T313;

KL)
C\
6
D

This is to certify that the

thesis entitled

Determinants of Comorbidity at the time of
Death in Four Asian American Groups

presented by

Mira M. Hidaj at

has been accepted towards fulfillment
of the requirements for

M°A- Jegree in m

 

 

k—Mr professor

 

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDue.p85.p.15

 

DETERMINANTS OF COMORBIDITY AT THE TIME OF DEATH
IN FOUR ASIAN AMERICAN GROUPS

By

Mira M. Hidajat

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS
Department of Sociology

2002

ABSTRACT

DETERMINANTS OF COMORBIDITY AT THE TIME OF DEATH IN FOUR ASIAN
AMERICAN GROUPS

By
Mira M. Hidajat

This study examined the intersection between morbidity and mortality in four
Asian American populations, namely Japanese, Korean, Chinese, and Vietnamese
Americans. It uses the total cause approach developed by Johnson and Christenson
(1998) of summing the total causes of deaths mentioned in the death certificate as an
indicator of comorbidity. Data were taken from the 1997 United States Multiple Cause
of Death file produced by the NCHS. The results of the observed means analysis show
that the effects of socio-demographic variables such as age, sex, education, marital status,
and nativity are generally small. Larger differences in average level of comorbidity are
found by race, residence in the West, and underlying cause of death (UC). The results of
the multivariate analysis show different effects of these socio-demographic variables on
the risk of having high comorbidity at the time of death. Unexpected results for
Japanese, Chinese, and Korean Americans, and married persons are found and alternative

explanations are discussed.

Copyright by
MIRA M. HIDAJ AT
2002

To Ina E. Muljono with love and gratitude

iv

ACKNOWLEDGEMENT

I would like to express my gratitude to my mentors and colleagues for supporting
me in the completion of this thesis. I am deeply indebted to my mentor, Dr. Nan E.
Johnson whose research, knowledge, and vision are the inspiration for this thesis.
Without her patience, guidance, and suggestions the completion of thesis would not be
possible. I thank my committee members, Dr. Brendan P. Mullan and Dr. Steven J. Gold
for closely reviewing the thesis and providing insightful comments, especially in the
theoretical and methodological sections.

I also thank my colleagues Dr. Susan Krouse, Sissi Foster, Robert Cook, and
Corey Ray for all their help, support, interest, and counsel in the writing process of this

thesis. Finally, I thank Mark Baxter for his support and invaluable editorial assistance.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................ Vii
BACKGROUND ................................................................................................................ 1
General Asian American Mortality Research ................................................... 3
Demographic and Socioeconomic Differentials in
Four Asian American Groups ........................................................................... 5
The Total Cause Approach to Assessing Comorbidity at Death ..................... 12
RESEARCH QUESTIONS AND HYPOTHESES .................................................................... 16
METHODOLOGY ........................................................................................................... 18
Data Source ..................................................................................................... l8
Dependent Variable ......................................................................................... 19
Independent Variables ..................................................................................... 20
Statistical Analysis .......................................................................................... 23
FINDINGS ...................................................................................................................... 25
Means Analysis ............................................................................................... 25
Multivariate Analyses ..................................................................................... 27
SUMMARY AND CONCLUSION ....................................................................................... 33
REFERENCES ................................................................................................................. 44

vi

LIST OF TABLES

TABLE 1. Distribution of Total Cause (TC) ............................................................... 37
TABLE 2. Distribution of Independent Variables by Race ......................................... 38
TABLE 3. Observed Mean TC for Independent Variables by Race ............................ 40

TABLE 4. Odds Ratios of the Effects of Sociodemographic Factors and Underlying
Medical Causes on Dying with Four or More TC ...................................... 42

TABLE 5. Odds Ratios of the Effects of Sociodemographic Factors and Underlying
Medical Causes on Dying with Four of More TC by Race ........................ 43

vii

BACKGROUND

Studies that examine demographic, socioeconomic, and other differentials in
mortality and morbidity at the national level have typically in the past used either
dichotomous (Black/White) or trichotomous (Black/White/Other) categorizations of race
(See for example Manton 1980; Menchik 1993). As many of these studies use vital
statistics or Census data for calculation of death rates, a partial explanation for such a
narrow definition of race could be attributed to the lack of opportunity for the respondent
to self-identify or choose an appropriate race category on these official forms. When the
rare and sporadic opportunity does occur, it is often motivated by the contemporaneous
political agenda and thus historically there was no systematic and consistent effort to
collect these data. For example, the Census Bureau recorded groups like Chinese and
Japanese Americans before the 19003 due to the large influx of immigrants during that
time. In contrast, the 1960 U. S. census only used a Nonwhitc/White dichotomy to
measure race. Because this was before the 1965 Immigration Act, which opened doors to
immigration, there was no motivation to count persons belonging to other racial or ethnic
groups than the dichotomy. The 1970 census continued this restrictive collection of race
data despite of the passage of the 1965 Immigration Act. The result is gross
misclassification of race, as some large immigrant groups such as Asian Indians were
added to the White population, while Vietnamese Americans were classified with the “all
other races” category (US. Census Bureau 2002).

However, as the United States becomes more diverse in race, ethnicity, and
nationality, after the 1965 Immigration Act, the importance of reporting detailed race

categories was seen; it enabled the creation of health and mortality profiles of many

previously uncounted minority groups. Moreover, these profiles could be used to
enhance the delivery of health and other community services. Although it is important to
examine many varieties of ethnic and race differentials, there are several reasons why the
Asian American population is particularly interesting. The Asian American population
numbered 11.9 million in 2000 (USCB 2002). This was almost a twofold increase from
the 1990 Census, which itself was also a twofold increase from the 1980 Census (USCB
1993). The rapidly increasing rate of the Asian American population has led researchers
to examine this population more closely to determine any existence of special health or
mortality risks. Although Asian Americans consistently had the highest median
household income and high educational attainment in the last decade, this group
unexpectedly had higher poverty rates than non-Hispanic whites (NCHS 1998).
Mortality research has demonstrated the negative consequences of poverty on health and
mortality (Menchik 1993). Despite higher poverty rates, death rates among Asian
Americans remain consistently lower than Whites. These facts call for mortality and
morbidity examinations at a national level using detailed Asian categories to uncover any
within-group differences.

This study will analyze determinants of comorbidity at death among Chinese,
Japanese, Korean, and Vietnamese Americans at the national level using 1997 mortality
data. This research uses the methodology developed by Johnson and Christenson (1998)
of analyzing sociodemographic differentials in the total causes of death mentioned in
death certificates of the decedents as a way of assessing the severity of comorbidity at
death. Because this methodology has not been previously applied to Asian American

mortality and morbidity research, the nature of this study is exploratory. The next section

will discuss previous research on Asian Americans in general and the four groups
respectively to guide the expectation and explanation of the findings.
General Asian American Mortality Research

Although the Asian American population is rapidly increasing, its proportion
relative to the rest of the population is still somewhat small, composing 4.2 percent of the
total population in 2000 (USCB 2002). Many reports and studies consequently either do
not disaggregate the Asian American category by specific nationality or ethnicity or they
do not include the Asian category at all. Out of the few studies that include a general
Asian American category, a study by Rogers et al. (2000) offered insights on
determinants of low mortality among adult Asian Americans. The authors used the
National Health Interview Survey (NHIS) from 1989 to 1994 linked with the National
Death Index (NDI) for 1989 to 1995. They found that Asian Americans have a 30
percent mortality advantage over Whites, controlling for age and sex. Although in
subsequent models that controlled for nativity, length of US residence, education, income
level, marital status, and employment status showed that Asian Americans still had a
lower mortality risk compared to Whites, the 30 percent advantage was reduced to 20
percent.

Further analysis showed that this mortality advantage was not observed across
three underlying causes of death. For example, the lower mortality risk among Asian
Americans became statistically insignificant after controlling for all demographic and
socioeconomic variables mentioned above for circulatory diseases, cancer, and residual
causes other than “social pathologies.” For social pathologies, which include accidents,

homicides, and suicides, however, the authors found a greater and statistically significant

mortality advantage among Asian Americans. There was a 50 percent lower risk of
mortality from social pathologies among Asian Americans compared to whites. After
controlling for demographic and socioeconomic factors, the mortality advantage
increased slightly to 56 percent. This study also found mortality advantages among the
foreign-bom population, higher educated, and married persons.

Mentioned briefly in this study is the existence of an epidemiologic paradox
among immigrants. The epidemiologic paradox essentially revolves around unexpectedly
low mortality rates among underprivileged immigrants. This is surprising given the
strong causal relationship between socioeconomic status and mortality. Some researchers
attribute this low mortality to the health selectivity in migration while others believe that
return migration leaves elderly persons statistically “immortal” (Abraido-Lanza et al.
1999). The healthy migrant effect means that the difficulty in the immigration selects
individuals who are not only physically strong but resourceful as well. This is why some
researchers believe that despite the low socioeconomic condition, immigrants have lower
mortality. The “salmon bias” on the other hand draws attention to a possible technical
issue in immigration research. If statistically uncontrolled, the possibility that frail
elderly immigrants return to their home countries to die will lower the death rate of
immigrants. Albeit that most work in this area involved Mexican immigrants, the
relevance extends to Asian immi grants as well.

