University of Southern California Dissertations and Theses

Social determinants of physiological health and mortality in China

(USC Thesis Other)

Social determinants of physiological health and mortality in China

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

SOCIAL DETERMINANTS OF PHYSIOLOGICAL HEALTH AND MORTALITY IN CHINA

By
Yuan Zhang

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
(GERONTOLOGY)

August 2019

Copyright 2019 Yuan Zhang
2
Dedication
This dissertation is dedicated

To my parents and grandparents whose life experiences inspire my work

3
Acknowledgments
I owe deep gratitude to many people. This dissertation would not have been possible
without the help of faculty, family, and friends who supported, guided, and inspired me during
my years at the University of Southern California’s Leonard Davis School of Gerontology.

First and foremost, I would like to thank the members of my dissertation committee, Drs.
Eileen Crimmins, Jennifer Ailshire, and John Strauss, for always being accessible, patient, and
supportive throughout this endeavor. Eileen, you have been an incredible mentor to me. Since
my first day in the Ph.D. program, you have encouraged and pushed me to become the best
scholar that I can be. You trusted my abilities and saw my potential where I did not. Thank you
for endlessly and meticulously reading and editing my manuscripts, seeking out opportunities for
me, and guiding me on the right path. You taught me how to think big but also to be practical. It
is impossible to count all the ways that you've helped me in my career. Jennifer, you have
always been generous with your time, guidance, and support. I thank you for your practical
advice on conducting literature reviews, developing research ideas, performing analysis, making
tables, writing papers, preparing presentations, and networking. I am fortunate to have benefited
from your critical thinking, excellent teaching, and methodological expertise. John Strauss, your
encouragement to pursue gerontological studies at USC led me to places I never dreamed of
going. I am deeply grateful for your support in the last decade, beginning with my tenure
working on the China Health and Retirement Longitudinal Study at Peking University and
culminating in this dissertation. Your thoughtful comments were invaluable.
I also benefited tremendously from many current and former members of the USC/UCLA
Center on Biodemography and Population Health. I thank my peer, Lauren Brown, who has seen
4
every step of my graduate student career. We formulated research topics, practiced presentations,
wrote papers, and even taught our first class together. I thank Morgan Levine who was my peer
mentor at USC. Morgan has always been generous with her time, willing to listen to my
questions and worries, and providing invaluable advice and support. I thank Hiram Beltran-
Sanchez who allowed me to sit in his demography methods classes at UCLA for two semesters,
from which I benefited tremendously. I would also like to thank Catherine Perez, Jung Ki Kim,
Joseph Saenz, and Uchechi Mitchell for their encouragement and support.
I would like to extend a heartfelt thank you to Jessica Ho for her friendship and
encouragement during the process of completing this dissertation. Diana Wang and Deborah Hoe
- thank you for all the memorable experiences we have had together.
I am also grateful for the opportunities and financial support to attend workshops/training
at the University of Michigan, University of California, Berkeley, the Wittgenstein Centre for
Demography and Global Human Capital, Max Planck Institute for Demographic Research, and
Higher School of Economics. These experiences have opened my eyes to a variety of exciting
research topics and connected me with many talented young scholars around the world.
Finally, I thank my parents for being an important source of motivation for me, and my
husband, Chunyuen Teng, who is always my primary proofreader and my sounding board, for
challenging and helping me. Thank you for your support – emotionally and professionally.
This work was supported by the National Institute on Aging (T32 AG 00037 and P30
AG017265).

5
Table of Contents
DEDICATION ................................................................................................................................ 2
ACKNOWLEDGMENTS .............................................................................................................. 3
CHAPTER 1: INTRODUCTION ................................................................................................... 8
REFERENCES ................................................................................................................. 14
CHAPTER 2: URBAN–RURAL DIFFERENTIALS IN AGE-RELATED BIOLOGICAL RISK
AMONG MIDDLE-AGED AND OLDER CHINESE................................................................. 19
INTRODUCTION ............................................................................................................ 21
METHODS ....................................................................................................................... 24
RESULTS ......................................................................................................................... 29
DISUCSSION ................................................................................................................... 32
REFERENCES ................................................................................................................. 36
TABLES/FIGURES .......................................................................................................... 40
APPENDIX ....................................................................................................................... 45
CHAPTER 3: CHILDHOOD ADVERSITY AND CARDIOVASCULAR AND METABOLIC
RISK IN OLD AGE: EVIDENCE FROM CHARLS .................................................................. 52
INTRODUCTION ............................................................................................................ 53
METHODS ....................................................................................................................... 59
6
RESULTS ......................................................................................................................... 63
CONCLUSIONS AND DISCUSSION ............................................................................. 66
REFERENCES ................................................................................................................. 70
TABLES/FIGURES .......................................................................................................... 76
CHAPTER 4: SOCIOECONOMIC, BIOLOGICAL, AND COMMUNITY ASSOCIATIONS
WITH OLD-AGE MORTALITY IN CHINA .............................................................................. 88
INTRODUCTION ............................................................................................................ 90
DATA AND METHODS .................................................................................................. 95
RESULTS ....................................................................................................................... 101
CONCLUSION ............................................................................................................... 103
REFERENCES ............................................................................................................... 107
APPENDIX ..................................................................................................................... 118
CHAPTER 5: ASCERTAINING CAUSE OF MORTALITY AMONG MIDDLE-AGED AND
OLDER PERSONS USING COMPUTER-CODED AND EXPERT REVIEW VERBAL
AUTOPSIES IN THE CHINA HEALTH AND RETIREMENT LONGITUDINAL STUDY
(CHARLS) .................................................................................................................................. 126
INTRODUCTION .......................................................................................................... 128
DATA AND METHODS ................................................................................................ 132
RESULTS ....................................................................................................................... 135
CONCLUSION AND DISCUSSION ............................................................................. 138
REFERENCES ............................................................................................................... 141
7
TABLES/FIGURES ........................................................................................................ 144
CHAPTER 6: CONCLUSION ................................................................................................... 150
REFERENCES ............................................................................................................... 155

8
Chapter 1: Introduction
The demand for understanding context-specific determinants of health and aging is
greater than ever. Although population aging is taking place in all parts of the world, individuals
in developing settings have a disproportionally large burden attributable to chronic diseases (Kyu
et al., 2018). Literature has documented substantial heterogeneity in disability, cognition,
morbidity, healthy expectancy, and mortality (or life expectancy) across countries (Avendano,
Glymour, Banks, & Mackenbach, 2009; Banks, Marmot, Oldfield, & Smith, 2006; Brønnum-
Hansen, 2014; Crimmins, Garcia, & Kim, 2010; Kyu et al., 2018; Lee et al., 2018), suggesting
that age-related health change takes place in a context and that the role of social, political, and
environmental determinants of health may vary with context. Most existing evidence on social
determinants of health and aging is from developed countries. However, individual risk factors
are contextualized – the exposure to and consequences of factors may depend on multiple
societal factors. China, as a developing and transitional country, is an opportune environment for
studying this topic as the unique historical, political, social, and economic circumstances may
result in context-specific determinants of health and mortality.

Life-course Socioeconomic Positions and the Links to Health and Mortality in China
Research on social determinants of health in the developing world has suggested that
certain socioeconomic indicators, such as living in urban areas, having better economic
resources, and prestigious occupations, may operate in different or even opposite ways in
developing and developed countries (Sudharsanan, 2017; Xu, 2018; Zhao et al., 2016). The
unique life experiences of older Chinese affect the links between socioeconomic factors and
health and mortality. The older Chinese lived their early lives in a socialist society in which
9
equality was emphasized, and hardships were almost universal at younger ages. They then
experienced changes in almost all aspects of life that lead to disparities in income, lifestyle, and
resources for managing later life health; therefore, the correlation between childhood and
adulthood SES could be weaker among older Chinese than in other places if childhood
experiences do not transmit to adult conditions. The current cohort of older Chinese adults has
experienced extreme childhood and early life conditions on a broad scale. Many individuals
experienced multiple wars, rampant infectious diseases, hunger or even famine, and a severe lack
of educational opportunities (Zeng, Gu, & Land, 2007). We may expect that significant mortality
selection could alter the associations between social factors, especially socioeconomic conditions
in early life, and health at old ages.
Unhealthy behaviors, such as smoking, poor diet, and lack of physical activity, are
critical contributors to the development and progression of chronic diseases. Contrary to the
well-documented inverse relationship between SES and unhealthy behaviors in Western
countries, empirical evidence from developing countries consistently shows that persons with
better economic resources have a higher likelihood of engaging unhealthy lifestyles, suggesting
that the social patterning of behavioral factors in China may flatten the SES gradient in health.
Further, because of traditional norms and deficiencies in state/social programs to support older
adults, Chinese older adults still heavily rely on family support which includes informal care
arrangement and financial assistance (Chen, Leeson, & Liu, 2017; Lee & Xiao, 1998; Sereny,
2011; Xie & Zhu, 2009). Therefore, the effects of individual-level socioeconomic factors could
be less pronounced.
Urban–rural environment is another crucial health determinant in China. Throughout the
long history of unbalanced development, urban and rural China have been shown to differ
10
strikingly in physical and social environments which influence people’s health and the aging
process. The urban Chinese population generally has more education, higher income, and greater
access to health services than the rural population (Gong et al., 2012). Additionally, rapid
urbanization complicates urban-rural health disparities. Rural-to-urban migrants cannot easily
change their hukou (the official household registration) to their new place of residence. Because
hukou officially determines the welfare support to which a person is entitled (Yip et al., 2012),
individuals who migrate to urban cities without an urban hukou are largely ineligible for social
programs and municipal services; therefore, could be more vulnerable to adverse health (Hu,
Cook, & Salazar, 2008).

Incorporating Biomarker and Physical Assessment Data in the Study of Health at Old Ages
Much of the existing literature on China relies on respondents’ self-reports of
downstream health outcomes. I argue that incorporating measures of biomarkers and physical
assessment can provide a valuable picture of the current health of the mid-aged and older
Chinese and advance scientific understanding of determinants of aging processes in the Chinese
context.
First, using objectively measured health indicators is of importance in China and many
other less developed settings where a significant proportion of older adults do not have access to
regular and high-quality health care. Because the awareness of chronic diseases is fairly poor and
largely determined by SES (Lu et al., 2017; Zhao et al., 2016), only using reports of health will
introduce bias in the relationship between social factors and health. However, in addition to self-
reports, biological and physical assessment data offer critical information on individual’s health
as individuals might not be able to report about their own health. More importantly, the
11
biological and performance assessment measures provide information on physiological health
and physical functioning that allows us to uncover how social, behavioral, and environmental
factors “get under the skin” to have consequences on downstream health outcomes and
ultimately lead to mortality (Crimmins & Vasunilashorn, 2011; Turra et al., 2005). Third,
studying physiological dysregulation and physical functioning changes allows us to risk social
and environmental circumstances to risk factors that occur prior to clinically diagnosed disease
and disability, therefore expanding our understanding of the process of health deterioration with
age.

Gaps in the Literature
Prior research on health and aging in China largely has used reports from individuals to
study social determinants of some downstream aspects of health, including physical functioning
and disability (Beydoun & Popkin, 2005; Zimmer, Wen, & Kaneda, 2010), chronic conditions
(Zimmer & Kwong, 2004), self-assessed health (Chen, Yang, & Liu, 2010; Xu & Xie, 2017;
Zimmer & Kwong, 2004), and mortality (Cheng & Elo, 2009; Wen & Gu, 2011; Zhu & Xie,
2007). However, due to data availability, most of the current literature has the following
limitations. First, most studies used non-representative samples so the populations results can be
generalized to varies by study – this could be one of the reasons for inconsistent findings across
studies. Second, as discussed previously, the differentials in the awareness of the presence of
chronic diseases are likely to introduce bias in the associations between socioeconomic factors
and health. Third, the lack of biological data limits the ability to identify physiological pathways
by which social factors “get under the skin” to affect downstream health outcomes. Last, but
importantly, analyzing determinants of mortality and patterns of causes of death are largely
12
constrained due to defective data in China. Previous research on old-age mortality in the
Chinese population primarily studied a relatively old population and did not use nationally
representative samples (Cheng & Elo, 2009; Shen & Zeng, 2014; Wen & Gu, 2011; Zeng et al.,
2007). Yet, the very old are a highly selective population and the determinants of mortality could
be different in the young old and the oldest old (Lee, Go, Lindquist, Bertenthal, & Covinsky,
2008; Zhu & Xie, 2007).
The present study
To address these limitations, the proposed research will use in-depth contextual
information on institutions and cultures as well as biomarker and physical assessment data to
clarify the context-specific determinants of physiological health and mortality using a nationally
representative sample of middle-aged and older Chinese. The aim of the dissertation is to
elucidate the unique underlying mechanisms that connect the life-time exposures to the processes
of aging in the Chinese context.
1. This study documents the recent profile of physiological health and mortality among mid-
aged and older Chinese using nationally representative data from the China Health and
Retirement Longitudinal Study. The representativeness of the sample ensures that results can
be generated to clarify the situations in China.
2. The CHARLS data collected a myriad of data in social, behavioral, economic aspects of lives
in China which have never been available in previous studies. The richness of the CHARLS
data allows us to explore how different dimensions of life-time experiences at individual and
community levels operate to influence health in old age.
13
3. The availability of biomarker and physical assessment data helps clarify how social and
behavioral factors get under the skin to affect physiological and physical functioning and
subsequent health outcomes. The objectively-measured health indicators also additional
information on health that may be not measured in the household survey. As many studies
have suggested, integrating both measured and reported health indicators would increase the
predictive capacities of the models predicting mortality (Levine, 2013; Turra et al., 2005).
4. The verbal autopsy interview in CHARLS offers a unique opportunity to ascertain cause of
death in the context of population-based surveys and possibly link cause of death information
to risk factors.

As a whole, this study advances the scientific understanding of the diverse aging
processes for people who have lived under different social circumstances and developmental
paths and emphasizes the complex patterns of social determinants of health.

14
REFERENCES
Avendano, M., Glymour, M. M., Banks, J., & Mackenbach, J. P. (2009). Health disadvantage in
US adults aged 50 to 74 years: a comparison of the health of rich and poor Americans with
that of Europeans. American Journal of Public Health, 99(3), 540–548.
https://doi.org/10.2105/AJPH.2008.139469
Banks, J., Marmot, M., Oldfield, Z., & Smith, J. P. (2006). Disease and disadvantage in the
United States and in England. JAMA, 295(17), 2037–2045.
https://doi.org/10.1001/jama.295.17.2037
Beydoun, M. A., & Popkin, B. M. (2005). The impact of socio-economic factors on functional
status decline among community-dwelling older adults in China. Social Science and
Medicine. https://doi.org/10.1016/j.socscimed.2004.08.063
Brønnum-Hansen, H. (2014). Ranking health between countries in international comparisons.
Scandinavian Journal of Public Health, 42(3), 242–244.
https://doi.org/10.1177/1403494814522555
Chen, F., Yang, Y., & Liu, G. (2010). Social change and socioeconomic disparities in health over
the life course in china: A cohort analysis. American Sociological Review.
https://doi.org/10.1177/0003122409359165
Chen, T., Leeson, G. W., & Liu, C. (2017). Living arrangements and intergenerational monetary
transfers of older Chinese. In Ageing and Society.
https://doi.org/10.1017/S0144686X16000623
Cheng, H., & Elo, I. T. (2009). Mortality of the oldest old Chinese: The role of early-life
nutritional status, socio-economic conditions, and sibling sex-composition. Population
Studies. https://doi.org/10.1080/00324720802626921
15
Crimmins, E. M., Garcia, K., & Kim, J. K. (2010). Are International Differences in Health
Similar to International Differences in Life Expectancy? In E. M. Crimmins, S. H. Preston,
& B. Cohen (Eds.), International Differences in Mortality at Older Ages: Dimensions and
Sources (pp. 88–102). Washington, D.C.: The National Academies Press.
Crimmins, E. M., & Vasunilashorn, S. (2011). Links Between Biomarkers and Mortality. In
International Handbook of Adult Mortality. https://doi.org/10.1007/978-90-481-9996-9
Gong, P., Liang, S., Carlton, E. J., Jiang, Q., Wu, J., Wang, L., & Remais, J. V. (2012).
Urbanisation and health in China. The Lancet, 379(9818), 843–852.
https://doi.org/https://doi.org/10.1016/S0140-6736(11)61878-3
Kyu, H. H., Abate, D., Abate, K. H., Abay, S. M., Abbafati, C., Abbasi, N., … Murray, C. J. L.
(2018). Global, regional, and national disability-adjusted life-years (DALYs) for 359
diseases and injuries and healthy life expectancy (HALE) for 195 countries and territories,
1990–2017: a systematic analysis for the Global Burden of Disease Study 2017.
The Lancet, 392(10159), 1859–1922. https://doi.org/10.1016/S0140-6736(18)32335-3
Lee, J., Phillips, D., Wilkens, J., Chien, S., Lin, Y.-C., Angrisani, M., & Crimmins, E. (2018).
Cross-Country Comparisons of Disability and Morbidity: Evidence from the Gateway to
Global Aging Data. The Journals of Gerontology. Series A, Biological Sciences and
Medical Sciences, 73(11), 1519–1524. https://doi.org/10.1093/gerona/glx224
Lee, S. J., Go, A. S., Lindquist, K., Bertenthal, D., & Covinsky, K. E. (2008). Chronic
Conditions and Mortality Among the Oldest Old. American Journal of Public Health, 98(7),
1209–1214. https://doi.org/10.2105/AJPH.2007.130955
Lee, Y. J., & Xiao, Z. (1998). Children’s support for elderly parents in urban and rural China:
Results from a national survey. Journal of Cross-Cultural Gerontology.
16
https://doi.org/10.1023/A:1006591608724
Levine, M. E. (2013). Modeling the rate of senescence: can estimated biological age predict
mortality more accurately than chronological age? The Journals of Gerontology. Series A,
Biological Sciences and Medical Sciences, 68(6), 667–674.
https://doi.org/10.1093/gerona/gls233
Lu, J., Lu, Y., Wang, X., Li, X., Linderman, G. C., Wu, C., … Jiang, L. (2017). Prevalence,
awareness, treatment, and control of hypertension in China: data from 1.7 million adults in a
population-based screening study (China PEACE Million Persons Project). Lancet (London,
England), 390(10112), 2549–2558. https://doi.org/10.1016/S0140-6736(17)32478-9
Sereny, M. (2011). Living arrangements of older adults in China: The interplay among
preferences, realities, and health. Research on Aging.
https://doi.org/10.1177/0164027510392387
Shen, K., & Zeng, Y. (2014). Direct and indirect effects of childhood conditions on survival and
health among male and female elderly in China. Social Science and Medicine.
https://doi.org/10.1016/j.socscimed.2014.07.003
Sudharsanan, N. (2017). The Association Between Socioeconomic Status and Adult Mortality in
a Developing Country: Evidence From a Nationally Representative Longitudinal Survey of
Indonesian Adults. The Journals of Gerontology: Series B, gbx061-gbx061. Retrieved from
http://dx.doi.org/10.1093/geronb/gbx061
Turra, C. M., Goldman, N., Seplaki, C. L., Glei, D. A., Lin, Y. H., & Weinstein, M. (2005).
Determinants of mortality at older ages: The role of biological markers of chronic disease.
Population and Development Review. https://doi.org/10.1111/j.1728-4457.2005.00093.x
Wen, M., & Gu, D. (2011). The Effects of Childhood, Adult, and Community Socioeconomic
17
Conditions on Health and Mortality among Older Adults in China. Demography, 48(1),
153–181. https://doi.org/10.1007/s13524-010-0003-2
Xie, Y., & Zhu, H. (2009). Do Sons or Daughters Give More Money to Parents in Urban China?
Journal of Marriage and Family, 71(1), 174–186. https://doi.org/10.1111/j.1741-
3737.2008.00588.x
Xu, H. (2018). Multilevel socioeconomic differentials in allostatic load among Chinese adults.
Health and Place. https://doi.org/10.1016/j.healthplace.2018.08.012
Xu, H., & Xie, Y. (2017). Socioeconomic Inequalities in Health in China: A Reassessment with
Data from the 2010–2012 China Family Panel Studies. Social Indicators Research.
https://doi.org/10.1007/s11205-016-1244-2
Zeng, Y., Gu, D., & Land, K. C. (2007). The association of childhood socioeconomic conditions
with healthy longevity at the oldest-old ages in China. Demography, 44(3), 497–518.
https://doi.org/10.1353/dem.2007.0033
Zhao, Y., Crimmins, E. M., Hu, P., Shen, Y., Smith, J. P., Strauss, J., … Zhang, Y. (2016).
Prevalence, diagnosis, and management of diabetes mellitus among older Chinese: results
from the China Health and Retirement Longitudinal Study. International Journal of Public
Health. https://doi.org/10.1007/s00038-015-0780-x
Zhu, H., & Xie, Y. (2007). Socioeconomic differentials in mortality among the oldest old in
China. Research on Aging. https://doi.org/10.1177/0164027506296758
Zimmer, Z., & Kwong, J. (2004). Socioeconomic Status and Health among Older Adults in Rural
and Urban China. Journal of Aging and Health. https://doi.org/10.1177/0898264303260440
Zimmer, Z., Wen, M., & Kaneda, T. (2010). A multi-level analysis of urban/rural and
socioeconomic differences in functional health status transition among older Chinese.
18
Social Science & Medicine, 71(3), 559–567.
https://doi.org/https://doi.org/10.1016/j.socscimed.2010.03.048

19
Chapter 2: Urban –rural differentials in age-related biological risk among
middle-aged and older Chinese
Yuan S. Zhang & Eileen M. Crimmins

ABSTRACT
Objectives
To assess urban–rural differentials in age-related biological risk among middle-aged and
older Chinese and links to individual and community characteristics.
Methods
Data come from the national baseline survey of the China Health and Retirement
Longitudinal Study. Biological risk is assessed using a set of measured biomarkers that reflect
cardiovascular, metabolic, and inflammatory processes.
Results
Urban residents who are officially registered in urban areas have greater biological risk
than rural residents. Having junior school or higher education provides an independent and
persistent protective effect against biological risk and eliminates the effect of community-level
measures. The reduced physical activity of urban dwellers with urban origins explains a
substantial part of the difference in risk.
Conclusions
Urban dwellers with urban household registration have elevated risk compared with their
rural peers, indicating that lifetime exposure to urban areas is an important risk factor for
increased biological risk in China. The urban–rural differential in risk is accounted for by
adjusting for health behaviors, particularly physical activity. The reduced physical activity
20
among urban dwellers with urban household registration appears to be highly related to their
elevated risk. No significant associations between community-level characteristics and biological
risk are found beyond individual characteristics.

Keywords: Biomarkers, China, Health disparity, Urban–rural difference, CHARLS

*This chapter was published as “Zhang, Y.S. & Crimmins, E.M. Int J Public Health (2018).
https://doi.org/10.1007/s00038-018-1189-0”

21
INTRODUCTION
Urban–rural residency is a crucial health determinant in many developing countries
including China. Over the course of its long history of unbalanced development, urban and rural
China have been shown to differ strikingly in physical environment, education, income, diet, and
lifestyles, which influence people’s health and the aging process. The urban Chinese population
generally has more education, higher income, and greater access to health services than the rural
population (Gong et al. 2012) leading to an ‘‘urban advantage’’ in certain downstream aspects of
health including life expectancy (Li and Dorsten 2010), cognitive functioning (Jia et al. 2014),
depressive symptoms (Li et al. 2016), and disability (Kaneda et al. 2010). However, Western-
style diets, lack of physical labor, and pollution are also more prevalent in urban areas. Studies
have shown that urban residents are more likely to be overweight or obese and to have
hypertension and diabetes (Hou 2008; Zhao et al. 2016). This study investigates urban-rural
differences in multiple indicators of physiological dysregulation which are precursors or
indicators of chronic diseases associated with aging.

Lifetime residence is not fixed, however. China has experienced massive urban growth
from the movement of people to urban areas from rural areas. People who migrate to urban areas
tend to move in early adulthood so they spend their early lives in rural areas and later lives in
urban areas. Individuals who migrate to urban areas are likely to have physically demanding jobs
but limited access to education, job opportunities, social programs, and municipal services
because they cannot easily change their administrative registration, called ‘‘hukou,’’ to their new
place of residence (Hu et al. 2008). As a governmental household registration system, hukou
status is not easy to change and it is the way individuals connect to social programs provided by
22
the government and is a source of inequality between urban and rural China. An individual living
in an urban area who still has a rural hukou is usually a migrant worker born and raised in a rural
area who has moved to an urban area in adulthood. Because people living in different places are
exposed at different points in life to varying physical, social, and service environments,
differentials in health by urban–rural residency as well as hukou are likely to arise.

