University of Southern California Dissertations and Theses

Factors associated with long-term-care facility residence of patients with Alzheimer's disease from Alzheimer's disease research centers of California

(USC Thesis Other)

Factors associated with long-term-care facility residence of patients with Alzheimer's disease from Alzheimer's disease research centers of California

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content FACTORS ASSOCIATED WITH LONG-TERM-CARE FACILITY
RESIDENCE OF PATIENTS WITH ALZHEIMER’S DISEASE FROM
ALZHEIMER’S DISEASE RESEARCH CENTERS OF CALIFORNIA

by
Han Wang

A Thesis Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(BIOSTATISTICS)

August 2008

Copyright 2008 Han Wang

ii

Dedication
To my parents

iii

Acknowledgements
The author is grateful to Dr. Wendy Mack, Dr. Freddi Segal-Gidan, Dr. Lon
Schneider, Dr. Stanley Azen, and Scott Lyness for their guidance and substantive
help throughout this study.

iv

Table of Contents

Dedication

ii
Acknowledgements

iii
List of Tables

v
Abstract

vi
Introduction

1
Methods
Data Collection
Statistical Analysis

4
4
5
Results
Characteristics of Study Population
Site Specific Characteristics
Explanatory Variables
Predictor Variables
Supplemental Analysis

11
11
14
15
15
17
Discussion

20
Bibliography

25

v

List of Tables

Table 1: Characteristics of ARCC Study Population (n = 10,244)

12

Table 2: Prevalence of Long-Term-Care Facility Residence Across Sites

15
Table 3: Multivariate Logistic Regression Model Predicting Long-Term-
Care Facility Residence

16
Table 4: Multivariate Logistic Regression Model Predicting Long-Term-
Care Facility Residence Based on Dataset with Income Information

19

vi

Abstract
The purpose of this cross-sectional study was to identify the predictors of
long-term-care facility residence of patients with Alzheimer’s disease (AD). The
sample included 10,421 AD patients who were evaluated at Alzheimer’s Disease
Research Centers in California (ARCCs) from 1985 to 2007. Based on logistic
regression analysis, the predictors of long-term-care facility residence included older
age, higher education, presence of depression, increased dependency in basic
activities of daily living (ADLs) and instrumental activities of daily living (IADLs).
The relationship between living situation and IADLs was modified by ethnicity and
marital status. In a supplemental analysis among the subgroup of patients whose
income information was available, white ethnicity, health care payment, income, and
greater impairment in cognition were also associated with long-term-care facility
residence. These factors could be used by health-care professionals in an attempt to
predict and develop interventions to delay AD patients’ long-term-care facility
residence.

1

Introduction
Alzheimer’s disease (AD), which is the most common cause of dementia
accounting for 60% to 80% of cases, is the seventh leading cause of death in the
United States and the fifth leading cause of death in Americans older than age of 65.
More than 5 million Americans are estimated to have AD (Alzheimer’s Association,
2008). More women than men have AD, due to women’s longer life expectancy,
rather than sex-specific risk factors (Hebert, Scherr, McCann et al., 2001). People
with fewer years of education are more likely to be diagnosed with AD than those
with more years of education. It is believed that higher education reduces the risk of
AD either by decreasing the ease of clinical detection or by providing a greater
cognitive reserve that delays the onset of AD (Ngandu, Strauss, Helkala et al., 2007;
Stern, Gurland, Tatemichi et al., 1994). Although the exact cause or causes of AD
are not known, most experts agree that a combination of inter-related factors cause
AD. The most important risk factor of AD is aging. Most patients with AD are age
65 or older. Other factors, such as Apolipoprotein E e4 (APOE e4) genotype and
traumatic head injury also increase the risk of AD.
The symptoms of AD include difficulty remembering names and recent
events, impaired judgment, disorientation to time and place, behavior changes, and in
the latter stages of the disease, trouble swallowing, speaking, and walking. The
symptoms worsen with time and AD is ultimately fatal. There is no current treatment
to halt the deterioration of brain cells in AD. The U.S. Food and Drug

2

Administration has approved five drugs that temporarily slow down the worsening of
symptoms for about 6 to 12 months, on average, for about half of the patients who
take these medications (Alzheimer’s Association, 2008).
Although residing at home maximizes patients’ independence, caring for
patients with AD at home is very difficult. Research shows that many family
caregivers experience high level of stress and depression (Yaffe, Fox, Newcomer et
al., 2002). Even if families generally try to keep the patient living with the family as
long as possible, many patients must be eventually placed in a long-term-care facility
for medical and safety reasons. Extensive longitudinal studies have been done to
identify factors that predict long-term-care facility placement for elderly adults.
Variables consistently found to predict long-term-care facility placement included
sociodemographic characteristics such as older age, white ethnicity, female gender,
higher education, lack of spouse, and poverty status (Bauer, 1996; Liu & Tinker,
2001; Miller & Weissert, 2000; Osterweil, Martin, & Syndulko, 1995). A second
group of disease-related variables including cognitive impairment and functional
dependence are also related to a higher likelihood of long-term-care facility
placement (Gaugler, Duval, Anderson et al., 2007; Miller & Weissert, 2000;
Osterweil et al., 1995; Scott, Edwards, Davis et al., 1997; Sevenson, Smith, Tangalos
et al., 1994). Other studies focused more specifically on risk factors for
institutionalization in people with dementia. In addition to the factors mentioned
above, a third category of caregiver-related variables, including caregiver burden,

