressive symptoms or wellbeing in adjusted models. While prior research
61
indicates an effect of socioeconomic inequalities on mental health in children and adolescents,
lasting effects on adult mental health and wellbeing in our sample were not observed.(Reiss,
2013) Thus, these outcomes are more strongly associated with other concurrent factors. This is
supported by the finding that educational attainment (a component of/proxy for adult
socioeconomic status) was significantly associated with both depressive symptoms (negatively
associated) and wellbeing (positively associated). While SES is well known to be associated with
mental health through numerous pathways, contextual SES is often less strongly associated in
models adjusted for individual-level SES.(Pickett & Pearl, 2001) Similarly, our sample may not
have been large enough to detect a weak association between the contextual environment and
psychosocial health outcomes.
Wellbeing also differed by gender and treatment intensity. Females reported greater
wellbeing, which may be due to greater coping strategies, social relationships and support, or
other personal resources.(Rosenfield & Mouzon, 2013) Treatment intensity in our sample was
associated with decreased wellbeing. Prior research on the effects of cancer treatment on
psychosocial health in adulthood among survivors of pediatric cancer has shown that higher
levels of treatment intensity were associated with fewer positive health beliefs and greater
anxiety, which may contribute to a diminished sense of wellbeing.[kazak 2010]
Late effects
None of the neighborhood SES exposure variables were significantly associated with late
effects after fully adjusting for covariates. This is largely due to the fact that late effects are
primarily driven by treatment exposure, which is accounted for in our models by the inclusion of
treatment intensity. Additionally, late effects are strongly associated with current age, as older
survivors have had a greater time since diagnosis in which they were susceptible to late effects.
62
Late effects also differed significantly by gender, with females reporting greater late effects.
Gender differences in adverse events from multiple treatment regimens have been documented,
and may be explained by differential likelihood to report late effects, adherence, or
pharmacogenomics.(J. M. Unger et al., 2019)
Education was also significantly negatively associated with late effects, which is
consistent with the broad literature base demonstrating the health protective effects of
socioeconomic status (SES). SES impacts health through numerous pathways. For example,
stress is associated with low SES, and is directly related to adverse health effects, and also
exacerbates existing health problems.(Santiago, Wadsworth, & Stump, 2011) SES is also
associated with numerous health behaviors such as tobacco use, physical inactivity, and poor
nutrition, all of which may increase the likelihood of comorbidities among cancer
survivors.(Pampel, Krueger, & Denney, 2010)
Follow up care
No significant main effects of neighborhood SES exposure variables on follow up care
were observed in fully adjusted models. However, the association with number of addresses per
year (“crude mobility”) differed by time since diagnosis. For those less than 16 years post-
diagnosis, there was no association, while for those at least 16 years from diagnosis, mobility
was negatively associated with receipt of recent cancer-related follow up care. This is consistent
with expectation, as preventive healthcare utilization is associated with having a usual source of
care- an ongoing relationship with a health service facility or individual provider.(DeVoe, Fryer,
Phillips, & Green, 2003) Having such an ongoing connection is complicated with frequent
changes in residence.
63
Some amount of mobility is attributed to education, as young adults often move away to
attend college and may move annually throughout college. Indeed, in our sample, those with
higher education reportedly significantly more residences per year. The maximum and the
variance of number of addresses is greater among those more recently diagnosed, likely due to
the fact that those further from diagnosis are also older and more likely to have experienced
greater residential stability in the years after college, bringing down their average number of
addresses per year since diagnosis. These survivors further out from diagnosis who have not
found residential stability are at greater risk of not receiving the recommended cancer related
follow-up care.
This has important implications for survivorship care management. For example, the
transition out of pediatric care into independent adult risk-based care is a vulnerable period in
which survivors must learn to navigate their own health and healthcare. This challenge is
compounded by other challenging aspects of emerging adulthood often experienced during the
same period, such as separation from the nuclear family, clarification of goals, and
experimentation.(Arnett, 2000; Henderson, Friedman, & Meadows, 2010) While residential
instability tends to be higher during this phase, our findings suggest that the number of residence
changes was not associated with likelihood of receiving recent follow up care among this group.
The chronological proximity to diagnosis may signify continued connections to healthcare
systems or providers and possibly the continuing involvement of family to assist in managing
care. Consistent with this notion, despite the relatively increased stability of life at older ages,
previous research shows that cancer-related follow up care declines with age.(Milam et al., 2015;
Nathan et al., 2009)
64
Our analyses indicate that residential instability is one possible explanation for this
decline in follow up care with increasing time since diagnosis, a decline not fully explained by
individual or neighborhood SES or insurance status. This association may instead be attributed to
lack of a usual source of cancer-related care. It has been documented that having a stable
provider-patient relationship can improve receipt of preventive healthcare services.(Bustamante,
Chen, Rodriguez, Rizzo, & Ortega, 2010; Xu, 2002) For some, having a usual source of care
may actually be a more salient factor than having health insurance in determining healthcare
utilization.(DeVoe et al., 2003) The impact of having a usual source of care, and distinguishing
the relative importance of having a regular provider versus a regular service site, warrants further
investigation among survivors of childhood cancer in order to inform efforts aimed at retention
among this population at high risk of morbidity and early mortality. While our survey did not
include a measure of usual source of cancer care, we did capture usual source of non-cancer care.
While not a perfect proxy (53% of those who reported a usual source of care indicated this
provider was the same one seen for cancer-related follow up care), it is a related indicator. As
expected, among those with a usual source of care, mean addresses per year was significantly
lower (data not shown). Interventions focused on patient education that rely on more stable
contact information (relative to residential address) such as email may support sustained
connection between survivors and providers and increase retention in follow up care among
aging survivors of pediatric cancer. Further, alternative methods of engagement such as social
media groups or apps may provide additional means for keeping survivors connected to
healthcare providers and cognizant of recommended follow up guidelines.
65
Self-rated health
Analyses of mobility type and self-rated health revealed that stable high SES was
associated with significantly higher self-rated health compared to stable low SES. While this
may seem to suggest that time spent in high SES areas is linearly positively associated with
health, analysis of average nSES per year (accounting for cumulative exposure to neighborhood
SES) showed only a marginal association with self-rated health in fully adjusted models.
Similarly, neither of the mobile profiles (upward or downward) were significantly different than
the stable low SES group, suggesting that even among those for whom residence a low SES area
was discontinuous, the risk of poorer self-rated health was not mitigated. Results of the diagonal
reference model also support this notion, as the weights for mobile individuals suggest the time
period spent in a low SES neighborhood was more influential on adult self-rated health than
residence in a high SES neighborhood, regardless of whether the exposure was in childhood or
adulthood.
The lack of significant difference in health between the stable low SES and either of the
mobile groups may be reflective of the health risks associated with residence in a disadvantaged
neighborhood at any time of life. Even among those survivors who currently reside in more
advantaged neighborhoods, the influence of childhood disadvantage may persist, imparting direct
effects or exacerbating existing health problems. Prior research has demonstrated that
neighborhood disadvantage predicts poor self-rated health longitudinally (ruling out the question
of selection processes whereby people in poor health move to disadvantaged areas), supporting
our finding that residence in a low SES neighborhood, even if only experienced in childhood,
may adversely impact adult health.(Glymour, Mujahid, Wu, White, & Tchetgen, 2010) Lower
SES neighborhoods may convey numerous risks to residents such as a reduced sense of safety,
increased stress and subsequent inflammation, decreased promotion of physical activity, social
66
connectedness, and overall access to health-supportive institutions.(Diez Roux & Mair, 2010;
Meyer, Castro-Schilo, & Aguilar-Gaxiola, 2014) Conversely, a strong compound advantage is
experienced by those who reside in high SES neighborhoods in childhood as well as adulthood.
Missingness
To evaluate the quality of residential history data, rates of missingness were assessed in
relation to demographic factors. Among cases eligible for analysis related to nSES in this study
(residences in California), 65% provided complete residential history since diagnosis, 90%
provided at least 75% of residential history, and nearly 98% of addresses provided included at
least street-level detail for more precise geocoding, suggesting that cumulative exposures can be
reasonably reconstructed for young adult survivors based on this self-reported data. Differences
in missingness in residential history by race/ethnicity and education indicate that these covariates
must be accounted for in analysis of variables derived from self-reported residential history to
reduce bias.
To assess the impact of missingness on estimates of association between exposure
variables derived from the residential history data and the primary outcomes of this study,
regression coefficients from adjusted models were compared across samples with differing levels
of restriction for missing data, showing little variation across samples. This demonstrates that
bias due to differing levels of completeness or precision of addresses did not have a significant
effect on statistical results and conclusions.
Limitations and strengths
Residential history data in our sample may be biased based on mobility itself. It is quite
possible that the most highly mobile cancer survivors in our eligible pool of cases were also
those for whom we were unable to obtain accurate contact information through tracing efforts,
67
which would bias our sample toward the less mobile. If this were the case, the variability in
crude mobility would be underestimated in our sample, biasing results toward the null. An
additional limitation in our collection of residential history data is that we did not capture history
dating back to birth but rather to diagnosis, so given the variability in age at diagnosis and age at
survey, we were unable to assess cumulative exposure across the duration of childhood for all
cases. However, the utilization of a time-weighted average nSES allowed standardization across
cases with variable lengths of exposure, as has been done in previous epidemiological studies
focused on cumulative exposure.(Cockburn et al., 2011; Weinberg et al., 1996)
This study only assessed childhood nSES as an index based on financial capital. However,
other types of capital such as human capital (e.g. skills and capabilities), and social capital (e.g.
strong social relationships) also predict health outcomes and provide a broader account of
contextual influences.(Vable, Gilsanz, Nguyen, Kawachi, & Glymour, 2017) Future work should
explore the relative importance of such alternative forms of capital among childhood cancer
survivors, and how this knowledge can be used to increase support services. For example,
survivors lacking social capital may be identified to participate in virtual support groups aimed at
increasing social support and wellbeing.(McLaughlin et al., 2012)
Our findings related to self-rated health must be interpreted with some caution, as its
meaning may differ by age and SES.(Dowd & Zajacova, 2010; Layes, Asada, & Kephart, 2012)
Prior research has shown that higher socioeconomic position and younger age is associated with
a pessimistic perception of health, while the opposite is evident among those from lower
socioeconomic position. These differences may be attributed to differences in true underlying
health status as well as differences in reporting behavior. The latter may be related to changing
optimism with increasing age.(Leinonen, Heikkinen, & Jylhä, 2001) Perceptions of health is also
68
influenced by comparisons within one’s context, which may explain why individuals from
disadvantaged circumstances may still perceive themselves in good health.(Whitehead, Drever,
& Doran, 2005) However, exploration of the association between self-rated health and other
measures of health in our sample (data not shown) suggest that self-rated health is an accurate
reflection of overall health (i.e. those who reported the greatest self-rated health also had the
lowest levels of depressive symptoms, highest levels of wellbeing, and the least late effects),
supporting the construct validity of the self-rated health measure within our sample.
Conclusion
Health may be conceptualized as a form of life course capital to be preserved or depleted
as a function of numerous factors, some of which systematically differ by neighborhood
socioeconomic status.(Shuey & Willson, 2014) This study utilized self-reported residential
history to gain insight into the residential stability and socioeconomic mobility of childhood
cancer survivors and their associations with a range of health outcomes. While most outcomes do
not appear to be strongly driven by neighborhood socioeconomic status, some findings suggest
that residential instability may be a risk factor for maintaining follow up care among survivors
farther from diagnosis, and that residing in low SES neighborhoods at any point may negatively
impact adult health compared to those with continuous residence in more advantaged
neighborhoods. These findings may inform future efforts to identify CCS at heightened risk of
poor health and disengagement with preventive healthcare.
69
Chapter 2: Multilevel demographic and cultural factors associated
with mental health, physical health, and healthcare utilization
among childhood cancer survivors
Overview and Specific Aims
Follow up care is critical for long-term surveillance and management of health conditions
among childhood cancer survivors. Limited research has identified individual-level clinical and
demographic factors associated with health status and healthcare utilization, although cultural
factors have not been as thoroughly explored. This study will assess the relationship of mental
health, physical health, and healthcare utilization with individual-level ethnicity as well as
individual- and neighborhood-level acculturation among a sample of young adult survivors of
childhood cancer. A conceptual model of these associations is presented in Figure 11.
Aim 1. Assess the association of neighborhood acculturation with physical and mental health,
and health behaviors (late effects, self-rated health, cancer-related and non-cancer-related health
care use, depressive symptoms, wellbeing), as well as the possible interaction between ethnicity
and neighborhood acculturation.
Hypothesis 1. Neighborhood acculturation will be moderated by individual ethnicity, such that
Hispanics living in low acculturated neighborhoods (i.e. with more Hispanic residents) will be
more likely to report higher self-rated health, healthcare use, and wellbeing, and less likely to
report depressive symptoms and late effects compared to Hispanics in more acculturated
neighborhoods.
70
Aim 2. Among Hispanics, assess the association between individual acculturation and health
outcomes (same as Aim 1)
Hypothesis 2. High Anglo orientation will be positively associated with healthcare use, self-rated
health, and depressive symptoms, and negatively associated with late effects.
Figure 11. Study 2 conceptual model
Macro-level factors
Neighborhood SES &
ethnic enclave
Micro-level factors
Demographic,
clinical,
socioeconomic,
cultural
Follow up care
Psychosocial health
Physical health
71
Introduction
Follow up care and health status among childhood cancer survivors
While survival has improved in recent decades for childhood cancer survivors (CCS), the
majority will experience an adverse health condition later in life. Previous studies have shown
that about 75% of CCS will develop some late effect (cancer treatment-related sequelae) in
adulthood, with about 25% facing severe or life-threatening conditions.(Kremer et al., 2013)
CCS who survived five or more years have been shown to be 2.5 times as likely to report adverse
general health than their non-cancer survivor siblings. (M. M. Hudson et al., 2003) Issues may
range from the physical, such as organ dysfunction, metabolic disorders, and cardiovascular
disease, to the psychological, such as anxiety or depression, (Jacobs & Pucci, 2013; Nathan et
al., 2009)
Due to the high rates of morbidity among CCS, risk-based follow up care in adulthood is
recommended in order to identify and manage late effects or secondary malignancies that may
not become apparent until many years after completing cancer treatment.(Kremer et al., 2013;
Landier et al., 2004) Despite this, less than 75% of CCS obtain cancer-related follow up care,
with some studies reporting as low as 31%. (Milam et al., 2015; Nathan et al., 2008; Kevin C
Oeffinger et al., 2004) Reasons for not getting recommended care may include a lack of
knowledge about the likelihood of late effects, (Bhatia & Meadows, 2006) or financial or
insurance limitations. (Henderson et al., 2010) Improved understanding of factors associated
with receipt of follow up care may inform future interventions aimed at appropriately educating
and supporting CCS as they transition out of the pediatric oncology care setting.