The urban/rural setting in China is related to socioeconomic (SES) differences. In China,
the urban population has better education than the rural population. There is ample evidence that
education is linked to better health (Cutler and Lleras-Muney 2006; Brown et al. 2012).
Education protects against adverse health outcomes by influencing individuals’ health behaviors
and access to health care, by improving ability to use health-related information to manage
health conditions, and by facilitating the acquisition of social and psychological resources
(Mirowsky and Ross 2003; Ross and Wu 1995). Education is also linked to better economic
situations. Higher income and wealth are paths through which socioeconomic conditions affect
health (Sorlie et al. 1995). Individuals with greater economic resources may have better access to
highquality food and health care and suffer from fewer stressful life events (Seeman et al. 1997).
Alternatively, in a developing country setting, higher income is also positively linked to
unhealthier behaviors, such as smoking, drinking, and diets high in fat, sugar, and refined
carbohydrates (Chen et al. 2010). We expect to find that the Chinese with higher SES will have
lower biological risk and that higher SES will partially explain the urban–rural differentials. We
further hypothesize that education will provide stronger protective effects than economic
resources, because education reflects life-cycle socioeconomic circumstances and provides a
more appropriate measure of cumulative and longstanding SES (Hayward et al. 2000).
23
While both rural and urban China have gone through epidemiological transitions
resulting in chronic diseases being the leading causes of death (National Bureau of Statistics of
China 2016; Yang et al. 2008), urban and rural residents face different risks of developing
chronic conditions and different resources for management of conditions. The urban population
has experienced greater changes in diet, lifestyle, and physical environment, which may place
them at higher risk of chronic diseases. However, the lack of development in some rural
environments is also likely to negatively impact heath as it offers less infrastructure for housing,
sanitation, and health care. Older persons in China living in less developed communities, i.e.,
poor waste management, reliance on hay or coal for cooking fuel, and no tap water, have worse
health (Smith et al. 2013). In this study, we expect to find those whose behavior is healthier (i.e.,
physically active) and who live in communities with better infrastructure will have lower
biological risk. We also hypothesize that behavioral factors and community characteristics will
partly explain urban–rural differentials in biological risk.
This study is unique in addressing differentials in biological risk by urban–rural residence
and origin using data from a nationally representative survey in China. The acceleration of the
chronic disease epidemic along with rapid population aging in China results in an urgent need to
better document the biological risk profiles for middle-aged and older Chinese using data from a
recent representative sample. As China ages rapidly, understanding behavioral and
environmental factors that contribute to urban–rural differentials is crucial for planning public
health initiatives and allocating health care to address the growing epidemic of chronic diseases.
This study incorporates information on hukou and current residency to help separate the effects
of lifetime urban residence and later urban residence. It is also unique in incorporating indicators
of individual and current community characteristics as most research among older Chinese
24
includes individual characteristics but not neighborhood environmental aspects as factors
influencing health and aging. By using measured physiological dysregulation, this study can
provide a valuable picture of the current health of the Chinese population in urban and rural
settings and clarify differentials in the aging process as well as the likely consequences of further
urbanization. China provides a context for this study which is likely to be informative for other
countries because it has a very significant rural population that is undergoing rapid urbanization
as is common in many places in the developing world.

METHODS

Dataset
Data for this study come from the first wave of the China Health and Retirement
Longitudinal Study (CHARLS), a nationally representative survey of the Chinese population
aged 45 and older living in households, conducted by Peking University in 2011–2012.
CHARLS adopted a stratified multistage probability sampling design using all counties in
mainland China except for those in Tibet. These were stratified by region, urban–rural, and
county per capita GDP. From these, 150 rural counties/urban districts were randomly selected,
and three rural villages/urban communities were randomly chosen from each county/ district,
resulting in 450 rural villages/urban communities in total. Sampling at the household level was
done from a list all dwelling units in all residential buildings. After applying sampling weights,
CHARLS baseline sample demographics match closely those of the population census in 2010
(Zhao et al. 2014b). The response rate for the baseline survey was 80.5% (94% in rural areas and
25
69% in urban areas) (Zhao et al. 2013). The study protocol was approved by the ethical review
committee (IRB) of Peking University.
CHARLS questionnaires reflect information for individuals, households, and
communities/villages. The community survey was administered with the person in charge of the
community/village committee (Zhao et al. 2013). Anthropometric measurements and non-blood-
based biomarkers, including blood pressure and pulse rate, were collected in the household along
with the main household surveys. Blood-based biomarkers were collected in a subsequent visit to
a local health facility. Medically trained staff from the Chinese Center for Disease Control and
Prevention (China CDC) collected three tubes of venous blood based on a standard protocol.
Respondents were asked to fast overnight before the blood draw, but a blood sample was
collected even if a respondent did not fast. Blood was prepared at the local health facilities and
shipped to Beijing where blood-based biomarkers were assayed from plasma. Further details of
study design, collection, and processing procedures are available in Zhao et al. (2014a).

Analytic sample
The survey collected information from 17,290 respondents aged 45 and above, of whom
9928 provided a blood sample and a physical assessment. We excluded participants who did not
provide a blood sample at a subsequent clinic visit (N = 7362), did not fast or for whom fasting
status was not recorded (N = 1009), those with missing data on one or more of the biomarker
measures (N = 407), and 13 respondents who did not provide information on gender or hukou.
The final analytic sample consists of 8499 individuals. Respondents who were excluded from the
analytic sample are more likely to be male and urban residents. No significant age or educational
differences are observed. Women and rural respondents were more likely to provide blood (69%
26
for women, 65% for men; 71% for rural respondents, 60% for urban respondents). A biomarker
weight was created by the CHARLS team to correct for both initial non-participation in the
survey and non-participation in blood collection. It combines the individual-level sampling
weight with a correction for non-participation in the blood sample collection using an inverse
probability weighting factor constructed from a logit regression of whether the individuals in the
blood collection (Zhao et al. 2014a). Details on sample selection and missing data analysis are
given in “Appendix of Electronic Supplementary Material.”

Measures
Biological risk
We capture age-related physiological dysregulation using a set of biomarkers that reflect
cardiovascular, metabolic, and inflammatory processes. These include systolic blood pressure,
diastolic blood pressure, pulse rate, total cholesterol, high-density lipoprotein (HDL) cholesterol,
low-density lipoprotein (LDL) cholesterol, triglycerides, plasma glucose, body mass index
(BMI), and C-reactive protein (CRP). These indicators of biological risk are recognized as risk
factors for major health outcomes in old age: cardiovascular diseases, metabolic diseases, loss of
cognitive and physical function, and mortality (Seeman et al. 2001). A dichotomous indicator for
each biomarker indicates ‘‘high risk’’ or ‘‘not high risk’’ based on clinical cutoff values (Table
1). The prevalence of high-risk levels for each biomarker is presented in Table 1. More than a
quarter of the sample has high systolic blood pressure (29.0%) and low HDL cholesterol
(25.8%). High CRP is found in 18.7% of the sample; 13.6% of the sample has high diastolic
blood pressure; and about 14% has high triglycerides. There are relatively few individuals with a
rapid pulse (5.5%) or who are obese (4.9%). The total number of biological risks is calculated by
27
summing the numbers of biomarkers in the high-risk range. Measures such as this have been
used to describe differences in risk within populations (Beltrán-Sánchez and Crimmins 2013;
Crimmins et al. 2009; Geronimus et al. 2006; Seeman et al. 2001).

Urban–rural
`Urban–rural participants are categorized based on their usual place of residence and their
assigned hukou. Three categories were created: rural residency, urban residency and rural hukou,
and urban residency and urban hukou. The small percentage (<2%) who live in rural areas but
have urban hukou are included in the rural residency category. This three-category variable
separates those living in urban areas by their ability to use services and obtain benefits.

Socioeconomic status (SES)
Individual educational attainment and an indicator of household resources—tertile of per
capita expenditures (PCE) (determined before weighting)—indicate the availability and
accessibility of socioeconomic resources (Zhao et al. 2016). Four education categories are
denoted: illiterate, literate without schooling (can read and write but no formal schooling),
primary school, and junior/secondary school or higher.

Health behaviors
We control for three health behavior indicators: smoking, drinking, and physical activity.
Non-smokers are those who never smoked; respondents who reported they do/did smoke are
classified as smokers. Drinking indicates how often the respondents drank liquor, wine, rice
wine, and beer per month in the last year. We use a gender-specific cutoff point to construct the
28
measure of alcohol consumption (US Department of Health and Human Services 2015). Those
who did not drink in the last year are classified as non-drinkers; women who had up to one drink
per day and men who had up to two drinks per day are classified as moderate drinkers; women
who drank more than one drink and men who drank more than two drinks per day are classified
as heavy drinkers. Dummy variables are constructed that represent the level of daily physical
activity: no regular physical activity, performing moderate physical activities at least 10 min
routinely, and performing vigorous activities at least 10 min routinely.

Community characteristics
A growing literature has reported association between neighborhood or community
environment and health (Li et al. 2016; Smith et al. 2013; Yen et al. 2009). Following the
approach of Li et al. (2016), we develop a community infrastructure deficiency index based on a
principal components analysis of seven indicators of community infrastructure, including road
type, days annually with unpassable roads, community accessible by bus, community having a
sewer system, main type of toilet in the community, main type of cooking fuel, and main source
of drinking water. Communities are divided into four quartiles based on the infrastructure
deficiency index scores.

Analysis
We use STATA 14.1 for all analyses. Because individuals are nested within communities
and our dependent variable, the number of biological risks, is an over-dispersed count variable,
we use multilevel negative binomial regression models. To understand the overall urban–rural
differentials in biological risk, we present the urban–rural differentials with only age and gender
29
controlled in Model 1. In Model 2, indicators of individual and family SES, education and
household expenditure, are added to examine how much of the urban–rural difference is due to
SES variation. We subsequently add indicators of health behaviors (Model 3) and community
characteristics (Model 4). Analyses are weighted by the CHARLS biomarker weight to adjust for
the complex sampling design and for being in the biomarker sample to ensure the
representativeness of results for the country. Robust standard errors for the regression
coefficients are computed that allow for clustering at the community level.
Because the questions on physical activities were asked only of a random half sample,
Models 3 and 4 which include these variables have a smaller N. Since the selection process was
random, the sample should remain nationally representative. In ‘‘Appendix of Electronic
Supplementary Material,’’ we relate the variables used in the analysis to the likelihood of being
in this half sample. We find only being in the top tertile of wealth is linked to differential—
higher—response to this question. To test the robustness of our results, we rerun Models 1 and 2
using the same half sample and compared the results with the results from the full sample. The
coefficients are very close, and the statistical significance is the same.

RESULTS
Descriptive analysis
Descriptive characteristics of the sample are given in Table 2. Just over half (54%) of the
sample resides in rural areas. The urban population is equally split between those with urban
hukou and rural hukou (23% vs. 23%). Overall, individuals in the sample have a mean biological
risk score of 1.5 and urban Chinese have a higher level of biological risk than rural Chinese. The
30
average number of biological risks is 1.8 for urban residents with urban hukou, 1.6 for urban
residents with rural hukou, and 1.4 for rural residents.
Our descriptive results show marked urban–rural gradients in many aspects, especially in
SES and physical activities. In this sample, educational attainment ranges from illiterate (25.8%),
literate without schooling (17.4%), primary school educated (22.1%) to junior/secondary school
and higher (34.7%). About 50% of rural residents but only about 17% of urban residents with
urban hukou did not complete primary education. The percentage with a junior/secondary school
or higher education is 23.3% for rural residents, 33.8% for urban residents with rural hukou, and
61.8% for urban residents with urban hukou. Similarly, a large urban–rural difference in
household expenditures is found. Urban residents with urban hukou are more likely to be in the
top tertile of PCE (71.3%) and much less likely to be in the bottom tertile (8.8%), whereas the
percentage of individuals who belong to the bottom tertile is much higher for rural residents and
urban residents with rural hukou (38.2% and 25.5%, respectively). About 40% of the sample
has smoked and 17.6% drinks moderately. The percentage of heavy drinkers is very low,
regardless of residency or hukou status. Rural residents have the highest levels of
physical activity. About 30.0% and 43.4% of rural residents routinely engage in moderate and
vigorous physical activities, respectively. The percentages engaging in vigorous activities are
lower for urbanites and individuals with urban hukou are the most sedentary (54.1%). The
indicators of the community environment differ significantly across urban–rural settings.
Infrastructure is more deficient for rural villages.

31
Regression analysis
Results from regression models are presented in Table 3. From Model 1 where only age
and gender are controlled, the level of biological risk for individuals who live in urban
areas with urban hukou is elevated (rate ratio = 1.32, significant at the 0.1% level) relative to
rural residents. When we add education and per capita expenditure in Model 2, the urban–rural
differential in biological risk persists (rate ratio = 1.36, significant at the 0.1% level). Individuals
who completed junior/secondary school have a 10% reduced risk compared with the illiterate
(rate ratio = 0.90 in Model 2, significant at the 5% level).
We find strong impacts of behavioral factors on biological risk. Smoking is associated
with an about 11% higher relative risk (rate ratio = 1.11 in Model 3 and Model 4, significant at
the 5% level). Performing vigorous physical activities is correlated with about a 20% reduced
relative risk (rate ratio = 0.81 in Model 3, rate ratio = 0.82 in Model 4, significant at the 0.1%
level). Performing moderate physical activity also displays a negative association with biological
risk, but the effect is small and not significant.
Once behavioral factors are controlled, the urban– rural differences are significantly
reduced (rate ratio is reduced from 1.36 to 1.19 from Model 2 to Model 3). This means
behavioral factors explain a substantial part of the initial higher urban risk observed in Models 1
and 2. In Model 4, we control for community infrastructure. The effects of community measures
are small and not significant, they eliminate the significant urban–rural difference in biological
risk when controlled, but the coefficient is not reduced.

32
DISUCSSION
CHARLS provides measured biological risk profiles for the middle-aged and older
Chinese. The nationally representative sample allows us to clarify the situation in China as a
whole as well as for representative population subgroups. In addition, the use of objective
measures to define biological risk provides objective information for a population in which
exposure to the medical system and knowledge of physiological measures of health may be
limited and differential (Zhao et al. 2016). Our approach to defining biological risk reflects an
absolute level of clinical risk rather than relative level of risk as is used in some studies. This
index which includes indicators of cardiovascular, metabolic, and inflammatory risk is desirable
for examining change over time or across places. This study focuses on urban–rural differences
in biological risk and aims to identify individual characteristics and environmental factors at the
community level associated with age-related biological risk. We find urban dwellers with urban
hukou have elevated risk compared with their rural peers. The fact that those who have rural
origins and live in urban areas do not differ from the rural group may mean that the lifetime
exposure to urban areas is an important risk factor for the increased biological risk. The higher
status of urban dwellers is not the source of their poorer physiology. However, their reduced
physical activity appears to be highly related to their elevated risk. The urban–rural differential
in biological risk is accounted for by adjusting for health behaviors, particularly physical
activity. Our descriptive analysis shows that the rural population is more physically active than
urban residents with urban hukou, and urban residents with rural hukou are in-between. Worse
physiological dysregulation in urban areas differs from the association often found in high-
income countries but it is found in other developing countries (Crimmins 2015). The difference
in the social patterning of behavioral factors in various settings suggests that the association
33
between risk factors for chronic conditions and socioeconomic status depends on the stage of the
epidemiological transition and the level of development and differentials may change with
development (Stringhini and Bovet 2017). In many developing countries, individuals with high
socioeconomic status are less physically active and consume more fats, salt, and processed food
(Allen et al. 2017), resulting in higher risk of chronic conditions. Our findings indicate that
continued urbanization is likely to present major public health challenges in China. As rural
areas continue to urbanize, migrants continue to move from rural to urban areas, and fewer
people work in physically demanding occupations, more Chinese will adopt Western style diets
and sedentary lifestyles. Although a variety of campaigns promoting physical activity and
healthy lifestyle have been implemented in the last 10 years (Wang and Zhai 2013), it is a
challenge to change people’s habits and activity levels if they have been sedentary for years.
Our study suggests that higher education is a strong and independent protective factor
against developing high biological risk. The effect of education persists after behavioral factors
and community characteristics are accounted for. This could reflect more knowledge of health
management and better access to higher-quality healthcare for highly educated older adults. We
do not find a significant correlation between per capita expenditure and biological risk. This is
consistent with some recent literature which has been skeptical about the links between financial
measures of SES and health (Smith 2004). Smith (2004) suggests that economic circumstances
during childhood might set the stage for the adult SES health gradient and have a bearing on
health later in life.
One thing we should note is that we do not find sex differences in physiological
dysregulation in China. In many countries, differences in physiological dysregulation between
men and women are found that reflect their social roles and health behaviors (Crimmins 2015).
34
The equality of dysregulation in this study may indicate relative similarity in life experiences for
this cohort of Chinese men and women.
While the percentage of obese persons in China is still relatively low, many studies have
shown a rapidly rising prevalence of overweight and obesity (Wang et al. 2007; Yang et al.
2008). Because of the well-known strong links between obesity and other cardiometabolic risks,
the epidemic of obesity will become an increasing public health challenge in China.
One limitation of our study is that the study sample is reduced by initial non-response and
non-participation in the blood sample collection. However, by applying the biomarker weight
correcting for this non-response, our analytic sample is representative of middle-aged and older
Chinese adults. Since Models 3 and Model 4 are based on the random half of the sample, we
examined whether respondents who were asked the questions on physical activities differ from
those who did not receive the questions. We find respondents who were excluded from Models 3
and 4 were not significantly different from those who were in the first two models, except that
they were more likely to be in the top tertile of per capita expenditures; however, as given in
Table 3, per capita expenditure is not significantly correlated with biological risk. Reassuringly,
when we use the same half sample for Models 1 and 2, the results are similar to those shown.
Steps for further research deserve mention. Mechanisms behind the urban–rural
differentials in biological risk need further exploration including how lifelong living
environments might be a potential mediator of the differences we observe. In moving forward,
taking advantage of the longitudinal design of CHARLS and a recent life history survey to better
explore potential causal effects will provide a next step for this research. Our findings not only
enhance our understanding of the health conditions in the older Chinese population and the
challenges China is facing in maintaining a healthy population, but also indicate that improving
35
education and promoting active lifestyles can help Chinese people reduce physiological
dysregulation and achieve healthy aging.

Compliance with ethical standards:
Ethical approval All procedures performed in studies involving human participants were
in accordance with the ethical standards of the institutional and/or national research committee
and with the 1964 Helsinki Declaration and its later amendments or comparable ethical
standards.
Informed consent Informed consent was obtained from all individual participants
included in the study.

36
REFERENCES
Allen L, Williams J, Townsend N, Mikkelsen B, Roberts N, Foster C, Wickramasinghe K (2017)
Socioeconomic status and noncommunicable disease behavioural risk factors in low-income
and lower-middle-income countries: a systematic review. Lancet Glob Health 5:e277–e289.
https://doi.org/10.1016/S2214-109X(17)30058-X
Beltrán-Sánchez H, Crimmins EM (2013) Biological risk in the Mexican population at the turn
of the 21st century. J Cross Cult Gerontol 28:299–316. https://doi.org/10.1007/s10823-013-
9199-0
Brown DC, Hayward MD, Montez JK, Hummer RA, Chiu CT, Hidajat MM (2012) The
significance of education for mortality compression in the United States. Demography
49:819–840. https://doi.org/10.1007/s13524-012-0104-1
Chen F, Yang Y, Liu G (2010) Social change and socioeconomic disparities in health over the
life course in China: a cohort analysis. Am Sociol Rev 75:126–150. https://doi.org/10.1177/
0003122409359165
Crimmins EM (2015) Physiological differences across populations reflecting early life and later
life nutritional status and later life risk for chronic disease. J Popul Ageing 8(1–2):51–69.
https:// doi.org/10.1007/s12062-014-9109-4
Crimmins EM, Kim JK, Seeman TE (2009) Poverty and biological risk: the earlier ‘‘aging’’ of
the poor. J Gerontol A Biol Sci Med Sci 64:286–292. https://doi.org/10.1093/gerona/gln010
Cutler DM, Lleras-Muney A (2006) Education and health: evaluating theories and evidence.
National Bureau of Economic Research, Cambridge. https://doi.org/10.3386/w12352
Geronimus AT, Hicken M, Keene D, Bound J (2006) ‘‘Weathering’’ and age patterns of
allostatic load scores among blacks and whites in the United States. Am J Public Health
37
96:826–833. https://doi.org/10.2105/AJPH.2004.060749
Gong P, Liang S, Carlton EJ, Jiang Q, Wu J, Wang L, Remais JV (2012) Urbanization and health
in China. Lancet 379:843–852. https://doi.org/10.1016/S0140-6736(11)61878-3
Hayward MD, Miles TP, Crimmins EM, Yang Y (2000) The significance of socioeconomic
status in explaining the racial gap in chronic health conditions. Am Sociol Rev 65:910–930.
https://doi.org/10.2307/2657519
Hou X (2008) Urban–rural disparity of overweight, hypertension, undiagnosed hypertension, and
untreated hypertension in China. Asia Pac J Public Health 20:159–169.
https://doi.org/10.1177/ 1010539507312306
Hu X, Cook S, Salazar MA (2008) Internal migration and health in China. Lancet 372:1717–
1719. https://doi.org/10.1016/S0140- 6736(08)61360-4
Jia J, Wang F, Wei C, Zhou A, Jia X, Li F et al (2014) The prevalence of dementia in urban and
rural areas of China. J Alzheimer’s Dis 10:1–9. https://doi.org/10.1016/j.jalz.2013.01.012
Kaneda T, Zimmer Z, Zhang K (2010) Explaining the urban–rural divergence in disability
among older Chinese using a large scale national dataset. Institute of Public and
International Affairs (IPIA) Working Paper Series. IPIA, University of Utah, Salt Lake
City, Utah
Li Y, Dorsten L (2010) Regional and urban/rural differences of public health in China. Glob J
Health Sci 2(1):20–30. https://doi.org/10. 5539/gjhs.v2n1p20
Li LW, Liu J, Xu H, Zhang Z (2016) Understanding rural–urban differences in depressive
symptoms among older adults in China. J Aging Health 28(2):341–362.
https://doi.org/10.1177/ 0898264315591003
Mirowsky J, Ross CE (2003) Education, social status, and health. Transaction Publishers, New
38
York
National Bureau of Statistics of China (2016) China statistical yearbook 2016. China Statistics
Press, Beijing
Ross CE, Wu C (1995) The links between education and health. Am Sociol Rev 60:719–745.
https://doi.org/10.2307/2096319
Seeman TE, Singer BH, Rowe JW, Horwitz RI, McEwen BS (1997) Price of adaptation—
allostatic load and its health consequences: MacArthur studies of successful aging. Arch
Intern Med 157:2259–2268. https://doi.org/10.1001/archinte.1997. 00440400111013
Seeman TE, McEwen BS, Rowe JW, Singer BH (2001) Allostatic load as a marker of
cumulative biological risk: MacArthur studies of successful aging. Proc Natl Acad Sci
98:4770–4775. https://doi.org/10.1073/pnas.081072698
Smith JP (2004) Unraveling the SES: health connection. Popul Dev Rev 30:108–132.
Smith JP, Tian M, Zhao Y (2013) Community effects on elderly health: evidence from CHARLS
national baseline. J Econ Ageing 1:50–59. https://doi.org/10.1016/j.jeoa.2013.07.001
Sorlie PD, Backlund E, Keller JB (1995) US mortality by economic, demographic, and social
characteristics: the national longitudinal mortality study. Am J Public Health 85(7):949–956
Stringhini S, Bovet P (2017) Socioeconomic status and risk factors for non-communicable
diseases in low-income and lower-middle income countries. Lancet Glob Health 5:e230–
e231. https://doi. org/10.1016/S2214-109X(17)30054-2
U.S. Department of Health and Human Services, U.S. Department of Agriculture (2015) Dietary
guidelines for Americans 2015–2020. https://health.gov/dietaryguidelines/2015/guidelines/.
Accessed 29 Oct 2017
Wang H, Zhai F (2013) Programme and policy options for preventing obesity in China. Obes
39
Rev 14:134–140. https://doi.org/10.1111/ obr.12106
Wang Y, Mi J, Shan XY, Wang QJ, Ge KY (2007) Is China facing an obesity epidemic and the
consequences? The trends in obesity and chronic disease in China. Int J Obes 31:177–188.
https://doi. org/10.1038/sj.ijo.0803354
Yang G, Kong L, Zhao W, Wan X, Zhai Y, Chen LC, Koplan JP (2008) Emergence of chronic
non-communicable diseases in China. Lancet 372:1697–1705.
https://doi.org/10.1016/S0140- 6736(08)61366-5
Yen IH, Michael YL, Perdue L (2009) Neighborhood environment in studies of health of older
adults: a systematic review. Am J Prev Med 37:455–463.
https://doi.org/10.1016/j.amepre.2009.06.022
Zhao Y, Strauss J, Yang G, Giles J, Hu P, Hu Y, Lei X, Liu M, Park A, Smith PJ, Wang Y
(2013) China Health and Retirement Longitudinal Study 2011–2012 national baseline users’
guide. National School of Development, Peking University, Beijing
Zhao Y, Crimmins E, Hu P, Hu Y, Ge T, Kim J-K, Strauss J, Yang G, Yin X, Wang Y (2014a)
China Health and Retirement Longitudinal Study—2011–2012 national baseline blood data
user’s guide. School of National Development, Peking University, Beijing
Zhao Y, Hu Y, Smith JP, Strauss J, Yang G (2014b) Cohort profile: the China Health and
Retirement Longitudinal Study (CHARLS). Int J Epidemiol 43:61–68.
https://doi.org/10.1093/ ije/dys203 Zhao Y, Crimmins EM, Hu P, Shen Y, Smith JP, Strauss
J, Wang Y, Zhang Y (2016) Prevalence, diagnosis, and management of diabetes mellitus
among older Chinese: results from the China Health and Retirement Longitudinal Study. Int
J Public Health 61:347–356. https://doi.org/10.1007/s00038-015-0780-x
40
TABLES/FIGURES