3

depressive mood, and low quality of life of family caregivers, have been shown to be
associated with long-term-care facility placement (Argimon, Limon, Vila et al., 2005;
Hébert, Dubois, Wolfson et al., 2001).
The aim of this study was to identify the factors associated with AD patients’
residence in a long-term-care facility, and to quantify the associations. The dataset
we used was collected by eleven Alzheimer’s Disease Research Centers of California
(ARCCs) through the Alzheimer’s Disease Program.

4

Methods
Data Collection: The Alzheimer’s Disease Program was established in 1984 in
California. The mission of the program is to reduce the burden and economic cost
associated with AD, and to assist in ultimately discovering the cause and cure of AD
(Alzheimer’s Disease Program of California, 2008). To accomplish the objectives,
eleven ARCCs were established at university medical centers throughout California.
The ten ARCCs that exist today are located in San Francisco, Irvine, Los Angeles,
Los Angeles/Downey, Los Angeles/San Fernando Valley, Martinez, Palo Alto,
Sacramento, San Diego, and San Francisco. These sites are operated by University of
California (UC) San Francisco, UC Irvine, University of Southern California (USC),
UC Los Angeles, UC Davis, Stanford University/VA Palo Alto Health Care System,
UC Davis, and UC San Diego. The site at USC/St. Barnabas was closed in 1998.
Patients have been recruited from the ARCCs since 1985. Each site contributed
around 100 patients per year on average, with a combined total of 23,542 initial
intakes from 1985 – 2007. 16,413 (69.7%) of these patients have been followed up
for reassessment approximately on an annual basis at least once. Patients were
interviewed by clinicians, and patients’ demographic data, caregiver data,
medical/diagnostic information, etc. were collected by a common instrument used in
all ARCCs, the Minimum Uniform Data Set (MUDS). There are two parts of the
MUDS instrument, Part 1 for initial assessment and Part 2 for reassessment. In this
analysis, we focused on modeling the living arrangement of AD patients based on

5

the dataset from the initial assessment only, i.e., we used a cross-sectional analysis to
identify correlates of long-term-care facility residence at the time at the initial
assessment.
Eight different versions of the MUDS have been used, the 1988, 1992, 1993,
1995, 1998, 1999, 2001 and 2006 versions, and the instruments have been constantly
modified. First, we identified and matched the common variables across all versions.
Then the dataset was thoroughly examined for data errors. Errors identified included
coding errors and inappropriate format. Errors were corrected after verification with
each ARCC site. We also retrieved some of the missing data from the hard copies of
the instruments. The original sample included patients with cognitive problem due to
any etiology. For this analysis, we only included the patients who had been
diagnosed with AD (possible or probable AD). The sample size was 10,421 (44.3%)
out of a total of 23,542.

Statistical Analysis: Our goal was to identify the factors that were associated with
long-term-care facility residence in patients with AD and to quantify these
associations. We combined living in a health-related facility and living in a group
quarters as living in a long-term-care facility; living alone, living in a household with
spouse or spouse equivalent only, living in a household with spouse or spouse
equivalent and others, living in a household with relatives, and living in a household
with non-relatives only as living at home. We used logistic regression with living

6

situation (long-term-care facility residence vs. home residence) as the outcome
variable to estimate the adjusted odds ratios (ORs) and their corresponding 95%
confidence intervals (CIs) associated with potential correlates of long-term-care
facility residence. We also considered effect modifiers and reported separate ORs for
each level of the effect modifier when we found significant interactions.
Independent variables of interest included patients’ age, education, gender,
ethnicity, marital status, whether the patient had Medicare, whether the patient had
Medi-Cal (Medicaid), medical history (including heart disease, stroke, transient
ischemic attack, seizure disorder, traumatic brain injury, hypertension, diabetes,
thyroid disorder, alcohol abuse, drug abuse, chronic obstructive pulmonary disease,
renal insufficiency, visual impairment, hearing impairment, and cancer), causal
factors of the patients’ cognitive syndrome (including cerebrovascular disease,
Parkinson’s disease, normal pressure hydrocephalus, depression, past alcohol abuse,
metabolic disorder, head trauma, CNS infection, and space-occupying lesion), basic
ADLs (including eating, dressing, and bladder & bowel control, scored from 0 to 9
points, with 0 point indicating perfect independence and 9 points total dependence),
IADLs (including ability to find way around familiar streets, perform household
tasks, cope with small sums of money, remember short lists of items, find way about
indoors, interpret surroundings, recall recent events, and tendency to dwell in the
past, scored from 0 to 8 points, with 0 point indicating perfect independence and 8
points total dependence), and MMSE (Mini-Mental Status Examination, scored from