In sum, nativity, education, and marital status are important determinants in the
mortality risk of Asian Americans. Examination of multiple cause-of—death data is
needed because the mortality advantage of Asian Americans was only observed for

particular causes. Competing explanations offered by previous researchers for the low

mortality rate of immigrants, namely the healthy migrant effect and the salmon bias, need
to be considered for the Asian American population as well.
Demographic and Socioeconomic Differentials in Four Asian American Groups

The main problem of using a broad Asian category in health and mortality
research is that the category captures too numerous and diverse nationalities, ethnicities,
and cultures that inferences to this population becomes somewhat ambiguous. The
diversity found among Asian Americans demonstrates the pressing need to conduct
mortality and morbidity analyses that are disaggregated by detailed Asian categories. On
the other hand, this problem may also compel the researcher to select wisely the
categorization of detailed Asian groups to retain continuity among the categories chosen
and preserve the focus of the study. In this paper, I selected four Asian groups, namely
Japanese, Chinese, Korean, and Vietnamese Americans. I have also included non-
Hispanic Whites for comparative purposes. Although the data were also available for
such large groups as Asian Indians and Filipino Americans, I pooled these groups into the
“Other Asian/Pacific Islander (API)” category to preserve focus and continuity in the
study.

There are several reasons for the selecting Japanese, Chinese, Korean, and
Vietnamese American groups. One reason is the sizes of the respective groups. The four
detailed Asian American groups comprised 56.6 percent of the total Asian and Pacific
Islander (API) population in 1990. The largest among these groups were Chinese
Americans with 23.8 percent, followed by Japanese Americans with 12.3 percent, then
Korean Americans with 11.6 percent, and Vietnamese Americans with 8.9 percent (U .8.

Census Bureau 1993). Another continuing factor among these groups is the similar

cultural roots. All four groups share similar Confucian roots that some studies found
significant in determining crucial elements in health and longevity such as
intergenerational relations and elder care (Ishii-Kuntz 1997). A third continuing factor is
similar settlement patterns. With the exception of New York City, 49 percent of the
Asian American population lives in the West (USCB 2002). Furthermore, in nine
metropolitan areas in California, the proportion of Asian Americans relative to the rest of
the population ranged from 29 to 54 percent.

One study by King and Locke (1987) in particular examined the effect of
residence in metropolitan areas that are heavily populated with Asian Americans, namely
New York City and San Francisco, on the health of Chinese Americans. Using data from
the New York City Health Department and the National Cancer Institute from 1969-71,
the authors observed higher mortality levels in Chinatown areas than in the rest of the
Chinese American population who did not reside in the Chinatowns in New York,
California, and Hawaii. Further analysis showed twofold increases in mortality from
tuberculosis, nephritis, and homicides among Chinese Americans who resided in
Chinatown than their non-Chinatown counterparts. This study suggests that the
environment of the Chinatowns examined was responsible for the detriment to the health
of Chinese Americans. The quality of the physical environment in Chinatown was poor.
Living in these contaminated areas heightened the risk of death from cancer as it relates
to environmental quality and dietary practices. The working environment of Chinatown
is also deprived. Many women who worked in the garment industries were exploited by

their employers. The increased depression and stress from the abuse and long working

hours heightened the suicide risk of female garment workers. The effect of Chinatown
residence is thus important to be measured.

Since this study was published, however, many ethnic suburbs emerged alongside
the rapid increase of the Asian American population. This is most evident in California
where the Asian American population grew not only in the metropolitan areas of Los
Angeles, San Diego, San Jose, and San Francisco, but also in their surrounding suburban
areas. Similar to the suburbanization movement of Whites, this may mean that many
Asian Americans take residence in the suburbs and work in Chinatown. Moreover,
although Chinatown residence has been found to be detrimental to the health of Chinese
Americans and thus emphasizes the need to examine the metropolitan residence of Asian
Americans, recent developments suggest that other non-Chinatown places with high
concentration of Asian Americans should be examined as well. Thus, as most of the
Asian American population resides in the West, it is reasonable to use a West/non-West
dichotomy.l

There are also well-known differences among the four groups. Some of these
differences are directly related to health and mortality such as the nativity composition of
the group, immigration status, immigration history, and demographic/socioeconomic
statuses. These differences influence one another, as we shall see later in the paper. I
will discuss trends and findings from previous research about each group in relation to
the four differences and how they relate to health and mortality. These trends and

findings will partially contribute to expected directions of relationships between various

 

' A separate analysis with a metropolitan/non-metropolitan residence variable proved the dichotomy to be
inefficacious in predicting higher comorbidity. The national level of this study (and thus effectively
including numerous metropolitan areas with low proportions of Asian Americans) is a possible cause for
this finding.

demographic, socioeconomic, and cause-of-death variables and comorbidity at death in
the multivariate analysis.

The first difference is the nativity composition of the four groups. According to
the 1990 Census, the proportions of foreign-bom persons are similar among Vietnamese
(80 percent), Korean (73 percent), and Chinese Americans (69 percent) (USCB 1993).
However, 68 percent of Japanese Americans are native-bom. Many studies examine the
relationship between various generations of Asian immigrants and their health and
mortality (Gordon 1989; Ishii-Kuntz 1997; Le 1997; LeClere et al. 1994; LeClere et al.
1997; Montero 1981). First-generation adult immigrants have been found to be healthier
than second and subsequent generations (Rumbaut 1999). Theorists believe that as
younger generations assimilate to the broader American culture, they lose some of the
healthy behaviors promoted by their native cultures and acquire unhealthy habits
promoted by the host culture, particularly dietary practices (Fujimoto et al. 1996; King
and Locke 1987; Rumbaut 1999). Other researchers find such groups as Japanese
Americans that assimilate faster and in greater degrees had higher income levels and
educational attainment (Montero 1980). Even children of recent immigrants who
assimilate faster than their parents indicate high educational achievements (Caplan et al.
1992). These scholars thus see assimilation as beneficial. Yet a third perspective argues
that there are segmented patterns of assimilation among recent immigrant groups, thus
resulting in positive or negative outcomes depending on what part of the American
culture immigrants assimilate into (Zhou 1997).

The second difference is found in the immigration history of these groups.

Chinese and Japanese Americans have had a long migration history dating back to the

mid-to-late 18003, while immigration flows from South Korea and Vietnam occurred
more recently after 1965 (Kuo 1998). The long migration history in the Japanese and
Chinese American populations led them to have different experiences in the US.
Japanese Americans, for example, experienced extreme racial discrimination and were
placed in concentration camps during World War 11. These experiences may have
adversely affected the health of Japanese Americans today. Chinese Americans have also
experienced extreme racial discrimination during the early part of their migration history.

The lives of early waves of Chinese immigrants were affected by the unbalanced
sex structure of this group. Immigration policies made it extremely difficult for Chinese
women to join their husbands in the US (Ishii-Kuntz 1997). Many of these men remained
in the US and stayed single for the rest of their lives. The disruption of Chinese families
caused by this migration history could also have adversely affected the health of Chinese
American men (Ishii-Kuntz 1997). More recent waves of Chinese immigration occurred
after the 1965 Immigration and Naturalization Act (Kuo 1998).

The other two groups, Korean and Vietnamese Americans, had shorter migration
histories. Korean Americans tend to bring their spouses and elderly parents into the US
as part of the 1965 Immigration and Naturalization Act. As the adult children work,
elderly Korean American parents tend to stay at home. Many of these persons have
language barriers and thus feel isolated from the rest of the community. Researchers
found that mental health, especially depression, is especially a concern among this group
(Kim and Grant 1997). Vietnamese immigrants began to enter the US in the late 19703.
This group is the only one among the four that primarily immigrated due to political

turmoil in their home country.

Another result of these unique immigration experiences is a similar demographic
composition in each immi grant group. The age structure of the Asian American
population, for example, tends to be younger. Because age increase has been found to be
concurrently related to higher mortality in the general population, the Asian American
population may exhibit lower mortality. The sex ratio in these four groups favors men to
women whereas the reversed trend is found in rest of the population. In relation to
mortality, the general population tends to have a higher male mortality than female. In
addition, Johnson and Christenson (1998) found that men tend to die with more comorbid
conditions than women. Although there is an absence of previous research in this area
for Asian Americans, the pattern found by mortality researchers on the general
population may still be applicable to these four groups.

The third difference among the four groups is experience with refugee status. As
mentioned earlier, Vietnamese Americans differ from the other three groups because they
were mainly involuntary migrants. At first the wealthiest Vietnamese fled their home
country to the United States. A few years later followed those with lower socioeconomic
statuses. This second group of immigrants experienced greater trauma in their home
countries. Some of them even spent time in refugee camps prior to entering the US. The
journey into the US was difficult. Coming often by boat, many of the second wave
immigrants from Vietnam became ill or even died along the way (Le 1997). Starting in
1985, a third wave of Vietnamese immigrants, consisting mainly of immediate family
members of Vietnamese Americans, entered the US. Some researchers hypothesize that
the socioeconomic and political conditions in the sending country along with the

experience of displacement among Vietnamese Americans may have caused stress and

10

other health-related problems (Oliver-Smith and Hansen 1982; Scudder and Colson
1982). Although the rough conditions suffered by survivors in the second wave of
immigrants from Vietnam may also have brought the healthiest ones, the ‘healthy
migrant effect’ does not apply to this population by virtue of its unique experience as a
refugee population.

Observing the trends found in the Japanese, Chinese, Korean, and Vietnamese
communities, the cumulative effect of immigration history and statuses as voluntary
migrants or refugees leads also to socioeconomic differences. In 1990 Japanese
Americans were the oldest group with the median age of 36.3, followed by Chinese
Americans (32.1) and Korean Americans (29.1). Vietnamese Americans are the youngest
with the median age of 25.2 years (USCB 1993). Educational attainment among adult
men in 1990 also differed among these four groups, with 89.9 percent of Japanese
Americans having had high school degrees, followed by Korean Americans (89.1
percent), Chinese Americans (77.2 percent), and Vietnamese Americans (68.5 percent).
Per capita income levels in 1990 show a greater disparity among these groups. Japanese
Americans had the highest with $19,373 and Vietnamese Americans had the lowest with
$9,032. The lower income levels in the Vietnamese American community are reflected
in their poverty rate, which is about 3.5 times the lowest group, Japanese Americans.