Table 1. Biological Risk Indicators, China Health and Retirement Longitudinal Study,
China, 2011-12
High-Risk Criteria % High Risk, weighted
High systolic blood pressure ≥ 140 mm Hg 29.0
High diastolic blood pressure ≥ 90 mm Hg 13.6
Low diastolic blood pressure < 60 mm Hg 7.7
Rapid pulse rate at 60sec ≥ 90 5.5
High total cholesterol ≥ 240 mg/dL 10.9
Low HDL cholesterol < 40 mg/dL 25.8
High LDL cholesterol ≥ 160 mg/dL 10.5
High triglycerides ≥ 200 mg/dL 13.9
High plasma glucose ≥ 126 mg/dL 11.8
Obesity (High BMI) ≥ 30 4.9
High C-reactive protein ≥ 3.0 mg/L 18.7
41

Table 2. Sample Characteristics, China Health and Retirement Longitudinal Study, China, 2011-12 (Weighted)

Full Sample

Rural Residency

Urban Residency,
Rural Hukou

Urban Residency,
Urban Hukou

N=8,499

N=5,452 (54%)

N=1,760 (23%)

N=1,287 (23%)

% 95% CI

% 95% CI

% 95% CI

% 95% CI
Age, Mean (SD) 59.7 (9.9)

59.9 (10.0)

58.1 (9.9)

60.8 (9.4)
Female 52.2 50.8 53.7

51.3 50.1 52.6

54.4 52.0 56.8

52.2 46.5 57.8
Education (N=8,497)

Illiterate 25.8 23.4 28.4

34.7 32.3 37.2

23.5 19.3 28.3

7.7 5.6 10.4
Literate, no formal schooling 17.4 15.7 19.2

20.2 18.6 22.0

19.0 15.7 22.7

9.3 6.7 12.8
Primary school 22.1 20.5 23.8

21.8 20.1 23.6

23.8 20.1 27.9

21.1 14.7 29.4
Junior/secondary school or higher 34.7 31.7 37.9

23.3 21.3 25.4

33.8 25.7 42.9

61.8 54.8 68.5
Per capita expenditure (N= 8,462)

Top tertile 41.3 36.8 45.6

27.9 25.4 30.5

41.5 33.6 49.9

71.3 64.4 77.3
2nd tertile 30.4 28.0 33.0

33.9 31.8 36.0

33.0 28.0 38.4

20.0 15.8 24.9
42
Bottom tertile 28.5 25.7 31.4

38.2 35.5 41.1

25.5 20.6 31.2

8.8 6.3 12.0
Ever smoke (N= 8,497) 39.5 37.5 41.5

40.8 39.2 42.5

40.6 37.1 44.2

35.4 27.9 43.7
Drinking (N=8,497)

Non-drinkers 80.4 77.9 79.9

79.7 78.0 81.3

82.5 78.9 85.6

79.9 70.7 86.8
Moderate drinkers 17.6 15.5 17.4

17.6 16.3 19.1

15.7 12.8 19.0

19.5 12.7 28.8
Heavy drinkers 2.0 1.65 2.7

2.7 2.2 3.4

1.9 1.3 2.7

0.6 0.3 1.1
Physical activity (N= 3,619)
Sedentary 37.7 33.0 42.7 28.2 25.1 31.6 44.1 30.0 59.3 54.1 47.6 60.5
Moderate 30.0 27.1 33.1 28.3 25.6 31.3 28.6 21.0 37.6 36.0 30.3 42.2
Vigorous 32.3 28.9 35.8

43.4 39.9 47.0

27.3 19.8 36.4

9.9 6.8 14.1
Community infrastructure

Least deficient 27.2 21.2 34.2

2.0 0.6 6.1

35.6 24.2 48.8

75.6 66.2 83.0
Somewhat deficient 23.9 19.5 29.0

19.7 14.6 26.2

37.6 27.3 49.1

20.4 13.8 29.1
Deficient 25.2 20.7 30.3

39.2 32.4 46.5

16.9 10.4 26.3

1.8 0.7 4.1
Most deficient 23.7 19.3 28.7

39.1 32.3 46.4

10.0 5.1 18.4

2.2 0.5 9.6
Number of biological risk factors, Mean (SD) 1.5 (1.4)

1.4 (1.3)

1.6 (1.5)

1.8 (1.5)

43

Table 3. Rate Ratios from Multilevel Negative Binomial Models Estimating Number of Biological Risks
China Health and Retirement Longitudinal Study, China, 2011-12

Model 1

Model 2

Model 3

Model 4

Rate ratio S.E.

Rate ratio S.E.

Rate ratio S.E.

Rate ratio S.E.
Urban-Rural (ref=rural residency)

Urban residency, rural hukou 1.09

(0.10)

1.11

(0.09)

1.11

(0.07)

1.09

(0.06)
Urban residency, urban hukou 1.32 *** (0.01)

1.36 *** (0.01)

1.19 * (0.08)

1.16

(0.12)
Age 1.01 *** (0.00)

1.01 *** (0.00)

1.01 ** (0.00)

1.01 ** (0.00)
Female 1.04

(0.03)

1.03

(0.04)

1.06

(0.06)

1.07

(0.06)
Education (ref=Illiterate)

Literate without schooling

0.97

(0.04)

0.98

(0.05)

0.98

(0.05)
Primary school

1.01

(0.04)

0.98

(0.05)

0.98

(0.05)
Junior/secondary school or higher

0.90 ** (0.03)

0.88 * (0.05)

0.88 * (0.05)
Per capita expenditure (ref=bottom tertile)

Top tertile

1.00

(0.03)

0.98

(0.04)

0.98

(0.04)
44
2nd tertile

1.01

(0.03)

0.99

(0.05)

0.99

(0.05)
Ever smoke

1.11 * (0.06)

1.11 * (0.06)
Drinking (ref= Non-drinkers)

Moderate drinkers

1.03

(0.06)

1.03

(0.06)
Heavy drinkers

0.97

(0.12)

0.96

(0.12)
Physical activities (ref=sedentary)
Moderate physical activities

0.94

(0.04)

0.94

(0.04)
Vigorous physical activities

0.81 *** (0.04)

0.82 *** (0.04)
Community infrastructure (ref=least deficient)

Somewhat deficient

1.14

(0.11)
Deficient

1.07

(0.11)
Most deficient

1.02

(0.10)
Community-level variance 0.05

(0.01)

0.05

(0.01)

0.10

(0.05)

0.09

(0.04)
Note: Model 4 includes a category indicating respondents were missing on the community infrastructure variable.
Significance: *p<0.05, **p<0.01, ***p<0.001

45
APPENDIX
Selection of Analytic Sample and Missing Data Analysis
For the China Health and Retirement Longitudinal Study (CHARLS), 17,708 respondents
were interviewed in 2011. Our analytic sample for this study includes 8,499 persons, 50% of the
total respondents. The reasons respondents are missing from our analytic sample are shown in
Figure A below. CHARLS interviews individuals 45+ and their spouses regardless of age. We
first excluded those who were below age 45 when interviewed and those who did not provide age
or their birth dates (N=418). Among the 17,290 age-eligible respondents, the primary reason that
individuals were not included in the analytic sample was that they did not provide blood in a
subsequent visit to a clinic (N=7,362). In addition, the analytic sample includes only those who
fasted overnight (a reduction of 1,009), and those with no data missing on any of the biomarkers
(N=407) and key covariates (N=13).

46

Figure A. Sample Inclusion Criteria

All Respondents in China Health and
Retirement Longitudinal Study 2011-12
(N=17,708)
Respondents aged 45+
(N=17,290)
Respondents who provided blood samples
(N=9,928)
Fasting samples
(N=8,919)
Analytic sample
(N=8,499)
407 respondents have missing
values for at least one of the
included biomarkers;

13 respondents have missing
values for gender and urban-rural.
47
We ran a weighted logistic regression using the CHARLS 2011 person-level weights to
examine whether the respondents who were excluded from this study differed from the analytic
sample. Odds ratios and significance are reported in Table A1.
Table A1. Weighted Logistic Regression Model Predicting Being Excluded from the Analytic
Sample, China Health and Retirement Longitudinal Study, China, 2011-12 (N=17,708)

Odds Ratios (SE)
Age 1.00 (0.00) ***
Female 0.85 (0.05) *
Education (ref=Illiterate)

Literate without schooling 0.88 (0.07)

Primary school 0.90 (0.08)

Junior school or higher 1.05 (0.09)

Urban-Rural (Ref=Rural residency)

Urban residency, rural hukou 1.44 (0.20) **
Urban residency, urban hukou 2.25 (0.23) ***
Significance: * p<0.05, **p<0.01, ***p<0.001

Respondents who were excluded from the analytic sample are more likely to be male and
urban residents. No significant age or educational differences are observed.
Individuals were excluded from the analytic sample primarily because they did not
provide a blood sample. In our analysis, we applied the biomarker weight to account for the fact
that men and urban respondents were less likely to provide blood so that they are more likely to
be excluded from the final analytic sample.
48
We then excluded those who did not provide a blood sample and compared the
characteristics of those who were in the sample to the characteristics of those who were excluded
from the analytic sample. We applied the biomarker weights to account for nonresponse in blood
data collection. We also conducted a stratified analysis of missing data by urban-rural status.
Results are reported in Table A2.

Table A2. Weighted Logistic Regression Model Predicting Being Excluded from Analytic
Sample, China Health and Retirement Longitudinal Study, 2011-12
Full Sample
Rural
residency
Urban
residency
rural hukou
Urban
residency
urban hukou
OR (SE) OR (SE) OR (SE) OR (SE)
Age 0.99 (0.01)* 0.99 (0.01)* 0.99 (0.02) 1.00 (0.01)
Female 0.94 (0.07) 0.99 (0.08) 0.98 (0.13) 0.93 (0.15)
Education (ref=Illiterate)
Literate without schooling 0.93 (0.11) 1.01 (0.12) 0.65 (0.32) 0.67 (0.32)
Primary school 0.88 (0.14) 1.05 (0.15) 0.78 (0.25) 0.59 (0.37)
Junior school or higher 0.90 (0.14) 1.00 (0.14) 0.64 (0.33) 0.79 (0.36)
Urban-Rural (Ref=Rural residency)
Urban residency, rural hukou 0.89 (0.16)
Urban residency, urban hukou 1.73 (0.26) ***
Significance: * p<0.05, **p<0.01, ***p<0.001
49
After applying the biomarker weights, we find that advanced age is associated with
somewhat lower probability of being excluded; urban residents with urban hukou are more likely
to be excluded compared to rural residents. Among the rural population, only being older is
associated with a slightly lower probability of being excluded from the analytic sample
(OR=0.99, significant at 5% level). Among the two urban-residency groups, the illiterate persons
have a lower probability of being excluded compared to those who were illiterate, but the
differences are not statistically significant due to relatively smaller sample size and large
standard errors.

Because questions on physical activities were asked only for a random half sample,
Models 3 and 4 use respondents in this random half sample. We conducted a logistic regression
to examine whether respondents who answered the question on physical activities differ from
those who did not receive the questions and thus were excluded from Models 3 and 4. Odds
ratios are reported in Table A3.

50
Table A3. Weighted Logistic Regression Model Predicting Being Excluded from Models 3 & 4,
China Health and Retirement Longitudinal Study, China, 2011-12
Odds Ratio (SE)
Age 1.00 (0.00)

Female 0.98 (0.08)

Education (ref=Illiterate)

Literate without schooling 1.02 (0.09)

Primary school 1.06 (0.13)

Junior school or higher 1.04 (0.11)

Urban-Rural (Ref=Rural residency)

Urban residency, rural hukou 0.86 (0.11)

Urban residency, urban hukou 1.14 (0.24)

Per capita expenditure (ref=bottom tertile)
Top tertile 0.84 (0.07) *
2nd tertile 0.97 (0.08)

Ever smoke 1.06 (0.13)

Drink (ref=nondrinkers)

Moderate drinkers 0.98 (0.12)

Heavy drinkers 1.22 (0.22)

Community infrastructure (ref=least deficient)
Somewhat deficient 0.81 (0.12)

Deficient 0.91 (0.13)

Most deficient 1.01 (0.15)

51
Number of biological risks 1.06 (0.04)
Significance: *p<0.05

Respondents who were excluded from Models 3 and 4 were not significantly different
from those who were in the first two models, except that those in the top tertile of per capita
expenditures were more likely to provide information on functioning. However, as shown in
Table 3 (Models 3 and 4), per capita expenditure is not significantly correlated with biological
risk.

52
Chapter 3: Childhood Adversity and Cardiovascular and Metabolic Risk in
Old Age: Evidence from CHARLS

Yuan S. Zhang, John Strauss, Peifeng Hu, Xinxin Chen, Qinqin Meng, Yafeng Wang, Yaohui
Zhao, Eileen M. Crimmins

ABSTRACT
Childhood adversity may contribute to the development of cardiovascular and metabolic
diseases in later life. However, most existing evidence is from developed countries, with little
attention to developing countries where the life experiences of older persons have differed from
those of their counterparts in developed countries. This study utilizes data on persons aged 45
and above from the China Health and Retirement Longitudinal Study (CHARLS) to examine the
associations of a number of childhood experiences on cardiovascular and metabolic risk in later
life. We do not find convincing evidence for strong links between childhood experiences and
cardiovascular and metabolic risk in China. We do find adult life circumstances, particularly
exposure to urban environments, relate to both cardiovascular and metabolic risk. Our focus on
China provides a more global perspective on the long-lasting effects of early life experiences on
old-age health. Our findings highlight the complexity of this relationship, suggesting the
examination of the effects of childhood conditions on old-age health in developing countries
deserves deeper scrutiny.

53
INTRODUCTION
An epidemic of cardiovascular and metabolic diseases is placing a large and growing
social and economic burden on low- and middle- income countries (LMIC) (Gaziano, 2007).
Cardiovascular and metabolic health at older ages results from a lifetime of experiences and
exposures to a variety of physical, social, and psychosocial factors (Kuh, Ben-Shlomo, Lynch,
Hallqvist, & Power, 2003). Understanding the early-life origins as well as the effect of adult life
conditions on cardiovascular and metabolic health in a developing country context has
significant implications for promoting healthy aging where cohorts of the aging population have
experienced harsher conditions throughout life course and are facing greater disease burden than
their counterparts in developed countries. This study aims to examine the associations of a
comprehensive set of indicators of childhood and later life experiences to cardiovascular and
metabolic risk in old age in China which has had a unique history among developing countries
but has gone through similar economic and epidemiological change as much of the developing
world.

Early-life Adversities and Later-life Health
Adversities and disadvantages in early-life haven been associated with lower
socioeconomic status (SES), greater exposure to adversity, and poor health behaviors in adult life
(Ben-Shlomo & Kuh, 2002). Significant research has provided evidence that disadvantage across
the life course is linked to poorer health and higher mortality in adult life (Hayward & Gorman,
2004; O’Rand & Hamil-Luker, 2005). Recent research from Western countries has focused on
exploring the underlying physiological pathways whereby life-course experiences are translated
into biological risk for developing chronic diseases, disability, and mortality in later-life.
54
Evidence has mounted that adverse childhood experiences can cause enduring changes in
cardiovascular, nervous, endocrine, and immune systems, that have lasting effects on
downstream health outcomes (Danese & McEwen, 2012). For instance, retarded fetal and infant
growth has been shown to be associated with elevated blood pressure and impaired glucose
tolerance in adulthood (Barker, Bull, Osmond, & Simmonds, 1990; Hales et al., 1991).
Undernutrition in childhood can change metabolism and central endocrine regulatory
mechanisms, resulting in an increased risk of metabolic syndrome (Gluckman & Hanson, 2004).
Low socioeconomic background and exposure to adverse childhood experiences have been
related to increases at older ages in expression of proinflammatory genes and reduced expression
of antiviral genes (Cole et al., 2015; Miller et al., 2009) as well as increased allostatic load and
cardiometabolic risk (Friedman, Karlamangla, Gruenewald, Koretz, & Seeman, 2015; Non et al.,
2014). Historical data indicate that high environmental exposure to infections and inflammation
early in life damages organs during development, affects physiological development, and can
elevate the risk of the development of chronic diseases in old age as high infection requires more
resources for maintenance and repair leaving fewer for physical growth and development
(Crimmins, 2015). It is suggested that the long-term consequences of adverse experiences in
early life may be more profound in LMIC as adversities are likely to be frequent and severe, and
the resources for remediation are more limited compared to high-income countries (Currie &
Vogl, 2013).
The dynamic interplay between social and biological processes across the lifecycle is
likely to vary with the historical and social context in which people live and age. Previous
findings from developed countries may not be reproduced in developing countries where
economic, demographic, and social experiences differ and where change has been relatively
55
rapid. There are a number of reasons to expect a different and more complicated picture in a
setting such as China. First, the economic, epidemiological, and nutritional conditions older
Chinese persons have experienced over their lives dramatically differ from those in developed
countries. Older persons in China generally lived their early years in a highly infectious
environment without adequate sanitation and health care. They then may have experienced rapid
socio-economic and epidemiological transitions throughout later childhood and through
adulthood. Subsequently increased food intake resulting from economic growth certainly has a
positive impact on health, but it may also have adverse consequences for those who were
deprived of adequate nutrition in childhood (Gluckman et al., 2009). We may expect that those
who grew up in families without sufficient food (indicated by having family members
experienced starvation) in their childhood could have a higher metabolic risk in later life.
Second, it is also true that the severity and types of early-life adversity differ across developing
and developed countries. The current cohort of older adults in LMIC has had much worse
childhood conditions on a broad scale. Many individuals have experienced multiple wars,
rampant infectious diseases, hunger or even famine, and a severe lack of educational
opportunities (Zeng, Gu, & Land 2007). Notably, almost all older Chinese experienced the 1959-
1961 Chinese Great Famine at some point in their earlier life, but then subsequently experienced
rapid improvement in nutrition. Poor nutrition at certain points in early life and overnutrition in
later life may place the older Chinese at a higher level of risk for cardiovascular and metabolic
health (Monteiro, Conde, & Popkin, 2004). We expect that these exposures will lead to some
uniqueness in the importance of childhood factors for later-life health among older Chinese
compared to those reported for other countries. For instance, the political status of a family is an
important dimension of socioeconomic status in China (Zhao, Zhou, Tan, & Lin, 2018). Older
56
Chinese have gone through a variety of social-political movements. Those who were born into
families with unfavorable backgrounds have generally experienced more stressful events and
adverse psychosocial experiences which have been associated with subsequent adult
inflammation (Danese, Pariante, Caspi, Taylor, & Poulton, 2007). In this study, we expect to find
that individuals whose parents were classified as members of “bad” or “undesirable classes” in
the 1950s to the1970s may have higher cardiovascular and metabolic risk. Distinct pattern of
associations between risk factors is the third reason we expect findings for China to differ from
those of other countries. Unlike high-income countries, sedentary lifestyles, smoking, drinking,
and diets high in fat, sugar, and refined carbohydrates are more prevalent among individuals with
higher socioeconomic status in China as in many LMIC (Chen et al. 2010; Kim et al. 2004);
therefore, we expect to find that a weak or even positive association between SES and
cardiovascular and metabolic risk. Fourth, older Chinese have experienced a transition from a
socialist society to an emerging capitalist society. In Maoist China when equality was
emphasized, differentials in income and social services were significantly reduced, resulting in
low socioeconomic stratification by international standards (Chen, Yang, & Liu, 2010). After
China’s economic reform in the 1980s, inequality in income, education, and access to health
services sharply increased. This transition may suggest a much weaker correlation between
childhood and adult SES; therefore, it is possible that adult SES is more consequential for later-
life health than childhood SES. China may also differ from other countries in the effects of
gender on health. The older population examined in this study were born and grew up in a
climate where women were largely devalued due to long-standing cultural values derived from
Confucius and Mencius (Yu & Sarri, 1997). These older women have much lower levels of
education than their age contemporary males which may result in less knowledge about risk
57
factors, awareness of chronic conditions, and ability to manage health. As a result, the
consequences of childhood adversity may be more pronounced in women.
It is worth noting the true relationship between early-life experiences and later-life health
may be difficult to measure, given significant mortality selection. Individuals who had severe
adversities in early years of life and those with poor childhood health may have experienced
mortality prior to this study. Those in the study sample are a group of selected individuals who
managed to survive hardships, such as wars, famine, and political movements. The long-term
health consequences of severe adversities in early life may not be fully observed. Given that
much of the population is exposed to adverse conditions and mortality selection could be
significant, the observed relationship early-life experiences and later-life health could be less
pronounced.

Early-life Experiences and Later-life Health among the Older Chinese Population
The uniqueness of the Chinese context suggests value in extending the investigation of
childhood experiences and later-life health to the elderly in China. The literature on the
associations of childhood conditions on later-life health in China has been inconsistent,
depending on measures of early-life circumstances, child health and health at old ages. Shen &
Zeng (2014) documented that childhood adversity, in terms of birthplace and father’s occupation
and education, is directly associated with increased survival at old ages, suggesting mortality
selection overpowers the fetal origin hypothesis and may offset the adverse effects of childhood
adversity on health. Smith et al. (2012) used self-rated child health and adult height as measures
of childhood health and found self-rated child health is associated with self-rated current health
among Chinese women but not for men; and for both men and women height is unexpectedly
58
positively associated with hypertension. But Huang et al (2007) failed to find a relationship
between height and hypertension. Although height is suggested as a good proxy of early-life
health and environment, it does not allow identification of the specific aspects of childhood
circumstances with long-term effects. Wen and Gu (2011) found having both parents alive at age
10 was associated with a lower risk of disability and mortality; being born in urban areas is
associated with a lower likelihood of reporting poor health; arm length is negatively associated
with mortality, and having access to health care is not related to any other health outcomes
examined but only poor self-rated health; no relationships of father’s occupation and being
hungry when going to bed to any health outcomes were found. Huang and Elo (2007) found
childhood nutritional status but not childhood SES is predictive of lower mortality at old ages.
(Schooling et al., 2011) studied parental death during childhood and adult cardiovascular risk
and found parental death did not associated with blood pressure, glucose, and cholesterol; but
associated with lower body mass index. Taken together, many existing studies investigated the
links between different types of childhood adversity and a variety of health outcomes and the
results indicate that the associations between childhood circumstances and health at old ages are
not uniform across different types of childhood indicators and various dimensions of health.
Existing literature on childhood experiences and physiological health in adulthood in China
mainly focuses on the effects of fetal and infant exposure to the Chinese famine of 1959 to 1961
with the exception of work by Crimmins (2015) who found being stunted was related to lower
levels of a ratio of lipids (HDL/Total cholesterol) but higher levels of pulse pressure. Numerous
studies have documented that those who were exposed to the Chinese famine in early-life have a
higher risk of overweight (Wang, Wang, Kong, Zhang, & Zeng, 2010), hypertension (Li et al.,
2011), diabetes (Xu, Zhang, Li, & Liu, 2018), hyperglycemia (Li et al., 2010), dyslipidemia
59
(Wang, Li, Yang, Ma, & Zou, 2017), and metabolic syndrome (Zheng et al., 2011) in adulthood;
however, the reliability of these estimates is questionable because the reported effects of famine
are have not properly controlled for age differences between famine and post-famine births (Li
& Lumey, 2017). Despite the fact that exposure to famine is an important aspect of early-life
experience for the current older cohort of Chinese, it is only one dimension of childhood
exposures. In this study, we used additional indicators of childhood that are culture, cohort, and
context- specific. In addition, the use of measured biological markers for physiological health is
particularly important when studying a population such as China with both limited and
differential awareness of the presence of chronic diseases (Lewington et al., 2016; Lu et al.,
2017; Y. Zhao et al., 2016). To our knowledge, this is the first study examining the relationship
between multiple dimensions of childhood experiences and physiological health using a
nationally representative sample in China.

METHODS
Study population
Data come from the China Health and Retirement Longitudinal Study (CHARLS), an
ongoing nationally representative longitudinal study of Chinese residents aged 45 or older
administered by Peking University. CHARLS adopted a stratified multistage probability
sampling design in fielding a national baseline survey between 2011 and 2012 (Zhao, Hu,
Smith, Strauss, & Yang, 2014). After applying sampling weights, the baseline sample
demographics closely match those of the population census in 2010 (Zhao et al., 2014). Follow-
up waves occur approximately every two years. The study protocol was approved by the ethical
60
review committee (IRB) of Peking University. Details on the CHARLS survey design and
sampling procedures are available from Peking University (http://charls.pku.edu.cn/en).