7

0 to 30 points, with 30 points indicating perfect cognitive status and 0 point total
cognitive impairment) score. Five of these potential explanatory variables were
continuous measures: patients’ age at first visit, education, ADLs, IADLs, and
MMSE score. The assumption of linearity was checked for each continuous variable.
We first cut the continuous variable into four categories, then evaluated the OR for
each level of the variable, relative to the lowest level, and finally plotted the log OR
against the midpoint of each category. If the plot was linear, we kept the variable as
continuous. According to results from the above analysis, we treated patients’ age at
first visit, ADLs, and IADLs, as continuous variables. Patients’ education was
categorized as a binary variable (> 12 years vs. ≤ 12 years), because the relationship
between living situation and education lower than 12 years and that between living
situation and education higher than 12 years was found to satisfy the assumption of
linearity, respectively. We categorized the MMSE score according to clinical
practice guidelines: ≤ 9 points, severe dementia; 10 – 14 points, moderate severe
dementia; 15 – 19 points, moderate dementia; 20 – 30 points, mild to very mild
dementia. We categorized ethnicity as white vs. non-white, because counts in non-
white categories (American Indian or Alaskan Native, Asian, Pacific Islander,
African American, Filipino, Hispanic, and other ethnicity) were sparse. We
categorized patients’ marital status as married vs. non-married (never married,
widowed, divorced, and separated).

8

In step one, we analyzed each variable univariately for its relationship with
living situation. T-tests were used to compare the mean of each continuous
independent variable, and chi-squared tests were used to compare the frequency of
each categorical independent variable with regard to patients’ living situation.
Variables significant at p < 0.15 were included as candidates for a multivariate
model. We were conservative in this step and chose a higher p-value than 0.05,
because certain variables might have significant effects when modeled with other
variables due to uncontrolled confounders in univariate analysis.
Before doing multivariate analyses, we tested correlations between each pair
of univariately statistically significant independent variables from step one. If any
pair of variables were highly correlated with each other (r > 0.90), then one of them
was dropped.
In step two, because there were a large number of candidate variables in our
dataset, we tested the significance of each variable in the multivariate models by
major group. We divided the independent variables into three main groups:
demographic variables, medical history, and causal factors. ADLs, IADLs, and
MMSE score were each considered separately as candidates in the final multivariate
model. For each group of variables, we fitted separate multivariate models to
determine the multivariately significant variables (p < 0.10) in that group. We
excluded non-significant variables one by one and compared the coefficients of the
remaining variables with that of the full model. If the coefficients changed by more

9

than 25%, indicating that the variable excluded confounded the relationship between
living situation and other exposures, we retained the variable in the model. If the
coefficients did not change much, but the variables were generally believed to
contribute to a patient’s living situation (for example, MMSE score and whether
patient had Medicare), we also kept these variables in the multivariate model. We
dropped the variable if it failed to satisfy either one of the conditions above (i.e. p <
0.10, a confounder, or a generally accepted covariate). Here we chose the cutoff p =
0.10, so that we could restrict the number of the variables that entered the final
model, and at the same time be conservative enough to consider the fact that certain
variables might be significant at p < 0.05 level when modeled with variables in other
groups.
In the third step, we modeled together those variables that were significant at
p < 0.10 level in the second step and selected the significant ones (p < 0.05) as the
final predictors based on the same methods described in step two.
In both steps two and three, we used backward selection, because our sample
size was large enough compared to the number of predictors, to provide acceptable
degrees of freedom in each step of the analysis.
Before testing for interactions, we added the variables that were not in the
current model back one by one. If adding them back changed the estimate of one or
more coefficients by more than 25%, or the new variables were significant at p <

10

0.05, we kept them in the model. All statistical testing in the multivariate analyses
utilized Wald’s tests.
After considering these main effects, effect modifiers were checked by
likelihood ratio tests between the multivariate model with only the main effects and
the model with the interaction term. We focused on the interactions between
demographic variables and clinical measurements. All the univariately significant (p
< 0.05) interaction terms were then put together in the multivariate model and p-
value of Wald’s test for the interaction terms were checked. Those that were no
longer significant were excluded.