Socioeconomic differences are directly related to health and mortality. More
disadvantaged groups like Vietnamese Americans will be more likely to have higher
mortality and morbidity compared to Japanese Americans who have higher
socioeconomic levels. Although the Korean and Chinese Americans have lower SES

than Japanese Americans, the healthy migrant effect may be stronger than the

11

socioeconomic forces that influence mortality. Vietnamese Americans, on the other
hand, have a double disadvantage with low SES and refugee status.

In sum, mortality research on the broader Asian American group exposed
mortality advantages in this group compared to whites. Yet examination of Korean,
Vietnamese, Japanese, and Chinese Americans shows great diversity in the nativity
composition of these populations, immigration statuses and histories, as well as
demographic and socioeconomic characteristics. Because these differences directly relate
to health and mortality, they call researchers to reconsider the findings from the general
Asian American group and use detailed Asian categories wherever possible. General
and national mortality research has greater advantage of generalizability than studies on
specific Asian categories, which oftentimes use regional or local data. Nevertheless, both
levels of studies find that certain demographic factors such as age, sex, region of
residence, and nativity along with socioeconomic factors such as education and marital
status are important determinants of mortality and morbidity (Rogers 1995; Rogers, et a1
1996; Verbrugge and Jette 1994). Some studies also found mortality differentials among
various causes of death (King and Locke 1987; Rogers, et. al 2000). This suggests that
by incorporating a more complex view of mortality, researchers could examine the
variety of ways that sociodemographic differentials affect mortality conditions and risks.
In the next section I will discuss the approach taken by Johnson and Christenson (1998)
in examining the intersection between mortality and morbidity by using the total causes
mentioned in the death certificate.

The Total Cause Approach to Assessing Comorbidity at Death

12

General mortality researchers are moving towards multiple cause-of-death
analyses because there is more to mortality differentials than the variety of underlying
medical cause of mortality. Charles Nam (1990) points out,

“More and more, we have come to realize that the complexity of mortality
analysis is reflected in the fact that death is best viewed not just as a momentary
vital event in time but also, and more meaningfully, as the end of a life-long
process during which the individual has experienced some brief and some
prolonged episodes of illness and disability which have combined with the
person’s genetic and physiological stock to alter the risk of survival” (p. 334).

One approach suggested by Nam and developed by Johnson and Christenson (1998) is to
examine the magnitude of the total causes of deaths mentioned in the death certificate
(TC) in addition to the underlying cause of death (UC). The National Center for Health
Statistics (NCHS) (1997) defines UC as “(a) the disease or injury which initiated the train
of events leading directly to death, or (b) the circumstances of the accident or violence
which produced the fatal injury” (p. 3). UC is thus the primary cause of death while TC,
according to Nam, captures both the diagnoses it triggers and “other significant
conditions contributing to the death but not directly related to the [underlying cause]” (p.
334). Thus, more TCs mentioned imply higher degrees of comorbidity at the time of
death.

Johnson and Christenson (1998) used 1989-91 death certificate data from
Michigan to examine whether sociodemographic differentials in comorbidity at the time
of death were real or artifactual. Utilizing an epidemiologic transition framework

(Omran 1971), the authors found support for the hypothesis that age was more

significantly related to variations in mean TC than other demographic factors such as

13

race, sex, and education. They also found that various UCs have varying means of TC.2
This is also consistent with the epidemiologic transition framework in that there will be
greater comorbidities in chronic UCs than acute ones. For example, the authors found
that persons dying from such diseases as diabetes and hypertension had higher mean TC
than persons dying from social pathologies such as accidents, suicides, and homicides.
Thus, older persons dying from chronic diseases have more comorbidity than others. In
regards to race, Blacks died with more comorbid conditions than Whites. For
unautopsied deaths, the difference is small for chronic diseases and larger (0.4 more cases
for black females) for social pathologies.

The epidemiological transition framework may not apply well to a group with
high proportion of immigrants such as Chinese, Korean, and Vietnamese Americans.
Because members Of these groups originated from countries currently on varying earlier
stages of the epidemiologic transition, the immigration process disrupted the smooth
epidemiologic transitions that they were supposed to experience had they remained in
their home countries. Thus, the ordering of comorbid conditions from low to high
according to the stages of epidemiologic transition may not be found in these populations.

One cautionary comment offered by Nam (1990) in regards to this approach of
measuring comorbidity is the reliability of the TCs mentioned in the death certificate.
Studies conducted by the National Center for Health Statistics (NCHS) found systematic
variation among the recording of multiple causes of deaths by the certifier (Nam 1990).
There may be more conditions discovered when autopsies are performed, for example,

than when no autopsies are performed. The performance of autopsies thus increases

 

2 Underlying medical causes and Total causes were coded according to the fifteen leading causes of deaths
using the Ninth International Classification of Diseases (ICD-9).

l4

greater accuracy in the count of TCs. Furthermore, the type of certifier also affects the
accuracy of TCs. Johnson and Christenson (1998) pointed out that autopsies performed
by private physicians for “nonlegal reasons (i.e., the scientific interest of the case or the
request by other family members) are expected to yield the most thorough counts of all
medical conditions that contributed to the death of an individual, as well as confirmation
of the UC” (p. 264). Thus, Johnson and Christenson controlled for autopsy performance
and the type of certifier and found a 1.35 difference in the mean TC performed by
Medical Examiners and private physicians, with the latter group reporting higher mean
TC than the earlier group. This strengthens the study’s claim that sociodemographic
differentials in comorbidity at death are real and not an artifact.

The national mortality data set, unfortunately, does not contain an item on autopsy
status because it does not appear on the death certificate of every state. However, there
are two procedures undertaken by the NCHS to ensure maximum accuracy of the national
Multiple Cause of Death (MCD) file of these death certificates. First, let us look back to
1968 when the computerization of MCD data began. The NCHS instituted the use of a
computer program called Automated Classification of Medical Entities (ACME). The
data entry personnel first inputs all the causes of deaths listed on the death certificate into
ACME, which then applies World Health Organization (WHO) rules in selecting one
underlying cause of death. The system then creates two separate variables to store the
rest of the causes mentioned in the death certificate: record and entity axes. The entity
axis variable preserves all causes listed by the physician for multiple cause analyses.
However, Charles Name (1990) points out that the record axis variable “involves an ‘axis

translation’ of the original cause data to create a set of codes, which are free of

15

contradictions among causes reported, devoid of repetitive codes, and the most precise
within the ICD-9 classification structure and medical info on the recor ” (p. 334).

The ACME software was gradually replaced by fully automated MCD coding
software called Medical Indexing, Class, and Race (MICAR). MICAR eliminated
potential data entry errors by fully automating WHO rules for the selection of UC from
the array of causes mentioned. In 1993, yet another software was implemented which
improves upon MICAR’S coding ability by allowing the actual written responses of the
examiner to be entered into the computer. This software was appropriately named
SuperMICAR. By 1995, twelve states used SuperMICAR, which processed about 75
percent of the death certificates in these states. The rest of the death certificates (74
percent) were coded using MICAR. Death certificates from 1997 were processed with
SuperMICAR and MICAR at the same rates.

The second procedure taken by the NCHS to ensure maximum accuracy in the
coding of MCD data involves continuous monthly sampling of eighty death certificates
from states sending mortality records in electronic form (U .S. DHHS 1997). NCHS staff
then code the original certificates in a separate file and compare them to the data obtained
from the states. Depending on the discrepancies found between NCHS and state-coded
data, the NCHS makes ongoing decisions for the need of corrective actions. For medical
items on death certificates sent to the NCHS via non-electronic methods, NCHS staff
recodes a 1 percent sample of the death certificates and check the results against the state-
coded data.

RESEARCH QUESTIONS AND HYPOTHESES

l6

Previous research found that socioeconomic factors influence health and mortality
regardless of race, ethnicity, or nativity. Also, it is important to investigate the state of
health at death because death, especially at older ages, results from an accumulation of
diseases and disabilities. This area of research has not been extended to Asian
Americans. The rapidly growing Asian American population is unique in that a large
proportion is immigrants with health-related needs specific to this status. Thus, it is
important to explore the following questions:

1. How do sociodemographic factors influence comorbidity at death in Chinese,

Japanese, Korean, and Vietnamese populations?

2. How do these influences compare with Whites?
With the latest available mortality dataset and appropriate statistical methods, this study
seeks to answer those questions.

Drawing on theories developed for mortality and health research on the general
population as well as Asian Americans, the following hypotheses are formulated:

1. There will be small sociodemographic differentials in the average level of
comorbidity, except for age and underlying cause of death (UC).
2. There will be different risks for having higher comorbidity for each
sociodemographic factor.
a. Older ages will increase one’s risks for high comorbidity.
b. Women will have decreased risks for high comorbidity.
c. High educational level will decrease risks for high comorbidity.
(1. Marriage will decrease risks for high comorbidity.

e. Foreign-bom persons will have decreased risks for high comorbidity.

17

f. Residence in the West will increase risks for high comorbidity.
g. Korean, Chinese, and Japanese Americans will have decreased risks for
high comorbidity.
h. Vietnamese Americans will have increased risks for high comorbidity.
i. Some UCs will increase the risk for high comorbidity while others will
decrease this risk.
3. Sociodemographic factors will also operate differently within each group due to
the uniqueness of each group resulting in varying risks for higher comorbidity.3
METHODOLOGY
Data Source
The data used in this study were taken from the Multiple Cause of Death file in
1997, a compilation of all death certificates issued in the United States within that year.
All states in the United States submit these records to the National Center for Health
Statistics, which then creates the MCD file and disseminates the data through the
Interuniversity Consortium for Political and Social Research (ICPSR). Because this
compilation consists of common elements in the US Standard Death Certificate, the
socioeconomic variables included are quite limited. Important information such as
performance of autopsies, which serves as a possible control for the accuracy of the total
mentioned causes of death (Johnson and Christenson 1998), may be required by some
states and not others and thus is not included in the national dataset.
When data are compiled from the whole population, the number of cases becomes

extremely large and several problems may arise. One problem often found in population

 

3 This hypothesis is exploratory, given that previous research is absent. The results will open doors for
future researchers to investigate deeper within each ethnic group.