For this analysis, we used data from the 2015 National Survey and the 2014 Life History
Survey. The 2015 national survey interviewed 19,566 individuals aged 45 and above.
Anthropometric measurements and blood pressure were collected in the household at the time of
individual interviews. Respondents were asked to then come to centralized locations for the
collection of blood. For those who could not go to a central laboratory, blood was collected at
their homes. Most respondents who completed the survey, 94%, provided a blood sample.
Respondents were asked to fast overnight, but blood was taken even if respondents did not fast.
Blood was prepared at the local health facilities and shipped to Beijing where blood-based
biomarkers were assayed from plasma. Further details of blood collection and processing
procedures are available from the 2015 National Wave 3 Blood Data Users’ Guide (Zhao et al.
2019).
Information about early life comes from the 2014 Life History Survey, a supplemental
wave to retrospectively collect information about family background and circumstances
including wealth and poverty, as well as health and health care when the respondent was a child.
Information about historical events specific to the Chinese context was also included, such as
experiences during the Cultural Revolution and the Great Famine of 1959-1961 (Chen, Smith,
Strauss, Wang, & Zhao, 2015).

61
Measures
Cardiovascular risk was measured by four biological indicators of cardiovascular health
including systolic blood pressure, diastolic blood pressure, pulse rate at 60 seconds, and C-
reactive protein. Metabolic risk was measured by seven metabolic components including
elevated levels of total cholesterol, high- and low- density lipoprotein cholesterol, plasma
glucose, triglycerides, glycosylated hemoglobin (HbA1c), large waist circumference, and
obesity, defined as Body Mass Index (BMI) ≥ 30 kg/m
2
. A dichotomous indicator for each
indicator indicates “high risk” or “not high risk” based on clinical cutoff values (Table 1). The
prevalence of high-risk levels for each biomarker is presented in the right two columns in Table
1. More than a quarter of men and women have high systolic blood pressure (26.8% for women,
29.7% for men); about 9% of women and 14% of men have high diastolic blood pressure; about
7% have a rapid pulse rate, and one in five has elevated C-reactive protein. Regarding metabolic
risk, about 9% of women and 5% of men have high total cholesterol. The percentage with low
HDL cholesterol is about 20% in men but 10% in women. There are relatively few individuals
with high LDL cholesterol (3-4%) or who are obese (6.5% in women and 4% in men). Both high
triglycerides and high HbA1c are found in about 15% of the sample. Plasma glucose is high in
17% of women and 15% of men. Women have a much higher likelihood of having large waist
circumference than men (44% for women and 8% for men). Summary measures of risk were
calculated by summing the numbers of markers indicating high risk (cardiovascular risk ranges
from 0 to 4, and metabolic risk ranges from 0 to 6).
[Table 1 here]
Childhood experiences were measured using the retrospective report of
conditions/events in childhood and early adolescence respondents reported in the CHARLS Life
62
History Survey in 2014. Our measures reflect three domains of childhood experiences: family
conditions, health related infrastructure before the age of 15, and health and healthcare when a
child. Four dummy variables are used that measure other aspects of family. These are whether
mother is illiterate, parent belongs to the bad/undesirable political classes, defined as landlords,
rich farmers, capitalists, counter-revolutionaries, bad elements, and rightists; whether parent had
drug, alcoholic, or gambling problems or was depressed for most of the respondent’s childhood,
or hit respondent very often; and whether family members experienced starvation and respondent
was 3 or younger during 1958 to 1962, considering the in utero exposure during the Chinese
Famine of 1959-1961. Environmental hazards to health are indicated by not having clean water,
flush toilet, and electricity in the home before age 15.

Adult circumstances were captured by individual educational attainment: no formal
education (reference group), primary school, junior school, and high school or higher. Health
behaviors were indicated by smoking status: never smoker (reference group), former smoker,
current smoker. Usual place of residence combined with assigned official place of registration
(called “hukou”) divides that population into the rural residents, the urban residents with urban
hukou, and urban residents with rural hukou. The small percentage who have urban hukou but
live in rural areas are classified with the rural residents category. Our urban-rural variable
separates those who live in urban areas by their hukou which defines their access to public
services and benefits. Body Mass Index (BMI) is calculated as measured weight divided by
measured height, and then classified into four groups: underweight (BMI < 18.5), normal weight
(18.5 kg/m
2
≤ BMI ≤ 24.9 kg/m
2
), overweight (25 ≤ BMI ≤ 29.9), and obese (BMI  30).
63
Because obesity is included in the metabolic risk measure, BMI is only in the final model
predicting cardiovascular risk, but not in the model predicting metabolic risk.

Statistical Analysis
We used STATA 14.1 for all analyses. Sample weights were applied to account for the
complex sample design of the CHARLS, nonresponses to the main household interview, and
nonparticipation of the collection of blood sample. Because our dependent variables are over-
dispersed count variables, we employed negative binomial regression models to examine the
associations between childhood and adult circumstances and cardiovascular and metabolic risk.
We split the sample by gender and estimate the models for men and women separately. We first
examine the age-patterns of cardiovascular and metabolic risk. To account for potential nonlinear
relationships between age and cardiovascular and metabolic risks, we grouped individuals into
10-years age groups. We added measures of childhood adversity in the second model. The final
model added adult circumstances to examine whether the associations between childhood
experiences and cardiovascular and metabolic risks are changed by including adult
circumstances. In all models, we estimated robust standard errors to account for sample
clustering at the community-level.

RESULTS
Table 2 presents descriptive statistics for the analytic sample. More than 60% of
respondents were between 45 and 64, about a quarter were between 65 and 74, and about 10%
were 75 and older. The majority (66.8% among women and 65.6% among men) live in rural
areas, about 12% live in urban areas but still hold a rural hukou, about 22% live in urban areas
64
with urban hukou. Approximately 52.2% are female. Women have significantly less education
than men: no formal education characterized 47.8% of women versus 21.5% for men. About
20% of men, but only about 11% of women, have a high school or more. A substantial gender
difference in smoking status is found: 93% of women are never smokers while more than half of
men are current smokers. The majority had a normal BMI (55.9% among women, 61.0% among
men), but women are more likely to be overweight and obese.

Most adverse childhood experiences do not differ between men and women, except that
men are more likely to report they had missed school because of health conditions, confined to
bed or hospitalized for more than a month or more than three times before they were age 15.
Men also have a higher probability of reporting that a family member experienced starvation
during 1958-1962 when they were three or younger (22.5% for men and 19.5 for women,
p=0.03). 80% of subjects reported their mother was illiterate, about 8% had parents who
belonged to bad/undesirable classes, about 20% reported that their parents had drug, alcohol or
gambling problems, or were depressed for most of their childhood, or hit them very often. About
60% did not live with clean water, flush toilet and electricity before age 15.
Table 3 reports rate ratios estimated from negative binomial models predicting
cardiovascular risk. Model 1 shows that elevated cardiovascular risk is associated with advanced
age. However, the age-pattern is somewhat different between men and women. For men,
cardiovascular risk increases as age increases; however, for women, cardiovascular risk for those
aged 55 to 74 is increased 20% compared to individuals who were 45 to 54. This gender-specific
pattern persists after childhood and adulthood indicators were added.
65
We found little association of childhood circumstances with cardiovascular risk.
Cardiovascular risk was somewhat elevated among those whose mother was illiterate, but this
association is only significant at the 10% level for women (Rate ratio = 1.15) before adult
circumstances were added to the model and for men (Rate ratio = 1.13) after adult circumstances
were controlled. Having parents belonged to bad/undesirable classes is positively related to
cardiovascular risk among men but the direct of the association is the opposite among women.
Parents had behavioral problems or was depressed is associated with higher cardiovascular risk
in both men and women, but the associations are not statistically significant.
Among indicators of adult circumstances, urban-rural residence and obesity status were
important for cardiovascular health. Those living in urban areas with rural hukou have an
increased cardiovascular risk relative to rural residents (Rate ratio = 1.23 for women, 1.18 for
men). Men and women who were overweight and obese have significantly increased
cardiovascular risk. (Rate ratio = 1.38 for overweight men, 1.50 for overweight women, 1.91 for
obese men, and 1.94 for obese women, p<0.01). We did not find a statistically significant
association between education and cardiovascular risk, although those who have higher school or
higher education have a reduced cardiovascular risk than those without formal education (Rate
ratio = 0.86 for men and 0.96 for women).
Table 4 reported rate ratios estimated from models predicting metabolic risk. We found a
gender-specific pattern of the association of age and metabolic risk. Compared to men 45 to 54,
those 55-64, 65-74, and 75+ have about a 30% increased metabolic risk; but the risk is not
increased continuously with age after age 64. We did not find a significant association between
age and metabolic risk among women. Among indicators of childhood adversity, metabolic risk
is increased among men whose parents belonged to the bad/undesirable classes (Ratio ratio =
66
1.13 in Model 2, 1.12 in Model 3). Among women, those who did not receive a vaccination or
did not have a usual source of care before age 15 have a reduced metabolic risk (Rate ratio =
0.84 in Model 2 with p < 0.05, and 0.87 in Model 3 with p < 0.1). Before we controlled
adulthood circumstances, having an illiterate mother was associated with a lower metabolic risk
among men (Rate ratio = 0.84, p < 0.05); lack of access to clean water, flushing toilet or
electricity was linked to a reduced metabolic risk (Rate ratio = 0.92, p < 0.1) among women.
Compared to childhood experiences, adulthood circumstances are more consequential for
metabolic risk in old age. Education is positively associated with metabolic risk among women –
those who have higher educational attainment have a higher metabolic risk (Rate ratio = 1.12 for
individuals with primary school, 1.25 for individuals with at least a high school education), but
the relationship was very different for men. Men with a high school education have a lower
metabolic risk (Rate ratio = 0.84, p < 0.1). We did not find a significant association between
smoking and metabolic risk among men, but former women smokers have an elevated metabolic
risk compared to nonsmoking women (Rate ratio = 1.48, p<0.01).

CONCLUSIONS AND DISCUSSION
Our study, among middle-aged and older adults in China – a developing and transitional
country, found little association of childhood experiences with cardiovascular and metabolic risk
after age 45 whereas exposure to the urban environment has important associations to
cardiovascular and metabolic risk among the older Chinese population. The weak links of
childhood adversity to cardiovascular and metabolic health differ from the associations often
found in both high-income as well as other developing countries (McEniry & McDermott, 2015;
Moore, Halsall, Howarth, Poskitt, & Prentice, 2001; Schooling et al., 2011).We did not find
67
direct associations between childhood circumstances and cardiovascular and metabolic risk, and
this relationship did not change after adding measures of adult life. Results from this paper
emphasizes the complexity of the consequences of childhood experiences on health. The
inconsistent relationships across populations suggest that life experiences get under the skin
through a combination of social and biological pathways which are substantially influenced by
the historical contexts in which the population ages . As discussed earlier, as China has gone
through dramatic changes in almost all aspects of society, most older Chinese experienced
hardships at younger ages and then experienced China had a phase of unbalanced development
which? led to disparities in income, lifestyle, and resources one could use for managing later life
health. Therefore, childhood and adulthood socioeconomic status have a much weaker
relationship with each other among older Chinese than in other settings. Moreover, because SES
is positively associated with many unhealthy behaviors and obesity in LMIC (Chen et al. 2010;
Kim et al. 2004), those who lived in poor conditions throughout life course are less likely to
engage in unhealthy behaviors such as smoking and a sedentary life style which are risk factors
for cardiovascular and metabolic health in later life. In our analyses, some indicators of
childhood adversity, for instance, not receiving vaccinations and not having a usual place for
health care, are linked to a lower metabolic risk. Certainly, inadequate immunization and
healthcare are not the sources of reduced metabolic risk. We believe this reflects the better
metabolic health of those who are exposed to the rural environment. The different results among
older Chinese relative to other settings might be result from the unique past development
strategies in China (Crimmins, 2015). It is also essential to consider the role of survival selection
in explaining the results. Those with few medical resources and those who experienced extreme
hardships in early life and died at young ages as a result are not in our study sample; therefore,
68
the impacts of the early-life experiences are not measured for these people. Indeed, we only
measure the associations among those who have survived to age 45 and above; so the effects of
early-life adversity are undoubtedly underestimated.
In contrast, adult circumstances seem more influential. Exposure to urban environments
has significant but differential consequences on cardiovascular and metabolic health. Our
analyses showed that urban residents with urban hukou and rural residents have a similar level of
cardiovascular risk whereas subjects living in urban areas but still holding rural hukou have
significantly higher cardiovascular risk. Those who reside in urban areas with rural hukou are
rural-to-urban migrants who likely have less current access to and may underuse health services
(Gong et al., 2012). This could mean that awareness, treatment, and control of cardiovascular
risk is particularly poor among those living in urban areas with rural hukou. The significantly
higher metabolic risk of urban residents with urban hukou may reflect more sedentary behavior
and unhealthy diets (Xi, He, Hu, & Zhou, 2013).
In a rapidly changing country, the impacts of life course experiences and exposure on
health could also be changing across cohorts. As China has experienced significant mortality
reduction, particularly young-age mortality, mortality selection earlier in the life course has
become much less severe among later cohorts. In addition, the social patterning of risk factors
for health is also changing over time in China. As the Chinese population is going through the
nutritional transition, those with higher SES will be the first to engage healthy lifestyles whereas
those with lower SES are likely to adopt unhealthy behaviors and become overweight (Strauss &
Thomas, 2008). Thus, future research should explore the changing impact of social and
behavioral factors in childhood and adulthood on health at old ages.
69
The present study has a few limitations. First, because the individuals, especially the
older individuals, we studied in this paper are survivors of a series of wars, famines, infectious
diseases, and political movements, mortality selection is likely to be significant. Therefore, our
estimates of the associations of childhood adversity and later-life health tend to be conservative.
Second, the potential recall bias of the self-reported childhood circumstances might be larger
among the old and low educated individuals because they are more likely to be cognitively
impaired. Despite these limitations, our study used a large nationally representative sample of
middle-aged and older persons to study the childhood origins of cardiovascular and metabolic
risk. The focus on China provides a more global perspective on the long-lasting effects of early
life experiences on old-age health. Our findings highlight the complexity of this relationship,
suggesting the impacts of childhood adversity is largely affected by social and historical context.

70
REFERENCES
Barker, D. J., Bull, A. R., Osmond, C., & Simmonds, S. J. (1990). Fetal and placental size and
risk of hypertension in adult life. BMJ, 301(6746), 259–262.
https://doi.org/10.1136/bmj.301.6746.259
Chen, X., Smith, J., Strauss, J., Wang, Y., & Zhao, Y. (2015). China Health and Retirement
Longitudinal Study. In Encyclopedia of Geropsychology (pp. 1–8). Springer Singapore.
https://doi.org/10.1007/978-981-287-080-3_42-1
Cole, S. W., Levine, M. E., Arevalo, J. M. G., Ma, J., Weir, D. R., & Crimmins, E. M. (2015).
Loneliness, eudaimonia, and the human conserved transcriptional response to adversity.
Psychoneuroendocrinology, 62, 11–17.
https://doi.org/https://doi.org/10.1016/j.psyneuen.2015.07.001
Crimmins, E. M. (2015). Physiological Differences Across Populations Reflecting Early Life and
Later Life Nutritional Status and Later Life Risk for Chronic Disease. Journal of Population
Ageing. https://doi.org/10.1007/s12062-014-9109-4
Currie, J., & Vogl, T. (2013). Early-Life Health and Adult Circumstance in Developing
Countries. Annual Review of Economics, 5(1), 1–36. https://doi.org/10.1146/annurev-
economics-081412-103704
Danese, A., & McEwen, B. S. (2012). Adverse childhood experiences, allostasis, allostatic load,
and age-related disease. Physiology & Behavior, 106(1), 29–39.
https://doi.org/10.1016/j.physbeh.2011.08.019
Danese, A., Pariante, C. M., Caspi, A., Taylor, A., & Poulton, R. (2007). Childhood
maltreatment predicts adult inflammation in a life-course study. Proceedings of the National
Academy of Sciences, 104(4), 1319–1324. https://doi.org/10.1073/pnas.0610362104
71
Friedman, E. M., Karlamangla, A. S., Gruenewald, T. L., Koretz, B., & Seeman, T. E. (2015).
Early life adversity and adult biological risk profiles. Psychosomatic Medicine, 77(2), 176–
185. https://doi.org/10.1097/PSY.0000000000000147
Gaziano, T. A. (2007). Reducing The Growing Burden Of Cardiovascular Disease In The
Developing World. Health Affairs, 26(1), 13–24. https://doi.org/10.1377/hlthaff.26.1.13
Gluckman, P. D., & Hanson, M. A. (2004). Developmental Origins of Disease Paradigm: A
Mechanistic and Evolutionary Perspective. Pediatric Research, 56, 311. Retrieved from
http://dx.doi.org/10.1203/01.PDR.0000135998.08025.FB
Gong, P., Liang, S., Carlton, E. J., Jiang, Q., Wu, J., Wang, L., & Remais, J. V. (2012).
Urbanisation and health in China. The Lancet, 379(9818), 843–852.
https://doi.org/https://doi.org/10.1016/S0140-6736(11)61878-3
Hales, C. N., Barker, D. J., Clark, P. M., Cox, L. J., Fall, C., Osmond, C., & Winter, P. D.
(1991). Fetal and infant growth and impaired glucose tolerance at age 64. BMJ, 303(6809),
1019–1022. https://doi.org/10.1136/bmj.303.6809.1019
Hayward, M. D., & Gorman, B. K. (2004). The long arm of childhood: The influence of early-
life social conditions on men’s mortality. Demography, 41(1), 87–107.
Kuh, D., Ben-Shlomo, Y., Lynch, J., Hallqvist, J., & Power, C. (2003). Life course
epidemiology. Journal of Epidemiology and Community Health, 57(10), 778 LP-783.
Retrieved from http://jech.bmj.com/content/57/10/778.abstract
Lewington, S., Lacey, B., Clarke, R., Guo, Y., Kong, X. L., Yang, L., … Chen, Z. (2016). The
Burden of Hypertension and Associated Risk for Cardiovascular Mortality in China. JAMA
Internal Medicine, 176(4), 524–532. https://doi.org/10.1001/jamainternmed.2016.0190
Li, C., & Lumey, L. H. (2017). Exposure to the Chinese famine of 1959-61 in early life and
72
long-term health conditions: a systematic review and meta-analysis. International Journal
of Epidemiology, 46(4), 1157–1170. https://doi.org/10.1093/ije/dyx013
Li, Y., He, Y., Qi, L., Jaddoe, V. W., Feskens, E. J. M., Yang, X., … Hu, F. B. (2010). Exposure
to the Chinese famine in early life and the risk of hyperglycemia and type 2 diabetes in
adulthood. Diabetes. https://doi.org/10.2337/db10-0385
Li, Y., Jaddoe, V. W., Qi, L., He, Y., Lai, J., Wang, J., … others. (2011). Exposure to the
Chinese famine in early life and the risk of hypertension in adulthood. Journal of
Hypertension, 29(6), 1085–1092.
Lu, J., Lu, Y., Wang, X., Li, X., Linderman, G. C., Wu, C., … Jiang, L. (2017). Prevalence,
awareness, treatment, and control of hypertension in China: data from 1.7 million adults in a
population-based screening study (China PEACE Million Persons Project). Lancet (London,
England), 390(10112), 2549–2558. https://doi.org/10.1016/S0140-6736(17)32478-9
McEniry, M., & McDermott, J. (2015). Early-Life Conditions, Rapid Demographic Changes, and
Older Adult Health in the Developing World. Biodemography and Social Biology, 61(2),
147–166. https://doi.org/10.1080/19485565.2015.1047488
Miller, G. E., Chen, E., Fok, A. K., Walker, H., Lim, A., Nicholls, E. F., … Kobor, M. S. (2009).
Low early-life social class leaves a biological residue manifested by decreased
glucocorticoid and increased proinflammatory signaling. Proceedings of the National
Academy of Sciences, 106(34), 14716–14721. https://doi.org/10.1073/pnas.0902971106
Monteiro, C. A., Conde, W. L., & Popkin, B. M. (2004). The Burden of Disease From
Undernutrition and Overnutrition in Countries Undergoing Rapid Nutrition Transition: A
View From Brazil. American Journal of Public Health, 94(3), 433–434.
https://doi.org/10.2105/AJPH.94.3.433
73
Moore, S. E., Halsall, I., Howarth, D., Poskitt, E. M., & Prentice, A. M. (2001). Glucose, insulin
and lipid metabolism in rural Gambians exposed to early malnutrition. Diabetic Medicine :
A Journal of the British Diabetic Association, 18(8), 646–653.
Non, A. L., Rewak, M., Kawachi, I., Gilman, S. E., Loucks, E. B., Appleton, A. A., …
Kubzansky, L. D. (2014). Childhood Social Disadvantage, Cardiometabolic Risk, and
Chronic Disease in Adulthood. American Journal of Epidemiology, 180(3), 263–271.
https://doi.org/10.1093/aje/kwu127
O’Rand, A. M., & Hamil-Luker, J. (2005). Processes of Cumulative Adversity: Childhood
Disadvantage and Increased Risk of Heart Attack Across the Life Course. The Journals of
Gerontology: Series B, 60(Special_Issue_2), S117–S124. Retrieved from
http://dx.doi.org/10.1093/geronb/60.Special_Issue_2.S117
Schooling, C. M., Jiang, C., Lam, T. H., Zhang, W., Cheng, K. K., & Leung, G. M. (2011).
Parental Death during Childhood and Adult Cardiovascular Risk in a Developing Country:
The Guangzhou Biobank Cohort Study. PLOS ONE, 6(5), e19675. Retrieved from
https://doi.org/10.1371/journal.pone.0019675
Shen, K., & Zeng, Y. (2014). Direct and indirect effects of childhood conditions on survival and
health among male and female elderly in China. Social Science and Medicine.
https://doi.org/10.1016/j.socscimed.2014.07.003
Strauss, J., & Thomas, D. (2008). Health Over the Life Course. In Handbook of Development
Economics. https://doi.org/10.1016/S1573-4471(07)04054-5
Wang, Y., Wang, X., Kong, Y., Zhang, J. H., & Zeng, Q. (2010). The Great Chinese Famine
leads to shorter and overweight females in Chongqing Chinese population after 50 years.
Obesity (Silver Spring, Md.), 18(3), 588–592. https://doi.org/10.1038/oby.2009.296
74
Wang, Z., Li, C., Yang, Z., Ma, J., & Zou, Z. (2017). Fetal and infant exposure to severe Chinese
famine increases the risk of adult dyslipidemia: Results from the China health and
retirement longitudinal study. BMC Public Health, 17(1), 488.
https://doi.org/10.1186/s12889-017-4421-6
Xi, B., He, D., Hu, Y., & Zhou, D. (2013). Prevalence of metabolic syndrome and its influencing
factors among the Chinese adults: The China Health and Nutrition Survey in 2009.
Preventive Medicine, 57(6), 867–871.
https://doi.org/https://doi.org/10.1016/j.ypmed.2013.09.023
Xu, H., Zhang, Z., Li, L., & Liu, J. (2018). Early life exposure to China’s 1959–61 famine and
midlife cognition. International Journal of Epidemiology, 47(1), 109–120.
https://doi.org/10.1093/ije/dyx222
Zhao, X., Zhou, Y., Tan, H., & Lin, H. (2018). Spillover effects of children{\textquoteright}s
political status on elderly parents{\textquoteright} health in China. Journal of Epidemiology
& Community Health, 72(11), 973–981. https://doi.org/10.1136/jech-2018-210770
Zhao, Y., Crimmins, E. M., Hu, P., Shen, Y., Smith, J. P., Strauss, J., … Zhang, Y. (2016).
Prevalence, diagnosis, and management of diabetes mellitus among older Chinese: results
from the China Health and Retirement Longitudinal Study. International Journal of Public
Health. https://doi.org/10.1007/s00038-015-0780-x
Zhao, Y., Hu, Y., Smith, J. P., Strauss, J., & Yang, G. (2014). Cohort Profile: The China Health
and Retirement Longitudinal Study (CHARLS). International Journal of Epidemiology,
43(1), 61–68. https://doi.org/10.1093/ije/dys203
Zheng, X., Wang, Y., Ren, W., Luo, R., Zhang, S., Zhang, J. H., & Zeng, Q. (2011). Risk of
metabolic syndrome in adults exposed to the great Chinese famine during the fetal life and
75
early childhood. European Journal Of Clinical
Nutritionfile:///Users/Yuanzhang/Downloads/Exposure-to-the-Chinese-Famine-in-Early-
Life-and-the-Risk-of-Hyperglycemia-and-Type-2-Diabetes-in-Adulthood.Bib, 66, 231.
Retrieved from http://dx.doi.org/10.1038/ejcn.2011.161

76
TABLES/FIGURES

Table 1. Biological risk indicators, China Health and Retirement Longitudinal Study,
2015 (N=8,458)

High-risk
criteria
% High risk
Biomarker Women Men
Cardiovascular
Risk
Systolic Blood Pressure ≥ 140 mmHg 26.79 29.69
Diastolic Blood Pressure ≥ 90 mmHg 8.93 13.70
Pulse Rate at 60s ≥ 90 7.16 7.94
C-reactive Protein ≥ 3 mg/L 20.59 21.18
Metabolic Risk
Total Cholesterol ≥ 240 mg/dL 8.56 4.60
HDL Cholesterol < 40 mg/dL 10.27 20.74
LDL Cholesterol ≥ 160 mg/dL 4.10 3.27
Triglycerides ≥ 200 mg/dL 17.57 14.89
Plasma Glucose ≥ 126 mg/dL 9.16 9.15
HbA1c ≥ 6.5% 15.76 14.27
Waist Circumference ≥ 88cm (female) 43.61 -
≥ 102cm(male) - 8.08
Obesity ≥ 30 kg/m
2
6.54 3.86

77
Table 2. Sample Characteristics, China Health and Retirement Longitudinal Study,
2015 (N=8,458)
Women Men p-value
Age, % <.001
45-54 32.2 25.2
55-64 35.4 38.1
65-74 23.1 26.1
75+ 9.3 10.7
Education, % <.001
No formal education 47.8 21.5
Primary school 20.9 27.3
Junior school 20.6 31.4
High school or higher 10.7 19.8
Urban-Rural, % 0.295
Living in rural 66.8 65.6
Rural hukou, living in urban 11.2 13.0
Urban hukou, living in urban 22.0 21.4
Smoking, % <.001
Never smokers 92.6 21.4
Former smokers 2.2 27.6
Current smokers 5.2 51.1
BMI, % <.001
Underweight 4.9 5.6
78
Normal 55.9 61.0
Overweight 32.7 29.5
Obesity 6.5 3.9
Family Background During Respondent's Childhood
Mother was illiterate, % 79.7 80.3 0.605
Parent belonged to the bad/undesirable classes 7.0 8.9 0.103

Parent had drug, alcoholic, or gambling problems/was
depressed/hit R very often 20.2 22.3 0.122

Family members experienced starvation & R<=3
between 1958 to 1962 19.5 22.5 0.030
Infrastructure in Childhood (Before 15)
No clean water/flush toilet/electricity 57.7 61.4 0.054
Health and Health Care in Childhood (Before 15)

Did not receive vaccination/ Did not have a usual source
of care 24.2 21.6 0.170

Missed school because of health issues/confined to bed
or hospitalized for 1mo+ or more than 3times 7.9 10.4 0.005
Note: p-values from the weighted Chi-square tests are reported.