11

Results
Characteristics of Study Population: Characteristics of the study population are
shown in Table 1. Among the 10,421 patients with AD, 977 (9.4%) lived in a long-
term-care facility and 9,267 (88.9%) lived at home. Information on living situation
was missing for the remainder 177 (1.7%), and these subjects were excluded from
the analysis. Patients living in a long-term-care facility had a higher mean age at
first clinic visit, age at symptom onset, BRDRS, ADLs, IADLs, and lower mean
MMSE score than patients living at home. For both patient groups, the initial clinic
visit occurred about 4.5 years after the symptom onset. Among the study population,
two-thirds (67%) were females. The percentage of males living in a long-term-care
facility was lower than that of males living at home. Whites and Hispanics were the
two major ethnic groups represented. A higher percentage of whites and lower
percentage of Hispanics lived in a long-term-care facility compared to living at home.
Married and widowed patients were the two major categories of patients’ marital
status. Among patients living in a long-term-care facility, 16.2% were married and
63.9% were widowed. In contrast, among patients living at home, 51.6% were
married and 35.3% were widowed. The percentage of patients who had Medicare or
Medi-Cal (Medicaid) was higher in those who lived in a long-term-care facility
compared to those living at home. Table 1 displays patients’ medical history and
causal factors of the cognitive syndrome only for those factors with prevalence
higher than 10%. All variables shown in table 1 are statistically significantly

12

associated with living situation except for education level, whether patient had Medi-
Cal (Medicaid), medical history of hypertension, diabetes, and cancer at p < 0.15
level.
Table 1. Characteristics of ARCC Study Population (n = 10,244)
Living in a long-term-care facility Living at home
Variables (n = 977) (n = 9,267) P value
Age at first clinic visit 80.8 (7.4) 76.9 (8.4) <.0001
Age at symptom onset 75.4 (8.1) 72.1 (8.9) <.0001
MMSE
a
score 15.0 (7.1) 17.3 (7.0) <.0001
BRDRS
b
score 7.1 (3.7) 5.0 (3.4) <.0001
ADLs
c
2.1 (2.3) 1.2 (1.8) <.0001
IADLs
d
4.9 (1.8) 3.7 (1.9) <.0001
Gender <.0001
Male 229 (23.4%) 3153 (34.1%)
Female 748 (76.6%) 6107 (66.0%)
Education level 0.86
>12 Years 416 (42.6%) 3974 (42.9%)
≤ 12 Years 561 (57.4%) 5293 (57.1%)
Ethnicity <.0001
e

White 848 (87.2%) 6738 (72.8%)
Hispanic 47 (4.8%) 1146 (12.4%)
African American 32 (3.3%) 674 (7.3%)
Asian 40 (4.1%) 419 (4.5%)
Filipino 2 (0.2%) 88 (1.0%)
American Indian or Alaskan Native 2 (0.2%) 26 (0.3%)
Pacific Islander 0 (0%) 27 (0.3%)
Other 2 (0.2%) 134 (1.5%)
Marital status <.0001
f

Married 156 (16.2%) 4717 (51.6%)
Widowed 617 (63.9%) 3231 (35.3%)
Divorced 115 (11.9%) 821 (9.0%)
Never married 56 (5.8%) 240 (2.6%)
Separated 21 (2.2%) 109 (1.2%)
Other 0 (0%) 30 (0.3%)

13

Table 1. Continued
Living in a long-term-care facility Living at home
Variables (n = 977) (n = 9,267) P value
Medicare <.0001
Yes 881 (93.2%) 7672 (84.8%)
No 64 (6.8%) 1372 (15.2%)
Medi-Cal (Medicaid) 0.62
Yes 160 (17.2%) 1476 (16.5%)
No 773 (82.8%) 7457 (83.5%)
Medical history of heart disease 0.002
Yes 273 (34.0%) 2167 (28.8%)
No 529 (66.0%) 5367 (71.2%)
Medical history of hypertension 0.91
Yes 407 (48.5%) 3835 (48.3%)
No 432 (51.5%) 4105 (51.7%)
Medical history of diabetes 0.33
Yes 94 (11.8%) 966 (13.0%)
No 705 (88.2%) 6481 (87.0%)
Medical history of thyroid disorder 0.02
Yes 183 (22.9%) 1445 (19.4%)
No 615 (77.1%) 5992 (80.6%)
Medical history of visual impairment 0.0003
Yes 375 (44.8%) 2972 (38.3%)
No 463 (55.2%) 4782 (61.7%)
Medical history of hearing impairment 0.01
Yes 256 (31.0%) 2033 (26.9%)
No 570 (69.0%) 5538 (73.1%)
Medical history of cancer 0.38
Yes 116 (14.6%) 993 (13.5%)
No 678 (85.4%) 6368 (86.5%)
Medical history of depression <.0001
Yes 269 (34.3%) 1904 (26.2%)
No 516 (65.7%) 5355 (73.8%)
Causal factor: cerebrovascular disease 0.15
Yes 192 (21.0%) 1634 (19.0%)
No 724 (79.0%) 6976 (81.0%)

14

Table 1. Continued
Living in a long-term-care facility Living at home
Variables (n = 977) (n = 9,267) P value
Causal factor: depression 0.02
Yes 150 (16.4%) 1171 (13.6%)
No 766 (83.6%) 7435 (86.4%)

Note. Data are M(SD) except for n(%). P value for continuous variable was obtained by t-test, and all others by
χ
2
test.
a
Mini-Mental Status Examination.
b
Blessed-Roth Dementia Rating Scale.
c
Activities of Daily Living.
d
Instrumental Activities of Daily Living.
e
White vs. Non-white.
f
Married vs. Non-married.