72
Ethnic disparities in childhood cancer survivorship
Despite having lower incidence rates, Hispanic CCS in Los Angeles County have
exhibited greater mortality rates compared to non-Hispanic whites. (Liu, Zhang, & Deapen,
2010) While the causes of this disparity are not fully known, unequal access to care is likely a
significant contributor. Ethnic disparities in healthcare utilization have been demonstrated among
CCS, with Hispanic ethnicity being less likely to remain engaged in care. (Milam et al., 2015;
Rokitka et al., 2017) Cultural factors may partially explain such disparities, as well as account
for heterogeneity within Hispanics. However, this has not been thoroughly explored among CCS.
The majority of previous research on CCS has predominantly focused on older, less ethnically
diverse cohorts, so improved understanding is needed of factors that may impact access to care
among minority populations.
Acculturation
Acculturation, a multi-dimensional adaptation resulting from contact with a new culture,
may impact health status and healthcare utilization through a range of mediating pathways.
Though there is a dearth of evidence on this association in the cancer survivorship literature,
acculturation has been found to be associated with healthcare access among other non-childhood
cancer populations, with mixed findings related to the use of preventive screenings for cancer,
HIV, and cardiovascular disease. Most previous studies have found that acculturation among
immigrants in the US was positively associated with getting screened, (Echeverria &
Carrasquillo, 2006; Jurkowski & Johnson, 2005; Kinsler et al., 2009; O'malley, Kerner, Johnson,
& Mandelblatt, 1999) while another found that certain domains of acculturation (e.g. loss
traditional family ties) were negatively associated with healthcare use. (Suarez, 1994)
73
A number of potential mechanisms may explain these associations, including factors such
as health literacy, defined as the ability to understand and/or act on medical instructions or
information, and reflects more than just education level or reading skills. (S. J. Shaw et al., 2009)
Cultural beliefs about disease etiology, patient-provider communication, or treatment may
influence health literacy.(S. J. Shaw et al., 2009) Poor or inaccurate disease knowledge and low
perceived risk of late effects from cancer treatment may undermine healthcare self-efficacy and
utilization. Other beliefs may also impact healthcare utilization; for example, fatalism, or the
belief that there is little one can do to alter fate, has been found to deter preventive healthcare
visits among minority populations.(Jun & Oh, 2013; Shelton, Jandorf, Ellison, Villagra, &
DuHamel, 2011) Less acculturated Hispanics are also likely to have more limited English
proficiency which can pose a barrier to accessing healthcare services.(Cheng, Chen, &
Cunningham, 2007; DuBard & Gizlice, 2008; Ponce, Hays, & Cunningham, 2006) While the
majority of the extant literature on the relation between acculturation and healthcare access has
predominantly focused on non-clinical or non-cancer populations, similar mechanisms may also
be salient among CCS.
With respect to physical and psychological health status, associations with acculturation
are mediated by different factors. Findings on the association with mental health among
Hispanics in the US suggest that symptoms of depression increase with acculturation to the US
culture (Gonzales, Deardorff, Formoso, Barr, & Barrera, 2006; Lorenzo-Blanco, Unger, Ritt-
Olson, Soto, & Baezconde-Garbanati, 2011) This may be due to loss of protective cultural
values such as close family ties. However with regard to self-rated health, one study found that
those who were more Mexican-oriented (less acculturated) were over three times as likely to
74
self-report fair or poor health compared to those who were more Anglo-oriented (more
acculturated). (Johnson et al., 2010) These associations may be explained by poorer access to
healthcare, or by perceived discrimination and the associated chronic stress. Discrimination has
been directly linked to self-rated health and depression among Latinos in the US, and may be
moderated by protective factors such as social support.(Finch, Hummer, Kol, & Vega, 2001;
Finch & Vega, 2003; Flores et al., 2008) However, consideration of such factors is lacking from
the cancer survivorship literature, despite the fact that acculturation may be a unique contributor
to health in minority CCS given their existing physical vulnerability as well as their potentially
increased resiliency gained from the cancer experience.
Neighborhood-level effects
Though individual- and neighborhood-level factors are correlated, they impart distinct
effects on individual health outcomes. Associations between the social environment and
healthcare access and utilization have been well documented and include issues such as
information networks, social capital/collective efficacy, and behavior norms.(Hendryx, Ahern,
Lovrich, & McCurdy, 2002; Prentice, 2006) The availability of healthcare providers and
services also differs by neighborhood characteristics. More segregated, ethnically dense
geographic areas are often characterized by lower SES and tend to have poorer access to care.
This may be due to a limited supply of health facilities and less access to primary care
offices.(Gaskin et al., 2012; Hussein, Diez Roux, & Field, 2016; Ko & Ponce, 2013)
Despite the abundance of evidence on differences in the distribution of healthcare
resources by the ethnic makeup of neighborhoods, there is limited research on the independent
effects of neighborhood acculturation on healthcare access and utilization. Neighborhood
acculturation has been measured as an area’s status as an ethnic enclave, an area with widely
75
held cultural norms and practices and that is culturally and/or ethnically distinct from
surrounding areas.(S. L. Gomez et al., 2011) This may reflect such factors as language, beliefs,
infrastructures, and dense social networks, all of which can impact access to and utilization of
medical care. In addition to direct effects, these neighborhood characteristics may also moderate
disparities in healthcare utilization by individual race/ethnicity, given that concordance between
one’s own cultural identity and the predominant culture of one’s neighborhood is likely to
influence susceptibility to neighborhood influences.(Chang & Chan, 2015; Gaskin et al., 2012;
Haas et al., 2004)
This study will assess the relationship of individual- and neighborhood-level ethnicity
and acculturation with measures of behavioral, physical, and psychological health among a
multi-ethnic sample of CCS diagnosed in Los Angeles County. It is hypothesized that the effect
of neighborhood acculturation will be moderated by individual ethnicity, such that Hispanics
living in low acculturated neighborhoods (i.e. with more Hispanic residents) will be more likely
to report higher self-rated health, healthcare use, and wellbeing, and less likely to report
depressive symptoms and late effects compared to Hispanics in more acculturated
neighborhoods. With respect to acculturation, it is hypothesized that high Anglo orientation will
be positively associated with healthcare use, self-rated health, and depressive symptoms, and
negatively associated with late effects.
Methods
Participants
Participants were recruited to Project Forward, a study of follow up care and health
among CCS. Eligible participants were recruited from the Los Angeles Cancer Surveillance
76
Program, the Surveillance, Epidemiology, and End Results (SEER) cancer registry for Los
Angeles County, and included CCS diagnosed at age 19 or younger who were between the ages
of 18 and 39 at the time of the study in 2015. Participants who had received cancer treatment less
than two years prior as well as those living outside California at the time of the survey were
excluded, because neighborhood-level variables were created from California census data.
Procedures
Treating physicians of eligible CCS were informed of the study and given opportunity to
deny permission to contact their patient for any reason. Paper surveys were mailed to eligible
CCS either in English (or Spanish upon request or selection online), with the option to also
complete the survey online, over the phone, or in person if requested. Follow-up calls and
mailers were completed for non-respondents. Participants received $20 cash and entry into a
lottery for a $300 prize for participating in the survey. All procedures were approved by the
California Committee for the Protection of Human Subjects, California Cancer Registry, and by
human subjects committees at the University of Southern California.
Measures
Demographic information including sex, age at diagnosis, age at survey, and were
obtained from the cancer registry. Hispanic versus non-Hispanic ethnicity, education, and health
insurance status were self-reported in the survey.
77
Residential Address. Current residential address was ascertained either from cancer
registry follow up data or from tracing efforts using databases of credit-reporting history (e.g.
LexisNexis, Experian). Addresses were then confirmed through recruitment efforts such as
mailings or phone calls during the study. These addresses were geocoded using Texas A&M’s
geoservices. (Texas A&M University GeoInnovation Center, 2018) Census tract determined in
geocoding was then used to match participants with their neighborhood-level census data (see
Neighborhood Acculturation below).
Treatment Intensity. The Intensity of Treatment Rating Scale 2.0 (ITR-2) uses clinical
and treatment characteristics obtained from a combination of cancer registry data and data
collected from medical charts to categorize cancer cases into four levels of treatment intensity,
where 1=least intensive (e.g. surgery only) 2=moderately intensive (e.g. chemotherapy or
radiation), 3=very intensive (e.g. 2+ treatment modalities), and 4=most intensive (e.g. relapse
protocols). (Kazak et al., 2012) However, large scale registry-based studies are often unable to
access the medical charts of every participant, so a novel method of calculating treatment
intensity was developed using exclusively cancer registry data as a proxy for chart data.
Using our pilot study sample, for which treatment intensity had been previously
determined using medical charts, concordance between treatment intensity estimated by our
method and treatment intensity estimated by the original chart-based method was assessed with
Cohen’s Kappa statistic to validate this approach, showing reasonable agreement between
methods. Full methods on our method of estimation for treatment intensity are described
elsewhere. (Tobin et al., under review)
78
Neighborhood Acculturation. A composite index of Hispanic neighborhood
acculturation, or ‘ethnic enclave,’ was created through principal components analysis based on
US Census data for California census tracts. Variables included in the index score included
percent linguistically isolated, percent linguistically isolated who speak Spanish, percent
speaking limited English, percent Spanish speaking who speak limited English, percent recent
immigrants, percent Hispanic, and percent foreign born. These variables were chosen to recreate
the composite Hispanic ethnic enclave index created by Gomez and colleagues in previous
research on cancer incidence and survival. (N. Gomez et al., 2015; Keegan et al., 2010) Higher
scores represent higher ethnic density and limited English language proficiency, or lower
neighborhood acculturation. This variable was dichotomized for analysis as the bottom three
quintiles versus the top two due to a skewed distribution, consistent with prior research assessing
neighborhood acculturation and cancer outcomes.(N. Gomez et al., 2015)
Individual Acculturation. Participants responded to a 13-item measure of acculturation
if they self-identified as Hispanic. The scale was adapted from the Acculturation Rating Scale for
Mexican Americans-II.(Cuellar, Arnold, & Maldonado, 1995) This shortened scale was taken
from a larger cohort study of Latino adolescents in Los Angeles County.(J. B. Unger, 2014)
Questions comprised two subscales, one for each culture (Hispanic and Anglo), based on the
conceptualization of acculturation as a bi-dimensional process.(Berry, 1997) This adapted scale
included items focused predominantly on linguistic preference and behavioral aspects of
acculturation such as “I enjoy speaking Spanish,” “I enjoy Spanish language TV,” and “My
friends are of Anglo or White origin.” Possible responses ranged from 1 (“Not at all”) to 5
(“Almost always/extremely often”). Scores were summed for each subscale, ranging from 6-30
for the Hispanic subscale (based on six items) and 6-35 for the Anglo subscale (based on seven
79
items). Chronbach’s alpha for Hispanic and Anglo orientation subscales for this sample were
0.92 and 0.71, respectively.
Neighborhood SES. Neighborhood SES (nSES) was calculated using seven census tract-
level indicator items from the US Census, including education, ratio of household income to
poverty line, employment, blue collar employment, median rental, median value of owner-
occupied housing, and median household income.(Yang, 2014; Yost et al., 2001) For the earliest
diagnosis years in our sample (1996-2004), nSES at diagnosis was calculated using 2000 census
data, while for later years of diagnosis (2005-2010), 2010 census data were used. Quintiles were
assigned to each tract based on the California statewide distribution.
Late effects. Participants self-reported in the survey which late effects they had already
experienced from a given list, including issues such as heart problems, weight gain, liver
damage, lung problems, or infertility. Late effects were analyzed as the total number of late
effects reported.
Self-Rated Health. Self-rated health was measured by a single item from the SF-36
asking participants to rate their general health overall.(Ware & Gandek, 1998) Responses ranged
from 0–“Poor” to 4–“Excellent.”
Cancer-related follow up care. Participants were coded as a 1 for cancer related follow
up care if they reported having received care within the past 2 years, and 0 if they reported
receiving care never or more than 2 years ago.
80
Depressive Symptoms. The Center for Epidemiologic Studies Depression Scale was
used to assess past week depressive symptoms. (Radloff, 1977) This scale includes 20 items
about how often participants experienced symptoms in the past week, such as depressed mood,
sleep disruption, and feelings of hopelessness. Response options range from 0-“Rarely/none of
the time” to 3-“Most or all of the time.” For analyses, scores were summed across items with a
possible range of 0-60. Chronbach’s alpha was 0.91.
Wellbeing. Wellbeing was assessed using the 14-item Mental Health Continuum- Short
Form.(Lamers et al., 2011) Participants were asked to indicate how often in the past month their
felt a certain way, such as “interested in life,” “that you belonged to a community,” “good at
managing the responsibilities of your daily life,” or “that your life has a sense of direction or
meaning to it.” Response options ranged from 0-“Never” to 5-“Every day.” Chronbach’s alpha
was 0.94.
Statistical Analysis
Dependent variables were assessed using linear regression for continuous outcomes
(depressive symptoms, wellbeing, and self-rated health), logistic regression for binary outcomes
(cancer-related follow up care), and negative binomial Poisson regression models for count
variables (late effects). For each outcome, initial models assessed ethnicity, neighborhood
acculturation, and the interaction between the two to assess potential effect modification while
controlling for covariates. Secondary models explored associations with individual acculturation,
which were limited to self-identified Hispanics because non-Hispanics did not complete the
acculturation scale in the survey. The following variables were included as covariates based on
previous literature, (Jacobs & Pucci, 2013; Landier et al., 2004; Milam et al., 2015; Rokitka et
81
al., 2017): sex, age at diagnosis, age at survey, treatment intensity, education (as a proxy for
individual SES), nSES. Health insurance was also controlled for in the follow up care models.
All analyses were conducted using SAS statistical software (Version 9.4) (SAS Institute, Cary,
North Carolina).
Results
The complete analytic sample consisted of 989 CCS, roughly half Hispanic. Complete
descriptive statistics are presented in Table 45. Survey responders differed from non-responders
in the full sample (not restricted to those living in California) by gender, Hispanic ethnicity, and
socioeconomic status. However, because current address could not be confirmed for non-
responders we are unable to compare responders and non-responders in this analytic subsample
and thus are unable to create and apply survey weights to analyses.
Table 45.