79
Table 3. Relative Ratios from Negative Binomial Models Predicting Cardiovascular Risk, China Health and Retirement Longitudinal Study,
2015 (N=8,458)
Age Only Age + Childhood
Age + Childhood
+Adulthood
Women Men Women Men Women Men
Age (ref=45-54)

55-64
1.02
[0.11]
1.20
[0.11] **

1.01
[0.11]
1.22
[0.11] **

1.02
[0.12]
1.23
[0.11] **

65-74
1.34
[0.14] ***

1.20
[0.10] **

1.28
[0.14] **

1.22
[0.11] **

1.32
[0.15]
** 1.26
[0.12] **

75+
1.67
[0.18] ***

1.15
[0.10]
1.57
[0.18]
**
*

1.18
[0.12] *

1.69
[0.21]
**
*

1.26
[0.14] **

Childhood Adversity
Family Background During Respondent's Childhood

Mother was illiterate

1.15
[0.18] *

1.12
[0.08]
1.12
[0.09]
1.13
[0.07] *

Parent belonged to the bad/undesirable classes

0.85
[0.17]
1.07
[0.11]
0.86
[0.11]
1.03
[0.08]
80

Parent had drug, alcoholic, or gambling problems/was depressed/hit R very often

1.08
[0.07]
1.10
[0.07]
1.05
[0.06]
1.10
[0.07]

Family members experienced starvation & R<=3 between 1958 and 1962

0.97
[0.08]
0.95
[0.07]
0.97
[0.08]
0.94
[0.07]
Infrastructure in Childhood (Before 15)

No clean water/flush toilet/electricity

1.01
[0.05]
0.91
[0.06]
0.96
[0.05]
0.92
[0.06]
Health and Health Care in Childhood (Before15)

Did not receive vaccination/ Did not have a usual source of care

1.07
[0.06]
0.90
[0.08]
1.07
[0.06]
0.93
[0.09]

Missed school for health issues/confined to bed or hospitalized for 1mo+ or more
than 3times
0.95
[0.09]
1.01
[0.10]
0.96
[0.10]
1.05
[0.09]
Adulthood Circumstances
Education (ref=No formal education)

Primary school

0.98
[0.07]
0.99
[0.07]

Junior school

0.89
[0.08]
1.00
[0.08]
81

High school or higher

0.86
[0.16]
0.96
[0.08]
Urban-Rural (ref=living in rural)

Rural hukou, living in urban

1.23
[0.12] **

1.18
[0.10] *

Urban hukou, living in urban

0.91
[0.07]
1.01
[0.09]
Smoking (ref=nonsmokers)

Former smokers

1.11
[0.12]
1.08
[0.08]

Current smokers

0.80
[0.09] *

1.08
[0.09]
BMI (ref=normal weight)

Underweight

0.81
[0.09] *

1.17
[0.11]

Overweight

1.38
[0.10]
**
*

1.50
[0.09] ***

Obesity

1.91
[0.18]
**
*

1.94
[0.22] ***

82
P-value: age groups <.001 0.135 <.001 0.118 <.001 0.072
P-value: age 55-64 = age 65-74 <.001 0.953 0.001 0.964 <.001 0.734
P-value: age 55-64 = age 75+ <.001 0.632 <.001 0.720 <.001 0.834
P-value: age 65-74 = age 75+ 0.002 0.647 <.001 0.728 <.001 0.987
P-value: all indicators of family background 0.099 0.214 0.297 0.151
P-value: all indicators of health and healthcare in
childhood 0.401 0.318 0.428 0.548
P-value: all indicators of adulthood circumstances <.001 <.001
P-value: all groups of education 0.532 0.956
P-value: primary school = junior school 0.233 0.957
P-value: primary school = high school or higher 0.485 0.662
P-value: junior school = high school or higher 0.884 0.642
P-value: all groups of urban-rural 0.050 0.138
P-value: rural hukou, living urban = urban hukou, living in
urban 0.015 0.113
P-value: all groups of smoking 0.082 0.535
P-value: former smokers = current smokers 0.031 0.989
P-value: all groups of BMI <.001 <.001
P-value: underweight = overweight <.001 0.013
83
P-value: underweight = obesity <.001 <.001
P-value: overweight = obesity <.001 0.037
# of Observations 4521 3937 4521 3937 4521 3937
Note: Standard errors in the parenthesis. Standard errors are robust and clustered at the community level.
*** p<0.01, **p<0.05, * p<0.1.

84
Table 4. Relative Ratios from Negative Binomial Models Predicting Metabolic Risk, China Health and Retirement Longitudinal Study,
2014/2015 (N=8458)
Age Only Age + Childhood
Age + Childhood
+Adulthood
Women Men Women Men Women Men
Age (ref=45-54)

55-64
1.28
[0.08] ***

1.18
[0.14]
1.29
[0.09] ***

1.16
[0.11]
1.27
[0.09] ***

1.05
[0.10]

65-74
1.29
[0.08] ***

0.93
[0.09]
1.33
[0.08] ***

1.04
[0.10]
1.28
[0.08] ***

0.92
[0.08]

75+
1.27
[0.11] ***

0.83
[0.16]
1.32
[0.12] ***

0.98
[0.21]
1.25
[1.11] **

0.83
[0.14]
Childhood Adversity
Family Background During Respondent's Childhood

Mother was illiterate

0.95
[0.05]
0.84
[0.07]
**

0.97
[0.05]
0.90
[0.06]

Parent belonged to the bad/undesirable classes

1.13
[0.08]
* 1.14
[0.15]
1.12
[0.08]
* 1.08
[0.12]
85

Parent had drug, alcoholic, or gambling problems/was depressed/hit R
very often

1.00
[0.05]
1.15
[0.11]
1.01
[0.05]
1.15
[0.09]
*

Family members experienced starvation & R<=3 between 1958 and
1962

1.03
[0.07]
1.18
[0.14]
1.03
[0.07]
1.16
[0.12]
Infrastructure in Childhood (Before 15)

No clean water/flush toilet/electricity

0.92
[0.04]
* 0.91
[0.07]
0.95
[0.04]
1.01
[0.08]
Health and Health Care in Childhood (Before15)

Did not receive vaccination/Did not have a usual source of care

1.03
[0.06]
0.84
[0.06]
**

1.04
[0.05]
0.87
[0.07]
*

Missed school for health issues/confined to bed or hospitalized for
1mo+ or more than 3times
1.00
[0.12]
0.97
[0.09]
0.98
[0.10]
0.90
[0.09]
Adulthood Circumstances
Education (ref=No formal education)

Primary school

1.06
[0.05]
1.12
[0.10]
**
86

Junior school

0.98
[0.07]
1.15
[0.11]

High school or higher

0.84
[0.09]
* 1.25
[0.14]
**
Urban-Rural (ref=living in rural)

Rural hukou, living in urban

1.12
[0.08]
1.27
[0.13]
**

Urban hukou, living in urban

1.25
[0.09]
*** 1.53
[0.14]
***
Smoking (ref=nonsmokers)

Former smokers

1.14
[0.14]
1.48
[0.17]
***

Current smokers

0.88
[0.08]
1.09
[0.10]
P-value: age groups <.001 0.003 <.001 0.293 <.001 0.256
P-value: age 55-64 = age 65-74 0.888 0.015 0.569 0.224 0.876 0.108
P-value: age 55-64 = age 75+ 0.971 0.009 0.804 0.350 0.839 0.104
P-value: age 65-74 = age 75+ 0.912 0.515 0.904 0.725 0.761 0.479
P-value: all indicators of family background 0.359 0.014 0.506 0.089
87
P-value: all indicators of health and healthcare in
childhood 0.773 0.034 0.658 0.130
P-value: all indicators of adulthood circumstances 0.042 <.001
P-value: all groups of education 0.158 0.171
P-value: primary school = junior school 0.290 0.679
P-value: primary school = high school or higher 0.028 0.617
P-value: junior school = high school or higher 0.143 0.454
P-value: all groups of urban-rural 0.004 <.001
P-value: rural hukou, living urban = urban hukou, living in urban 0.231 0.123
P-value: all groups of smoking 0.236 <.001
P-value: former smokers = current smokers 0.101 <.001
# of Observations 4521 3937 4521 3937 4521 3937
Note: Standard errors in the parenthesis. Standard errors are robust and clustered at the community level.
*** p<0.01, **p<0.05, * p<0.1

88
Chapter 4: Socioeconomic, Biological, and Community Associations with Old-age
Mortality in China

Yuan S. Zhang
1
, John A. Strauss
2
, Peifeng Hu
3
, Yaohui Zhao
4
,
Eileen M. Crimmins
1

1. Davis School of Gerontology, University of Southern California, USA
2. Department of Economics, University of Southern California, USA
3. David Geffen School of Medicine, University of California, Los Angeles, USA
4. National School of Development, Peking University, China

Funding: This work was supported by the National Institute on Aging (grant number NIA P30
AG17265, T32 AG000037). The National Institute on Aging also supported the data collection
of the China Health and Retirement Longitudinal Study.

89
ABSTRACT
Determinants of mortality may differ depending on context. This study uses a nationally
representative sample of persons aged 60 and over in China to determine whether socioeconomic
factors, early life conditions, community characteristics, biological and physical functioning, and
disease burden predict four-year mortality. We find that current education and place of residence,
assessments of physical functioning, uncontrolled hypertension, diabetes, cancer, a high level of
systemic inflammation, and poor kidney functioning are strong predictors of mortality among
older Chinese. We find almost no linkages between early-life experiences or community
infrastructure to mortality at older ages. Results from this study highlight the value of
incorporating biological and performance measurements and the importance of social and
historical context in studying old-age mortality.

90
INTRODUCTION
Mortality at older ages results from health deterioration over the lifecycle which can be
impacted by individual-level characteristics as well as the social, economic, epidemiological, and
political context to which a person has been exposed. A large body of work on the determinants
of old-age mortality in developed countries has yielded insights on the relationship of mortality
to numerous demographic, social, biological, behavioral, and environmental factors (Rogers,
Everett, Onge, & Krueger, 2010; Seeman & Crimmins, 2006); however, these findings may not
directly translate to developing countries which have distinct socioeconomic and epidemiological
circumstances, leading to different patterning of risk factors (Sudharsanan, 2017; Wang, Kong,
Wu, Bai, & Burton, 2005). China provides an important context in which to study the
determinants of mortality in old age not only because of its large aging population but also
because of its unique social, economic, and epidemiological circumstances, which may result in
different determinants of mortality among the current cohort of aging people in China. In this
study, we use panel data collected in four waves of the China Health and Retirement
Longitudinal Study to explore determinants of mortality employing a life course perspective
among older Chinese. Understanding determinants of mortality will help identify important links
to social, economic, and epidemiological circumstances across the lifecycle and help improve
current public health programs in developing countries which face emerging aging challenges
but have limited resources.

Background
There are compelling reasons to expect the determinants of mortality to be unique in the
Chinese context. First, socioeconomic status (SES) can affect mortality through psychosocial and
91
behavioral mechanisms and environmental exposure (Adler & Newman, 2002; Evans &
Kantrowitz, 2002; Kristenson, Eriksen, Sluiter, Starke, & Ursin, 2004; Seeman & Crimmins,
2006). Contrary to patterns in Western developed countries where higher status is almost always
linked to poorer health and higher risk, SES differences in risk factors are very inconsistent in
developing countries including China. For example, higher SES is positively linked to
unhealthier behaviors such as smoking, drinking, obesity, as well as diets rich in fat, sugar and
refined carbohydrates (Chen et al., 2010; Monteiro et al., 2004). On the other hand, individuals
with higher SES also have greater resources, leading to advantages in many downstream health
outcomes including functioning and disability (Beydoun & Popkin, 2005; Zimmer & Kwong,
2004), chronic conditions (Zimmer & Kwong, 2004), and cognition (Huang & Zhou, 2013; Wen
& Gu, 2011). Most of the current cohort of aging people in China were born and matured in a
collective society, in which equality was more emphasized than in many countries; and they are
more likely to rely on family support at older ages than persons in developed countries (Zhu &
Xie, 2007). Taken together, it is reasonable to suspect a weaker relationship between SES and
mortality in China. We hypothesize that education is a protective factor for old-age mortality but
individual’s own economic status may have little effect on mortality among older Chinese.

Urban-rural differences further complicate the relationship between SES and mortality in
China. Throughout its long history of unbalanced development, urban and rural China have been
shown to differ strikingly in the physical environment, education, income, diet, and lifestyles,
which could lead to differentials in health and mortality. Generally, the urban Chinese population
has more education, higher income, better knowledge about health, and greater access to high-
quality health services than the rural population (Gong et al. 2012), leading to advantages in
92
mortality (Zhu & Xie, 2007). However, urban residents who migrate from rural areas may not
receive urban advantages since their access to social programs and public health services are
limited by their administrative registration, called hukou. As a governmental household
registration, hukou status officially defines a person as an agricultural or nonagricultural resident
but is difficult to change if a person has moved from a rural area to an urban city. In general,
people with rural hukou living in urban areas are severely disadvantaged in education, job
opportunity, and the eligibility for health insurance, compared to residents with urban hukou. We
expect to find that urban residents, especially those who are officially registered in urban areas,
have lower mortality risk compared to rural residents.

Second, the awareness and control of chronic disease are generally deficient in China,
particularly in less developed regions; therefore, chronic diseases, especially conditions that
people are unaware of can be more life-threatening. Older Chinese generally have very limited
education, limited access to health care, and limited resources to manage their conditions. A
study involving 1.7 million Chinese found that nearly half of the sample had hypertension, but
less than half have been diagnosed, only a third were being treated and fewer than 8% had their
blood pressure controlled (Lu et al., 2017). A study on diabetes among older Chinese based on
the CHARLS sample showed that among people with diabetes, 59.3% had measured high
HbA1c (≥6.5%) and/or high plasma glucose (≥126 mmHg) but were undiagnosed, representing
10.3% of the entire Chinese middle-aged and elderly population (Zhao et al., 2016). It has been
suggested that poor management of chronic disease is associated with elevated mortality at older
ages (Lewington et al., 2016). We expect that those who had measured high biological risk (i.e.,
high blood pressure, high HbA1c, high plasma glucose, and high C-reactive protein) would have
93
an increased mortality risk, reflecting the adverse effects of physiological dysregulation as well
as the consequences of poor health literacy and inadequate disease control. In addition,
performance-based physical assessments could better discriminate functional ability (Ailshire &
Crimmins, 2013), and reflect the health status that individuals may not be able to report;
therefore we would expect physical assessments to be predictive of mortality.

Abundant evidence from Western countries documents that early life circumstances have
long-term consequences for health and mortality for older adults (Haas, 2008; Hayward &
Gorman, 2004). However, previous studies on early-life conditions and mortality in the Chinese
population have not provided strong evidence of early life effects on late life mortality. Zeng,
Gu, and Land (2007) used the China Healthy Longevity Survey and found some association of
lower mortality for those whose father’s occupation was professional/administrative and for
those born in urban areas. Wen and Gu (2011) showed arm length, an indicator of childhood
health and development, to be negatively associated with mortality before controlling for other
health measures. Shen and Zeng (2014) found that favorable childhood conditions is directly
associated with a higher mortality risk, reflecting mortality selection; whereas advantageous
childhood also linked to improved SES in adulthood and thus indirectly promote health and
survival at older ages. Huang and Elo (2009) used the same data set and reported that place of
birth was not significantly linked to old-age mortality and childhood socioeconomic status was
only marginally related to mortality. The inconsistent findings suggest the link between
childhood conditions and mortality at old ages may be different and somewhat complicated in
China. Older Chinese have experienced a transition from a socialist society to an emerging
capitalist society where inequality in income, education, and access to health services sharply
94
increased in their middle and older ages. This transition may suggest a weaker association of
childhood conditions to adulthood circumstances and health at older ages. Thus, we expect to
find that adult SES may be more consequential for mortality in later life than childhood SES.

Gaps in the Literature
Previous research on old-age mortality in the Chinese population primarily focuses on
socioeconomic differences in mortality and has provided significant insights about the links of
social and environmental factors to health and mortality. Due to data availability, almost all
studies studied a relatively old population. Several studies, for instance Huang and Elo (2009),
only focused on the oldest old (aged 80+). Other studies (i.e., Shen and Zeng (2014), Wen and
Gu (2011)) studied a wider age range but their samples primarily consist of the 80+ population.
The oldest old is a highly selective population and the determinants of mortality could be
different in the young old and the oldest old (Lee, Go, Lindquist, Bertenthal, & Covinsky, 2008;
Zhu & Xie, 2007).

This study is unique in studying mortality at ages sixty and over using data from a recent
nationally representative study, the China Health and Retirement Longitudinal Study. The
baseline demographics of the CHARLS sample matches closely those of the population census in
2010 (Zhao, Hu, Smith, Strauss, & Yang, 2014). The use of a nationally representative sample
makes our estimates representative of the entire older Chinese population. By using rich data
from CHARLS, the present study will also advance our understanding of old-age mortality in
China in the following important ways. First, our study conceptualizes health as
multidimensional. Elucidating the links between the multidimensional health measures and
95
mortality is important for understanding the mechanisms of age-related health changes and
therefore has great implications for intervention. Second, incorporating measured biomarkers and
physical functioning is of importance in China where a significant proportion of older Chinese
have limited access to regular, high-quality health care and unawareness of chronic diseases is
common. Many studies have provided evidence for the value of incorporating indicators of
measured functioning and biomarkers in mortality prediction (Glei et al., 2014; Ailshire &
Crimmins, 2013; Vasunilashorn et al., 2014). More important, the biological measure and
performance-based functioning measures provide important information on physiological health
and physical functioning that allows us to uncover how social, behavioral, and environmental
factors “get under the skin” to have consequences on health and lead to mortality (Crimmins &
Vasunilashorn, 2011; Turra et al., 2005).

DATA AND METHODS
Data
Data come from the China Health and Retirement Longitudinal Study (CHARLS), a
nationally representative survey of those aged 45 and older in China. The baseline national
survey was conducted by Peking University from June 2011 to March 2012. As described in
greater detail elsewhere (Zhao et al., 2014), the household survey collected information about
individuals and households. A set of standardized physical, anthropometric, and blood pressure
measurements were collected in the household along with the household surveys. Blood was
collected in a subsequent visit to a township hospital or a local office of China Center for
Disease Prevention and Control (China CDC). Respondents were asked to fast overnight before
the blood draw, but blood was still collected even if a respondent did not fast. Written consent
96
forms were obtained from all respondents. The study protocol was approved by the ethical
review committee of Peking University.
Because the goal of this study is to investigate determinants of mortality in the older
population in which chronic diseases are the major causes of death, age is restricted to 60 years
and older to eliminate mortality at younger ages which are likely to have different causes-of-
death. Of the 7,724 subjects aged 60 and older at the initial interview, 4,262 (55.2%) provided a
blood sample and a physical assessment. Because the use of non-fasting samples could introduce
imprecision, our analytic sample included only those who fasted (N=4,176). We excluded an
additional 325 subjects because they did not provide information on chronic diseases or did not
participate in some physical assessments, or did not have information on adult socioeconomic
status. The final analytic sample consists of 3,851 persons. We compared characteristics of the
3,851 individuals in our analytic sample to the 3,873 individuals who were excluded from the
sample. Advanced age and living in an urban area were associated with a higher likelihood of
being excluded from the final analytic sample. No educational differences were observed. The
Appendix provides details on sample selection and analysis of missing data and loss-to-follow-
up (Table A1 and A2).
Measures
Mortality
CHARLS has followed respondents from the baseline survey with interviews in 2013,
2014, and 2015. At each interview wave, interviews were sought for earlier respondents. If the
event of respondent death is reported, CHARLS attempts an interview with a surviving family
member, relative, or other informant to obtain information about the death, including date of
death. We assessed four-year all-cause mortality in this paper. Deaths after 4 years were
97
censored. During the four-year-follow-up, 291 individuals (7.40%) died. The average survival
time is 3.73 years.

Predictors of Mortality
Age and sex were included as covariates in all analyses. Age is measured in years. Sex
was coded as 1 for females and 0 for males. Three variables representing current socioeconomic
status were created. Education is categorized as no formal schooling, primary school, and junior
school and above. Respondents were classified based on their usual place of residence and their
official household registration (hukou). Three dummy variables were created: rural residency,
urban residency and rural hukou, urban residency and urban hukou. The small percentage (<2%)
who lived in rural areas but had urban hukou were classified with other rural residents. This
three-category variable separates urban residents by their hukou which determines their ability to
use services and obtain benefits. Household per capita expenditures (PCE) was discretized into
terciles. We used PCE instead of income because it is a preferred measure of household resource
in settings like China where significant economic activity does not pass through markets
(Deaton, 1997).

Physical capacity
The assessment of physical functioning through performance-based tests is an essential
component of the evaluation of physical functioning among CHARLS respondents. We used
four performance-based physical functioning measures which have been shown to be predictive
of health outcomes associated with aging and mortality (Cooper, Kuh, & Hardy, 2010; Roberts
& Mapel, 2012; Sasaki, Kasagi, Yamada, & Fujita, 2007; Studenski, Perera, Patel, & al, 2011).
98
Grip strength was measured using a hand dynamometer. Two measures were taken from each
hand. Lung function was assessed using peak expiratory flow. Three measurements were taken,
at 30 second intervals. Gait speed (meters/second) was measured with a timed walk of 2.5
meters, completed twice. For each of these functioning assessments, we took the maximum value
as the functioning score and classified those in the worst 25% for each measure within each sex
as having poor functioning. Balance was assessed using the semi-full tandem timed balance test.
Individuals who were unable to hold semi- tandem stand for 10 seconds were considered as
having poor balance. For each assessment, those who were unable to perform the tests, or those
who did not complete the tests because either they or their interviewers thought it was unsafe
were also classified as having poor functioning.