Site Specific Characteristics: As mentioned above, patients were recruited from
eleven sites throughout California. The characteristics of the population differed
among sites. For example, from 1985 to 2007, the ARCC site at Stanford University
recruited 1,245 patients with AD, among which 1,000 (80.3%) were white and only
60 (4.8%) were Hispanic. During the same time period, the ARCC site at
USC/Rancho Los Amigos recruited 941 patients with AD, among which 532 (56.5%)
were white and 273 (29.0%) were Hispanic. Again for these two ARCC sites,
patients’ mean education was 13.4 years at Stanford, and only 10.6 years at
USC/Rancho Los Amigos. From table 2, the prevalence of long-term-care residence
was statistically significantly different across sites (χ
2
(10) = 45.08, p < .0001),
ranging from 5.1% to 12.4%. Therefore to control for the site effects, we created
indicator variables for each site, and included them in the model.

15

Table 2. Prevalence of Long-Term-Care Facility Residence Across Sites

Living in a long-term-
care facility Living at home
Site n (%) n (%)
University of California, San Diego 109 (8.7) 1145 (91.3)
University of Southern California/Rancho Los Amigos 63 (6.7) 875 (93.3)
University of Southern California/St. Barnabas (88’ – 98’) 36 (11.3) 283 (88.7)
University of California, Davis/Sacramento 160 (10.8) 1327 (89.2)
University of California, Davis/Berkeley, Martinez 192 (12.4) 1354 (87.6)
University of California, San Francisco 109 (9.5) 1041 (90.5)
Stanford University 107 (8.6) 1130 (91.4)
University of California, San Francisco/Fresno 112 (10.6) 944 (89.4)
University of California, Irvine 57 (9.0) 576 (91.0)
University of California, Los Angeles 13 (5.2) 236 (94.8)
University of Southern California/Los Angeles 19 (5.1) 356 (94.9)

Note. χ
2
(10) = 45.08, p < .0001.

Explanatory Variables: Explanatory variables in the final multivariate model
included ethnicity, marital status, age, education, presence of depression, ADLs,
IADLs, MMSE score, and whether patients had Medicare as health care payment.
MMSE score and whether patients had Medicare were included because they are
generally considered important predictors of long-term-care facility residence. We
found ethnicity (p = 0.0004) and marital status (p = 0.01) to be effect modifiers in the
relationship between patients’ living situation and IADLs score.

Predictor Variables: Table 3 shows the ORs and 95% CIs in the multivariate model.
After adjusting for all the variables listed above, with each 5 years increase in age,
patients were 1.20 times as likely to live in a long-term-care facility (95% CI = 1.13

16

– 1.27). Patients who had more than 12 years of education were 1.33 times as likely
to live in a long-term-care facility as those who had 12 or fewer years of education
(95% CI = 1.12 – 1.58). Patients who had depression were 1.55 times as likely to live
in a long-term-care facility as those who did not have depression (95% CI = 1.24 –
1.94). With every point increase in ADLs score (increased ADLs dependency),
patients were 1.09 times as likely to live in a long-term-care facility (95% CI = 1.04
– 1.15). For those who are white and married, with every point increase in IADLs
score (increased IADLs dependency), patients were 1.50 times as likely to live in a
long-term-care facility (95% CI = 1.33 – 1.68), while for those who are white and
non-married, the odds ratio decreased to 1.28 (95% CI = 1.20 – 1.37), and for those
who are non-white and married, 1.19 (95% CI = 1.01 – 1.40), finally for those who
are non-white and non-married, 1.02 (95% CI = 0.90 – 1.16).
Table 3. Multivariate Logistic Regression Model Predicting Long-Term-Care Facility Residence
Variable OR 95% CI
Age
a
1.20 1.13 - 1.27
Education: > 12 years vs. ≤ 12 years 1.33 1.12 - 1.58
Depression 1.55 1.24 - 1.58
ADLs
b
1.09 1.04 - 1.15
IADLs
b
: white married 1.50 1.33 - 1.68
IADLs
b
: white non-married 1.28 1.20 - 1.37
IADLs
b
: non-white married 1.19 1.01 - 1.40
IADLs
b
: non-white non-married 1.02 0.90 - 1.16

Note. Controlling for Medicare, MMSE and site effects. For abbreviation, see table 1.
a
Odds ratio for 5 years increase in age.
b
Odds ratio for 1 point increase.