Sample Descriptive Statistics, n=989 Overall
N
missing(%)
N(%)
Overall
Non-
Hispanic/NR
1
Hispanic
p-value
Self-reported
ethnicity
Hispanic
Non-Hispanic
36(3.6)
538(54.4)
415(42.0)
Female
Male
0(0.0) 501(50.7)
488(49.3)
213(51.3)
202(48.7)
253(47.0)
285(53.0)
0.188
Foreign born
US born
19(1.9) 110(11.1)
860(87.0)
44(10.6)
370(89.4)
64(12.0)
469(88.0)
0.508
Treatment
Intensity
0(0.0) <.0001
Least Intensive
Moderately
Intensive
Very Intensive
Most Intensive
66(6.7)
303(30.6)
489(48.5)
140(14.2)
51(12.3)
152(36.6)
187(45.1)
25(6.0)
20(3.7)
197(36.6)
283(52.6)
38(7.1)
82
Last cancer
treatment
22(2.2) 0.218
<2 years ago
>2 years ago
Don’t know
8(0.8)
2
925(93.5)
34(3.4)
5(1.2)
395(96.1)
11(2.7)
(0.4)
504(95.8)
20(3.8)
Cancer site group 0(0.0) <.0001
Leukemia
Lymphoma
Brain & Other
Nervous System
Endocrine
Skin
Other
360(36.4)
206(20.8)
147(14.9)
56(5.7)
37(3.7)
183(18.5)
120(28.9)
99(23.9)
79(19.0)
22(5.3)
30(7.2)
65(15.7)
223(41.5)
103(19.1)
59(11.0)
34(6.3)
4(0.7)
115(21.4)
Follow up care in
past 2 years
19(1.9) 0.965
Never/2+ years
ago
<2 years ago
404(40.9)
566(57.2)
170(41.5)
240(58.5)
219(41.3)
311(58.7)
Self-rated Health
Poor
Fair
Good
Very Good
Excellent
28(2.8)
29(2.9)
177(17.9)
352(35.6)
267(27)
136(13.8)
8(1.9)
58(14.1)
139(33.7)
137(22.2)
70(17.0)
20(3.8)
113(21.6)
207(39.6)
122(23.3)
61(11.7)
<.0001
Late effects
0
1
2
3+
19(1.9)
596(60.3)
181(18.3)
98(9.9)
95(9.6)
Insured
No
Yes
19(1.9)
93(9.4)
877(88.7)
28(6.8)
386(93.2)
63(11.8)
471(88.2)
0.009
Insurance type
Uninsured
Private
Public
Other
33(3.34)
93(9.4)
546(55.2)
301(30.4)
16(1.6)
28(6.8)
299(72.6)
77(18.7)
8(1.9)
63(12.1)
239(45.8)
213(40.8)
7(1.3)
<0.0001
Ethnic enclave
(quintile)
3
0
1
2
3
4
0(0.0)
92(9.3)
169(17.1)
190(19.2)
238(24.1)
300(30.3)
73(17.6)
125(30.1)
115(27.7)
72(17.4)
30(7.2)
14(2.6)
39(7.3)
65(12.1)
157(29.2)
263(48.9)
<.0001
Education
Grade school
38(3.8)
4(0.4)
1(0.2)
3(0.6)
<.0001
83
Some high school
High school
grad/GED
Some
college/training
Associate degree
College graduate
Post graduate
degree
46(4.8)
177(18.6)
368(38.7)
99(9.3)
211(22.2)
57(6.0)
10(2.4)
54(13.1)
143(34.6)
38(9.2)
126(30.5)
41(9.9)
36(6.7)
123(22.9)
225(41.8)
50(9.3)
85(15.8)
16(3.0)
Overall
mean(SD)
Range
Mean(SD):
Non-
Hispanic/NR
Mean(SD):
Hispanic
p-value
Age at diagnosis 0(0.0) 11.7(5.3) 0-19 12.1(5.3) 11.5(5.3) 0.095
Age at survey 0(0.0) 26.1(4.9) 18-41 26.6(5.0) 25.8(4.7) 0.006
Years since
diagnosis
0(0.0) 14.5(4.4) 5-22 14.6(4.3) 14.3(4.5) 0.291
MHC
4
68(6.9) 47.3(15.0) 4-70 48.3(14.3) 46.6(15.3) 0.092
CESD Summary
Score
87(8.8) 14.0(10.9) 0-58 13.0(10.8) 14.5(11.0) 0.046
Total Late Effects 19(1.9) 0.8(1.3) 0-10 0.7(1.2) 0.8(1.3) 0.100
Acculturation
5
Hispanic
orientation
Anglo orientation
16(3.0)
20(3.7)
18.5(6.6)
29.3(3.9)
6-30
12-35
--
-- --
1
Ethnicity not self-reported
2
Only chronic myeloid leukemia cases currently on treatment. All other cases treated less than 2 years prior were
excluded from analyses
3
Neighborhood Hispanic ethnic enclave quintiles created based on the California statewide distribution of census
tracts, where 0 is the lowest enclave (i.e. more English proficient, least percent Hispanics and foreign born) and 5 is
the highest enclave
4
Mental Health Continuum sum score
5
Only completed by self-identified Hispanic participants. Total missing reported is out of 538 self-identified
Hispanics
Follow-up care
In unadjusted analyses, female gender, Hispanic orientation, treatment intensity, and
being insured were significantly positively associated with cancer-related follow up care, while
84
age was significantly negatively associated. Residence in a high ethnic enclave was marginally
negatively associated (p=0.09). Unadjusted estimates are presented in Table 46.
In the adjusted initial model, females were 38% more likely to have received recent
follow up care (95% CI: 1.04-1.83). Odds of follow up care increased by 12% (95% CI: 1.07-
1.18) with each increasing year of age at diagnosis, while they decreased by 17% for each
increasing year of current age (95% CI: 0.80-0.87). Having any insurance was associated with
greater odds of follow up care. Ethnic enclave was not significantly associated, nor was its
interaction with ethnicity. Initial model estimates are presented in Tables 47 and 48.
In the secondary model restricted to Hispanics, age at diagnosis, current age, and
insurance remained significantly associated in the same direction as in the initial model, and
Hispanic orientation was also significantly positively associated with follow up care. While the
interaction term of ethnic enclave and Hispanic orientation was not significant, in models
stratified by ethnic enclave, Hispanic orientation was significantly associated among high
enclaves but not significantly associated among low enclaves. Among Hispanics living in high
ethnic enclaves, a one-point increase in Hispanic orientation was associated with a 4% increase
in the probability of having had recent follow up care (95% CI 1.00-1.08) Secondary model
estimates are presented in Tables 49-52.
Table 46.
Unadjusted Associations With Cancer-Related Follow-Up Care
N Estimate Std Error t p-value
Female
Hispanic
Hispanic orientation
Anglo orientation
Treatment intensity 1
Treatment intensity 2
970
940
514
510
970
0.074
0.001
0.007
-0.001
---
0.101
0.032
0.032
0.003
0.006
---
0.067
2.34
0.04
2.08
-0.15
---
1.50
<0.0001
0.965
0.038
0.880
---
0.135
85
Treatment intensity 3
Treatment intensity 4
Age
Education
Age at diagnosis
Uninsured
Public insurance
Private insurance
Other insured
Ethnic enclave
nSES
970
962
970
943
970
970
0.132
0.189
-0.021
-0.011
0.003
--
0.315
0.310
0.380
-0.054
0.014
0.065
0.074
0.003
0.012
0.003
--
0.058
0.055
0.131
0.032
0.011
2.03
2.55
-6.63
-0.91
1.10
--
5.44
5.66
2.90
-1.71
1.27
0.042
0.011
<0.0001
0.364
0.272
--
<0.0001
<0.0001
0.004
0.087
0.205
Table 47.
Multivariable Model of Follow-Up Care, n=931
OR 95% CI
Ethnic enclave
nSES
Hispanic
Female
Age at diagnosis
Age
Treatment intensity
Education
Uninsured
Other insurance
Private insurance
Public insured
0.84
1.03
1.13
1.38
1.12
0.83
1.25
0.99
--
4.23
3.74
3.27
0.52-1.35
0.81-1.30
0.82-1.56
1.04-1.83
1.07-1.18
0.80-0.87
0.92-1.70
0.83-1.19
--
1.98-9.03
2.34-5.98
1.81-5.90
Table 48.
Multivariable Model of Follow-Up Care, Including Interaction Between Ethnicity and Enclave.
Results Not in Odds Ratio Form to Provide Estimates for Interaction Terms, n=931
Estimate Std. Error T value p-value
Ethnic enclave
Hispanic
Hispanic*Enclave
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
Uninsured
-0.08
0.06
-0.12
0.02
0.33
0.12
-0.19
0.23
-0.01
--
0.12
0.08
0.11
0.12
0.14
0.03
0.02
0.15
0.09
--
-0.70
0.73
-1.09
0.20
2.30
4.66
-9.18
1.46
-0.11
--
0.48
0.47
0.28
0.84
0.03
<.0001
<.0001
0.15
0.91
--
86
Other insurance
Private insurance
Public insured
0.46
0.34
0.21
0.26
0.12
0.16
1.76
2.80
1.29
0.08
0.01
0.20
Table 49.
Multivariable Model of Follow-Up Care, Assessing Acculturation and Enclave Main Effects
(Restricted to Hispanics), n=481
OR 95% CI
Ethnic enclave
Hispanic orientation
Anglo orientation
nSES
Uninsured
Other insurance
Private insurance
Public insured
Female
Age at diagnosis
Age
Treatment intensity
Education
0.64
1.04
0.98
1.02
--
2.95
3.77
3.01
1.17
1.12
0.82
1.33
1.13
0.33-1.23
1.00-1.08
0.95-1.01
0.77-1.35
--
1.42-6.14
2.22-6.40
1.71-5.33
0.85-1.61
1.05-1.20
0.78-0.86
0.80-2.19
0.89-1.45
Table 50.
Multivariable Model of Follow-Up Care, Assessing Acculturation*Enclave Interaction
(Restricted to Hispanics) Results not in Odds Ratio Form to Provide Estimates for Interaction
Terms, n=481
Estimate Std. Error T value p-value
Ethnic enclave
Hispanic orientation
Enclave*Hisp. orient
Anglo orientation
Enclave*Anglo orient
nSES
Uninsured
Other insurance
Private insurance
Public insured
Female
Age at diagnosis
Age
Treatment intensity
Education
0.22
0.02
0.02
0.01
-0.03
0.02
--
0.19
0.45
0.23
0.13
0.11
-0.20
0.28
0.12
1.76
0.03
0.03
0.05
0.06
0.13
--
0.29
0.18
0.16
0.17
0.03
0.02
0.25
0.12
0.13
0.77
0.59
0.14
-0.56
0.19
--
0.65
2.49
1.39
0.79
3.56
-8.45
1.15
1.03
0.90
0.45
0.56
0.89
0.58
0.85
--
0.52
0.02
0.17
0.43
<0.001
<.0001
0.26
0.31
87
Table 51.
Stratified Models of Follow Up Care: Among High Ethnic Enclaves (Top 2 Quintiles) n=372
OR 95% CI
Female
Hispanic orientation
Anglo orientation
Treatment intensity
Age
Education
Age at diagnosis
nSES
Uninsured
Other insurance
Private insurance
Public insured
1.15
1.04
0.98
1.26
0.82
1.13
1.11
0.91
--
2.57
3.46
2.74
0.81-1.64
1.00-1.08
0.94-1.01
0.69-2.29
0.76-0.88
0.84-1.53
1.05-1.18
0.67-1.22
--
1.23-5.37
1.81-6.61
1.50-5.00
Table 52.
Stratified Models of Follow Up Care: Among Low Ethnic Enclaves (Bottom 3 Quintiles) n=111
OR 95% CI
Female
Hispanic orientation
Anglo orientation
Treatment intensity
Age
Education
Age at diagnosis
nSES
Uninsured
Private insurance
Public insured
1.20
1.04
1.01
1.89
0.79
1.13
1.19
1.50
--
5.98
5.83
0.39-3.65
0.98-1.09
0.84-1.20
1.32-2.70
0.68-0.91
0.79-1.62
1.03-1.39
0.99-2.26
--
1.93-18.60
1.12-30.20
Note: No cases with ‘other’ insurance in this stratum
Depressive symptoms
In unadjusted analyses, Hispanic ethnicity was marginally significantly positively
associated with depressive symptoms (p=0.05), while education, age, and Hispanic orientation
were significantly negatively associated. Hispanic orientation was marginally significant and was
negatively associated with depressive symptoms (p=0.08). Unadjusted estimates are presented in
Table 53.
88
In the adjusted initial model, only education remained significantly negatively associated
with depressive symptoms. Female gender was marginally significant. Ethnic enclave was not
significantly associated, nor was its interaction with ethnicity. Initial model estimates are
presented in Tables 54-55.
In the secondary model restricted to Hispanics, Hispanic orientation was significantly
negatively associated with depressive symptoms (p=0.03). While the interaction term of ethnic
enclave and Hispanic orientation was not significant, in models stratified by ethnic enclave,
Hispanic orientation was significantly negatively associated among high enclaves (p<0.001) but
not significantly associated among low enclaves. Secondary model estimates are presented in
Tables 56-59. Associations between Hispanic orientation and depressive symptoms, stratified by
ethnic enclave presented in Figure 12.
Table 53.
Unadjusted Associations with Depressive Symptoms (CESD sum)
N Estimate Std Error t p-value
Female
Hispanic
Hispanic orientation
Anglo orientation
Treatment intensity 1
Treatment intensity 2
Treatment intensity 3
Treatment intensity 4
Age
Education
Age at diagnosis
Ethnic enclave
nSES
902
878
479
475
902
902
897
902
902
900
0.634
1.479
-0.132
-0.144
---
1.823
1.013
1.739
-0.133
-1.447
-0.060
1.170
-0.267
0.727
0.739
0.075
0.133
---
1.567
1.515
1.717
0.075
0.272
0.068
0.729
0.254
0.87
2.00
-1.76
-1.08
---
1.16
0.67
1.01
-1.78
-5.3
-0.88
1.61
-1.05
0.383
0.046
0.079
0.280
---
0.245
0.504
0.311
0.075
<0.0001
0.382
0.109
0.293
Table 54.
Multivariable Model of Depressive Symptoms, n=876
Estimate Std Error t p-value
Female 0.98 0.49 1.99 0.05
89
Hispanic
Treatment intensity
Age
Education
Age at diagnosis
Ethnic enclave
nSES
0.48
0.02
-0.06
-1.40
0.08
1.18
0.42
0.77
0.49
0.07
0.28
0.05
1.38
0.47
0.62
0.03
-0.93
-4.92
1.55
0.86
0.89
0.54
0.97
0.36
<0.0001
0.13
0.40
0.38
Table 55.