Biomarkers
Biomarkers measure one’s physiological functioning, indicate health status, and predict
mortality (Crimmins & Vasunilashorn, 2011). The biomarker measurements used in this analysis
were obtained from physical assessments and the collection of a venous blood sample (Zhao et
al., 2014) and reflect health of multiple physiological systems, including cardiovascular
functioning – systolic blood pressure, diastolic blood pressure, and pulse rate; metabolic
functioning – glycated hemoglobin (HbA1c), plasma glucose, total cholesterol, and triglycerides,
kidney functioning – creatinine and blood urea nitrogen (BUN), and immune functioning – C-
reactive protein (CRP). For each biomarker, a dichotomous indicator was created indicating
“high-risk” and “not high-risk” based on clinical cutoff values for high risk (shown in Table 1).
Those who had elevated systolic/diastolic blood pressure are defined as having hypertension (Lu
99
et al., 2017); those who had high HbA1c or high plasma glucose were defined as having diabetes
(Zhao et al., 2016).

The Presence, Awareness, and Control of Chronic Diseases
Numerous studies show chronic conditions are powerful independent predictors of
mortality in older populations (Fried et al., 1998; Lee et al., 2008). Respondents were asked if
they had ever been diagnosed with a chronic disease – hypertension, diabetes, cancer, chronic
lung disease, heart problems, stroke, and kidney disease. However, the actual prevalence of
chronic diseases among older Chinese is believed to be higher because of underreporting (Lu et
al., 2017; Zhao et al., 2016). As indicated above, because CHARLS measured blood pressure,
HbA1c, and plasma glucose, we combined measured biomarkers and self-reported diagnosis to
create categories for indicators of hypertension (no hypertension, undiagnosed, uncontrolled, and
controlled), and diabetes (no diabetes, undiagnosed diabetes, and diagnosed diabetes).

Behavioral Factor - Smoking
A key behavioral indicator—smoking status was included as a predictor. A dummy
variable for smoking was coded 1 if respondents reported ever smoking, and 0 if respondents
never smoked.

Community Infrastructure
We adopted the method used in Li, Liu, Xu, & Zhang (2016) that utilized the same
dataset to construct a community infrastructure deficiency index based on a principal
components analysis of seven community infrastructure measures including road type, annual
100
days with unpassable roads, community accessible by bus, community has a sewer system, main
type of toilet in the community, main type of cooking fuel, and main source of drinking water.
We then divide communities into tertiles based on the infrastructure deficiency index.

Early-life Conditions
Knee height is an important proxy for early childhood growth and development which
reflects health at a young age (Huang et al., 2008). We used sex-specific quartiles of knee height
and classified those in the bottom 25% as having short knee height. Whether an individual was
living in a rural area before age 16 was also included as an indicator of the early-life
environment.

Statistical Analyses
Table 1 presents the descriptive statistics for the study sample. All analyses were
weighted using the biomarker weights to account for the complex sample design of CHARLS,
nonresponses to the main household interview, and nonparticipation in the blood collection. We
employed a series of Weibull hazard models based on exact survival time to predict 4-year all-
cause mortality. Model 1 examined age, gender, and SES differentials in mortality. We then
successively added multiple indicators of health – physical capacity, chronic diseases, and
biomarkers of physiological functioning - were successively added in Models 2, 3, and 4. We
then subsequently added smoking (Model 5), community infrastructure (Model 6), and indicators
of early-life conditions (Model 7).

101
RESULTS
Sample Description
As shown in Table 1, age at baseline ranges from 60 to 98, with a mean of 68.58. The
study sample was almost equally split between males and females. 42.16% individuals did not
have any schooling. 57.79% lived in a rural area, 18.76% lived in an urban area but still had a
rural hukou, 23.45% lived in an urban area but had urban hukou. Almost 90% of individuals
lived in rural areas before age 16. About 40% have smoked. The percentage with low grip
strength, poor lung function, and slow walking is about 25%, because they were coded to
represent the bottom 25% (before weighting) in physical functioning. We include an indicator of
missing data for the timed walk test; 3.78% did not take the timed walk test.
46% did not have hypertension; 14.12% have controlled hypertension, those who had
been diagnosed with hypertension but had measured blood pressure below 140/90 mm Hg; about
20% had been diagnosed but still measured as having high blood pressure. 20.16% of the sample
had currently undiagnosed hypertension. Diabetes was found in about 20% of the sample –
7.57% was diagnosed but 11.26% was undiagnosed. 14.65% reported chronic lung disease and
17.61% reported heart problems. The prevalence of reported cancer, stroke and kidney disease is
low. High CRP was found in 22.49% of the sample, 18.06% of the sample had high BUN,
13.27% had high triglycerides, 10.99% had high total cholesterol. There are relatively few
individuals with a rapid pulse (4.80%) or high creatinine (1.42%).
Regression Analyses
Table 2 presents hazard rate ratios for mortality. Advanced age is associated with higher
mortality (HR=1.08, p<.001) and this result is consistent across all models (HR =1.05-1.06,
p<.001 in Models 2-7). Being female is associated with about fifty percent less mortality
102
(HR=0.49, p<.001). This significant female advantage in mortality is reduced and becomes
insignificant when smoking is included in the model, suggesting the sex difference in mortality is
attributed to smoking. We also observe significant SES differences in mortality between those
with some schooling and those with no schooling. Individuals who had only a primary school
education and those who had junior school or higher education have about 50% reduced
mortality risk compared to those without formal schooling (HR=0.53, p<.01 for primary school,
and HR = 0.51, p<.05 for junior school and above). Being an urban resident with urban hukou is
associated with lower mortality (HR = 0.63, p<0.1). We find little association of household
expenditures with mortality.

Among the indicators of physical functioning, low grip strength, poor lung function, and
poor balance are predictive of mortality. The relative hazard ratio for mortality for those with
poor grip strength relative to having normal/strong grip strength is about 1.4. The hazard ratio for
mortality with poor lung function relative to normal lung function is around 1.8. Those who
failed the balance test have about 75%- 90% higher mortality risk (HR = 1.75-1.91 for poor
balance). Slow walking is related to higher mortality but the association is not statistically
significant when these other functioning measures are in the equation. Uncontrolled hypertension
is related to an excess risk of mortality. Those who have been diagnosed with diabetes have an
about 80% higher mortality risk than those who do not have diabetes. Undiagnosed diabetes is
associated with about 30% higher risk but it is not statistically significant. The prevalence of
self-reported cancer is very low in China; however, reporting having been diagnosed with cancer
is associated with a substantially increased risk of mortality. The mortality risk for those who
103
reported having cancer is 4.5-5 times as those who did not reported cancer. Among the blood-
based biomarkers, high CRP and high creatinine are consistently predictive of mortality.

Smoking is significantly related to a higher mortality risk and eliminates the significant
male-female mortality difference. We do not observe links of community infrastructure, knee
height and living in a rural area in childhood to mortality after we control for adulthood SES and
a variety of health measures.

CONCLUSION
This study investigated socioeconomic, biological, community associations with
mortality in the older Chinese population. We began by estimating the association between
adulthood SES and mortality with age and sex adjusted. We used three indicators (i.e., education,
urban-rural residence, and household expenditures) to represent different dimensions of adult
SES. Our results show a strong inverse association between education and mortality – the risk of
dying in four years for the elderly with the highest education was about 50% less than the risk for
the uneducated. Those who have an urban hukou and live in an urban area also have a
substantively lower risk of dying compared to those living in rural villages. These educational
differences and urban-rural differences existed after adjustment for a variety of health measures
as well as community and early-life characteristics. Household expenditures were not related to
mortality risk. These findings are consistent with our hypotheses and existing literature on SES
and mortality in China (Wen & Gu, 2011b; Zhu & Xie, 2007; Zimmer & Kwong, 2004). The
weak relationship between economic resources and mortality is also reported in other developing
countries (Rosero-Bixby & Dow, 2009; Sudharsanan, 2017). We are optimistic about the
104
continuous decline in mortality at older ages as education in China has been increasing rapidly,
especially in rural areas. However, educational access and quality remain uneven between urban
and rural areas. Therefore, the urban advantage in life expectancy is likely to persist or even
increase in the future.

Importantly, our results indicated that the objectively-measured physical functioning
measures (except slow walking), elevated CRP and creatinine are independent and strong
predictors of mortality after accounting for a number of additional self-reported health measures.
These findings highlighted the value of incorporation of biological and performance
measurements in population health surveys to help us model the health change and aging
processes that lead to mortality (Crimmins, Kim, & Vasunilashorn, 2010). Because these
measured indicators of functional health and biomarkers allow for assessment of a full range of
physical capacity and physiological health particularly among persons who do not report
functional limitations, disabilities, or diseases; therefore, they can be a sensitive tool for
assessing the health risk in a relatively young and healthy sample (Ailshire & Crimmins,
2013).By combining measured blood pressure and self-reports of diagnosis, we could examine
the links between awareness and control of hypertension on mortality. Our analysis shows that
hypertensive persons who have adequately controlled their blood pressure have lower mortality
while hypertensive persons with uncontrolled blood pressure have a 50% higher probability of
dying in four years compared to non-hypertensive persons. The burden of chronic diseases is
rapidly increasing in China along with the improving economy, rapid urbanization, and
population aging; yet the awareness, treatment, and control is still limited (Lu et al., 2017;
Zengwu et al., 2018; Zhao et al., 2016). Su et al. (2017) assessed the availability, cost, and
105
prescription patterns of 62 antihypertensive medications at 3362 primary health-care sites across
31 provinces in China and found that 8% of primary care facilities did not stock any
antihypertensive medications and only a third stocked all classes of routinely-used anti
hypertensives. High-value drugs (defined as low cost and recommended by the Chinese
Guideline for Hypertension Management in Primary Health Care 2014) are only stocked in 33%
of facilities and are not preferentially used. These findings highlight the role of insufficient
education and screening, as well as marked deficiencies in the availability, cost and prescription
of medications are important reasons for the high burden of chronic diseases in China. Another
strong mortality predictor is cancer which is the leading cause of death in China. (Chen et al.,
2016). Diabetics also have a much higher risk of dying. Although the prevalence of cancer
reported by respondents is very low, the mortality risk for those who have been diagnosed with
cancer is almost 400% higher than individuals who did not report cancer diagnosis. The low
diagnosis and substantial mortality risk certainly reflect both late diagnosis and poor treatment of
cancer.

We did not observe significant associations of community infrastructure and early-life
indicators with mortality. The literature associating childhood conditions with mortality in China
has been inconsistent. Although it has been suggested that the adverse effects of early-life
adversity on adult health could be larger in developing countries because adversity shocks are
more frequent and interact with limited capacity for remediation, the relationships may be more
difficult to measure given that much of the population is exposed to adverse conditions and
mortality selection could be significant (Currie & Vogl, 2013). The older Chinese we studied in
this paper are survivors of a series of wars, famines, infectious diseases, and political
106
movements, likely resulting in significant mortality selection. Therefore, the long-term effects of
early-life could appear less pronounced. It is also essential to note the importance of the
transitions that the current cohort of older Chinese has gone through. The older Chinese lived in
a socialist society in which equality was emphasized, generally experienced hardships at younger
ages, and then experienced change that could lead to disparities in income, lifestyle, and
resources one could use for managing later life health; therefore, the correlation between
childhood and adulthood SES could be weaker among older Chinese than in other places. Thus,
childhood experiences may have less impact on mortality at older ages.

One limitation of the present study is that like most longitudinal cohort studies, CHARLS
suffers from loss to follow-up and some of those who were lost to follow-up died and some did
not. We conducted a number of analyses to examine the characteristics of those lost-to-follow-up
and used a variety of techniques for handling loss to follow-up to assess the robustness of our
results. Our analyses (Appendix Table A3) shows that our results from different models are
almost identical. Second, our analytic sample is significantly reduced by non-participation in the
blood sample collection and physical assessments. However, by applying the biomarker weight
correcting for this non-response, our analytic sample is representative of middle-aged and older
Chinese adults. Despite these limitations, our study used a large nationally representative sample
of middle-aged and older persons to study the determinants of old-age mortality. Our study
highlighted the necessity of incorporation of biological and performance measurements and the
importance of social and historical context in studying old-age mortality.

107
REFERENCES
Adler, N. E., & Newman, K. (2002). Socioeconomic disparities in health: Pathways and policies.
Health Affairs. https://doi.org/10.1377/hlthaff.21.2.60
Beydoun, M. A., & Popkin, B. M. (2005). The impact of socio-economic factors on functional
status decline among community-dwelling older adults in China. Social Science and
Medicine. https://doi.org/10.1016/j.socscimed.2004.08.063
Chen, W., Zheng, R., Baade, P. D., Zhang, S., Zeng, H., Bray, F., … He, J. (2016). Cancer
statistics in China, 2015. CA: A Cancer Journal for Clinicians, 66(2), 115–132.
https://doi.org/10.3322/caac.21338
Cooper, R., Kuh, D., & Hardy, R. (2010). Objectively measured physical capability levels and
mortality: systematic review and meta-analysis. BMJ, 341.
https://doi.org/10.1136/bmj.c4467
Ailshire, J. & Crimmins, E. (2013). Physical and biological indicators of health and functioning
in US oldest old. Annual Review of Gerontology and Geriatrics, 33(1), 193-215.
Crimmins, E. M., & Vasunilashorn, S. (2011). Links Between Biomarkers and Mortality. In
International Handbook of Adult Mortality. https://doi.org/10.1007/978-90-481-9996-9
Currie, J., & Vogl, T. (2013). Early-Life Health and Adult Circumstance in Developing
Countries. Annual Review of Economics, 5(1), 1–36. https://doi.org/10.1146/annurev-
economics-081412-103704
Deaton, A. (1997). The analysis of household surveys. The World Bank.
https://doi.org/doi:10.1596/0-8018-5254-4
Evans, G. W., & Kantrowitz, E. (2002). Socioeconomic Status and Health: The Potential Role of
Environmental Risk Exposure. Annual Review of Public Health.
108
https://doi.org/10.1146/annurev.publhealth.23.112001.112349
Fried, L. P., Kronmal, R. A., Newman, A. B., Bild, D. E., Mittelmark, M. B., Polak, J. F., …
Group, for the C. H. S. C. R. (1998). Risk Factors for 5-Year Mortality in Older AdultsThe
Cardiovascular Health Study. JAMA, 279(8), 585–592.
https://doi.org/10.1001/jama.279.8.585
Huang, C., & Elo, I. T. (2009). Mortality of the oldest old Chinese: The role of early-life
nutritional status, socio-economic conditions, and sibling sex-composition. Population
Studies, 63(1), 7–20. https://doi.org/10.1080/00324720802626921
Huang, T. L., Carlson, M. C., Fitzpatrick, A. L., Kuller, L. H., Fried, L. P., & Zandi, P. P.
(2008). Knee height and arm span: a reflection of early life environment and risk of
dementia. Neurology, 70(19 Pt 2), 1818–1826.
https://doi.org/10.1212/01.wnl.0000311444.20490.98
Huang, W., & Zhou, Y. (2013). Effects of education on cognition at older ages: Evidence from
China’s Great Famine. Social Science and Medicine.
https://doi.org/10.1016/j.socscimed.2013.08.021
Kristenson, M., Eriksen, H. R., Sluiter, J. K., Starke, D., & Ursin, H. (2004). Psychobiological
mechanisms of socioeconomic differences in health. Social Science & Medicine, 58(8),
1511–1522. https://doi.org/https://doi.org/10.1016/S0277-9536(03)00353-8
Lee, S. J., Go, A. S., Lindquist, K., Bertenthal, D., & Covinsky, K. E. (2008). Chronic
Conditions and Mortality Among the Oldest Old. American Journal of Public Health, 98(7),
1209–1214. https://doi.org/10.2105/AJPH.2007.130955
Li, L. W., Liu, J., Xu, H., & Zhang, Z. (2016). Understanding Rural–Urban Differences in
Depressive Symptoms Among Older Adults in China. Journal of Aging and Health, 28(2),
109
341–362. https://doi.org/10.1177/0898264315591003
Lu, J., Lu, Y., Wang, X., Li, X., Linderman, G. C., Wu, C., … Jiang, L. (2017). Prevalence,
awareness, treatment, and control of hypertension in China: data from 1.7 million adults in a
population-based screening study (China PEACE Million Persons Project). Lancet (London,
England), 390(10112), 2549–2558. https://doi.org/10.1016/S0140-6736(17)32478-9
Roberts, M. H., & Mapel, D. W. (2012). Limited Lung Function: Impact of Reduced Peak
Expiratory Flow on Health Status, Health-Care Utilization, and Expected Survival in Older
Adults. American Journal of Epidemiology, 176(2), 127–134. Retrieved from
http://dx.doi.org/10.1093/aje/kwr503
Rogers, R. G., Everett, B. G., Onge, J. M. S., & Krueger, P. M. (2010). Social, behavioral, and
biological factors, and sex differences in mortality. Demography.
https://doi.org/10.1353/dem.0.0119
Rosero-Bixby, L., & Dow, W. H. (2009). Surprising SES Gradients in Mortality, Health, and
Biomarkers in a Latin American Population of Adults. The Journals of Gerontology: Series
B, 64B(1), 105–117.
Sasaki, H., Kasagi, F., Yamada, M., & Fujita, S. (2007). Grip Strength Predicts Cause-Specific
Mortality in Middle-Aged and Elderly Persons. The American Journal of Medicine, 120(4),
337–342. https://doi.org/https://doi.org/10.1016/j.amjmed.2006.04.018
Seeman, T. E., & Crimmins, E. M. (2006). Social Environment Effects on Health and Aging.
Annals of the New York Academy of Sciences. https://doi.org/10.1111/j.1749-
6632.2001.tb02749.x
Shen, K., & Zeng, Y. (2014). Direct and indirect effects of childhood conditions on survival and
health among male and female elderly in China. Social Science and Medicine.
110
https://doi.org/10.1016/j.socscimed.2014.07.003
Studenski, S., Perera, S., Patel, K., & al, et. (2011). Gait speed and survival in older adults.
JAMA, 305(1), 50–58. Retrieved from http://dx.doi.org/10.1001/jama.2010.1923
Su, M., Zhang, Q., Bai, X., Wu, C., Li, Y., Mossialos, E., … Jiang, L. (2017). Availability, cost,
and prescription patterns of antihypertensive medications in primary health care in China: a
nationwide cross-sectional survey. The Lancet, 390(10112), 2559–2568.
https://doi.org/https://doi.org/10.1016/S0140-6736(17)32476-5
Sudharsanan, N. (2017). The Association Between Socioeconomic Status and Adult Mortality in
a Developing Country: Evidence From a Nationally Representative Longitudinal Survey of
Indonesian Adults. The Journals of Gerontology: Series B, gbx061-gbx061.
Turra, C. M., Goldman, N., Seplaki, C. L., Glei, D. A., Lin, Y. H., & Weinstein, M. (2005).
Determinants of mortality at older ages: The role of biological markers of chronic disease.
Population and Development Review. https://doi.org/10.1111/j.1728-4457.2005.00093.x
Wang, L., Kong, L., Wu, F., Bai, Y., & Burton, R. (2005). Preventing chronic diseases in China.
Lancet. https://doi.org/10.1016/S0140-6736(05)67344-8
Wen, M., & Gu, D. (2011a). The Effects of Childhood, Adult, and Community Socioeconomic
Conditions on Health and Mortality among Older Adults in China. Demography, 48(1),
153–181. https://doi.org/10.1007/s13524-010-0003-2
Wen, M., & Gu, D. (2011b). The Effects of Childhood, Adult, and Community Socioeconomic
Conditions on Health and Mortality among Older Adults in China. Demography, 48(1),
153–181. https://doi.org/10.1007/s13524-010-0003-2
Zeng, Y., Gu, D., & Land, K. C. (2007). The association of childhood socioeconomic conditions
with healthy longevity at the oldest-old ages in China. Demography, 44(3), 497–518.
111
https://doi.org/10.1353/dem.2007.0033
Zengwu, W., Zuo, C., Linfeng, Z., Xin, W., Guang, H., Zugui, Z., … null, null. (2018). Status of
Hypertension in China. Circulation, 137(22), 2344–2356.
https://doi.org/10.1161/CIRCULATIONAHA.117.032380
Zhao, Y., Crimmins, E. M., Hu, P., Shen, Y., Smith, J. P., Strauss, J., … Zhang, Y. (2016).
Prevalence, diagnosis, and management of diabetes mellitus among older Chinese: results
from the China Health and Retirement Longitudinal Study. International Journal of Public
Health. https://doi.org/10.1007/s00038-015-0780-x
Zhao, Y., Hu, Y., Smith, J. P., Strauss, J., & Yang, G. (2014). Cohort Profile: The China Health
and Retirement Longitudinal Study (CHARLS). International Journal of Epidemiology,
43(1), 61–68. https://doi.org/10.1093/ije/dys203
Zhu, H., & Xie, Y. (2007). Socioeconomic differentials in mortality among the oldest old in
China. Research on Aging. https://doi.org/10.1177/0164027506296758
Zimmer, Z., & Kwong, J. (2004). Socioeconomic Status and Health among Older Adults in Rural
and Urban China. Journal of Aging and Health. https://doi.org/10.1177/0898264303260440

112
Table 1. Sample Characteristics, China Health and Retirement Longitudinal Study, 2011-15
4-year all-cause mortality, N (weighted %) 291 (7.40%)
Baseline Characteristics
Age, Mean(SD) 68.58 (6.78)
Female 50.37
Adulthood SES
Education

No formal schooling 42.16
Primary school 37.50
Junior school+ 20.35
Urban-Rural
Rural residency 57.79
Urban residency, rural hukou 18.76
Urban residency, urban hukou 23.45
Household Expenditure
Bottom tertile 37.12
Middle tertile 31.82
Top tertile 30.06
Smokers 40.97
Physical Assessments
Low grip strength 25.88
Poor lung function 26.48
113
Poor balance 7.25
Slow walking 26.41
Missing 3.78
Disease
Hypertension
No hypertension 46.03
Undiagnosed 20.16
Controlled 14.12
Uncontrolled 19.70
Diabetes
No diabetes 81.17
Undiagnosed 11.26
Diagnosed 7.57
Cancer 0.75
Chronic lung disease 14.65
Heart problems 17.61
Stroke 4.05
Kidney disease 6.48
Biomarkers
Rapid pulse 4.80
High total cholesterol 10.99
High triglycerides 13.27
High CRP 22.49
114

High Creatinine 1.42
High BUN 18.06
Poor infrastructure community 22.47
Missing on community characteristics 4.27
Early-life Indicators
knee height, bottom 25% 23.97
living in rural before age 16 89.09
# Observations 3,851
115
Table 2. Hazard Ratios from Weighted Weibull Hazard Regression Models Predicting 4-year Mortality, N=3,851
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

Age 1.08 *** 1.06 *** 1.06 *** 1.05 *** 1.05 *** 1.05 *** 1.05 ***
Female 0.49 *** 0.47 *** 0.44 *** 0.48 *** 0.69 + 0.69 + 0.69 +
Adulthood SES
Education (ref=no formal schooling)
Primary school 0.53 ** 0.60 * 0.58 * 0.59 * 0.62 * 0.62 * 0.62 *
Junior school+ 0.51 * 0.58 + 0.56 + 0.60 0.60 0.60 + 0.59
Urban-Rural (ref=rural residency)

Urban residency, rural
hukou 0.93 0.89 0.90 0.84 0.86 0.85 0.86

Urban residency, urban
hukou 0.63 + 0.70 0.67 0.64 + 0.69 0.68 0.64 +
Household Expenditure (ref=bottom
tertile)
Middle tertile 0.87 0.87 0.85 0.86 0.84 0.84 0.84
116
Top tertile 1.16 1.15 1.16 1.14 1.10 1.10 1.09
Physical Assessments
Low grip strength 1.44 ** 1.43 * 1.42 * 1.42 * 1.41 * 1.42 *
Poor lung function 1.88 *** 1.87 *** 1.78 *** 1.78 *** 1.78 *** 1.77 ***
Poor balance 1.91 ** 1.88 ** 1.80 * 1.75 * 1.75 * 1.78 *
Slow walking 1.14 1.17 1.15 1.19 1.20 1.20
Disease
Hypertension (ref=no)
Undiagnosed 1.12 1.05 1.05 1.06 1.06
Controlled 0.68 0.64 0.65 0.65 0.64
Uncontrolled 1.57 * 1.49 * 1.47 * 1.47 * 1.46 *
Diabetes (ref=no)
Undiagnosed 1.30 1.29 1.31 1.31 1.31
Diagnosed 1.79 ** 1.81 * 1.86 ** 1.85 ** 1.86 **
Cancer 4.50 ** 4.72 ** 4.69 ** 4.73 ** 4.95 **
Chronic lung disease 1.26 1.22 1.14 1.14 1.16
117
Heart problems 0.91 0.91 0.90 0.90 0.90
Stroke 0.85 0.86 0.84 0.84 0.84
Kidney disease 0.94 0.80 0.78 0.78 0.80
Biomarkers
Rapid pulse 1.27 1.30 1.31 1.29
High total cholesterol 1.03 1.02 1.02 1.02
High triglycerides 0.86 0.84 0.84 0.83
High CRP 1.81 *** 1.79 *** 1.79 *** 1.78 ***
High Creatinine 2.82 ** 2.66 ** 2.65 ** 2.67 **
High BUN 0.90 0.88 0.88 0.87
Smoking (ref=nonsmoking) 1.77 ** 1.76 ** 1.78 **
Poor infrastructure community 0.97 0.97
Early-life Indicators
bottom 10% knee height 0.89
living in rural before age 16 0.81
Significance: +p<0.1, *p<0.05, **p<0.01, ***p<0.001

118
APPENDIX
Sample Selection and Missing Data Analysis
The China Health and Retirement Longitudinal Study (CHARLS) interviewed 7,724
individuals aged 60 and older at the national baseline interview. Our analytic sample for this
study consists of 3,826 subjects, 50% of the 60+ respondents. The primary reason for sample
reduction is that 3,462 individuals did not provide a blood sample or a physical assessment at the
initial interview. Figure A illustrates our how the analytic sample was determined.