17

Supplemental Analysis: Because the data was collected from eight different
versions of the data collection instruments, we could only focus on the common
variables available in all versions. Since the instruments changed substantially over
the 23 years, we had difficulty matching up certain variables, and some potentially
important variables had to be dropped from the final dataset. For example, a factor
that should be very important in determining patients’ long-term-care facility
placement is their socioeconomic status. Those who are more economically stable
should be more likely to afford medical insurance and thus more likely to be placed
in a long-term-care facility. However, the household annual income information was
only collected in 5 versions of the instrument, and was not included in the later
versions. To verify that income information is important in predicting AD patients’
living situation, we performed a supplemental analysis based on the partial dataset
from the 5 versions following the same methods.
The sample size was 5,853 patients, among which 605 (10.3%) lived in a
long-term-care facility, and 5197 (88.8%) lived at home. Information on living
situation was missing for the remainder 51 (0.9%), therefore they were excluded
from the analyses. We categorized the household annual income information as:
under $10,000 vs. $10,000 and above. The explanatory variables that entered the
final model included: ethnicity, marital status, whether patient had Medicare,
whether patient had Medi-Cal (Medicaid), age at first clinic visit, education, income,
whether patient had depression, MMSE score, and IADLs, controlling for ADLs and

18

site effects. Consistent with the full-dataset model, the relationship between living
situation and IADLs was modified by ethnicity (p = 0.001).
Table 4 shows the ORs and 95% CIs in this reduced multivariate model, with
the income information as one of the independent variables. Controlling for all the
other explanatory variables, patients who were married were 0.18 time as likely to
live in a long-term-care facility as those who were not married (95% CI = 0.13 –
0.25). With every 5 years increase in age, patients were 1.22 times as likely to live in
a long-term-care facility (95% CI = 1.12 – 1.34). Patients who had more than 12
years of education were 1.50 times as likely to live in a long-term-care facility as
those who had 12 or fewer years of education (95% CI = 1.15 – 1.94). Patients
whose household income was $10,000 or more were 1.49 times as likely to live in a
long-term-care facility as those whose household income was lower than $10,000
(95% CI = 1.08 – 2.05). Patients who had Medi-Cal (Medicaid) were 1.47 times as
likely to live in a long-term-care facility as those who did not have Medi-Cal
(Medicaid) (95% CI = 1.03 – 2.11). Patients who had Medicare were 2.30 times as
likely to live in a long-term-care facility as those who did not have Medicare (95%
CI = 1.21 – 4.37). Patients who had depression were 1.62 times as likely to live in a
long-term-care facility as those who did not have depression (95% CI = 1.16 – 2.26).
Patients who had severe dementia were 1.65 times (95% CI = 1.09 – 2.50), and
patients who had moderate dementia were 1.49 times (95% CI = 1.07 – 2.08) as
likely to live in a long-term-care facility as those who had mild dementia. With every

19

point increase in IADLs score (increased IADLs dependency), white patients were
1.34 times as likely to live in a long-term-care facility (95% CI = 1.22 – 1.47); the
OR for the association of IADLs with long-term-care facility residence among non-
white was not statistically significant.
Table 4. Multivariate Logistic Regression Model Predicting Long-Term-Care Facility Residence Based on
Dataset with Income Information
Variable OR 95% CI
Age
a
1.22 1.12 - 1.34
Education: > 12 years vs. ≤ 12 years 1.50 1.15 - 1.94
Married 0.18 0.13 - 0.25
Income: ≥ $10,000 vs. < $10,000 1.49 1.08 - 2.05
Medi-Cal (Medicaid) 1.47 1.03 - 2.11
Medicare 2.30 1.21 - 4.37
Depression 1.62 1.16 - 2.26
MMSE: severe vs. mild dementia 1.65 1.09 - 2.50
MMSE: moderate severe vs. mild dementia 1.18 0.79 - 1.75
MMSE: moderate vs. mild dementia 1.49 1.07 - 2.08
IADLs
b
: white 1.34 1.22 - 1.47
IADLs
b
: non-white 0.96 0.79 - 1.17

Note. Controlling for ADLs and site effects. For abbreviation, see table 1.
a
Odds ratio for 5 years increase in age .
b
Odds ratio for 1 point increase.

20

Discussion
Characteristics of patients recruited at Stanford and USC/Rancho Los
Amigos were compared in the Results part. These site differences reflected
differences among the local residence populations. From the 2000 census, in Palo
Alto, CA, where Stanford University is located, 72.8% of the population was white,
and only 4.6% was Hispanic or Latino of any race. For educational attainment
among persons 25 years and older, 31.4% had a bachelor’s degree, and 43.0% had a
graduate or professional degree. In Downey, where USC/Rancho Los Amigos is
located, 28.7% of the population was white, and 57.9% were Hispanic or Latino of
any race. For educational attainment among persons 25 years and older, only 11.9%
had a bachelor’s degree, and 5.4% had a graduate or professional degree. Other
information which we could not retrieve from the MUDS, for example,
socioeconomic status, also differed. Again from the 2000 census, in Palo Alto, the
median household income was $90,377. In Downey, the median household income
was only $45,667.
The results from the multivariate analysis suggest that for patients with AD,
older age, higher education, presence of depression, and increased ADLs dependency,
were related to a higher likelihood of living in a long-term-care facility. The greatest
independent risk factor (defined by biggest OR) for long-term-care facility residence
was presence of depression, with presence of depression increasing the likelihood of
long-term-care facility residence by 1.55 times. Thus, demographic factors, patients’