18

datasets is missing data. Many times researchers resort to imputation to avoid problems
with missing data. The variables used in this study, however, are generally complete
with minimal imputations. Nevertheless, multivariate combinations of missing data
could occur and thus increase the number of cases excluded in statistical analyses.
Therefore, “unknown” categories in several socioeconomic variables are included to
minimize missing data. The issue of missing data along with other exclusions made in
this study will be discussed in detail as appropriate for the respective variables.

Another problem usually found in studies addressing particular ethnic minority
groups is the small sample size. Although this study included the entire population, the
number of deaths found in the Asian American population are miniscule compared to
other ethnic groups. For example, the fewest adult deaths is found in the Vietnamese
population with 1,308 in 1997. Nevertheless, this magnitude proves sufficient for the
analyses performed in this study, even with the use of multiple-category independent
variables such as underlying cause of death.

Dependent Variable

The dependent variable is the total number of causes mentioned in the death
certificate (TC). It does not discriminate between specific values of the causes of death
mentioned. The values ranged from 1 to 14 causes, with a total average of 2.69 TCs (see
Table One). This variable was not normally distributed, with 75.2 percent of the cases
having 1 to 3 TCs and the rest having 4 or more. The assumption that TC was normally
distributed around the mean is violated and thus the properties of the central limit
theorem are not valid (Agresti and Finlay 1997). Because of this, the multivariate

analysis will use a dichotomized version of TC chosen at the natural break in the

19

distribution, namely 4 TCs. This departs from Johnson and Christensen’s study (1998)
which used a continuous measure of TC.
Independent Variables

The independent variables are divided into three types: demographic,
socioeconomic, and cause of death. The first type of variables is demographic. They
include sex, age, nativity, region of residence, and the test variable race. The data on sex,
recorded as male and female, contained no missing values and the two groups are
proportionate to each other. There is a higher proportion of men to women in the four
Asian groups, while the reversed is true for Whites (see Table 2).

Missing data are also not found in the age variable. Nevertheless, following the
convention of Johnson and Christenson (1998), this study is restricted to adults 25 years
and older. Korean and Vietnamese decedents are on average younger by 6 to 10 years
than the others.

Although the dataset recorded nativity by US State in addition to several countries
such as Mexico and Canada, this variable has been recoded into a trichotomy of native-
bom, foreign-Dom, and unknown nativity. Although the problem of missing data is not
serious (ranging from 0.1 to 0.5 percent missing across the various racial categories),
unknown nativity is included as a separate category to maximize the number of cases
included in the analysis. Most Chinese, Korean, and Vietnamese Americans are foreign-
bom while most Japanese Americans and Whites are native-bom.

There are several issues involved with the definition and selection of racial and
ethnic groups of interest. One is missing values, which was slightly problematic in the

race variable. Although the numbers of missing values themselves are not large, NCHS

20

imputed the racial categories of 1,377 Whites, 34 Chinese Americans, 2 Japanese and
Korean Americans respectively, 1 Vietnamese American, and 20 Other Asian and Pacific
Islanders. The problematic aspect of the imputation of these values arises because the
NCHS provided no information on its accuracy. As race is an important focus in this
study, the aforementioned cases with imputed racial categories were excluded from the
study. Another issue is the selection of persons with certain racial backgrounds.
Consistent with its focus, this study excluded persons with racial backgrounds other than
Asian American such as African and Native American. Nevertheless, Whites, which
comprise over 80 percent of all deaths, were included to serve as the comparative group.
Persons with Hispanic origin were also excluded from this study. The death certificate
reports Hispanic origin separately from race because persons with Hispanic origin could
technically be of any race. Because they lie at the intersection of race, ethnicity, and
even nationality, there may be cultural and social complexities involved that are beyond
the scope of this study. Thus, the exclusion of persons with Hispanic origin focuses the
study to a more homogenous population.

Finally, as mentioned earlier, the Asian American category is divided into
Japanese, Chinese, Korean, Vietnamese, and other Asian or Pacific Islander. The final
population sizes are quite uneven: Whites with 1,860,939; Chinese Americans with
6,904; Japanese Americans with 5,005; Korean Americans with 1,796; Vietnamese
Americans with 1,308; and Other API with 11,841.

The second type of independent variables is socioeconomic. The dataset
contained few socioeconomic variables. A measurement of assets, which was not

included in the dataset, is of course the most direct assessment of socioeconomic status.

21

Usually the accumulation of assets occurs over a lifetime and is associated with
occupational status. Although occupational status is an important socioeconomic status
indicator and included in the dataset, its use in the analysis of mortality becomes plagued
with potential problems. The first and most significant problem is that occupational
status is not recorded by all states. Second, many people are already in retirement when
they die at older ages. This decreases the chance of ascertaining the person’s occupation
before they reach retirement. Third, occupational status is not constant over a lifetime.
Some people ascend or descend the occupational ladder and others earn no wages for
their work as homemakers. From these reasons, occupational status was not included as a
predictor variable.

An indirect yet strong assessment of earnings and social status is education.
Education’s strength in predicting socioeconomic status is that it is tied to other
socioeconomic variables. For example, many occupations require certain skills attained
through higher education, thus without education higher occupational status may be
unattainable. Secondly, educational levels tend to stay constant over a lifetime for adults
aged 25 and older. After the initial compulsory education and perhaps postsecondary
education, most people never return to obtain a higher degree. Thus, education is
arguably the strongest and most stable indicator of a person’s socioeconomic status. In
this study, education is coded as primary, secondary, postsecondary, and unknown.
Persons with unknown educational levels are included in the analysis to minimize
missing data (total missing is 7.29 percent). The average education level of each group is

similar, ranging from 10.4 to 11.6 years.

22

A social support variable included in the study is marital status. Studies have
found marital status important in assessing one’s health and longevity (Rogers et. al.
2000). Marriage, in particular, is associated with better health and longer lives among
Americans (Rogers et. al. 2000). To capture the benefits of marriage as well as the
complexity of marital status, this variable was coded into never-married single, married,
widowed, divorced, and unknown. Missing data are not problematic for this variable, as
the total percentage missing is 0.28 percent. Nevertheless, it has been included as a
separate category to enhance the analysis. Across all groups, the proportion married is
highest followed by proportion widowed.

The third type of independent variable is underlying cause of death (UC).
Following the convention of Johnson and Christenson (1998), UC is divided into sixteen
categories using the ninth revision of the International Classification of Diseases.
Although this categorization indicated the leading causes of death in 1990, it is still
highly significant in 1997. The categories are: diabetes, hypertension, septicemia,
chronic obstructive pulmonary disease, nephritic diseases, pneumonia and influenza,
heart disease, cerebrovascular disease, chronic liver disease and cirrhosis, atherosclerosis,
HIV, malignant neoplasm, accident, suicide, homicide, and residual causes. There were
no missing data in this variable.

Statistical Analysis

The methodology chosen for this study is somewhat unusual, as it only analyzes
only the numerator of the more conventional mortality rate. Because of the detailed
study population and combination of independent variables, it was impossible to compute

mortality rates. Although the numerator exists, the Census Bureau does not publish

23

denominators containing counts of the living population to such a fine detail.
Nevertheless, a numerator analysis answers an entirely different and important question
Of the quality of near-death life. Furthermore, it is also a largely unexplored topic in the
demography literature.

The analysis is divided into three sections. First, a means analysis was conducted
to test the first hypothesis that there will be small sociodemographic differentials in the
average level of comorbidity, except for age and underlying cause of death (UC).
Second, two logistic regression models were constructed to test the second hypothesis
that there will be different risks for having higher comorbidity for each sociodemographic
factor. The first model incorporates demographic, social support, and socioeconomic
variables. The second model adds UC to model one. Third, model two of the previous
analysis is separately reanalyzed for Whites, Chinese and Japanese Americans, and Other
API.

High comorbidity is defined as 4 or more TCs on the decedent’s death certificate
and low as having 1 to 3 TCs. These multivariate analyses will determine the odds of
having higher comorbidity in comparison to the reference group within each independent
variable while controlling for the other independent variables. The value 1.0 indicates
equal odds as the reference group for having high comorbidity. Odds lower than 1.0
indicates less risk for high comorbidity. Conversely, odds higher than 1.0 indicates more
risk for high comorbidity.

Tests for statistical significance are not strictly applicable, as this study contains
the universe of the selected Asian American decedent groups. A study that samples a

section of the universe needs to rely on tests of statistical significance to determine

24

whether or not its findings are due to chance. The present study, however, does not need
to rely on such tests because the study population contains a universe which, unlike a
sample, does not have a sampling error. Thus, all observed findings from this study
reflect the true characteristics of the universe, regardless of the results of the tests of
statistical significance. However, tests of statistical significance will be presented with
the sole purpose to indicate which independent variables are more salient than others in
predicting higher comorbidity.

FINDINGS
Means Analysis

Table 3 presents the observed mean TC for the independent variables by race.
The total mean of cause-of-death items in the death certificate is 2.69. This is slightly
higher than the total TC mean observed by Johnson and Christenson (1998). Because the
increase is small (0.21 cause) it could have been a result of better recording of TCs or the
effect of the national level of the analysis which included states with lower overall SES
and poorer health care infrastructure than Michigan.