Multivariable Model of Depressive Symptoms, Including Interaction Between Ethnicity and
Enclave, n=876
Estimate Std Error t p-value
Female
Hispanic
Treatment intensity
Age
Education
Age at diagnosis
Ethnic enclave
Enclave*Hispanic
nSES
0.99
0.59
0.02
-0.06
-1.40
0.08
1.30
-0.23
0.41
0.50
0.91
0.49
0.07
0.29
0.05
1.57
1.08
0.46
1.98
0.64
0.04
-0.93
-4.90
1.54
0.83
-0.21
0.89
0.05
0.52
0.97
0.36
<.0001
0.13
0.41
0.83
0.38
Table 56.
Multivariable Model of Depressive Symptoms, Assessing Acculturation and Enclave Main
Effects (Restricted to Hispanics), n=460
Estimate Std Error t p-value
Female
Treatment intensity
Age
Education
Age at diagnosis
Ethnic enclave
Hispanic orientation
Anglo orientation
nSES
-0.10
-0.25
-0.07
-0.90
0.05
2.15
-0.14
-0.12
0.63
0.68
0.52
0.15
0.44
0.12
1.25
0.06
0.10
0.51
-0.15
-0.47
-0.49
-2.06
0.43
1.72
-2.33
-1.28
1.22
0.89
0.64
0.63
0.05
0.67
0.09
0.03
0.21
0.23
Table 57.
Multivariable Model of Depressive Symptoms, Assessing Acculturation*Enclave Interaction
(Restricted to Hispanics), n=460
Estimate Std Error t p-value
Female 0.08 0.63 0.13 0.90
90
Treatment intensity
Age
Education
Age at diagnosis
Ethnic enclave
Hispanic orientation
Enclave*Hisp orient.
Anglo orientation
Enclave*Anglo orient
nSES
-0.18
-0.07
-0.94
0.05
15.18
0.08
-0.30
0.08
-0.27
0.58
0.54
0.16
0.46
0.12
11.63
0.22
0.25
0.30
0.35
0.54
-0.33
-0.45
-2.04
0.40
1.31
0.38
-1.22
0.27
-0.77
1.08
0.74
0.66
0.05
0.69
0.20
0.71
0.23
0.79
0.45
0.29
Table 58.
Multivariable Model of Depressive Symptoms, Assessing Acculturation Effect Among Low Ethnic
Enclaves, n=104
Estimate Std Error t p-value
Female
Treatment intensity
Age
Education
Age at diagnosis
Hispanic orientation
Anglo orientation
nSES
-0.35
-1.87
-0.01
-1.91
0.12
0.07
0.17
0.25
2.63
1.04
0.36
0.82
0.20
0.21
0.32
0.61
-0.13
-1.80
-0.02
-2.33
0.62
0.32
0.54
0.40
0.90
0.08
0.98
0.03
0.54
0.75
0.59
0.69
Table 59.
Multivariable Model of Depressive Symptoms, Assessing Acculturation Effect Among High
Ethnic Enclaves, n=356
Estimate Std Error t p-value
Female
Treatment intensity
Age
Education
Age at diagnosis
Hispanic orientation
Anglo orientation
nSES
0.01
0.42
-0.09
-0.59
0.02
-0.21
-0.20
0.74
0.81
0.72
0.18
0.49
0.15
0.07
0.11
0.73
0.01
0.58
-0.50
-1.22
0.11
-3.03
-1.74
1.01
0.99
0.57
0.62
0.23
0.91
<0.001
0.09
0.32
Figure 12.
Interaction Between Ethnic Enclave and Hispanic Orientation on Depressive Symptoms
91
*Fit computed at the means of all covariates
Wellbeing
In unadjusted analyses, female gender, Hispanic and Anglo orientation, and education
were significantly positively associated with wellbeing. Hispanic orientation was marginally
significant and was negatively associated with wellbeing. Unadjusted estimates are presented in
Table 60.
In the adjusted initial model, female gender and education remained significantly
positively associated with wellbeing, while nSES was significantly negatively associated. Ethnic
enclave was significantly negatively associated with wellbeing in the full sample. Initial model
estimates are presented in Tables 61-62.
In the secondary model restricted to Hispanics, education, female gender, Hispanic and
Anglo orientation were significantly positively associated with wellbeing. Though the interaction
terms between ethnic enclave and Hispanic and Anglo orientation were not significant, in
stratified models, Hispanic and Anglo orientation were significantly protective of wellbeing
92
among high ethnic enclaves but were not significantly associated among low ethnic enclaves.
Secondary model estimates are presented in Tables 63-66. Associations between Hispanic
orientation and wellbeing, stratified by ethnic enclave presented in Figure 13.
Table 60.
Unadjusted Associations with Wellbeing (Mental Health Continuum total score)
N Estimate Std Error t p-value
Female
Hispanic
Hispanic orientation
Anglo orientation
Treatment intensity 1
Treatment intensity 2
Treatment intensity 3
Treatment intensity 4
Age
Education
Age at diagnosis
Ethnic enclave
nSES
921
899
492
489
921
921
917
921
921
921
1.96
-1.68
0.29
0.66
--
-1.90
-2.20
-4.03
0.11
2.11
0.14
1.17
0.23
0.98
1.00
0.10
0.18
--
2.11
2.04
2.31
0.10
0.37
0.09
0.73
0.34
2.00
-1.69
2.84
3.78
--
-0.90
-1.08
-1.74
1.04
5.74
1.46
1.61
0.68
0.046
0.092
0.005
0.0002
--
0.370
0.280
0.082
0.297
<0.0001
0.145
0.109
0.500
Table 61.
Multivariable Model of Wellbeing, n=896
Estimate Std Error t p-value
Ethnic enclave
Hispanic
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-3.03
-0.28
-1.06
1.50
0.05
-0.10
-0.82
2.15
1.45
1.11
0.48
0.67
0.12
0.10
0.47
0.35
-2.08
-0.25
-2.19
2.23
0.40
-0.95
-1.74
6.09
0.04
0.80
0.03
0.03
0.69
0.35
0.09
<.0001
Table 62.
Multivariable Model of Wellbeing, With Hispanic Enclave Interaction, n=896
Estimate Std Error t p-value
Ethnic enclave
Hispanic
-3.17
-0.41
1.99
1.33
-1.59
-0.31
0.12
0.76
93
Enclave*Hispanic
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
0.26
-1.06
1.50
0.05
-0.10
-0.82
2.15
1.48
0.48
0.68
0.12
0.10
0.47
0.36
0.18
-2.21
2.19
0.40
-0.95
-1.74
6.05
0.86
0.03
0.03
0.69
0.35
0.09
<.0001
Table 63.
Multivariable Model of Wellbeing, Assessing Acculturation and Enclave Main Effects (Restricted
to Hispanics), n=475
Estimate Std Error t p-value
Ethnic enclave
Hispanic orientation
Anglo orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-2.80
0.31
0.73
-0.85
2.35
0.24
-0.01
-1.01
1.15
1.60
0.10
0.11
0.69
0.86
0.16
0.15
1.28
0.55
-1.75
3.16
6.70
-1.23
2.73
1.50
-0.05
-0.78
2.09
0.09
<0.001
<.0001
0.23
0.01
0.14
0.96
0.44
0.04
Table 64.
Multivariable Model of Wellbeing, Assessing Acculturation*Enclave Interaction (Restricted to
Hispanics), n=475
Table 65.
Stratified Models of Wellbeing: Among High Ethnic Enclave (Top 2 Quintiles), n=369
Estimate Std Error t p-value
Estimate Std Error t p-value
Ethnic enclave
Hispanic orientation
Enclave*Hisp. orientation
Anglo orientation
Enclave*Anglo orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-15.35
-0.09
0.53
0.65
0.12
-0.75
1.98
0.25
-0.02
-1.08
1.15
19.15
0.29
0.30
0.43
0.53
0.68
0.87
0.15
0.16
1.35
0.56
-0.80
-0.32
1.74
1.49
0.22
-1.10
2.27
1.63
-0.10
-0.80
2.05
0.43
0.75
0.09
0.14
0.83
0.28
0.03
0.11
0.92
0.43
0.05
94
Hispanic orientation
Anglo orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
0.42
0.78
-1.10
1.85
0.27
0.03
-2.10
0.96
0.11
0.15
0.89
1.17
0.17
0.18
1.91
0.56
3.94
5.06
-1.24
1.58
1.58
0.16
-1.10
1.70
<0.001
<.0001
0.22
0.12
0.12
0.88
0.28
0.10
Table 66.
Stratified Models of Wellbeing: Among Low Ethnic Enclave (Bottom 3 Quintiles) n=106
Estimate Std Error t p-value
Hispanic orientation
Anglo orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-0.03
0.63
0.33
3.12
0.28
-0.22
2.02
1.45
0.26
0.39
0.64
2.75
0.12
0.32
0.83
0.87
-0.13
1.61
0.52
1.14
2.28
-0.70
2.43
1.67
0.90
0.12
0.61
0.27
0.03
0.49
0.02
0.11
Figure 13.
Interaction Between Ethnic Enclave and Hispanic Orientation on Wellbeing
*Fit computed at the mean of all covariates
95
Late effects
In unadjusted analyses, female gender, treatment intensity, having public insurance (vs.
uninsured), and age were significantly positively associated with late effects, while education
was significantly negatively associated. Unadjusted estimates are presented in Table 67.
In the adjusted initial model, female gender, treatment intensity, and age remained
significantly positively associated and education remained significantly negatively associated
with late effects. Hispanic ethnicity was marginally significant and was positively associated.
Neither ethnic enclave nor nSES were significant and this did not differ when exploring a
potential interaction between enclave and ethnicity. Initial model estimates are presented in
Tables 68-69.
In the secondary model restricted to Hispanics, gender, age, and treatment intensity
remained significant, and Hispanic orientation was also significantly negatively associated with
late effects, and this did not differ by ethnic enclave (b=-0.018, p=0.008). No main effect of
ethnic enclave was observed. While no main effect of Anglo orientation was observed, an
interaction with ethnic enclave was found such that Anglo orientation was significantly
negatively associated with late effects among high ethnic enclaves and not significantly
associated among low ethnic enclaves. Secondary model estimates are presented in Tables 70-73.
Table 67.
Unadjusted Associations with Total Late Effects
N Estimate Std Error t p-value
Female
Hispanic
Hispanic orientation
Anglo orientation
Treatment intensity 1
Treatment intensity 2
970
944
518
514
970
0.206
0.134
-0.011
-0.021
---
-0.015
0.083
0.083
0.009
0.015
---
0.173
2.49
1.62
-1.28
-1.37
---
-0.08
0.013
0.105
0.202
0.172
---
0.933
96
Treatment intensity 3
Treatment intensity 4
Age
Education
Age at diagnosis
Uninsured
Public
Private
Other
Ethnic enclave
nSES
970
965
970
949
970
970
0.236
0.875
0.022
-0.099
0.007
--
0.396
0.178
-0.288
-0.022
-0.010
0.167
0.189
0.009
0.031
0.008
--
0.152
0.143
0.354
0.083
0.029
1.42
4.62
2.55
-3.17
0.87
--
2.61
1.24
-0.81
-0.26
-0.33
0.156
<0.0001
0.011
0.002
0.385
--
0.009
0.215
0.417
0.793
0.743
Table 68.
Multivariable Negative Binomial Poisson Regression Model of Late Effects, n=940
Estimate Std Error Z p-value
Ethnic enclave
Hispanic
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-0.18
0.16
0.03
0.37
-0.02
0.05
0.42
-0.19
0.16
0.09
0.04
0.12
0.02
0.02
0.07
0.03
-1.16
1.67
0.64
3.13
-0.68
2.56
5.80
-5.92
0.25
0.09
0.52
<0.001
0.50
0.01
<.0001
<.0001
Table 69.
Multivariable Negative Binomial Poisson Regression Model of Late Effects, Interaction Between
Hispanic and Ethnic Enclave n=940
Estimate Std Error Z p-value
Ethnic enclave
Hispanic
Enclave*Hispanic
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-0.24
0.11
0.10
0.03
0.37
-0.01
0.05
0.42
-0.19
0.19
0.11
0.18
0.04
0.12
0.02
0.02
0.07
0.03
-1.26
1.02
0.57
0.65
3.06
-0.67
2.56
5.80
-5.97
0.21
0.31
0.57
0.52
<0.001
0.50
0.01
<.0001
<.0001
Table 70.
Multivariable Model of Late Effects, Assessing Acculturation and Enclave Main effects
97
(Restricted to Hispanics), n=497
Estimate Std. Error Z p-value
Ethnic enclave
Hispanic orientation
Anglo orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-0.18
-0.02
-0.03
0.00
0.40
-0.02
0.05
0.45
-0.06
0.17
0.01
0.02
0.07
0.15
0.03
0.02
0.08
0.05
-1.02
-2.97
-1.85
-0.06
2.66
-0.76
2.01
5.52
-1.36
0.31
0.00
0.06
0.95
0.01
0.45
0.04
<.0001
0.17
Table 71.
Multivariable Model of Late Effects, Assessing Acculturation*Enclave Interaction (Restricted to
Hispanics) Results Not in Odds Ratio Form to Provide Estimates for Interaction Terms, n=497
Estimate Std. Error Z p-value
Ethnic enclave
Hispanic orientation
Enclave*Hisp orient
Anglo orientation
Enclave*Anglo orient
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
3.67
-0.01
-0.01
0.07
-0.12
-0.01
0.39
-0.02
0.05
0.47
-0.07
1.44
0.03
0.04
0.05
0.05
0.08
0.14
0.03
0.03
0.09
0.04
2.55
-0.33
-0.39
1.33
-2.33
-0.15
2.84
-0.70
1.81
5.33
-1.58
0.01
0.74
0.70
0.18
0.02
0.88
<0.001
0.48
0.07
<.0001
0.11
Table 72.
Stratified Models of Late Effects: Among High Ethnic Enclave (Top 2 Quintiles) n=386
Estimate Std. Error T value p-value
Hispanic orientation
Anglo orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-0.02
-0.05
0.03
0.35
-0.04
0.06
0.52
-0.07
0.01
0.02
0.09
0.13
0.03
0.03
0.12
0.07
-3.28
-2.67
0.29
2.75
-1.29
2.15
4.42
-0.98
<0.001
0.01
0.77
0.01
0.20
0.03
<.0001
0.33
Table 73.
98
Stratified Models of Late Effects: Among Low Ethnic Enclave (bottom 3 quintiles) n=111
Estimate Std. Error T value p-value
Hispanic orientation
Anglo orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-0.01
0.07
-0.03
0.64
0.06
0.01
0.41
-0.07
0.03
0.05
0.13
0.33
0.02
0.02
0.16
0.10
-0.22
1.33
-0.26
1.96
2.48
0.42
2.50
-0.71
0.82
0.18
0.80
0.05
0.01
0.68
0.01
0.48
Self-rated health
In unadjusted analyses, Hispanic ethnicity, treatment intensity, age, and residence in an
ethnic enclave were significantly negatively associated with self-rated health, while Anglo
orientation, education, and nSES were significantly positively associated. Unadjusted estimates
are presented in Table 74.