Figure A. Sample Selection

60+ Respondents in China Health and
Retirement Longitudinal Study 2011-12
(N=7,724)
Respondents provided blood samples
and physical assessments
(N=4,262)
Respondents fasted overnight
(N=4,176)
325 subjects missing on one of the
biomarkers or physical
assessments, or did not provide
information on at least one of the
chronic diseases, or did not have
information on adult SES
Analytic sample
(N=3,851)

119

Differences between the final sample of 3851 and the 3873 who were not included in the
analysis are indicated by the odds ratios presented in Table A1. Advanced age is associated with
a higher probability of being missing from the analysis; urban residents, especially those with
urban hukou, are more likely to be excluded compared to rural residents. Being in the top tercile
of household per capita expenditure is also associated with a higher probability of being
excluded from the analytic sample.

Table A1. Odds Ratios from Weighted Logistic Regression Model Predicting Being
Excluded from the Analytic Sample Among the 60+, CHARLS (N=7,724)
Age 1.02
***
Female 0.97

Education (ref=illiterate)

No formal schooling

Primary school 0.87

Junior school+ 1.07

Urban-Rural (ref=rural residency)

Urban residency, rural hukou 1.39
*
Urban residency, urban hukou 2.09
***
Household Expenditure (ref=bottom tertile)

Middle tertile 1.09

Top tertile 1.20
+
Significance: +p<0.1, *p<0.05, **p<0.01, ***p<0.001

120
The mortality outcome for subjects in our analytic sample can be classified into three
groups – those who were alive and re-interviewed in 2015, those who died by 2015, and
individuals lost to follow-up. We compare the characteristics of these three groups in Table A2.
Regression results confirmed that those who lived in urban areas, especially those who have
urban hukou are more likely to be lost to follow-up.

121
Table A2. Relative Risks from Multinomial Logistic Regression Models Predicting Being Alive
by 2015, N=3851

Known to have died
compared to known
to be alive by 2015

Lost-to-follow-up
compared to known
to be alive by 2015
Age 1.06 *** 1.01
Female 0.68 1.18
Adulthood SES
Education (ref=no formal schooling)
Primary school 0.60 * 0.56 **
Junior school+ 0.56 + 0.77
Urban-Rural (ref=rural residency)
Urban residency, rural hukou 0.84 2.83 **
Urban residency, urban hukou 0.64 4.46 ***
Household Expenditure (ref=bottom tertile)
Middle tertile 0.85 1.58 +
Top tertile 1.04 1.24
Physical Assessments
Low grip strength 1.47 * 0.95
Poor lung function 1.88 *** 1.34
Poor balance 1.91 * 0.76
Slow walking 1.16 1.32
Disease
Hypertension (ref=no)

122
Undiagnosed 1.14 0.95
Controlled 0.65 0.83
Uncontrolled 1.52 * 1.68 *
Diabetes (ref=no)
Undiagnosed 1.48 + 1.14
Diagnosed 1.88 * 0.46 *
Cancer 5.17 ** /
Chronic lung disease 1.13 0.87
Heart problems 0.92 0.82
Stroke 0.78 0.67
Kidney disease 0.9 1.27
Biomarkers
Rapid pulse 1.37 0.56
High total cholesterol 0.99 0.93
High triglycerides 0.79 0.85
High CRP 1.97 *** 0.77
High Creatinine 2.97 ** 1.69
High BUN 0.91 0.87
Smoking (ref=nonsmoking) 1.91 ** 0.99
Poor infrastructure community 0.93 0.89
Early-life Indicators
bottom 10% knee height 0.85 0.92
living in rural before age 16 0.77 0.65
Significance: +p<0.1, *p<0.05, **p<0.01, ***p<0.001

123

Previous studies used different ways to handle those lost to follow-up. We used different
models to examine whether our results are robust to different models. Results from three
different models are reported in Table A3.

124
Table A3. Hazard Ratios from Weighted Weibull Hazard Regression Models Predicting 4-year Mortality

Assign those
who were only
observed at
baseline 1-year
study time

Drop those
who were only
observed at
baseline

Only Include
Those Who
were known as
alive/dead by
2015
Age 1.05 *** 1.05 *** 1.05 ***
Female 0.69 + 0.69 + 0.70 +
Adulthood SES
Education (ref=no formal schooling)
Primary school 0.62 * 0.62 * 0.61 *
Junior school+ 0.59 0.59 0.58
Urban-Rural (ref=rural residency)
Urban residency, rural hukou 0.86 0.86 0.88
Urban residency, urban hukou 0.63 + 0.64 + 0.68 +
Household Expenditure (ref=bottom tertile)
Middle tertile 0.84 0.84 0.84
Top tertile 1.09 1.09 1.09
Physical Assessments
Low grip strength 1.42 * 1.42 * 1.41 *
Poor lung function 1.77 *** 1.77 *** 1.76 ***
Poor balance 1.78 * 1.78 * 1.73 *
Slow walking 1.20 1.20 1.22
Disease
Hypertension (ref=no)
Undiagnosed 1.06 1.06 1.05
Controlled 0.64 0.64 0.63
Uncontrolled 1.46 * 1.46 * 1.46 *
Diabetes (ref=no)
Undiagnosed 1.31 1.31 1.32

125
Diagnosed 1.86 ** 1.86 ** 1.82 **
Cancer 4.96 ** 4.95 ** 4.76 **
Chronic lung disease 1.16 1.16 1.16
Heart problems 0.90 0.90 0.90
Stroke 0.84 0.84 0.84
Kidney disease 0.80 0.80 0.80
Biomarkers
Rapid pulse 1.30 1.29 1.27
High total cholesterol 1.01 1.02 1.00
High triglycerides 0.83 0.83 0.84
High CRP 1.78 ** 1.78 *** 1.75 ***
High Creatinine 2.67 ** 2.67 ** 2.64 **
High BUN 0.87 0.87 0.86
Smoking (ref=nonsmoking) 1.78 ** 1.78 ** 1.80 **
Poor infrastructure community 0.97 0.97 0.97
Early-life Indicators
bottom 10% knee height 0.89 0.89 0.88
living in rural before age 16 0.81 0.81 0.80
Significance: +p<0.1, *p<0.05, **p<0.01, ***p<0.001

126
Chapter 5: Ascertaining Cause of Mortality Among Middle-Aged and Older Persons Using
Computer-coded and Expert Review Verbal Autopsies in the China Health and Retirement
Longitudinal Study (CHARLS)

Yuan S. Zhang
1
, Peifeng Hu
2
, John A. Strauss
3
, Yaohui Zhao
4
, Yafeng Wang
5
,
Eileen M. Crimmins
1
1. Davis School of Gerontology, University of Southern California, USA
2. David Geffen School of Medicine, University of California, Los Angeles, USA
3. Department of Economics, University of Southern California, USA
4. National School of Development, Peking University, China
5. Institute of Social Science Surveys, Peking University, China

Funding: This work was supported by the National Institute on Aging (grant number NIA P30
AG17265, T32 AG000037). The National Institute on Aging also supported the data collection
of the China Health and Retirement Longitudinal Study.

127
ABSTRACT
Background: Verbal autopsy (VA) can provide information on cause of death (COD) for places
where vital statistics systems are unreliable and lacking. However, no existing research has
addressed the use of VA in an older population.
Objective: This paper compares COD assignments by experts and from InterVA4, computer
software for analyzing VA data, to ascertaining COD in the older Chinese population.
Methods: Data come from the China Health and Retirement Longitudinal Study, a nationally
representative longitudinal survey of middle-aged and older Chinese. We compared COD
derived from InterVA4 with assignments by experts and evaluated how characteristics of the
deceased and VA interview affect the comparison.
Results: Neoplasms, cardiac disease, and stroke are the leading COD determined by both
methods. The percentage matching from the two approaches is about 50% at the individual level.
The primary reason for the lack of a match is with expert review no cause could be assigned for
more than 25% of the sample. A higher likelihood of mismatch is associated with advanced age
and deaths occurring at home. Indeterminate deaths by expert review occurred primarily among
frail older individuals.
Conclusions: Both approaches identify the same leading causes at the aggregate level, but the
matching is poor at the individual level. InterVA performs well when COD is characterized by
distinctive signs and symptoms that do not overlap with those common to other causes.
Contribution: This study provides important empirical evidence of the value of conducting VA
interviews in the context of population-based surveys of older persons.

128
INTRODUCTION
Cause of mortality information is crucial for understanding why mortality risk differs
across populations or subgroups. Vital statistics systems are relatively unreliable in providing
cause of death (COD) ascertainment in China and many other less-developed countries
(Mahapatra et al., 2007). One approach to determining death rates by cause in places where COD
is not available is to conduct verbal autopsy (VA) interviews with persons who were close to the
deceased in their final illness and who can provide information on the death. Applying VA in the
context of population-based surveys may allow researchers to determine COD for individual
survey respondents and link COD to risk factors, such as early-life experiences, socioeconomic
conditions, living circumstances, health behaviors, and disease history.

VA interviews collect rich personal information, information on death registration and
certification, medical history of the deceased, signs, symptoms, and circumstances prior to death,
as well as injuries and accidents that lead to death. It is evident that VA can provide relatively
valid and reliable estimates of causes of neonatal death (Aggarwal, Kumar, Pandit, & Kumar,
2013), pneumonia (Setel et al., 2006), HIV/AIDS (Tensou et al., 2010), and external causes of
death (Setel et al., 2006). However, most existing studies have focused on the value of VA in
determining causes of death from infectious diseases and causes of childhood and maternal
deaths in countries where they are still major public health concerns, such as Africa and South
Asian countries (Byass et al., 2015; Leitao et al., 2013). Research is limited on the value of VA
in older populations in which chronic diseases are the major causes of death. Classifying deaths
for older adults from chronic causes pose many different issues. First, the potential causes of
adult deaths are numerous in contrast to the limited number of causes for childhood deaths and

129
maternal deaths (Gajalakshmi & Peto, 2006). Second, some chronic diseases are difficult to
diagnose even in clinical settings (Yang et al., 2006). Finally, even with use of advanced medical
care, it is difficult to determine the underlying cause of death among very old and frail persons.
Understanding the value of verbal autopsy when used among older survey respondents has
implications for the development of cause of death data in the developing world and for using
data from surveys to understand mortality differentials and trends. To our best knowledge, no
existing research has addressed the use of VA in an older population in a large-scale national
study. This paper compared approaches to ascertaining cause of death in a sample representative
of the older Chinese population.

VA results can be significantly influenced by instruments, respondents, interviewers,
recall periods and the methods used to interpret the data (Fottrell & Byass, 2010). The
standardized VA instrument developed by the World Health Organization (WHO), uses a
combination of open-ended questions asking about the illness/events leading to the death and a
series of closed-ended questions asking about whether specific symptoms and signs were present
(Byass et al., 2012). Either computer-coded VA or physician review is then used to determine
cause of death. The advantage of computer-coded VA over physician review is that it makes the
coding faster and cheaper and improves inter-observer consistency and comparability (Byass et
al., 2012; Leitao et al., 2014); however, computer-coded VA can only use closed-question data
which may not adequately capture potentially relevant information, especially details about
chronic conditions. Physician review usually incorporates written narrative text along with
responses to closed questions to make ascertainment. Open-ended narratives from survivors may

130
be particularly important in assigning diagnoses when they provide information from hospital
reports and other circumstances surrounding death.
VA collects information retrospectively from family members/relatives who may not
always have complete information about the deceased. When VA is applied in longitudinal
household surveys, critical information on health provided by the deceased before death can be
incorporated in the cause ascertainment. In this study, we combine data from the VA interview,
both closed and open-ended questions, and the baseline survey of respondents to make a COD
ascertainment. We compare these results to those derived using the closed ended questions and
InterVA, a computer-based model providing a probabilistic ascertainment of cause of death
developed by Peter Byass and colleagues (InterVA-4, 2007). In addition, we evaluate how
characteristics of the deceased, respondents, and VA interview affect the comparison. The results
from this study are important for understanding the value of including survivor VA interviews in
the context of population-based surveys.

Causes of Old-age Mortality in China
Like many other developing countries, China has gone through an epidemiological
transition in the 20th century, resulting in chronic diseases becoming the leading causes of death
with mortality increasingly concentrated in the older ages. This estimate of the distribution of
causes of death comes from Chinese government data which are based on data from a selective
surveillance area rather than the entire country. Much of the data even in the surveillance area
for attributing causes to deaths occurring at home are from family members and responses to a
VA questionnaire. Yang et al. (2006) conducted a VA validation study using 3290 deaths (90%
of deaths occurred after age 45) among residents in six cities in urban China. The results suggest

131
that VA performed well for detecting deaths from stroke, cancer, and transportation, but was less
reliable for ascertaining deaths due to heart disease, chronic pulmonary disease, diabetes, and
kidney disease. However, generalizability of the results from this study to all of China is limited
by their criteria for sample selection. The deceased in the study had to have been a resident of the
city and the death had to have occurred in a tertiary care health facility. Yet, by design, VA is
designed to ascertain cause-of-mortality in settings where a substantial proportion of deaths
occur outside of the hospital system.
Urban/rural setting in China is another critical factor affecting the accuracy of cause of
death determination. The urban population, in general, has better education, higher income, as
well as greater access to health services and health insurance than the rural population; therefore,
it is likely that the performance of VA would be worse when deaths occur at home and in rural
populations because families of the deceased would have less information when the death occurs
without interaction with the medical care system. Wang et al. (2007) used data from 14 Disease
Surveillance Points in rural China to compare the causes identified by VA with diagnoses in
registration data and found that VA identified more chronic obstructive pulmonary disease
(COPD) and tuberculosis (TB) and resulted in fewer deaths from ill-defined causes in rural
China. In fact, a large proportion of deaths coded as “other cardiac diseases” in the routine
system were diagnosed as COPD or TB by VA. However, because of the lack of a gold standard,
formal validation in this study was not possible.

132
DATA AND METHODS
Data
Data for this study come from the baseline (2011) wave of the China Health and
Retirement Longitudinal Study (CHARLS) and information from VA interviews collected from
survivors of those who died in the interval between the 2011 and 2013 interviews. CHARLS is a
nationally representative longitudinal survey of Chinese aged 45 years and older and their
spouses regardless of age, conducted by Peking University (Zhao et al., 2013). The national
baseline survey interviewed 17,708 individuals between June 2011 and March 2012; 431 of these
individuals died before the follow-up interview in 2013. A “proxy informant” provided details
on the illness/events that led to death, the cause of death indicated by a physician if available, as
well as questions from the VA interview. VA interviews were obtained for 404 deaths (93.7%)
between July and November of 2013.

Methods
Ascertaining Causes of Death in CHARLS
Respondents to the CHARLS VA interview were primarily spouses, children, or relatives
of the deceased. CHARLS interviewers administered the 2012 WHO verbal autopsy instrument,
which was translated to Chinese by the CHARLS team and then back-translated to English by
staff at the China Center for Disease Control and Prevention (China CDC) to ensure accuracy.
Before adopting the VA instrument, the China CDC conducted a pre-test of the Chinese VA
questionnaire for 14 deaths of persons 50 or older, comparing causes of death determined by VA
with those determined by China CDC experts based on clinical information as well as proxy
interview. The pre-test showed a match of cause of death in over half (57%) of deaths. There was

133
no evidence to suggest that the discrepancies were due to questionnaire translation from English
to Chinese.
The CHARLS VA questionnaire consists of questions that cover basic sociodemographic
information of the deceased, the time, place and date of death, history of medical conditions,
history of injury or accident, treatment and health service use during the period of final illness, as
well as information on death certificates if available. Besides closed-ended questions,
respondents were asked to provide the three most likely causes of death as part of the VA. More
than half of respondents provided the names of the diseases that led to mortality based on their
knowledge (e.g. died of esophageal cancer). About a quarter of respondents described the
symptoms and circumstances preceding death in detail (e.g., “diagnosed with hypertension, had
stroke 6-7 years ago, had cerebral hemorrhage and died”). Some respondents (about 20%) could
not provide meaningful details about deaths (e.g., “Old age,” “no particular reason,” “died
peacefully”). Because we are applying the VA in a survey with prior information from the
respondent on health, we incorporate information on prior diagnosis of hypertension, diabetes,
cancer, asthma, COPD, stroke, kidney disease, and liver disease reported by the respondent in
the earlier household survey to indicate presence of diseases which are asked about in the VA
questionnaire. If the deceased reported having been diagnosed with these diseases when they
were interviewed, we code him/her as having the disease even if the respondents to the VA did
not report it.
Data from the 404 interviews were processed using the InterVA4 software (Version
4.04). Coding of the VA data with the software involves using “yes” answers to questions; no
distinction is made between “no” or “don’t know” responses to the questions. In addition,

134
InterVA4 requires pre-setting two basic epidemiological parameters, the prevalence of
HIV/AIDS and malaria. Both were coded as very low among older persons in China.

As an alternative approach to the coding of cause of death using software, three of the
authors of this paper, including a Chinese speaking physician and an expert on mortality among
older persons, reviewed the reported causes of death and the responses to the closed-ended VA
questions provided by respondents as well as the health conditions reported at the first survey
interview, and assigned a likely cause for each deceased person and then aggregated the causes
to the WHO simplified cause of death list (World Health Organization, 2011). Because the
number of deaths due to nutritional and endocrine disorders, gastrointestinal disorders, and renal
disorders is small, we grouped these causes into “Other Non-Communicable Diseases.”
We first compared the distributions of the most likely COD identified by the VA software
to the cause identified by the research team. To understand whether urban and rural residence
affects the value of VA interviews and how this would affect COD ascertainment, respondents
are classified based on both their usual place of residence and their official place of household
registration (hukou) when they were alive. Because official registration status can affect the use
of medical services, three categories were denoted: rural residency, urban residency with rural
hukou, and urban residency with urban hukou. The small percentage (N=7, 1.73%) who lived in
rural areas but had urban hukou are included in the rural residency category. This three-category
variable separates those living in urban areas by the availability and ability to use medical
services and obtain benefits.
To explore the pattern of mismatch between the two approaches to determining cause of
death and related factors, we conducted logistic regression models to predict whether the most

135
likely cause of death as determined by InterVA matched the cause determined by experts. The
predictors were demographics of the deceased, the characteristics of the VA interview, the
relationship between the respondent of the VA interview and the deceased, the time interval
between death and VA interview, and place of death. We also explored what factors related to
an inability of expert review to assign a cause of death using a logistic regression model.

RESULTS
Demographic characteristics of the deceased are presented in Table 1. The mean age at
death was 72.45 with a standard deviation of 10.9 years. Almost two-thirds (63.4%) of the
descendants resided in rural China. Urban deaths were almost equally split between those with
rural hukou and urban hukou (17.8% versus 18.8%). Females comprised 44.1% of the deaths.
However, among those who had rural hukou but resided in urban areas, more than half of the
deceased were females (56.9%). Most of the deceased did not complete primary school. More
urban residents with urban hukou reported having junior/secondary school or higher education
(43.4%), compared to 10.9% of rural residents and 9.7% of urban residents with rural hukou.
Overall, most of the deceased died at home; although the percentage dying in hospitals was
lowest among rural residents (8.2%) and highest among urban residents with urban hukou
(42.1%), showing a clear rural-urban gradient. 74.8% of the VA interviews were provided by a
child or spouse of the deceased; the remaining VA interviews were answered by
relatives/neighbors/friends of the deceased. 45.5% of the interviews were taken within a year
after death. The differences between residence categories in respondent and time were not
significant.

136
Table 2 reports the distributions of COD derived from two methods. Both InterVA and
expert review indicate neoplasms as the most frequent cause of death in this age group (26.5%
for InterVA diagnosis and 24.3% based on expert review). InterVA and expert review suggest
very similar proportions of deaths due to heart disease (17.1% versus 17.3%). InterVA classified
25.0% of deaths as due to stroke; 9% higher than by expert review and about 5% higher than in
official statistics.
InterVA classified more cases as due to infectious and parasitic diseases than the expert
review (8.4% versus 4.0%). Noticeably, InterVA attributed 4.7% of the deaths to pulmonary
tuberculosis while the percentage by expert review is only 1.2%. Both methods yield higher
proportions of TB deaths compared to the 0.3% TB deaths reported by the official mortality
statistics. Only 6.2% of deaths were identified as respiratory disorders by InterVA and 4.2% by
expert review. Deaths due to COPD were low according to both methods used (4.7% by InterVA
and 3.0% by expert review). It is notable that the cause of death for 22.8% of the deaths in this
sample could not be determined based on expert review.

Table 3 compared individual cause of death assignment by InterVA and by expert review.
The highest levels of matching for the two approaches were for external causes of death and
neoplasms (88.2% and 67.3%, respectively). Among the 17 deaths diagnosed as external causes
by InterVA, experts identified 15 of them as external causes. One of the nonmatching cases was
assigned to stroke and no cause was assigned for the other. Among the 107 deaths identified as
neoplasms by InterVA, 72 (67.3%) were also identified as such by expert review, and 18
(16.8%) could not be assigned a cause based on the responses provided by survivors. Among the
101 stroke deaths according to InterVA, 49 (48.5%) were classified as stroke and 17 (16.8%) as

137
cardiac disease by experts, whereas experts could not determine a cause in 27 (26.7%) cases. The
percentage matching between InterVA and expert review is 46.4% for cardiac disease. The
lowest levels of concordance were found for infectious and parasitic disease (20.6%) and other
NCDs (14.7%). Among the 34 deaths determined by InterVA as being due to other NCDs,
including nutritional and endocrine disorders, gastrointestinal disorders, and renal disorders, only
5 (14.7%) of them were classified as such; among the remaining deaths, cause for 9 deaths
(26.5%) could not be determined based on expert review. The reason for the low level of
matching is that a large proportion of deaths diagnosed as other NCDs by InterVA could not be
assigned a cause by experts because the families/relatives of the decedents did not provide
explicit and meaningful information about the deaths. This is also true for infectious and
parasitic diseases, and respiratory disorders which have low levels of matching. For the 92 deaths
where experts failed to determine a likely cause after reviewing the written narrative text and the
responses to the standard closed-ended VA questions, InterVA diagnosed 8 of them as infectious
and parasitic disease, 18 as neoplasms, 17 as cardiac disease, 27 as stroke, 3 as respiratory
disorders, 9 as other NCDs, 1 as external cause, and could not determine a cause for 9 deaths.
Based on the results from Table 3, we divided the sample into two groups, the matched
cases and the unmatched cases. Table 4 presents the results from logistic regression models
predicting the determinants of mismatch. There is greater mismatch among older deaths.
Compared to those who died before age 70, the odds ratio of being diagnosed with a different
cause of death if a death occurred at 80 or over is 2.5 (95% CI: 1.4-4.5). While only marginally
significant, deaths that occurred at home are almost twice as likely to be assigned different
causes compared to deaths occurring in hospitals (OR=1.8). The other indicators were not
significantly related to mismatched cases.

138
We also examined the characteristics of the deaths for whom causes were not assigned by
expert review (Table 5). Deaths for whom experts cannot assign a cause are more likely to occur
among those 70 and over. Compared to those who died before 70, the odds ratio of having an
indeterminate cause of death is 2.4 (95% CI: 1.2-5.1) for those who died in their 70s, and 7.0
(95% CI: 3.2-15.3) for those who died at age 80+.