There is a greater variability of mean TCs across all racial groups. Whites
exhibited the lowest (2.68). The mean TCs for Chinese, Vietnamese, and Korean
Americans were higher than whites. For Vietnamese Americans, this was expected given
the immigration status of this group. Korean and Chinese Americans, however, were
expected to have lower TCs because they would have canied the healthy migrant effect.
Instead, the data reflect otherwise. The highest mean TC is observed in the Japanese
population, with 3 mean TCs per decedent. This indicates greater observed morbidity at

the time of death in this population. This is contrary to expectation because Japanese

25

Americans have the highest socioeconomic statuses compared to the rest of the groups,
even to whites.

The mean TC for both sexes is very similar. It was expected that the difference
be smaller than for age and UC. Region produced a significant mean TC differential
among other independent variables. Those who reside in the West tend to have higher
mean TCs than those who do not. For Asian Americans, this was expected given the
findings of previous research that mortality rates among residents of ethnic communities
are higher than residents of non-ethnic communities. Whites who reside in the West are
likely to be dispersed and not as concentrated as Asian Americans. Thus, the mean TC
differential for Whites in terms of residence in the West is only half to a quarter the
differential of other groups. The nativity differential produced the expected foreign-bom
advantage over native-bom, indicating a healthy migrant effect. Nevertheless, the
differential is miniscule and warrants further analysis.

The differences in marital status mean TC are very small, reflecting slight
disadvantage for widowed persons over all other categories. Albeit modest, widowed
Vietnamese Americans exhibited the second largest differential in mean TC to Japanese
Americans with “unknown” marital statuses. Never-married singles had the lowest mean
TC of 2.7 while widowed decedents had a mean of 3.1 TCs. Education level produced a
greater differential in mean TC. There is a clear gradient in the mean TC of all the
groups with advantage for decedents with postsecondary education.

UC produced the greatest differentials in mean TC. Some diseases such as
diabetes are clearly associated with higher mean TC. This is consistent with the findings

in Johnson and Christenson (1998). However, mean TCs for acute diseases and social

26

 

pathologies were not as low as the authors found. Homicides, for example had the lowest
mean TC in their study while in this study persons who died from honricides had one of
the highest mean TCs. Another difference is the range of mean TCs for the various UCs.
The mean TCs observed in this study among whites ranged from 2.18 to 3.46. The range
observed by Johnson and Christensen (1998) is much greater, starting from 1.1 to 3.6.

In sum, examination of observed mean TCs shows only slight differentials among
all categories of the independent variables. Although the average TC is 2.69, since TCs
ranged from 1 to 14, there are some individuals who died with extremely comorbid
conditions. This leads the analysis away from the examination of means to ask a more
complicated question, what are the determinants of comorbidity at the time of death? For
this, I conducted logistic regressions predicting the odds of dying with 4 or more TCs,
indicating a morbidity condition higher than the average deceased population.
Multivariate Analyses

Table 4 shows the odds ratios from the logistic regression predicting the odds of
having 4 or more TCs in the death certificate. Model 1 controlled for all the demographic
and socioeconomic variables while Model 2 added UC to the first model. Age was found
to be statistically significant and had the effect of increasing the odds of higher
comorbidity in older ages. This is consistent with the findings of Johnson and
Christenson (1998) and the hypothesized direction. However, the influence of age on
greater comorbidity in this study was very small, only 0.8 percent. Women had a slight
advantage over men. Their odds of having greater comorbidity is 5 percent less than
men. This advantage was slightly reduced with the introduction of UCs. Albeit small,

this finding is consistent with the hypothesized direction. Foreign-bom persons also had

27

an advantage over native-bom persons. The odds of dying with greater comorbidity was
9 percent less than native—bom persons. The healthy migrant hypothesis cannot be
rejected by this finding.

Decedents from the West had a 24 percent increased risk for higher comorbidity.
This partly confirms the hypothesized relationship. Because race is controlled, the
residential concentration in ethnic communities associated with Asian Americans should
have been accounted for. This means that residing in the West is detrimental to one’s
health net of the independent variables. It would be interesting to examine whether the
advantage is equal among all racial groups.

Out of the various race categories, only Vietnamese Americans and other API
were statistically significant. Vietnamese Americans had 36.1 percent higher odds of
greater comorbidity at death than whites. This group was expected to have higher odds
due to the experience of displaced involuntary immigration.

Korean American decedents had 21.5 percent higher odds of dying from 4 or
more TCs than whites. This was unexpected because Korean Americans have higher
SES and as a group is largely composed of labor migrants. The healthy migrant effect
did not operate for this group. Previous studies documented the high number of elderly
Korean parents who joined their children through the Family Reunification Act. As
mentioned earlier, many studies also found high incidence of depression and loneliness
among these elderly parents due to isolation and inability to assimilate to the broader
society because of communication difficulties. Perhaps these factors allow the
accumulation of diseases among elderly Korean Americans to go unnoticed and result in

elevated risks for high comorbidity.

28

The third highest group is Japanese Americans with 20.9 percent increased odds
for high comorbidity compared to whites. Although Japanese Americans are not an
immigrant group (and thus the foreign-bom advantage does not apply to them), this
finding is still unexpected. This is because Japanese Americans have higher SES even
among whites and so at least their risk should have been the same as whites. Perhaps
high SES in the Japanese American population operates against them. As Johnson and
Christenson (1998) noted, lower quality health care could have the unintended effect of
lower TCs recorded in the death certificate. This is because the physician responsible
may not be familiar with the medical history of the decedent. Thus he or she may
overlook certain conditions that are not apparent. Conversely, decedents with high SES
could afford better quality health care. The unintended consequence of this, on the other
hand, is a greater number of TCs recorded by the physician familiar with the patient’s
entire medical history.

Chinese Americans had an 18 percent greater risk than whites from dying with 4
or more comorbid conditions. This finding is also unexpected given the healthy migrant
selectivity that should have protected them from these comorbid conditions. Previous
research reported several findings that could partly explain this unexpected finding. One
is the Chinatown settlement of this population. Although the analysis controlled for
Western residence, perhaps in the case of Chinese Americans this is not specific enough.
A separate analysis with only Chinese Americans would clarify this matter. Another
explanation is the medicinal practices among elderly Chinese Americans. The use of
traditional medicines that may alleviate but not cure diseases may allow the diseases to

progress to a more advance stage. Furthermore, the reliance on traditional Chinese

29

medicines may leave the hospital as a final resort. Thus, diseases may have accumulated
and progressed before being treated by hospital physicians. A third explanation supports
and has the same affect as the second one, namely the reported underuse of health care
facilities by elderly Asian Americans. Further research needs to be done to identify the
mechanisms that leave Chinese Americans at a greater risk for high comorbidity.

Education level shows a gradient with the advantage for hi gher-educated
decedents. The odds for dying with 4 or more TCs for primary-school educated
decedents were 17.8 percent higher than post-secondary educated decedents. This
gradient confirms the hypothesis that higher education decreases the risk for high
comorbidity.

Marital status showed advantage for currently married decedents. While the
actual observed means showed advantage for divorced decedents in some racial groups,
controlling for other demographic variables and education gave the advantage for married
decedents, as expected. Never-manied singles had the highest odds of dying with 4 or
more TCs than the other groups compared to manied decedents. It is important to note
that the marital advantage here is very small. Furthermore, after the addition of UCs, the
marital advantage disappeared altogether. This suggests that the marital advantage is
associated with lowered risk for certain diseases. Perhaps diseases that are related to
unhealthy lifestyles such as chronic liver disease, diabetes, and heart disease are the most
positively affected by marriage. Further research needs to be done to identify the specific
diseases that marriage alleviates the most.

In Model 2, the fifteen leading causes of deaths were compared to the residual

causes category. Almost all of them were statistically significant, with the exception of

30

hypertension and homicide. Another pattern is that almost all UCs are associated with
lower odds than the residual causes except for diabetes, chronic liver disease and
cirrhosis, and accidents. Persons dying from diabetes had a 72.2 percent higher odds of
having 4 or more TCs at the time of death than persons dying from residual causes. The
odds for those with chronic liver disease and cirrhosis was only slightly higher, with 6.8
percent. The odds for decedents from accidents was 62.8 percent higher as compared to
those dying from residual causes. The lowest Odds for greater morbidity is found among
decedents from cancer. Compared to those dying from residual causes, decedents from
cancer have a 57.2 percent lower odds from dying with 4 or more TCs.

These findings did not show any particular patterns, especially one that relates to
the epidemiologic transition framework. As the study population experienced differing
stages of the transition, perhaps an earlier stage in their home country and a more
advanced stage in the U. 8., it may not apply to them particularly. Another possible
explanation is the technological advances in emergency and end-of-life care. These
advances may save or allow persons dying from accidental causes to live a little longer
than a decade ago. During this short duration, the person dying from the ultimately fatal
accident could develop various complications. Since he or she is most likely under close
medical attention, all complications are also likely to be detected by the physician. Thus,
more TCs would be recorded in the death certificate. Data on the length of last
hospitalization prior to death may help researchers unravel this issue.

In sum, many of the sociodemographic variables only slightly influence one’s
odds of dying with 4 or more TCs. Race, however, was an exception. The odds that

Japanese Americans would die from at least four medical causes were 20.9 percent higher

31

than for Whites; Korean Americans, 21.5 percent; Vietnamese Americans, 36.1 percent.
Although one would expect higher odds among Vietnamese American decedents,
Japanese Americans also had unexpectedly higher odds of dying from 4 or more TCs.
Furthermore, these odds were increased when UC was added to the model. Race was
almost as strong in predicting this negative condition as UC. This suggests that
comorbidity is more closely related to causes of deaths than other socioeconomic and
demographic variables. Furthermore, there are different relationships between causes of
deaths and comorbidity. Some causes of deaths are associated with lower odds of dying
from 4 or more TCs while other had greatly higher odds than the residual category.