In the adjusted initial model, female gender, Hispanic ethnicity, treatment intensity, and
age were significantly negatively associated with self-rated health, while education was
significantly positively associated. The effect of ethnic enclave was not significant and did not
differ by ethnicity. Initial model estimates are presented in Tables 75-76.
In the secondary model restricted to Hispanics, Anglo orientation was significantly
positively associated with self-rated health. Age and treatment intensity remained significantly
negatively associated with self-rated health. While the interaction term for Anglo orientation and
ethnic enclave was not significant, Anglo orientation was significantly associated with self-rated
health among high enclaves (b=0.03, p=0.014) but not significantly associated among low
enclaves. There was no main effect for Hispanic orientation, though the interaction between
Hispanic orientation and ethnic enclave was marginally significant (p=0.07). Among high ethnic
99
enclaves, Hispanic orientation was positively associated (p=0.01), while it was not significantly
associated among low ethnic enclaves. Secondary model estimates are presented in Tables 77-80.
Table 74.
Unadjusted Associations with Self-Rated Health
N Estimate Std Error t p-value
Female
Hispanic
Hispanic orientation
Anglo orientation
Treatment intensity 1
Treatment intensity 2
Treatment intensity 3
Treatment intensity 4
Age
Education
Age at diagnosis
Ethnic enclave
nSES
961
935
508
505
961
961
956
961
961
960
-0.070
-0.319
0.001
0.037
--
-0.158
-0.297
-0.489
-0.022
0.088
-0.008
-0.326
0.113
0.066
0.067
0.007
0.012
--
0.141
0.136
0.155
0.007
0.025
0.006
0.066
0.023
-1.05
-4.79
0.20
3.19
--
-1.13
-2.19
-3.17
-3.21
3.55
-1.27
-4.98
4.95
0.292
<0.0001
0.843
0.002
--
0.260
0.029
0.002
0.001
<0.001
0.204
<0.0001
<0.0001
Table 75.
Multivariable Model of Self-Rated Health, n=931
Estimate Std. Error T value p-value
Female
Treatment intensity
Age
Education
Age at diagnosis
Ethnic enclave
Hispanic
nSES
-0.10
-0.11
-0.04
0.09
0.01
-0.14
-0.13
0.04
0.05
0.05
0.01
0.03
0.01
0.11
0.06
0.04
-2.09
-2.43
-6.99
3.05
1.63
-1.20
-2.25
0.81
0.04
0.02
<.0001
<0.001
0.11
0.23
0.03
0.42
Table 76.
Multivariable Model of Self-Rated Health, Interaction Between Ethnicity and Enclave, n=931
Estimate Std. Error T value p-value
Female
Treatment intensity
Age
Education
Age at diagnosis
-0.11
-0.11
-0.04
0.09
0.01
0.05
0.05
0.01
0.03
0.01
-2.17
-2.41
-7.10
3.06
1.62
0.03
0.02
<.0001
<0.001
0.11
100
Ethnic enclave
Hispanic
Hispanic*Enclave
nSES
-0.20
-0.19
0.12
0.04
0.15
0.09
0.11
0.04
-1.34
-2.19
1.12
0.84
0.19
0.03
0.27
0.40
Table 77.
Multivariable Model of Self-Rated Health, Assessing Acculturation and Enclave Main effects
(Restricted to Hispanics), n=488
Estimate Std. Error T value p-value
Ethnic enclave
Hispanic orientation
Anglo orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
-0.02
0.01
0.03
0.07
-0.11
0.03
-0.04
-0.14
0.01
0 0.16
0.00
0.01
0.07
0.09
0.01
0.01
0.06
0.05
-0.10
1.41
2.90
1.09
-1.26
3.15
-4.94
-2.15
0.26
0.92
0.17
0.01
0.28
0.21
<0.001
<.0001
0.04
0.80
Table 78.
Multivariable Model of Self-Rated Health, Assessing Acculturation*Enclave Interaction
(Restricted to Hispanics), n=488
Estimate Std. Error T value p-value
Ethnic enclave
Hispanic orientation
Enclave*Hisp. orient
Anglo orientation
Enclave*Anglo orient
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
0.05
-0.02
0.03
0.05
-0.02
0.08
-0.13
0.03
-0.04
-0.14
0.01
0.98
0.02
0.02
0.03
0.03
0.06
0.09
0.01
0.01
0.07
0.05
0.05
-1.20
1.85
1.98
-0.83
1.24
-1.56
3.28
-4.57
-2.18
0.22
0.96
0.24
0.07
0.05
0.41
0.22
0.13
<0.001
<.0001
0.04
0.83
Table 79.
Stratified Models of Self-Rated Health: Among High Ethnic Enclave (Top 2 Quintiles) n=379
Estimate Std. Error T value p-value
Anglo orientation
Hispanic orientation
nSES
0.03
0.02
0.06
0.01
0.01
0.10
3.16
2.84
0.61
<0.001
0.01
0.55
101
Female
Age at diagnosis
Age
Treatment intensity
Education
-0.18
0.03
-0.04
-0.12
-0.03
0.10
0.01
0.01
0.07
0.06
-1.85
3.18
-3.62
-1.66
-0.56
0.07
<0.001
<0.001
0.11
0.58
Table 80.
Stratified Models of Self-Rated Health: Among Low Ethnic Enclave (Bottom 3 Quintiles) n=109
Estimate Std. Error T value p-value
Anglo orientation
Hispanic orientation
nSES
Female
Age at diagnosis
Age
Treatment intensity
Education
0.03
-0.02
0.12
0.11
0.03
-0.06
-0.20
0.17
0.03
0.01
0.07
0.15
0.01
0.02
0.07
0.05
1.20
-1.51
1.68
0.76
2.32
-3.22
-2.71
3.30
0.24
0.14
0.11
0.46
0.03
<0.001
0.01
<0.001
Sensitivity analyses
As a sensitivity analysis, models were run with alternative coding of ethnic enclave: 5-
category continuous, dichotomized with top 3 vs. bottom 2 quintiles, dichotomized with top 2 vs.
bottom 3 quintiles, and categorical with bottom 2, vs. middle, vs. top 2 quintiles. Outcomes
appear to mainly differ across the top 2 vs. bottom 3 quintiles, this coding also creates a
somewhat even distribution given the skewed nature of residence in an ethnic enclave among
Hispanics, and is also consistent with the coding used in previous literature.(N. Gomez et al.,
2015)
As an alternative way of simultaneously accounting for ethnic enclave, nSES, and
individual acculturation while preserving the entire sample, a 6-category variable was also
explored which compared high vs. low acculturated Hispanics residing in high vs. low Hispanic
enclaves, and non-Hispanics residing in high vs. low Hispanic enclaves. With this coding, some
results were difficult to interpret. So for ease of interpretation, acculturation was assessed only in
102
Hispanic models and models were subsequently stratified by enclave to assess effect
modification, while controlling for nSES.
While nSES and ethnic enclave are correlated (r = -0.81, p<0.0001), nSES operates as a
suppressor of ethnic enclave for the wellbeing outcome. That is, without controlling for nSES,
the effect of ethnic enclave is not significant, whereas its coefficient is larger and statistically
significant with the inclusion of nSES in the model. This means that when not controlled for, the
variance in ethnic enclave explained by nSES is left unexplained in the model, reducing its
relationship with the outcome.(Cohen, West, & Aiken, 2014) The role of nSES as a suppressor
variable in the wellbeing models is further supported by the fact that the bivariate association
between nSES and wellbeing is weak (p=0.50), though it’s correlation with ethnic enclave is
highly significant as previously stated. This indicates that nSES explains additional variance in
ethnic enclave, clarifying its effect on wellbeing when controlled for, though nSES does not have
a direct effect on the outcome itself.(Friedman & Wall, 2005; MacKinnon, Krull, & Lockwood,
2000) For self-rated health, the effect of ethnic enclave is not significant with the inclusion of
nSES, and becomes significant after its exclusion. This is because nSES is highly correlated with
self-rated health (p<0.0001), as well as ethnic enclave, so excluding it as a covariate allows the
estimate for ethnic enclave to reflect the variance attributed to nSES. Thus, the more
conservative choice is to retain nSES so that neighborhood SES effects are not erroneously
attributed to ethnic enclave. For all remaining outcomes, the inclusion of nSES had no impact on
the magnitude or significance of the effect of ethnic enclave. Thus, nSES was retained as a
covariate in all models. The variance inflation factor for nSES ranged from 1.99-2.05 across
outcomes, well under the limit of 5 some have suggested as a threshold to consider exclusion or
103
alternative treatment of a collinear variable.(Akinwande, Dikko, & Samson, 2015; Craney &
Surles, 2002)
For analysis of depressive symptoms, models were additionally run using a dichotomous
indicator of depression as the outcome based on the CESD cut point of 16 for clinical depression.
Results did not meaningfully differ so the continuous summary score was used as the outcome in
all models of depressive symptoms.
Discussion
This study demonstrates the significant impact of concordance between an individual’s
cultural orientation and that of their neighborhood on a range of health outcomes. For Hispanics
residing in an ethnic enclave, higher Hispanic orientation is positively associated with mental
health and wellbeing as well as healthcare utilization.
Receipt of cancer-related follow up care decreased with age and was positively associated
with age at diagnosis, female gender, and having insurance, consistent with previous
literature.(Milam et al., 2015; Nathan et al., 2008) Among Hispanics, a significant main effect of
Hispanic orientation on follow up care was observed. The interaction of Hispanic orientation and
ethnic enclave was not significant, though stratified models revealed that Hispanic orientation
was significantly positively associated with receipt of recent follow up care for those residing in
high enclaves but was not significant among those living in low enclaves. However, the lack of
significance of Hispanic orientation among low ethnic enclaves was likely due to sample size
limitations. Just 111 Hispanics who responded to the responded to the cancer follow up care item
in the survey resided in low ethnic enclaves, compared to 378 living in high ethnic enclaves. The
odds ratio for Hispanic orientation was the same in each of the stratified models, with a wider
104
confidence interval that crossed zero in the low enclave model, explaining the lack of
significance for the Hispanic orientation term. Thus, this cultural orientation appears to be
protective of follow up care regardless of the cultural makeup of one’s neighborhood.
Hispanic participants who retained an orientation toward the Hispanic culture appear to
fare better in terms of maintenance of healthcare. This may be due to Hispanic socio-cultural
values such as familismo, the importance and prioritization of family. Some evidence suggests
that this commitment to family may guide health-related decision making and health
behaviors.(Davila, Reifsnider, & Pecina, 2011) CCS who are more closely connected with their
family may have responsive support systems and resources that aid in healthcare
maintenance.(Katiria Perez & Cruess, 2014; Teran, Baezconde-Garbanati, Marquez, Castellanos,
& BELKIĆ, 2007; Yamada et al., 2009)
For depressive symptoms, education was the only significant correlate among the full
sample, a well-documented association. Education was an individual-level proxy for SES
included in our models, and there is ample prior evidence that SES is associated with mental
health.(Ashing-Giwa & Lim, 2009; Reiss, 2013) Among Hispanics, a significant negative main
effect of Hispanic orientation on depressive symptoms was observed. Although the interaction
term between Hispanic orientation and ethnic enclave was not significant, it was protective of
mental health among high enclaves but were not significantly associated in low ethnic enclaves.
For wellbeing, female gender and education were significantly positively associated, while nSES
was significantly negatively associated among the full sample. Among Hispanics, a positive
main effect of both Hispanic and Anglo orientation on wellbeing was observed. As with
depressive symptoms, the interaction terms between Hispanic and Anglo orientation and ethnic
105
enclave were not significant for wellbeing, though both appear to be protective of mental health
in high enclaves but not significantly associated in low ethnic enclaves.
The lack of significance of the interaction term for cultural orientation and ethnic enclave
may have been a result of sample size. The effect of Hispanic orientation does appear to differ
substantially across levels of ethnic enclave when plotting marginal effects. However, just 106
and 104 Hispanics who responded to the wellbeing and depressive symptoms items, respectively,
resided in a low ethnic enclave neighborhood, whereas 369 and 356 resided in a high ethnic
enclave. As depicted in Figures 1, the association among low ethnic enclaves appears positive,
but the width of the confidence limits renders the findings in this subgroup difficult to interpret.
Conversely, among high enclaves the association is negative and significant. The low sample
size among low ethnic enclaves resulted in insufficient power to detect a possible effect of
Hispanic orientation in the low enclave subgroup, which may differ in direction from that of the
high enclave subgroup, or the association may simply be non-significant among low enclaves but
significant among high enclaves. However, the decision to explore ethnic enclave as a modifier
of the effect of individual acculturation on health outcomes was made a priori, and the theoretical
plausibility of the findings and the consistency of the results across correlated outcomes
(depressive symptoms and wellbeing) provides support for interpreting the interaction between
Hispanic orientation and residence in an ethnic enclave despite the lack of significance of the
interaction term.(Sun, Briel, Walter, & Guyatt, 2010)
The observed protective effect of Hispanic orientation for those residing in high ethnic
enclaves has been observed previously, though typically measured simply by ethnic density
rather than by a comprehensive measure of ethnic enclaves.(Gerst et al., 2011) This association
may be explained by several factors. First, concordance between one’s cultural orientation and
106
that of one’s neighborhood may signify a greater sense of social cohesion through shared
customs and values.(Stafford, Bécares, & Nazroo, 2010) The ‘ethnic density hypothesis’ posits
that minority groups experience better mental health when living in areas with greater
concentrations of others of the same ethnicity, theoretically due to social support and higher self-
evaluation, even in spite of the adverse impact of material disadvantage often characteristic areas
of higher ethnic density.(R. J. Shaw et al., 2012) However, these causal mechanisms are
theoretical with few studies addressing them directly. One study did find that the inclusion of
social support attenuated the association between ethnic density and mental health, however
another study found that social support was not associated with neighborhood composition.(Das-
Munshi, Becares, Dewey, Stansfeld, & Prince, 2010; Vogt Yuan, 2007) A longitudinal study
found that neighborhood ethnic density was negatively associated with mental health and was
partially mediated by social cohesion. However the association between ethnic density and social
cohesion was moderated by race/ethnicity, such that is was negative among Asians but positive
among Latinos.(Hong, Zhang, & Walton, 2014) Another study found that the association
between ethnic density and social capital health differed across ethnic groups, reflecting
structural differences in neighborhoods and their effects on the health of individual
residents.(Bécares & Nazroo, 2015) These mixed findings highlight the need to further explore
theories of social relationships that account for complex multilevel (individual- and
neighborhood-level) mechanisms and moderating factors. Inclusion of measures of social capital
such as neighborhood trust and social networks would improve understanding of the mechanisms
linking ethnic density and mental health and wellbeing.