21

functional and psychiatric status were found to be statistically significant predictors
of long-term-care facility residence. Both ethnicity and marital status modified the
relationship between IADLs dependency and living situation, with the IADLs-
residence association stronger among married subjects and white subjects. With
increased IADLs dependency, white married patients were most likely to be in a
long-term-care facility, with white non-married, non-white married following. Non-
white non-married were the least likely to be in a long-term-care facility. A possible
explanation for this is the relationship to socioeconomic status with white married
patients having greater financial resources and therefore the ability to afford
placement and care living in a long-term-care facility. Among non-white non-
married patients, the association of IADLs with living situation was not statistically
significantly different from 1.0.
From the results based on the partial dataset, marital status was no longer an
effect modifier between living situation and IADLs, but instead was the greatest risk
factor of long-term-care facility residence based on the OR. Medicare, Medi-Cal
(Medicaid) and MMSE became statistically significant, but ADLs did not. Most
noticeably, we found income to be a statistically significant predictor of living
situation. The association between living situation and age remained the same, while
the association between living situation and education and presence of depression
were both higher in the results from partial dataset. The major limitation of this
supplemental analysis was that patients whose information was collected in these

22

five versions of the instrument were enrolled from 1985 to 1999, and people’s
income increased steadily during this time period. Therefore ideally, income should
be categorized based on the relative income level rather than the absolute amount.
The factors we identified and their associations with living situation were in
accord with previous findings from cross-sectional studies (Trottier, Martel, Houle et
al., 2000; Weissert & Cready, 1989), including demographic variables (age, ethnicity,
marital status, and education), clinical factors (depression), and functional
impairment measures (ADLs and IADLs). Unfortunately there was no data on
caregivers’ burden and quality of life in the MUDS dataset, both of which have also
been found to be associated with long-term-care facility residence (Argimon et al.,
2005; Hébert et al., 2001). A new finding is that the relationship between IADLs and
living situation is modified by the combination of ethnicity and marital status.
The strengths of this study are: the sample size was large; patients were
recruited across the entire state of California; data collection was done by interviews
conducted by clinicians according to the same instruments; the instruments were not
designed for this particular study, therefore there should be no bias from clinicians
who were collecting the data.
The dataset had several limitations. First, it was a dataset with information
from patients recruited from 1985 to 2007. Diagnostic criteria for dementing illness
likely changed during this period of time due to new medical findings or new
technology, so some diagnosis were not consistent over time. This could have led to

23

over-diagnosis of AD in the earlier years as well under-diagnosis or misdiagnosis at
times. Also clinical practices may have varied over time as well as the availability of
resources such as nursing homes. Second, the information was collected from eleven
sites through clinicians’ interviews with patients. Although there was a manual
providing guidelines for the clinicians, it was difficult to guarantee that every
clinician followed exactly the same criteria on diagnosis. This is almost inevitable
for any large dataset collected by multiple groups at multiple sites over an extended
period of time.
A final limitation was the cross-sectional data, thereby we actually compared
non-institutionalized and institutionalized AD patients at one assessment point.
Patients were already living in a long-term-care-facility before they were recruited at
the ARCCs. Therefore we could not identify the factors that were related to the
placement of long-term-care facility, i.e., the factors influencing patients being
moved from home to long-term-care facility. Also this analysis only included the
patients in a long-term-care facility who have survived, thus the factors identified
were related to both long-term-care facility placement and survival once placed.
In spite of these limitations, the research provides some valuable findings
using this dataset from the Alzheimer’s Disease Program in California. The program
is ongoing, so more waves of data are added every year; this analysis was based on
the most up-to-date dataset (latest modification in Jan. 2008). In addition to the
results reported here, we found many data errors and had the data administrators

24

correct them. This research might also be used as a reference by the Alzheimer’s
Disease Program when they modify the instruments, so the information that is
important for statistical analyses could be kept or added in future versions.
All analyses reported here used the initial assessment. In addition, there is data from
reassessments conducted during the 23 years of research, which means repeated
measures data is available. Because we followed the patients’ living situation over
the years, in the future, we could focus our research on identifying the predictors of
time from diagnosis to long-term-care facility placement and death in patients with
AD.

25

Bibliography
Alzheimer’s Association. (2008). 2008 Alzheimer’s disease facts and figures.
Alzheimer’s & Dementia, 4(2), 110-133.

Alzheimer’s Disease Program of California. (2008) Retrieved June 27, 2008, from
http://www.dhs.ca.gov/CDIC/alzheimers/html/ARCCHome.htm

Argimon, J. M., Limon, E., Vila, J., & Cabezas, C. (2005). Health-related quality-of-
life of care-givers as a predictor of nursing-home placement of patients with
dementia. Alzheimer Disease and Associated Disorders, 19, 41-44.

Bauer, E. J. (1996). Transitions from home to nursing home in a capitated long-term
care program: The role of individual support systems. Health Services
Research, 31, 22-30.