The interpretation of some of the findings in the previously discussed analyses
could benefit from further analysis into each racial group. Table 5 displays the results of
logistic regressions on the odds for having 4 or more TC on the death certificate for
Whites, Chinese and Japanese Americans, and Other API.4 Although the purpose of this
final analysis was to clarify the findings of the previous analyses, the earlier produced
many intriguing results whose interpretations are beyond the scope of this study and may
instead be a task for future researchers. This analysis is intended purely for exploratory
purposes.

The analysis of region of residence shows that Chinese and Japanese Americans
benefit more than Whites from not residing in the Western part of the United States. This
would support the explanation given earlier that the residential concentration of these
Asian American groups in the West increases their risk for dying with higher

comorbidity. Perhaps residence in the non-West allows for positive assimilation of Asian

 

4 Separate regressions for Vietnamese and Korean Americans were also conducted. Due to the small
respective population sizes of these groups, some of the cells became empty. Thus, the models did not
converge and the results are highly unstable.

32

Americans into the broader culture. This explanation is also supported in the loss of the
foreign-born advantage in the Chinese American population net of the independent
variables. For the Chinese American group then, assimilation is more beneficial.
Nevertheless, future researchers should use more precise categories of Chinatown and
other areas with high Asian American concentration. Furthermore, these findings do not
explain why the non-West advantage for Whites remained. Perhaps the reverse is true for
Whites, that residential assimilation heightens their comorbidity risk.

Other findings for the Chinese American population present another puzzling
picture. While UC offsets the marriage advantage for Whites and Japanese Americans,
the risk for high comorbidity among widowed Chinese Americans remained slightly
higher. In opposition to this, divorced Chinese Americans have a 28 percent decreased
risk for high comorbidity. Although divorced Whites and Japanese Americans also have
lowered risk for high comorbidity after controlling for UC. the risk exhibited in the
Chinese American population is 13 to 18 percent lower than the two others. A task for
future researchers is to examine divorce in the Chinese American culture to uncover
whether men and women derive the same health benefits from divorce and why divorce
decreases high comorbidity risks in this population more than others.

SUMMARY AND CONCLUSION

The results of this study revealed differentials in sociodemographic
characteristics, means of total causes, and determinants of comorbidity within and
between Korean, Chinese, Vietnamese, and Japanese Americans. These observed
differentials truly reflect the characteristics of the population, as statistical analyses were

conducted on all US. deaths in 1997. Multivariate analysis revealed that Vietnamese

33

Americans and Other API had the greatest comorbidity at the time of death, after
controlling for various sociodemographic factors, as well as for the underlying cause of
death. The advantage of analyzing the entire universe is reflected in the higher odds of
comorbidity found in the Chinese, Japanese, and Korean Americans groups compared to
whites. Although these findings did not have p<0.05, they truly existed in the population
and were not due to chance. This means that there are other unobserved factors within
each group that increase their odds of dying with a higher level of comorbidity than
whites.

The findings regarding Vietnamese Americans support previous research on the
negative impact of political trauma and low socioeconomic status for this refugee-
immigrant group (Gordon 1989; Le 1997).

The findings on Japanese Americans were more puzzling. Discrimination during
the World War II era that exclusively targeted Japanese Americans likely had negative
mental and physical health consequences (Jensen 1999). More research needs to be
conducted to measure the magnitude of impact of these experiences on morbidity,
especially at the older ages. The findings of this study also invite research on the
intersection of morbidity and mortality for other Asian American groups such as Asian
Indians and Filipino Americans. These groups have different religious backgrounds—
Indians are mainly Hindu and Filipinos mainly Catholic-~than the four exarrrined in this
paper. They also have different migration statuses and histories. The results of research
on these populations will bring more insight on the diversity within the broad label,

“Asian American.”

34

There are several advantages to this study. First, it used the latest available US
mortality data from 1997, thus improving and expanding the scope of current mortality
and health literature for detailed Asian American groups. Second, this study improved
upon general Asian American studies by using four detailed Asian categories, namely
Japanese, Chinese, Korean, and Vietnamese Americans. Third, it used a unique approach
to the examination of comorbidity at the time of death by using the magnitude of total
causes mentioned in the death certificate, as developed by Johnson and Christensen
(1998). This approach has not been explored by other Asian American health and
mortality researchers.

There were also several limitations to the study. First is the accuracy of TC. In
spite of technological advancements in the recording of multiple cause of death data,
accuracy of the total causes mentioned may still be in question. Johnson and Christenson
(1998) overcame this issue by controlling for autopsy status and the type of physician
performing the autopsy. Nevertheless, these variables were unavailable in the national
mortality dataset. Because autopsy status is recommended on the US Standard Death
Certificate, I call for its inclusion in the national mortality dataset. Until this becomes
available, future research needs to incorporate an indirect measure for autopsy status to
ensure maximum accuracy of TC or conduct their analyses at the state level.

The second shortcoming to this study is the limited number of variables available
in the mortality dataset. Since the populations of interest were largely immigrant groups,
variables for immigration history and status would have contributed to the study. To
improve the assessment of comorbidity, variables such as type of health insurance,

medical history, behavioral health risks, and disability status would be important in future

35

 

research. The third limitation to the study is the cross-sectional aspect of the data. I only
used national mortality data from 1997 and in so doing I could not assess trends in
sociodemographic differentials in mortality and comorbidin in Asian American groups.
Future research overcoming these limitations will lead to a better understanding of

differentials in comorbidity at the time of death.

36

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

No.2 :58 do 8:3 mass: SEEM .28: Le 5:8 $5.52 «8.2% :25: new £8: to .38 .m. d 598m
8.2: BEBFIISSF 5a.: 8.8F 8N; 8.8F 85 8.8F mam 8.9: «86 8.8F «$.83 .98
Food NN Sod o 83 0 Bed o Sod o 80.0 o F85 N 3
83 mm Sod o coed o Sod o Sod o 83 o m8.o B NF
93 8F mood F coed o 890 F 036 N 39o F oFo.o 8F NF
wNod 8F. v8.0 F. 88 N 83 o 086 F Sod F mNod 8F. FF
NBd EN; mood m Sod F Bod F 83 m 890 N N56 83 0F
28 FFNN mNNd AN mNNd N ENd m 9.3 F 8N6 4F 88 Fm; a
Red 8N6 8nd 9. 33 0F Ammo 9 83 mm 83 MN 53 $3 a
mFFF m8.FN we: mt 83 8 83 NF 83 8 53 SF FFF.F 298 A
NEN 33m 53 ms. 83 Fe NE 9. SN... NFN 8nd FAN BAN momdm m
8% oooNNF BF.» 8m Nome mFF Sod 3F FNFN Ev Sm.» 9m 8% 33: m
92: FBBN 9.}: NSF REF mFN 8me FAN 33F 8A 552 v8; v8.9 $9va F.
oFmNN F89... oFoeN NEN NFo.mN m8 898 N4 EQNN BF; SEN 3K; 808 84.5. N
8mg BBB NFFfiN @de SEN mmm was one 968 «on; 9.0.8 #3 908 9658 N
BNNN 339. So? BFN FNFGF FFN NSF 3N NNEF NS 9%.? BF NmNNN 34.3 F
RN @3va Neon 33.. NF.FN SF. oawN oFm ANNN mFoF EN 83 SN N359. +4
NNBN AFNdeF 9.8 NNNB 8.8 F8 8.: SN; 2% 89m FN.FA 93. 8.2 3.38; N
2F
0»
as 2 e. z e. z e. 2 a. z e. z e. I z 85%.;
:30... E< .550 mmemgoS Emcee! owmcmmmfi omchO SEB
6.: 8:8 .33 Fe cozaeeen .F 22:.

37

«8:538.

 

.39 Soon .6 826 mi. .32 60.3.55 5.81 to. 3.50 .3932 £8.38 FEES: new £3613 .58 .m. .3 683%

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

omnfiooé Feo. F r mom; ooh; mood .90...
88 NnoKoF moo eNo.F moo Ne oe.e oo om.o o5 :26ch
Rem ooF.ooe Fooo ooo.e whom KN emeo emo 3.3 So; eNoN coo; ofiew ewodoe boocoooflmom
onoe oFN.ooo oo.oo oood eooe moo. No.5 Foo oooe ooe.~ Noon oood mode ontomo Emocooom
mooF oFo_Noo Boom oofio oo.oo moo oodw FFe «mom to; oo.o~ com; 3.? Rodeo beetn.
o.FF 99 e9 o.FF o.FF or? .9: Amen; 5. :mos.

_o>o._ coooguw
oe.o wood oe.o em oo.o F to o 36 x. owo oF oe.o eoo.o 565.5
mos NoFNeF KFo ooo.oF oooo eon; Keo Fon; «.on moo; ooNo oFo.o oeo oeodNF E09568”.
ooFo ooo.ooN.F Fox Foo owo o «to No oodn mood eo.NF ono 8.8 35th €099.sz

3.2sz
owo Boo...“ ooo we go o med o N; o go 9 omo Nomfi 565::
ood mood? moo ewo omo on So FFF oeo oNo ohm ooN 2H neoeeF 82020
oooo Nooeon oNNN ooN.o NFoo eoo oedo Foo oozoo Foo; Noeo FFe.N mode wooden 828:5
{Ne eFo.ooo oooo nooN Noeo BK oeoo eFo.F oeoo oNo.N too ohod Ede 9:2 3:55.
ems KodeF moo oNo oF.o oFF oe.o No and one Noe oeo mos wood: 06%
ooEmEeo>oz

35m .255.
5.5 Foo_ooo oo.Ne onoe oe.oe ooo oo.\.e ooo onoe ooeN Nooe auto oN.Fo Noodoo 295".
oooe NequNo oo.oo ooo_o Nooo ooh NFNo ooo Rom oFo.N ooeo who ehoe neoooo was.

xom
8.? wooden the meme No.NN NoN oo.oo So So oee oo.oo oom.m tnmo 33%; 625.52
Beéo FFoKoo; ewoo oooo oo.R oFo.F Nooo mew; ooFo oom.e Fooo {me 8.: 8.58 $25

552.

NMN oNo woo oNo k .olu oon ooln .28., co oo< :85.

co 2 .x. z o\o z oxo z o\o z o\o z o\o z mm.nm...m>

.90... #2850 omoEmcFoS 5.93. mm comma 035:0 3.55

 

 

 

 

00mm >9 wo_na.._n> «concaMME no cos—5.50.0 .N 03a...

38

 

 

 

 

 

 

 

 

 

 

 

 

NP». ...me .6 8.5.0 0.... .33. .§.m...m..mk...mm... 6135062262 $8.33. :mEo... bow ....mo... .0 ..Qon. .m. .... 62.6%

oon.nB . .eo. . . .moo. . oon.. moo... oo.eoo.: ooo.ooo.. Eon
oo.e. ooo..oo ooo. new. coo. eo. .3. no. 8o. .on oo... oon oo.... ono.nn~ monsooaopoom
om: ooo.o oo.. oo. No... o. moo n. ...: n on: oe no: 8:... $.2on
no... ooo.om 8.. no. .oo oo o...o .o o... om .o. ... .o. ooodm oposo
o..o .oo.oo ooe ooo o..o .o o..o ... 8o 3.. eoo .oo n..o e.o.oo 328,.
«So oeooo... .n.o nood oeoo moo oeoo ooo .n.n~ noo.. oooo ono.o no...o eoo.ne.. .850
ooo ooo.o we: on ooo o no: e 9.: o ooo ... om: 5...... 2:
ono on}: om: eo 3.: o ow: m 9.: oo ooo o. nno moo... $8568.22
5.. o...o. ooo .... oo.o o. oo. no end no 8.: oo .o.. mood. so
.52. 0.59.50
n..n ono.oo. eno om... moo. oo. moo .n. 3:. .om moo noo o..n oon.oo. asageoewo
oooo oon_.oo eooo o.o.o moon eno nnoo ooe nono 8.... So... ooo. .o.oo onmio 3:35:81
moo 8...: ooo .9. no... mo 6... on one ooo ooo ono eoo nnmon 2.
Dcm m_coE:wcn_
oo. oeoo. om. we. no. ... .oo .. No: .o oo. on 8.. onoo. .osooEooz
..No ooo.oo eoo moo o..e eo on.~ oe 8o 3.. moo ono noo eoooo 0:00
8.: ooo.o. 8.: no. E: n ow: o No: ... ooo ... oo.o mono. geoooooo
om: ooo_o moo nn :3 .. on: o. 8.: oo 8.: no on: 3.3 8.29.8.1
o...o oo..oe .n.o ooe ooo no moo .n oo.o om. ooo ..o oeo .oo.oe 8.89:

8350 9.2.8::

o... z .e. z a... z ... z a... z a... 2 on. LI z 853%

.90P E< .550 mmempzw; Cm.w._o¥ wwwcmmm... meCEO muck

 

 

 

 

 

 

 

 

 

 

 

.omoéooo .m 2%..

39

Table 3. Observed Mean TC tor lndepeflent Variables by Race

 

 

 

 

 

 

"Variables White Chinese Japanese Korean Vietnamese Other
Overall 2.68 2.88 3.00 2.85 2.97 2.91
Sex

Female 2.68 2.86 2.98 2.85 3.03 2.86
Male 2.68 2.91 3.02 2.85 2.92 2.93
Education Level
Primary 2.77 2.98 3.24 2.91 3.06 3.05
Secondary 2.68 2.91 2.99 2.89 3.01 2.90
Postsecondary 2.61 2.78 2.83 2.75 2.85 2.81
Unknown 2.69 2.68 2.99 2.87 2.43 2.82
Underlying Cause
Diabetes 3.46 3.73 3.79 3.68 4.00 3.69
Hypertension etc. 2.91 3.10 3.29 2.85 3.27 3.17
Septicemia 2.90 3.27 3.73 2.60 2.29 3.36
Chronic pulmonary 2.88 3.08 3.07 3.29 3.20 3.20
Nephritis etc. 2.87 3.18 2.84 3.82 3.93 3.57
Pneumonia and flu 2.71 3.14 3.22 3.11 3.1 1 3.32
Heart disease 2.79 3.00 3.12 3.05 3.12 3.01
Cerebrovascular 2.60 2.90 3.06 2.98 2.63 2.72
Chronir liver etc. 2.93 3.56 3.43 3.30 3.46 3.10
Atherosclerosis 2.66 3.50 2.83 2.80 3.00 2.88
HIV 2.24 2.71 3.00 2.00 2.40 2.40
Malignant neoplasms 2.18 2.41 2.41 2.37 2.46 2.36
Accident 3.43 3.12 3.33 3.17 3.72 3.05
Suicide 2.62 2.44 2.60 2.33 2.63 2.42
Homicide 2.85 3.02 3.57 3.35 2.50 2.91
Residual causes 2.89 3.31 3.46 3.06 3.51 3.31
Age Group
25-29 2.61 2.63 2.93 2.77 2.55 2.62
30-34 2.59 2.60 2.92 2.69 2.57 2.57
35-39 2.55 2.54 2.56 2.25 2.88 2.40
40-44 2.50 2.50 2.51 2.49 2.68 2.45
45-49 2.46 2.36 2.67 2.52 2.68 2.56
50-54 2.43 2.54 2.70 2.44 2.72 2.64
55-59 2.46 2.52 2.54 2.76 2.66 2.71
60-64 2.51 2.79 2.65 2.76 2.90 2.75
65-69 2.59 2.79 2.93 2.83 3.06 2.98
70-74 2.67 2.89 2.89 2.86 2.94 3.02
75-79 2.74 3.01 3.04 3.17 3.19 3.08
80-84 2.78 3.00 3.1 1 2.97 3.28 3.16
85+ 2.76 3.03 3.23 3.04 3.29 3.14

 

 

 

 

 

 

 

 

 

I ‘
Source: U. S. Dept. of Health and Human Services, National nter for Health Statistics. Multiple
Cause of Death 1997.
(Continued)

4O

(Table 3, Continued)

 

 

 

 

 

White hinese Japanese Korean Vietnamese 55m

Marital Status

Never-Married Single 2.67 2.72 2.89 2.92 2.71 2.75

Married 2.64 2.86 2.97 2.76 2.92 2.88

Widowed 2.74 2.99 3.1 1 2.98 3.12 3.03

Divorced 2.65 2.62 2.89 2.84 3.1 1 2.73

Unknown 2.55 2.31 2.00 2.75 2.33 2.40
Region

West 2.84 3.03 3.05 3.03 3.1 1 3.14

Non-West 2.65 2.64 2.49 2.43 2.48 2.58
Nativity

Native-Dom 2.68 2.90 3.02 3.12 3.00 2.73

Foreign-Dom 2.69 2.88 2.94 2.83 2.97 2.92

Unknown 2.68 2.28 3.29 2.00 1 .00 2.24

 

 

 

 

 

 

 

 

— —
Source: U. 8. Dept. of Health and Human Services, National Center for Health Statistics. Multiple

Cause 01 Death 1997.

41

Table 4. Odds Ratios of the Eflects oi Sociodemographic Factors and

Mllemcal Causes on With Four or More TC

 

 

 

 

 

Model 1 Model 2
Ago 1.008 “ 1.007 ...
Region
West ref. rel.
Non-West 0.758 “1 0.751 ”'
Sex
Male 1.052 ... 1.041 '“
Female rel. rel.
Race
White rei. rel.
Chinese 1 .182 1 .228
Japanese 1 .209 1 .241
Korean 1 .215 1 .245
Vietnamese 1.361 ' 1.390 “
Other API 1.378 ... 1.347 ""
Marital Status
Never-Married Single 1.033 *“ 0.943 “"
Married net. rel.
Widowed 1.019 "4 0.962
Divorced 1.022 "' 0.968 '“
Unknown 0.839 ... 0.749 ...
Nativity
Native-Dom rei. rel.
Foreign-pom 0.912 “' 0.913 '“
Unknown 1.007 0.984
Education Level
Primary 1.178 '“ 1.149 “'
Secondary 1.111 “ 1.099 ‘“
Postsecondary rel. rel.
Unknown 1.124 “' 1.101 “
Underlying Cause
Diabetes 1.722 “'
Hypertension 0.849
Septicemia 0.967 '“
COPD 0.891 ...
Nephn'tis, neph. syndrome, nephrosis 0.904 “"
Pneumonia and influenza 0.705 '“
Heart disease 0.846 '“
Cerebrovascuiar disease 0.671 '“
Chronic liver disease and cirrhosis 1.068 '“
Atherosclerosis 0.654 “"
HIV 0.518 '“
Malignant neoplasms 0.428 ...
Accident 1.628 “'
Suicide 0.468 ""
Homicide 0.796
Residual causes rel.

-2 Log-Likelihood 2,103.608.4 2,059,915]
Chi-Square 9,875.3 “' 54.0450 "‘
Degrees at ireedom 17 32

n 1 £87,793 1 .887,793

 

3'p<0.05;“p<0.01;“p<0.001

Source: U. S. Dept. of Health and Human Services. National Center tor Health Statistics.

Multiple Cause 01 Death 1997.