Improved mental health outcomes in the context of cultural concordance with one’s
neighborhood may also reflect reduced exposure to health risk factors such as cultural
107
intolerance and discrimination.(Arévalo, Tucker, & Falcón, 2015) Previous research has
identified that states with increased racial heterogeneity experienced higher mortality rates for all
races, possibly due to manifested racial antagonism which negatively affects social
cohesion.(Reidpath, 2003) The same has been observed in studies of ethnic and cultural
homogeneity, whereby areas that are more ethnically and culturally homogenous experience
protective health benefits due to reduced intercultural tensions.(Stafford et al., 2010) Thus the
harmony that may be experienced in ethnic enclaves due to the absence of ethnic antagonism
may improve social capital, imparting a number of health benefits to residents.
As a measure of physical health status, late effects were explored. Female gender, current
age, and treatment intensity were positively associated, while education was negatively
associated. Among Hispanics, Hispanic orientation was significantly negatively associated with
late effects and did not differ by ethnic enclave, while Anglo orientation was also found to be
protective of late effects only among ethnic enclaves. Individual acculturation may operate
through similar mechanisms previously discussed in regards to the findings on follow up care,
such as Hispanic cultural values that may be health promoting. The positive association between
Anglo orientation and late effects among high ethnic enclaves may reflect that improved health
literacy and understanding of the health system in the US is supportive of preventive health
behaviors, despite living in an area with potentially reduced healthcare resources. Those more
acculturated to the US are less likely to face linguistic and other healthcare barriers, which
affects the use of preventive health services as well as knowledge of lifestyle factors that affect
health outcomes.(Britigan, Murnan, & Rojas-Guyler, 2009; Katiria Perez & Cruess, 2014)
Finally, self-rated health was explored as an additional measure of health status. Overall,
females, Hispanics, and those with greater treatment intensity, and current age reported poorer
108
self-rated health, while those with higher education reported greater health. Among Hispanics,
Anglo orientation was significantly associated with greater self-rated health. The interaction term
for Anglo orientation and ethnic enclave was not significant, and although Anglo orientation was
significantly associated with self-rated health among high enclaves but not significantly
associated among low enclaves, the estimate was the same in both stratified models, suggesting
that the effect did not differ across enclave levels. The lack of significance of Anglo orientation
in the low enclave model was likely a sample size issue, given that just 109 Hispanics who
reported self-rated health resided in low ethnic enclaves. The positive association between self-
rated health and Anglo orientation likely reflects the same protective mechanisms as discussed in
relation to late effects. While there was no main effect of Hispanic acculturation among
Hispanics, its interaction with ethnic enclave was marginally significant (p=0.07). Indeed, among
high ethnic enclaves, Hispanic orientation was significantly positively related to self-rated
health, while it was non-significant in low ethnic enclaves. As has been previously discussed,
those who identify with the Hispanic culture are more likely to benefit from the health protective
aspects of Hispanic ethnic enclaves, such as social support.
Limitations and strengths
It can be difficult to disentangle individual-level effects from contextual effects, given
that individuals with shared characteristics are more likely to self-select into distinct types of
neighborhoods. Some supposed neighborhood effects are attributable simply to the fact that they
are comprised of similar types of people, which may be referred to as “compositional effects.”
However, a strength of this study was the ability to account for both individual- and
neighborhood-level factors, including acculturation, to estimate neighborhood effects net of
109
compositional effects. SES was also controlled for at the individual- and neighborhood-levels to
account for distinct effects. This study also benefited from the use of self-identified Hispanic
ethnicity. Registry-based studies without patient response data suffer from the potential
misclassification of ethnicity, which can bias results.
An additional limitation includes the lack of inclusion of nativity in multivariable models
and the lack of information about pre-migration factors for those who were foreign-born. While
nativity was captured in our survey, accounting for this in multivariable models would have
resulted in small cell sizes, so this variable was omitted. Information about pre-migration
conditions may also have influenced results, given that psychosocial adaptation to a new
environment is partially determined by the conditions that motivated the immigration.(Arévalo et
al., 2015)
Generalizability of these findings is also limited by the inability to properly weight
analyses for survey non-response. Given the unavailability of confirmed current residential
address for those who were eligible but did not participate in our survey, we were not able to
create survey weights for the analytic subsample in this study, so results may not be
generalizable to the entire CCS population in Los Angeles County given that certain
demographic subgroups were more likely to participate in our study. However, this study
provided good representation of Hispanic CCS, who have not been the focus of the majority of
prior survivorship research, despite the fact that they are a growing population that exhibits
poorer incidence and survival rates of certain pediatric cancers, relative to non-Hispanics whites.
The ability of this study to recruit a diverse sample from the state with the largest number of
Hispanics in the country, while also characterizing acculturation, represents an important
110
contribution to the field of cancer disparities research.(Pew Research Center: Ranking Latino
populations in the states, 2016)
Conclusion
Individual-level cultural orientation interacts with neighborhood-level acculturation
(ethnic enclaves) to impact a range of health outcomes among Hispanic CCS. Cultural
concordance with one’s neighborhood may impart health protective effects. Mechanisms
underlying this protective effect should be explored in future work, such as increased social
capital and reduced discrimination. Such health protective factors could be the target of
community-level interventions aimed at improving health outcomes in culturally heterogeneous
neighborhoods.
111
Conclusion
Survivors of pediatric cancer are a unique population whose health is determined by
numerous clinical, sociodemographic, developmental, and contextual factors. While prior
research has focused predominantly on clinical and demographic correlates of health outcomes
among CCS, contextual factors have not been as widely addressed in this population. This
dissertation addresses this gap by characterizing residential trajectories and the social
environment in association with a range of health outcomes among a diverse sample of CCS
diagnosed in Los Angeles County.
By addressing neighborhood cultural context and its interaction with individual acculturation,
this work highlights sources of heterogeneity in outcomes among Hispanic CCS, an under-
researched group despite experiencing disproportionately high incidence and mortality rates of
some pediatric cancers.(Liu et al., 2010) Results from study 2 indicate that both individual
acculturation and neighborhood ethnic enclave are associated with CCS health, and may interact
in determining health outcomes. While sample sizes were limited to fully characterize
associations between acculturation and health among Hispanics residing in low ethnic enclaves,
greater Hispanic orientation appears protective of health among those residing in high ethnic
enclaves. Those lacking a cultural identity may be at greater risk, possibly due to a lack of
protective cultural values or reduced social support. This finding may have implications for
identification of at-risk CCS. First, cultural differences must be acknowledged, as ethnic groups
are not homogenous. Culturally/linguistically tailored survivorship education services and
materials should be available to optimally reach minority CCS. Second, survivorship clinics
could aim to provide targeted support services to CCS who report a lack family or community
support.
112
In study 1, analyses of mobility and cumulative nSES revealed differences in self-rated
health between those who resided in high SES neighborhoods at diagnosis as well as at survey
compared to those who resided in low SES neighborhoods at diagnosis as well as at survey.
Notably, those who were mobile, either upwardly or downwardly, did not differ significantly
from the stable low SES individuals. This suggests that residence in a disadvantaged
neighborhood at any point whether in childhood or adulthood, even if temporary, may impart
risk for long term health. Said differently, moving out of a disadvantaged neighborhood did not
appear to result in significantly improved adult self-rated health in our sample. This reinforces
the importance of ensuring that survivors from, or currently residing in, disadvantaged areas
have adequate access to education and resources supportive of health.
By focusing on cultural and contextual factors associated with the health of young adult
CCS, these insights may guide further investigation into multilevel factors supportive of CCS
outcomes as well as inform the focus of support services aimed at retaining survivors in follow
up care and reducing disparities in lifelong health and wellbeing.
113
References
Akinwande, M. O., Dikko, H. G., & Samson, A. (2015). Variance inflation factor: as a condition
for the inclusion of suppressor variable (s) in regression analysis. Open Journal of
Statistics, 5(07), 754.
Arévalo, S. P., Tucker, K. L., & Falcón, L. M. (2015). Beyond cultural factors to understand
immigrant mental health: neighborhood ethnic density and the moderating role of pre-
migration and post-migration factors. Social Science & Medicine, 138, 91-100.
Armstrong, G. T., Chen, Y., Yasui, Y., Leisenring, W., Gibson, T. M., Mertens, A. C., . . . Krull,
K. R. (2016). Reduction in late mortality among 5-year survivors of childhood cancer.
New England Journal of Medicine, 374(9), 833-842.
Armstrong, G. T., Liu, Q., Yasui, Y., Neglia, J. P., Leisenring, W., Robison, L. L., & Mertens, A.
C. (2009). Late mortality among 5-year survivors of childhood cancer: a summary from
the Childhood Cancer Survivor Study. Journal of Clinical Oncology, 27(14), 2328.
Arnett, J. J. (2000). Emerging adulthood: A theory of development from the late teens through
the twenties. American Psychologist, 55(5), 469-480. doi:10.1037//0003-066X.55.5.469
Ashing-Giwa, K. T., & Lim, J.-w. (2009). Examining the impact of socioeconomic status and
socioecologic stress on physical and mental health quality of life among breast cancer
survivors. Paper presented at the Oncology Nursing Forum.
Béatrice, N., Lise, G., Victoria, Z. M., & Louise, S. (2012). Longitudinal patterns of poverty and
health in early childhood: exploring the influence of concurrent, previous, and cumulative
poverty on child health outcomes. BMC pediatrics, 12(1), 141.
Bécares, L., & Nazroo, J. (2015). Social capital, ethnic density and mental health among ethnic
minority people in England: a mixed-methods study. In Handbook of Research Methods
and Applications in Social Capital: Edward Elgar Publishing.
Ben-Shlomo, Y., & Kuh, D. (2002). A life course approach to chronic disease epidemiology:
conceptual models, empirical challenges and interdisciplinary perspectives. In: Oxford
University Press.
Berry, J. W. (1997). Immigration, Acculturation, and Adaptation. Applied Psychology, 46(1), 5-
34. doi:10.1111/j.1464-0597.1997.tb01087.x
Bhatia, S., & Meadows, A. T. (2006). Long‐term follow‐up of childhood cancer survivors: future
directions for clinical care and research. Pediatric blood & cancer, 46(2), 143-148.
Blalock Jr, H. M. (1966). The identification problem and theory building: the case of status
inconsistency. American sociological review, 52-61.
Braveman, P., & Barclay, C. (2009). Health disparities beginning in childhood: a life-course
perspective. Pediatrics, 124(Supplement 3), S163-S175.
Britigan, D. H., Murnan, J., & Rojas-Guyler, L. (2009). A qualitative study examining Latino
functional health literacy levels and sources of health information. Journal of Community
Health, 34(3), 222-230.
Busacker, A., & Kasehagen, L. (2012). Association of Residential Mobility with Child Health:
An Analysis of the 2007 National Survey of Children’s Health. Maternal and Child
Health Journal, 16(Supplement 1), 78-87. doi:10.1007/s10995-012-0997-8
Bustamante, A. V., Chen, J., Rodriguez, H. P., Rizzo, J. A., & Ortega, A. N. (2010). Use of
preventive care services among Latino subgroups. American Journal of Preventive
Medicine, 38(6), 610-619.
114
Chang, E., & Chan, K. S. (2015). Variations in Asian Americans: How Neighborhood
Concordance Is Associated With Health Care Access and Utilization. American Journal
of Public Health, 105(1), 66-68. doi:10.2105/AJPH.2014.302275
Cheng, E. M., Chen, A., & Cunningham, W. (2007). Primary language and receipt of
recommended health care among Hispanics in the United States. Journal of General
Internal Medicine, 22(2), 283-288.
Cockburn, M., Mills, P., Zhang, X., Zadnick, J., Goldberg, D., & Ritz, B. (2011). Prostate cancer
and ambient pesticide exposure in agriculturally intensive areas in California. American
Journal of Epidemiology, 173(11), 1280-1288.
Cohen, P., West, S. G., & Aiken, L. S. (2014). Applied multiple regression/correlation analysis
for the behavioral sciences: Psychology Press.
Conroy, K., Sandel, M., & Zuckerman, B. (2010). Poverty grown up: how childhood
socioeconomic status impacts adult health. Journal of Developmental & Behavioral
Pediatrics, 31(2), 154-160.
Council, N. R. (2003). Childhood cancer survivorship: improving care and quality of life:
National Academies Press.
Craney, T. A., & Surles, J. G. (2002). Model-dependent variance inflation factor cutoff values.
Quality Engineering, 14(3), 391-403.
Cuellar, I., Arnold, B., & Maldonado, R. (1995). Acculturation Rating Scale for Mexican
Americans-II: A Revision of the Original ARSMA Scale. Hispanic Journal of Behavioral
Sciences, 17(3), 275-304. doi:10.1177/07399863950173001
Das-Munshi, J., Becares, L., Dewey, M. E., Stansfeld, S. A., & Prince, M. J. (2010).
Understanding the effect of ethnic density on mental health: multi-level investigation of
survey data from England. Bmj, 341, c5367.
Davila, Y. R., Reifsnider, E., & Pecina, I. (2011). Familismo: influence on Hispanic health
behaviors. Applied Nursing Research, 24(4), e67-e72.
DeVoe, J. E., Fryer, G. E., Phillips, R., & Green, L. (2003). Receipt of preventive care among
adults: insurance status and usual source of care. American Journal of Public Health,
93(5), 786-791.
Diez Roux, A. V., & Mair, C. (2010). Neighborhoods and health. Annals of the New York
Academy of Sciences, 11861(1), 125-145. doi:10.1111/j.1749-6632.2009.05333.x
Do, D. P., Wang, L., & Elliott, M. R. (2013). Investigating the relationship between
neighborhood poverty and mortality risk: a marginal structural modeling approach. Social
Science & Medicine, 91, 58-66.
Dowd, J. B., & Zajacova, A. (2010). Does self-rated health mean the same thing across
socioeconomic groups? Evidence from biomarker data. Annals of epidemiology, 20(10),
743-749.
DuBard, C. A., & Gizlice, Z. (2008). Language spoken and differences in health status, access to
care, and receipt of preventive services among US Hispanics. American Journal of Public
Health, 98(11), 2021-2028.
Echeverria, S. E., & Carrasquillo, O. (2006). The roles of citizenship status, acculturation, and
health insurance in breast and cervical cancer screening among immigrant women.
Medical Care, 44(8), 788-792.
Evans, G. W., & Kim, P. (2007). Childhood Poverty and Health. Psychological Science, 18(11),
953-957. doi:10.1111/j.1467-9280.2007.02008.x
115
Fidler, M. M., Ziff, O. J., Wang, S., Cave, J., Janardhanan, P., Winter, D. L., . . . Frobisher, C.
(2015). Aspects of mental health dysfunction among survivors of childhood cancer. Br J
Cancer, 113(7), 1121.
Finch, B. K., Hummer, R. A., Kol, B., & Vega, W. A. (2001). The role of discrimination and
acculturative stress in the physical health of Mexican-origin adults. Hispanic Journal of
Behavioral Sciences, 23(4), 399-429.
Finch, B. K., & Vega, W. A. (2003). Acculturation stress, social support, and self-rated health
among Latinos in California. Journal of immigrant health, 5(3), 109-117.
Flores, E., Tschann, J. M., Dimas, J. M., Bachen, E. A., Pasch, L. A., & de Groat, C. L. (2008).
Perceived discrimination, perceived stress, and mental and physical health among
Mexican-origin adults. Hispanic Journal of Behavioral Sciences, 30(4), 401-424.
Friedman, L., & Wall, M. (2005). Graphical views of suppression and multicollinearity in
multiple linear regression. The American Statistician, 59(2), 127-136.
Gaskin, D. J., Dinwiddie, G. Y., Chan, K. S., & McCleary, R. (2012). Residential segregation
and disparities in health care services utilization. Medical Care Research and Review,
69(2), 158-175.
Gerst, K., Miranda, P. Y., Eschbach, K., Sheffield, K. M., Peek, M. K., & Markides, K. S.
(2011). Protective neighborhoods: neighborhood proportion of Mexican Americans and
depressive symptoms in very old Mexican Americans. Journal of the American
Geriatrics Society, 59(2), 353-358.
Glymour, M. M., Mujahid, M., Wu, Q., White, K., & Tchetgen, E. J. T. (2010). Neighborhood
disadvantage and self-assessed health, disability, and depressive symptoms: longitudinal
results from the health and retirement study. Annals of epidemiology, 20(11), 856-861.
Gomez, N., Guendelman, S., Harley, K. G., & Gomez, S. L. (2015). Nativity and neighborhood
characteristics and cervical cancer stage at diagnosis and survival outcomes among
Hispanic women in California. American Journal of Public Health, 105(3), 538-545.
doi:10.2105/AJPH.2014.302261
Gomez, S. L., Glaser, S. L., McClure, L. A., Shema, S. J., Kealey, M., Theresa, H. M. K., &
Satariano, W. A. (2011). The California Neighborhoods Data System: a new resource for
examining the impact of neighborhood characteristics on cancer incidence and outcomes
in populations. Cancer Causes & Control, 22(4), 631-647. doi:10.1007/s10552-011-
9736-5
Gomez, S. L., Shariff-Marco, S., DeRouen, M., Keegan, T. H. M., Yen, I. H., Mujahid, M., . . .
Glaser, S. L. (2015). The impact of neighborhood social and built environment factors
across the cancer continuum: Current research, methodological considerations, and future
directions: Neighborhoods and the Cancer Continuum. Cancer, 121(14), 2314-2330.
doi:10.1002/cncr.29345
Gonzales, N. A., Deardorff, J., Formoso, D., Barr, A., & Barrera, M. (2006). Family mediators of
the relation between acculturation and adolescent mental health. Family Relations, 55(3),
318-330.
Gordon-Larsen, P., Nelson, M. C., Page, P., & Popkin, B. M. (2006). Inequality in the built
environment underlies key health disparities in physical activity and obesity. Pediatrics,
117(2), 417-424.
Gupta, S., Wilejto, M., Pole, J. D., Guttmann, A., & Sung, L. (2014). Low socioeconomic status
is associated with worse survival in children with cancer: a systematic review. PloS One,
9(2), e89482.
116
Gurney, J. G., Krull, K. R., Kadan-Lottick, N., Nicholson, H. S., Nathan, P. C., Zebrack, B., . . .
Ness, K. K. (2009). Social outcomes in the childhood cancer survivor study cohort.
Journal of Clinical Oncology, 27(14), 2390.
Haas, J. S., Phillips, K. A., Sonneborn, D., McCulloch, C. E., Baker, L. C., Kaplan, C. P., . . .
Liang, S.-Y. (2004). Variation in access to health care for different racial/ethnic groups
by the racial/ethnic composition of an individual's county of residence. Medical Care,
42(7), 707-714.
Hamilton, A. S., Zhuang, X., Modjeski, D., Slaughter, R., Ritt-Olson, A., & Milam, J. (2019).
Population-based survey methods for reaching adolescent and young adult survivors of
pediatric cancer and their parents. Journal of Adolescent and Young Adult Oncology,
8(1), 40-48.
Henderson, T. O., Friedman, D. L., & Meadows, A. T. (2010). Childhood cancer survivors:
transition to adult-focused risk-based care. Pediatrics, 126(1), 129-136.
doi:10.1542/peds.2009-2802
Hendryx, M. S., Ahern, M. M., Lovrich, N. P., & McCurdy, A. H. (2002). Access to health care
and community social capital. Health services research, 37(1), 85.
Hertzman, C., & Boyce, T. (2010). How experience gets under the skin to create gradients in
developmental health. Annual Review of Public Health, 31, 329-347.
Hertzman, C., & Power, C. (2003). Health and human development: understandings from life-
course research. Developmental neuropsychology, 24(2-3), 719-744.
Hong, S., Zhang, W., & Walton, E. (2014). Neighborhoods and mental health: exploring ethnic
density, poverty, and social cohesion among Asian Americans and Latinos. Social
Science & Medicine, 111, 117-124.
Houle, J. N., & Martin, M. A. (2011). Does intergenerational mobility shape psychological
distress? Sorokin revisited. Research in social stratification and mobility, 29(2), 193-203.
Hudson, D. L., Puterman, E., Bibbins-Domingo, K., Matthews, K. A., & Adler, N. E. (2013).
Race, life course socioeconomic position, racial discrimination, depressive symptoms and
self-rated health. Social Science & Medicine, 97, 7-14.
Hudson, M. M., Mertens, A. C., Yasui, Y., Hobbie, W., Chen, H., Gurney, J. G., . . . Robison, L.
R. (2003). Health status of adult long-term survivors of childhood cancer: a report from
the Childhood Cancer Survivor Study. JAMA, 290(12), 1583-1592.
Hussein, M., Diez Roux, A. V., & Field, R. I. (2016). Neighborhood Socioeconomic Status and
Primary Health Care: Usual Points of Access and Temporal Trends in a Major US Urban
Area. Journal of Urban Health, 93(6), 1027-1045. doi:10.1007/s11524-016-0085-2
Jacobs, L. A., & Pucci, D. A. (2013). Adult Survivors of Childhood Cancer: The Medical and
Psychosocial Late Effects of Cancer Treatment and the Impact on Sexual and
Reproductive Health. The Journal of Sexual Medicine, 10, 120-126.
doi:10.1111/jsm.12050
Johnson, K. L., Carroll, J. F., Fulda, K. G., Cardarelli, K., & Cardarelli, R. (2010). Acculturation
and self-reported health among Hispanics using a socio-behavioral model: the North
Texas Healthy Heart Study. BMC public health, 10(1), 53.
Jun, J., & Oh, K. M. (2013). Asian and Hispanic Americans' cancer fatalism and colon cancer
screening. American Journal of Health Behavior, 37(2), 145-154.
Jurkowski, J. M., & Johnson, T. P. (2005). Acculturation and cardiovascular disease screening
practices among Mexican Americans living in Chicago. Ethnicity and Disease, 15(3),
411.
117
Kadan-Lottick, N. S., Ness, K. K., Bhatia, S., & Gurney, J. G. (2003). Survival variability by
race and ethnicity in childhood acute lymphoblastic leukemia. JAMA, 290(15), 2008-
2014.
Kahn, J. M., Keegan, T. H., Tao, L., Abrahão, R., Bleyer, A., & Viny, A. D. (2016). Racial
disparities in the survival of American children, adolescents, and young adults with acute
lymphoblastic leukemia, acute myelogenous leukemia, and Hodgkin lymphoma. Cancer,
122(17), 2723-2730.
Katiria Perez, G., & Cruess, D. (2014). The impact of familism on physical and mental health
among Hispanics in the United States. Health psychology review, 8(1), 95-127.
doi:10.1080/17437199.2011.569936
Kazak, A. E., Hocking, M. C., Ittenbach, R. F., Meadows, A. T., Hobbie, W., DeRosa, B. W., . . .
Reilly, A. (2012). A revision of the intensity of treatment rating scale: classifying the
intensity of pediatric cancer treatment. Pediatric blood & cancer, 59(1), 96-99.
Keegan, T. H., John, E. M., Fish, K. M., Alfaro-Velcamp, T., Clarke, C. A., & Gomez, S. L.
(2010). Breast cancer incidence patterns among California Hispanic women: differences
by nativity and residence in an enclave. Cancer Epidemiology and Prevention
Biomarkers, 19(5), 1208-1218.
Kinsler, J. J., Lee, S.-J., Sayles, J. N., Newman, P. A., Diamant, A., & Cunningham, W. (2009).
The impact of acculturation on utilization of HIV prevention services and access to care
among an at-risk Hispanic population. Journal of health care for the poor and
underserved, 20(4), 996.
Ko, M., & Ponce, N. A. (2013). Community residential segregation and the local supply of
federally qualified health centers. Health services research, 48(1), 253-270.
Kremer, L., Mulder, R. L., Oeffinger, K. C., Bhatia, S., Landier, W., Levitt, G., . . . Armenian, S.
H. (2013). A worldwide collaboration to harmonize guidelines for the long‐term follow‐
up of childhood and young adult cancer survivors: A report from the international late
effects of Childhood Cancer Guideline Harmonization Group. Pediatric blood & cancer,
60(4), 543-549.
Kuh, D., Ben-Shlomo, Y., Lynch, J., Hallqvist, J., & Power, C. (2003). Life course
epidemiology. Journal of Epidemiology & Community Health, 57(10), 778-783.
Kuh, D., & Shlomo, Y. B. (2004). A life course approach to chronic disease epidemiology:
Oxford University Press.
Lamers, S., Westerhof, G. J., Bohlmeijer, E. T., ten Klooster, P. M., & Keyes, C. L. (2011).
Evaluating the psychometric properties of the mental health continuum‐short form
(MHC‐SF). Journal of Clinical Psychology, 67(1), 99-110.
Landier, W., Bhatia, S., Eshelman, D. A., Forte, K. J., Sweeney, T., Hester, A. L., . . . Constine,
L. S. (2004). Development of risk-based guidelines for pediatric cancer survivors: the
Children's Oncology Group Long-term Follow-up Guidelines from the Children's
Oncology Group Late Effects Committee and Nursing Discipline. Journal of Clinical
Oncology, 22(24), 4979-4990.
Langeveld, N., Stam, H., Grootenhuis, M., & Last, B. (2002). Quality of life in young adult
survivors of childhood cancer. Supportive Care in Cancer, 10(8), 579-600.
Layes, A., Asada, Y., & Kephart, G. (2012). Whiners and deniers–What does self-rated health
measure? Social Science & Medicine, 75(1), 1-9.
118
Leinonen, R., Heikkinen, E., & Jylhä, M. (2001). Predictors of decline in self-assessments of
health among older people—A 5-year longitudinal study. Social Science & Medicine,
52(9), 1329-1341.
Liu, L., Zhang, J., & Deapen, D. (2010). Cancer in Los Angeles County, Incidence and Mortality
by Race/Ethnicity, 1988-2009. Retrieved from Los Angeles, CA:
Lorenzo-Blanco, E. I., Unger, J. B., Ritt-Olson, A., Soto, D., & Baezconde-Garbanati, L. (2011).
Acculturation, Gender, Depression, and Cigarette Smoking Among U.S. Hispanic Youth:
The Mediating Role of Perceived Discrimination. Journal of Youth and Adolescence,
40(11), 1519-1533. doi:10.1007/s10964-011-9633-y
MacKinnon, D. P., Krull, J. L., & Lockwood, C. M. (2000). Equivalence of the mediation,
confounding and suppression effect. Prevention Science, 1(4), 173-181.
Matheson, F. I., Moineddin, R., Dunn, J. R., Creatore, M. I., Gozdyra, P., & Glazier, R. H.
(2006). Urban neighborhoods, chronic stress, gender and depression. Social Science &
Medicine, 63(10), 2604-2616.
McDonough, P., Sacker, A., & Wiggins, R. D. (2005). Time on my side? Life course trajectories
of poverty and health. Social Science & Medicine, 61(8), 1795-1808.
McLaughlin, M., Nam, Y., Gould, J., Pade, C., Meeske, K. A., Ruccione, K. S., & Fulk, J.
(2012). A videosharing social networking intervention for young adult cancer survivors.
Computers in Human Behavior, 28(2), 631-641.
Meyer, O. L., Castro-Schilo, L., & Aguilar-Gaxiola, S. (2014). Determinants of mental health
and self-rated health: a model of socioeconomic status, neighborhood safety, and physical
activity. American Journal of Public Health, 104(9), 1734-1741.
Milam, J. E., Meeske, K., Slaughter, R. I., Sherman-Bien, S., Ritt-Olson, A., Kuperberg, A., . . .
Hamilton, A. S. (2015). Cancer-related follow-up care among Hispanic and non-Hispanic
childhood cancer survivors: The Project Forward study: Follow-Up Care Among Cancer
Survivors. Cancer, 121(4), 605-613. doi:10.1002/cncr.29105
Missinne, S., Daenekindt, S., & Bracke, P. (2015). The social gradient in preventive healthcare
use: what can we learn from socially mobile individuals? Sociology of health & illness,
37(6), 823-838.
Morenoff, J., & Lynch, J. W. (2002). What makes a place healthy. Neighborhood influences on
racial/ethnic disparities in health over the life course, 21-22.
Nathan, P. C., Ford, J. S., Henderson, T. O., Hudson, M. M., Emmons, K. M., Casillas, J. N., . . .
Oeffinger, K. C. (2009). Health behaviors, medical care, and interventions to promote
healthy living in the Childhood Cancer Survivor Study cohort. Journal of Clinical
Oncology, 27(14), 2363.
Nathan, P. C., Greenberg, M. L., Ness, K. K., Hudson, M. M., Mertens, A. C., Mahoney, M.
C., . . . Robison, L. L. (2008). Medical care in long-term survivors of childhood cancer: a
report from the childhood cancer survivor study. Journal of Clinical Oncology, 26(27),
4401.
O'malley, A. S., Kerner, J., Johnson, A. E., & Mandelblatt, J. (1999). Acculturation and breast
cancer screening among Hispanic women in New York City. American Journal of Public
Health, 89(2), 219-227.