CONCLUSION AND DISCUSSION
This study used data from the CHARLS VA interviews and initial interview to ascertain
cause of death in a sample representative of the older Chinese population. The representativeness
of the study sample ensures that the results can be generalized to both urban and rural China.
Neoplasms, cardiac disease, and stroke are the most frequent causes of death determined by both
InterVA and expert review, accounting for about 60-70% of deaths. For these three major causes
of death, the percentage matching between the two approaches is about two-thirds (67.3%) for
neoplasms, and half for stroke (48.5%,) and for cardiac disease (46.4%). The main reason for the
lack of a match between the two methods is that with expert review no cause was assigned for
more than a quarter of the sample. A higher likelihood of having mismatch is associated with
advanced age and deaths occurring at home. Indeterminate deaths occurred primarily among
those aged 80+ who were likely to be frail. The InterVA is designed to assign a cause for these
cases, but we are not confident about the accuracy of the InterVA assignments because
assignment is based on insufficient relevant details for the deaths. The InterVA appears to
perform well when the cause of death is characterized by distinctive signs and symptoms that do
not overlap with those common to other causes. If a cause is characterized by symptoms and

139
signs that are common to multiple causes, it is not appropriate to distinguish between the
different causes.
Interestingly, while the percentage of matching assignments for the InterVA and expert
review is only about 50% at the individual level, the two approaches yield similar percentages of
deaths for major causes of deaths such as neoplasms and cardiac disease at the population level.
Misclassification only affects the accuracy of the InterVA assignment at the population level
when there is an imbalance among misclassifications.
Ascertaining COD in older populations has specific challenges. Multimorbidity is highly
prevalent among older adults. Older decedents often suffer from multiple life-threatening
conditions at the time of death. The fact that InterVA can provide probabilities for multiple
causes of death may be an advantage as it to some extent reflects multimorbidity. Furthermore,
many chronic conditions not only have similar risk factors but also share common signs and
symptoms. For instance, breathlessness is a symptom for many possible causes including COPD,
lung cancer, pneumonia, as well as congestive heart failure. These make the identification of a
single cause for each death a challenge. Although assignment of multiple causes of death
requires further methodological thinking, a feasible approach is to use the broad-category
method (Fottrell & Byass, 2010; Yang et al., 2006). If causes with shared etiology or common
risk factors are considered together, agreement will be raised and the accuracy of the assignment
will be improved.

Validation studies usually use clinical diagnosis as the gold standard for cause of death
(Byass, 2014; Lozano et al., 2011; Setel et al., 2006; Yang et al., 2006). In this study, the absence
of a reference standard makes it impossible to compare the results to a gold standard to evaluate

140
which method would most closely match a clinical diagnosis. However, high levels of agreement
between the results as diagnosed by the two approaches provide greater confidence in the ability
of the software to correctly assign some causes; whereas low consistency indicates that more
investigation is needed to be confident in the software assigned causes and expert coded
assignments. Including a narrative section with the VA instrument can provide more evidence for
COD ascertainment among older adults. The coding of the Inter VA could be redesigned
specifically to be used with older populations. As it is it appears to be a useful tool, in
conjunction with expert review, for getting insight into all causes of death as well as useful for
some individual causes.
It is important to note several limitations of this study. First, as noted earlier, the lack of a
gold standard makes it impossible to draw a conclusion on the relative accuracy on the
performance of the two approaches used. Second, the sample size is relatively small, particularly
for rare causes, since only two-year mortality was assessed. It will be worthwhile to re-examine
the agreement between the results as diagnosed by experts and InterVA when more deaths are
observed over time. However, a major strength of this study is that it is the first study to ascertain
COD for participants of a nationally representative survey of the older Chinese population.
While many longitudinal household surveys are collecting verbal autopsy data in the developing
world (Streatfield et al., 2014), the results of this study provide important empirical evidence of
the value of conducting VA interviews in the context of population-based surveys of older
persons.

141
REFERENCES
Aggarwal, A. K., Kumar, P., Pandit, S., & Kumar, R. (2013). Accuracy of WHO Verbal Autopsy
Tool in Determining Major Causes of Neonatal Deaths in India. PLOS ONE, 8(1), e54865.
Retrieved from https://doi.org/10.1371/journal.pone.0054865
Byass, P. (2014). Usefulness of the Population Health Metrics Research Consortium gold
standard verbal autopsy data for general verbal autopsy methods. BMC Medicine.
https://doi.org/10.1186/1741-7015-12-23
Byass, P., Chandramohan, D., Clark, S. J., D’Ambruoso, L., Fottrell, E., Graham, W. J., …
Tollman, S. M. (2012). Strengthening standardised interpretation of verbal autopsy data: the
new InterVA-4 tool. Global Health Action. https://doi.org/10.3402/gha.v5i0.19281
Byass, P., Herbst, K., Fottrell, E., Ali, M. M., Odhiambo, F., Amek, N., … Tollman, S. M.
(2015). Comparing verbal autopsy cause of death findings as determined by physician
coding and probabilistic modelling: a public health analysis of 54 000 deaths in Africa and
Asia. Journal of Global Health, 5(1), 10402. https://doi.org/10.7189/jogh.05.010402
Fottrell, E., & Byass, P. (2010). Verbal autopsy: Methods in transition. Epidemiologic Reviews.
https://doi.org/10.1093/epirev/mxq003
Gajalakshmi, V., & Peto, R. (2006). Commentary: Verbal autopsy procedure for adult deaths.
International Journal of Epidemiology, 35(3), 748–750. https://doi.org/10.1093/ije/dyl112
InterVA-4. (2017). InterVA-4 User Guide, Version 4.04. Retrieved from http://www.interva.net/
Leitao, J., Chandramohan, D., Byass, P., Jakob, R., Bundhamcharoen, K., Choprapawon, C., …
Mathai, M. (2013). Revising the WHO verbal autopsy instrument to facilitate routine cause-
of-death monitoring. Global Health Action. https://doi.org/10.3402/gha.v6i0.21518
Leitao, J., Desai, N., Aleksandrowicz, L., Byass, P., Miasnikof, P., Tollman, S., … Jha, P.

142
(2014). Comparison of physician-certified verbal autopsy with computer-coded verbal
autopsy for cause of death assignment in hospitalized patients in low- and middle-income
countries: Systematic review. BMC Medicine. https://doi.org/10.1186/1741-7015-12-22
Lozano, R., Lopez, A. D., Atkinson, C., Naghavi, M., Flaxman, A. D., & Murray, C. J. L.
(2011). Performance of physician-certified verbal autopsies: Multisite validation study
using clinical diagnostic gold standards. Population Health Metrics.
https://doi.org/10.1186/1478-7954-9-32
Mahapatra, P., Shibuya, K., Lopez, A. D., Coullare, F., Notzon, F. C., Rao, C., & Szreter, S.
(2007). Civil registration systems and vital statistics: successes and missed opportunities.
The Lancet, 370(9599), 1653–1663. https://doi.org/https://doi.org/10.1016/S0140-
6736(07)61308-7
Setel, P. W., Whiting, D. R., Hemed, Y., Chandramohan, D., Wolfson, L. J., Alberti, K. G. M.
M., & Lopez, A. D. (2006). Validity of verbal autopsy procedures for determining cause of
death in Tanzania. Tropical Medicine & International Health : TM & IH, 11(5), 681–696.
https://doi.org/10.1111/j.1365-3156.2006.01603.x
Streatfield, P. K., Khan, W. A., Bhuiya, A., Hanifi, S. M. A., Alam, N., Ouattara, M., … Byass,
P. (2014). Cause-specific childhood mortality in Africa and Asia: Evidence from INDEPTH
health and demographic surveillance system sites. Global Health Action.
https://doi.org/10.3402/gha.v7.25363
Tensou, B., Araya, T., Telake, D. S., Byass, P., Berhane, Y., Kebebew, T., … Reniers, G. (2010).
Evaluating the InterVA model for determining AIDS mortality from verbal autopsies in the
adult population of Addis Ababa. Tropical Medicine & International Health : TM & IH,
15(5), 547–553. https://doi.org/10.1111/j.1365-3156.2010.02484.x

143
Wang, L., Yang, G., Ma, J., Rao, C., Wan, X., Dubrovsky, G., & Lopez, A. D. (2007).
Evaluation of the quality of cause of death statistics in rural China using verbal autopsies.
Journal of Epidemiology and Community Health. https://doi.org/10.1136/jech.2005.043927
World Health Organization. (2011). Review of the WHO Verbal Autopsy (VA)Instruments.
Retrieved from
http://www.who.int/healthinfo/statistics/WHO_VerbalAutopsy_MeetingReport_Dec2012.p
df?ua=1
Yang, G., Rao, C., Ma, J., Wang, L., Wan, X., Dubrovsky, G., & Lopez, A. D. (2006).
Validation of verbal autopsy procedures for adult deaths in China. International Journal of
Epidemiology. https://doi.org/10.1093/ije/dyi181
Zhao, Y., Strauss, J., Yang, G., Giles, J., Hu, P., Hu, Y., … Wang, Y. (2013). China health and
retirement longitudinal study–2011-2012 national baseline users’ guide. National School of
Development, Peking University.

144
TABLES/FIGURES
Table 1. Characteristics of the Deceased in CHARLS, 2011-2013

All
Rural
Residency
Urban Residency

Rural
Hukou
Urban
Hukou
N=404
N=256
63.4%
N=72
17.8%
N=76
18.8%
Chi-square test
Age at death, Mean (SD)
72.5
(10.9)
72.7
(10.7)
72.6
(10.8)
71.5
(11.6)

Female, % 44.1 43.4 56.9 34.2 p=0.019
Education of the deceased, % p<.001
No formal schooling 64.4 71.5 73.6 31.6
Primary school 18.3 16.8 16.7 25.0
Junior/Secondary + 16.8 10.9 9.7 43.4
Missing 0.5 0.8 0.00 0.00
Place of death, % p<0.001
Hospital 15.4 8.2 12.5 42.1
Home 81.9 90.6 80.6 54.0
Other places 2.7 1.2 6.9 4.0
Relationship of respondents to the deceased, % p=0.304
Children 34.7 37.1 34.7 26.3
Spouse 40.1 36.7 41.7 50.0
Others 25.3 26.2 23.6 23.7

145
Time interval between death and VA interview, % p=0.726
Less than 1 year 45.5 46.9 41.7 44.7

146
Table 2. Percentages of Causes of Death Determined by InterVA and Expert Review
InterVA Expert Review
Infectious and parasitic diseases 8.4 4.0
Acute respiratory infection, including pneumonia 2.2 2.5
HIV/ADIS related death 0.3

Diarrheal diseases 0.5

Pulmonary tuberculosis 4.7 1.2
Other and unspecified infectious disease 0.7 0.3
Neoplasms 26.5 24.3
Oral neoplasms 0.3 1.0
Digestive neoplasms 10.4 10.2
Respiratory neoplasms 6.4 5.5
Breast neoplasms 0.3 0.7
Reproductive neoplasms 2.5 0.5
Other and unspecified neoplasms 6.7 6.4
Stroke 25.0 16.1
Diseases of the circulatory system, excluding stroke 17.1 17.3
Acute cardiac disease 5.2 5.9
Other and unspecified cardiac diseases 11.9 11.4
Respiratory disorders 6.2 4.2
Chronic obstructive pulmonary disease 4.7 3.0
Asthma 1.5 1.2
Other NCDs 8.4 6.0

147
Nutritional and endocrine disorders 3.0 2.7
Gastrointestinal disorders 4.2 2.5
Renal disorders 1.2 0.7
External causes of death 4.2 5.5
Indeterminate 4.2 22.8

148
Table 3. Differential classification Matrix for Major Causes of Death, CHARLS 2011-13
Expert Review
InterVA
Assignment
Infectious
&
parasitic
diseases
Neoplasms
Cardiac
disease
Stroke
Respiratory
disorders
Others
NCD*
External
causes
Indeterminate Total
%
Matching
%
Indeterminate
Infectious &
parasitic diseases
7 9 3 2 4 1 0 8 34 20.6 23.5
Neoplasms 2 72 2 3 2 7 1 18 107 67.3 16.8
Cardiac Disease 2 7 32 4 2 5 0 17 69 46.4 24.6
Stroke 1 1 17 49 1 3 2 27 101 48.5 26.7
Respiratory
disorders
3 3 4 2 8 1 1 3 25 32.0 12.0
Others NCDs* 0 5 9 3 0 5 3 9 34 14.7 26.5
External causes 0 0 0 1 0 0 15 1 17 88.2 5.9
Indeterminate 1 1 3 1 0 2 0 9 17 45.0 /
Total 16 98 70 65 17 24 22 92 404

Note: Other NCDs include nutritional and endocrine disorders, gastrointestinal disorders, and renal disorders.

149
Table 4. Odds Ratios from the Logistic Model Predicting the Mismatch between the
Diagnosis of InterVA and Expert Review

OR 95% CI
Age (ref=less than 70 years)
70-79 1.4 (0.8, 2.2)
80+ 2.5 (1.4, 4.5)
Female 1.2 (0.8, 1.9)
Urban-Rural (ref=rural residency)

urban residency, rural hukou 1.0 (0.6, 1.7)
urban residency, urban hukou 0.8 (0.5, 1.5)
Education (ref=no formal schooling)

primary school 0.7 (0.4, 1.2)
junior school or above 1.6 (0.8, 3.0)
Place of death (ref=hospital)

Home 1.8 (1.0, 3.3)
Other 0.5 (0.1, 2.3)
Relationship to the deceased (ref=children)

Spouses 0.9 (0.5, 1.6)
Other 0.8 (0.5, 1.4)
Time duration between death and VA interview (ref= ≤1yr)

More than 1 year 1.0 (0.7, 1.5)

150
Chapter 6: Conclusion
The four papers in this dissertation have made significant contributions to our
understanding of the life-course processes leading to physiological dysregulations and mortality
in the current cohort of older Chinese and shed lights on the factors that may underlie the diverse
aging processes across populations. The first study examines urban-rural differences in multiple
indicators of physiological dysregulation which are precursors or indicators of chronic diseases
associated with aging and identifies individual characteristics and environmental factors at the
community level related to age-related biological risk. Lifetime exposure to urban areas is an
important risk factor for increased biological risk. The higher status of urban dwellers does not
appear to be the source of their poorer physiology. The urban–rural differential in biological risk
is accounted for by adjusting for health behaviors, particularly physical activity. The second
study takes advantages of a recent life history survey to investigate the early-life origins of later
life cardiovascular and metabolic risk. Contrary to the strong links between early-life adversity
and cardiometabolic risk documented in developed countries, we do not find convincing
evidence for strong connections between childhood experiences and cardiovascular and
metabolic risk in China. We do find adult life circumstances, particularly exposure to urban
environments, relate to both cardiovascular and metabolic risk. Our findings highlight the
complexity of this relationship, suggesting the impact of childhood adversity is affected mainly
by social and historical context. The third study investigates the determinants of mortality from
a life course perspective. We assessed associations between four-year mortality and
socioeconomic conditions, early life indicators, biological markers, physical functioning, disease
presence, and community characteristics. We find that low education and rural-residence, poor
physical functioning, uncontrolled hypertension, diabetes, cancer, a high level of systemic
151
inflammation, and poor kidney functioning are strong predictors of mortality among older
Chinese. We do not observe linkages between early-life experiences or community infrastructure
to mortality at older ages. The fourth study addresses ascertaining causes of death using verbal
autopsy in an older population, aiming to improve the mortality measurement in population-
based surveys. We compare the cause of death derived using computer-coded and expert review
verbal autopsies. Results show that both approaches yield similar percentages of deaths for major
causes at the population level, but the percentage of matching assignments at the individual level
is only about 50%. Advanced age and deaths occurring at home are associated with a higher
probability of having a mismatch between the two methods. The primary reason for the lack of a
match is that expert review did not assign a specific cause for more than a quarter of the sample.
This study provides empirical evidence of the value of conducting verbal autopsy interviews in
the context of population-based surveys of older persons.
Taken together, this dissertation uses newly-collected data from a nationally
representative sample to study physiological health and mortality in the unique Chinese context.
We examined the relationships of biological risk factors and mortality to a wide range of social
factors, many of which are culture, cohort, and context-specific. This dissertation adds to the
literature emphasizing the complexity of the consequences of life experiences on health and
highlights that societal factors exert substantial influences on the aging process. Therefore,
researchers should acknowledge the differential historical and societal contexts between
countries and the changing context over time when studying how factors at different levels shape
health and aging in later life through the life course.
152
Many results of from our analysis seem to contradict literature from many developed
countries; however, in fact, the inconsistent relationships across populations suggest the
pathways by which life experiences get under the skin is both biological and social, and largely
affected by the historical and social environments in which individuals become old. For instance,
in this work, we did not find strong evidence on the association between early-life conditions and
cardiometabolic health. This does not suggest that that the body forgets the hardships that have
been experienced. It is important to note that the current older Chinese are the survivors of a
series of wars, famines, infectious diseases, and social-political movements. It is estimated that
the Great Famine has led to 15-30 million excess deaths in China (Chen & Zhou, 2007). We still
have relatively little knowledge about the health and aging in this unique population. Because of
strong survival selection, those with extremely adverse early-life experiences and poor health
may have experienced mortality prior to the study, leaving individuals with high resilience who
managed to survive adversities thus distorting the true relationship between childhood hardships
and later-life health. However, because China has experienced an extraordinary mortality
reduction, especially from under-5 mortality and infectious diseases, between the 1950s and the
1980s (Zhao, Chen, & Jin, 2016), the effect of survival selection would have also decreased over
time. Thus, it is possible to observe stronger effects of some early-life experiences on later-life
health in the later-born cohorts. Studies from Mexico, another country which has undertaken
extensive changes in a relatively short time, find the association between education and mortality
differs across birth cohorts - low education was only linked to higher mortality risk in the
younger cohort but not the older cohort (Saens & Wong, 2015). Indeed, China and other
countries which have also undertaken rapid changes offer valuable opportunities to examine how
determinants of health and aging may differ across cohorts. For instance, results from this work

153
show that exposure to the urban environment is associated with an increased level of
physiological dysregulation. As the Chinese population moving through the nutrition transition,
the urban population first moved out of undernutrition and started to become overweight or
obese. However, it is also very likely that the most educated and highest income urban
population are also the first to adopt a healthy lifestyle to avoid the adverse effects of being
overweight/obese (Strauss and Thomas, 2008). We could expect to observe dynamic changes in
physiological health and disease burden between urban and rural Chinese older adults.
Understanding these dynamic changes across time and cohorts has public health and policy
implications. As the rural population is currently moving through the nutrition transition, they
may face a larger burden of chronic diseases because of the lack of health literacy, resources, and
information.
Another importance of this work is that it studies health from specific domains of health
through a wide array of health measures. Due to data availability, previous studies on health and
aging in older Chinese primarily focused on mortality, ADL/IADL disability, and self-reported
health. This study focused on more upstream measures, such as biological risk and physical
assessments, which are precursors of downstream health outcomes; therefore, the results are
valuable for disentangling associations between various lifetime exposures and specific aspects
of health. At the population level, morbidity process provides a comprehensive framework to
study aging that involves multiple health dimension (Crimmins, Kim, & Vasunilashorn, 2010). It
is worthwhile to note these indicators may not change in the same direction. In fact, studies in
this dissertation show a complex picture of health status in different populations. For example,
the urban population has worse physiological health but exhibits a strong survival advantage
compared to their rural peers, highlighting an important role of environments in the aging

154
process. Future work should further address differentials in the morbidity process across
populations or subpopulations.

Future Research
Building upon the work I presented in this dissertation, I plan to take advantage of the
longitudinal design of CHARLS to conduct longitudinal analysis to examine the changes of
psychological health and physical functioning since the second wave of biomarker data will be
available soon. Examining changes in biological risk can identify the physiological systems
driving worsening or improving health. Linking social, behavioral, and environmental factors to
health change will help us understand the heterogeneity in aging within the population. In the
long run, it is also feasible and essential to examine these associations across cohorts understand
the dynamic aging process.
Also, I am proposing to use the biomarker data from the U.S. Health and Retirement
Study and its sister studies around the world to examine cross-national differences in
pathophysiological processes of aging. The cross-calibrated biomarker data will allow me to
conduct direct comparisons on inflammation, glucose regulation, and lipid profiles. These data
are highly comparable and representing countries at different levels of development, thus
providing excellent opportunities to study the influences of social factors at different levels on
biological embodiment in aging. Results from this study will clarify biological pathways that
result in differentials in health and mortality across countries and help promote healthy aging
globally.

155
REFERENCES
Chen, Y., & Zhou, L. A. (2007). The long-term health and economic consequences of the 1959–
1961 famine in China. Journal of health economics, 26(4), 659-681.
Crimmins, E., Kim, J. K., & Vasunilashorn, S. (2010). Biodemography: New approaches to
understanding trendsand differences in population health and mortality. Demography, 47(1),
S41-S64.
Strauss, J., & Thomas, D. (2008). Health over the life course, Vol. 4 of Handbook of
Development Economics.
Saenz, J. L., & Wong, R. (2015). A life course approach to mortality in Mexico. salud pública de
méxico, 57, s46-s53.
Zhao, Z., Chen, W., & Jin, Y. (2016). Recent mortality trends in China. In Contemporary
Demographic Transformations in China, India and Indonesia (pp. 37-53). Springer, Cham.

Abstract (if available)

Abstract The demand for understanding context-specific determinants of health and aging is greater than ever. Literature has documented substantial heterogeneity in a variety of age-related health outcomes across countries, suggesting that age-related health change takes place in a context and that the role of social, political, and environmental determinants of health may vary with context. Most existing evidence on social determinants of health and aging is from developed countries. However, individual risk factors are operated under societal factors. China, as a developing and transitional country, is an opportune environment for studying this topic as the unique historical, political, social, and economic circumstances may result in context-specific determinants of health and mortality. ❧ This dissertation used newly-collected data from a nationally representative sample to study the life-course processes leading to physiological dysregulations and mortality in the current cohort of older Chinese and shed lights on the factors that may underlie the diverse aging processes across populations. The first study examines urban-rural differences in multiple indicators of physiological dysregulation which are precursors or indicators of chronic diseases associated with aging and identifies individual characteristics and environmental factors at the community level related to age-related biological risk. I find that exposure to urban areas is an important risk factor for increased biological risk among older Chinese. The urban–rural differential in biological risk is accounted for by adjusting for health behaviors, particularly physical activity. The second study examines the early-life origins of later life cardiovascular and metabolic risk. No convincing evidence for strong connections between childhood experiences and cardiovascular and metabolic risk in China is found. The results contradict the findings from many developed countries, suggesting the consequences of childhood adversity is affected mainly by social and historical context. The third and fourth studies focused on mortality. The third chapter investigates the determinants of mortality from a life course perspective. We find that current education and place of residence, assessments of physical functioning, uncontrolled hypertension, diabetes, cancer, a high level of systemic inflammation, and poor kidney functioning are strong predictors of mortality among older Chinese. The results from this study highlight the value of incorporating biological and performance measurements and the importance of social and historical context in studying old-age mortality. The last study addresses ascertaining causes of death using verbal autopsy in an older population, aiming to improve the mortality measurement in population-based surveys. In this study, I compared the cause of death assignments derived using computer-coded and expert review verbal autopsies. Both approaches yield similar percentages of deaths for major causes at the population level, but the percentage of matching assignments at the individual level is only about 50%. The primary reason for the lack of a match is that expert review did not assign a specific cause for more than a quarter of the sample. ❧ This dissertation emphasizes the complexity of the consequences of life experiences on health and highlights that societal factors exert substantial influences on the aging process. Some findings from this dissertation seem to contradict literature from developed countries

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Statistical algorithms for examining gene and environmental influences on human aging

PDF

Internet communication use, psychological functioning and social connectedness at older ages

PDF

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

PDF

Health differences between the elderly in Japan and the United States

PDF

A biodemographic approach to understanding sociodemographic disparities in kidney functioning on three dimensions: individual, population, and cross-national

PDF

Self-perceptions of Aging in the Context of Neighborhood and Their Interplay in Late-life Cognitive Health

PDF

International sex and age differences in physical function and disability

PDF

Three essays on modifiable determinants of shingles: risk factors for shingles incidence and factors affecting timing of vaccine uptake

PDF

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

PDF

Essays on health and aging with focus on the spillover of human capital

PDF

Estimating survival in the face of pain: evidence from the health and retirement study

PDF

Air pollution neurotoxicity throughout the lifespan: studies on the mechanism of toxicity and interactions with effects of sex and genetic background

PDF

Residential care in Los Angeles: policy and planning for an aging population

PDF

Computational approaches to identify genetic regulators of aging and late-life mortality

PDF

Age integration in late life: sociodemographic & psychosocial correlates of intergenerational-only, peer-only, and age-integrated social networks

PDF

Aging in a high infection society

PDF

Psychological and physiological pathways from social support exchanges to health: a lifespan perspective

PDF

The effect of body weight on mortality: different countries and age groups

PDF

Lifting up ourselves by lifting up others: examining cognitive-affective states linking generativity to well-being

PDF

Examining the longitudinal influence of the physical and social environments on social isolation and cognitive health: contextualizing the role of technology

Asset Metadata

Creator Zhang, Yuan (author)

Core Title Social determinants of physiological health and mortality in China

School Leonard Davis School of Gerontology

Degree Doctor of Philosophy

Degree Program Gerontology

Publication Date 08/12/2019

Defense Date 02/25/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag aging,biomarkers,China,mortality,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Crimmins, Eileen M. (committee chair), Ailshire, Jennifer A. (committee member), Strauss, John A. (committee member)

Creator Email yuan_zhang@unc.edu,yuanz1010@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-212563

Unique identifier UC11663432

Identifier etd-ZhangYuan-7772.pdf (filename),usctheses-c89-212563 (legacy record id)

Legacy Identifier etd-ZhangYuan-7772.pdf

Dmrecord 212563

Document Type Dissertation

Format application/pdf (imt)

Rights Zhang, Yuan

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