42

 

Table 5. Odds Ratios of the Effects oi Sociodemographic Factors and Underlying Medical Causes
on Eng with Four or More TC by Race

 

 

 

 

 

 

 

 

W—hite _criinesi_a Japanese Ot——herAPl
Age 1.007 “ 1.00 “ 1.006 ' 1.011 ‘“
Region
West rel. rei. rel. rel.
Non-West 0.756 "1 0.525 '" 0.599 ... 0.521 '“
Sex
Male 1.040 “' 1.132 ' 0.955 1.033 '
Female rei. rel. rel. rel.
Marital Status
Never-Married Single 0.943 “ 0.908 0.879 1.049
Married rat. rel. rei. rel.
Widowed 0.962 '" 1.040 0.919 0.982
Divorced 0.968 '" 0.781 0.913 0.998
Unknown 0.750 “1 0.887 <0.001 ' 0.856
Nativity
Native-bom rel. rel. rei. rel.
Foreign-hem 0.910 “" 1.180 0.892 1.270
Unknown 0.989 0.647 2.71 1 0.597
Education Level
Primary 1.148 '“ 1.023 1.274 0.987
Secondary 1.099 “1 1.038 1.184 1.024
Postsecondary rel. rat. rel. rel.
Unknown 1.102 “ 0.906 1.164 0.949
Underlying Cause
Diabetes 1.725 "t 1.356 1.304 1.452
Hypertension 0.854 0.522 0.321 0.686
Septicemia 0.968 ‘“ 0.549 1.123 ' 1.026 '
COPD 0.894 ... 0.633 0.483 0.705
Nephritis. neph. syndrome, nephrosis 0.903 “* 0.824 0.406 1.083 '
Pneumonia and influenza 0.702 '“ 0.682 0.751 0.881 ‘
Heart disease 0.848 “' 0.657 0.701 0.692
Cerebrovascular disease 0.67 "' 0.541 ‘ 0.524 0.4 '“
Chronic liver disease and cirrhosis 1.068 “1 1.358 ' 1.020 0.913
Atherosclerosis 0.657 “' 0.894 0.236 0.404
HIV 0.519 "4 1.295 0.443 0.390
Malignant neoplasms 0.4 "1 0.334 “1 0.305 “1 0.355
Accident 1.652 “‘ 0.625 0.754 0.638
Suicide 0.474 '“ 0.156 '“ 0.281 ' 0.215 "'
Homicide 0.797 0.625 1.254 0.704
Residual causes rel. rei. rel. rat.
-2 Log-Likelihood 2,027,782.2 7,889.5 6,004.5 13,735.9
Chi-Square 52,3362 “' 399.5 '“ 290.5 “" 833.5 ""
Degrees of ireedom 27 27 27 27
n 1,860,939 6,904 5,005 11,841

 

 

' p < 0.05; “ p < 0.01; “ p < 0.001
‘ Because the cell size is very small (n=6), the estimated odds ratio is not reliable

Source: U. S. Dept. of Health and Human Services,
National Center for Health Statistics. Multiple Cause at Death 1997.

43

REFERENCES

Abraido-Lanza, Ana F., Bruce P. Dorenwend, Daisy S. N g-Mak, and J. Blake Turner.
1999. “The Latino mortality paradox: A test of the ‘Salmon Bias’ and Healthy
Migrant hypothesis.” American Journal of Public Health 89(10): 1543-48.

Agresti, Alan, and Barbara Finlay. 1997. Statistical methods for the social sciences.
Upper Saddle River, NJ: Prentice Hall.

Caplan, Nathan, Marcella H. Choy, and John K. Whitmore. 1992. “Indochinese Refugee
Families and Academic Achievement.” Scientific American (February): 36-42.

Fujimoto, Wilfred, Edward Boyko, Donna Leonetti, Richard Bergstrom, Laura Nowell-
Morris, and Patricia W. Wahl. 1996. “Hypertension in Japanese Americans: The
Seattle J apanese-Americans Community Diabetes Study.” Public Health Reports
1 l 1: 56-8.

Gordon, Linda W. 1989. "The Missing Children: Mortality and Fertility in a Southeast
Asian Refugee Population.” International Migration Review 23: 219-37.
Ishii-Kuntz, Masako. 1997. “Intergenerational Relationships Among Chinese, Japanese,

and Korean Americans.” Family Relations 46: 23-32.

Jensen, G. M. 1999. “System failure: health-care deficiencies in the World War 11
Japanese American detention centers.” Bulletin of the history of medicine 73(4):
602-28.

Johnson, Nan E., and Bruce A. Christenson. 1998. "Socio-demographic Correlates of
Multiple Causes of Death: Real or artifactual?" Population Research and Policy

Review 17: 261-274.

Kim, Y., and Grant D. 1997. “Immigration patterns, social support, and adaptation
among Korean immigrant women and Korean American women.” Cultural
Diversity and Mental Health 3(4): 235-45.

King, Haitung, and Frances B. Locke. 1987. “Health effects of migration: U. S. Chinese
in and outside of Chinatown.” International Migration Review 21(3), Special
issue: Migration and health: 555-76.

Kuo, JoAnn. 1998. “Health Status of Asian Americans: United States, 1992-94.”
Advance Data 298: 1-16.

Le, Quyen Kim. 1997. “Mistreatrnent of Vietnamese Elderly by Their Families in the
United States” Journal of Elder Abuse & Neglect 9: 51-62.

LeClere, Felicia B., Leif Jensen, and Ann E. Biddlecom. 1994. “Health Care Utilization,
Family Contest, and Adaptation Among Immigrants to the United States.”
Journal of Health and Social Behavior 35: 370-84.

LeClere, Felicia 3., Richard G. Rogers, and Kimberley D. Peters. 1997. “Ethnicity and
Mortality in the United States: Individual and Community Correlates.” Social
Forces 76: 169-98.

Manton, Kenneth G. 1980. “Sex and Race Specific Mortality Differentials in Multiple
Cause of Death Data.” The Gerontologist 20: 480-92.

Menchik, Paul L. 1993. “Economic Status as a Determinant of Mortality Among Black
and White Older Men: Does Poverty Kill?” Population Studies 47: 427-36.

Montero, Darrel. 1981. “The Japanese Americans: Changing Patterns of Assimilation

Over Three Generations.” American Sociological Review 46: 829-39.

45

Nam, Charles B. 1990. “Mortality Differential from a Multiple-Cause-of-Death
Perspective.” Pp. 328-342 in Jacques Vallin, Stan D’Souza, and Alberto Palloni
(eds.), Measurement and Analysis of Mortality: New Approaches. Oxford:
Clarendon Press.

National Center for Health Statistics. 1998. Health, United States, 1998: With
Socioeconomic Status and Health Chartbook. Hyattsville, MA.

Oliver-Smith, Anthony and Art Hansen. 1982. “Involuntary Migration and
Resettlement: Causes and Contexts.” pp. 1-9 in Art Hansen and Anthony Oliver-
Srnith (eds.) Involuntary Migration and Resettlement: The Problems and
Responses of Dislocated People. Boulder: Westview.

Omran, Abdel R. 1971. “The Epidemiologic Transition: A Theory of the Epidemiology
of Population Change.” Millbank Memorial Fund Quarterly 49(4): 509-38.

Rogers, Richard G. 1995. “Sociodemographic Characteristics of Long-Lived and
Healthy Individuals.” Population and Development Review 21: 33-57.

Rogers, Richard 6., Robert A. Hummer, Charles B. Nam. 2000. Living and Dying in the
USA: Behavioral, Health, and Social Differentials of Adult Mortality. San Diego:
Academic Press.

Rogers, Richard G., Robert A. Hummer, Charles B. Nam, and Kimberley Peters. 1996.
“Demographic, Socioeconomic, and Behavioral Factors Affecting Ethnic
Mortality by Cause.” Social Forces 74: 1419-38.

Rumbaut, Ruben G. 1999. “Assimilation and its Discontents: Ironies and Paradoxes.” in
Josh Dewind, Charles Hirschman, and Philip Kasinitz (eds.), Becoming American,

America Becoming. New York: Russell Sage Foundation.

46

Scudder, Thayer and Elizabeth Colson. 1982. “From Welfare to Development: A
Conceptual Framework for Analysis of Dislocated People.” pp.267-87 in Art
Hansen and Anthony Oliver-Smith (eds.) Involuntary Migration and
Resettlement: The Problems and Responses of Dislocated People. Boulder:
Westview.

Tabachnick, Barbara G. and Linda S. Fidell. 2001. Using Multivariate Statistics.
Boston: Allyn and Bacon, fourth edition.

US. Census Bureau. 1993. We the Americans: Asians, Series WE-3. Washington, D.
C.: U. S. Government Printing Office.

_. 2002. A Profile of the Nation 's F oreign-Born Population From Asia (2000
Update), Census Brief: Current Population Survey. Washington, D. C.: U. 8.
Government Printing Office.

_. 2002. The Asian Population: 2000, Census 2000 Brief. Washington, D. C.: U.
S. Government Printing Office.

US. Dept. of Health and Human Services, National Center for Health Statistics. Multiple
Cause of Death, 1997 (Computer file). ICPSR version. Hyattsville, MD: US.
Dept. of Health and Human Services, National Center for Health Statistics
(producer), 1999. Ann Arbor, MI: Inter-university Consortium for Political and
Social Research (distributor), 2001.

Vega, William A. 1991. “Ethnic Minorities and Mental Health.” Annual Review of
Sociology 17: 351-83.

Verbrugge, Lois M. and Alan M. Jette. 1994. “The Disablement Process.” Social

Science and Medicine 38: 1-14.

47

Zhou, Min. 1997. "Segmented Assimilation: Issues, Controversies, and Recent Research

on the New Second Generation." International Migration Review 31: 975-1008.

48