Oeffinger, K. C., Mertens, A. C., Hudson, M. M., Gurney, J. G., Casillas, J., Chen, H., . . .
Robison, L. L. (2004). Health care of young adult survivors of childhood cancer: a report
from the Childhood Cancer Survivor Study. The Annals of Family Medicine, 2(1), 61-70.
119
Oeffinger, K. C., Mertens, A. C., Sklar, C. A., Kawashima, T., Hudson, M. M., Meadows, A.
T., . . . Childhood Cancer Survivor, S. (2006). Chronic Health Conditions in Adult
Survivors of Childhood Cancer. The New England Journal of Medicine, 355(15), 1572-
1582. doi:10.1056/NEJMsa060185
Pampel, F. C., Krueger, P. M., & Denney, J. T. (2010). Socioeconomic disparities in health
behaviors. Annual Review of Sociology, 36, 349-370.
Petridou, E. T., Sergentanis, T., Perlepe, C., Papathoma, P., Tsilimidos, G., Kontogeorgi, E., . . .
Polychronopoulou, S. (2014). Socioeconomic disparities in survival from childhood
leukemia in the United States and globally: a meta-analysis. Annals of Oncology, 26(3),
589-597.
Pew Research Center: Ranking Latino populations in the states. (2016). Retrieved from
https://www.pewresearch.org/hispanic/2016/09/08/4-ranking-the-latino-population-in-
the-states/
Pickett, K. E., & Pearl, M. (2001). Multilevel analyses of neighbourhood socioeconomic context
and health outcomes: a critical review. Journal of Epidemiology & Community Health,
55(2), 111-122.
Ponce, N. A., Hays, R. D., & Cunningham, W. E. (2006). Linguistic disparities in health care
access and health status among older adults. Journal of General Internal Medicine, 21(7),
786-791.
Poulton, R., Caspi, A., Milne, B. J., Thomson, W. M., Taylor, A., Sears, M. R., & Moffitt, T. E.
(2002). Association between children's experience of socioeconomic disadvantage and
adult health: a life-course study. The lancet, 360(9346), 1640-1645.
Prentice, J. C. (2006). Neighborhood effects on primary care access in Los Angeles. Social
Science & Medicine, 62(5), 1291-1303. doi:10.1016/j.socscimed.2005.07.029
Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general
population. Applied Psychological Measurement, 1(3), 385-401.
Reidpath, D. D. (2003). “Love thy neighbour”—it's good for your health: a study of racial
homogeneity, mortality and social cohesion in the United States. Social Science &
Medicine, 57(2), 253-261.
Reiss, F. (2013). Socioeconomic inequalities and mental health problems in children and
adolescents: a systematic review. Social Science & Medicine, 90, 24-31.
Reulen, R. C., Winter, D. L., Frobisher, C., Lancashire, E. R., Stiller, C. A., Jenney, M. E., . . .
Group, B. C. C. S. S. S. (2010). Long-term cause-specific mortality among survivors of
childhood cancer. JAMA, 304(2), 172-179.
Rokitka, D. A., Curtin, C., Heffler, J. E., Zevon, M. A., Attwood, K., & Mahoney, M. C. (2017).
Patterns of loss to follow-up care among childhood cancer survivors. Journal of
Adolescent and Young Adult Oncology, 6(1), 67-73.
Rosenfield, S., & Mouzon, D. (2013). Gender and mental health. In Handbook of the sociology
of mental health (pp. 277-296): Springer.
Santiago, C. D., Wadsworth, M. E., & Stump, J. (2011). Socioeconomic status, neighborhood
disadvantage, and poverty-related stress: Prospective effects on psychological syndromes
among diverse low-income families. Journal of Economic Psychology, 32(2), 218-230.
Shaw, R. J., Atkin, K., Bécares, L., Albor, C. B., Stafford, M., Kiernan, K. E., . . . Pickett, K. E.
(2012). Impact of ethnic density on adult mental disorders: narrative review. The British
Journal of Psychiatry, 201(1), 11-19.
120
Shaw, S. J., Huebner, C., Armin, J., Orzech, K., & Vivian, J. (2009). The role of culture in health
literacy and chronic disease screening and management. Journal of Immigrant and
Minority Health, 11(6), 460-467.
Shelton, R. C., Jandorf, L., Ellison, J., Villagra, C., & DuHamel, K. N. (2011). The influence of
sociocultural factors on colonoscopy and FOBT screening adherence among low-income
Hispanics. Journal of health care for the poor and underserved, 22(3), 925.
Short, S. E., & Mollborn, S. (2015). Social determinants and health behaviors: conceptual frames
and empirical advances. Current opinion in psychology, 5, 78-84.
Shuey, K. M., & Willson, A. E. (2014). Economic hardship in childhood and adult health
trajectories: an alternative approach to investigating life-course processes. Advances in
life course research, 22, 49-61.
Sobel, M. E. (1981). Diagonal mobility models: A substantively motivated class of designs for
the analysis of mobility effects. American sociological review, 893-906.
Sobel, M. E. (1985). Social mobility and fertility revisited: Some new models for the analysis of
the mobility effects hypothesis. American sociological review, 699-712.
Stafford, M., Bécares, L., & Nazroo, J. (2010). Racial discrimination and health: exploring the
possible protective effects of ethnic density. In Ethnicity and Integration (pp. 225-250):
Springer.
Stam, H., Grootenhuis, M., & Last, B. (2005). The course of life of survivors of childhood
cancer. Psycho‐Oncology, 14(3), 227-238.
Stansfeld, S. A., Clark, C., Rodgers, B., Caldwell, T., & Power, C. (2011). Repeated exposure to
socioeconomic disadvantage and health selection as life course pathways to mid-life
depressive and anxiety disorders. Social psychiatry and psychiatric epidemiology, 46(7),
549-558.
Steptoe, A., & Feldman, P. J. (2001). Neighborhood problems as sources of chronic stress:
development of a measure of neighborhood problems, and associations with
socioeconomic status and health. Annals of behavioral medicine : a publication of the
Society of Behavioral Medicine, 23(3), 177-185.
Suarez, L. (1994). Pap smear and mammogram screening in Mexican-American women: the
effects of acculturation. American Journal of Public Health, 84(5), 742-746.
Sun, X., Briel, M., Walter, S. D., & Guyatt, G. H. (2010). Is a subgroup effect believable?
Updating criteria to evaluate the credibility of subgroup analyses. Bmj, 340, c117.
Teran, L., Baezconde-Garbanati, L., Marquez, M., Castellanos, E., & BELKIĆ, K. (2007). On-
time mammography screening with a focus on Latinas with low income: a proposed
cultural model. Anticancer research, 27(6C), 4325-4338.
Texas A&M University GeoInnovation Center. (2018). TAMU Geoservices: Geocoding: Texas
A&M University. Retrieved from http://geoservices.tamu.edu/
Tukenova, M., Guibout, C., Oberlin, O., Doyon, F., Mousannif, A., Haddy, N., . . . Hawkins, M.
(2010). Role of cancer treatment in long-term overall and cardiovascular mortality after
childhood cancer. J Clin Oncol, 28(8), 1308-1315.
Turner, H., Firth, D., Ripley, B., Venables, B., Bates, D. M., & Maechler, M. (2007).
Generalized Nonlinear Models package in R (Version 1.1). Retrieved from
https://github.com/hturner/gnm
Unger, J. B. (2014). Cultural Influences on Substance Use Among Hispanic Adolescents and
Young Adults: Findings From Project RED. Child Development Perspectives, 8(1), 48-
53. doi:10.1111/cdep.12060
121
Unger, J. M., Vaidya, R., Albain, K. S., LeBlanc, M. L., Minasian, L. M., Gotay, C., . . . Blanke,
C. D. (2019). Sex differences in adverse event reporting in SWOG chemotherapy,
biologic/immunotherapy, and targeted agent cancer clinical trials. In: American Society
of Clinical Oncology.
Vable, A. M., Gilsanz, P., & Kawachi, I. (2019). Is it possible to overcome the ‘long arm’of
childhood socioeconomic disadvantage through upward socioeconomic mobility?
Journal of Public Health, 41(3), 566-574.
Vable, A. M., Gilsanz, P., Nguyen, T. T., Kawachi, I., & Glymour, M. M. (2017). Validation of a
theoretically motivated approach to measuring childhood socioeconomic circumstances
in the Health and Retirement Study. PloS One, 12(10).
van der Waal, J., Daenekindt, S., & de Koster, W. (2017). Statistical challenges in modelling the
health consequences of social mobility: the need for diagonal reference models.
International journal of public health, 62(9), 1029-1037.
Vogt Yuan, A. S. (2007). Racial composition of neighborhood and emotional well-being.
Sociological Spectrum, 28(1), 105-129.
Walsemann, K. M., Geronimus, A. T., & Gee, G. C. (2008). Accumulating disadvantage over the
life course: evidence from a longitudinal study investigating the relationship between
educational advantage in youth and health in middle age. Research on Aging, 30(2), 169-
199.
Ward, E., DeSantis, C., Robbins, A., Kohler, B., & Jemal, A. (2014). Childhood and adolescent
cancer statistics, 2014. CA: A Cancer Journal for Clinicians, 64(2), 83-103.
doi:10.3322/caac.21219
Ware, J. E., & Gandek, B. (1998). Overview of the SF-36 health survey and the international
quality of life assessment (IQOLA) project. Journal of clinical epidemiology, 51(11),
903-912.
Weinberg, C. R., Moledor, E. S., Umbach, D. M., & Sandler, D. P. (1996). Imputation for
exposure histories with gaps, under an excess relative risk model. Epidemiology, 490-
497.
Whitehead, M., Drever, F., & Doran, T. (2005). Is the health of the long-term unemployed better
or worse in high unemployment areas? Health statistics quarterly, 25, 12.
Xu, K. T. (2002). Usual source of care in preventive service use: a regular doctor versus a
regular site. Health services research, 37(6), 1509-1529.
Yamada, A.-M., Barrio, C., Harding, C. M., Webster, D., Atuel, H., & Hough, R. L. (2009).
Cultural themes associated with pathways to initial use of psychiatric services in a tri-
ethnic community sample of adults with schizophrenia. Schizophrenia research, 115(2-
3), 375.
Yang, J., Schupp, CW, Harrati, A, Clarke, C, Keegan, THM, Gomez, SL. (2014). Developing an
area-based socioeconomic measure from American Community Survey data. Cancer
Prevention Institute of California, Fremont, California.
Yost, K., Perkins, C., Cohen, R., Morris, C., & Wright, W. (2001). Socioeconomic status and
breast cancer incidence in California for different race/ethnic groups. Cancer Causes &
Control, 12(8), 703-711.
Zebrack, B. J., Gurney, J. G., Oeffinger, K., Whitton, J., Packer, R. J., Mertens, A., . . . Robison,
L. L. (2004). Psychological outcomes in long-term survivors of childhood brain cancer: a
report from the childhood cancer survivor study. Journal of Clinical Oncology, 22(6),
999-1006.
122
Zebrack, B. J., Zeltzer, L. K., Whitton, J., Mertens, A. C., Odom, L., Berkow, R., & Robison, L.
L. (2002). Psychological Outcomes in Long-Term Survivors of Childhood Leukemia,
Hodgkin's Disease, and Non-Hodgkin's Lymphoma: A Report From the Childhood
Cancer Survivor Study. Pediatrics, 110(1), 42-52. doi:10.1542/peds.110.1.42
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Childhood cancer survivorship: parental factors associated with survivor's follow-up care behavior and mental health
PDF
Multilevel influences of care engagement and long-term survival among childhood, adolescent, and young adult cancer survivors
PDF
The role of dyadic and triadic factors on psychosocial wellbeing and healthcare interactions among childhood cancer survivors, parents, and medical providers
PDF
Psychosocial and cultural factors in the primary prevention of melanoma targeted to multiethnic children
PDF
Transfusional iron, anthracyclines and cardiac outcomes in childhood cancer survivors
PDF
The role of acculturation in the sun-safe behaviors of US Latino adults in the United States
PDF
Melanoma in children, adolescents, and young adults
PDF
The role of social support in the relationship between adverse childhood experiences and addictive behaviors across adolescence and young adulthood
PDF
Factors associated with agreement between childhood cancer survivors (CCS) and their parents on knowledge and attitudes about health-related quality of life, treatment and follow-up care
PDF
Exploring geography and multi-level factors associated with disparities in HPV vaccination among Latino adolescents in Los Angeles County
PDF
Investigating racial and ethnic disparities in patient experiences with care and health services use following colorectal cancer diagnosis among older adults with comorbid chronic conditions
PDF
Genetic and environmental risk factors for childhood cancer
PDF
Applications of mindfulness toward promoting psychosocial health in diverse communities
PDF
Survival trends and related outcomes of survivors of childhood and young adult cancer
PDF
Association between informed decision-making and mental health-related quality of life in long term prostate cancer survivors
PDF
Racial/ethnic differences in colorectal cancer patient experiences, health care utilization and their association with mortality: findings from the SEER-CAHPS data
PDF
Understanding normative influence of neighborhoods: a multilevel approach to promoting Latinas’ cervical cancer prevention behaviors in urban ethnic communities
PDF
The effects of tobacco exposure on hormone levels and breast cancer risk among young women
PDF
The impact of childhood trauma on substance use and mental health during the SARS-CoV-2 pandemic among young adults
PDF
Racial/ethnic differences in patient experiences with health care in association with the Gleason’s score level at prostate cancer diagnosis: findings from the SEER-CAHPS data
Asset Metadata
Creator
Tobin, Jessica Lyn
(author)
Core Title
Multilevel sociodemographic correlates of the health and healthcare utilization of childhood cancer survivors
School
Keck School of Medicine
Degree
Doctor of Philosophy
Degree Program
Preventive Medicine (Health Behavior Research)
Publication Date
07/19/2020
Defense Date
06/02/2020
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
acculturation,childhood cancer,disparities,Neighborhoods,OAI-PMH Harvest,socioeconomic status,survivorship
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Unger, Jennifer (
committee chair
), Cockburn, Myles (
committee member
), Finch, Brian (
committee member
), Hamilton, Ann (
committee member
), Milam, Joel (
committee member
)
Creator Email
jessi.tobin@gmail.com,tobinj@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-335386
Unique identifier
UC11663577
Identifier
etd-TobinJessi-8701.pdf (filename),usctheses-c89-335386 (legacy record id)
Legacy Identifier
etd-TobinJessi-8701.pdf
Dmrecord
335386
Document Type
Dissertation
Rights
Tobin, Jessica Lyn
Type
texts
Source
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 a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
acculturation
childhood cancer
disparities
socioeconomic status
survivorship