Gaugler, J. E., Duval, S., Anderson, K. A., & Kane, R. L. (2007). Predicting nursing
home admission in the U.S.: a meta-analysis. BMC Geriatrics, 7: 13.

Hebert, L. E., Scherr, P. A., McCann, J. J., Beckett, L. A., & Evans, D. A. (2001). Is
the risk of developing Alzheimer’s disease greater for women than for men?
American Journal of Epidemiology, 153(2), 132-136.

Hébert, R., Dubois, M., Wolfson, C., Chambers, L., & Cohen, C. (2001). Factors
associated with long-term institutionalization of older people with dementia:
data from the Canadian Study of Health and Aging. Journal of Gerontology,
56A(11), M693-M699.

Liu, L. F., & Tinker, A. (2001). Factors associated with nursing home entry for older
people in Taiwan, Republic of China. Journal of Interprofessional Care, 15,
245-255.

Miller, E. A., & Weissert, W. G. (2000). Predicting elderly people’s risk for nursing
home placement, hospitalization, functional impairment, and mortality: A
synthesis. Medical Care Research and Review, 57, 259-297.

Ngandu, T., Strauss, E., Helkala, E. L., Winblad, B., Nissinen, A., Tuomilehto, J., et
al. (2007). Education and dementia: What lies behind the association?
Neurology, 69(14), 1442-1450.

26

Osterweil, D., Martin, M., & Syndulko, K. (1995). Predictors of skilled nursing
placement in a multilevel long-term-care facility. Journal of the American
Geriatrics Society, 43(2), 108-112.

Scott, W. K., Edwards, K. B., Davis, D. R., Cornman, C. B., & Macera, C. A. (1997).
Risk of institutionalization among community long-term care clients with
dementia. The Gerontologist, 37(1), 46-51.

Severson, M. A., Smith, G. E., Tangalos, E. G., Peterson, R. C., Kokmen, E., Ivnik,
R. J., et al. (1994). Patterns and predictors of institutionalization in
community-based dementia patients. Journal of the American Geriatrics
Society, 42(2), 181-185.

Stern, Y., Gurland, B., Tatemichi, T. K., Tang, M. X., Wilder, D., & Mayeux, R.
(1994). Influence of education and occupation on the incidence of Alzheimer’s
disease. JAMA, 271(13), 1004-1010.

Trottier, H., Martel, L., Houle, C., Berthelot, J., & Légaré, J. (2000). Living at home
or in an institution: What makes the difference for seniors? Health Reports,
11(4), 49-61.

Weissert, W. G., & Cready, C. M. (1989). Toward a model for improved targeting of
aged at risk of institutionalization. Health Services Research, 24(4), 485-510.

Yaffe, K., Fox, P., Newcomer, R., Sands, L., Lindquist, K., Dane, K., et al. (2002).
Patients and caregiver characteristics and nursing home placement in patients
with dementia. JAMA, 287(16), 2090-2097.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Racial/ethnic differences in patient experiences with health care in association with the Gleason’s score level at prostate cancer diagnosis: findings from the SEER-CAHPS data

PDF

The effect of renal function on toxicity of E7389 (eribulin) among patients with bladder cancer

PDF

Racial/ethnic differences in colorectal cancer patient experiences, health care utilization and their association with mortality: findings from the SEER-CAHPS data

PDF

Factors associated with agreement between childhood cancer survivors (CCS) and their parents on knowledge and attitudes about health-related quality of life, treatment and follow-up care

PDF

Eliciting perspectives on palliative care: outpatient visits, advance care planning, and the impact of COVID-19

PDF

The evaluation of the long-term effectiveness of zero/low fluoroscopy workflow in ablation procedures for the treatment of paroxysmal and persistent atrial fibrillation

PDF

Association of carotid artery stiffness with measures of cognition in older adults

PDF

Age related macular degeneration in Latinos: risk factors and impact on quality of life

PDF

Chronic eye disease epidemiology in the multiethnic ophthalmology cohorts of California study

PDF

Associations of ambient air pollution exposures with perceived stress in the MADRES cohort

Asset Metadata

Creator Wang, Han (author)

Core Title Factors associated with long-term-care facility residence of patients with Alzheimer's disease from Alzheimer's disease research centers of California

School Keck School of Medicine

Degree Master of Science

Degree Program Biostatistics

Publication Date 07/20/2008

Defense Date 06/25/2008

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

Tag Alzheimer's disease,cross-sectional study,long-term-care facility residence,OAI-PMH Harvest

Language English

Advisor Mack, Wendy J. (committee chair), Azen, Stanley Paul (committee member), Schneider, Lon (committee member), Segal-Gidan, Freddi (committee member)

Creator Email hanwang@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1361

Unique identifier UC180449

Identifier etd-Wang-20080720 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-84849 (legacy record id),usctheses-m1361 (legacy record id)

Legacy Identifier etd-Wang-20080720.pdf

Dmrecord 84849

Document Type Thesis

Rights Wang, Han

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu